Advances in Cognitive Science Volume 1
Advances in Cognitive Science Volume 1
Edited by
Narayanan Srinivasan A.K. Gupta Janak Pandey
Copyright © Narayanan Srinivasan, A.K. Gupta, and Janak Pandey, 2008
All rights reserved. No part of this book may be reproduced or utilized in any form or by any means, electronic or mechanical, including photocopying, recording or by any information storage or retrieval system, without permission in writing from the publisher.
First published in 2008 by SAGE Publications India Pvt Ltd B1/I-1 Mohan Cooperative Industrial Area Mathura Road, New Delhi 110 044, India www.sagepub.in SAGE Publications Inc 2455 Teller Road Thousand Oaks, California 91320, USA SAGE Publications Ltd 1 Oliver’s Yard, 55 City Road London EC1Y 1SP, United Kindom SAGE Publications Asia-Pacific Pte Ltd 33 Pekin Street #02-01 Far East Square Singapore 048763
Published by Vivek Mehra for SAGE Publications India Pvt Ltd, typeset in Stone Serif 10/13 pt. by Innovative Processors, New Delhi, and printed at Chaman Enterprises, New Delhi. Library of Congress Cataloging-in-Publication Data Advances in cognitive science / edited by Narayanan Srinivasan, A.K. Gupta, Janak Pandey. p. cm. Includes bibliographical references and indexes. 1. Cognition. 2. Cognitive science. I. Srinivasan, Narayanan. II. Gupta, A.K. III. Pandey, Janak, 1945– BF311.A31
153—dc22
ISBN: 978-0-7619-3649-7 (HB)
2008
2008017471 978-81-7829-814-6 (India-HB)
The SAGE Team: Sugata Ghosh, Jasmeet Singh, Anju Saxena and Trinankur Banerjee
Contents List of Tables List of Figures List of Abbreviations Preface
ix xi xv xvii
Section I
Cognitive Processes
CHAPTER 1 Hierarchical Organization of Complex Visuo-Motor Sequences V.S. Chandrasekhar Pammi, S. Bapi Raju, Ahmed, K.P. Miyapuram and Kenji Doya
7
CHAPTER 2 Orienting Attention and Cued Sustained Attention Indramani L. Singh, Pamela M. Greenwood and Raja Parasuraman
18
CHAPTER 3 Why Does Foveal Bias Decrease in the Presence of Additional Element? Muhammad Kamal Uddin, Takahiro Kawabe and Sachio Nakamizo
31
CHAPTER 4 New Associative Learning in Amnesia Suparna Rajaram and H. Branch Coslett
43
CHAPTER 5 The Coordinated Processing of Scene and Utterance: Evidence from Eye-Tracking in Depicted Events Pia Knoeferle and Matthew W. Crocker
50
CHAPTER 6 Script Indices Richard Sproat and Prakash Padakannaya
62
CHAPTER 7 How Do We Parse Compound Words? Gary Libben
71
vi
Advances in Cognitive Science
CHAPTER 8 Other Minds: Social Cognition in Wild Bonnet Macaques Anindya Sinha
Section II
Cognitive Neuroscience
CHAPTER 9 A Survey of Molecular Mapping as Applied to Studies of the Visual System Avi Chaudhuri CHAPTER 10 Neural Substrates of Language Processing in Bilinguals: Imagi(ni)ng the Possibilities Jyotsna Vaid CHAPTER 11 Side Bias in Human Behaviour Manas K. Mandal, Hari S. Asthana and Ramakrishna Biswal
Section III
87
111
120
135
Computational Modelling
CHAPTER 12 Non-linear Dynamical Analysis of Point Neuron Models and Signal Propagation along Axon Deepak Mishra, Abhishek Yadav, Sudipta Ray and Prem Kumar Kalra CHAPTER 13 Smoke Without Fire: What Do Virtual Experiments in Cognitive Science Really Tell Us? Peter R. Krebs
159
177
CHAPTER 14 Complex Primitives and Their Linguistic and Processing Relevance Aravind K. Joshi
188
CHAPTER 15 Dissecting the Frog: Computational Approaches to Humour Perception Narayanan Srinivasan and Vani Pariyadath
197
Contents
Section IV
vii
Culture and Cognition
CHAPTER 16 Sources of Evidence and Levels of Interpretation in Culture-andCognition Research Ype H. Poortinga CHAPTER 17 Spatial Language and Concept Development: Theoretical Background and Overview R.C. Mishra and Pierre R. Dasen CHAPTER 18 Spatial Encoding: A Comparison of Sanskrit- and Hindi-Medium Schools Aparna Vajpayee, Pierre R. Dasen and R.C. Mishra CHAPTER 19 A Cross-Cultural Comparison of Spatial Language and Encoding in Bali and Geneva Pierre R. Dasen and Jürg Wassmann CHAPTER 20 Culture, Language, Spatial Frames of Reference and Hemispheric Dominance R.C. Mishra and Pierre R. Dasen CHAPTER 21 Cultural Adaptations and Cognitive Processes of Tribal Children in Chotanagpur R.C. Mishra and John W. Berry CHAPTER 22 An Eco-cultural Perspective on Cognitive Competence John W. Berry
Section V
221
240
253
264
277
287
300
Cognitive Development and Intervention
CHAPTER 23 Experimental Approaches to Specific Disabilities in Learning to Read: The Case of Symmetry Generalization in Developmental Dyslexia Thomas Lachmann
319
viii Advances in Cognitive Science CHAPTER 24 Cognitive Profiles of Children with Dyslexia Bhoomika R. Kar and Nishi Tripathi CHAPTER 25 Emergence of Social Play and Numeracy: A Related Development with Young At-Risk Students? Geerdina M. van der Aalsvoort, Arjette M. Karemaker and Mieke P. Ketelaars CHAPTER 26 Cognitive Stimulation of Rural School Children in India: An Evaluative Study Malavika Kapur
341
355
369
Section VI Consciousness CHAPTER 27 Taxonomy of Consciousness K. Ramakrishna Rao
383
About the Editors and Contributors Subject Index Name Index
424 434 447
List of Tables 1.1 1.2
The percentage change of SR for the four sessions The key-press RTs for the four sessions
11 15
11.1 11.2
Incidence of left handedness across countries Theoretical notions behind the incidence of left handedness
143 145
15.1
A comparative study of theories of humour perception
208
16.1
An overview of four levels of psychometric equivalence of data and three levels of inclusiveness of interpretations Extent to which validity of cross-cultural differences in score levels is open to empirical control (ruling out alternative explanations)
16.2
17.1
225 230
Spatial frames of reference in developmental psychology and in linguistics
241
18.1 18.2
Sample characteristics One-way ANOVA comparing Hindi-medium (H) and Sanskrit schools (S)
256 261
19.1 19.2
Sample characteristics of studies in Bali 2002 and 1994 Pearson correlation coefficients between language, encoding and background variables Pearson correlations between acculturation, language and encoding
265
Sample characteristics Pearson correlations between language, encoding and FDI Partial correlations controlling for age, gender, preschooling, grade, years of schooling and school type ANOVA outcomes on brain lateralization measures, G and E encoding groups
279 282
21.1 21.2 21.3 21.4 21.5
Mean score of groups on the measures of cultural dimensions Mean score of groups on differentiation measures Mean score of groups on contextualization measures Mean score of groups on integration measure Factor analysis outcomes on core cognitive measures
293 294 295 296 297
25.1
Design of the longitudinal study
359
19.3 20.1 20.2 20.3 20.4
271 272
282 283
x Advances in Cognitive Science 25.2 25.3 25.4
The means and standard deviations of the instruments used to select the subjects listed per school and per research condition The means and standard deviations of the ECERS ratings per category and the SES per school and per research condition The means and standard deviations on number sense, counting and ordering per school and per research condition
363 364 365
List of Figures 1.1 1.2 1.3 1.4 1.5 1.6
The 2 × 12 task procedure The block-wise combination graph Block-wise improvements of SR for all the three experiments Session-wise improvements of SR for all the three experiments Block-wise data of key-press RTs for all the three experiments Session-wise data of key-press RTs for all the three experiments
10 12 13 13 15 15
2.1 2.2 2.3 2.4
19 23 24
2.6
Percentage of correct detection as function of two-hour time period Perceptual sensitivity as function of event rate and cue validity Sensitivity index scores as a function of cue validity Sensitivity index scores as function of cue validity and block in low event rate Sensitivity index scores as function of cue validity and block in high event rate Mean sensitivity index scores under 300 and 450 SOA
3.1 3.2 3.3
Schematic representation of the experimental protocol Mean displacements plotted as a function of 5 experimental conditions Mean displacements plotted as a function of 5 experimental conditions
34 36 39
5.1 5.2 5.3 5.4
The Jackendovian architecture of the language system A sketch of the competitive-integra Example image from Experiment 1 Example item from Knoeferle and Crocker (2004)
52 53 56 57
6.1 6.2 6.3 6.4
Catenation operators, after Sproat 2000 Schematic illustration of script layout catenators Illustration of the layout catenators in Chinese Full and diacritic forms for Devanagari vowels, classified by catenator inherent to the diacritic forms Forms for diacritic Kannada vowels, classified by catenator Anusvara in Devanagari (left) and Kannada for the word /pensil/ Layout details for /pensil/ in Devanagari Layout details for /eešvaikya/ in Kannada Layout details for /lakšmiiša/ in Kannada Feature vector slots Feature vector slots for Devanagari /peMsil/
63 63 63
2.5
6.5 6.6 6.7 6.8 6.9 6.10 6.11
25 25 26
64 64 66 67 67 67 68 68
xii 6.12 6.13 7.1 7.2 7.3 7.4 7.5 7.6 8.1 8.2
12.1 12.2 12.3 12.4 12.5 12.6 12.7 12.8 12.9 12.10 12.11 12.12
Advances in Cognitive Science
Feature vector slots for Kannada /eešvaikya/ Feature vector slots for Kannada /lakšmiiša/
68 69
A simplistic view of compound parsing A more adequate view of compound parsing (The APPLE II model) Left branching and right branching structures for English triconstituent compounds The parsing of a left-branching triconstituent word in German The parsing of a right-branching triconstituent word in German A schematic representation for the parsing of ambiguous triconstituent compounds containing the morphemes 1, 2, and 3
75 76
Distribution of deceptive acts across different categories of tactical deception exhibited by Troops G I, G II and B I Correlation between the number of deceptive acts and the number of categories of tactical deception in which they were performed by males in the three study troops Time response and phase portraits for HH model Time responses and phase portraits for FitzHugh–Nagumo neuron model at (a) I = 0.5 (b) I = 1.5 (c) I = 2.35 Nullclines for FitzHugh–Nagumo neuron model Bifurcation diagram for FitzHugh–Nagumo neuron model Time response and phase portrait for Wilson–Cowan model (a) ry = –2 (b) ry = –3 (c) ry = –9.5 Bifurcation diagram for Wilson–Cowan model with ry as the bifurcation parameter Time response and phase portrait for cortical neuron model at (a) I = –3 (b) I = 1.5 and (c) I = 6 Bifurcation diagram for cortical neuron model Response of Morris–Lecar model when current is injected only at first compartment Response of Morris–Lecar model when current injection varies inversely with fourth root of axon length Response of Morris–Lecar model when current injection varies inversely with square root of axon length Response of Morris–Lecar model when current injection varies inversely with square root of axon length
14.1 Domain of locality of a context-free grammar 14.2 Substitution 14.3 Adjoining
78 80 81 83 97
99 161 163 164 164 166 167 168 169 172 173 173 174 189 189 190
List of Figures 14.4 14.5 14.6 14.7 14.8
xiii
An LTAG example An LTAG derivation An LTAG derived tree An LTAG derivation tree Two supertags for with
17.1 Results of previous studies (Bali, India, Nepal), absolute encoding for Animals task (3 animals only) and Steve’s Maze 18.1 Animals: number of items with completely geocentric encoding, out of seven items, using four animals 18.2 Chips: number of items with completely geocentric encoding, out of seven items 18.3 Steve’s Maze: number of items with completely geocentric encoding, out of five items 19.1 Mean items with absolute encoding for four animals (seven items) in three samples of 2002 study 19.2 Absolute encoding on Animals and Chips, in Bali (2002 study) and in Geneva 22.1 Eco-cultural framework linking ecology, cultural adaptation and individual behaviour 22.2 Cultural transmission: Linkages between contexts and outcomes 23.1 Letter and shape stimuli used in different blocks in the experiment by Brendler and Lachmann (2001) 23.2 Relationship between ‘b–d’ reversals in word reading and errors in the same–different task with physical instruction and lexical material 23.3 Mean RT (ms) as a function of angle of rotation for dyslexics and controls 23.4 Accuracy rate (%) as a function of angle of rotation in the dyslexic and normal reader group 24.1 24.2 24.3 24.4 24.5 24.6 24.7
Errors in reading English Errors in reading Hindi Writing errors—English Writing errors—Hindi Mean standard scores on PASS scales of CAS Mean performance of children with dyslexia on subtests of CAS Cognitive processes in decoding
27.1 Consciousness as awareness 27.2 States of awareness
191 191 192 192 194 246 259 260 260 268 274 304 305 328 329 332 333 346 346 348 349 350 350 351 384 384
xiv 27.3 27.4 27.5 27.6 27.7 27.8 27.9 27.10
Advances in Cognitive Science
Waking states of consciousness Altered states of consciousness Pure conscious states Content of awareness States of consciousness in Hinduism Buddhism: four planes of consciousness Ethical classification of consciousness in Buddhism Roots of mental states
385 389 393 395 399 414 414 416
List of Abbreviations AI Artificial Intelligence ANCOVA Analysis of Covariance ANN Artificial Neural Nets ANOVA Analysis of Variance CARG Cognitive Anthropology Research Group CAS Cognitive Assessment System CBCS Centre for Behavioural and Cognitive Sciences CBL Constraint-Based Lexicalist CFG Context-Free Grammar CMS Computer Models and Simulations CPM Coloured Progressive Matrices CS Conditioned Stimulus CT Computational Topography DA Dry Agriculture DID Dissociative Identity Disorder DIF Differential Item Functioning DPEP District Primary Education Project EC Extra-unit Connectedness EEG Electroencephalogram EPDA Embedded Push-down Automata Excitatory Post-Synaptic Potential EPSP ERPs Event Related Potentials ESP Extra-Sensory Perception FCD Functional Coordination Deficit fTCD Functional Transcranial Doppler Sonography FHN FitzHugh–Nagumo fMRI Functional Magnetic Resonance Imaging FoR Frames of Reference GraPHIA Graphical Phonological Humor Identification Algorithm
GTVH HG HWT IA ICC ID IEGs IQs IR ITFs JAPE
Global Theory for Verbal Humor Hunting–Gathering Hidden Words Test Irrigation Agriculture Immuno-Cytochemical Intra-unit Distinctiveness Immediate Early Genes Intelligent Quotients Incongruity-Resolution Inducible Transcription Factors Joke Analysis and Production Engine KE Knowledge Extraction KR Knowledge Resources LGN Lateral Geniculate Nuclei LH Left Hemisphere LIBJOG Light Bulb Joke Generator LOT Locating Objects Test LRFB Left-Right-Front-Back LTAG Lexicalized Tree-Adjoining Grammar LTM Long-term Memory MEG Magneto-encephalography MRI Magnetic Resonance Imaging mV Membrane Potential NIMHANS National Institute of Mental Health and Neurosciences NIRS Near Infrared Spectrometry NMDA N-methyl D-aspartate NSEW North-South-East-West OET Object Enumeration Test OVS Object Verb Subject PDA Pushdown Automata PET Positron Emission Tomography RA Relative Absolute RH Right Hemisphere
xvi RHD RLFB RPM RT SES SFB SLD SOA SPEFT SPM
List of Abbreviations
Right Hemisphere Damage Right, Left, Front and Back Raven's Progressive Matrices Reaction time socio-economic status Senguin Form Board Specific Learning Disability Stimulus Onset Asynchrony Story–Pictorial Embedded Figures Test Standard Progressive Matrices
SR SRT SSTH STM TMS TRN TTX UWT VCT VSCCs WM
Success Rate Syllogistic Reasoning Test Semantic Script Theory of Humor Short-term Memory Transcranial Magnetic Stimulation Thalamic Reticular Nucleus Tetrodotoxin Unfamiliar Words Test Visual Closure Test Voltage-sensitive Calcium Channels Working Memory
Preface
I
n the last four decades, cognitive science has established itself as a truly interdisciplinary science. Cognitive science is an intellectual enterprise that studies cognition and seeks to answer many fundamental and long-standing questions about the nature of mind and mental processes. The central assumption for cognitive science can be phrased as follows: ‘The human mind is a complex system that receives, stores, retrieves, transforms and transmits information’. Here the mind is conceived as an information processor. Herbert Simon, a pioneer in cognitive science and artificial intelligence defines cognitive science as ‘recognition of a fundamental set of common concerns shared by the disciplines of psychology, computer science, linguistics, economics, epistemology and social sciences generally’. Cognitive science is the multidisciplinary scientific study of cognition and its role in intelligent agency. It examines what cognition is, what it does, and how it works. An important way in which cognitive science approaches the mind is to view the scientific study of the mind in terms of three levels of analysis proposed by David Marr for understanding cognition, namely, computational, algorithmic and implementational levels. While there have been disagreements about the three levels of analysis and the way they are related to each other, these levels of analysis provide a framework for understanding and studying cognition. Palmer and Kinchi discuss three assumptions of information processing approach: informational description, recursive decomposition and physical embodiment. These assumptions enable us to discuss mental states as informational events, specify an informational event at one level in terms of component informational events at a lower level, and use the concept of representations (states of the system that carry information) and processes (changes in states/representations). Cognitive science employs qualitatively different research tools such as formal methods used to develop computational proofs, the programming techniques of computer science, the experimental practices of psychology, and a variety of paradigms (single-cell studies, lesions, neuroimaging) of neuroscience. Given the current advances, it is expected that cognitive science will become even more inter-disciplinary. Cognitive science is not yet a flourishing discipline in India. Under the Universities Grants Commission (UGC) Scheme of Universities with Potential for Excellence, the University of Allahabad was selected for developing ‘Behavioural and Cognitive Sciences’ as an Island of Excellence. As a follow-up, the University established the Centre for Behavioural and Cognitive Sciences (CBCS) in 2002, for providing education of merit and distinction in line with new developments and challenges, as a constructive opportunity for advancement of scientific knowledge through basic and applied research and teaching as well as outreach programmes. The objectives of the academic programme are to provide
xviii Advances in Cognitive Science comprehensive training and prepare the students for a professional/research/academic career, to develop a richer understanding of mental processes and neural mechanisms underlying cognition using behavioural, computational and neurophysiological techniques. All aspects of behavioural and cognitive sciences are explored to understand the nature of cognitive and information processing system as well as to explore possible applications for the individual and the society. The faculty and students at the Centre are involved in research programmes pertaining to vision, attention, perception, linguistics, cognitive neuroscience, consciousness, cognitive disorders, cognitive modelling and human computer interaction. There is a strong emphasis on research projects and exposure to various theoretical and experimental studies in cognitive science. The Centre and the University provide an ideal environment for study and research in cognitive science. The International Conference on Cognitive Science was held on 16–18 December 2004 at the CBCS. The conference was the first of its kind in focussing on all aspects of cognitive science. The mission of the conference was to explore the truly inter-disciplinary nature of cognitive science and create awareness of cognitive science among the interested students and researchers. The conference served as the meeting point for scientists from interfacing disciplines like psychology, neuroscience, computer science, linguistics and philosophy. The conference elucidated current research on significant areas interfacing cognitive science such as language processing, culture and cognition, perception, cognitive disorders, consciousness, computational neuroscience, memory, social cognition and so on. The technical programme comprised six keynote lectures, 59 oral presentations and 22 poster presentations. The keynote lectures were presented by six prominent experts in cognitive science like Prof. Aravind K. Joshi, University of Pennsylvania, USA, Prof. Jyotsna Vaid, Texas A and M University, USA, Prof. Ype Poortinga, University of Tilburg, Netherlands, Prof. Avi Chaudhuri, McGill University, Canada, and Prof. John W. Berry, Queen’s University, Canada. In addition to faculty members from various Indian universities and abroad, research scholars from various Indian and foreign universities attended the conference. Based on the initial submissions of papers and abstracts for the conference, the editors requested the authors to submit full papers for the volume. All the editors reviewed the papers and 27 contributions were selected for publication in the current volume. Some contributions had to be excluded due to the stringent review process. The contributors are senior leading as well as young talented cognitive scientists from various countries including the USA, Canada, the UK, Germany, the Netherlands, Belgium, Switzerland, Denmark, Australia, New Zealand, Lebanon, Japan and India. The volume contains research articles addressing the challenges faced in cognitive science requiring crosslinking of different interfacing disciplines like psychology, neuroscience, computer science, linguistics and philosophy. The recent findings from cognitive science presented in the volume will serve as a useful resource for scientists working in the area of cognitive science. The volume represents a
Preface
xix
good sample of the current trends in major sub-disciplines in cognitive science. The book contains six sections: (1) Cognitive processes (2) Cognitive neuroscience (3) Computational modelling (4) Culture and cognition (5) Cognitive development and intervention, and (6) Consciousness. The first section contains eight chapters and focuses on the study of various cognitive processes ranging from vision and social cognition. The second section on cognitive neuroscience contains three chapters on cognitive neuroscience of vision and language as well as brain-behavioural asymmetry. The third section on computational modelling has four chapters spanning computational neuroscience and computational linguistics. The fourth section focuses on culture and cognition. This section has seven chapters discussing methodological and theoretical issues as well as related set of cross-cultural studies on spatial development. The fifth section on cognitive development includes four chapters discussing abnormal cognitive development and cognitive rehabilitation. The final section appropriately concludes with a chapter by Prof. K.R. Rao on consciousness with an Indian perspective. The editors would like to acknowledge the efforts of many people who have contributed so much to the conference and preparation of this volume. Our colleague Dr Bhoomika R. Kar helped tremendously in organizing the conference and without her dedication the conference would not have been such a success. We would like to thank all the CBCS staff and students as well as research scholars from the Department of Psychology who worked very hard for the conference. The editors thank SAGE Publications for bringing out this volume. Narayanan Srinivasan A.K. Gupta Janak Pandey
SECTION
I
Cognitive Processes
T
he domain of cognitive science includes all the mental processes that are grouped under the term ‘cognition’ (Bechtel and Graham, 1998). These processes enable us to perform all the tasks that we do from mundane tasks to complex problem solving. Performing even a simple task involves perceiving and identifying stimuli, retrieving relevant information and making appropriate decisions. Perceptual processes are used to analyse and categorize relevant stimuli, attentional processes are used to select stimuli for further processing or action due to limited capacity, events are encoded and remembered, and various behaviours are performed depending on motivation and context. Scientific approaches to study cognition include various methodologies such as behavioural experiments, computational modelling and simulation, and cognitive neuroscience (ibid., 1998). This section focuses mostly on cognitive processes including attention, memory, language and social cognition through behavioural experiments, observation and stimulus analysis. One of the important aspects of performing any complex behaviour is to sequence events in an appropriate manner. Even at early stages, cognitive scientists focussed on sequencing of information and planning of action (Miller et al., 1960). Sequencing of information and actions is a fundamental human ability that underlies several intelligent activities and behaviours. It is known that acquiring a complex sequential skill involves chaining a number of primitive actions to make the complete sequence. The first chapter by Pammi et al. discusses learning of sequences using behavioural experiments demonstrating the hierarchical organization. They present the results obtained by analysing behavioural parameters measured, while subjects performed complex sequential skill by trial and error process. The behavioural results point out spontaneous reorganizational differences during the performance of complex sequence learning tasks. The Pammi et al. study reveals that hierarchical organization of complex movement sequences is dependent on working memory (WM). Hierarchical organization is more likely when the amount of information processed at any point of time is well within the WM capacity. Our visual system utilizes selective mechanisms that help to direct the resources towards relevant stimuli. These selection mechanisms form a significant aspect of visual attention and affect the processing and perception of visual stimuli (Johnson and Proctor, 2004). Initial attempts explored the loci of selection, and theories of attention were proposed that were characterized as early or late selection theories of attention. Attention has been also characterized as selection for perception or selection for action. It has been a critical and well-researched topic in psychology. Several theories have been proposed to explain the attentional phenomena. One of the most famous early theories was the filter theory, which argued that attention works like a filter and processes the relevant information while excluding the others. There were two major positions on the stage at which the filter operated in cognition. One position argued that filtering/selection occurs in the early stages of processing and is based on the physical properties of information, whereas the other view argues that filtering occurs only after the semantic analysis of the information.
4 Advances in Cognitive Science All the theories of attention focus on the selective aspect of attention and therefore it is important to study the types of selection processes in attention. Traditionally, it was believed that the stimulus is selected on the basis of the location only which is called ‘space-based attention’. But recent research has shown that a stimulus can be selected on the basis of the object itself with or without consideration of location and this is termed as ‘object-based attention’. In space-based selection, the selection occurs on the basis of position or location in the space and in object-based selection, the selection takes place on the basis of object properties such as colour, shape and size. Another major aspect of attention is sustained attention or vigilance. Sustained attention is needed for maintaining and operating attentional processes over a period of time. The chapter by Singh et al. focuses on the relationship between selective attention and sustained attention using a cuing paradigm. Singh et al. report results from two experiments examining this relationship. The effects of attention were shown to depend on event rates and also dependent on the age of the participants indicating that attentional performance changes with age. Singh et al. argue in favour of Posner‘s inhibition theory, which proposes the perceptual decrement in vigilance tasks are the result of inhibition generated due to orienting of attention. There is a strong relationship between attention and WM. Working memory corresponds to the short-term maintenance of information as well as processing of information in the WM. The chapter by Kamal Uddin et al. explores the interaction between attention and memory by focussing on foveal bias. Foveal bias is a phenomenon in which the memory for location of a peripheral target is biased towards the fovea. They examine two explanations for foveal bias, namely, attention shift and memory averaging. Kamal Uddin discusses two experiments in which shifts in attention towards a landmark (a black bar) was induced by flashing (Experiment 1) and vanishing (Experiment 2) the bar nearest the target with a variable stimulus onset asynchrony (SOA). In both the experiments, the task was to point to the remembered location of the target. They found that foveal bias was dependent on the landmark conditions with smaller foveal bias in the flashed and vanished landmark conditions compared to the landmark absent condition. Kamal Uddin et al. conclude that the results support the ‘attention shift’ explanation rather than the ‘memory averaging’ mechanism for foveal bias pointing to the importance of attentional processes in memory tasks. One of the significant problems in learning and memory is related to the issue of stability and plasticity in the context of new learning. Stability is related to the maintenance of existent knowledge and plasticity is related to the ability to learn new information. How does new information from the world integrate with the already existing knowledge? Associative learning is one of the mechanisms that have been proposed for integration of new information with concepts previously stored in memory. Memory can be characterized in various ways (Baddeley, 1990). A significant aspect of a memory is whether the memory is explicit or implicit. Explicit or procedural memory refers to the conscious recollection
Cognitive Processes
5
of events and is composed of two systems (episodic and semantic memory systems). Explicit memory plays a critical role in our day-to-day life and enables us to remember and perform various tasks. Implicit memory is memory without accompanying awareness, and is typically shown in fairly automatized skills, conditioning and priming. Compared to explicit memory, implicit memory is less understood, and amnesic patients continue to retain their implicit memory even after losing their explicit memory. The contribution by Rajaram and Coslett focuses on new verbal associative learning through implicit memory in amnesic patients. The research enables the understanding of the role played by implicit memory, as well as identifying the brain areas involved in new verbal learning. The next three contributions focus on language and language processing. Various aspects of language processing interact with other perceptual and action systems for effective performance. Reading is one such aspect of language processing, in which the visual system is used to obtain linguistic information. Researchers on reading have focussed on the interaction of the visual system, especially the eye-movement system with reading passages. The chapter by Knoeferle and Crocker focuses on the close link between fixations and utterances using eye-tracking. While it is accepted that there are interactions between visual and linguistic processing, it is not clear how the visual and linguistic processing are coordinated temporally. Questions also remain on the relative role played by visual and language processes for sentence comprehension. Knoeferle and her colleagues have shown tight synchronization between utterance comprehension, attention in the scene, and the influence of scene information on comprehension using eye-tracking experiments. It has been shown that scene information plays a larger role than linguistic knowledge in sentence comprehension. Knoeferle and Crocker, through their eye tracking experiments, provide a view on the nature of interaction between visual and linguistic processes in sentence comprehension. While Knoeferle and Crocker (this volume) focus on sentence comprehension and scene information, the chapter by Sproat and Padakannaya explores reading by focussing on various language scripts. They develop a script index for Indian languages, especially Devanagari and Kannada scripts. This is based on earlier work on the formal model of layout in writing systems, in which graphic elements are conjoined by two dimensional catenation operators. It has been shown that segmental awareness and reading with Indian writing systems are dependent on layout and diacritization of symbols. In the context of these findings, Sproat and Padakannaya explore the performance on metaphonological awareness tasks based on the script index. All the languages of the world contain compound words and compound words present in one way of producing new words in a language. The chapter by Libben focuses on language comprehension by looking at compound words from various languages, including European and Asian languages. Libben argues that research on compound words have implications for the way humans process morphological information. Libben focuses on two key aspects related to compound words. One is the way in the individual constituent
6 Advances in Cognitive Science words present in a compound word are activated and the other is the way morphemes are arranged within compound words. Libben claims that many of the key properties of compound comprehension derive from the manner in which constituents are isolated in online parsing. He argues that this parsing allows for the activation of all potential compound constituents, but favours left-branching morphological structures over rightbranching ones. So far, all the contributions in this section have focussed on empirical research in the laboratory, and the chapter by Sinha looks at important aspects of social cognition among animals in the wild. Social cognition is an important aspect of all cognition and behaviour among social animals (Kunda, 1999). Social cognition studies processes that are involved in how we make sense of social events and how we use cognitive processes to interact with other members of the society. Some important topics include stereotypes and knowledge about others’ minds. Not only humans, but other animals also live in societies, and knowledge about others’ minds plays a large role in cognition and behaviour. Sinha focuses on social cognition in primates and uses the theoretical framework of intentional stance for his research on animal cognition. According to this framework, each individual is an intentional system capable of mental states like beliefs (Dennett, 1987). This framework is utilized for research on social cognition in wild bonnet macaques (Macaca radiata). Humans are said to have a ‘theory of mind’, which enables attribution of mental states to oneself and others. Sinha investigates the possible presence of the ‘theory of mind’ mechanism among primates by focussing on processes involved in social cognition (decision-making and deception) and the development and social relationships among primates. Sinha questions whether the primates truly show higher order intentionality based on his observation of these primates in the wild.
REFERENCES Baddeley, A. 1990. Human Memory: Theory and Practice. Needham Heights, Massachusetts: Allyn & Bacon. Bechtel, W. and G. Graham. 1998. A Companion to Cognitive Science. Oxford: Basil Blackwell. Dennett, D.C. 1987. The Intentional Stance. Cambridge, Massachusetts: MIT/Bradford Books. Johnson, A. and R. Proctor. 2004. Attention: Theory and Practice. Thousand Oaks, California: Sage Publications. Kunda, Z. 1999. Social Cognition: Making Sense of People. Cambridge, Massachusetts: MIT Press. Miller, G.A., E. Galanter and K.H. Pribram. 1960. Plans and the Structure of Behavior. New York: Henry Holt & Co.
Chapter 1 Hierarchical Organization of Complex Visuo-Motor Sequences V.S. Chandrasekhar Pammi, S. Bapi Raju, Ahmed, K.P. Miyapuram and Kenji Doya
INTRODUCTION
A
Karl Lashley pointed out way back in 1951, understanding the problem of serial order is of utmost importance for studies of cognition. It is well known that when humans acquire skills, initially they are slow and deliberate and eventually become fast and automatic (Fitts, 1964). Sequence learning is amenable for experimentation and hence taken as a paradigm to probe various aspects of skill acquisition (Clegg et al., 1998). It is known that acquiring a complex sequential skill involves chaining a number of primitive actions to make the complete sequence. The notion of chunking in the context of limited capacity of short-term working memory (WM) was introduced by Miller, back in 1956. Miller (1956) suggested that the unrelated list items became reorganized into chunks in order to overcome the limitation of the size of immediate memory buffer. In the current terminology, this buffer is called as short-term or WM buffer (Baddeley, 1986, 1992). Miller suggested that the capacity of short-term memory (STM) or WM is 7 ± 2 chunks and argued that the amount of information that is retained in STM/WM is independent of the information contained in each chunk. The hierarchical organization of movement sequences has been suggested by several researchers. Rosenbaum et al. (1983) first demonstrated the evidence for hierarchical control of the execution of movement sequences. A sequence might consist of several sub-sequences and these sub-sequences in turn can contain sub-sub-sequences. In their experiment, subjects were required to perform a previously memorized finger-tapping sequence of responses as quickly and as accurately possible. Their latency and error data of the movement sequences suggested a hierarchical model, which resembled a tree-traversal-process. Though Karl Lashley in his classic paper on serial order (1951) also argued that the sequential responses that appear to be organized S
8 V.S. Chandrasekhar Pammi et al. in linear and flat fashion concealed an underlying hierarchical structure, Rosenbaum et al. (1983) made notable contributions by actually demonstrating hierarchical control of execution of movement sequences. In a review, Conway and Christiansen (2001) argued that the humans outperform non-human primates on more complex sequential learning tasks—in particular the learning and processing of hierarchically organized temporal sequences. The concept of chunking in sequential behaviour has been studied in animals (see Terrace, 2001 for a review) and in humans (for example, see Koch and Hoffmann, 2000; Pammi et al., 2004; Sakai et al., 2003; Verwey, 2001). Terrace (2001) reviewed mainly the chunking phenomenon in list learning by pigeons, monkeys and humans. He also argued for an operational definition of chunks. He suggested a distinction between the notions of input and output chunks from the ideas of short-term and long-term memory (LTM). Input chunks reflect the limitation of WM during the encoding of new information, that is, how new information is stored in LTM, and how it is retrieved during subsequent recall. Output chunks reflect the organization of over-learned motor programmes that are generated online in WM. He also suggested that during the sequence performance stage (already learnt stage), subjects download list items as chunks during pauses and that chunk size for order information is approximately three items. Koch and Hoffmann (2000) examined the influence of relational structures in the stimulus and in the response sequences on sequence learning using serial reaction task paradigm. In three experiments, they separately studied patterns, chunks and the higher-order relations (that is, hierarchical structures). Verwey (2001), using two highly practised sequences, investigated the robustness of motor chunks (motor representations) in different situations. He suggested a model consisting of dual-processors, that is, cognitive and motor may be involved in the execution of sequences. The familiar sequences are carried out by a dedicated motor processor and the forth coming sequence predictions are done by cognitive processor. Based on earlier studies, they stated that the dorsal prefrontal cortex and anterior cingulate might have a role in the cognitive processor, and the motor processor comprises of the supplementary motor area, the basal ganglia and the lateral cerebellum. Using a 2 × 10 sequence task, in a recent study Sakai et al. (2003) have demonstrated that different subjects chunked the same sequence of movements differently. They have also shown that performance on a shuffled sequence after learning was less accurate and slower when the chunk patterns were disrupted than when they were preserved. This clearly suggests an operational role for the chunks as a single memory unit that facilitates efficient performance of the sequence. Most of the studies reviewed in this section suggest that the shorter reaction time (RT) between successive responses / actions during the sequence performance stage could possibly be due to the spontaneous reorganization (in the form of chunks) across the elements of sequence. On trial-wise data on the same paradigm used in this chapter, Pammi et al. (2004) recently demonstrated the chunking phenomenon using graphical visualization and quantification through clustering analysis. The usefulness of simple clustering analysis on the cumulative reaction times (RTs) in revealing the complete hierarchical structure is also an important contribution of this study.
Hierarchical Organization of Complex Visuo-Motor Sequences
9
The current study specifically addresses the differences in spontaneous reorganization (chunking process) while subjects performed complex sequential skills. In this study, the amount of information to be processed at a time forms a set. We hypothesized that while spontaneous reorganization across sets facilitates efficient performance of the sequence, increasing the set-size would limit across-set reorganization of the sequence. The main aim of this investigation is to study the hierarchical organization of sequential skills. Studies such as these may have an impact in understanding higher cognitive functions such as language, speech, and so on, all of which involve sequence processing.
MATERIALS AND METHODS Visual stimuli consisting of illuminated squares on a 3 × 3 grid were projected on a mirror in front of the subject while they lay supine inside functional Magnetic Resonance Images (fMRI) scanner. Visual cues were presented on a grid formed by nine square cells with black border arranged into a 3 × 3 matrix and shown on a dark grey background. Generation and presentation of the stimuli were controlled by custom-developed software running on a Macintosh computer. Subjects responded by pressing corresponding keys on a keypad. We utilized an on–off (boxcar) design in which subjects performed alternating control and test conditions. In the control condition subjects followed randomly generated visual targets and thus there was no learning involved. In the test condition, they learned a sequence of key-presses arranged either as a 2 × 12 or 4 × 6 or 2 × 6 sequence. In this study, we used the explicit m × n visuo-motor sequence learning paradigm (Bapi et al., 2000; Hikosaka et al., 1995) in which subjects progressively learn by trial and error the correct order of pressing m keys (called a set) successively for n times (called a hyperset). Eighteen human subjects were tested on three sequence tasks—2 × 12, 4 × 6 and 2 × 6. We thus varied sequence complexity along two dimensions m and n, reflecting short and long-range prediction loads. Here m varied from 2 to 4 between the 2 × 6 and 4 × 6 tasks, whereas n varied from 6 to 12 between the 2 × 6 and 2 × 12 tasks.
Experimental Paradigm Subjects performed 2 × 12, 4 × 6 and 2 × 6 sequence learning tasks (see Figure 1.1 for 2 × 12 task procedure). Each experiment consisted of four sessions and a session comprised 13 epochs of alternating control (seven) and test (six) conditions, each lasting 36 and 18 sec, respectively. Each epoch began with relevant instruction screen lasting for 6 sec. A random hyperset was generated for each subject that remained fixed during the experiment. To reduce the possibility of any explicit structure or patterns in the sequence, the hyperset was generated such that any repetition or transposition of sets did not occur. Subjects were given 0.8 sec on an average per key-press. However, they were allowed to proceed
10
V.S. Chandrasekhar Pammi et al.
immediately to the next set as soon as they completed one set. The presentation was reset to the beginning of the hyperset upon an error. Moreover, subjects were instructed to use their index, middle and ring fingers for the three columns of the keypad—left, middle and right columns, respectively. The generation of sequences, presentation of visual stimuli and recording of user responses were carried out with custom-built software running on a Macintosh computer. FIGURE 1.1
The 2 × 12 task procedure
Note: Subjects learned by trial and error to press m (=2) keys successively for n (=12) times in response to visual stimuli on a 3 × 3 grid display. On successful completion of a set, subjects were allowed to progress to the next set and the trial is reset to the beginning of the hyperset upon error.
Data Analysis While subjects performed the tasks inside an fMRI scanner, two behavioural measures, namely, the success rate (SR) and the key-press RT were recorded throughout the experiment. For each epoch, the software calculated SR as ratio of the number of successful hypersets or part thereof completed to the total number of hypersets attempted (expressed as percentage) and key-press RT as the average time required to complete a key-press in a successfully completed set. Repeated-measures analysis of variance (ANOVA) was performed on these two behavioural measures. Data from one subject was not included in the analysis, as significant improvements in key-press RTs were not observed in one of the experimental tasks.
RESULTS The average learning-related improvements in all the tasks and the effect of increasing complexity (based on RTs) are described in a single plot with the two behavioural measures
11
Hierarchical Organization of Complex Visuo-Motor Sequences
(Figure 1.2). This plot also indicates the complexity effect as resulted in slower learning in both the complex tasks (2 × 12 and 4 × 6) as compared to the simple task (2 × 6). Repeated-measures ANOVA was performed, entering the number of experiments (3: 2 × 12, 4 × 6 and 2 × 6), number of sessions (4) and number of blocks (6) as within-subject factors.
Success Rate A significant main effect was observed for experiment F(2,32) = 37.01, p < 0.0001, sessions F(3,48) = 159.21, p < 0.0001 and blocks F(5,80) = 21.19, p < 0.0001. The Experiment x Session interaction term was significant F(6,96) = 8.36, p < 0.0001. The post hoc means comparison revealed significant differences between the 2 × 6 and the 2 × 12 [F(1,32) = 37.98, p < 0.0001], 2 × 6 and 4 × 6 [F(1,32) = 68.57, p < 0.0001] tasks. On the other hand, there was marginally significant difference between the 2 × 12 and 4 × 6 tasks [F(1,32) = 4.49, p = 0.042]. The session-wise post hoc means comparison was done to further probe the above mentioned differences across the three experimental tasks. For each of the four sessions (S1, S2, S3, S4) there was significant difference between the 2 × 6 and 2 × 12 tasks and between 2 × 6 and 4 × 6 tasks (p < 0.0005). Interestingly, the two complex tasks 2 × 12 and 4 × 6 were different in the first session [F(1,96) = 13.77, p < 0.0005] but were similar in sessions 2 [F(1,96) = 3.687, p = 0.0577], 3 [F(1,96) = 1.794, p = 0.183] and 4 [F(1,96) = 0.203, p = 0.653]. The percentage of success obtained for the three experiments in each of the sessions is tabulated in Table 1.1 and the block and session-wise improvements are presented in Figures 1.3 and 1.4 respectively. It is apparent from Table 1.1 that SR attained almost similar values in 2 × 12 and 4 × 6 tasks by the late sessions. It is interesting to note that the SR in the first session is more in 2 × 12 than compared to the success in 4 × 6. The difficulty subjects faced in the first session of 4 × 6 task clearly points out that the short-range prediction load imposed by the 4 × 6 task might have slowed them down the performance. TABLE 1.1
The percentage change of SR for the four sessions
Session
2 × 12
4×6
2×6
1
44.19 ± 13.68
34.03 ± 14.66
69.51 + 15.91
2
71.64 ± 6.26
66.38 ± 5.14
85.57 ± 3.56
3
77.19 ± 3.01
73.52 + 2.29
87.86 ± 1.75
4
78.84 ± 3.21
77.61 + 1.92
88.09 ± 1.08
Note: The standard deviation indicated is computed as variability across each session.
12 FIGURE 1.2
V.S. Chandrasekhar Pammi et al.
The block-wise combination graph
Note: Contains both the average SR and key-press RTs separately for the three experiments with the error-bars of standard deviation across the subjects.
Hierarchical Organization of Complex Visuo-Motor Sequences FIGURE 1.3
13
Block-wise improvements of SR for all the three experiments
Note: The error-bar indicates the standard deviation across subjects. FIGURE 1.4
Session-wise improvements of SR for all the three experiments
Note: The error-bar indicates the standard deviation across sessions. The statistical significance values are also shown (*** is highly significant p < 0.0001; * is significant p < 0.001 and NS is non-significant.
Key-Press RT A significant main effect was observed for experiment F(2,32) = 41.68, p < 0.0001, sessions F(3,48) = 152.47, p < 0.0001 and blocks F(5,80) = 28.34, p < 0.0001. The Experiment x Session interaction term was significant F(6,96) = 4.67, p = 0.0003. The post hoc means comparison revealed significant differences between the 2 × 6 and the 2 × 12 [F(1,32) =
14
V.S. Chandrasekhar Pammi et al.
41.662, p < 0.0001], 2 × 6 and 4 × 6 [F(1,32) = 77.783, p < 0.0001] tasks. On the other hand, there was marginally significant difference between the 2 × 12 and 4 × 6 tasks [F(1,32) = 5.592, p = 0.024]. The session-wise post hoc means comparison was done to further probe the earlier mentioned differences across the three experimental tasks. For each of the four sessions (S1, S2, S3, S4) there was significant difference between the 2 × 6 and 2 × 12 tasks and between 2 × 6 and 4 × 6 tasks (p < 0.0001). Interestingly, the two complex tasks 2 × 12 and 4 × 6 were similar in the first session [F(1,96) = 0.000535, p = 0.982], but were different in sessions 2 [F(1,96) = 7.19, p < 0.01], 3 [F(1,96) = 11.567, p < 0.001] and 4 [F(1,96) = 28.360, p < 0.0001]. The key-press RT improvements in the three experiments for each of the sessions is tabulated in Table 1.2 and the block and session-wise changes are presented in Figures 1.5 and 1.6, respectively. It is interesting to note that the RTs in 2 × 12 and 4 × 6 were similar in the first session. It is evident from Table 1.2 that RTs became dissimilar between 2 × 12 and 4 × 6 tasks by the last session. The RT value in 2 × 12 is also observed to be lesser compared to the RTs of 4 × 6 task. So when 24 movements are arranged as 2 × 12, the single key-press RTs were shorter compared to when they are arranged as 4 × 6. These differences may have potential implication in organizational differences across 2 × 12 and 4 × 6 tasks. To test for any differences in the chunking phenomenon across different sequence tasks, we performed ANOVA on average set completion times taking the data from all trials in the last session in which subjects completed the hyperset successfully. We assumed that the chunking patterns would have stabilized by the fourth session. ANOVA was performed for each experiment separately with set as main effect. ANOVA of set completion times (across subjects) revealed significant main effect for set in the 2 × 6 [F(5,96) = 7.93, p < 0.0001] and 2 × 12 [F(11,192) = 2.32, p = 0.01] experiments, but not for the 4 × 6 experiment [F(5,96) = 0.23, p = 0.95]. It is interesting to note that there are significant pauses in 2 × 12 and 2 × 6 but not in 4 × 6, reinforcing again the fact that chunking was facilitated in 2 × 6 and 2 × 12 tasks, but not in the 4 × 6 task. In summary, the SR and the key-press RT in complex sequence tasks (2 × 12 and 4 × 6) demonstrated the complexity effects when compared with simple task (2 × 6) and the differential effects across the complex tasks. As the learning progressed, the SR in 2 × 12 and 4 × 6 tasks were different in the first session, and finally became similar by the last session. This may point out the similar level of difficulty while performing the complex tasks, and the differences in the first session may indicate the short-range WM load imposed by the 4 × 6 task. The key-press RT also behaved differently in the first session, but as the time progressed, RTs eventually became significantly different. The RT values in the last session of 2 × 12 task appeared to be significantly smaller as compared to RT in the 4 × 6 task. This may point out reorganization differences, while the subjects performing complex tasks arranged in two ways. As the number of sets to be internalized (12) is larger than the STM capacity, it appears that subjects compressed information into a number of chunks.
Hierarchical Organization of Complex Visuo-Motor Sequences TABLE 1.2
15
The key-press RTs for the four sessions
Session
2 × 12
4×6
2×6
1
0.362 ± 0.011
0.362 ± 0.011
0.304 ± 0.047
2
0.306 ± 0.018
0.329 ± 0.013
0.224 ± 0.009
3
0.268 ± 0.013
0.297 ± 0.008
0.195 ± 0.007
4
0.228 ± 0.007
0.274 ± 0.010
0.175 ± 0.003
Note: The standard deviation indicated is computed as variability across each session. FIGURE 1.5
Block-wise data of key-press RTs for all the three experiments 0.6
Key-press RT
0.5 0.4 2 × 12 0.3
4×6 2×6
0.2 0.1 0 1
3
5
7
9
11
13
15
17
19
21
23
Sequence blocks
Note: The error-bar indicates the standard deviation across subjects. FIGURE 1.6
Session-wise data of key-press RTs for all the three experiments
Note: The error-bar indicates the standard deviation across session. The statistical significance values are also shown (*** is highly significant p < 0.0001; ** is more significant p < 0.0001; * is significant p < 0.001 and NS is non-significant).
16
V.S. Chandrasekhar Pammi et al.
These results may suggest long-range optimization process in the 2 × 12 task by chaining/spontaneous reorganization across the sets in the form of chunks. This result also suggests that the chunking phenomenon across sets is less likely to happen in the 4 × 6 task, possibly due to increased short-range load.
CONCLUSIONS AND FUTURE WORK In the current study, we presented behavioural results revealing the differential behaviour of spontaneous reorganization (chunking) while performing complex sequential skill. The behavioural results of the current study and the clustering analysis on RTs from our earlier study (Pammi et al., 2004) point out that a hierarchical model that learns sequences using a limited capacity WM would need to optimize in two different ways depending on the amount of information to be processed at any instance of time. If the amount stretches to the limit of WM, then optimization process needs to operate within the logical unit (set). If the amount is well within the WM capacity, optimization across the logical units (sets) would facilitate efficient performance. Recent fMRI investigations by Owen (2004) and Todd and Marois (2004) also support the effects of the limit of WM capacity that we pointed out in our results. In the current study, we have also acquired fMRI from all the subjects while they performed the m × n tasks. The analysis of fMRI data comparing the 2 × 12, 4 × 6 and 2 × 6 tasks, which is in progress, is expected to reveal brain mechanisms underlying differential cognitive requirements of complexity. Based on findings from the behavioural and neuroimaging studies, we are planning to augment these efforts by applying Machine Learning and Neural Network techniques for building functional models that elucidate the brain function. Our current computational hypothesis (inspired from our recent behavioural results, see Pammi et al., 2004) is that a hierarchical neural network with two levels, one that extracts first-order markov chains and the other that clusters these at a higher level using Reinforcement Learning, is a plausible model of how the brain might deal with the problem of learning complex sequences. The limitation of the current study is that we could not record the RTs for every key-press, but we only measured at the resolution of a set. In future, we would like to experiment with recording each individual key-press times along with the choice and movement times also. This will enable us to understand the differential chunking phenomenon in a comprehensive fashion.
ACKNOWLEDGEMENTS We thank Dr Kiranmayi S. Bapi for helpful discussions on statistical analysis. We also would like to thank Dr Kazuyuki Samejima, Computational neurobiology lab, ATR computational
Hierarchical Organization of Complex Visuo-Motor Sequences
17
neuroscience laboratories, Kyoto, Japan for help with conducting the experiments. The grants from Japan Science and Technology Agency (JST), Japan under the ERATO and CREST schemes for conducting the experiments are gratefully acknowledged. Pammi Chandrasekhar and Ahmed would like to thank CSIR, New Delhi, India for the Senior Research Fellowship.
REFERENCES Baddeley, A. 1986. Working Memory. Oxford: Claredon Press. ———. 1992. ‘Is Working Memory Working? The Fifteenth Bartlett Lecture’. Quarterly Journal of Experimental Psychology, 44A: 1–31. Bapi, R.S., K. Doya and A.M. Harner. 2000. ‘Evidence for Effector Independent and Dependent Representations and their Differential Time Course of Acquisition During Motor Sequence Learning’. Experimental Brain Research, 132: 149–62. Clegg, B.A., G.J. DiGirolamo and S.W. Keele. 1998. ‘Sequence Learning’. Trends in Cognitive Sciences, 2(8): 275–81. Conway, C.M. and M.H. Christiansen. 2001. ‘Sequential Learning in Non-Human Primates’. Trends in Cognitive Science, 5(12): 539–46. Fitts, P.M. 1964. ‘Perceptual Motor Skill Learning’. In A.W. Melton (ed.), Categories of Human Learning (pp. 243–85). New York: Academic Press. Hikosaka, O., M.K. Rand, S. Miyachi and K. Miyashita. 1995. ‘Learning of Sequential Movements in the Monkey: Process of Learning and Retention of Memory’. Journal of Neurophysiology, 74: 1652–61. Koch, I. and J. Hoffmann. 2000. ‘Patterns, Chunks and Hierarchies in Serial Reaction-Time Tasks’. Psychological Research, 63: 22–35. Lashley, K.S. 1951. ‘The Problem of Serial Order in Behavior’. In L.A. Jeffress (ed.), Cerebral Mechanisms in Behavior (pp. 112–36). New York: Wiley. Miller, G.A. 1956. ‘The Magical Number Seven Plus or Minus Two: Some Limits on Our Capacity for Processing Information’. The Psychological Review, 63: 81–97. Owen, A.M. 2004. ‘Working Memory: Imaging the Magic Number Four’. Current Biology, 14: R573–74. Pammi, V.S.C., K.P. Miyapuram, R.S. Bapi and K. Doya. 2004. ‘Chunking Phenomenon in Complex Sequential Skill Learning in Humans’. In N.R. Pal, N Kasabov, R.K. Mudi, S. Pal and S.K. Parui (eds), Lecture Notes in Computer Science, (3316: 294–99). Heidelberg: Springer-Verlag. Rosenbaum, D.A., S.B. Kenny and M.A. Derr. 1983. ‘Hierarchical Control of Rapid Movement Sequences’. Journal of Experimental Psychology: Human Perception and Performance, 9: 86–102. Sakai, K., K. Kitaguchi and O. Hikosaka. 2003. ‘Chunking During Human Visuomotor Sequence Learning’. Experimental Brain Research, 132: 149–62. Terrace, H. 2001. ‘Chunking and Serially Organized Behavior in Pigeons, Monkeys and Humans’. In R.G. Cook (ed.), Avian Visual Cognition. Medford, Massachusetts: Comparative Cognition Press. Todd, J.J. and Marois R. 2004. ‘Capacity Limit of Visual Short-Term Memory in Human Posterior Parietal Cortex’. Nature, 428: 751–54. Verwey, W.B. 2001. ‘Concatenating Familiar Movement Sequences: The Versatile Cognitive Processor’. Acta Psychologica, 106: 69–95.
Chapter 2 Orienting Attention and Cued Sustained Attention Indramani L. Singh, Pamela M. Greenwood and Raja Parasuraman
INTRODUCTION
A
ttention is an important area in cognitive psychology. The ability of human beings to focus their attention on a specific location in a visual field in order to detect a critical signal has been the subject of many studies in the area of attention. Several attempts have been made to study complex cognitive phenomena in terms of basic mental operations in the domain of attention. Researchers proposed three aspects of attention, viz., selective attention (the ability of a subject to select a target stimulus among many other competing stimuli), divided attention (the ability to share attention efforts between two tasks or sources of stimuli) and sustained attention (the capacity of a subject to stay alert for a specific target event for a prolonged period of time). These aspects of attention are not considered independent of one another; they are closely related to the point of affecting each other (Bahri, 1990). The inter-relationship between selective attention and sustained attention has been examined in this chapter.
PROBLEM OF VIGILANCE N.H. Mackworth (1950), a pioneer researcher of sustained attention, suggested decrement function, or the progressive decline in vigilance performance over time periods (see Figure 2.1). He used a simulated radar display, which was known as the ‘clock test’. It consisted of a rotating black pointer, 6" long. The pointer moved 0.3" in either direction, but occasionally it moved double, that is, 0.6" in both sides. The subjects were required to detect the double jump of pointer on either side, and report it by pressing a switch.
Orienting Attention and Cued Sustained Attention
19
The experimental duration was four 30-minutes duration and 12 signals (double jump of pointer) occurred in a 30-min block (Davies and Parasuraman, 1982). Several other studies also revealed that the decline in performance was completed in approximately 20 to 35 minutes, and at least half of final loss occurred during the first 15 minutes (Teichner, 1974). Parasuraman and Davies (1977) identified two tasks, that is, event rate and target discrimination type factors, through a taxonomic analysis of vigilance tasks that could be a cause for decrement in vigilance performance across time. FIGURE 2.1
Percentage of correct detection as function of two-hour time period
FACTORS AND THEORIES OF VIGILANCE DECREMENT Vigilance tasks have often been regarded as boring and monotonous. However, Hart and Staveland (1988) found that vigilance tasks were more demanding in terms of mental workload. According to Dember and Warm (1979), vigilance performance is affected by first-order psychological factors such as sense modality, signal conspicuity and event rate, as also by second-order factors that is temporal-spatial uncertainty and the knowledge of results. Studies have shown that vigilance decrement was more pronounced under high event with successive discrimination type of tasks (Davies and Parasuraman, 1982; Parasuraman, 1979; See et al., 1995). Nonetheless, perceptual sensitivity decrement could also be obtained in high event with simultaneous discrimination type task (Nuechterlein et al., 1983). The vigilance performance of an observer can be analysed into three task-relevant phases, viz., (i) storing background information; (ii) observing and decision-making; and (iii) functioning of neural attention units (Jerison and Pickett, 1963). The various theories
20
Indramani L. Singh et al.
of sustained attention have approached the vigilance decrement function in terms of one or more of these phases. Most theories of vigilance attempt to explain the underlying mechanism of sustained attention, causing perceptual decrement. Mackworth (1950) attributed the vigilance decrement as the accumulation of inhibition due to lack of reinforcement or negative conditioning (Matthews et al., 2000). Similarly, J.F. Mackworth (1969) suggested that performance decrement occurs because of neural habituation process due to stimulus repetition. Broadbent (1958) proposed filter theory and he attributed vigilance decrement as the rejection of repetitive information. Observing rate hypothesis theory (Schroeder and Holland, 1968) suggested that head and eye movements could be a cause for perceptual vigilance decrement. Baker (1963) proposed expectancy theory to explain perceptual decrement in vigilance performance. According to him, observers act as temporal averaging instruments and form expectancies as to the approximate time course of critical signal appearances. Thus, regular signal presentation would facilitate vigilance performance. The arousal theory (Duffy, 1962) of sustained attention suggests that vigilance task of monotonous nature reduces the arousal level, which could lead to a decline in the efficiency of an observer. Broadbent and Gregory (1963) showed that vigilance decrement in hit rate was due to a change in the criterion shift () rather than in the perceptual sensitivity (' ). Swets (1977) suggested that sensitivity decrement occurs under high event rate in presence of other factors like fatigue, boredom, stress, and so on. Similarly, Parasuraman (1979) demonstrated that perceptual sensitivity declines under high event rate with successive discrimination task type.
ORIENTING OF ATTENTION AND SUSTAINED ATTENTION Posner et al. (1984) endeavoured to explain the possible relationship between selective attention and sustained attentions. They conducted two sets of experiments. In the first set of experiments, a cue was displayed at the central location and targets and non-targets were presented at the left- or right-periphery. Posner et al. were interested to examine whether a passive process could maintain the selectivity where targets/non-targets appeared at a non-cued location (periphery)? The same type of task and design was used in the second set of experiments. In this set, subjects were required to orient their attention to the cued side and return to the central fixation point after detection. Targets followed either the peripheral cue or shifting of the subject’s attention to the fixation point. Posner et al. suggested that inhibition was caused by repeating the display of the same target in the same location, which could create a hindrance in selectivity phenomenon of selective attention. They further concluded that the sustained attention in which stimuli and responses were presented in blocks was both inhibited and facilitated under the same conditions in which trial-by-trial cues were used. Parasuraman (1985) suggested a decrement in vigilance performance as a result of sharing ‘primary vigilance’ with other sources of activating events over time periods under a high event rate condition. However, such sharing process is not required at a low
Orienting Attention and Cued Sustained Attention
21
event rate, as the observer can easily process both sources of events. Parasuraman also suggested a close relationship between selective and sustained attention, based on the Posner et al. (1984) findings that sustained attention was changing in terms of facilitation and inhibition under the same conditions. According to these two points of view, the relation between sustained and selective attention is like two different sides of the same phenomenon. A cue that facilitates and inhibits sustained attention actually improves selective attention. Thus, human subjects cannot select a critical signal without being vigilant (sustained attention). Contrarily, one cannot be vigilant, if he is not aware (selective attention) about the multiple external stimuli which surround him. If it happens to be so, a question arises: is there any vigilance decrement over a prolonged period of time, if sustained attention task is cued? Does the vigilance decrement function differ for the un-cued and cued tasks? Does the facilitation or inhibition of sustained attention improve selective attention? These issues have been addressed by combining the covert orienting and vigilance paradigms in the two studies that are described below.
EYE MOVEMENTS AND ATTENTION Several researchers reported that orienting of attention might be independent of eye movements (Eriksen and Hoffman, 1973; Posner et al., 1978; Shaw, 1978). Recording of eye movements showed that subjects detected the critical signals displayed at the periphery while they fixated their eyes on the centre of the visual field. Kosnik et al. (1985) further supported this finding, which allowed their subjects to track a moving object in order to make discrimination. Results revealed that while tracking the movement of stimuli, the subjects’ eye movements were similar to fixating the eye at the centre. Thus, results of covert attention indicated that visual attention shifted even without necessarily moving the eyes (Posner and Cohen, 1982). Eye movements are also not found to affect sustained attention performance over time periods (Coates et al., 1972). In sum, most studies showed that covert orienting and eye movements are independent phenomena (Maylor, 1985; Posner, 1978, 1980; Prinzemetal et al., 1986).
FACILITATION AND INHIBITION Facilitation refers to an enhancement in detection accuracy and/or decrease in reaction time (RT). Inhibition is characterized by a slowing in the response latency (Maylor, 1985; Posner, 1978, 1980; Posner and Cohen, 1982; Prinzemetal et al., 1986). Facilitation occurs when subjects detect more targets with fewer errors and with speeded RT when attention is directed by a valid cue. However, if the target does not occur after 300 ms or more from the display of the visual cue, the facilitation effect is then changed to an inhibitory one and this inhibition lasts about 1 to 1.5 sec (Posner and Cohen, 1982). Posner and Cohen
22
Indramani L. Singh et al.
concluded that inhibition is an automatic process, which occurs inevitably in response to any visual stimulus presented in the periphery and without any deliberate behaviour on the part of the subject. However, Maylor (1985: 189) argued against the theory that facilitation is an attentional process, whereas inhibition is an automatic process, which is sensory rather than attentional in nature. Maylor further stated that ‘the facilitatory and inhibitory components of externally controlled orienting appear to act together to direct the eye movement system and to maintain selectivity in visual space’. These studies also suggested that cues had both facilitatory and inhibitory effects depending on temporal factors. Moreover, Remington and Pierce (1984) also found similar facilitation effects for cued targets, which developed at early stage than did costs (inhibition) for un-cued target locations. They found high RT for neutral and un-cued target than for validly cued target at 150 ms and 200 ms stimulus onset asynchrony (SOA). Bahri (1990), a pioneer researcher in the area of cued vigilance, examined the relationship of orienting of attention and sustained attention. He conducted two experiments and tested 40 subjects in the first experiment and 45 subjects in the second experiment. Subjects performed a three 10-minute block of sensory vigilance task in both the experiments. In experiment 1, two event rates, a high of 30 events, and a low of 15 events, were presented with a 350 ms SOA. The second experiment was a replication of the first experiment, the only exception being that slow event rate and three levels of SOA were used. Results revealed facilitation effect on sensitivity criterion in the 30-minute task with valid cue at low event rate condition (see Figure 2.2). No cue validity benefit was found at high event rate. He further found no significant difference between valid and invalid cues at the short interval (150 ms), but greater sensitivity decrement over time was found with valid than with invalid cues at the 350 SOA. However, there was more decrement with invalid than with valid cues at the long SOA that is, 550 ms. Parasuraman et al. (1989) have suggested that the elderly experienced a greater vigilance decrement than that of the young on hit rate performance. They also reported that old adults showed high RT in the processing of information at invalidly cued location than young adults (Greenwood and Parasuraman, 1994). Recently, Singh et al. (2004) examined the effect of ageing on the relationship of covert and sustained attention on sensory vigil task over prolonged periods of time. In this study, two basic issues of cognitive processes were examined: (i) whether directed attention (valid cues) was facilitated (Bahri, 1990), or (ii) inhibited (Posner et al., 1984) by sustained attention performance with two cue target intervals. Two experiments were conducted on young and old adults. Besides these two experimental conditions, two control conditions were also completed. Four hypotheses were tested in this study: (i) the young subjects would receive more benefits of cued-target than would old adults; and (ii) the detection of targets over non-targets would be facilitated in cued vigilance task; (iii) the phenomenon of facilitation would change over time periods; and (iv) the vigilance decrement would be greater with validly cued than it would be with invalidly cued targets at 450 ms as compared to that of low 300 ms.
Orienting Attention and Cued Sustained Attention FIGURE 2.2
23
Perceptual sensitivity as function of event rate and cue validity
Twenty young students of the Catholic University of America, USA, with a mean age of 21.75 years and 20 old adults with a mean age of 67.81 years participated in this study. Besides these 40 subjects, 29 more participants were also assigned in two control conditions that is, neutral and no cue condition. Fifteen subjects (seven young with mean age of 20.5 years and eight old adults with mean age of 67 years) participated in neutral cue and 14 subjects (seven young with mean age of 21.5 years and seven old adults with mean age of 65.5 years) were employed in no cue control condition. Two squares of different sizes were used as target and non-target. The target square was 3.30 cm2 and the non-target was 2.80 cm2. Subjects were required to detect only bigger square (target) over non-target square immediately after presentation by pressing the space bar. Target was displayed randomly either to the left or right periphery of the screen with a probability of 0.2 (20 per cent). An arrow oriented towards left or right indicated the location of target 80 per cent of the time called valid cue and other cue which could not provide the location of the targets 10 per cent of the time was named invalid. The arrow with both directions also did not indicate the location of targets 10 per cent of the time, and was called neutral cue. All visual cues were presented at the centre of the screen prior to the presentation of targets or non-targets. A 2(young and old adults) × 2(low and high event rates) × 3(valid, invalid and neutral cues) × 3(time periods) mixed factorial design was employed with repeated measures on the last three factors. Two event rates, that is, 15 events and 30 events per minute were used as low and high event conditions, respectively. Three types of visual cues, namely valid, invalid and neutral were manipulated in this study. The time periods consisted of three 10-minute blocks in each low and high event rate condition. Correct detection (hit rates), incorrect detection (F.As.) and RT were recorded for each of the six 10-minute blocks. The same procedures were followed for other two control conditions.
24
Indramani L. Singh et al.
The analysis of variance (ANOVA) of the sensitivity index showed that old adults were poor in detecting the targets with valid, invalid and neutral cues than their young counterparts. However, both young and old subjects were significantly benefited with valid cues more than with invalid or neutral cues (see Figure 2.3). The main effect of the event rate was significant, which demonstrated more cue benefits (facilitation) in low event rate than it did in high event rate. The interaction of event rate × cue validity and event rate × cue validity × time periods indicated that the sensitivity index performance was higher in low event rate with validly cued targets than it was in high event rate. The cued target facilitation at low event was maintained over time periods (see Figures 2.4 and 2.5). FIGURE 2.3
Sensitivity index scores as a function of cue validity
The main effect of event rate on RT performance further manifested that cue benefits were more pronounced at low event than they were at high event rate condition, and that this effect was maintained across the blocks. Results further revealed the superiority of low event over high event rate with validly cued target than invalidly and neutral cued targets. The obtained perceptual sensitivity index in neutral control condition demonstrated that subjects showed high sensitivity performance at low event rate than they did at high event rate. Similarly, RT performance was also low at low event rate than high event rate and an increment in RT performance was observed over time periods. However, these differences were not significant. No significant difference was observed between low and high event condition across sessions for young and old adults on sensitivity index scores in no cue condition (control condition 2). Similar results were also found on RT performance measure.
Orienting Attention and Cued Sustained Attention
25
FIGURE 2.4
Sensitivity index scores as function of cue validity and block in low event rate
FIGURE 2.5
Sensitivity index scores as function of cue validity and block in high event rate
26
Indramani L. Singh et al.
A second experiment was conducted to determine the effect of SOA, the time interval between the cue and the stimulus, with directed or undirected conditions on vigil task detection performance. It was predicted that more decrement in sensitivity index scores would emerge in low SOA over time periods than would in high SOA level. Experiment 2 was a replication of Experiment 1, with two exceptions: (i) only the low event rate was used, since Experiment 1 showed that cue validity effects were maximized at low event rate; and (ii) two different SOAs were used in young and old adults. Correct detection (hit rates), incorrect detection (F.As.) and RT responses were recorded for each six 10-minute blocks. Forty subjects, 20 young and 20 old adults participated in this experiment. The 40 subjects were randomly assigned in two SOA groups: short SOA of 300 ms and long SOA of 450 ms. The procedure and the apparatus were exactly the same as in Experiment 1, with the exception of the variation across groups in time interval between the cue and the target/non-target stimuli. The design in Experiment 2 was a 2(young and old adults) × 2(low SOA and high SOA) × 3(valid, invalid and neutral cue validity) × 2(two 30-minute blocks) × 3(three 10-minute time periods) with repeated measures on the last three factors. ANOVA revealed that age group showed more facilitation of cued target in young than it did in old adults. This trend of results was maintained across six 10-minute time periods, which revealed that young subjects performed better than did old adults over time. Similarly, more benefit of cue validity was associated with young subjects across six blocks. The main effect of SOA showed higher sensitivity index scores at 300 ms SOA than it did at 450 ms (see Figure 2.6). A similar trend on RT performance was also observed. FIGURE 2.6
Mean sensitivity index scores under 300 and 450 SOA
Orienting Attention and Cued Sustained Attention
27
The general findings of the orienting attention suggested that subjects moved their attention to the cued location and generated an expectation (focussed attention). In cued target condition, subjects engage attention, which facilitates the detection accuracy of targets over non-targets. And if the target appears at the un-cued location, the subject has to disengage attention and shift to the target location, which results in slowing in speed or increasing accuracy in discriminating targets from non-targets (Posner et al., 1978). The above findings confirm our prediction that the cue validity effect would also stand in for old subjects on sustained attention tasks. However, the maximum benefit of validity was associated with young subjects. It was also found that validly cued target enhanced sensitivity index (d’), while invalidly and neutral cued targets lowered sensitivity. Results also supported our hypothesis that a validly cued target would facilitate sensitivity performance (increase in hit rates and decrease in false alarms), and invalidly and neutrally cued targets would lead to inhibition on vigilance tasks. The obtained findings also confirm our hypothesis that sensitivity decrement would be higher with a valid cue than with a invalid and/or neutral cue across time periods. The results of the high event rate are consistent with Parasuraman’s (1985) prediction, which stated that sensitivity decrement occurred for both validly and invalidly cued targets than for low event rate. Results further support Posner et al. (1984) theory for the low event rate and the Parasuraman (1985) theory for the high event rate conditions. Finally, cued facilitation at short SOA and inhibition at long SOA also confirm our last prediction. The obtained results are consistent with other researchers (Posner, 1980; Posner et al., 1984), who suggested that cognitive factors provide some logical criteria that assist in the selection and, therefore, the detection of the sensory evidence. These cognitive factors (central factors) help sensory processes, which suggest an interactive relationship between the two processes, that is, selective and sustained attention. Results also indicated that facilitation and inhibition are relatively robust.
CONCLUSIONS The results of the present studies add to our knowledge about the role of pre-cueing in sustained attention task performance. Results showed that once attention was oriented to the target location (allocation) detection of signals was enhanced (facilitated). However, an inhibitory effect was created when subjects were cued to the opposite side of target location (de-allocation), resulting in decreased detection rate and increased false alarms. These findings also support the contention of Bashinski and Bacharach (1980), who suggested that a de-allocation of attention resulted in an inhibition of sensitivity in unattended spatial location. The findings further provide evidence for greater decrement under facilitation than under inhibition on sustained attention task. The cue validity effects in terms of facilitation and inhibition seem to change over time but more decrement observed under the facilitation effect. However, facilitation generally appears in short
28
Indramani L. Singh et al.
cue target intervals than it shows in long intervals. The present study offers evidence that supports Posner et al. (1984) and Parasuraman’s (1985) claims suggesting a close relationship between selective and sustained attentions. However, this claim was confirmed only at the low event rate. In sum, the present research has contributed to the enrichment of different theories of attention in general and sustained attention in particular. This study has shed light on three important issues of attention: (i) the orienting of sustained attention; (ii) the effects of facilitation and inhibition on vigilance task performance; and finally (iii) the ageing effect on cued sustained attention task performance.
ACKNOWLEDGEMENTS This research was supported by the research grants from the National Institute of Health, and the National Institute on Aging to Pamela Greenwood and Raja Parasuraman and by the University Grants Commission and the Defence Research and Development Organisation (DRDO) to Indramani L. Singh. A part of the results was presented by the first author in the 3rd International Symposium on Cognition and Education, organized by the Department of Psychology, Banaras Hindu University, Varanasi, India on 14–18 December 1995. Authors wish to thank Prof. C.B. Dwivedi, Department of Psychology, Banaras Hindu University for his comments and valuable suggestions.
REFERENCES Bahri, T. 1990. Orienting of Attention and Vigilance. Unpublished Doctoral Dissertation, Catholic University of America, Washington, DC. Baker, C.H. 1963. ‘Further Toward a Theory of Vigilance’. In B.N. Buckner and J.J. McGrath (eds), Vigilance: A Symposium (pp. 127–70). New York: McGraw Hill. Bashinski, H.S. and V.R. Bacharach. 1980. ‘Enhancement of Perceptual Sensitivity as the Result of Selectively Attending to Spatial Locations’. Perception and Psychophysics, 3: 241–48. Broadbent, D.E. 1958. Perception and Communication. London: Pergamon. Broadbent, D.E. and M. Gregory. 1963. ‘Vigilance Considered as Statistical Decision’. British Journal of Psychology, 54: 309–23. Coates, G.D., M. Loeb and E.A. Alluisi. 1972. ‘Influence of Observing Strategies and Stimulus Variable on Watch Keeping Performance’. Ergonomics, 15: 379–86. Davies, D.R. and R. Parasuraman. 1982. The Psychology of Vigilance. London: Academic. Dember, W.N. and J.S. Warm. 1979. The Psychology of Perception. New York: Holt, Rinchart and Winston. Duffy, E. 1962. Activation and Behaviour. New York: John Wiley. Eriksen, C.W. and J.E. Hoffman. 1973. ‘Selective Attention: Noise Suppression or Signal Enhancement?’ Bulletin of the Psychonomic Society, 4: 587–89.
Orienting Attention and Cued Sustained Attention
29
Greenwood, P.M. and R. Parasuraman. 1994. ‘Attentional Disengagement Deficit in Non-Demented Elderly Over 75 Years of Age’. Aging and Cognition, 1: 188–202. Hart, S.G. and L.E. Staveland. 1988. ‘Development of NASA-TLX (Task Load Index): Results of Empirical and Theoretical Research’. In P.A. Hancock and N. Meshkati (eds), Human Mental Workload (pp. 139–83). Amsterdam: Elsevier. Jerison, H.J. and R.M. Pickett. 1963. ‘Vigilance: A Review and Re-Evaluation’. Human Factors, 5: 211–38. Kosnik, W., J. Fikre and R. Sekuler. 1985. ‘Improvement in Direction Discrimination: No Role for Eye Movements’. Perception & Psychophysics, 38: 554–58. Mackworth, J.F. 1969. Vigilance and Habituation: A Neuropsychological Approach. Harmondsworth: Penguin. Mackworth, N.H. 1950. ‘Researches on the Measurement of Human Performance’. In H.W. Sinaiko (ed.), Selected Papers on Human Factors in the Design and Use of Control System (pp. 174–331). New York: Dover. Matthews, G., D. Davies, S.J. Westerman and R.B. Stammers. 2000. Human Performance: Cognition, Stress and Individual Differences. UK: Psychology Press. Maylor, E.A. 1985. ‘Facilitatory and Inhibitory Components of Orienting in Visual Space’. In M.I. Posner and O.M. Marin (eds), Attention and Performance XI (pp. 189–204). Hillsdale, New Jersey: Erlbaum. Nuechterlein, K., R. Parasuraman, and Q. Jiang. 1983. ‘Visual Sustained Attention: Image Degradation Produces Rapid Sensitivity Decrement Over Time’. Science, 20: 327–29. Parasuraman, R. 1979. ‘Memory Load and Event Rate Control Sensitivity Decrements in Sustained Attention’. Science, 205: 31. ———. 1985. ‘Sustained Attention: A Multifactorial Approach’. In O.M. Mavin and M.I. Posner (eds), Attention and Performance XI (pp. 493–511). Hillsdale, New Jersey: Erlbaum. Parasuraman, R. and D.R. Davies. 1977. ‘A Taxonomic Analysis of Vigilance Performance’. In R.R. Mackie (ed.), Vigilance: Theory, Operational Performance and Physiological Correlates (pp. 559–74). New York: Plenum. Parasuraman, R., P. Nestor and P.M. Greenwood. 1989. ‘Sustained Attention Capacity in Young and Older Adults’. Psychology and Aging, 4: 339–45. Posner, M.I. 1978. Chronometric Explorations of Mind. Hillsdale, New Jersey: Erlbaum. ———. 1980. ‘Orienting of Attention’. Quarterly Journal of Experimental Psychology, 32: 3–25. Posner, M.I. and Y. Cohen. 1982. ‘Components of Visual Orienting’. In H. Bouma and D. Bowhuis (eds), Attention and Performance VIII (pp. 531–56). Hillsdale, New Jersey: Erlbaum. Posner, M.I., M.J. Nissen and W.C. Ogden. 1978. ‘Attended and Unattended Processing Modes: The Role of Set for Spatial Location’. In H.L. Pick, Jr. and E. Saltzman (eds), Modes of Perceiving and Processing Information (pp. 137–57). Hillsdale New Jersey: Erlbaum. Posner, M.I., Y. Cohen, L.S. Choate, R. Hockey, and E. Maylor. 1984. ‘Sustained Concentration: Passive Filtering or Active Orienting’? In S. Kornblum and J. Requin (eds), Preparatory States and Processes (pp. 49–68). Hillsdale New Jersey: Erlbaum. Prinzemetal, W., D.F. Presti and M.I. Posner. 1986. ‘Does Attention Affect Visual Feature Integration?’ Journal of Experimental Psychology: Human Perception and Performance, 12: 361–69. Remington, R. and L. Pierce. 1984. ‘Moving Attention: Evidence for Time-Invariant Shifts of Visual Selective Attention’. Perception and Psychophysics, 35: 393–99. Schroeder, S.R. and J.G. Holland. 1968. ‘Operant Control of Eye Movement During Human Vigilance’. Science, 161: 292–93. See, J.E., S.R. Howe, J.S. Warm and W.N. Dember. 1995. ‘Meta-Analysis of the Sensitivity Decrement in Vigilance’. Psychological Bulletin, 117: 230–49.
30
Indramani L. Singh et al.
Shaw, M.L. 1978. ‘A Capacity Allocation Model for Reaction Time’. Journal of Experimental Psychology: Human Perception and Performance, 4: 586–98. Singh I. L., P.M. Greenwood and R. Parasuraman. 2004. ‘Effects of Visual Cues Sustained Attention Task Performance in Young and Old Adults’. Proceedings of the International Conference on Cognitive Science (pp. 206–10) 16–18 December, University of Allahabad. Swets, J.A. 1977. ‘Signal Detection Theory Applied to Vigilance’. In R.R. Mackie (ed.), Vigilance: Theory, Operational Performance and Physiological Correlates (pp. 705–18). New York: Plenum. Teichner, W.H. 1974. ‘The Detection of Simple Visual Signal as a Function of Time on Watch’. Human Factors, 16: 339–53.
Chapter 3 Why Does Foveal Bias Decrease in the Presence of Additional Element? Muhammad Kamal Uddin, Takahiro Kawabe and Sachio Nakamizo
INTRODUCTION
I
dentification and localization of objects are two fundamental aspects of vision. The former is concerned with identifying an object, while the latter is concerned with determining its location in space. Even though the sense of location is a fundamental aspect of the perception of objects, vision research was primarily concerned with the processes of identification. It was not until the late 1960s of the 20th century that an explicit distinction was proposed between two anatomically different neural systems that process either identity or location information (Held, 1968). This early lack of concern with the role of spatial information in visuomotor behaviour has been ascribed to the result of a passive stimulus-response model of perception (Butter et al., 1982). According to the model, the perceiver is a passive recipient of environmental stimuli, and hence not concerned with the processes of localization that guides the active search for new information. The rate at which identity and location information is processed was examined in a number of studies. For example, Dick and Dick (1969) tachistoscopically presented one single letter in one of the four corners of an imaginary square centred on the fixation point. Observers were asked to report either the identity or the location of this letter. Their results showed that observers were more accurate in localizing the letter than in identifying it. Based on this finding, they suggested that perception should be viewed as a hierarchical process where location information is processed before identity information. In a similar study, Smythe and Finkel (1974) found that observers recalled more spatial information than identity information; hence, they argued that spatial information is processed more quickly than identity information, and that different mechanisms underlie spatial and object vision.
32
Muhammad Kamal Uddin et al.
Later studies show that the reproduced location of an object is different from its actual location. Specifically, the memory for location of an object is subject to a number of distortions. An object transiently presented in the retinal periphery is reproduced closer to the retinal centre; a distortion referred to as ‘foveal bias’ (for example, Kerzel, 2002b; Mateeff and Gourevich, 1983, 1984; Musseler et al., 1999; Olivers, 2004; Sheth and Shimojo, 2001; van der Heijden et al., 1999). It has been also reported that the memory is biased towards an additional element; a distortion termed as ‘landmark attraction’ (for example, Hubbard and Ruppel, 2000; Kerzel, 2002a; Sheth and Shimojo, 2001). ‘Landmark repulsion’ as opposed to landmark attraction, is another distortion in which memory for location is biased away from the landmark or distractor (for example, Kerzel, 2002b; Sheth and Shimojo, 2004; Werner and Diedrichsen, 2002). Finally, a distortion reported in the literature as ‘representational gravity’ or ‘cognitive analogs of gravity’ is a tendency to memorize location too far down (for example, Hubbard and Ruppel, 2000; Kerzel, 2002c). Recent studies show that foveal bias does not result from memory averaging,* whereas landmark bias does. van der Heijden et al. (1999) reported that foveal bias occurred regardless of the presence or absence of an actual fixation point. This implies that the fixation point served merely as a cue for fixation and not as a visual landmark (Sheth and Shimojo, 2001; see also Deubel, 2004) suggesting that foveal bias does not stem from memory averaging between the fixation point and the target. Sheth and Shimojo (2001) reported that the frequency of foveal bias decreased in the presence of an additional display element. This implies that the additional element served as a landmark (Hubbard and Ruppel, 2000) suggesting that the memory averaging between target and landmark reduces foveal bias. However, Kerzel (2002b) reported that memory averaging between a target and neighbouring object did not occur. Rather, the target was reproduced away from the object (Werner and Diedrichsen, 2002). Thus, reproduced location or, in other words, memory for location of a target does not appear to be consistently biased towards nontargets. Therefore, we can neither accept nor refute memory averaging as an underlying mechanism for the reduction of foveal bias. Another possible mechanism underlying the reduction in foveal bias is the attention shift towards non-targets. Kerzel (2002a) showed that the vanishing position of a moving target was reproduced towards an object that had abruptly appeared at the time of target disappearance or thereafter. He suggested that attention shifts towards the object might underlie the memory bias. However, contributions of attention shift and memory averaging *When two adjacent objects are to be memorized, each location of the two objects is associated. During association in memory, spatial averaging of the positions of the two objects takes place. Consequently, two objects may be reproduced towards each other. We call this localization bias ‘memory averaging’. Memory averaging predicts that the position of a briefly presented target will be associated with one of the permanently visible landmarks resulting in the localization bias of the target towards it.
Why Does Foveal Bias Decrease in the Presence of Additional Element?
33
were not segregated in his experiments. Moreover, it is unsure whether the same suggestion of attention shift (or memory averaging) modifying the localization performance could be extended to a stationary target as well. Thus, it appears imperative to dissociate attention shift from memory averaging to specify the mechanism underlying reduction in foveal bias. The present study tries to do so by examining the condition in which the landmark nearest the target was spatially cued. In Experiment 1, the landmark was flashed on for 80 ms with a variable stimulus onset asynchrony (SOA). In Experiment 2, the landmark was suddenly vanished instead with the same variable SOA as in Experiment 1. The results provided supporting evidence for attention shift against memory averaging as an underlying mechanism reducing foveal bias.
EXPERIMENT 1 We examined whether or not a shift of visual attention towards the landmark reduces foveal bias by comparing responses in three experimental conditions: with or without flashed landmarks, and without a landmark. We expected that landmark conditions (flashed and non-flashed) would yield lower foveal bias than without landmark conditions. We further expected that the flashed conditions would result in lower foveal bias than non-flashed conditions.
Method Participants Nine graduate psychology students (ST, YT, TS, YY, YM, DK, RI, NN and SH; four females and five males) of Kyushu University volunteered as observers. They were between 23 and 28 years old with a mean age of 26 years and all had normal or corrected-to-normal vision. All observers were naive of the purpose of the experiment.
Apparatus and Stimuli The stimuli were programmed in Delphi 6 with DirectX and displayed on a 19-inch colour CRT monitor (Nanao, Flex Scan T761) with a pixel resolution of 1024 × 768 and refresh rate of 75 Hz. A Sony Video Audio Integrated Operation (VAIO) PC interfaced with the monitor and controlled stimuli presentation and data collection. The target was a black dot with 10 cd/m² luminance subtended 0.33º in diameter. Landmarks were four identical bars with 10 cd/m² luminance subtended 2º in length and 0.2º in width. They were 12º eccentric and placed to the left, right, top and bottom to avoid predictability of the target location. Left and right bars, and top and bottom bars were vertically and horizontally aligned with the fixation mark, respectively. The fixation mark with 1.32 cd/m² luminance
34
Muhammad Kamal Uddin et al.
subtended 1º in length and 0.04º in width and was centred on the screen. The background was black (luminance 0.1 cd/m²). The target was presented randomly at an eccentricity of 3º, 6º or 9º from the fixation mark which is, respectively, 9º, 6º and 3º from the landmark. The target directions were predetermined at 0º, 90º, 180º and 270º in polar angle, where 0º represented the right horizontal direction and the values increased counterclockwise. The two main experimental conditions were ‘with’ and ‘without’ a landmark. The ‘with landmark’ condition was manipulated in two ways by causing the landmark to flash and not flash. In the flashing condition, the landmark was flashed (100 cd/m²) on for about 80 ms with a SOA of 0, 106.4 or 212.8 ms. Thus, a total of five experimental conditions were included in this experiment, which followed a two-factor within the group design (See Figure 3.1). FIGURE 3.1
Schematic representation of the experimental protocol
Time
ON A
OFF
B
C
D
80 ms SOA
80 ms 500 ms Time
Note: A: Four bars serving as landmarks were presented. B: A target (dot) appeared 50 ms after the space key was pressed. C: The landmark nearest the target was flashed for 80 ms (Experiment 1) or vanished until a response was given (Experiment 2). D: The mouse cursor appeared 500 ms after target offset.
Why Does Foveal Bias Decrease in the Presence of Additional Element?
35
Procedure The experiment was conducted in a darkened room. Observers sat 50 cm away from the CRT display and the viewing was binocular. A chin-and-head rest was used to stabilize their visual field and to match their eye level to that of the fixation mark. The experiment was self-paced; observers initiated each trial by pressing the space key while maintaining fixation on the fixation mark. Fifty minutes later a target appeared for 80 ms, during which time observers were required to continue maintaining fixation while memorizing the location of the target. Observers were instructed to continue maintaining fixation until a mouse cursor appeared. After a retention interval of 500 ms following target offset, the mouse cursor identical to the target in all respects appeared at a random location within an imaginary square of 4º sides concentric with the centre of the target. The observers’ task was to position the cursor in the remembered location of the target, then to press the left button of the mouse to record the screen coordinates. During localization, eye movements were allowed. After pressing the mouse button, the trial was terminated and observers again fixed their gaze for the following trial. Observers received six blocks of 48 trials each in a single session lasting about 30 minutes, including breaks between blocks. The first five blocks each consisted of randomly intermixed conditions of four ‘with landmark’ conditions, and the last one only of the ‘without landmark’ condition. The first block was regarded as practice, and hence discarded from the statistical analysis. Thus, each observer performed a total of 240 experimental trials (5 experimental conditions × 3 target eccentricities × 4 target directions × 4 repetitions).
Results and Discussion Basic Data The x and y coordinates of the target presented on the horizontal and vertical axes, respectively, were subtracted from the corresponding coordinates of the respective responses to obtain the magnitudes of the displacements. Positive and negative values are indicative of landmark and foveal bias, respectively. Displacements for each eccentricity were averaged over four target directions and four repetitions to obtain the mean displacement per condition per observer. The mean displacements thus obtained constituted the basic data for the analysis. Thus, a total of 135 (5 experimental conditions × 3 target eccentricities × 9 observers) basic data were available for further analyses. Mean displacements as a function of experimental condition and target eccentricity are plotted in Figure 3.2. An analysis of variance (ANOVA) showed a significant main effect of experimental condition {F (4, 32) = 3.21, p < 0.05} and an interaction effect {F (8, 64) = 2.54, p < 0.05}. Post hoc tests (Ryan’s method) for the pair-wise comparisons of the main effect showed significantly lower foveal bias in the ‘flashed landmark’ condition with a SOA of 106.4 ms than in the ‘without landmark’ condition. Post hoc tests for the simple
36
Muhammad Kamal Uddin et al.
main effect of the interaction between experimental condition and target eccentricity showed that foveal bias in the ‘with landmark’ conditions (flashed and non-flashed) were significantly lower than in the ‘without landmark’ condition at 9º target eccentricity. An ANOVA for flashed landmark conditions showed significant {F (2, 16) = 3.919, p < 0.05} main effect of SOA; with the lowest foveal bias corresponding to 106.4 ms, followed by another with 212.8 ms and the highest foveal bias corresponding to 0 ms SOA. FIGURE 3.2
Mean displacements plotted as a function of 5 experimental conditions
Note: ‘With landmark’ (flashed with SOAs of 0, 106.4 and 212.8 ms, and ‘non-flashed’) and ‘Without landmark’ conditions and 3 target eccentricities (3: filled circle, 6: open circle, and 9: triangle). Each data point was obtained by averaging 144 measurements (4 target directions × 4 repetitions × 9 observers). Vertical bars denote one standard error of mean.
The results are in agreement with the expectation that landmark conditions (flashed and non-flashed) would result in a lower foveal bias than those without landmark conditions. However, they are not in agreement with the expectation that flashed conditions would result in a lower foveal bias than non-flashed conditions. In fact, the magnitude of foveal bias in the flashed conditions was smaller, but not statistically different from that in the non-flashed conditions. This signifies that the mere presence of a landmark drew attention, as did the flashed landmark and reduced foveal bias; however, the flash added an insignificant magnitude of reduction. The results also showed that the effect of
Why Does Foveal Bias Decrease in the Presence of Additional Element?
37
landmark was largest when target was presented closest to it. The results are consistent with the findings reported by Deubel (2004)—that the efficiency of landmarks disappears beyond a horizontal range of three degrees from the target. Moreover, the SOA showing significant reduction of foveal bias was congruent with that causing large cueing effects in the cost-benefit paradigm (Posner, 1980). The results unequivocally suggest that landmark biases localization towards it. However, at this stage, our objective to specify which of attention shift and memory averaging was critical in biasing localization towards landmark is not achieved. A critical drawback to this experimental paradigm was the visibility of landmark in both flashed and nonflashed conditions after the target had disappeared. Accordingly, we cannot argue for the possibility that attention shift to the landmark itself induced spatial shift that was purely the source of the reduction nor can we argue for the alternative that the position of landmark was spatially ‘averaged’ with the target in memory. To resolve the above issue, we employed in the next experiment a new sequence of stimuli in which the landmark nearest the target was suddenly caused to disappear. By examining this condition, we tried to provide evidence for/against attention shift or memory averaging reducing foveal bias. We expected that both disappeared and nondisappeared conditions would yield a lower foveal bias than in cases without landmark conditions, while the former conditions would not differ from themselves if the attention shift account was valid. On the other hand, non-disappeared conditions would yield a lower foveal bias than both disappeared and without landmark conditions, while the latter conditions would not differ from themselves if memory averaging account was valid.
EXPERIMENT 2 Method Participants Nine graduate psychology students (ST, TS, HS, SS, MI, SR, KS, DK and AY; six females and three males) of Kyushu University volunteered as observers. ST, TS and DK also participated in Experiment 1. The participants’ ages ranged from 21 to 28 years with a mean age of 26 years, and all had normal or corrected-to-normal vision. All were extensively experienced in psychophysical experiments; however, they were naive of the purpose of the experiment.
Experimental Conditions The experimental conditions were identical to those in Experiment 1 except for the following: the landmark was made to disappear with the three SOAs used in the ‘flashed landmark’ conditions.
38
Muhammad Kamal Uddin et al.
Apparatus and Stimuli The stimuli were identical to those used in Experiment 1 except for the following: the luminance of the fixation mark, landmark and target was 1.32 cd/m2, while that of the background was 19 cd/m2.
Procedure All procedures were identical to Experiment 1.
Results and Discussion Mean displacements as a function of experimental condition and target eccentricity are plotted in Figure 3.3. An ANOVA showed significant main effects of experimental condition {F (4, 32) = 6.203, p < 0.01} and target eccentricity {F (2, 16) = 11.547, p < 0.015}. Post hoc pair-wise comparisons (Ryan’s method) of the main effect of experimental condition showed that foveal bias in the ‘disappeared landmark’ conditions with a SOA of 106.4 ms and 212.8 ms was significantly lower than in the ‘without landmark’ condition. Pairwise comparisons of the main effect of eccentricity showed a significantly lower foveal bias for targets at 9° eccentricity than those at 3° and 6°. An ANOVA for disappeared landmark conditions showed significant main effect of eccentricity {F (2, 16) = 9.204, p < 0.01} and non-significant main effect of SOA {F (2, 16) = 3.320, p > 0.05}. Following the non-significant SOA effect, we collapsed the data across all disappeared conditions and ran a two-way [3 (mean of disappeared, non-disappeared and no-landmark) × 3 (3, 6, 9 degree of eccentricity)] repeated measures ANOVA, which yielded significant main effects of experimental condition {F (2, 16) = 9.089, p < 0.005} and target eccentricity {F (2, 16) = 12.311, p < 0.001}. Pair-wise comparisons showed that disappeared conditions did not differ significantly from the non-disappeared condition, whereas they both differ significantly from the no-landmark condition (t16 = 4.102, p < 0.001; t16 = 3.057, p < 0.001, respectively). The results clearly show that ‘attention shift’ was the crucial factor that reduced foveal bias. As in Experiment 1, foveal bias was significantly reduced when the SOA was 106.4 ms and additionally when 212.8 ms. In the disappeared landmark conditions, the landmark was no longer available after the disappearance. Therefore, we cannot attribute the significant reduction of foveal bias observed in the disappeared conditions to memory averaging of the target and attended landmark. In addition, both disappeared and nondisappeared conditions differed significantly from the no-landmark condition in this experiment. This latter finding suggests that the common mechanism in disappeared and non-disappeared conditions was attention shift that biased localization towards the landmark, thus reducing the foveal bias.
Why Does Foveal Bias Decrease in the Presence of Additional Element? FIGURE 3.3
39
Mean displacements plotted as a function of 5 experimental conditions
Note: ‘With landmark’ (‘disappeared’ with SOAs of 0, 106.4 and 212.8 ms, and ‘non-disappeared’) and ‘Without landmark’ conditions and 3 target eccentricities (3: filled circle, 6: open circle, and 9: triangle). Each data point was obtained by averaging 144 measurements (4 target directions × 4 repetitions × 9 observers). Vertical bars denote one standard error of mean.
GENERAL DISCUSSION The main purpose of the present study was to clarify how an additional element reduces foveal bias in a manual localization task. We hypothesized that an abrupt change in landmark would draw observers’ attention and hence reduce foveal bias. The results of Experiment 1 showed that an abrupt flash in the landmark nearest the target significantly reduced foveal bias. However, we could not differentiate contributions of attention shift from those of memory averaging. The results of Experiment 2 showed that a sudden disappearance of the landmark, which seemed to draw the observer’s attention, significantly reduced foveal bias, suggesting that attention shift, not memory averaging, plays a key role in reducing foveal bias. The results are consistent with previous reports that an additional element reduces foveal bias. In Sheth and Shimojo (2001), the distracter near the target reduced foveal
40
Muhammad Kamal Uddin et al.
bias. This finding and our results can be commonly explained in terms of attention shift towards the distracter. On the other hand, Kerzel (2002b) reported no reduction in foveal bias. In contrast, they actually found a repulsion effect in which the target location was estimated to be away from the distracter. The discrepancies between the results regarding the reduction of foveal bias might be due to the relative positioning of fixation, target and landmark. In our experiment, the landmark was placed on a line passing through the target and fixation. A previous study demonstrated that such a configuration can reduce the frequency of foveal bias (Sheth and Shimojo, 2001). On the other hand, the distracter placed obliquely to the virtual line cannot affect foveal bias (Kerzel, 2002b). These discrepancies in the results nicely fit with the findings of Tse et al. (2003), which state that in comparison to the no-cue case, the attended region was significantly elongated along the line passing through the cue (with SOA of 106 ms or more) and fixation. The distracter in Kerzel’s study (2002b) was placed orthogonally to the target; as a result the attended region elongated by the abrupt appearance of the distracter was unlikely to encompass the target, and hence had no effects on foveal bias. Thus, the relative positioning of the distracter, the target and the fixation seemed to better explain the discrepancies between the two streams of studies. A different line of thinking is that disruption in the balance of visual space due to changes in saliency might explain the reduction in foveal bias. It has been suggested that the luminance change of the object is salient enough to draw visual attention (Nothdurft, 2002). In our experimental paradigm, both the flash and the disappearance of landmark involved luminance changes that entailed changes in saliency leading to disruption in the balance of visual scene. In the first experiment, we observed a SOA effect which had similar time course as reported by Posner (1980). However, in the second experiment, a prolonged effect of disappeared landmarks was observed. The difference in the time course of landmark effects between the two experiments can be explained by the existence of two types of spatially directed attentions: a transient attention in flashed conditions and a sustained attention in disappeared conditions (Nakayama and Mackeben, 1989). Specifically, the cue employed in Experiment 1 was transiently flashed, while that employed in Experiment 2 remained disappeared until the response. Hence, the flash and disappeared cues might have entailed the transient and sustained attentions, respectively. Thus, the two experimental conditions differing in temporal saliency exhibited differential SOA effects, that is, a significant SOA effect in Experiment 1 and a non-significant SOA effect in Experiment 2. Why was the attention shift so effective in reducing foveal bias? Here, we speculate that a reorganization of visuo-spatial coordinates takes place around an attended salient distracter in visual space. As described in the introduction, foveal bias generally occurs even when observers are not provided with a fixation mark (van der Heijden et al., 1999). It has been proposed that this is because visual space in memory is coded and remapped with a focussed location as a centre of representation (Kerzel, 2002a). In our experiments, the
Why Does Foveal Bias Decrease in the Presence of Additional Element?
41
observers’ attention was shifted towards the landmark that became salient due to flashing and vanishing. Therefore, it was likely that memory of visual space was reorganized with a focussed location (that is the distracter position) as a centre of representation. This idea is consistent with that of Werner and Diedrichsen (2002) that spatial memory was re-mapped on the basis of the distracters’ position. Here, our results newly indicated that the trigger of the remapping might be an attention shift towards the distracter, resulting in the reduction of foveal bias. One may contend that our results might have originated from an artefact of involuntary eye movements to transient changes in the distracter. The transient change is a bottom-up signal that automatically necessitates saccadic eye movements. Although observers were instructed to fixate on the central cross, they might have made saccadic eye movements towards the flashed or disappeared landmark. However, if the eye movements were the source of the reduction in foveal bias, a similar pattern would have been observed across three eccentricities. As it was, a significant reduction was observed only in the 9º eccentricity condition. Therefore, it seems untenable that eye movements were involved in modulating the magnitude of foveal bias in our study. Nonetheless, since eye position is a strong cue for accurate manual localization (Adam et al., 1993; Uddin et al., 2004), it is imperative to address this issue in future research.
ACKNOWLEDGEMENTS This chapter is reprinted from Vision Research, (Vol. 45,pp. 3301–3306) with permission from Elsevier.
REFERENCES Adam, J.J., M. Ketelaars, H. Kingma and T. Hoek. 1993. ‘On the Time Course and Accuracy of Spatial Localization: Basic Data and a Two-Process Model’. Acta Psychologica, 84: 135–59. Butter, C.M., D. Kurtz, C.C. Leiby and A. Campbell. 1982. ‘Contrasting Behavioral Methods in the Analysis of Vision in Monkeys with Lesions of the Striate Cortex or the Superior Colliculus’. In D.J. Ingle, M.A. Goodale and R.J.W. Mansfield (eds), Analysis of Visual Behaviour (pp. 301–34). Boston: MIT Press. Deubel, H. 2004. ‘Localization of Targets Across Saccdaes: Role of Landmark Objects’. Visual Cognition, 11: 173–202. Dick, A.O. and S.O. Dick. 1969. ‘An Analysis of Hierarchical Processing in Visual Perception’. Canadian Journal of Psychology, 23: 203–11. Held, R. 1968. ‘Dissociation of Visual Functions by Deprivation and Rearrangement’. Psychologische Forschung, 31: 339–48. Hubbard, T.L. and S.E. Ruppel. 2000. ‘Spatial Memory Averaging, the Landmark Attraction Effect, and Representational Gravity’. Psychological Research, 64: 41–55.
42
Muhammad Kamal Uddin et al.
Kerzel, D. 2002a. ‘Attention Shifts and Memory Averaging’. The Quarterly Journal of Experimental Psychology. Section A: Human Psychology, 55: 425–43. ———. 2002b. ‘Memory for the Position of Stationary Objects: Disentangling Foveal Bias and Memory Averaging’. Vision Research, 42: 159–67. ———. 2002c. ‘The Locus of ‘Memory Displacement’ is at least Partially Perceptual: Effects of Velocity, Expectation, Friction, Memory Averaging, and Weight’. Perception & Psychophysics, 64: 680–92. Mateeff, S. and A. Gourevich. 1983. ‘Peripheral Vision and Perceived Visual Direction’. Biological Cybernetics, 49: 111–18. ———. 1984. ‘Brief Stimuli Localization in Visual Periphery’. Acta Physiologica et Pharmacologica Bulgarica, 10: 64–71. Musseler, J., A.H.C. van der Heijden, S.H. Mahmud, H. Deubel, and S. Ertsey. 1999. ‘Relative Mislocalization of Briefly Presented Stimuli in the Retinal Periphery’. Perception & Psychophysics, 61: 1646–61. Nakayama, K. and M. Mackeben. 1989. ‘Sustained and Transient Components of Focal Visual Attention’. Vision Research, 29: 1631–47. Nothdurft, H.C. 2002. ‘Attention Shifts to Salient Targets’. Vision Research, 42: 1287–306. Olivers, C.N.L. 2004. ‘Blink and Shrink: The Effect of the Attentional Blink on Spatial Processing’. Journal of Experimental Psychology: Human Perception and Performance, 30: 613–31. Posner, M.I. 1980. ‘Orienting of Attention’. The Quarterly Journal of Experimental Psychology, 32: 3–25. Sheth, B.R. and S. Shimojo. 2001. ‘Compression of Space in Visual Memory’. Vision Research, 41: 329–41. ———. 2004. ‘Extrinsic Cues Suppress the Encoding of Intrinsic Cues’. Journal of Cognitive Neuroscience, 16: 339–50. Smythe, L., and D.L. Finkel. 1974. ‘Masking of Spatial and Identity Information from Geometric Forms by a Visual Noise Field’. Canadian Journal of Psychology, 28: 399–408. Tse, P.U., D.L. Sheinberg, and N.K. Logothetis. 2003. ‘Attentional Enhancement Opposite a Peripheral Flash Revealed Using Change Blindness’. Psychological Science, 14: 91–99. Uddin, M.K., Y. Ninose, and S. Nakamizo. 2004. ‘Accuracy and Precision of Spatial Localization with and without Saccadic Eye Movements: A Test of the Two-Process Model’. Psychologia, 47: 28–34. ———. 2005. Attention Shift not Memory Averaging Reduces Foveal Bias. Vision Research, 45:3301–06. van der Heijden, A.H.C., J.N. van der Geest, F. de Leeuw, K. Krikke, and J. Musseler. 1999. ‘Sources of Position-Perception Error for Small Isolated Targets’. Psychological Research, 62: 20–35. Werner, S. and J. Diedrichsen. 2002. ‘The Time Course of Spatial Memory Distortions’. Memory & Cognition, 30: 718–30.
Chapter 4 New Associative Learning in Amnesia Suparna Rajaram and H. Branch Coslett
INTRODUCTION
M
emory for past experience and future plans underlies much of human existence. This cognitive function connects the past to the present and the present to the future. It also serves as the measure of how we acquire information from the episodes of our daily lives, incorporate it into our existing knowledge, and presumably use it to guide our decisions and behaviour. Therefore, it is no surprise that the loss of memory function has been shown—both in anecdotal reports and in scientific evidence on amnesia—to have devastating consequences on one’s way of life. Human memory is shaped by a number of factors that operate during encoding and retrieval. For instance, factors such as repetition, interference and retrieval cues play a large role in modulating long-term memory (LTM). An understanding of how these factors exert their influence on different forms of memory has occupied centre-stage in the study of human memory. In this field of study, explicit memory tasks such as recall and recognition have long served as measures of investigation. In these tasks, experiment participants are instructed to consciously and deliberately think back on the first encounter, or the study phase, during which the to-be-remembered information was presented (for example, words such as banana, snake or pictures such as strawberry, monkey). In recall, no retrieval cues are available for such retrieval whereas in recognition, studied information is intermixed with non-studied information and participants are asked to identify the studied information. In contrast to explicit measures of memory, a relatively newer procedure involves the use of implicit memory tasks (Graf and Schacter, 1985). As the name implies, the retrieval of studied information in these tasks is indirect because no reference is made to the study phase; instead, participants are provided with different cues depending on the type of task (for example, the word stem completion task—mon______; the word fragment completion
44
Suparna Rajaram and H. Branch Coslett
task—b_n_ _ a; the general knowledge task—what fruit wears its seeds on its skin? the category verification task—Does this belong in a restaurant? Hammer (yes/no?)), and are asked to simply complete or respond to the cues presented to them with the first answer that comes to mind. The advantage for correctly completing or answering the cues for studied information relative to non-studied information serves as the measure of implicit memory or priming. A consideration of different classes of memory tasks—explicit and implicit—is important because the nature of memory and its durability depend not only on the process by which information was encoded during the first encounter, but also by the task used to measure memory. Furthermore, memory tasks can differ not only in their explicit/implicit retrieval demands, but also in the extent to which performance on these tasks relies on different types of processes—conceptual or perceptual (Roediger, 1990; Roediger et al., 1989, for reviews). Thus, any investigation of memory for new associations needs to take into consideration all of these properties of memory tasks. Some of the key distinctions between explicit and implicit memory performance are particularly relevant here (see Roediger and McDermott, 1993, for a review of task distinctions and the influence of independent variables on different classes of tasks.) Explicit memory is vulnerable to numerous factors such as passage of time and interference from other information even though people presumably rely on this form of memory all the time. In contrast, implicit memory, which is by definition not available to conscious experience, is surprisingly immune to the ravages of many variables that impair explicit memory. It usually survives longer, it can often be robust despite interference during encoding, and it underlies performance in a wide variety of situations. These properties make implicit memory a fundamentally important cognitive function and call for a deeper understanding of its nature and its influence on human behaviour. Another striking property of implicit memory is its resistance to brain damage. This is illustrated in a dramatic fashion in cases of human anterograde amnesia. This form of amnesia is characterized by the impairment of LTM for day-to-day events while leaving other cognitive functions such as language, reasoning, problem solving, short-term memory (STM) and skill learning relatively intact. This cognitive disorder is typically caused by neural damage to certain circumscribed regions of the brain. Although amnesics are grossly impaired on explicit memory functions, they perform normally on a variety of implicit memory tests (Warrington and Weiskrantz, 1968, 1970). Intact priming in amnesia is routinely reported when the implicit task provides perceptual cues (mon______; b_n_ _ a), but also frequently reported when the task provides only a conceptual link to the target (what fruit wears its seeds on its skin?) (see Moscovitch et al., 1993 for a review). These features of the amnesic syndrome provide a unique opportunity to better understand the nature of both perceptual and conceptual forms of implicit memory. Conversely, a better understanding of implicit memory has the potential to inform rehabilitative measures of amnesia.
New Associative Learning in Amnesia
45
An impressive amount of empirical evidence has accumulated in the last 30 years on the status and nature of implicit memory in amnesia. But much of this evidence and the associated theoretical advances largely elucidate the nature of implicit memory for information acquired prior to the onset of amnesia. Evidence on the nature of implicit memory for new information encountered after the onset of amnesia remains limited. Since the latter question directly relates to the process by which new information may be incorporated into existing knowledge, the focus of research presented here is on implicit memory for new verbal associations. The use of the amnesic population to study the development of new implicit memory is particularly relevant, because it is notoriously difficult to rule out contamination from explicit memory in the performance of memoryintact individuals. Thus, this investigation has the potential to mutually inform the role of implicit memory in the development of new verbal associations both in the amnesic and the memory-intact populations. In our work (Rajaram and Coslett, 2000a, 2000b), we have investigated the roles of repetition, interference and retrieval cues in new associative learning. Early research on this issue suggested that while amnesic individuals do show intact priming for new wordpair associations (for example, window-reason), this measure is often contaminated by the use of residual explicit memory in many cases of amnesia (Graf and Schacter, 1985; Schacter and Graf, 1986). More recently, evidence for new learning was reported in a profound case of amnesia where the involvement of explicit memory was unlikely (Tulving et al., 1991). In this study, novel sentences (for example, MEDICINE cured HICCUP) were presented to the amnesic patient, KC, over a large number of sessions spaced across several weeks. Despite his failure to consciously remember these sentences later, patient KC produced a substantial proportion of targets (HICCUP) in response to partial sentence cues (MEDICINE cured ???) when his task was to simply complete the sentences with the first answer that came to mind. Such positive evidence for learning in this case study stood in contrast to previously reported null findings (for example, Gabrieli et al., 1988). These contradictory findings could have resulted from several important differences in tasks and procedures across studies. We were motivated in our studies to select the experimental conditions from these reports that would optimally isolate the relevant cognitive processes. In addition to variations in the experimental conditions, differences in previous findings could also have resulted from the differences in the loci of neural damage in patients who participated in these studies. A number of neural regions in the brain have been implicated in mediating explicit memory function, and these include the medial temporal lobes (including the hippocampus), the diencephalon and the basal forebrain region. The medial temporal lobes are specifically implicated in binding the elements of new associations (Cohen et al., 1997; Johnson and Chalfonte, 1994; McClelland et al., 1995). These neurocognitive theories suggest that new associative learning of verbal information is less likely to occur in medial temporal lobe amnesia than in amnesia from
46
Suparna Rajaram and H. Branch Coslett
other aetiologies. We addressed this issue by comparing the performance of amnesic patients with two different aetiologies. In one study, we examined the performance of two amnesic patients, one with selective but extensive medial temporal lobe damage (patient RH) and the other with selective basal forebrain damage (patient CC) across several experiments (Rajaram and Coslett, 2000a). Each experiment involved multiple sessions, each conducted at least one week apart. Within each session, the experiment consisted of three phases—a study phase where novel sentences (MEDICINE cured HICCUP, TRAIN frightened KANGAROO) were presented with instructions to indicate the extent to which each sentence made sense, a five-minute distractor phase, and a test phase where the nature of retrieval cues varied. In some sessions, only the fragment of the target was presented (_ I _ C _ P; _ _ N_A R_O) to assess perceptual priming as a measure of control because this form of priming is usually preserved in amnesia. The most relevant test conditions consisted of sentence cues where perceptual support for the target was either present (MEDICINE cured _ I _ C _ P) or absent (TRAIN frightened ???). The Sentence+??? cues provided the critical test of the question whether implicit memory measures can detect new conceptual, associative learning in all forms of amnesia. Both amnesic patients, RH and CC, showed intact perceptual priming (_ I _ C _ P) and learning with perceptual support (MEDICINE cured _ I _ C _ P) for studied targets. In contrast to these findings, only patient CC with basal forebrain damage demonstrated steady increase in priming for sentence cues that provided no perceptual cues for the target itself (TRAIN frightened ???). Strikingly, the medial temporal lobe patient RH failed to show priming in this condition even after 12 sessions of training. This lack of learning stood in sharp contrast to substantial priming for the same targets he showed when perceptual support for the target was present. These findings demonstrate the critical role of the medial temporal lobes in mediating new associative learning. Of further interest is the nature of stimuli—novel sentences that require semantic processing—that were used in these experiments. As humans are in a unique position to demonstrate acquisition and transfer of verbal, semantic content, these neurocognitive findings with human subjects are particularly significant for our understanding of the role of the medial temporal lobes in binding conceptual associations. At least two inter-related reasons may account for the absence of new associative learning in medial temporal lobe amnesia. Conceptual processes involved in learning and memory are particularly sensitive to effects of interference. For instance, recall—a task that heavily relies on conceptual processes—is impaired by proactive interference caused by information learned prior to the encoding of to-be-remembered information or retroactive interference caused by information learned subsequent to the encoding of to-be-tested information. To the extent that conceptual processes underlie new associative
New Associative Learning in Amnesia
47
learning, a variety of input occurring naturally during the time between different sessions can impair learning. This possibility is especially likely with compromised medial temporal lobe functions, because the suppression of interference is assumed to be a prominent function of these neural structures (Shapiro and Olton, 1994). To address these related issues, we conducted the previously described experiment with a different amnesic patient CV with presumed medial temporal lobe damage. We made one critical change in this study—instead of conducting each study-distractor-test cycle one–three weeks apart, we conducted 10 such cycles in back-to-back sessions (Rajaram and Coslett, 2000b). The successive administration of study-test cycles was intended to reduce interference from intervening events. Despite the critical change that was introduced to reduce interference, patient CV failed to show any learning with Sentence+??? retrieval cues. This outcome converges on our previous findings with patient RH and provides strong support for the proposal that the medial temporal lobes play a special role in mediating the learning of new verbal associations in humans. A different method for reducing interference involves training where no response is required of subjects during the repeated presentations of novel sentences across sessions for encoding. In this Study-Only method, subjects—particularly amnesic patients—cannot produce the incorrect response during early sessions. As a result, incorrect responses do not interfere with the gradual learning of the correct responses. Evidence from amnesic patients indicates that this errorless learning method improves learning in amnesia compared to the Test-Study method where subjects are asked to produce responses in every training session before the correct target is shown (Hamann and Squire, 1995; Hayman et al., 1992). In light of patient CV’s inability to benefit from back-to-back training sessions, we examined whether his performance would improve in the Study-Only condition (that further reduces interference) compared to the Test-Study and Study-Immediate Test (as in the previous experiment) conditions (Rajaram and Coslett, 2000b). Patient CV showed robust perceptual priming for targets in this experiment, but once again failed to produce the targets in response to Sentence+??? cues in all conditions including the critical StudyOnly condition. Together, this series of studies highlights two fundamental features of implicit new associative learning in humans: One, at the cognitive level, perceptual support (provided by the Sentence + Fragment cues) facilitates learning, and two, at the neurocognitive level, the medial temporal lobes play a critical role in mediating relatively more conceptual associative learning. Successful production of targets with Sentence+Fragment retrieval cues by all amnesic patients in our studies highlights the importance of perceptual processes. It remains to be determined whether the elements are also unitized at a conceptual level. In a recent study, we (Rajaram et al., 2005) addressed this question in a heterogeneous group of
48
Suparna Rajaram and H. Branch Coslett
four amnesic patients. We compared the level of target production in three conditions. In two of the three conditions, the perceptual match between the study and test stimuli was held constant, but the conceptual information was varied. For example, in one condition, subjects studied only the target (HICCUP) and were tested on its fragment (_ I _ C _ P) whereas in another condition they studied the entire sentence (MEDICINE cured HICCUP) and were tested with a matching cue (MEDICINE cured _ I _ C _ P). Better completion performance was observed in the latter condition, indicating that performance is indeed improved by the conceptual information provided by the sentence structure. Interestingly, even when the test cue was limited to target fragment alone ((_ I _ C _ P), having studied the target in the context of a sentence (MEDICINE cured HICCUP) led to better performance than in the first condition. Together, this evidence suggests that novel, verbal associative learning involves a conceptual component. The findings from this series of studies have implications for understanding amnesia as well as normal cognition. These findings delimit the cognitive and mnemonic processes that are compromised or preserved in different forms of amnesia. As such, a better understanding of the precise conditions under which verbal associative learning occurs has implications for the development of rehabilitative programmes. With respect to normal cognition, these findings provide a window into the nature of implicit memory for verbal associative learning when explicit memory contributions are minimized. Findings from these studies reveal the need for assessing the distinct contributions of perceptual and conceptual processes in normal associative learning and in the acquisition of knowledge.
REFERENCES Cohen, N.J., R.A. Poldrack and H.E. Eichenbaum. 1997. ‘Memory for Items and Memory for Relations in the Procedural/Declarative Memory Framework’. In A.R. Mayes and J.J. Downes (eds), Theories of Organic Amnesia: A Special Issue of Memory (pp. 131–78). UK: Psychology Press. Gabrieli, J.D.E., N.J. Cohen and S. Corkin. 1988. ‘The Impaired Learning of Semantic Knowledge following Bilateral Medial Temporal-Lobe Resection’. Brain and Cognition, 7: 157–77. Graf, P. and D.L. Schacter. 1985. ‘Implicit and Explicit Memory for New Associations in Normal and Amnesic Subjects’. Journal of Experimental Psychology: Learning, Memory and Cognition, 11: 501–18. Hamann, S.B. and L.R. Squire. 1995. ‘On the Acquisition of New Declarative Knowledge in Amnesia’. Behavioral Neuroscience, 109: 1027–44. Hayman, C.A.G., C.A. Macdonald and E. Tulving. 1992. ‘The Role of Repetition and Associative Learning in Amnesia: A Case Experiment’. The Journal of Cognitive Neuroscience, 5: 376–89. Johnson, M.K. and B.L. Chalfonte. 1994. ‘Binding of Complex Memories: The Role of Reactivation and the Hippocampus’. In D.L. Schacter and E. Tulving (eds), Memory Systems 1994 (pp. 311–50). Cambridge, Massachusetts: MIT Press. McClelland, J.L., B.L. McNaughton and R.C. O’Reilly. 1995. ‘Why There are Complementary Learning Systems in the Hippocampus and Neocortex: Insights from the Successes and Failures of Connectionist Models of Learning and Memory’. Psychological Review, 102: 419–57.
New Associative Learning in Amnesia
49
Moscovitch, M., E. Vriezen and Y. Goshen-Gottstein. 1993. ‘Implicit Tests of Memory in Patients with Focal Lesions and Degenerative Brain Disorders’. In H. Spinnler and F. Boller (eds), Handbook of Neuropsychology (Vol. 8: 133–74.). Amsterdam: Elsevier. Rajaram, S. and H.B. Coslett. 2000a. ‘Acquisition and Transfer of New Verbal Information in Amnesia: Retrieval and Neuroanatomical Constraints’. Neuropsychology, 14: 427–55. ———. 2000b. ‘New Conceptual Associative Learning in Amnesia: A Case Study’. Journal of Memory and Language, 43: 291–315. Rajaram, S., S. Travers, M. Hamilton, M. Jefferson and A. Bolton. 2005. ‘New Associative Learning in Amnesia: The Roles of Encoding Context and Retrieval Cues’, in preparation. Roediger, H.L. 1990. ‘Implicit Memory: Retention without Remembering’. American Psychologist, 45: 1043–56. Roediger, H.L. and K.B. McDermott. 1993. ‘Implicit Memory in Normal Human Subjects’. In J. Grafman and F. Boller (eds), Handbook of Neuropsychology (Vol. 8: 63–131). Amsterdam: Elsevier. Roediger, H.L., M.S. Weldon and B.H. Challis. 1989. ‘Explaining Dissociations between Implicit and Explicit Measures of Retention: A Processing Account’. In H.L. Roediger and. F.I.M. Craik (eds), Varieties of Memory and Consciousness: Essays on Honour Endel Tulving (pp. 3–41). Hillsdale, New Jersey: Lawrence Erlbaum Associates. Schacter, D. L., and Graf, P. 1986. ‘Preserved Learning in Amnesic Patients: Perspectives from Research on Direct Priming’. Journal of Clinical and Experimental Neuropsychology, 8: 727–43. Shapiro, M.L. and D.S. Olton. 1994. ‘Hippocampal Function and Interference’. In D.L. Schacter and E. Tulving (eds), Memory Systems 1994 (pp. 87–117). Cambridge, Massachusetts: MIT Press. Tulving, E., C.A.G. Hayman and C. Macdonald. 1991. ‘Long-Lasting Perceptual Priming and Semantic Learning in Amnesia: A Case Experiment’. Journal of Experimental Psychology: Learning, Memory, and Cognition, 7: 595–617. Warrington, E.K. and L. Weiskrantz. 1968. ‘A New Method of Testing Long-Term Retention with Special Reference to Amnesic Patients’. Nature, 217: 972–74. ———. 1970. ‘The Amnesic Syndrome: Consolidation or Retrieval?’ Nature, 228: 628–30.
Chapter 5 The Coordinated Processing of Scene and Utterance: Evidence from Eye-Tracking in Depicted Events Pia Knoeferle and Matthew W. Crocker
INTRODUCTION
P
eople often find themselves in situations where both spoken language and an immediate scene context are available and relevant. When watching movies, for instance, people are able to rapidly integrate both the utterance they hear, and the events they see. The rapid integration of scene and utterance has been demonstrated experimentally in numerous psycholinguistic investigations. Tanenhaus et al. (1995) demonstrated that a visual referential context influences the initial structuring of an utterance. In instructions such as Put the apple on the towel in the box, the phrase on the towel can be temporarily analysed as modifier and interpreted as the location of the apple (identifying which apple) or it can be attached to the verb phrase and interpreted as a destination (where to put the apple). In a scene containing one apple, the phrase on the towel was preferentially interpreted as destination. In a scene with two apples, people preferentially analysed the phrase on the towel as a modifier of the apple, interpreting it as a location, as evidenced by eye-movement patterns. Crucially, fixation patterns to objects differed from the onset of the utterance depending on the type of referential context (that is, a one-apple versus two-apple context, supporting a destination versus modifier interpretation, respectively). The early difference in gaze patterns suggests that the scene influenced the initial structuring of the sentence very early during comprehension of the utterance. Importantly, the findings by Tanenhaus and colleagues have shown that the informational integration between the language and vision systems is not informationally encapsulated in the Fodorian sense (Fodor, 1983). Fodor postulated strong architectural
The Coordinated Processing of Scene and Utterance
51
restrictions on the informational interaction between distinct cognitive systems such as language and vision. In his model of the mind, distinct input modules such as language and vision have access only to the output of another distinct module, but cannot influence its internal processes. The findings by Tanenhaus et al. (1995) provide strong evidence for the view that processes internal to the language system, such as the structuring of an utterance, can be rapidly influenced by a perceived visual referential context.
PSYCHOLINGUISTIC ACCOUNTS OF LANGUAGE PROCESSING Despite much evidence against a strong version of Fodor’s model of the mind, Fodorian views have influenced psycholinguistic theories of online language comprehension. Scene information has, for instance, not been explicitly included in most psycholinguistic theories or frameworks of language comprehension (Forster, 1979; Frazier and Fodor, 1979; MacDonald et al., 1994). Furthermore, these approaches do not explicitly specify the visual perceptual system as a cognitive system that might proffer important information for comprehension processes. We argue that as a result, such theories do not provide a complete account of language comprehension for situations in which language relates to a scene. A sub-group of these theories by definition excludes the influence of scene information on the initial structuring of a sentence through restricting the informational sources that can influence this process to syntactic information (Forster, 1979; Frazier and Fodor, 1979). In contrast to this restricted position, scene information could, in principle, influence the initial structuring of a sentence in unrestricted interactionist frameworks (MacDonald et al., 1994; Tanenhaus et al., 2000). These frameworks propose that any available and relevant informational source can influence the initial structuring of a sentence. However, just as restricted psycholinguistic theories, they still do not explicitly include scene information as an informational source.
EMBEDDING THE LANGUAGE SYSTEM WITH VISION Recent research on the language system has, in contrast, begun to take into account the fact that scene information can influence core comprehension processes such as the structuring of an utterance, and explicitly embeds the language system in relation to the other perceptual systems (Jackendoff, 2002) (see Figure 5.1) Spatial representations provide information about the shape and location of objects, and are the ‘upper end’ of the visual system (Jackendoff, 2002: 346). The arrows represent interface modules that provide for the communication between distinct structures.
52
Pia Knoeferle and Matthew W. Crocker
Through communication via the interface modules, information from the immediate visual scene can influence language comprehension and the structuring of a sentence. Thus, such a framework should, in principle, be able to account for the influence of visual scenes on the structuring of an utterance. FIGURE 5.1
The Jackendovian architecture of the language system
Source: Jackendoff, 1997: 44.
A PROCEDURAL ACCOUNT OF LANGUAGE PROCESSING While providing representational interfaces between distinct cognitive systems, the Jackendovian framework makes no explicit predictions about precisely how utterance and scene information are integrated. We thus consider whether other frameworks provide a more fully specified procedural account of online sentence comprehension. One such type of framework is the interactionist competitive-integration model (SpiveyKnowlton, 1994; Tanenhaus et al., 2000). While current implementations do not model the integration of scene information, we propose that such a model might, in principle, be extended to incorporate information that has been derived from a visual scene. A sketch of this type of model is provided in Figure 5.2. Processing steps in the model detail how the biases (A, B) are combined, a process that simulates structural ambiguity resolution (between structures A, B). In the first processing step, the activations for the two nodes (A, B) of each bias are normalized. In a second cycle, the activation of the two structures (A, B) is determined by integrating their respective corresponding bias nodes using a weighted sum. In a third step, structure nodes send feedback to the biases depending on how strongly a bias activated that structure. This type of model thus provides a specification of how diverse informational biases combine, and a linking hypothesis, which predicts processing difficulty when constraints that are very similar in strength compete. We think, however, that while the model details how informational biases combine, it is not a model of sentence comprehension. The algorithm in the competitive-integration
The Coordinated Processing of Scene and Utterance
53
model does not determine the construction of an interpretation, for example, how grammatical function and thematic role are assigned to a currently processed phrase. If, instead of modelling how to decide between two alternative structural analyses we aim at actively building a structure, then we need a theory that in some form or other permits us to specify conceptual structures, and mechanisms for how conceptual structures combine. FIGURE 5.2
A sketch of the competitive-integra
A framework that offers a linguistic formalism for the analysis of meaning, and that still provides for the procedural integration of perceptual and linguistic information is Embodied Construction Grammar (Bergen and Chang, 2005). Bergen and Chang show how this linguistic formalism can be integrated into a simulation-based model of language understanding. In their model, sentence comprehension takes place via the activation of embodied schemas (cognitive experience-based structures), and the simulation of motor/perceptual experiences that are associated with these schemas. The Embodied Construction Grammar framework provides a less detailed procedural account than the competitive-integration model. However, it has a stronger linguistic component, which makes it suitable as a framework for a theory of full sentence comprehension. Neither of these two models, however, has so far specified the nature of the interaction between linguistic and visual processes. We therefore take as our starting point the following observation: ‘We agree that constraint-based models need to be more explicit about the nature of the constraints and how they combine [...]’ (Tanenhaus et al., 2000: 94). Specifically, we propose that exploring the precise nature of the interplay between the visual system and comprehension could lead to a more fully specified account (and ultimately theory) of online sentence comprehension in visual scenes.
54
Pia Knoeferle and Matthew W. Crocker
THE NATURE OF THE INTERPLAY BETWEEN SCENE AND UTTERANCE PROCESSING One way of specifying what we term the ‘nature’ of the interplay between visual perception and comprehension is by detailing how mental representations of utterance and scene derive and combine. An important aspect is whether the mechanisms of the systems that process utterance and scene are coordinated. Linguistic and visual processing might, for instance, be asynchronous (that is, not require a common timing reference in order to communicate). Alternatively, the interaction of distinct cognitive processes might be of a more synchronous nature (that is, requiring a common time signal to coordinate them). Support for a tight coordination of linguistic and visual processing comes from Tanenhaus et al. (1995), and Zelinsky and Murphy (2000). Their results show linguistic processing as serving as a ‘signal’ for other cognitive processes such as attention in the scene (for example, Tanenhaus et al., 1995), and duration/frequency of object inspection (for example, Zelinsky and Murphy, 2000). To investigate the time-course of scene–sentence integration during the structuring of an utterance, we reconsider findings by Tanenhaus et al. (1995). Recall that eye-movements to objects differed from the onset of the utterance depending on the type of referential context in their studies. Eye-movements showed, furthermore, that shortly after people had heard a word in the utterance they inspected the real-world referent of that word. The fixation patterns in their studies thus provide evidence for the view that the utterance directs attention in the scene. Second, they demonstrate that there is an early influence of the scene on the initial structuring of an utterance. What the fixation patterns in the studies by Tanenhaus et al. (1995) do not permit to determine is when exactly the scene influenced structuring and interpretation of the utterance. This is because eye-movements for the two context conditions (one apple, two apples) differed from the very start of the utterance. As a result, there are at least two ways of interpreting their data. One interpretation is that people acquired the referential context prior to hearing the utterance, and that there was from the start only one way in which the referential context permitted structuring of the utterance. The second possible interpretation is that comprehension of words during presentation of the utterance directed attention in the scene, and that this guiding function of the utterance was necessary for making scene information available in the first place. Under this view, only after the relevant scene information had been identified through the utterance was it able to influence the structuring of the utterance. While the findings by Tanenhaus and colleagues are compatible with the second interpretation, they do not permit teasing apart the two interpretations. Further to the coordination of the interaction between scene and utterance processing, a fundamental and little-studied question is how important scene and utterance processing
The Coordinated Processing of Scene and Utterance
55
are relative to one another in the integration process. Teasing apart the relative effects of perceived scene information and stored linguistic/world knowledge in online sentence comprehension is of relevance for frameworks of the language system (Bergen and Chang 2005; Jackendoff, 2002). Such an endeavour may ultimately allow us to propose a more fully specified model of processing mechanisms for real-time sentence comprehension in visual scenes.
EXPERIMENTAL EVIDENCE FROM EYE-TRACKING IN DEPICTED EVENTS Experimental findings have indeed revealed a close temporal coordination between mention of relevant scene information in the utterance and the point in time when that scene information influences structural disambiguation. Further, there is evidence that suggests a greater relative importance of immediately depicted events than linguistic/world knowledge during the thematic interpretation of an utterance. Findings by Knoeferle et al. (2004) revealed the coordinated influence of depicted events on structural disambiguation and incremental thematic role assignment by monitoring people’s eye-movements in visual scenes during comprehension of an utterance that related to the scene. Specifically, they examined the time-course with which depicted events (for example, princess-washing-pirate, fencer-painting-princess, see Figure 5.3) enabled structural disambiguation and incremental thematic role assignment of initially structurally ambiguous German subject-verb-object (SVO)/object-verb-subject (OVS) sentences. Listeners heard Die Prinzessin wäscht/malt den Pirat/der Fechter (‘The princess [amb.] washes/paints the pirate [obj./patient]/the fencer [subj./agent]’). Once the verb had identified the relevant depicted action, anticipatory eye-movements in the event scenes provided evidence for expectations of a patient (the pirate) and agent role filler (the fencer) for the initially ambiguous SVO and OVS sentences, respectively. Findings from the third experiment reported by Knoeferle et al. (2004) demonstrated that even when the main verb was sentence-final and did not establish early reference to the depicted events, linguistic cues (temporal adverbs) that appeared prior to the verb still enabled disambiguation. Crucially, this shows that the rapid coordinated influence of depicted agent-action-patient events on online utterance comprehension does not depend upon reference by the main verb. Even when the verb did not make the depicted actions available for early disambiguation and incremental role assignment, soft adverbial cues were sufficient to make the relevant scene information accessible. The rapid, verb-mediated influence of depicted events on incremental thematic role assignment importantly generalizes to another language and sentence structure as revealed by Knoeferle et al. (2003). Results reported in their paper demonstrated the early verbmediated influence of depicted events for online disambiguation of the English main verb/reduced relative ambiguity, thus extending and generalizing findings from the comprehension of German SVO/OVS sentences.
56 FIGURE 5.3
Pia Knoeferle and Matthew W. Crocker
Example image from Experiment 1
Source: Knoeferle et al., 2005.
Further to examining the temporal coordination of scene and utterance processing, Knoeferle and Crocker (2004) investigated the relative importance of depicted events by directly comparing their influence on incremental thematic role assignment to that of stored thematic role knowledge. It has been found that these two informational sources each affect incremental thematic role assignment online. Prior research by Kamide et al. (2003) demonstrated the rapid influence of verb-based thematic role knowledge on incremental thematic role assignment. As detailed above, studies by Knoeferle and colleagues showed the rapid verb-mediated influence of depicted events on incremental thematic role assignment and structural disambiguation. Based on these two sets of findings, Knoeferle and Crocker (2004) carried out an experiment that examined two research questions. The first aim was to see whether we could replicate findings from two studies that had found eye movement evidence for incremental thematic role assignment through stereotypical thematic role knowledge (Kamide et al., 2003) and non-stereotypical depicted events (Knoeferle et al., 2005) within one experiment. To investigate this issue, participants were presented an utterance that uniquely identified only either a stereotypical agent or an agent of a depicted event. An example scene showed a wizard, a pilot and a detective serving food (see Figure 5.4). Sentences were all in OVS order. When people had heard ‘The pilot (object/patient) serves food to …’, perception of the detective who was depicted as serving food to the pilot enabled early incremental thematic interpretation of the detective as the agent of the food-serving action after the verb (no other agent in the scene was a plausible agent for the same, or depicted as serving food). In contrast, after people had heard the beginning of a comparison sentence
The Coordinated Processing of Scene and Utterance
57
(‘The pilot [object/patient] jinxes …’) eye-movements revealed that the wizard was identified as the most plausible agent on the basis of stored thematic role knowledge (no other agent in the scene was a plausible agent for a jinxing action or depicted as performing such an action). This was revealed by a higher proportion of anticipatory eye-movements to the stereotypical agent (wizard) than to the other agent when the verb was ‘jinx’. In contrast, when the verb was ‘serve-food-to’, there were more inspections to the agent of the depicted event (detective) than to the other agent. FIGURE 5.4
Example item from Knoeferle and Crocker (2004)
The second goal of this study was to determine the relative importance of depicted events and verb-based thematic role knowledge. To test this issue, participants heard an utterance in which the verb did not determine uniquely whether the comprehension system should rely on verb-based thematic role knowledge (identifying a stereotypical agent), or on depicted events (identifying an alternative, agent of a depicted event) for thematic interpretation of the utterance (‘The pilot [object/patient] spies-on …’, see Figure 5.4). Both stereotypical thematic role knowledge and scene events provide relevant information about thematic relations. The detective is a stereotypical agent of the spying action identified by the verb (but is not depicted as performing such an action). The wizard, in contrast, is depicted as involved in a spying action but is not a stereotypical agent for a spying action. In this case, there was a strong preference of the comprehension
58
Pia Knoeferle and Matthew W. Crocker
system to rapidly rely on depicted events over stored thematic knowledge for processes of incremental thematic role assignment. Evidence for this came from a higher proportion of anticipatory eye-movements to the depicted agent (the wizard) in comparison with the stereotypical agent (the detective) shortly after people had heard the verb.
THE CASE FOR COORDINATED PROCESSING In sum, the two main findings reported above are first that shortly after a cue in the utterance mediated a relevant depicted event, the eye-gaze pattern provided evidence for rapid structural disambiguation and incremental thematic role assignment. Second, depicted scene events when identified by a verb have a greater relative importance in assigning thematic role relations than stereotypical thematic role knowledge associated with a verb. In what follows, we discuss the implications of these two findings for accounts of online incremental utterance comprehension in situations when both utterance and scene are relevant for comprehension. Recall that while findings by Tanenhaus et al. (1995) reveal a close time-lock between the mention of a word and the time when listeners establish reference to relevant scene objects, they do not allow to clearly determine a closely time-locked reciprocal influence of the perceived scene in determining the structuring of the utterance. Results from Knoeferle and colleagues thus importantly extend the findings by Tanenhaus et al. (1995). The rapid verb-mediated influence of the depicted actions on online sentence comprehension importantly revealed a tight coupling of visual and linguistic processing. This was apparent from the eye-movement records: Unlike in the Tanenhaus et al. (1995) studies, fixation patterns between the SVO and OVS conditions did not differ from the start of the utterance. Rather, fixation patterns in the experiments by Knoeferle et al. (2005) only diverged after people had heard the verb that identified the relevant depicted action. The finding of a close coordination of structural disambiguation with the time when the verb identified the action and its associated thematic relations allows us to exclude a procedural account in which the timing of the influence of scene events is underspecified (Tanenhaus et al., 1995). Let us now consider in more detail the extent to which the findings from experiments by Knoeferle et al., 2005, and Knoeferle and Crocker (2004) are compatible with the other lines of research that we briefly introduced further above (for example, Bergen and Chang, 2005; Jackendoff, 2002; Tanenhaus et al., 2000). The finding that depicted events rapidly influence online structural disambiguation and incremental thematic role assignment seems to be broadly compatible with all of these frameworks. Note, however, that none of them predicts a close synchronization in the integration of visual and linguistic processing. Results from Knoeferle and Crocker (2004) (that is, the preferred reliance on depicted events as opposed to stored thematic role knowledge for incremental thematic
The Coordinated Processing of Scene and Utterance
59
interpretation) do not appear to be straightforwardly accounted for by these frameworks. The relative priority of depicted events during comprehension cannot, for instance, be fully explained by a Jackendovian or Embodied Construction Grammar framework. Neither of these predicts a greater relative importance of depicted events than of verb-based thematic role knowledge during comprehension. Interactionist models can also not account for the greater relative importance of depicted events (Tanenhaus et al., 2000). Recall that this type of model would predict processing difficulty if constraints that are comparable in strength compete. Findings from Knoeferle and Crocker (2004), importantly, showed that the competing informational sources— stored thematic role knowledge and depicted events—were comparable in strength: Each of these two types of information enabled thematic interpretation rapidly when the verb uniquely identified them. The model further would appear to predict that when equally strong constraints compete, they cannot rapidly be applied during processing. This was clearly not the case in the experiment by Knoeferle and Crocker (2004). When the verb identified two informational sources (verb-based thematic role knowledge and depicted actions) that were comparable in strength, comprehension processes rapidly relied on depicted events for incremental thematic interpretation. Since existing frameworks do not predict findings of the relative importance of depicted events, the first step is to identify the factor(s) that caused the preferred reliance of the comprehension system on depicted events in comparison with stored thematic role knowledge. The origin of the preferential reliance on scene events might, for instance, derive from developmental and/or evolutionary comprehension strategies. The rapid impact of explicitly depicted thematic relations between entities identifies the comprehension system to be highly adapted towards acquiring new information from its environment rather than always relying on linguistic and world knowledge. Such a proposal is compatible with theories of language acquisition, such as that of Gleitman (1990). In her account of language acquisition, Gleitman argues that a child can extract event structure from the world around it. When the child perceives an event, the structural information it extracts from it can determine the interpretation of a sentence that describes that event. The interpretation of a sentence can in turn direct the child’s attention within the visual environment. The fact that the child can draw on two informational sources (sentence and scene) enables it to infer information that it has not yet acquired from what it already knows. Assume, for instance, the child perceives a girl throwing a ball, and hears the sentence The girl is throwing the ball. Assume further, the child knows the word girl. Even if it does not know what throwing means, perception of the event can enable it to deduce the meaning of the unknown verb it just heard. It can further identify girl as the thing performing the throwing-action. Such a developmental account offers one explanation for why the comprehension system relies in preference on depicted events over stored thematic role knowledge. Indeed, the observed priority of depicted events may have developed in the course of language acquisition. There is, furthermore, clearly an important interplay between the function
60
Pia Knoeferle and Matthew W. Crocker
of language in guiding attention in the immediate scene during language acquisition, and the influence of scene information on language understanding. The experiments that we have discussed in exploring the interplay of linguistic and visual processing share one fundamental aspect: the utterances are about the immediate visual scene. In language acquisition, when parents talk to their children, language is likely often about the immediate scene which children typically explore. In these situations, it serves a specific function, namely making the immediate scene accessible for the child, and identifying objects that it perceives (see, Roy and Pentland, 2002, for related research in modelling). During adult language comprehension, however, this is not true to the same extent. Language often sub-serves other functions, and is only used to refer to entities in the immediate scene for part of the communication we engage in. Much of our day is spent talking, reading or writing about things not immediately present. Examples that come to mind are the expression of abstract ideas, or the narration of past events. As a result of this observation, we deem it necessary to qualify the findings by Knoeferle and colleagues. We acknowledge that in situations where the utterance does not directly refer to the immediate visual environment, depicted events will almost certainly not have the importance that is suggested by the findings discussed earlier on the influence of depicted events, since the scene is irrelevant. It is the immediate presence and relevance of both utterance and scene during comprehension which enables the rapid interplay between these two informational sources. We do expect, however, that in situations where the utterance is about the immediate environment, the findings of the priority of depicted events over stored thematic role knowledge in thematic interpretation will hold true. We propose that while language is often not about the immediate scene in adult life, we have spent a substantial part of our lives acquiring language. We suggest that this period may indeed have shaped both our cognitive architecture (that is, providing for rapid closely temporally coordinated interaction between cognitive systems such as language and vision), and comprehension mechanisms (for example, we rapidly and in preference avail ourselves of information from the immediate scene when the utterance identifies it). Specifying in more detail the nature of multi-modal comprehension (the temporal coordination and relative importance of distinct cognitive processes) may eventually lead to a more fully specified procedural account within a framework of comprehension, such as Jackendoff (2002) or Bergen and Chang, 2005.
ACKNOWLEDGEMENTS This research was funded by a Ph.D. scholarship to the first author, and by SFB 378 project ‘ALPHA’ to the second author, both awarded by the German Research Foundation (DFG).
The Coordinated Processing of Scene and Utterance
61
REFERENCES Bergen, B. and Chang, N. (2005). Embodied Construction Grammar in Simulation-Based Language Understanding. In J. Ostman and M. Fried (eds), Construction Grammar(s): Cognitive and Cross-Language Dimensions (pp. 147–90). Amsterdam, NL: John Benjamins. Forster, K. 1979. ‘Levels of Processing and the Structure of the Language Processor’. In W.E. Cooper and E.C.T. Waler (eds), Sentence Processing: Psycholinguistic Studies Presented to Merrill Garrett (pp. 27–85). Hillsdale, New Jersey: Lawrence Erlbaum. Fodor, J. 1983. The Modularity of Mind. Cambridge. Massachusetts: MIT Press. Frazier, L. and J.D. Fodor. 1979. ‘The Sausage Machine: A New Two-Stage Parsing Model’. Cognition, 6: 291–325. Gleitman, L. 1990. ‘The Structural Sources of Verb Meanings’. Language Acquisition, 1: 3–55. Jackendoff, R. 1997. The Architecture of the Language Faculty. Cambridge, Massachusetts: MIT Press. ———. 2002. Foundations of Language: Brain, Meaning, Grammar, Evolution. Oxford, UK: Oxford University Press. Kamide, Y., C. Scheepers and G.T.M. Altmann. 2003. ‘Integration of Syntactic and Semantic Information in Predictive Processing: Cross-Linguistic Evidence from German and English’. Journal of Psycholinguistic Research, 32: 37–55. Knoeferle, P. and M.W. Crocker. 2004. ‘Stored Knowledge Versus Depicted Events: What Guides Auditory Sentence Comprehension?’ In K. Forster, D. Gentner and T. Regier (eds), Proceedings of the 26th Annual Conference of the Cognitive Science Society. (pp. 714–19). Mahwah, New Jersey: Lawrence Erlbaum. Knoeferle, P., M.W. Crocker, C. Scheepers and M.J. Pickering. 2003. ‘Actions and Roles: Using Depicted Events for Disambiguation and Reinterpretation in German and English’. In Proceedings of the 25th Annual Conference of the Cognitive Science Society (pp. 681–86). Boston, Massachusetts. ———. 2005. ‘The Influence of the Immediate Visual Context on Incremental Thematic Role-Assignment: Evidence from Eye-Movements in Depicted Events’. Cognition, 95: 95–127. MacDonald, M.C., N.J. Pearlmutter and M.S. Seidenberg. 1994. ‘The Lexical Nature of Syntactic Ambiguity Resolution’. Psychological Review, 101: 676–703. Roy, D. and A. Pentland. 2002. ‘Learning Words from Sights and Sounds: A Computational Model’. Cognitive Science, 26: 113–46. Spivey-Knowlton, M.J. 1994. Quantitative Predictions from a Constraint-Based Theory of Syntactic Ambiguity Resolution. In M. Mozer, J. Elman, D. Smolensky, P. Touretzky, and A. Weigand (eds), The 1993 Connectionist Models Summer School (pp. 130–37). Hillsdale, New Jersey: Erlbaum. Tanenhaus, M.K., M.J. Spivey-Knowlton, K. Eberhard and J.C. Sedivy. 1995. ‘Integration of Visual and Linguistic Information in Spoken Language Comprehension’. Science, 268: 632–34. Tanenhaus, M.K., J.C. Trueswell and J.E. Hanna. 2000. ‘Modeling Thematic and Discourse Context Effects with a Multiple Constraints Approach: Implications for the Architecture of the Language Comprehension System’. In M.W. Crocker, M.J. Pickering and C. Clifto (eds), Architectures and Mechanism for Language Processing (pp. 90–118). Cambridge: Cambridge University Press. Zelinsky, G.J. and G.L. Murphy. 2000. ‘Synchronizing Visual and Language Processing: An Effect of Object Name Length on Eye Movements’. Psychological Science, 11: 125–31.
Chapter 6 Script Indices Richard Sproat and Prakash Padakannaya
INTRODUCTION
W
hen letters are combined into words in English or other languages that use the Latin alphabet, the basic rule for doing this combination is simple: add the next letter to the right of the previous letter. More formally put, in English one (con)catenates letters together in a left-to-right fashion. Other scripts, such as Hebrew, switch the direction to right-to-left, but overall the process is equally simple. This simple method of combination found in Latin-based and other canonical alphabetic scripts is in stark contrast to writing systems such as those of India, where glyphs are combined in a variety of manners. Thus, in the Devanagari script, for example, while consonants are typically arranged left-to-right, post-consonantal vowels may be written after, above, below, or even before the consonant(s) they logically follow.
SCRIPT LAYOUT In previous work (Sproat, 2000, 2003), a model of script layout was developed wherein the various manners of combining glyphs were expressed in terms of directional catenation operators. In particular, five types of catenation were countenanced, as shown in Figure 6.1. The semantics of the different catenation operators is shown schematically in Figure 6.2. In Figure 6.3, we illustrate each of the catenators with examples from Chinese. Sproat (2003) developed the system further for Brahmi-derived scripts of India, and provided formal analysis of the Devanagari, Oriya, Kannada, Malayalam and Tamil scripts. Among the important properties of these scripts, as described in the literature on writing systems (for example, Daniels and Bright, 1996) are the following:
Script Indices FIGURE 6.1 Sproat 2000
Catenation operators, after
FIGURE 6.2 catenators
FIGURE 6.3
Illustration of the layout catenators in Chinese
63
Schematic illustration of script layout
Note: Approximately 95 per cent of all Chinese characters are composed of a so-called semantic radical, shown in light grey in the above, and a so-called phonetic radical, shown in dark grey. The semantic radical often gives a rough semantic sense of the character; the phonetic component gives a hint as to the pronunciation. As shown above, the semantic and phonetic components of characters can be combined using any of the five catenator types defined previously. Usually, the particular catenator used is determined by the semantic radical.
64
Richard Sproat and Prakash Padakannaya
• Glyphs are arranged into orthographic syllables termed aksara. An aksara begins at the left edge of a consonant sequence and continues until the end of the vowel of the syllable, possibly including a postvocalic nasal. Thus a sequence VCCCVCCVC will be divided into V.CCCV.CCV.C, with the first vowel being in an aksara by itself, and subsequent consonant plus vowel combinations forming their own aksara. It is important to note that the orthographic syllables do not in general correspond to phonological syllables. • Within an aksara, consonants are combined using various catenators depending upon the script. Similarly, vowels are written around the consonantal core using various catenators. Figures 6.4 and 6.5 show the layout properties of various glyphs in different scripts. Full vowel forms are used aksara-initially, otherwise reduced (diacritic) forms are used. (In Hindi, full vowel forms can appear at non-initial word position too.) Similarly consonants, when combined into the consonant sequences in aksara, may include various kinds of reductions from their full ‘stand-alone’ variant. Finally, varying degrees of fusion may be seen with different glyph combinations. • In many Brahmi-derived scripts, some vowels are written as complexes of other more basic symbols. Often these more basic symbols are themselves indicators of other vowels. This is illustrated for Kannada in Figure 6.5. Thus ‘o’ in Kannada is FIGURE 6.4 Full and diacritic forms for Devanagari vowels, classified by catenator inherent to the diacritic forms
FIGURE 6.5 Forms for diacritic Kannada vowels, classified by catenator
Note: Complex vowels are at the bottom of the table.
Script Indices
65
written as a combination of the glyphs for ‘e’ and ‘uu’. Each of these combines with the consonants within the aksara using the catenators that the independent symbol would normally have. Thus the ‘e’ portion is written above and ligatured to the consonant sequence, the ‘uu’ portion after and ligatured to the consonant sequence. In the other complex vowel cases listed in Figure 6.5, special symbols (‘LONG’, ‘ai’) are used that are not used except in complex vowels. • Aksaras always have an inherent vowel, usually a schwa-like vowel, which is the vowel sound that is understood if no explicit vowel mark is used. • Tamil differs from other Brahmi-derived scripts in that consonant sequences have become linearized. A consequence of this is that vowel symbols are written around the last consonant in a prevocalic sequence. So, if a vowel mark uses rightwards catenation and thus is written before what it logically follows, in Tamil it will show up before the last consonant of the sequence. This contrasts with the situation in other scripts where such a vowel mark would show up before the first consonant of a sequence. Similarly in Kannada, where consonant sequences tend to be stacked using the downwards catenator, some vowels, such as the ‘ulke’ diacritic for /e/, are ligatured above the first consonant in the sequence; this again is in contrast to the situation in Tamil. In Sproat 2003, it is shown that this difference between Tamil and other scripts follows from the theory of Sproat 2000 as a consequence of Tamil having lost the traditional aksara. This discussion motivates a set of features or questions that can be asked of a phonological element in Indian languages, vis-à-vis its graphic expression. In particular, for a particular phoneme and corresponding glyph (possibly null), one can ask the following questions. Possible answers are given in parentheses: • Is expressed at all? (Yes/No) • What catenator is used to combine with its logical predecessor? (One of four possibilities, since is not used in Indian scripts.) • Is reduced in form relative to its stand-alone variant? (Yes/No) • Are and an adjacent glyph fused? (Yes/No) • Is a complex glyph as with the Kannada vowels from Figure 6.5? (Yes/No) Finally, recognizing that in some Indic writing systems (for example, Tamil) the spelling is less than fully phonemic (that is in Tamil, voiced/voiceless distinctions are not made in the script though they are significant in the language), one can add a feature of transparency: • Is transparently encoded?
66
Richard Sproat and Prakash Padakannaya
SCRIPT INDICES Based on the above features, we propose vector of feature values that we will term a Script Index, which can be applied to any phoneme and its graphical expression (including allographs). The purpose of this index will be to correlate features of the graphical expression of phonemes with readers’ performance on phoneme awareness tasks in psycholinguistic experiments. It has been shown in previous work (Jyotsna and Gupta, 2002; Prakash, 1999, 2000, 2001; Prakash et al., 1993) that issues of layout (catenation direction), and degree of fusion or diacriticity have a measurable effect on readers’ abilities to show awareness of the phoneme corresponding to the glyph in question. A striking example of this is in the differential awareness of postvocalic nasals, written with the anusvara symbol, in Hindi (Devanagari script) versus Kannada (Prakash et al., 1993). Hindi speakers have great difficulty in treating the postvocalic nasal as a separate segment, whereas this is relatively easy for Kannada speakers. This correlates with the fact that in Devanagari the anusvara is written as a small diacritic dot above the righthand edge of the aksara, whereas in Kannada script the anusvara is a large circle written inline with the main consonant of the aksara (Figure 6.6). FIGURE 6.6
Anusvara in Devanagari (left) and Kannada for the word /pensil/
Note: In Devanagari the anusvara is the small dot above the initial element in the word. In Kannada, the anusvara is the large circle in the second position. Note that the actual word ‘pencil’ in Kannada is not written using anusvara; the spelling shown is merely a theoretically possible spelling for illustrative purposes.
We illustrate Script Indices by means of the examples in Figures 6.7–6.9. Each phoneme is associated with a vector, where each of the elements of the vector corresponds to an answer to the questions outlined above. Thus each vector will have the form in Figure 6.10. Consider Figure 6.7, which shows the graphical expression of /peMsil/ (here /M/ represents a postvocalic nasal) in (Hindi) Devanagari. The phonemes /peMsil/ map to a formal expression of the graphical layout represented by the middle line in the figure. In this expression, the aksaras are delimited with square brackets. The spellout of each phoneme is expressed, following Sproat 2000, in the form () where is the phoneme and represents a relation that maps phonemes to their graphical expression. A special version of , namely , represents a reduced form. Within each aksara, the catenators
Script Indices
67
used to combine the elements are explicitly given, with the exception of the logically first element of the aksara, which is always arranged using the script’s macroscopic catenator (Sproat, 2000), which is left catenation in Indic scripts. Not explicitly shown here is whether the elements are fused with other elements, or whether they are phonologically transparent. Figure 6.7 also shows how the formal layout is actually expressed in the form /peMsil/, with arrows showing the correspondence with the formal elements and the glyphs. In Figure 6.11, we list the Script Index vector for each phoneme. Note that /p/ is marked as fused since it is fused with the logically following /e/. Also, in Devanagari, most glyphs are joined together with the headbar: we are not considering the headbar to constitute fusion. FIGURE 6.7 Devanagari
Layout details for /pensil/ in
FIGURE 6.9
Layout details for /lakšmiiša/ in Kannada
FIGURE 6.8
Layout details for /eešvaikya/ in Kannada
Turning now to Figure 6.8, we see a similar diagram for the Kannada word /eešvaikya/. In this case, one of the vowels, /ai/, is expressed as a complex symbol consisting of /e/, and a symbol that is only used with /ai/ (and therefore glossed here as /ai/). In the formal expression of this mapping, this complex symbol is expressed as a set in curly braces, with the individual components being elements of the set. The inherent vowel is not expressed
68
Richard Sproat and Prakash Padakannaya
graphically, and this is represented in the formal representation with the expression (a), and in the graphical representation with the ‘kill’ symbol of a skull-and-crossbones. The Script Index values for each phoneme in this case are as in Figure 6.12. Note that if a phoneme is not expressed, none of the other features are relevant; this is why the values for all features except expr are 0 for /a/. The
component of the expression of /ai/ is written to the right of the expression of /v/ rather than below it as one might expect from its use of the downwards catenator. Sproat 2003 argues for a rule of Kannada linearization whereby all but the first of a sequence of downwards catenators are converted to left catenators. Thus, the first glyph in a sequence of glyphs that use downwards catenation is written below the glyph it logically follows, then all subsequent glyphs in the sequence are written to the right. Finally notice that the features for cat, red and fused are ordered sets when the compl has the value yes. Figure 6.9 depicts the graphic expression of Kannada /lakšmiiša/. The Script Indices for the phonemes in this case are as in Figure 6.13. FIGURE 6.10
Feature vector slots [expressed, catenator, reduced, fused, complex, transparent]
FIGURE 6.11
Feature vector slots for Devanagari /peMsil/ /p/ [expr=yes, cat=left, red=no, fused=yes, compl=no, trans=yes] /e/ [expr=yes, cat=up, red=yes, fused=yes, compl=no, trans=yes] /M/ [expr=yes, cat=up, red=yes, fused=no, compl=no, trans=yes] /s/ [expr=yes, cat=left, red=no, fused=no, compl=no, trans=yes] /i/ [expr=yes, cat=right, red=yes, fused=no, compl=no, trans=yes] /l/ [expr=yes, cat=left, red=no, fused=no, compl=no, trans=yes]
FIGURE 6.12
Feature vector slots for Kannada /eešvaikya/ /ee/ [expr=yes, cat=left, red=no, fused=no, compl=no, trans=yes] /sh/ [expr=yes, cat=left, red=no, fused=yes, compl=no, trans=yes] /v/ [expr=yes, cat=down, red=yes, fused=no, compl=no, trans=yes] /ai/ [expr=yes, cat={up,down}, red={yes,yes}, fused={yes, no}, compl=yes, trans=yes] /k/ [expr=yes, cat=left, red=no, fused=no, compl=no, trans=yes] /y/ [expr=yes, cat=down, red=yes, fused=no, compl=no, trans=yes] /a/ [expr=no, cat=0, red=0, fused=0, compl=0, trans=0]
Script Indices FIGURE 6.13
69
Feature vector slots for Kannada /lakšmiiša/ /l/ [expr=yes, cat=left, red=no, fused=no, compl=no, trans=yes] /a/ [expr=no, cat=0, red=0, fused=0, compl=0, trans=yes] /k/ [expr=yes, cat=left, red=no, fused=yes, compl=no, trans=yes] /s/ [expr=yes, cat=down, red=yes, fused=no, compl=no, trans=yes] /m/ [expr=yes, cat=down, red=yes, fused=no, compl=no, trans=yes] /ii/ [expr=yes, cat={up, left}, red={yes,yes}, fused={yes,no}, compl=yes, trans=yes] /s/ [expr=yes, cat=left, red=no, fused=no, compl=no, trans=yes] /a/ [expr=no, cat=0, red=0, fused=0, compl=0, trans=yes]
SCRIPT INDICES AND PHONOLOGICAL AWARENESS: PRELIMINARY EVIDENCE Previous research mentioned earlier showed that issues of layout (catenation direction), and degree of fusion in Indian alphasyllabaries do have a significant effect on one’s performance on metaphonological tasks. In all these studies, phonemic tests had corresponding syllabic tests controlled for the general task demands. In general, participants performed consistently better on syllabic tests than corresponding phonemic tests. However, qualitative analysis of performance on various phonemic tasks reveals the following. In the phoneme recognition test, performance is better when the target is a consonant part of aksara (orthographic syllable) rather than a vowel. Phoneme oddity test results also demonstrate an advantage of consonant targets over the vowel targets. Children who find phonemic tasks very difficult are almost 100 per cent successful in deleting phonemes (for example, arka /r/, and anusvara ‘nasal postvocalic sounds’) that have independent glyphs written inline with other aksaras. The second consonant in a sequence that has a reduced but not fused glyph also posed no problem in the deletion task. On the other hand, the SR in deleting consonants with full but fused forms is very low. In addition, the non-linearity of some of the features (for example, arka in Kannada) also pose problems in such tasks as aksara deletion. Thus, catenations, their explicitness, independence and linearity are the factors that determine one’s success in metaphonological manipulations of the glyphs in question. The overall performance of individuals in metaphonological tasks should be a function of these factors. Our aim in future work is to arrive at a script index based on the total vector values of individual characters and to compare the performance on metaphonological tests with script indices in the Hindi (Devanagari), Kannada, Malayalam, Oriya and Tamil languages.
70
Richard Sproat and Prakash Padakannaya
REFERENCES Daniels, P. and W. Bright. 1996. The World’s Writing Systems. New York: Oxford University Press. Jyotsna, V. and A. Gupta. 2002. ‘Exploring Word Recognition in a Semialphabetic Script: The Case of Devanagari’. Brain and Language, 81: 679–90. Prakash, P. 1999. ‘Reading Disability and Knowledge of Orthographic Principles’. Psychological Studies, 44: 59–64. ———. 2000. ‘Is Phonemic Awareness an Artefact of Alphabetic Literacy?!’ Presented at ARMADILLO 11, Texas, A&M University, 13–14 October. ———. 2001. Syllabic and Phonemic Awareness in Children Acquiring Literacy in a Semisyllabic Script. Mysore: Department of Psychology, University of Mysore. Prakash, P., D. Rekha, R. Nigam and P. Karanth. 1993. ‘Phonological Awareness, Orthography and Literacy’. In Robert Scholes (ed.), Literacy and Language Analysis (pp. 55–70). Hillsdale, New Jersey: Lawrence Erlbaum Associates. Sproat, R. 2000. A Computational Theory of Writing Systems, (ACL Studies in Natural Language Processing Series). Cambridge. UK: Cambridge University Press. ———. 2003. ‘A Formal Computational Analysis of Indic Scripts’, International Symposium on Indic Scripts: Past and Future, Tokyo, December. Available at: http://compling.ai.uiuc.edu/rws/newindex/indic. pdf.
Chapter 7 How Do We Parse Compound Words? Gary Libben
INTRODUCTION AND OVERVIEW
I
t has been estimated that a typical educated adult who speaks English as a mother tongue will have a vocabulary of about 100,000 words. If this person also speaks other languages, his or her total vocabulary set could be quite easily doubled. These are rather large numbers, particularly if one considers that each item in a vocabulary set must be acquired and organized in the mind so that it can be accessed effortlessly in about a tenth of a second for the purposes of online language production and comprehension. To be sure, the maintenance of an adult vocabulary, its organization in the mind, and its availability in the timeframe required by online lexical processing are all rather amazing. Yet, it must be noted that most words are not entirely unique. Rather, they are typically multimorphemic recombinations of two or more semantic units. This greatly facilitates the understanding of novel words and enables words with similar meanings to be related by formal structure. Thus, to take an English example, the agentive suffix—er creates new predictable meanings when attached to verbs, so that the word teacher means ‘somebody who teaches’, the word farmer means ‘somebody who farms’ and the word blogger must mean ‘somebody who blogs’ (providing that blog is interpreted to be a verb). Although suffixation of the sort just described for English is a powerful means of word formation, it is by no means as powerful as compounding, which constitutes the focus of research in this report. Compounding is the most widespread of all morphological processes found in the world’s languages, and allows an almost unconstrained ability for morphemes to be recombined to create new meaningful structures. So, for example, whereas the suffix—er must occur following a root, the compound element—key—may occur in any position. Moreover, because it will carry with it a lexical meaning rather than a morphological function (of the type which is more associated with affixes), a compound formed from a combination of roots will have a greater range of meaning
72
Gary Libben
relations. The element—key—can thus combine with other roots to form compounds such as keyboard, keyhole, keynote or roomkey. It is important to note, however, that in compound structures such as these, the meaning of the whole word is typically not at all predictable from the meanings of its parts. Thus, although the meaning of keyboard (the computer input device) is certainly related to the meanings of its elements—key—and—board—the meaning of the word is by no means determined by the meanings of those elements. A keyboard could, quite reasonably, refer to a board shaped like a key or perhaps a board mounted on a wall onto which keys are attached. Indeed, if a speaker of English were presented with the compound keyboard, and for some reason were unfamiliar with its conventionalized meaning, he or she would surely construct some meaning, and indeed perhaps the conventionalized one, or one of the others I have just proposed. The considerations above form the backdrop to the research that my colleagues and I have been carrying out on the mechanisms of compound word processing across languages. In this chapter, I focus on two key aspects of this research. These are: 1. The manner in which the constituents of compound words are activated. 2. The manner in which morphemes are arranged within compound structures. Before proceeding to provide an overview of our findings so far, I would like to also present a brief overview of the motivation for this research in the context of the cognitive sciences. Here the issue is plainly: why would the study of compound processing stand to advance our general understanding of how words are processed and of the cognitive structures and activities that enable that processing? In my view, much of the answer to this question does not come from the psycholinguistic research itself, but rather from the characteristics of the object of that research—specifically the place of compound words in human language and their properties as morphological objects. As I have mentioned above, compound words are almost ubiquitous across the world’s languages. As Dressler (2006) notes, there exists a cross-linguistic implicational hierarchy of morphological word formation processes such that if a language has prefixed or suffixed words it will almost certainly have compound words. However, if it has compound words, it may or may not have any other morphological means of creating new words in the language (Chinese could be considered to be an example of such a ‘compounding-only’ language). As has been suggested by Jackendoff (2002), this prevalence of compounding may be historical as well as synchronic. He argues that there could be grounds for considering compounding morphology as a type of protolinguistic fossil, one that survives in languages as a vestige of earlier forms of human communication. In Libben (2006), I argue that whether or not compounding extends back to some protolinguistic period, it seems quite probable that compounding would have been the first human morphological process. Moreover, it is important to note that compound structures are not simply fossils; they are also linguistic survivors with an outstanding track record of success over time, across languages and within languages. This success is related most probably to the
How Do We Parse Compound Words?
73
second property, which makes compounds particularly revealing of principles of lexical processing and cognition. This second property concerns the relatively unconstrained productivity of compounding as a morphological process. Compound words were not only the preferred past means of word formation, they are the preferred present means of word formation. Unlike affixation which is constrained positionally in a lexical string, syntactically in terms of the lexical category (for example, noun, verb, adjective) to which it can attach, and semantically in terms of the kinds of meanings that it can be produced, compounding allows practically anything to be possible. Indeed, in principle, compounding would allow a set of 100 roots to be expanded into a set of 9,900 bimorphemic compounds because every root can occur in both first and second position and can combine with every other element. Moreover, the meanings that would be produced would be quite unconstrained semantically. This can be seen by considering a word such as ball in English. As the morphological head of compounds having the structure __X__ball, one would think that it would retain a relatively constant semantic relation to its modifiers. This, however, is not the case. Consider words such as football, basketball, baseball and broomball. These are all compounds that refer to both a sport and a particular ball used in that sport. Yet even within this highly constrained semantic sphere, the relations between the compound head and its noun modifiers are quite variable, both in terms of the concept to which the noun refers (for example, a body part or an object) and in terms of its semantic function in relation to the head noun (for example, instrument or goal). The overall unconstrained productivity of compounding as a word formation process has substantial consequences for the manner in which compound words are comprehended. The fact that new compound words are coined both easily often suggests that a speaker of English will encounter a novel compound with relative frequency. This brings us to the last characteristic of compounding to be considered before discussing our key research questions. This characteristic derives from the fact that, in compounding, everything can combine with everything. Unlike prefixed or suffixed words in which an open-class root combines with a closed-class affix form, compound words are created through the union of two or more roots, which are typically open-class structures. The consequence of this is that, in compounding, the constituents must be discovered from a potentially limitless set of items rather than identified as members of a limited set of closed-class affixes. This poses a relatively substantial challenge for online morphological parsing. For novel compounds such as trainball, the identification of constituents is the only means through which the word can be interpreted. But if anything can combine with anything, morphological parsing procedures must be powerful enough to find constituents from an open-class set, but also robust enough to recover from spurious word forms within a parse that are not, in fact, morphological constituents (as in *re-ally instead of really or *bar-king instead of barking), and mechanisms that allow the identification of constituents within compound words. These issues are discussed under our first research question below.
74
Gary Libben
THE MANNER IN WHICH THE CONSTITUENTS OF COMPOUND WORDS ARE ACTIVATED Our research in this domain has proceeded under the assumption that the parsing of compound words into their constituents involves a beginning to end scan of the word (for example, Libben et al., 2003; Mandelis and Tharp, 1977; Taft and Forster, 1976). Thus under this view, a novel compound such as trainball will be processed by activation of the constituent train first and then by the activation of the constituent ball. One of the first questions that we posed in our research was: how does this beginningto-end activation take place? It seemed reasonable to suppose that because, as has been stated above, there is no short list of potential compound initial segments that could be looked up. So, the processing system must scan the compound word starting from the beginning and continuing until a morpheme is found. This might seem simple enough for a novel compound such as trainball. Indeed, the parsing of the compound word could be modelled in a sequential and linear manner as represented in Figure 7.1. For the purposes of ease of explication, we can make the simplifying assumption that elements of a compound string are scanned one letter at a time (although we have good reasons to suspect that digraphs such as ‘ai’ and ‘ll’ are treated as single elements, and the steps may not be discrete). The key feature of this model is that scanning continues across a putative compound string and gets re-initialized once a constituent has been activated. The key shortcoming of the model in Figure 7.1, is that human languages are not quite as obliging to compound parsing as one might hope. Compounds can frequently contain false initial constituents. Consider, for example, the novel compound tenderball, both ten and tend are false initial constituents. If compound parsing simply searched for the first real word form in a putative compound string, then a compound such as tenderball would not be appropriately parsed. Yet, simple observation suggests that it is, in fact, correctly parsed with ease. One possible reason for why the false initial constituents do not stop successful parsing is that the strings that follow the false initial elements are not meaningful elements. This would allow ten-derball or tend-erball to be rejected very quickly as candidate parses. Another possibility is that our key assumption that compound parsing searches for the first possible constituent is false. It could, perhaps just as plausibly, search for the last possible initial constituent. In this case, ten and tend would be passed over in favour of tender, which is the point beyond which there is no additional possible initial constituent. In Libben (1994), I constructed a set of novel compound stimuli that could, in principle, allow us to see whether compound parsing preferentially isolates the first possible constituent or the last possible constituent. These ambiguous novel compound stimuli are words such as clamprod, which can be parsed as either clam-prod or clamprod. If morphological parsing preferentially isolates the first possible initial constituent,
How Do We Parse Compound Words?
75
then the word clamprod should be parsed as clam-prod. If, on the other hand, the parsing system preferentially advances as far as it can through a compound word, then clamp-rod should be the preferred parse. The results of this research were surprising. They revealed that, in fact, all possible parses were considered by the system in the initial phases of word recognition. Subsequent research by Libben et al. (1999) supported this conclusion with evidence that an ambiguous compound such as clamprod, facilitated recognition of the semantic associates of all its constituents (for example, sea associated to clam; hold associated to clamp; push associated to prod; stick associated to rod). We reasoned that such activation was likely to be computationally intensive and, indeed, a subsequent lexical decision experiment found that ambiguous novel compounds such as clamprod took longer to reject than their unambiguous reversals (for example, prodclam or rodclamp). These findings formed the background to the view that the morphological processing system does not necessarily find a single ‘correct’ parse of a compound stimulus, but rather, is constrained to find all possible parses of that stimulus. FIGURE 7.1
A simplistic view of compound parsing
Note: The model assumes that parsing occurs from left-to-right one letter at a time. It is also assumed that the activation of an existing root within the string triggers a new start to the parsing mechanism.
In Figure 7.2, a mechanism is presented in which such exhaustive compound parsing can be accomplished. As with the model shown in Figure 7.1, beginning-to-end parsing
76
Gary Libben
is assumed. The difference in Figure 7.2 is that the primary beginning-to-end scan of the putative compound word is exactly the same as the scan that would be employed for any long word of the language. Since clamprod is a novel compound, however, the final step of the primary parse (the one on the far left) does not yield an existing word and is therefore not represented in boldface type. The model provides an account for how all constituent morphemes of an ambiguous string are activated. The beginning-to-end parsing approach also predicts that embedded strings will be activated only insofar as they are consistent with an incremental parsing. Thus, although the model captures the fact that the constituents, clam, clamp, prod and rod are activated; it disallows the activation of the embedded string lamp, because it is not isolated as a potential constituent at any of the parsing steps. In Libben (1994) and Libben et al. (1999) this approach to compound parsing is referred to as the APPLE model, standing for Automatic Progressive Parsing and Lexical Excitation. FIGURE 7.2
A more adequate view of compound parsing (The APPLE II model)
Note: As in Figure 7.1, letter by letter scanning is assumed for the sake of simplicity. In this approach, however, each morphemic activation triggers a new morphological parse that is conducted in parallel with the primary parse.
A consideration of the representation in Figure 7.2 brings us to another noteworthy characteristic of compound processing. If indeed, the parser will isolate all potentially useful constituents of a compound string, there must be a mechanism whereby the system settles into a preferred parse for ambiguous strings and deactivates dispreferred interpretations. As we noted above, ambiguous novel compounds take longer to process
How Do We Parse Compound Words?
77
than ambiguous ones. In our experiments, we also found that participants in the experiments were not aware of the ambiguity of strings such as clamprod. Rather, they produced a single parse of the string. We suspect that the elevated RTs for ambiguous strings does not result from extra time required to activate multiple constituents. As is shown in Figure 7.2, we assume that such activation is carried out in parallel. Rather, it seems likely that the extra processing time is associated with the need to deactivate nonpreferred parses. The consequences of this deactivation are discussed in Libben and de Almeida (2001) for multimorphemic lexical processing and in Libben et al. (2004) with respect to its consequences for impairments of compound processing in aphasia. If indeed, the lexical processing system is organized such that all possible constituents of compounds must be activated and if it is also organized so that dispreferred parses need to be deactivated (at some cost to processing time and resources), then one would imagine that there would be considerable pressure against maintaining such structurally ambiguous compounds in the lexical inventory. To the best of our knowledge ambiguous compounds such as clamprod are very rare across languages.
THE MANNER IN WHICH MORPHEMES ARE ARRANGED WITHIN COMPOUND STRUCTURES Although ambiguous novel compounds of the type we have discussed above are indeed rare in languages, there is another type of compound that is also susceptible to ambiguity and therefore to the creation of alternate parses. This is the class of compounds that contain three, rather than two constituents. In our research, we have investigated these structures, because they possess the minimum complexity to study how morphemes may be arranged within compound words. English, however, is not the ideal language for such an investigation. The primary reason for this is that it contains relatively few triconstituent compounds and, when it does, their constituent structure is marked in the orthography by the presence of a space at the major constituent boundary. Consider, for example, the two compounds that could be constructed from the morphemes teddy, bear and puppet. As is shown in Figure 7.3, the two resulting compounds can be described as having distinct morphological arrangements, one that is left-branching (that is, teddybear puppet) and one that is right-branching (that is, puppet teddybear). In English, in contrast to the languages that we will consider below, this branching structure has already been coded in the orthography. As is shown in Figure 7.3, there is an orthographic space at the major constituent boundary, but no space at the minor constituent boundary. In fact, the pattern for such compounds in English is that (i) at least one space must be present when three morphemes are concatenated, and (ii) one of those spaces must occur at the major constituent boundary.
78 FIGURE 7.3
Gary Libben
Left branching and right branching structures for English triconstituent compounds
Languages such as Dutch and German differ from English in two ways with respect to triconstituent compounding. First, this form of complex compound is much more frequent in Dutch and German and, second, triconstituent compounds are written without orthographic spaces in those languages. Thus, in Dutch or German, the compounds in Figure 7.3 would be written as single words. What follows from this is that, for Dutch and German, the morphemes within the compound are not pre-arranged in print. Rather, the reader of these languages has to first isolate the constituents within the compound and then create (or identify) at least one compound within a compound such that the whole structure can be given a hierarchical structure. It should be noted that the creation of such a hierarchical structure is essential to the comprehension of a triconstituent compound because, to the best of our knowledge, all compounds containing three elements in these languages are only interpretable as being composed of a compound plus a monomorphemic word. That is, the languages do not possess any triple morpheme compounds that have a flat morphological structure (for example, teddy-bear-puppet). Our research questions were thus: which sorts of hierarchical structures are preferred by speakers of these languages and how could such preferences be accounted for psycholinguistically? Krott et al. (2004) report an experimental and lexical statistical analysis of triconstituent compounding in the two languages. It was found that a compound’s hierarchical branching structure is significantly skewed towards the left in both languages (65 per cent of German compounds are left-branching; 64 per cent of Dutch compounds are left-branching. Interestingly, these patterns were mirrored in an off-line nonsense word task in which native speakers of Dutch and German were presented with three-syllable nonsense strings (for example, mapgarwef ) as potential triconstituent compounds and were asked, in separate experiments, to indicate (i) where they would place a hyphen if this triconstituent compound were to occur at the end of a line of text, and (ii) where they would consider the major constituent boundary to be. In both types of experiments, the results showed a significant preference for left-branching structures (66 per cent in German; 61 per cent in Dutch). Now, why would we see such a structural preference? Here we might look to three major possibilities. The first derives from the overall semantics of compounding, in which, for
How Do We Parse Compound Words?
79
languages such as English, Dutch and German, the constituent on the left modifies the constituent on the right. Thus, for right-branching triconstituent compounds, the head (for example, bear in puppet teddybear) would be receiving double modification. Moreover, it would be embedded in a sub-compound and, as such, might be less available semantically. It should be noted, of course, that such constraints could not apply to a nonsense word paradigm directly. Rather, they would have to be interpreted as semantic pressures that form the character of a trimorphemic morphological template that, in turn, influences nonsense word parsing. A second possibility is somewhat related to the first. It may be the case that languages in general prefer the morphological heads of compound words to the simple rather than complex. In English, Dutch and German, compounds can be described linguistically as right-headed. That is, the last constituent of the compound word functions as its morphological head. So, for example, if the compound is composed of an adjective and a noun (for example, blackboard) the whole word is a noun, taking its grammatical properties from the last (rightmost) element. This headedness also has semantic properties. Even if a speaker of English does not know the meanings of the compounds boathouse and houseboat, that speaker will know, simply from headedness patterns in the language, that the first one must be a type of boat, whereas the second one must be a type of house. It may be the case that the mobilization of this knowledge is easier if the morphological head is a morphologically simple element. A third reason for left-branching preference derives from the notion of parsing directly. If we assume, as has been claimed above, that beginning-to-end parsing of the type shown in the APPLE II model is a fundamental characteristic of the lexical processing system, then it might be the case that this parsing advantages left-branching interpretation in some way. As is shown in Figures 7.4 and 7.5, there are reasons to believe that this might be the case. Let us take as examples, two German triconstituent compounds. The compound Handschuhleder is composed of the morphemes hand (meaning ‘hand’), schuh (meaning ‘shoe’), and leder (meaning ‘leather’). Together, hand+schuh (literally ‘hand-shoe’) means ‘glove’. So, the entire compound is a left-branching structure meaning ‘glove leather’. In Figure 7.5, the same morphemes are re-arranged to form another existing German compound Lederhandschuh. In this case, the compound is composed of the right branching structure leder+handshuh which means ‘leather glove’. As can be seen by contrasting Figures 7.4 and 7.5, the left branching structure in Figures 7.4 results in a parse that produces an early activation of the initial compound constituent handschuh at Step 9 of the parse. At this stage, both the constituents of handschuh are also activated. In the final step of parsing, the simple head leder receives double activation from the two sub-parses that are conducted in parallel. This general tendency for early bimorphemic activation and double head activation would apply to the parsing of any left-branching triconstituent word. Similarly, any right-branching
80
Gary Libben
triconstituent word would show the parsing pattern represented in Figure 7.5. In this case, the initial monomorphemic element is activated early, but the compound substring is not activated until the last stage of parsing. We conclude from this that the functional architecture of morphological parsing plays a role in advantaging left-branching interpretations of compound words. We must be cautious, however, in carrying this view too far because the reality is that native speakers of German derived a uniquely left-branching interpretation for handschuhleder and a uniquely right-branching interpretation for lederhandschuh. Thus, although morphological parsing might advantage left-branching analyses (as found by Krott et al., 2004), it does not force a left-branching analysis. FIGURE 7.4
The parsing of a left-branching triconstituent word in German
Note: The initial compound substring as well as its constituents receive early activation.
How Do We Parse Compound Words? FIGURE 7.5
81
The parsing of a right-branching triconstituent word in German
Note: Here the embedded compound substring is not activated until the final parsing step.
It seemed to us, then, that the best way to explore the extent to which morphological parsing influences the manner in which morphemes are arranged within words would be to investigate ambiguous structures—those in which both a left-branching and rightbranching interpretations are possible for the same linear sequence of morphemes.
82
Gary Libben
To do this, we turned to Chinese. This is a language that has two key properties that promised to be very useful in addressing the question of the role of morphological parsing on morpheme arrangement within compounds. The first is that compounding is the central, if not the only, morphological process in Chinese (Myers, 2006). The second is that, unlike German and Dutch, Chinese represents triconstituent characters as a linear sequence of three characters with equal spacing among them. This allowed us to check whether the patterns of left-branching preference that were observed for Dutch and German might be a special adaptation to the fact that, in both those languages, triconstituent compounds are represented orthographically as single words. This presents a substantial ‘find the morpheme’ problem to the reader. In Chinese, the morphemes are provided as discrete units. Thus, the investigation of compounding in Chinese allowed us to examine whether the left-branching preference that we observed might disappear when such ‘find the morpheme’ challenges are removed. In Yin et al. (2004), we used a variety of off line and online tasks, including a segmentation task which was methodologically similar to that reported in Krott et al. (2004). For Chinese, however, real triconstituent compounds were used, rather than strings of nonsense words. These included: (a) Structurally ambiguous triconstituent compounds such as ‘big-meeting-hall’, which could either be a ‘hall for a big meeting’ or a ‘meeting hall that is big’. (b) Compounds that have a flat semantic structure, such as ‘real-kind-beautiful’ (zhenshan-mei), meaning ‘virtue’. (c) Trisyllabic loan words such mai-ke-feng, which are made up of the three characters ‘wheat-gram-wind’, chosen for their combined sound similarity to the English word microphone, which is the meaning of the entire string. The segmentation results for categories (a) and (b) above yielded patterns (61 and 65 per cent, respectively) that were almost identical to both the experimental data and lexical statistical data for German and Dutch. Category (c) showed a significantly higher proportion of left-branching segmentation (77 per cent). Taken together with the findings from Dutch and German, this latter finding points to the conclusion that the left-branching preference is not dependent on the semantics of triconstituent compounding, but rather is a structural preference. As the string of Chinese characters ‘wheat-gram-wind’ has nothing to do with the overall meaning of the compound, there is no arrangement of the constituent morphemes that would make more sense than any other. On the contrary, we may conclude that, for Chinese, the leftbranching preference is enhanced when segmentation is freed from semantic constraints. Thus, the general left-branching preference in Chinese offers further support for the view that multimorphemic structures such as these are processed in a beginning-to-end manner and that branching preference is related to the manner in which compounds are parsed.
How Do We Parse Compound Words?
83
FIGURE 7.6 A schematic representation for the parsing of ambiguous triconstituent compounds containing the morphemes 1,2, and 3 Isolated substring 1
1 1
Parsing step
3
2
2
12
12
2
12 3
3
23
3
(a)
(b)
(c)
(d)
Note: Shaded boxes represent lexical activation.
These findings allow us perhaps to consider the relation between the question of how morphemes are isolated within compounds and how they are arranged in a more general way. In Figure 7.6, I provide a schematic representation of the principles of automatic progressive parsing and lexical excitation (APPLE) that would apply to any ambiguous trimorphemic string in any language. The use of boxes rather than strings underlines the claim that it does not make a difference whether compounds are written orthographically (as in Dutch and German) or in morphemic characters (as in Chinese), or whether they are parsed auditorily (in which case, the differences in orthographic representation would be irrelevant). The representation in Figure 7.6 suggests that in all these cases, left-branching possibilities will be preferred over right-branching possibilities. In this figure, the three morphemes of a trimorphemic word are represented as the numbers 1, 2 and 3. Lexical activation is indicated by the shading of a box. As can be seen in this figure, initial substring compounds are activated before final ones (yielding the left-branching preference). I propose that the representation in Figure 7.6 might represent a universal summary of the course of events in the parsing of ambiguous triconstituent words, and therefore also words with less complexity. So, simple monomorphemic words would be represented as having only the sequence shown in (a) on the left of Figure 7.6. A non-ambiguous leftbranching compound would have all sequences (a), (b), (c) and (d), with the last block of (c) unshaded. A non-ambiguous right-branching compound would have the sequences (a), (c) and (d). The model suggests that a universal feature of compound processing is that all single constituents and multimorphemic strings that become activated in a progressive scan of an input string initiate new parallel parses. For right-headed languages, this means that left-branching structures will result in early activation of embedded compounds and will also result in potentially multiple activation of a simple morphological head. For languages with left-headed compounds (for example, Italian, Hebrew), we predict that there will be less preference for left-branching compounds because in that case the advantages of early activation of the embedded compound might be offset by the disadvantages of having a morphologically complex head.
84
Gary Libben
Finally, I would like to move from the more technical to the more general implications of the representation in Figure 7.6: specifically, I wish to underline my view that the preference for left-branching morphological structure shown in Figure 7.6 is simply something that falls out from beginning-to-end parsing. It is not a design feature of the system. In fact, even the beginning-to-end nature of morphological parsing is not a design feature. This, most probably, is a simple consequence of the fact that, especially in auditory processing, the beginnings of compound words are available before the ends. Therefore, the lexical processing system works first on what it first has available to it. In my view, if there is a design feature at all present in morphological parsing it is that the parsing system is constrained to not rule out any parses that could potentially be useful. This feature and its implications for the overall functional architecture of language processing, are discussed below.
DISCUSSION This report began with the observation that compound words are the most common morphological structures in the world’s languages and perhaps the oldest morphological structures that humans have used to create new meanings from existing ones. It was noted at the outset that perhaps one of the reasons for this is that compounding is almost completely unconstrained in terms of which linguistic elements can occur in a string and where those elements can occur. Thus, in compounding, almost anything can combine with anything. It was also noted that compounding is characterized by great productivity, such that speakers of a language construct new compounds freely and that the new compounds are interpreted by hearers in terms of their morphological constituents. In the sections above, I have discussed two of the foundations for this interpretative ability: constituents within compounds must be isolated and they must be arranged in a manner that will allow interpretation. The picture that emerges from this research might appear paradoxical. The morphological parsing system has as its goal the development of a single interpretation for a word, yet it yields multiple, and often contradictory parses. One might wonder, then, why it has not developed to become more effective at isolating the right answer, rather than all possible answers to the question ‘What is this long word made of?’. As I have claimed elsewhere (for example, Libben, 2006) it seems most likely to me that the lexical processing system does not have as its goal the isolation of any correct answer. Indeed, considering that the ‘correct answer’ for any language comprehension event is figuring out what the speaker or writer had in mind, it is difficult to imagine how a computationally encapsulated lexical parsing system could ever get the right answer. So, it does not even try to do so. Rather, its goal is the ‘maximization of opportunity’ for comprehension. This involves creating a candidate set for as many interpretations as
How Do We Parse Compound Words?
85
possible, which, in the case of compounds, means as many potential morphemes and their combinations as possible. This possibility emerged in fact from our early investigations of ambiguous novel compounds such as clamprod, which are summarized above. As I have noted, although these words are structurally ambiguous, participants in our experiments did not perceive them to be ambiguous. Rather, they arrived as single interpretations. In separate experiments, it was discovered that the compound parsing preferences corresponded most closely to which version of the ambiguous string made most sense (for example, clamp-rod or clamprod). We reasoned that this relative sensibility could only play a role in the determination of the final interpretation if both possible interpretations were initially considered. In my view, this overall functional architecture, one in which lower language procedures make all possible interpretations available to higher-level interpretive procedures might be an overarching characteristic of human language processing. As such, compound words might be best considered not as having meaning, but as providing springboards to meaning.
ACKNOWLEDGEMENTS This research was supported by a Major Collaborative Research Initiative Grant from the Social Sciences and Humanities Research Council of Canada to Gary Libben, Gonia Jarema, Eva Kehayia, Lori Buchanan, Bruce Derwing and Roberto de Almeida.
REFERENCES Dressler, W.U. 2006. ‘Compound Types’. In G. Libben, and G. Jarema (eds), The Representation and Processing of Compound Words (pp. 23–44). Oxford, UK: Oxford University Press. Jackendoff, R. 2002. Foundations of Knowledge. Oxford, UK: Oxford University Press. Krott, A., G. Libben, G. Jarema, W. Dressler, R. Schreuder and H. Baayen. 2004. ‘Probability in the Grammar of German and Dutch: Interfixation in Triconstituent Compounds’. Language and Speech, 47: 83–106. Libben, G. 1994. ‘How is Morphological Decomposition Achieved?’ Language and Cognitive Processes, 9: 369–91. ———. 2006. ‘Why Study Compounds? An Overview of the Issues’. In G. Libben and G. Jarema (eds), The Representation and Processing of Compound Words (pp. 1–21). Oxford, UK: Oxford University Press. Libben, G. and R.G. de Almeida. 2001. ‘Is there a Morphological Parser?’ In W.U. Dressler and S. Bendjaballah (eds), Morphology 2000. Amsterdam, NL: John Benjamins. Libben, G., B. Derwing and R. de Almeida. 1999. ‘Ambiguous Novel Compounds and Models of Morphological Parsing’. Brain and Language, 68: 378–86. Libben, G., M. Gibson, Y.B. Yoon and D. Sandra. 2003. ‘Compound Fracture: The Role of Semantic Transparency and Morphological Headedness’. Brain and Language, 84: 26–43.
86
Gary Libben
Libben, G., L. Buchanan and A. Colangelo. 2004. ‘Morphology, Semantics, and the Mental Lexicon: The Failure of Inhibition Hypothesis’. Logos and Language, 4: 45–53. Mandelis, L. and D.A. Tharp. 1977. ‘The Processing of Affixed Words’. Memory and Cognition, 5: 690–95. Myers, J. 2006. ‘Compounding in Chinese’. In G. Libben and G. Jarema (eds), The Representation and Processing of Compound Words (pp. 169–96). Oxford, UK: Oxford University Press. Taft, M. and K.I. Forster. 1976. ‘Lexical Storage and Retrieval of Polymorphemic and Polysyllabic Words’. Journal of Verbal Learning and Verbal Behavior, 15: 607–20. Yin, H., B.L. Derwing and G. Libben. 2004. Branching Preferences for Large Lexical Structures in Chinese. Paper Presented at the Fourth International Conference on the Mental Lexicon, Windsor, Ontario, Canada.
Chapter 8 Other Minds: Social Cognition in Wild Bonnet Macaques Anindya Sinha
INTRODUCTION Empirical studies on the cognitive abilities of non-human primates and their underlying mechanisms developed primarily because we assume that their intelligence and, if one may use the term, minds are most like our own. Through our understanding of them, we would possibly one day understand what it is like to be essentially human. However, this view that they are most like us also coexists in our minds with the equally pervasive idea that non-human primates differ fundamentally from us because they lack sophisticated language, and may, thus, also lack some of the capacities necessary for reasoning and abstract thought. Given the recent developments in our understanding of the cognitive abilities of many primates, including the possible existence of rudimentary semantic communication in some species (Cheney and Seyfarth, 1990; Tomasello and Call, 1997), it is likely that comparative studies on primate taxa may yet throw light on the nature and evolution of different human cognitive abilities, including that holy grail of current cognitive research—consciousness (Griffin, 2001). This chapter first briefly reviews some theoretical approaches that utilize observations of behaviour to examine the phenomenon of the animal mind. Two specific examples of social behaviour, knowledge-based decision-making and tactical deception, displayed by bonnet macaques, are then examined in terms of the possible underlying cognitive processes in an effort to obtain some glimpses into the non-human primate mind.
THE PRIMATE MIND A feature that commonly characterizes most primates, including the great apes and humans, is the presence of a complex society in which individuals spend most of their lives.
88
Anindya Sinha
Extensive social interactions between individuals of different ages, sexes, dominance ranks and kinship are typical of many of these societies (Smuts et al., 1987). The development and maintenance of such complex social relationships—each different in its own way—are believed to have placed unusual demands and selected for enhanced cognitive abilities in individuals living in such societies (Chance and Mead, 1953; Humphrey, 1976; Jolly, 1966). If this is true and if indeed there has been a general increase in social complexity—in at least some of its dimensions—during the course of primate evolution, does this provide at least indirect evidence that there has been a progressive evolution of the primate mind, culminating in the human mind, as well? Although there is now increasing belief that primate minds can be rather complex, the question of whether non-human primates can be considered truly conscious continues to be a controversial one. Related to this problem is perhaps one of the most difficult aspects of studying consciousness—that of providing an objective scientific definition of the phenomenon. This definition obviously has to be functional in order for it to be dissected analytically. And it becomes an even greater problem when studying non-human primates—because consciousness then has to manifest itself in behaviour that can be unambiguously ascribed to being an effect of being conscious. Of the numerous definitions of consciousness that exist in literature, perhaps the most functional that have been proposed are perceptual consciousness, the ability to possess certain mental states including emotions, thoughts, beliefs, desires or memories, and reflective consciousness, the recognition by the thinking subject of its own perceptions and mental states (Griffin, 2001). In other words, if an animal were perceptually conscious, it would be able to possess certain mental states—it might, for example, be able to believe, think or remember. If, in addition, it were reflectively conscious, it would be aware of its own mental states—whether they are beliefs, thoughts or memories. Current thinking holds that some of the higher primates may indeed be perceptually conscious, but are extremely unlikely to be reflectively so (Cheney and Seyfarth, 1990). The principal reason for this bias against the belief that primates can reflect on their thoughts and actions may, however, largely be methodological: people can tell us what they are aware of, monkeys cannot.
INTENTIONALITY AND ATTRIBUTION Functionally, an elegant theoretical framework to investigate higher cognitive processes in non-human primates in terms of mentalistic notions is that of Dennett’s intentional stance (Dennett, 1987, 1988). If one assumes that animals are intentional systems capable of mental states like beliefs, desires and emotions, it is possible to consider them beings with different levels of intentionality. Note that a particular individual of any species can be in different intentional states depending on the cognitive basis of the particular behaviour performed. Under this framework, however, each species has an unique position with regard to the highest order of intentionality that it can ever achieve, although lower-order intentional behaviours can always be exhibited.
Other Minds: Social Cognition in Wild Bonnet Macaques
89
To give an example (adapted from Cheney and Seyfarth, 1990), bonnet macaques typically give alarm calls to potential predators such as leopards or wild dogs. On hearing an alarm call given by a particular individual, the other troop members immediately run up trees and then scan for the predator. Theoretically, this behaviour could be considered under different orders of intentionality, as explained next, although studies in cognitive psychology will aim to determine exactly which order it belongs to.
Zero-order An individual has no beliefs or desires at all. All behavioural actions in this category are thus instinctive, invariably evoked in response to specific stimuli. If the bonnet macaque alarm call truly belongs to this category, it must be hypothesized that bonnet macaques give alarm calls as a mere response to a stimulus—the sight of a predator—and no actual desires or beliefs are involved in this reaction.
First-order An individual has beliefs or desires, but no beliefs about beliefs. Behavioural acts can thus be generated intentionally by the actor who, however, need not have any conception of the audience’s mental states. In this case, therefore, bonnet macaque alarm calls are given because the caller believes that there is a predator nearby, although it may have no comprehension of the belief system of its troop members.
Second-order Some conception exists about both one’s own and other individuals’ states of mind. An individual may thus behave in a particular way because it wants others to believe in something. A bonnet macaque may thus give alarm calls because it wants its troop members to believe that there is a predator lurking nearby.
Third-order At this level, an individual may want others to believe that it itself has a particular belief or is in a specific emotional state, or that it wants others to believe that it wants them to respond in a particular manner.
90
Anindya Sinha
If bonnet macaques are truly third-order intentional systems, an alarm call may be given because the caller wants the other individuals to believe that it wants them to rush up the trees. Human beings are typically third-order intentional systems exhibiting a wide variety of behaviours that can be classified under different orders of intentionality. When a human subject removes a finger from a pinprick or a flame involuntarily, for example, it is a zero-order intentional system, since there are no desires or beliefs associated with this behavioural act. Human linguistic communication, on the other hand, is a notable example of a system where the actor (or speaker) makes its own mental states apparent to the audience; this clearly and essentially requires third-order intentionality. Higher-order intentionality (including second- and third-order levels) is interesting because it requires the ability to represent simultaneously two different states of mind— that of the actor and of the audience. To do this, an individual must recognize, for example, that it has knowledge, others have knowledge, and that there may be a discrepancy between them—or, for that matter, between any of the intentional states held by these two minds. Unfortunately, very few studies—either in the wild or in captivity—have so far extensively tested for these alternative capacities of intentionality in primates. A very important functional manifestation of higher-order intentionality, and also of perceptual consciousness, is attribution, whereby an individual is capable of attributing thoughts, emotions and desires to another individual (reviewed in Cheney and Seyfarth, 1990). It is evident that primates are knowledgeable about each other’s behaviour, to the extent that they can often predict and act upon this knowledge even before a behavioural interaction has occurred (Sinha, 1998). But do primates know as much about each other’s beliefs, emotions and intentions? To attribute beliefs, knowledge and emotions to both oneself and to others is to have a theory of mind, first outlined by Premack and Woodruff (1978). And if indeed primates are able to attribute a mind—or more functionally, mental states—to each other, are they capable of recognizing the similarity and differences between their own and others’ states of mind as well? The principal advantage that an animal enjoys if it is able to recognize that other individuals have beliefs which might be different from its own, is that it becomes capable of immensely more flexible and adaptive behaviour. It might then be able to manipulate another individual’s actions and beliefs in a great variety of social situations. Furthermore, if it can recognize ignorance in others, it can selectively reveal and withhold information from them. Again, novel information can be transmitted across individuals by active teaching rather than by the relatively slow process of observational learning. However, there has almost been no such systematic studies of attribution of mental states in social animals, including non-human primates.
Other Minds: Social Cognition in Wild Bonnet Macaques
91
PREDICTING BEHAVIOUR OR PREDICTING MENTAL STATES? Perhaps the most difficult problem in understanding cognitive processes in non-verbal subjects—be they pre-verbal human infants or truly non-verbal primates—is the question of whether an individual is discriminating between others’ states of mind or simply reacting to differences in their behaviour patterns. While it is evident that true mind-reading can only be achieved through some form of behavioural analysis and can, therefore, perhaps be considered a sub-category of behaviour-reading (Whiten, 1993, 1996), it becomes important in certain situations, as, for example, in the analysis of deceptive interactions, to differentiate between actual behaviour analysis and the more cognitively sophisticated (as well as evolutionarily advanced?) mentalism. Although many philosophers of mind have argued that these two processes represent mutually exclusive phenomena (for example, Fodor, 1968), it can also be argued and examples provided from human cognitive processes to demonstrate that they represent two positions on a possible continuum (Whiten, 1994). A theoretical concept of how mental states could be considered intervening variables facilitating the recognition of a number of otherwise complex stimulus–response links (each of which could independently form the basis of a behaviour-reading process) has been elegantly proposed by Whiten (1993, 1996). Drawing from an earlier concept in psychology, it suggests that any number of observable conditions could lead an individual to recognize a certain specific mental state in another individual, and once this state has been attributed, to predict a definable number of behavioural outcomes depending on the ambient situation. A crucial advantage of this model is that the coding of each of these intervening variables (or the so-called mental states) would be more ‘economic of representational resources’ (Whiten, 1996) than would be the multitude of the stimulus– response links that each now represents. This is particularly true for mental states which are achievable by very many different conditions and which can, in turn, affect a number of different outcomes. Mind-reading or the recognition of mental states, by such a definition, could thus constitute a more neurologically economic strategy than would a collection of independent stimulus–response pathways that represent behaviour-reading. Note also that, according to this concept, mentalism arises gradually from behaviour analysis—if the intervening variable mediates the recognition of and response to only a single stimulus–response link, it is virtually indistinguishable from behaviour-reading. Yet another line of evidence that can potentially argue for mentalism as an underlying cognitive process rather than simple behaviour analysis, at least in certain situations, is that of projection of experience (Povinelli et al., 1992b, succinctly reviewed in Whiten, 1996). This has stemmed from studies on role reversal in cooperative tasks in which an individual primate was first trained to perform a definite task to aid another individual in reaching a desired goal, following which it was asked to take on the role of the other
92
Anindya Sinha
individual. These experiments have suggested that the great apes, notably chimpanzees, are able to master new roles with ease and perform novel tasks; and perhaps because they can attribute beliefs and desires to one another; their performance cannot be explained by simple learning of the behaviour of their partner before role reversal. Rhesus macaques, on the other hand, appear to lack empathy and, in order to successfully complete such cooperative tasks, have to learn their new roles afresh (Povinelli et al., 1992a). A philosophical assumption that seems to be implicit in all discussions over whether individual primates are able to recognize mental states or simply perform behaviour analysis is that principles of parsimony are violated when mind-reading is invoked in nonverbal, non-human primates. Such an assumption perhaps owes its origin to the subtle influences that Biblical tradition and Cartesian philosophy seem to have had on western scientific ideology, which has, often implicitly, valued the inherent superiority of man over all other forms of life. Although outside the scope of this chapter, it is important to stress here that it is perhaps now time to re-evaluate such an assumption, and concepts such as that of mental states as economical intervening variables, discussed above, are important steps in this direction (Bennett, 1991). In the remaining section of the chapter, the possible cognitive mechanisms involved in two complex social processes displayed by wild bonnet macaques—social knowledge-based decision-making and tactical deception—will be analysed. Particular attempts will also be made to explore the conceptual contribution that attribution of mental states, as well as orders of intentionality, could offer towards an understanding of these mechanisms.
BONNET MACAQUES—THE SPECIES AND THE TROOPS The bonnet macaque (Macaca radiata), a cercopithecine primate found only in peninsular India, usually lives in large troops of eight to 60 individuals; such multimale troops typically contain several adult males and females, as well as juveniles and infants of both sexes (reviewed in Sinha, 2001). Female bonnet macaques, like many other cercopithecines, usually remain in their natal group throughout their lives, and during adulthood, form strong, linear dominance hierarchies with daughters occupying dominance ranks just below those of their mothers. Adult females develop strong social bonds and display extensive allogrooming and other affiliative behaviour towards one another. Juvenile and adult males, on the other hand, usually emigrate from their natal troops; but bonnet macaque males appear to be unique in being rather unpredictable in this regard, with some individuals even staying back to become the most dominant males in their respective natal troops. Adult males form unstable dominance hierarchies through direct aggression and coalitions, and, unusually for most cercopithecines, exhibit extensive affiliative interactions with one another. Our insights into the social knowledge underlying decision-making processes and tactical deception in this species come from a long-term study involving behavioural
Other Minds: Social Cognition in Wild Bonnet Macaques
93
observations on individually identified animals in three troops (Troops G I, G II and B I) of bonnet macaques inhabiting dry deciduous scrubland and mixed forests around Bangalore city (described in Sinha, 1998, 2003).
SOCIAL KNOWLEDGE A very important component of social cognition is the social knowledge that individual primates might possess with regard to certain attributes of other individuals that they regularly interact with within their social group. In addition to the obvious recognition of each animal as a distinct individual, the possible attributes that such knowledge might encompass could include their dominance ranks and affiliative relationships—factors that seem to influence much of the social behaviour observed in primate societies. Our study of Troop G I led to the documentation of a frequent interaction between adult females—allogrooming supplants—in which a dominant female displaces one member of a pair of grooming females, both subordinate to her (Sinha, 1998). In a majority of these observed supplants (~80 per cent), the most subordinate of the three individuals left her grooming partner as soon as she noticed the dominant female approaching them—such females were thus clearly aware of their own subordinate status relative to the other two individuals. On about 20 per cent of these occasions, however, it was the other female (the more dominant of the two allogrooming individuals) that left—and the factor that most significantly appeared to influence this decision was the social attractiveness of her grooming partner, defined in terms of the amount and consistency of allogrooming that this individual received from all the other adult females in the troop. These dominant females thus retreated when their grooming partners, though subordinate in rank, were more socially attractive, receiving relatively higher levels of allogrooming more consistently from all the other adult females in the troop. Bonnet macaque females, therefore, are clearly aware of the social attractiveness of their grooming companions, and thus seem to be knowledgeable of the social relationships maintained by the other females in the troop. That individual females might also know the relative dominance ranks of their troop members was revealed by the typical patterns of aggressive behaviour and allogrooming choices that occurred during other, similar triadic interactions. If neither of the two allogrooming subordinate females retreated when the third dominant female approached them, for example, the latter usually displayed aggression towards the more subordinate of the two females (Sinha, 1998). Occasionally, however, she did not display any agonistic behaviour, but proceeded to directly allogroom one of the two individuals—and, in the majority of these cases, she groomed the more dominant female. Approaching females thus seemed to be aware of the relative dominance ranks of the two other females, both subordinate to her.
94
Anindya Sinha
Logistic regression analysis of the decisions made by the females indicated that three factors were taken into consideration when they decided to either remain behind or retreat during allogrooming supplants: knowledge of the subject’s own dominance rank, her rank difference with the approaching dominant female, and rank difference with her grooming companion (Sinha, 1998). Individuals are thus clearly aware not only of their own positions in the rank hierarchy, but also of that of the other females in the troop. A model which incorporated the absolute dominance ranks of the latter, however, failed to explain the observed behavioural patterns. Knowledge of another individual’s dominance rank is, therefore, egotistical in that it seems to be acquired only relative to one’s own; a female knows of her rank difference with another female, but does not appear to be aware of the absolute position of her adversary in the rank hierarchy. The observation that rank difference with the approaching dominant female and that with the grooming companion both influenced the decision-making process indicates that a bonnet macaque’s knowledge system is integrative in nature—females are able to simultaneously process information about all their interacting companions and use this knowledge effectively during social interactions. The decisions made in this particular situation were, in reality, even more complex: the intermediate female in a grooming supplant chose to retreat as the approaching individual became relatively more dominant to her while her grooming companion became comparatively more subordinate (as also more socially attractive).
MENTAL REPRESENTATION OF INDIVIDUAL ATTRIBUTES A noteworthy observation in this study was that individual macaques seem to be knowledgeable about the general social attractiveness of particular females in terms of the allogrooming that they receive from other individuals, rather than remember specific pair-wise affiliative relationships (Sinha, 1998). Since, as mentioned earlier, they also know the relative dominance rank of each adult female in the troop, this seems to constitute a clear example of recognition of individuality and individual attributes by these animals. Furthermore, the decision to retreat or remain behind during allogrooming supplants also depended on the absolute position of the actor in the dominance hierarchy—the more subordinate an individual, the more likely she was to retreat. Clearly then, each bonnet macaque female has knowledge of some of her own individual attributes as well. Although all of these abilities must obviously call for some form of fairly sophisticated mental representation of particular individuals, including themselves, associated with their specific properties, what remains unclear is how exactly such information is categorized and coded in the non-verbal cognitive architecture of the macaque mind. It is also important to note that, during triadic interactions, the integrative property of the bonnet macaque’s knowledge system allows her to respond appropriately to the relative dominance ranks of the other interacting individuals. It is striking, therefore, that whatever be the
Other Minds: Social Cognition in Wild Bonnet Macaques
95
stored imagery of the individual attributes of the two females she is interacting with, it is possible for her to access both these sources and integrate them when finally making a socially complex decision.
ATTRIBUTION OF MOTIVES AND FORMATION OF A BELIEF SYSTEM Since during allogrooming supplants, the dominant member of the grooming dyad is more likely to retreat if her grooming partner is very socially attractive, these females behave as if they ‘believe’ that the approaching individual is targeting their subordinate, but usually more socially attractive, companion. Bonnet macaques thus seem to be capable of attributing motives to other individuals within their social matrix, suggesting that they may be able to develop beliefs about such motives. It would appear that this decision to retreat was taken on the basis of a belief that a highly socially attractive individual is more likely, in general, to be the preferred target for affiliative interactions, even if she holds a relatively low position in the dominance hierarchy. That such a belief might indeed be valid is supported by our earlier observation that there was a very strong positive correlation between the number of approaches that the subordinate female of the allogrooming dyad received from other females and her social attractiveness (Sinha, 1998). The nature of this belief and the attribution of a corresponding motive to the approaching individual also seem to be rather pragmatic, since bonnet macaque females evaluate social attractiveness of an individual on the basis of the levels of allogrooming received, and the consistency with which such grooming is received from other females in the troop.
PROJECTION OF EXPERIENCE? An interesting insight into the nature of this particular belief system comes from the actual choices that the troop females made in their display of aggression and grooming preferences during triadic interactions. Thus, when a grooming supplant did not actually occur, as described above, the approaching individuals were most likely to display aggression and chase away the more subordinate of the two individuals, while on other occasions, if they did not demonstrate any aggression, they almost invariably preferred to groom the more dominant member of the dyad. Why then occasionally did the more dominant member of the allogrooming dyad retreat? One possible answer to this question is that bonnet macaque females may form a general belief system regarding the social attractiveness of the other females in the troop and motives may be attributed to the approaching females during triadic interactions in accordance with such a system. This belief system may, however, be an erroneous one—and this is revealed by the fact that although the approaching females usually chose the dominant member of the
96
Anindya Sinha
allogrooming dyad as a grooming companion during allogrooming supplants, these same females retreated when they were, in turn, approached while grooming a socially attractive subordinate. In other words, individual females exhibited this erroneous behaviour even though, on several occasions, they themselves had preferentially allogroomed the dominant member of a grooming pair after approaching such dyads. Could this be considered a failure, in some sense, of macaque females to project their own past experiences and thus to adopt different, but suitable, behavioural strategies under changing situations? Thus, is it possible that a bonnet macaque female, as the dominant member of a grooming pair, is unable to attribute the correct motive to an approaching individual although she herself had had such a motive earlier as an approaching individual? If this is indeed true, bonnet macaques are similar to rhesus macaques, which were unable to empathize with and understand the motivations of their partners in a laboratory cooperative task, although they themselves had taken up similar roles earlier; as mentioned earlier, successful role reversal in these experiments was necessarily accompanied by fresh trial-and-error learning.
TACTICAL DECEPTION Human-like deception requires that an individual who signals information create a false belief in another individual, the audience. The signaller thus needs to recognize that the audience’s mind can be in a state of knowledge that is different from one’s own, and that it is possible to alter and hence, control others’ mental states without necessarily changing one’s own. Such manipulations are usually tactical in that they involve the use of acts from the normal repertoire of the actor in situations where they are likely to be misinterpreted by the audience—leading to some tangible benefit for the actor with or without some corresponding cost to the audience (Whiten and Byrne, 1988; Byrne and Whiten, 1990). All such acts of tactical deception are thus functional, and most cases of deception documented in primates can be included in this category (Byrne and Whiten, 1990). However, is primate deception truly intentional, attributable to a theory of mind (see Byrne and Whiten, 1991 for a theoretical discussion)? Does the deceiver actually attempt to alter the beliefs of another individual when it actively suppresses some information or signals false information to the other? Or, has experience simply taught the deceiver the use of certain behavioural strategies in particular situations, leading to predictable responses from the audience and thus allowing the actor to achieve a desired goal?
MIND-READING OR BEHAVIOUR-READING? We observed 128 events of social interactions in the three study troops of bonnet macaques that could potentially be considered deceptive, and the overwhelming majority
Other Minds: Social Cognition in Wild Bonnet Macaques
97
of these provide clear evidence for tactical deception over other competing explanations (Sinha, 2003). It is also noteworthy that although individuals in all the troops exhibited comparable levels of deception overall, the troops differed widely with regard to the social situations—competition for food, mates and grooming partners, as well as aggressive interactions—during which tactical deception was displayed (Sinha, 2003). There were also striking differences in the distribution of deceptive acts across 15 broad categories of deception commonly used by individuals in these troops (Figure 8.1). FIGURE 8.1 Distribution of deceptive acts across different categories of tactical deception exhibited by Troops G I, G II and B I 0.35
Troop G I Troop G II Troop B I
0.30
Proportion of acts
0.25 0.20 0.15 0.10 0.05 0.00
IA
IB
II
III
IV
V
VI
VII
VIII
IX
X
XI
XII
XIII XIV
Category of deception
Note: The categories of deception are: Simple—IA, Concealment by hiding behind a physical barrier; IB, Concealment by hiding away from the troop; II, Concealment by inhibiting interest in object; III, Concealment by ignoring; IV, Distraction by calling; V, Distraction by threat; VI, Distraction by close-range behaviour; VII, Creating a neutral image; VIII, Creating a positive image; IX, Deflection to third party; and Complex—X, Concealment by hiding behind a physical barrier and Concealment by inhibiting interest in object; XI, Concealment by inhibiting interest in object and Distraction by close-range behaviour; XII, Distraction by leading and Distraction by calling; XIII, Distraction by threat and Distraction by close-range behaviour; XIV, Concealment by hiding behind a physical barrier, Distraction by close-range behaviour and Distraction by threat.
A striking feature of the deception displayed by bonnet macaques is the remarkable individual variation in the performance of these acts. Certain individuals thus exhibited deceptive acts with very high frequency at levels significantly greater than that shown by other individuals within the troop; moreover, such deceptive abilities appeared to be
98
Anindya Sinha
independent of age categories and the dominance ranks of the actors. The fact that certain individuals are more adept at deception than others, and that the ability to deceive is independent of other individual attributes, including age, is an indication that many of these acts could involve mentalism on the part of the actor rather than simple behaviour analysis (since the latter would usually imply that rates of deception would increase with age and/or experience). It could, of course, be true that these particular individuals are good social learners and, therefore, more efficient behaviour analysts. This, however, seems unlikely, since it would require complex behavioural contingencies (given the type of situations where tactical deception is actually shown) to occur with high probabilities for individuals to learn such associations successfully; such contingencies appeared to be relatively rare in the social sphere of the study troops. If, on the one hand, macaques are indeed better social learners than mentalists while, on the other, complex social situations where deceptive behaviour could potentially be learnt are rare, it might be predicted that individuals who exhibit high levels of deception should perform the same acts repeatedly. However, for the deception displayed by males in all the study troops, there was significant positive correlation between the frequency of deceptive acts and the functional categories to which these acts belonged (Figure 8.2, Sinha, 2003); in other words, individuals who deceived at relatively higher levels, did so in many more different ways. This is an indication that these individuals may have indeed been better cheaters with perhaps greater insights into the power of manipulative behaviour than other individuals in the same troop. A particularly illuminating example comes from Troop B I, in which nine of the 16 acts of deception observed among the eight resident males were performed within a period of eight months by a single young subadult male who had recently joined the troop; remarkably, these nine acts belonged to nine different categories of deception! Moreover, certain rare acts of tactical deception displayed by the study individuals were extremely complex and involved several simple categories of deceptive acts juxtaposed together and performed in rapid succession to achieve one particular desired goal. The probability that these individuals had experienced an earlier identical behavioural contingency for them to learn all the constituent deceptive acts is indeed very low; moreover, virtually all these complex deceptive behavioural sequences were performed only on a single occasion each during the entire study period. If the argument put forward regarding the involvement of the mind in at least some of the acts of tactical deception displayed by bonnet macaques can be accepted, it would seem logical that such manipulation must necessarily involve at least second-order intentionality. This would mean, in simple terms, that an individual performs a deceptive act in order to change the belief system of the audience—and then takes advantage of the false belief, which has been generated, to achieve a particular personal goal.
Other Minds: Social Cognition in Wild Bonnet Macaques
99
FIGURE 8.2 Correlation between the number of deceptive acts and the number of categories of tactical deception in which they were performed by males in the three study troops 10 9
Categories of deception
8 7 6 5 4 3 2 1 0
0
1
2
3
4
5
6
7
8
9
10
Acts of deception
Note: Each point represents a single individual. Kendall’s rank correlation: = 0.6352, n = 17, p < 0.001.
VISUAL PERSPECTIVE-TAKING Several events of deception shown by individuals in two different troops involved acts of physical concealment in which the actor either simply hid from the target behind some physical object, or performed a behaviour surreptitiously behind a barrier, occasionally leaning out to inspect whether the target individual was still present. Since all these individuals repeatedly performed this exercise—in different contexts and using different objects or barriers to hide behind—these acts would appear to represent a genuine tactic, and were possibly not simply events coincident by chance alone. This kind of visual perspective-taking, estimating what would be visible from another individual’s point of view, has earlier been seen in other primates, notably chimpanzees and baboons (Whiten and Byrne, 1988). Such an ability to recognize and utilize the geometric perspective of another individual has been equated to being able to correctly represent another individual’s mental representation in one’s own mind, although there have also been dissenting views on such an identity (see Whiten, 1991 for a detailed discussion).
100 Anindya Sinha ‘INTENTION TO DECEIVE’ AS AN INTERVENING VARIABLE? Another characteristic feature of the tactical deception exhibited by bonnet macaques was that individuals did not invariably use deceptive strategies in apparently identical situations, a result not expected if these acts were being performed in response to certain behavioural contingencies alone. What is difficult to rule out, of course, is that there were subtle differences in these apparently identical situations—and these may have triggered off the deceptive acts in some of them, but not in others. A related finding to this form of volitional control of deception was that of some individual adult males who changed their repertoire of deceptive acts following changes in the social environment. This happened when two particular males emigrated out of one troop and joined a neighbouring one; following this movement, they displayed very different categories of deceptive acts, and one of them even exhibited a five-fold increase in the frequency of his deception. A major difference that these individuals faced in the two situations was that of their dominance ranks, which fell drastically once they had joined the new troop. It is, therefore, entirely possible that the perception of their specific positions in the rank hierarchy in the respective troops, as well as the changing demands of the new social milieu, may have triggered both a completely different repertoire and increased rates of deceptive acts in these two males. It is perhaps possible to model a complex set of stimulus–response links in the different social situations outlined above, leading to differential responses in terms of the deceptive acts displayed by specific individuals. It might, however, be more parsimonious to consider ‘intention to deceive’ as an intervening variable in these different situations, as outlined by Whiten (1996). This would mean that a variety of perceptual changes under different social conditions would be translated into either the presence or absence of potentially deceptive acts or into different forms of deceptive acts, all of these mediated by an intention or lack thereof to deceive. In addition to simplifying a possibly complex web of conditional stimulus–response chains, such an intervening variable would also be compatible with the notion of second-order intentionality underlying deception, outlined above.
AN INCOMPLETE THEORY OF MIND? Subordinate adult bonnet macaque males often give out loud predator alarm calls when they are chased by more dominant males—even if there are no predators in the vicinity. An extremely intriguing variant of this deceptive act was observed in one of the study troops. A victim of aggression emitted a false predator alarm call on being chased, but continued to give this call even as he descended from the tree and continued to walk on the ground—behaviour that would never have been performed if there was truly a predator around.
Other Minds: Social Cognition in Wild Bonnet Macaques
101
Deceiving individuals thus occasionally exhibit behavioural components that are not compatible with their own apparent ‘belief’ system, as communicated to others. An important point here is that notwithstanding its incompleteness, such a belief system must have been generated to alter the belief state of the audience—a return to secondorder intentionality. What is also noteworthy is that the aggressor did not appear to have read the internal inconsistency of the victim’s deceptive act; interestingly, this may have been due to his own theory of mind being similarly incomplete.
CONCLUSION Cercopithecine or Old World monkey societies are typically characterized by social relationships established between individuals belonging to different age cohorts, dominance ranks and kinship groups. Given the unique nature of each and every relationship that individuals need to develop and maintain, it is perhaps not surprising that bonnet macaques may be inherently capable of solving many complex social problems. These monkeys may, for example, observe the social interactions of other individuals in the troop and acquire knowledge of different attributes of these individuals, thus aiding their own decision-making during social interactions. Many individuals are also potentially capable of developing strategies of tactical deception; these strategies not only encompass different categories of deceptive acts, but are also employed in a variety of social situations, including agonistic interactions and competition over food, allogrooming companions and sexual partners. Underlying these complex social strategies may be the ability of individual macaques to form rudimentary mental representations of their social interactants and their various attributes, including their relative dominance ranks and social attractiveness. Interestingly, an elaborate example of tool manufacture and use by a bonnet macaque, documented earlier, indicated the possibility that the individual was able to perceive the underlying causality of its actions and also form a mental model of the tool to which it could repeatedly refer (Sinha, 1997). The cognitive ability to form mental representations could thus underlie the bonnet macaque’s interactions with both the mechanical as well as the social components of its environment. Analyses of the decision-making processes that bonnet macaques employ during social interactions indicate that individuals appear to attribute distinct motives to other individuals, a clear example of second-order intentionality. Moreover, several acts of tactical deception provide evidence that the macaques are capable of attributing visual perspectives to another individual, thus being able to perceive what would be visible from that particular individual’s point of view. This arguably constitutes another way in which a monkey is able to comprehend another monkey’s mental representation of the world—again a prime cognitive candidate for second-order intentionality.
102 Anindya Sinha Bonnet macaques, it can be argued, may thus have some degree of comprehension of the mental world of other individuals and are able to attribute distinct individuality to each other, including themselves. But does this imply that they have a theory of mind? It has been discussed above that, during social interactions, individuals may fail to project their own experiences on to others and are thus often unable to correctly predict the true motives of other individuals. Moreover, even in instances of tactical deception where the macaques communicate their apparent ‘beliefs’ to others, they exhibit behavioural components incompatible with their own beliefs. Extensive observational studies on the study troops have also so far failed to turn up any clear evidence for unambiguous thirdorder intentionality, which could be considered evidence for a true theory of mind. In conclusion, therefore, even if bonnet macaques do have a rudimentary theory of mind, it is a construct incomplete in many ways, some of which have been outlined here and some that still remain to be discovered.
ACKNOWLEDGEMENTS The writing of this paper was facilitated by a research grant from the Wenner-Gren Foundation for Anthropological Research, New York. I would like to express my gratitude to Anirban Datta-Roy for acquiring data on tactical deception from Troop B I in Bannerghata National Park, and to Shobini L Rao for inviting me to the ICCS2004 conference and for her infectious enthusiasm and stimulating discussions.
REFERENCES Bennett, J. 1991. ‘How to Read Minds in Behaviour: A Suggestion from a Philosopher’. In A. Whiten (ed.), Natural Theories of Mind: Evolution, Development and Simulation of Everyday Mindreading (pp. 97–108). Oxford: Basil Blackwell. Byrne, R.W. and A. Whiten. 1990. ‘Tactical Deception in Primates: The 1990 Database’. Primate Report, 27: 1–101. ———. 1991. ‘Computation and Mindreading in Primate Tactical Deception’. In A. Whiten (ed.), Natural Theories of Mind: Evolution, Development and Simulation of Everyday Mindreading (pp. 127–41). Oxford: Basil Blackwell. Chance, M.R.A. and A.P. Mead. 1953. ‘Social Behaviour and Primate Evolution’. Symposia of the Society for Experimental Biology, 7: 395–439. Cheney, D.L. and R.M. Seyfarth. 1990. How Monkeys See the World. Chicago: University of Chicago Press. Dennett, D.C. 1987. The Intentional Stance. MIT/Bradford Books, Cambridge: Massachusetts. ———. 1988. ‘The Intentional Stance in Theory and Practice’. In R.W. Byrne and A. Whiten (eds), Machiavellian Intelligence: Social Expertise and the Evolution of Intellect in Monkeys, Apes, and Humans (pp. 180–202). Oxford: Oxford University Press.
Other Minds: Social Cognition in Wild Bonnet Macaques
103
Fodor, J. 1968. Psychological Explanation. New York: Random House. Griffin, D.R. 2001. Animal Minds: Beyond Cognition to Consciousness. Chicago: University of Chicago Press. Humphrey, N.K. 1976. ‘The Social Function of Intellect’. In P.P.G. Bateson and R. Hinde (eds), Growing Points in Ethology (pp. 303–17). Cambridge: Cambridge University Press. Jolly, A. 1966. ‘Lemur Social Behaviour and Primate Intelligence’. Science 153: 501–06. Povinelli, D.J., K.A. Parks and M.A. Novak. 1992a. ‘Role Reversal by Rhesus Monkeys, But No Evidence of Empathy’. Animal Behaviour, 43: 269–81. Povinelli, D.J., K.E. Nelson and S.T. Boysen. 1992b. ‘Comprehension of Role Reversal in Chimpanzees: Evidence of Empathy?’ Animal Behaviour, 43: 633–40. Premack, D., and G. Woodruff. 1978. ‘Does the Chimpanzee have a Theory of Mind?’ Behavioural and Brain Sciences, 1: 515–26. Sinha, A. 1997. ‘Complex Tool Manufacture by a Wild Bonnet Macaque’. Macaca Radiata, Folia Primatologica, 68: 23–25. ———. 1998. ‘Knowledge Acquired and Decisions Made: Triadic Interactions During Allogrooming in Wild Bonnet Macaques’. Macaca Radiata, Philosophical Transactions of the Royal Society, London, Series B, 353: 619–31. ———. 2001. The Monkey in the Town’s Commons: A Natural History of the Indian Bonnet Macaque (NIAS Report R 2-01). Bangalore: National Institute of Advanced Studies. ———. 2003. ‘A Beautiful Mind: Attribution and Intentionality in Wild Bonnet Macaques’. Current Science, 85: 1021–30. Smuts, B., D.L. Cheney, R.M. Seyfarth, R.W. Wrangham and T. Struhsaker. 1987. Primate Societies. Chicago: University of Chicago Press. Tomasello, M. and J. Call. 1997. Primate Cognition. New York: Oxford University Press. Whiten, A. 1991. Natural Theories of Mind: Evolution, Development and Simulation of Everyday Mindreading. Oxford: Basil Blackwell. ———. 1993. ‘Evolving Theories of Mind: The Nature of Non-Verbal Mentalism in Other Non-Primates’. In S. Baron-Cohen, H. Tager-Flusberg and D.J. Cohen (eds), Understanding Other Minds: Perspectives from Autism (pp. 367–96). Oxford: Oxford University Press. ———. 1994. ‘Grades of Mindreading’. In C. Lewis and P. Mitchell (eds), Children’s Early Understanding of Mind: Origins and Development (pp. 47–70). Hove, UK: Erlbaum. ———. 1996. ‘When does smart Behaviour-reading become Mind-reading?’ In P. Carruthers and P.K. Smith (eds), Theories of Theories of Mind (pp. 227–92). Cambridge: Cambridge University Press. Whiten, A. and R.W. Byrne. 1988. ‘Tactical Deception in Primates’. Behavioural and Brain Sciences, 11: 233–73.
SECTION
II
Cognitive Neuroscience
C
ognitive neuroscience is an inter-disciplinary field of study linking the brain and other aspects of the nervous system to cognitive processing and, ultimately, to behaviour (Gazzaniga et al., 2000). It involves the study of biological foundations of mental phenomenon and is seen as a part of wide inter-disciplinary study of cognitive science. There are many reasons for the recent focus on cognitive neuroscience. Findings from cognitive neuroscience provide constraints for cognitive processes. Cognitive neuroscience provided the neuroscientists of previous generations to observe various cognitive functions in terms of brain processes, and its emergence was fueled by new methods that utilize high-technology tools that enabled them to measure brain activity. Significant progress has been made in the last 25 years to study the neural bases of perception, memory and language. The methods of cognitive neuroscience employ sophisticated techniques for measuring the activity of the brain such as single-cell recordings, positron emission tomography (PET), functional magnetic resonance imaging (fMRI), electroencephalogram (EEG), and so on. Brain lesions can also be localized with amazing precision, owing to such methods of imaging the brain. These techniques provide vital information for clinical investigations, neurological diagnosis, as well as research in the domain of cognitive science. Brain activity is primarily a measure of either electrochemical processes, or changes in blood flow and metabolism in the brain. It is believed that the human brain functions via specific anatomical relays to circumscribed areas of the brain in the form of well-defined pathways. These pathways get highlighted as a result of modulation of electrical activity in the brain in response to a particular task. EEG provides a continuous measure of global electrical activity. The embedded evoked response can be extracted from the global signal using various methods. The most exciting methodological advances for cognitive neuroscience have been provided by techniques that identify anatomical correlates of cognitive processes such as PET and fMRI. These techniques measure metabolic changes that occur as a result of changes in the energy requirements of the nervous system in the form of oxygen and glucose to sustain their cellular integrity and to perform specialized functions. These metabolic changes can be correlated with the neuronal activity that techniques such as EEG attempt to measure indirectly. These techniques represent a remarkably successful synergy of developments in imaging technology, neuroscience and cognitive science, which has led to the development of new theories on human cognition. Cognitive neuroscience also aims to understand cognition by decomposing the mental operations and their neural substrates. It has been established that the decomposition of mental events can be united with an understanding of mental and emotional computations carried out by human brain (Posner and DiGirolamo, 2000). Cognitive neuroscience has been applied to understand various cognitive domains such as perception, language, attention, memory, and so on. This is an altogether different approach towards understanding cognitive functions in the brain. It provides valuable insight into the understanding of basic capabilities of cortical as well sub-cortical
108 Advances in Cognitive Science structures. The functional asymmetries and interactions between brain regions help to understand cognitive processes better. It addresses problems as basic as those related to the human language system and the organization of conceptual knowledge in the brain. The neuropsychological lesion approach epitomized by research on language disorders represents only one way to investigate biological bases of language. Additional evidences from functional neuroimaging, studies involving stimulation of the human cortex, and recordings related to language-related brain potentials from healthy and brain-damaged patients yield information about normal language system that may not be obvious from the neuropsychological approaches alone. Cognitive neuroscience also addresses subtleties of cognitive function such as role of attention in modulation of information processing, emotional processing and decisionmaking. This has led to the evolution of critical structures and derivation of networks in the brain that are responsible for emotional processing. It also elucidates the neural activity that underlies basic cognitive processes such as memory. Thus, it attempts to understand the neural concomitants of memory processes such as encoding, retrieval and consolidation. Cognitive neuroscience also derives the relationships between neural structures that work in coordination to enhance certain cognitive functions. This emerging shift in our approach to study human cognition also provides vital information about hemispheric specialization of cognitive functions. Given the recent advances in cognitive neuroscience, this section focuses on this approach to study various cognitive processes. One sub-discipline in which the integration of cognitive psychology and neuroscience has been achieved significantly is visual perception. The chapter by Chaudhuri discusses techniques based on molecular mapping and its effective use in studying the visual system. Chaudhuri discusses the strengths and weaknesses of the molecular mapping technique followed by a discussion of stimulation or deprivation studies using molecular mapping to study the visual system. Chaudhuri ends with a discussion on the recent application of molecular mapping in understanding complex visual function. Cognitive neuroscience of language has grown tremendously in the past decade or so with the advent of brain imaging using PET and fMRI. This is in conjunction with studies on Event Related Potentials (ERPs) as well as lesion-based studies, which have thrown light on the neural mechanisms and processes underlying language. Patients with various brain disorders have been used to investigate the mechanisms underlying all cognitive processes, including language. Language-related processes have been found to be localized in different areas like Broca’s area and Wernicke’s area. Broca’s area has been implicated in language production, whereas Wernicke’s area has been implicated in language comprehension. Vaid provides a critical overview of the evidences pertaining to the bilingual brain to identify points of convergence, and suggests directions for further research. He critically looks at imaging experiments to understand the neural processes underlying language in bilinguals, and has discussed evidences on brain-injured and
Cognitive Neuroscience
109
normal bilinguals, behavioural laterality and neurobehavioural evidences. Her chapter highlights the need for better conceptualization of bilingualism in the context of its very dynamic nature, which is the first requirement for a better understanding of the neural mechanisms of bilingualism. Another interesting aspect of the brain and behaviour is the asymmetry. The chapter by Mandal discusses various forms of functional asymmetry in behaviour and possible causes that may give rise to such asymmetries. Mandal et al. examine various types of asymmetry, which they refer to as ‘side bias’. They discuss different forms of side bias such as motor bias and cognitive bias. Motor bias includes handedness, footedness, earedness, eyedness and facedness. Cognitive biases include visual field bias, auditory bias and attention bias. The discussion of cognitive biases is followed by an extensive discussion on their research on bias in facial expression. The chapter ends with a discussion on handedness and also discusses findings from different cultures on the prevalence on handedness. The results show the prevalence of functional asymmetry in behaviour and point to the possibility of cerebral hemispheres in side bias.
REFERENCES Gazzaniga, M.S., R.B. Ivry and G.R. Mangun. 2000. Cognitive Neuroscience: The Biology of Mind. New York, USA: W.W. Norton & Company, Inc. Posner, M.I. and G.J. DiGirolamo. 2000. ‘Cognitive Neuroscience: Origins and Promise’. Psychological Bulletin, 126(6): 873–89.
Chapter 9 A Survey of Molecular Mapping as Applied to Studies of the Visual System Avi Chaudhuri
INTRODUCTION
T
he immediate early genes (IEGs) c-fos and zif268 have become popular neurobiological tools for mapping functional activity in the brain. We now know that the expression of these genes is strongly, though not exclusively, linked to synaptic stimulation, and that their products may be involved in the key aspects of normal cellular function. Although much progress has been made towards understanding the intracellular processes that guide the expression of these genes, the precise physiological roles of the proteins encoded by them remain largely unknown. Many members of the immediate early genes (IEG) family are activated shortly after neuronal stimulation and without the requirement for de novo protein synthesis. The products of many of these genes have been observed in a wide range of tissues and under a variety of stimulation conditions (reviewed in Kaczmarek and Chaudhuri, 1997). The rapid accumulation of IEG products in activated neurons, combined with histological methods that offer detection at the cellular level, are key features that have led to their wide use in visualizing activated neurons.
CHARACTERISTICS OF INDUCIBLE TRANSCRIPTION FACTORS The various members of the IEG family are known to encode proteins with diverse cellular functions. In the brain, IEGs that are linked to neural activity and that have mapping applications generally encode proteins that serve as transcription factors, that is, complexes that bind to the promoter regions of certain genes, and then either activate or repress their expressions. These so-called inducible transcription factors (ITFs) are distinguished from other proteins that normally reside in the nucleus and which also serve as regulators of ongoing transcriptional activity.
112 Avi Chaudhuri Both the ITFs c-fos and zif268 are induced in neurons after extracellular stimulation by neurotransmitters and trophic substances. The sequence of events that leads to ITF induction is largely coordinated by Ca2+ influx into the cell (Ghosh et al., 1994). This can occur either through the N-methyl D-aspartate (NMDA) receptor–Ca2+ ionophore complex after glutamate binding, or through voltage-sensitive calcium channels (VSCCs) following membrane depolarization. Thereafter, several different enzyme systems are marshalled by Ca2+. After transcription is completed, the ITF mRNAs are translated into a protein product (c-fos and zif268) in the cytosol. These products rapidly migrate into the nucleus where they themselves influence the expression of another set of genes, the late-response genes. In the case of c-fos, it must dimerize with a member of the Jun phosphoprotein family (c-Jun, JunB, or JunD) to produce a functional transcription factor that is called activating-protein 1 (AP-1). By regulating the expression of a host of late-response genes, both AP-1 and zif268 are able to have a commanding influence on short- and long-term cellular homeostasis. Regardless of what the specific molecular roles of ITFs may be, their accumulation in the neuron generally signifies a prior state of activity and thus forms the logic for obtaining functional maps based on ITF staining.
STRENGTHS AND CAVEATS OF MOLECULAR MAPPING Molecular mapping strategies have distinct advantages over other functional imaging approaches, and also suffer from some unique problems. The following are brief accounts of the key advantages offered by IEG activity mapping.
Rapid Induction One of the striking advantages of IEG use in activity mapping is the short time course of induction. It appears that the intracellular mechanisms operate in a very rapid manner, such that mRNA levels are detectable in neurons within 20 minutes after stimulation. The time course for IEG protein expression is somewhat greater, most likely due to the added requirement for translation to be completed. However, depending on the particular brain structure examined, protein induction can be detected as early as 30 minutes after the onset of stimulation and appears to peak at 60–90 minutes. The rapid induction of IEGs means that short stimulation times are sufficient, allowing its use in many behavioural situations.
Cellular Resolution Mapping Another key advantage is the superb spatial resolution of IEG activity maps. Histological labelling of IEG proteins, such as c-fos and zif268, produces staining that is confined to
A Survey of Molecular Mapping as Applied to Studies of the Visual System
113
the nucleus. This provides a punctate visual display of immunostained neurons that actually extends the resolution even beyond the cellular level. This feature allows precise laminar and spatial localization, especially in complex structures or small nuclei within the brain.
Application with Other Histological Techniques A particular advantage of the nuclear locus of the immunostained product is that doublelabelling procedures can be employed by counterstaining the same tissue section for various products that are confined to the cytoplasm. This allows one to correlate the expression of a particular IEG product with endogenous markers to reveal physiological or morphological features of activated neurons. Furthermore, molecular mapping can be used in tandem with tract-tracing procedures to identify input-output relationships.
Large-Scale Analysis Molecular mapping techniques require analysis of post-mortem tissue. Therefore, it is possible to undertake analysis on a large-scale basis to precisely identify the various cortical and subcortical brain structures that may be activated after sensory, behavioural or pharmacological manipulation. The following are some of the disadvantages of molecular mapping approaches: 1. Temporal Resolution: The IEG induction process is not very dynamic, and therefore much less sensitive to temporal changes than some other methods, such as electrophysiological recording and optical imaging. As a result, subtle changes in stimulation or behavioural treatment during an experiment can go undetected and become blurred in the resulting activity map. 2. Quantitative Analysis: While immuno-cytochemical (ICC) staining provides good qualitative results, quantitative analysis is difficult because of the non-linear nature of substrate amplification and technical problems of accurately determining the signal strength. Some quantization, however, is still possible by means of cell counts of immunostained or mRNA-stained neurons, allowing for a comparison of the effects of different stimulation or behavioural treatments. 3. Neuronal Expression Specificity: It has been found that not all neurons express IEG products. For example, both c-fos and zif268 are sparsely expressed in numerous subcortical structures, and even the neocortex appear to be largely restricted to excitatory neurons. Thus, activity maps do not reflect accumulation of induced IEG products across a broad range of neurons, but rather only in a restricted set. The absence of labelling in a particular neural structure therefore cannot merely
114 Avi Chaudhuri be taken as a lack of activation because certain IEGs may not be inducible in that structure. 4. False Positives: There have been several demonstrations of dissociation between IEG expression and behavioural or physiological assay. In some cases, a positive IEG label has not been supported by other mapping techniques (for example, 2-deoxyglucose accumulation), or behavioural condition of the animal (for example, nociceptive response). False positive results have generally been attributed to other extraneous events, leading to the notion that neural activity does not have an exclusive association with IEG expression. Other factors, such as endocrine responses, can sometimes be implicated in IEG induction that may either be in addition to or in lieu of the primary stimulus effects being mapped. 5. Multiple Stimulation Condition: It is extremely difficult to obtain activity maps in response to more than one stimulation sequence. This is because IEG products are unitary activity markers. If two compounds or successive stimuli are applied, then neurons equally sensitive to either will express the IEG products in equal measure, making it difficult to differentiate the separate response to each stimulation sequence. Recent developments in double-labelling that exploit the differential time course of mRNA versus protein induction provide one solution to this problem.
STUDIES OF THE VISUAL SYSTEM Molecular mapping efforts to study visual function can be broadly classified into four areas—activity-mapping of sensory function in the mammalian central nervous system; the influence of photoperiod on immediate-early gene induction; maps of developmental change and their postulated relationship to plasticity; and exploring the neural substrates of complex factors on visual processing. The basal distribution of several immediateearly genes is known for several visual structures in different species (Kaminska et al., 1996, 1999). A number of studies have conclusively shown that the expression of several immediate-early genes is rapidly modulated by either application or removal of visual input.
MOLECULAR MAPS OF VISUAL FUNCTION The study of visual function with immediate-early gene products has been approached by way of either visual stimulation or deprivation (reviewed in Chaudhuri, 1997; Chaudhuri et al., 2000; Kaczmarek and Chaudhuri, 1997). Deprivation studies have been undertaken by way of simple eyelid suture, administration of a sodium channel blocker such as tetrodotoxin (TTX) into the eye, or complete unilateral or bilateral enucleation. The latter
A Survey of Molecular Mapping as Applied to Studies of the Visual System
115
approach raises the possibility that sensory effects in visual neurons may be compromised by response to nerve injury. In general, enucleation is not appropriate for short-term deprivation studies because of the associated trauma and possible upstream degenerative complications that may follow severing of the optic nerve. Among the immediate-early genes, zif268 has been a favoured candidate for deprivation studies because of its normally high level of expression in neocortical areas (Chaudhuri et al., 2000; Kaczmarek and Chaudhuri, 1997). It has also been shown that visual disruption in one eye leads to rapid down-regulation of zif268 basal levels in the contralateral rodent visual neocortex, being most pronounced in the thalamorecepient layer IV. Similar results have been found in cat neocortical areas, where visual deafferentation produces a timedependent decrease of zif268 mRNA (Zhang et al., 1994), and prolonged dark-rearing leads to a marked reduction in zif268 protein levels (Kaplan et al., 1995, 996). To the contrary, products of the c-fos gene are rather poor markers of deprivation effects because of their generally low level of basal expression (Kaczmarek and Chaudhuri, 1997). However, some decline in c-fos mRNA levels have been reported after visual differentation in the cat (Zhang et al., 1994). One species where visual deprivation has been used to map cortical activity is the monkey. It has been shown that brief monocular deprivation, as little as two hours, is sufficient to reveal the details of ocular dominance architecture at cellular levels (Chaudhuri and Cynader, 1993; Chaudhuri et al., 1995). The inter-digitated pattern of ocular dominance columns within each hemisphere provides a valuable internal control because it permits a more effective comparison between activity-induced expression, and neural quiescence upon immediate-early gene expression. Molecular maps of zif268 expression within primate ocular dominance columns have been used to determine the temporal details of induction, assess the cellular morphology of immunopositive neurons, and explore developmental patterns and their possible relationship to ongoing neuroplastic events (Chaudhuri et al., 1997; Markstahler et al., 1998). A second, and more common, way to obtain IEG maps is by inducing their expression through visual stimulation. In general, these experiments require a period of darkadaptation prior to stimulation in order to reduce basal expression and background staining. Furthermore, the transition from a quiescent period to intense synaptic stimulation affects the expression of ITFs, revealing the extent to which these genes are up-regulated in response. Visual stimulation experiments in rats have provided laminar details of c-fos expression in response to patterned visual stimulation (Montero and Jian, 1995), neural activity profiles in response to ultraviolet light (Amir and Robinson, 1996), effects of altering the temporal frequency of light presentation (Correa-Lacarcel et al., 2000), and the role of central noradrenergic input in guiding plastic change through modulation of c-fos expression (Yamada et al., 1999). One interesting application of c-fos mapping has been to establish that functional connectivities are made by transplanted foetal retinae onto the dorsal midbrain of neonatal rats. Craner et al. (1992) showed that
116 Avi Chaudhuri transplanted retinae exposed to light stimulation produce patterns of c-fos expression in visual neural structures, which were similar to those seen upon stimulation of the intact eyes. Induction experiments in higher animals have largely revolved around the issue of developmental sequences of cortical maturation and their corresponding effects upon IEG expression. Mower and colleagues (Kaplan et al., 1996; Mower, 1994; Mower and Kaplan, 1999) have shown that different laminar patterns of c-fos expression are evident in kittens versus adult cats after brief visual stimulation. Furthermore, the induction effects in general were found to be greater in young animals, a result that was consistent with that found in rodents (Worley et al., 1990). A similar result has been reported for the zif268 gene as well. Both mRNA and protein products are elevated after visual stimulation, the increase being more notable in young animals (Nedivi et al., 1996; Rosen et al., 1992). zif268 induction after dark adaptation has also been used to show ocular dominance profiles in the striate cortex of developing and adult primates (Chaudhuri et al., 1995; Kaczmarek et al., 1999; Markstahler et al., 1998).
NEURAL SUBSTRATES OF COMPLEX VISUAL FUNCTION The use of molecular mapping strategies to observe activation in response to complex visual stimuli represents a new and exciting direction for this field (Broad et al., 2000; Miyashita et al., 1998; Zangenehpour and Chaudhuri, 2005). Miyashita et al. (1998) have used zif268 mapping in monkey inferotemporal cortex to show that visual paired associate learning produces increased immunostaining in superficial and deep layers. The study by Broad et al. (2000) examined c-fos mRNA expression in several brain areas, including inferotemporal cortex, amygdala and hippocampus of sheep after they were exposed to normal and inverted facial stimuli. They found that face recognition in sheep preferentially engages the right temporal cortex. There also appeared to be activation of the amygdala and hippocampus that reflected downstream activation after active choice between upright faces. Interestingly, inverted faces, which are known to be perceived in a less veridical manner in humans, did not show enhanced activation of either the inferotemporal cortex, amygdala or hippocampus. The lateralization effects observed in this animal model are consistent with human data. Behavioural experiments suggest that IEG induction in visual cortex may be linked to stimulus novelty, and can often be critical for c-fos activation. Quantitative analysis of zif268 expression in the visual cortex has shown that it is expressed at significantly higher levels in rats exposed to a complex housing environment in comparison to those housed individually and without any handling (Wallace et al., 1995). It may be observed that the accumulation of zif268 mRNA in animals reared in an enriched environment reflects a higher level of neuronal activity that correlates with plastic changes of sensory cortex.
A Survey of Molecular Mapping as Applied to Studies of the Visual System
117
Several research groups have investigated the possible effects of behavioural training with complex stimuli on IEG expression. A single training session of two-way active avoidance behaviour with darkness being used as a conditioned stimulus (CS) resulted in a large accumulation of c-fos and zif268 mRNA levels in various brain regions, including the visual cortex. Long-term training with darkness as CS for nine days up to an asymptotic level of performance produced negligible c-fos activation. Active exploration has been shown to produce marked increase in c-fos mRNA in visual cortex of rats, a result that has been attributed to arousal, stress and stimulus novelty (Hess et al., 1995a). The role of stimulus novelty in selective cortical activation has been explored in further detail by Montero (1997, 1999, 2000). He has suggested that attentional factors play a dominant role in the enhancement of c-fos induction that is seen in rat visual and somatosensory cortices after exploring a novel environment. Furthermore, the known interaction between cortical and subcortical sites was postulated to produce an attentiondependent gating of thalamocortical transmission. Attentional activation of the thalamic reticular nucleus (TRN) was significantly diminished in animals with monocular postnatal deprivation (Montero, 1999). Furthermore, c-fos mapping showed that activation of the TRN was reduced after unilateral visual cortex lesions, but sparing activity in the Lateral Geniculate Nuclei (LGN). This suggests that top–down control from the sensory neocortex is responsible for the activation of TRN by attentive exploration of a novel complex environment (Montero, 2000).
REFERENCES Amir, S. and B. Robinson. 1996. ‘Fos Expression in Rat Visual Cortex Induced by Ocular Input of Ultraviolet Light’. Brain Research, 716: 213–18. Broad, K.D., M.L. Mimmack and K.M. Kendrick. 2000. ‘Is Right Hemisphere Specialization for Face Discrimination Specific to Humans?’ European Journal of Neuroscience, 12: 731–41. Chaudhuri, A. 1997. ‘Neural Activity Mapping with Inducible Transcription Factors’. Neuroreport, 8: v–ix. Chaudhuri, A. and M.S. Cynader. 1993. ‘Activity-Dependent Expression of the Transcription Factor Zif268 Reveals Ocular Dominance Columns in Monkey Visual Cortex’. Brain Research, 605: 349–53. Chaudhuri, A., J.A. Matsubara and M.S. Cynader. 1995. ‘Neuronal Activity in Primate Visual Cortex Assessed by Immunodetection for the Transcription Factor Zif268’. Visual Neuroscience, 12: 35–50. Chaudhuri, A., J. Nissanov, S. Larocque and L. Rioux. 1997. ‘Dual Activity Maps in Primate Visual Cortex Produced by Different Temporal Patterns of zif268 mRNA and Protein Expression’. Proceedings of the National Academy of Science of the United States of America, 94: 2671–75. Chaudhuri, A., S. Zangenehpour, F. Rahbar-Dehgan and F. Ye. 2000. ‘Molecular Maps of Neural Activity and Quiescence’. Acta Neurobiologiae Experimentalis, 60: 403–10. Correa-Lacarcel, J., M.J. Pujante, F.F. Terol, V. Almenar-Garcia, A. Puchades-Orts, J.J. Ballesta, J. Lloret, J.A. Robles and F. Sanchez-del-Campo. 2000. ‘Stimulus Frequency Affects c-fos Expression in the Rat Visual System’. Journal of Chemical Neuroanatomy, 18: 135–46.
118 Avi Chaudhuri Craner, S.L., G.E. Hoffman, J.S. Lund, A.L. Humphrey and R.D. Lund. 1992. ‘cFos Labeling in Rat Superior Colliculus: Activation by Normal Retinal Pathways and Pathways from Intracranial Retinal Transplants’. Experimental Neurology, 117: 219–29. Ghosh, A., D.D. Ginty, H. Bading and M.E. Greenberg. 1994. ‘Calcium Regulation of Gene Expression in Neuronal Cells’. Journal of Neurobiology, 25: 294–303. Hess, U.S., G Lynch and C.M. Gall. 1995. ‘Changes in c-fos mRNA Expression in Rat Brain during Odor Discrimination Learning: Differential Involvement of Hippocampal Subfields CA1 and CA3’. Journal of Neuroscience, 15: 4786–795. Kaczmarek, L. and A. Chaudhuri. 1997. ‘Sensory Regulation of Immediate-Early Gene Expression in Mammalian Visual Cortex: Implications for Functional Mapping and Neural Plasticity’. Brain Research. Brain Research Reviews, 23: 237–56. Kaczmarek, L., S. Zangenehpour and A. Chaudhuri. 1999. ‘Sensory Regulation of Immediate-Early Genes c-fos and zif268 in Monkey Visual Cortex at Birth and Throughout the Critical Period’. Cerebral Cortex, 9: 179–87. Kaminska, B., L. Kaczmarek and A. Chaudhuri. 1996. ‘Visual Stimulation Regulates the Expression of Transcription Factors and Modulates the Composition of AP-1 in Visual Cortex’. Journal of Neuroscience, 16: 3968–78. Kaminska, B., L. Kaczmarek, S. Zangenehpour and A. Chaudhuri. 1999. ‘Rapid Phosphorylation of Elk-1 Transcription Factor and Activation of MAP Kinase Signal Transduction Pathways in Response to Visual Stimulation’. Molecular and Cellular Neuroscience, 13: 405–14. Kaplan, I.V., Y. Guo and G.D. Mower. 1995. ‘Developmental Expression of the Immediate Early Gene EGR-1 Mirrors the Critical Period in Cat Visual Cortex’. Brain Research. Developmental Brain Research, 90: 174–79. ———. 1996. ‘Immediate Early Gene Expression in Cat Visual Cortex During and After the Critical Period: Differences Between EGR-1 and Fos Proteins’. Brain Research. Molecular Brain Research, 36: 12–22. Markstahler, U., M. Bach and W.B. Spatz. 1998. ‘Transient Molecular Visualization of Ocular Dominance Columns (ODCs) in Normal Adult Marmosets Despite the Desegregated Termination of the RetinoGeniculo-Cortical Pathways’. Journal of Comparative Neurology, 393: 118–34. Miyashita, Y., M. Kameyama, I. Hasegawa and T. Fukushima. 1998. ‘Consolidation of Visual Associative Long-Term Memory in the Temporal Cortex of Primates’. Neurobiology of Learning and Memory, 70: 197–211. Montero, V.M. 1997. ‘c-fos Induction in Sensory Pathways of Rats Exploring a Novel Complex Environment: Shifts of Active Thalamic Reticular Sectors by predominant Sensory Cues’. Neuroscience, 76: 1069–81. ———. 1999. ‘Amblyopia Decreases Activation of the Corticogeniculate Pathway and Visual Thalamic Reticularis in Attentive Rats: A ‘Focal Attention’ Hypothesis’. Neuroscience, 91: 805–17. ———. 2000. ‘Attentional Activation of the Visual Thalamic Reticular Nucleus Depends on ‘Top-Down’ Inputs from the Primary Visual Cortex Via Corticogeniculate Pathways’. Brain Research, 864: 95–104. Montero, V.M. and S. Jian. 1995. ‘Induction of c-fos Protein by Patterned Visual Stimulation in Central Visual Pathways of the Rat’. Brain Research, 690: 189–99. Mower, G.D. 1994. ‘Differences in the Induction of Fos Protein in Cat Visual Cortex During and After the Critical Period’. Brain Research. Molecular Brain Research, 21: 47–54. Mower, G.D. and I.V. Kaplan. 1999. ‘Fos Expression During the Critical Period in Visual Cortex: Differences Between Normal and Dark Reared Cats’. Brain Research. Molecular Brain Research, 64: 264–69.
A Survey of Molecular Mapping as Applied to Studies of the Visual System
119
Nedivi, E., S. Fieldust, L.E. Theill and D. Hevroni. 1996. ‘A Set of Genes Expressed in Response to Light in the Adult Cerebral Cortex and Regulated During Development’. Proceedings of the National Academy of Science USA, 93: 2048–53. Rosen, K.M., M.A. McCormack, L. Villa-Komaroff and G.D. Mower. 1992. ‘Brief Visual Experience Induces Immediate Early Gene Expression in the Cat Visual Cortex’. Proceedings of the National Academy of Science of the United States of America, 89: 5437–41. Wallace, C.S., G.S. Withers, I.J. Weiler, J.M. George, D.F. Clayton and W.T. Greenough. 1995. ‘Correspondence Between Sites of NGFI-A Induction and Sites of Morphological Plasticity following Exposure to Environmental Complexity’. Brain Research. Molecular Brain Research, 32: 211–20. Worley, P.F., A.J. Cole, T.H. Murphy, B.A. Christy, Y. Nakabeppu and J.M. Baraban. 1990. ‘Synaptic Regulation of Immediate-early Genes in Brain’. Cold Spring Harbor Symposium on Quantitative Biology, 55: 213–23. Yamada, Y., Y. Hada, K. Imamura, N. Mataga, Y. Watanabe and M. Yamamoto. 1999. ‘Differential Expression of Immediate-Early Genes, c-fos and zif268, in the Visual Cortex of Young Rats: Effects of a Noradrenergic Neurotoxin on their Expression. Neuroscience, 92: 473–84. Zangenehpour, S. and A. Chaudhuri. 1999. ‘Multisensory Neurons of the Superior Colliculus Revealed by Dual Fluorescent ICC/ISH Labeling Using c-fos and zif268 Expression Profiles’. Society for Neuroscience Abstracts, 25: 1414. ———. 2005. ‘Patchy Organization and Asymmetric Distribution of the Neural Correlates of Face Processing in Monkey Inferotemporal Cortex’. Current Biology, 15: 993–1005. Zhang, F., P. Halleux, L. Arckens, W. Vanduffel, L. Van Bree, P. Mailleux, F. Vandesande, G.A. Orban and J.J. Vanderhaeghen. 1994. ‘Distribution of Immediate Early Gene zif-268, c-fos, c-jun and jun-D mRNAs in the Adult Cat with Special References to Brain Region Related to Vision’. Neuroscience Letters, 176: 137–41.
Chapter 10 Neural Substrates of Language Processing in Bilinguals: Imagi(ni)ng the Possibilities Jyotsna Vaid
INTRODUCTION
E
ver since the association between the left cerebral hemisphere and language was discovered in the late 1800s, questions about the neural substrates of multiple language experience have been the subject of much speculation. There have been three waves of research on this topic. The first was triggered by the publication of a monograph reviewing behavioural and clinical evidence on the bilingual brain (Albert and Obler, 1978), and by a subsequent anthology of polyglot aphasia case reports from European journals (Paradis, 1983). The second was triggered by studies of cortical electrical stimulation to localize language regions in bilingual patients undergoing removal of seizure-inducing brain regions (Ojemann and Whitaker, 1978), and studies of language lateralization in brain-intact bilinguals (Vaid and Genesee, 1980; Vaid and Hall, 1991). The third wave has been characterized by haemodynamic and electrophysiological studies of intra-hemispheric neural activation patterns during language processing in healthy bilinguals (Vaid and Hull, 2002; Hull and Vaid, 2005). In considering the vast and complex literature on neural substrates of bilingualism, it is important first to consider the theoretical significance of this topic.
WHY STUDY THE BILINGUAL BRAIN? The bilingual brain is of interest for the simple reason that the majority of the world’s language users are bi- or multilingual, rather than monolingual, and thus research and theory should acknowledge this fact rather than frame bilingualism as a deviation from the norm. Until recently, research on language and the brain largely focussed on monolinguals
Neural Substrates of Language Processing in Bilinguals
121
and thus addressed only a fraction of actual language users. Given that multiple language users differ widely in when, how and how well they acquire language, the study of bilinguals provides a way of addressing issues of neural plasticity, that is, functional and/or structural changes in the brain in response to variations in experience. Studies of synaptic density and brain glucose metabolism point to the changes in the organization of certain neural structures following alterations in sensory experience, for example, blindness or deafness. Whereas some neural systems appear to change very little even under extreme changes in early experience, other systems undergo considerable change, particularly during certain limited time periods, and still other neural systems appear to be modifiable by experience throughout life (Neville and Bavelier, 2002). Recent studies suggest that differences in musical experience alter neural sensitivity (Munte et al., 2002). For example, violinists who had learned to play music from early childhood show greater somatosensory representation of the fingers of their left hand as compared to those who learned to play later in life (Elbert et al., 1995), and there is evidence for greater grey matter in skilled musicians as compared to less-skilled musicians (Gaser and Schlaug, 2003). To the extent that bilingualism covers a broad spectrum of language experience, the study of bilingual individuals may provide critical insights into brain plasticity in response to variations in language experience.
ORGANIZATION OF THE CHAPTER The literature on the bilingual brain is characterized by a diversity of theoretical and methodological approaches and a multiplicity of research outcomes (see Ijalba et al., 2004; Paradis, 2004; Vaid, 2002, 2008; Vaid and Hall, 1991, for reviews). A core issue in both clinical and experimental research has been to determine whether or under what circumstances acquiring and using more than one language has distinct neural repercussions. For both methodological and interpretive reasons, there is no clear consensus on this issue to date. The present overview seeks to consider all available sources of evidence pertaining to the bilingual brain, assess their strengths and limitations, identify points of convergence, and suggest directions for further research.
STUDIES WITH BRAIN-INJURED BILINGUAL POPULATIONS Aphasia Clinical sources constitute the earliest evidence pertaining to the organization of multiple languages in the brain. In particular, case studies of patterns of language loss and recovery in individuals with aphasia have fueled much interest for their potential bearing on issues
122 Jyotsna Vaid of language localization within the brain. In so-called polyglot aphasics, a variety of patterns of language impairment and recovery following aphasia have been documented, ranging from completely parallel loss or recovery to selective or differential loss or recovery, that is, where the post-morbid pattern of impairment or recovery of one language relative to the other does not parallel pre-morbid use patterns. One interpretation of differential loss or recovery has been that the languages are differentially localized within the languagedominant hemisphere, or across the hemispheres. An alternative view of non-parallel recovery has also been entertained, in terms of an impairment in access to one or another language, rather than an actual impairment in representation (Green, 1998). Since the majority of the 300 or so reported cases of polyglot aphasia in the early neurological literature have been drawn from single, selected reports rather than unselected group studies, there is a real question as to their representativeness. A focus of more recent research on polyglot aphasia, thus, has been to develop repositories of unselected cases, using standardized aphasia assessment of the patients in each of their languages (for example, Paradis, 1987), with the goal of eventually determining the relative frequency of the different types of recovery patterns observed and factors that could account for the occurrence of specific patterns. Based on the few group-level studies of unselected cases of bilingual aphasia to date, the consensus appears to be that instances of differential patterns of impairment and instances of crossed aphasia in bilinguals (that is, aphasia following damage to the right hemisphere (RH) in right-handers) are fairly uncommon (Fabbro, 2001; Paradis, 2001). However, it should be noted that the relevant database is still fairly limited, and that regression or other statistical analyses based on data from large-scale clinical studies have typically not been undertaken.
Cortical Stimulation Another important source of evidence from clinical populations has been studies of cortical electrical stimulation of the exposed cortex in epileptic individuals, performed in order to map critical language areas prior to surgical removal of seizure-inducing foci. These studies, conducted with a few dozen patients to date, mostly late bilinguals, have reported a pattern of partially overlapping language areas for object naming, counting, reading or writing the tasks most typically used (Lubrano et al., 2004; Lucas et al., 2004; Roux and Tremoulet, 2002; Roux et al., 2004). That is, some regions that are stimulated result in disruption of these tasks in both languages of the patient, whereas others result in disruption in only one or the other language. Furthermore, sites where there was greater disruption in the less-proficient language tend to span a broader region compared to those in which naming disruption was in the more proficient language. However, this pattern is not consistently observed across studies, and the specific locations of regions that show differential naming impairment vary across patients, making generalizations difficult.
Neural Substrates of Language Processing in Bilinguals
123
Taken together, findings from the aphasia literature and cortical stimulation studies are important in that they document the variety of patterns that can occur following aphasia or following cortical stimulation. However, studies of brain bases of language functioning based on brain-injured populations have distinct limitations for drawing conclusions about neural organization of language in the intact brain. This is because of the possibility that in patients with a long history of seizures or those experiencing traumatic injuries, there is compensatory reorganization of function. A recent Magnetoencephalography (MEG) study, for example, showed that a majority of patients with early onset of seizures (before age five) showed atypical language lateralization, and a third of patients with left hemisphere (H) lesions showed intrahemispheric reorganization of receptive language function (Pataraia et al., 2004). While such results caution against extrapolating findings from lesion-deficit studies to studies of the healthy brain, evidence from patients with lesions complements that obtained from other sources and, as such, should not be completely discounted (Rorden and Karnath, 2004).
STUDIES WITH BRAIN-INTACT BILINGUALS Behavioural Laterality Studies of cerebral functional asymmetry constitute the most extensive empirical body of evidence on language and the brain in brain-intact bilinguals. Exploiting the contralateral sensory organization of the brain by presenting linguistic input lateralized to the left or right visual field or ear, these studies allow for an examination of the relative pattern of asymmetry across the bilingual’s two languages, between either languages of certain bilinguals compared to other bilinguals, and between bilinguals and monolinguals. The canonical pattern observed with monolinguals has been H dominance for language, particularly for phonological and syntactic components of language, and RH involvement in mediating discourse, semantic, and prosodic aspects of language. The question of interest has been whether bilingualism engages the RH to a greater extent than is the case in monolinguals, or whether certain subgroups of bilinguals show a differential lateralization pattern as compared to other bilingual subgroups (for example, early versus late bilinguals). Taking into consideration the findings from behavioural studies of language-processing strategies characterizing early versus late bilinguals, and fluent versus less fluent language learners, Vaid and Genesee (1980) developed a model of language lateralization in bilinguals. According to this model, hemispheric involvement in bilinguals was hypothesized to resemble that of monolinguals to the extent that the two languages of bilinguals were acquired simultaneously and/or in similar contexts. Right hemisphere involvement in either language of bilinguals was expected to be more likely the later the second language (L2) was acquired, the more informal the exposure to the L2, and the earlier the stage of L2 acquisition.
124 Jyotsna Vaid To evaluate evidence bearing on the hypotheses arising from this view, Vaid and Hall (1991) undertook a meta-analysis of 59 published and unpublished laterality studies of bilinguals of which 11 studies directly compared bilinguals with monolinguals. Although no clear difference between bilinguals and monolinguals emerged (perhaps in part because monolinguals were sometimes compared on the first language of bilinguals and sometimes on the second language of bilinguals), age of second language acquisition emerged as a significant variable among the bilinguals. Somewhat unexpectedly, early bilinguals were found to be significantly less left lateralized than late bilinguals (Vaid and Hall, 1991). A subsequent meta-analysis of 28 laterality studies that directly compared early and late bilinguals with monolingual users of the bilinguals’ first language (L1) corroborated this finding, and further noted that early bilinguals were also less lateralized than monolinguals (Hull and Vaid, 2006). Finally, a more comprehensive meta-analysis, involving nearly 70 studies of bilinguals tested on both languages, confirmed the early–late difference (Hull and Vaid, 2007). In all cases, no difference in lateralization was found between the L1 versus L2 of either group, arguing against differential neural representation of L1 and L2 (for a similar argument based on clinical evidence, see the convergence hypothesis discussed by Green, 2003). Since laterality approaches, by definition, only allow investigations of inter-hemispheric differences, they leave open the question of whether there may be differential intrahemispheric organization of language associated with bilinguality. To address this question, other approaches, such as neural imaging, are better suited.
Neurobehavioural Studies Imaging studies have been widely hailed as providing a direct window into the workings of the intact brain. In actuality, they are not necessarily a direct measure of neural activity. The number of brain imaging studies conducted in the past 15 years has been impressive; there has also been an increase in the technological and statistical sophistication and precision of localizing brain activity. The different imaging techniques vary in their extent of spatial versus temporal resolution and in how direct or indirect a measure of neural activity they provide. However, all the techniques share an assumption that to understand cognitive processes, it is important to understand where and when they occur in the brain and further, that cognitive processes involved in language are in some sense localizable to discrete brain regions, rather than being distributed throughout the brain.
Neuroimaging Techniques Techniques that permit examination of neural activity in vivo to examine language functions have made use of (i) electrophysiological measures, such as event-related brain
Neural Substrates of Language Processing in Bilinguals
125
potentials, (ii) magnetocephalographic measures, such as magnetic source imaging, (iii) transcranial magnetic stimulation and (iv) haemodynamic measures, such as PET, fMRI, near infrared spectrometry (NIRS) and functional transcranial Doppler sonography (fTCD). In general, language imaging research with monolinguals has found that the set of regions activated in functional imaging studies of language is larger and more variable than that observed with lesion-deficit data. Electroencephalographic research is based on recordings of electrical activity of neurons measured at the surface of the scalp. Event-related brain wave components elicited by specific sensory, motor or cognitive events are measured, and differences in the amplitude and latency of these components (such as the N400, which indexes an unexpected semantic event) allow researchers to infer the degree and timing of electrical activity in the brain during a specific cognitive activity. ERP measures provide a direct measure of neuronal electrical activity and have a high degree of temporal resolution; as such, they provide a real-time measure of stages of cognitive processing. However, their spatial precision is fairly limited due to distortion of electrical fields produced by brain tissue and the skull. Furthermore, it remains to be established if certain brain wave components that have been the subject of extensive study in the language domain uniquely specify linguistic events rather than cognitive events more generally. Nevertheless, there has been a veritable explosion of interest in ERP studies of language processing in second language users, with over three dozen ERP studies conducted to date, most of them in the past decade. One focus of interest in bilingual ERP research has been to compare grammatical versus semantic processing in the first versus the second (less proficient) language of late learners to test notions of a sensitive period for the acquisition of grammar. Unfortunately, the use in many of these studies of semantic and syntactic anomalies as the method of choice has made the interpretation somewhat problematic, as one does not know whether the way the brain responds to anomalous constructions may be extrapolated to make claims about normal semantic or syntactic processing. More research is needed with a range of linguistic tasks before conclusions can be drawn about the issue of a critical period in the acquisition of a second grammar. One aspect of this research that is of interest in the light of the laterality findings is that among certain subgroups of individuals tested in the ERP studies (namely, early learners of ASL, including deaf and hearing individuals), there is evidence of bilateral hemispheric involvement in language, consistent with the laterality findings (see Hull and Vaid, 2005, for further discussion of this point). Unfortunately, ERPs have limited spatial resolution, so it is not possible to consider issues of differential intra-hemispheric organization related to bilingualism using this technique by itself. A second source of evidence involves MEG. Like ERPs, MEG also affords a direct measure of cortical neural activity. MEG is based on recordings of magnetic flux produced by intracellular currents in populations of neurons time-locked to the presentation of an external stimulus. However, since magnetic fields do not undergo distortion by intervening brain tissue, MEG offers a high level of accuracy in measurement. With a millisecond-level temporal resolution and a spatial resolution comparable to that of fMRI, MEG is a useful tool for measuring real-time changes in cognitive processes.
126 Jyotsna Vaid To date, only a few studies with bilinguals have been conducted using MEG (for example, Kubota et al., 2003; Lipski and Mathiak, 2007). Magnetic source imaging activation profiles based on a simple word recognition protocol used with monolinguals have recently been extended to examine late bilingual speakers of Spanish and English. Simos et al. (2001) found an overlap of regions across the bilinguals’ languages in the left superior temporal gyrus, but language-specific activation in other regions. Another study using MEG with Finnish monolinguals and Swedish–Finnish late bilinguals in a passive listening task with speech and non-speech sounds (vowels and tones) reported greater inter-hemispheric differences in bilinguals and a different signal morphology in the RH among bilinguals than monolinguals (Vihla et al., 2002). Transcranial magnetic stimulation (TMS) is a technique in which a coil is placed near the participant’s head and a brief magnetic pulse of current runs through it, which temporarily inhibits processing of the affected brain area. A common variant of this technique, known as repetitive transcranial magnetic stimulation, involves administering several magnetic pulses in a short period of time. The effect of this stimulation is to induce a temporary lesion in a targeted brain area. If there is impairment of performance on a task, one can infer that the brain area affected by the TMS is necessary for task performance. Thus, TMS can be a powerful tool in showing that a brain region is necessary for the performance of a task rather than being simply involved in some way. However, there are some safety concerns with TMS as it can induce seizures in some individuals. Moreover, it has limited spatial resolution and cannot be reliably used to access brain areas where there is overlying muscle. To date, only one study with bilinguals has used TMS (Holtzheimer et al., 2005); it found an unexpected language switching in two bilinguals given repeated TMS on the left dorsolateral prefrontal cortex as treatment for depression. Haemodynamic imaging techniques, such as PET, fMRI and NIRS, provide an indirect measure of neuronal activity as they monitor changes in blood flow elevation levels that accompany changes in neural activity. Changes in neuronal activity produce changes in local cerebral blood flow which are measured directly by PET and indirectly by fMRI or NIRS. PET uses radioactive tracers to image blood flow or metabolism; areas that have a higher blood flow will have a larger amount of tracer and will thus emit a stronger signal. fMRI and NIRS measure changes in the relative amounts of oxygenated versus deoxygenated haemoglobin present in neural regions that are metabolically active during a task. Since they do not require the use of radionuclides, participants in an fMRI or NIRS study can undergo many more trials than would be feasible in a PET study. fMRI offers better temporal and spatial resolution than PET. Functional transcranial Doppler sonography measures cerebral perfusion changes related to neural activation, and permits comparison of averaged, event-related blood flow velocity changes within two cerebral arteries (such as the left versus right middle cerebral artery), thereby allowing the evaluation of lateralization differences. As such, it provides a non-invasive alternative to the sodium amytal procedure, or the so-called Wada test, for determining language lateralization in
Neural Substrates of Language Processing in Bilinguals
127
patients prior to brain surgery, and can be used with normal individuals to determine atypical patterns of lateralization.
Further Considerations with PET and fMRI Research The typical paradigm in haemodynamic studies involves participants performing a cognitive task for some period of time and contrasting neural activity during that task period with activity during a baseline task. It is assumed that brain activity will change between the two tasks and will reflect task demands. Brain areas that show increases in activation are identified by noting which specific brain regions change their signal intensity as the task state changes from the baseline condition to the target task. However, interpretation of activation data using haemodynamic methods is an uncertain endeavour, because the sites activated usually overestimate the actual number of sites that are integral to a particular cognitive function. Several regions may show elevated activation reflecting task-irrelevant, ancillary mental activity. In addition, because of the subtractive logic of most designs using fMRI or PET procedures, patterns of activation are interpreted in terms of changes in signal intensity between the two task states. As such, an activation pattern associated with a particular cognitive function more strictly reflects a change in activation related to a particular baseline activity. The same function could, in principle, elicit a different activation if paired with a different baseline task. For these reasons, assignment of regions to linguistic functions is usually tentative, and converging evidence across different studies is needed before one can have confidence in assigning functions to specific regions. Another problem that mars the interpretation of neuroimaging studies is the use of spatial smoothing or averaging techniques. Although a standard practice in functional imaging studies, this has the effect of removing individual differences, which can be particularly problematic for studies in which group differences are of theoretical interest. To date, there have been hardly any NIRS studies with bilinguals (but see Chen et al., 2008). The majority of bilingual imaging studies have been conducted using fMRI and PET measures. There are several dozen studies of these two types conducted with bilingual participants (Vaid and Hull, 2002) including over 35 fMRI studies alone (Vaid, 2008). In fMRI research with bilinguals, the central question of interest is whether the neural regions activated when participants perform a task in their second language are spatially distinct from those activated during first language performance. A related question concerns whether the degree of overlap and the amount of activation are influenced by such factors as L2 proficiency level or age of onset of bilingualism. The outcomes of the bilingual imaging studies have varied. Some studies report overlapping activation across the two languages of bilinguals (Frenck-Mestre et al., 2005; Illes et al., 1999), whereas others show evidence of distinct regions activated across the two
128 Jyotsna Vaid languages (Dehaene et al., 1997; Kim et al., 1997; Perani et al., 1996; Tham et al., 2005). Far more studies show overlapping activation profiles than distinct language activation. Consideration of proficiency or age of onset of bilingualism variables has not clarified the picture as some studies have found a difference in activation of the L1 versus L2 in late bilinguals only (Kim et al., 1997), whereas other studies report an identical pattern of activation for the two languages in early and late bilinguals alike (Frenck-Mestre et al., 2005). Differences in activation found across languages need not be interpreted as differential underlying neural representation of the languages, but may reflect differences in cognitive load when performing a task in a less versus more proficient language (for example, Hasegawa et al., 2002). Alternatively, differences in activation across languages could reflect an artefact of the usual variability observed across imaging runs even in a single language when subjects are tested in more than one session (for example, Mahendra et al., 2003). A further problem is that when overlapping activation across languages is observed, it need not reflect overlapping neural representation, but simply the fact that the task was underspecified. Many of the tasks used in haemodynamic studies with bilinguals have employed covert word generation, typically in order to minimize movement artefacts produced by articulation. However, use of covert generation tasks in the absence of behavioural measures validating how the task was actually performed makes it impossible to know the extent to which participants were actually carrying out the task as per instructions. This issue clouds the interpretation of the much cited Kim et al. (1977) study. In this study, bilingual participants were to think silently about what they had eaten the day before, and to use one or their other language as instructed by the experimenter. There were no behavioural or other independent measures to indicate that participants were, in fact, doing the task as instructed. Furthermore, it appears that covert and overt versions of a task do not produce the same pattern of activation, casting further doubt on the validity of covert tasks.
TOWARDS CONVERGENCE The typical language user in psycholinguistic or neurocognitive research on language conducted in the west continues to be conceptualized as monolingual. Taking monolinguals as the norm has meant that research on bilingualism has been shaped by an implicit monolingual perspective (Grosjean, 1989). This perspective has had distinct repercussions on the kinds of questions posed by researchers of language and the brain. For example, neuroimaging studies of the bilingual brain have focussed almost exclusively on the question of whether the brain regions or circuitry associated with functioning in one language are spatially segregated from those associated with functioning in the other
Neural Substrates of Language Processing in Bilinguals
129
language, consistent with an underlying view that regards bilinguals’ two languages as being separate, intact entities rather than operating as an interacting, integrated system. It is, therefore, interesting that one point of convergence between electrophysiological, laterality and haemodynamic methods has been a resounding lack of evidence for differential representation of the two languages of bilinguals. Instead, where differences appear, they appear to characterize the performance of both languages similarly. The cortical stimulation data and lesion data, on the other hand, do point to some variability in language representation. However, the pattern varies from individual to individual, making generalizations about specific regions untenable. The general conclusion that tentatively emerges is that the languages of bilinguals operate as a unified system and are not spatially segregated or selectively disturbed. Points of difference between the behavioural laterality and imaging methods have been that the latter have tended not to systematically address inter-hemispheric activation (Chee et al., 1999; Dehaene et al., 1997; Illes et al., 1999), and hardly any imaging studies have included monolingual comparison groups or have compared across different bilingual subgroups (Frenck-Mestre et al., 2005; Mahendra et al., 2003; Wartenburger et al., 2003). As noted at the outset, a core question motivating both clinical and experimental research on the bilingual brain has been whether brain functioning or structure is altered by the experience of acquiring and using more than one language. This question in turn has generated two further questions: (1) Does knowing two or more languages make sufficiently different linguistic and cognitive processing demands relative to knowing a single language such that a greater number or different neural structures are enlisted to support bilingual (or multilingual) language functioning than are needed to support unilingual functioning? and (2) Given the variation within multiple language users in the particular languages known, and in the contexts in which each language is acquired, might one expect differences in neural circuitry within bilingual populations related to any of these aspects of language experience? With regard to the first question, the evidence to date suggests that early onset of bilingualism appears to deviate from the typical pattern of language lateralization noted in research on single language users in that greater use is made of both cerebral hemispheres for language functioning. With regard to differences within bilingual populations, the evidence is less clear, with some studies supporting a difference between early and late bilinguals, and others a difference between proficient and non-proficient bilinguals. However, most studies to date have confounded age and proficiency effects and have not systematically addressed inter-individual differences, focussing instead on withinindividual cross-language comparisons. As such, it is too early to arrive at definitive answers. Functional imaging methods using cerebral blood flow changes offer a number of advantages over traditional tools in neuropsychology. They are non-invasive, they afford
130 Jyotsna Vaid a greater spatial scope and they can be repeated to establish reliability. Even so, they are not without interpretive problems, the most serious of which is that they at best point to brain regions that are involved in a particular function, but do not allow inferences about the areas that are necessary for that function. In future studies, progress will be achieved with research designs that use a combination of methods, for example, NIRS coupled with fCTD, or fMRI coupled with TMS, or divided visual field methods coupled with ERPs.
FURTHER DIRECTIONS Future work in the bilingual brain should acknowledge that regarding the languages of bilinguals as separable, autonomous entities may not be an accurate reflection of how language is actually used by many bilinguals. Bilinguals tend to use only one language when addressing monolinguals, but when interacting with other bilinguals code-switching appears to be the norm. Yet, to date, there has been very little research on the neural correlates of code switching (Franceschini et al., 2004; Moreno et al., 2002). A foregrounding of code-switching behaviour by bilinguals would also direct more attention to the dynamic and interactive aspects of bilingual language use. More generally, further research on bilingualism should examine a range of linguistic tasks and a range of individual difference variables to acknowledge and exploit the variety in the types of language combinations in use, and the contexts in which languages are acquired and used. A second area in which more research may prove fruitful is to consider the neural correlates of the bilingual cognitive advantage. If bilingualism, in fact, confers greater cognitive flexibility, metalinguistic awareness and ability to inhibit irrelevant information, perhaps one should be looking at frontal lobe function in relation to bilingualism (Bialystok, 2004; Bialystok et al., 2005). Furthermore, one should not restrict the scope of research to linguistic tasks, but consider how bilingualism may influence processing in non-linguistic tasks, given that the cognitive repercussions of bilingualism are thought to extend beyond the linguistic domain (see Hausmann et al., 2004, for evidence for reduced laterality in bilinguals on a non-verbal task). Some imaging studies have begun to demonstrate alterations in neural functioning associated with the learning of new sounds (Golestani and Zatorre, 2003), or the attainment of proficiency in a second language (Chee et al., 2004). If acquiring a second language indeed alters brain functioning, it would be plausible to expect differences between bilinguals and monolinguals at the level of brain structure as well. Indeed, two recent studies suggest that there may well be structural differences associated with the acquisition of another language. Coggins et al. (2004) found that the anterior midbody to total corpus callosum midsagittal area ratio was significantly larger in a group of late bilinguals compared to monolingual controls. They interpreted this result in terms of an adaptive response to bilingual capacity. Somewhat surprisingly, however, no differences
Neural Substrates of Language Processing in Bilinguals
131
between bilinguals and monolinguals were found in corpus callosum regions that subserve frontal lobe or perisylvian regions. More recently, a voxel-based morphometry study using whole brain magnetic resonance imaging (MRI) to examine structural plasticity in the bilingual brain compared Englishspeaking monolinguals with early bilingual and late bilingual speakers of English and a second European language (Mechelli et al., 2004). The study found a significantly greater grey matter density in the left inferior parietal cortex in bilinguals as compared to monolinguals. Furthermore, in a sample of Italian speakers who had learnt English as a second language at varying ages, with age of L2 acquisition correlating negatively with overall L2 proficiency, Mechelli et al. found that the density of grey matter in the left inferior parietal cortex increased with second language proficiency and decreased with age of acquisition. Whether the increase in grey matter density observed in bilinguals is due to changes in neuronal size, or dendritic or axonal arborization could not be determined by the method used. Nevertheless, the results are clearly provocative in their implication that ‘the structure of the human brain is altered by the experience of acquiring a second language’ (Mechelli et al., 2004: 757). What is particularly noteworthy is that the difference in brain density was most pronounced in early bilinguals, who showed greater grey matter density in both the left and the RH, a result that converges nicely with the findings from the laterality literature showing greater bilateral hemispheric involvement in language processing in early bilinguals as compared to monolinguals. Further insights into the neural substrates of bilingualism will require the use of more inclusive, multilevel research designs in which converging evidence is sought using concurrent methods. In addition to improvements in methodology, though, a more nuanced conceptualization of bilingualism is needed to advance our understanding of language and the brain in bilinguals. In this newer conceptualization, the interactive and dynamic aspects of language processing by bilinguals should be foregrounded rather than persisting with a view that implicitly regards the two languages of bilinguals as being static, autonomous and intact entities.
ACKNOWLEDGEMENTS This chapter is based on a keynote address delivered at the International Conference on Cognitive Science, Allahabad, December 2004. Preparation of this manuscript was supported in part by an International Research Travel Award to the author by the College of Liberal Arts, Texas A&M University.
REFERENCES Albert, M. and L.K. Obler. 1978. The Bilingual Brain. New York: Academic Press. Bialystok, E. 2004. ‘The Impact of Bilingualism on Language and Literacy Development’. In T.K. Bhatia and W.C. Ritchie (eds), Handbook of Bilingualism (pp. 577–602). Malden. Massachusetts: Blackwell.
132 Jyotsna Vaid Bialystok, E., F. Craik, C. Grady, W. Chau, R. Ishii, A. Gunji and C. Pantev. 2005. ‘Effect of Bilingualism on Cognitive Control in the Simon Task: Evidence from MEG’. NeuroImage, 24: 40–49. Chee, M., E. Tan and T. Thiel. 1999. ‘Mandarin and English Single Word Processing Studied with Functional Magnetic Resonance Imaging’. The Journal of Neuroscience, 19: 3050–56. Chee, M., C.S. Soon, H.C. Lee and C. Pallier. 2004. ‘Left Insula Activation: a Marker for Language Attainment in Bilinguals’. Proceedings of the National Academy of Sciences, 101: 15265–70. Chen, H.C., J. Vaid, , H. Bortfeld, and D. Boas. 2008. Optical imaging of phonological processing in two distinct orthographies’. Experimental Brain Research, 184: 427–33. Coggins, P., T. Kennedy and T. Armstrong. 2004. ‘Bilingual Corpus Callosum Variability’. Brain and Language, 89: 69–75. Dehaene, S., E. Dupoux, J. Mehler, L. Cohen, E. Paulesu, D. Perani, P. van de Moortel, S. Lehericy and D. Le Bihan. 1997. ‘Anatomical Variability in the Cortical Representation of First and Second Language’. Neuroreport, 8: 3809–15. Elbert, T., C. Pantev, C. Wienbruch, B. Rockstroh and E. Taub. 1995. ‘Increased Use of the Left Hand in String Players Associated with Increased Cortical Representation of the Fingers’. Science, 270: 305–07. Fabbro, F. 2001. ‘The Bilingual Brain: Bilingual Aphasia’. Brain and Language, 79: 201–10. Franceschini, R., C. Krick, S. Behrent and W. Reith. 2004. ‘The Neurobiology of Code-Switching’. In Multilingual Communication (pp. 179–93). Amsterdam: John Benjamins. Frenck-Mestre, C., J.L. Anton, M. Roth, J. Vaid and F. Viallet. 2005. ‘Articulation in Early and Late Bilinguals’ Two Languages: Evidence from Functional Magnetic Resonance Imaging’. NeuroReport, 16: 761–65. Gaser, C. and G. Schlaug. 2003. ‘Brain Structures Differ Between Musicians and Non-Musicians’. Journal of Neuroscience, 23: 9240–45. Golestani, N. and R. Zatorre. 2003. ‘Learning New Sounds of Speech: Reallocation of Neural Substrates’. NeuroImage, 21: 494–506. Green, D. 1998. ‘Mental Control of the Bilingual Lexico-Semantic System’. Bilingualism: Language and Cognition, 1: 67–81. ———. 2003. ‘Neural Basis of Lexicon and Grammar in L2 Acquisition: The Convergence Hypothesis’. In R. van Hout, A. Hulk, F. Kuiken and R. Towell (eds), The Interface Between Syntax and the Lexicon in Second Language Acquisition (pp. 197–208). Amsterdam: John Benjamins. Grosjean, F. 1989. ‘Neurolinguists Beware! The Bilingual is not Two Monolinguals in One Person’. Brain and Language, 36: 3–15. Hasegawa, M, P.A. Carpenter, M.A. Just. 2002. ‘An fMRI Study of Bilingual Sentence Comprehension and Workload’. NeuroImage, 15: 647–60. Hausmann, M., G. Durmusoglu, Y. Yazgan and O. Güntürkün. 2004. ‘Evidence for Reduced Hemispheric Asymmetries in Non-Verbal Functions in Bilinguals’. Journal of Neurolinguistics, 17: 285–99. Holtzheimer, P., W. Fawaz, C. Wilson and D. Avery. 2005. ‘Repetitive Transcranial Magnetic Stimulation may Induce Language Switching in Bilingual Patients’. Brain and Language, 94(3): 274–77. Hull, R. and J. Vaid. 2005. ‘Clearing the Cobwebs from the Study of the Bilingual Brain: Converging Evidence from Laterality and ERP Studies’. In J. Kroll and A.M. de Groot (eds), Handbook of Bilingualism: Psycholinguistic Approaches (pp. 480–96). Oxford: Oxford University Press. ———. 2006. ‘Laterality and Language Experience’. Laterality, 11(5): 436–64. ———. 2007. ‘Language Lateralization in Bilinguals: A Meta-Analytic Tale of Two Hemispheres’. Neuropsychologia, 45(9): 1987–2008.
Neural Substrates of Language Processing in Bilinguals
133
Ijalba, E., L.K. Obler and S. Chengappa. 2004. ‘Bilingual Aphasia’. In T.K. Bhatia and W.C. Ritchie (eds), The Handbook of Bilingualism (pp. 71–89). Malden, MA: Blackwell. Illes, J., W. Francis, J. Desmond, J. Gabrieli, G. Glover, R. Poldrack, C. Lee and A. Wagner. 1999. ‘Convergent Cortical Representation of Semantic Processing in Bilinguals’. Brain and Language, 70(3): 347–63. Kim, K., N. Relkin, K. Lee and J. Hirsch. 1997. ‘Distinct Cortical Areas Associated with Native and Second Languages’. Nature, 388: 171–74. Kubota, M., P. Ferrari and T. Roberts. 2003. ‘Magnetoencephalography Detection of Early Syntactic Processing in Humans: Comparison Between L1 Speakers and L2 Learners of English’. Neuroscience Letters, 353: 107–10. Lipski, S.C. and K. Mathiak. 2007. ‘A Magnetoencephalographic Study on Auditory Processing of Native and Nonnative Fricative Contrasts in Polish and German Listeners. Neuroscience Letters, 415(1): 90–95. Lubrano, V., F.E. Roux and J.F. Demonet. 2004. ‘Writing-Specific Sites in Frontal Areas: A Cortical Stimulation Study’. Journal of Neurosurgery, 101: 787–98. Lucas, T., G. McKhann, G. Ojemann. 2004. ‘Functional Separation of Languages in the Bilingual Brain: A Comparison of Electrical Stimulation Language Mapping in 25 Bilingual Patients and 117 Monolingual Control Patients’. Journal of Neurosurgery, 101: 449–57. Mahendra, N., E. Plante, J. Magloire, L. Milman, T.P. Trouard. 2003. ‘fMRI Variability and the Localization of Languages in the Bilingual Brain’. NeuroReport, 14: 1225–28. Mechelli, A., J. Crinion, U. Noppeney, J. O’Doherty, J. Asburner, R. Frackowiak and C. Price. 2004. ‘Structural Plasticity in the Bilingual Brain’. Nature, 431: 757. Moreno, E., K. Federmeier and M. Kutas. 2002. ‘Switching Languages, Switching Palabras (Words): An Electrophysiological Study of Code Switching’. Brain and Language, 80: 188–207. Munte, T., F. Altenmuller and I. Jancke. 2002. ‘The Musician’s Brain as a Model of Neuroplasticity’. Nature Review Neuroscience, 3: 473–78. Neville, H. and D. Bavelier. 2002. ‘Human Brain Plasticity: Evidence from Sensory Deprivation and Altered Language Experience’. In M. Hofman, G. Boer, A. Holtmaat, E. Van Someren, J. Verhaagen and D. Swaab (eds), Progress in Brain Research (pp. 177–88). Amsterdam: Elsevier Science. Ojemann, G. and H.A. Whitaker. 1978. ‘The Bilingual Brain’. Archives of Neurology, 35: 409–12. Paradis, M. 1983. Readings on Aphasia in Bilinguals and Polyglots. Montreal: Didier. ———. 1987. The Assessment of Bilingual Aphasia. Hillsdale, NJ: Erlbaum. ———. 2001. ‘Bilingual and Polyglot Aphasia’. In R. Berndt (ed.), Handbook of Neuropsychology, Vol. 3 (pp. 69–91). Amsterdam: Elsevier. ———. 2004. An Integrated Neurolinguistic Theory of Bilingualism. Amsterdam: John Benjamins. Pataraia, E., P.G. Simos, E.M. Castillo, R.L. Billingsley-Marshall, A.L. McGregor, J. Breier, S. Sarkari and A.C. Papanicolaou. 2004. ‘Reorganization of Language-Specific Cortex in Patients with Lesions or Mesial Temporal Epilepsy’. Neurology, 63: 1825–32. Perani, D., S. Dehaene, F. Grassi, S. Cohen, S. Cappa, Dupoux, F. Fazio and J. Mehler. 1996. ‘Brain Processing of Native and Foreign Languages’. Neuroreport, 7: 2439–44. Rorden, C. and H. Karnath. 2004. ‘Using Human Brain Lesions to Infer Function: A Relic from a Past Era in the fMRI Age?’ Nature Reviews Neuroscience, 5: 813–19. Roux, F.E. and M. Tremoulet. 2002. ‘Organization of Language Areas in Bilingual Patients: A Cortical Stimulation Study’. Journal of Neurosurgery, 97: 857–64. Roux, F.E., V. Lubrano, V. Lauwers-Cances, M. Tremoulet, C. Mascott and J.F. Demonet. 2004. ‘IntraOperative Mapping of Cortical Areas Involved in Reading in Mono- and Bilingual Patients’. Brain, 127: 1796–810.
134 Jyotsna Vaid Simos, P., E. Castillo, J. Fletcher, D. Francis, F. Maestud, J. Breier, W. Maggio and A. Papanicolaou. 2001. ‘Mapping of Receptive Language Cortex in Bilingual Volunteers by Using Magnetic Source Imaging’. Journal of Neurosurgery, 95(1): 76–81. Tham, W.P., S.J. Rickard Liow, J. Rajapakse, C.L.Tan, S.E.S. Ng, W.E.H. Lim and L.G. Ho. 2005. Phonological Processing in Chinese-English Bilingual Biscriptals: An fMRI Study. NeuroImage, 28: 579–87. Vaid, J. 2002. ‘Bilingualism’. In V.S. Ramachandran (ed.), Encyclopedia of the Human Brain, Vol. 1 (pp. 417–32). San Diego: Elsevier. ———. 2008. ‘The Bilingual Brain: What’s Right? What’s Left?’ In J. Altarriba and R. Heredia (eds), An Introduction to Bilingualism: Principles and Processes (pp.129–44). New York: Lawrence Erlbaum Associates. Vaid, J. and F. Genesee, 1980. ‘Neuropsychological Approaches to Bilingualism: A Critical Review’. Canadian Journal of Psychology, 34: 417–45. Vaid, J. and D.G. Hall. 1991. ‘Neuropsychological Perspectives on Bilingualism: Right, Left, and Center’. In A. Reynolds (ed.), Bilingualism, Multiculturalism and Second Language Learning: The McGill Conference in Honour of Wallace E. Lambert (pp. 81–112). Hillsdale, NJ: Lawrence Erlbaum Associates. Vaid, J. and R. Hull. 2002. ‘Re-Envisioning the Bilingual Brain Using Functional Neuroimaging: Methodological and Interpretive Issues’. In F. Fabbro (ed.), Advances in the Neurolinguistics of Bilingualism (pp. 315–55). Udine Forum: Udine University Press. Vihla, M., K. Kiviniemi and R. Salmelin. 2002. ‘Auditory Cortical Activation in Finnish and Swedish Speaking Finns: A Magnetoencephalographic Study’. Neuroscience Letters, 322: 141–44. Wartenburger, I., H.R. Heekeren, J. Abutalebi, S.F. Cappa, A. Villringer, D. Perani. 2003. ‘Early Setting of Grammatical Processing in the Bilingual Brain’. Neuron, 37: 159–70.
Chapter 11 Side Bias in Human Behaviour Manas K. Mandal, Hari S. Asthana and Ramakrishna Biswal
INTRODUCTION
C
ephalo-caudally, the human body is nearly symmetrical and may be halved equally (barring some visceral and other organs) including the cerebral cortex. Such anatomical symmetry, however, is not reflected in human behaviour involving cognition, emotion and motor action. Our daily experience suggests that human behaviour is rather asymmetrical. For example, for most unimanual activities, we more often use the right hand. Likewise our visual, auditory and facial behaviours, as well as simple motor activities (like cradling) are asymmetrical. In this chapter, we intend to examine the functional asymmetry in human behaviour, which we refer to as ‘side bias’. Functional asymmetry refers to the fact that certain higher functions are differentially represented in the two sides of the brain. The left hemisphere (LH) specializes in processing information that are analytic, linear and successive, while the right hemisphere (RH) specializes in functions that are synthetic, configurational, parallel and holistic in nature (Borod, 1992; Bryden, 1982). Since the relationship of functional asymmetry (that is, largely governed by cerebral hemispheres) and side bias (that may not always be directly influenced by cerebral hemispheres) has not been fully substantiated, the term ‘side bias’ will be used throughout the text to explain asymmetry of human behaviour. By side bias, we mean a predominant mode of response ranging from simple hand movements to complex behaviour patterns like facial expression, oriented to either the left or right side from the medial plane of the body. In operational terms, side bias may be viewed as the bias reflected in the motor expression of paired organs (like hand, foot, eye or ear) or non-paired organs (like face) as a function of preference/performance or attentional/intentional factors (Mandal et al., 2000).
136 Manas K. Mandal et al. FORMS OF SIDE BIAS Functional side bias may be broadly classified into two categories: motor and cognitive. While cognitive bias refers to bias in information processing, motoric side bias is reflected in paired as well as in non-paired organs. There are four forms of bias in paired organs, handedness, footedness, earedness and eyedness.
Motor Bias Handedness It is the form of side bias that is most clearly evident, and is reflected in the motor expression of hand activities. Most individuals prefer to use their right hand in performing unimanual tasks. However, it is also not uncommon for a sizeable number of individuals in any population to show the reverse pattern of hand preference who may be classified as left-handers. The incidence of left-handedness has been found to vary between 10 and 13 per cent in human population (Van Strien, 2000). Handedness is assessed by asking an individual to indicate her/his preferred hand to perform a task that may be carried out using either hand, for example, eating, writing, grasping a tennis racket, cutting with a knife and so on. Handedness is also assessed by examining the individual’s ability to perform a task skilfully with the right or the left hand. Performance and preference measures, however, have moderate relationship in terms of predicting handedness.
Footedness This form of side bias refers to the preference for one foot over another in a given task in which one foot will mobilize and the other will support the function. Defining the preference for a foot over another is a more difficult task than defining hand preference because of its stabilizing and mobilizing features. The foot that is used to manipulate an object is the preferred foot, whereas the foot that is used to support the activities of the preferred foot by lending postural and stabilizing support is the non-preferred foot (Peters, 1981). The phenomenon of footedness has been observed while performing activities such as kicking a football, mounting a bicycle or horse, and so on. Porac et al. (1980) reported a right foot bias in about 80–90 per cent of the world’s population. Mandal et al. (1992c) reported 87 per cent of the study sample as having a right foot bias, while Dittmar (2002) found 70.9 per cent people to be right-footed. Several reports and experimental observations support the general contention that humans are typically right-footed for achieving mobilization and left-footed for postural support (Gabbard and Iteya, 1996; Gentry and Gabbard, 1995).
Side Bias in Human Behaviour
137
Earedness By earedness, we mean the preference or frequency of use of one of the ears in situation where only one ear can be used, for example, hearing a telephone, orienting one ear towards the source of mild sound (Mandal and Singh, 1990). Mandal et al. (1992c) reported a clear right-ear bias (74 per cent) in the normal population. Ear dominance is poorly reflected compared to other forms of side bias (Coren et al., 1979). In comparison to hand, foot and eye biases, people with right-ear bias are less likely with increasing age in the population (Porac et al., 1980).
Eyedness Eyedness may be defined as preference for one eye in monocular viewing situation, such as looking through a microscope or telescope or shooting with a rifle (Coren and Kaplan, 1973). In a normal population, approximately 67 per cent of the individual show right-eye preference (Porac and Coren, 1973, cited in Bryden, 1982). Studies available on sighting dominance suggest that incidence of reported right-eyedness ranges from 60 to 70 per cent (Porac et al., 1980). Although eye dominance is free from societal pressure, the study by Porac and associates (Porac et al., 1980) revealed an increase in right-eyedness with increasing chronological age. Functional biases of non-paired organs are of various types, of which the important ones for scientific purposes are facial mobility/facedness and cradling bias.
Facedness Ordinarily, people are less aware of the differential involvement of the two sides of face. Lynn and Lynn (1938) used the term ‘facedness’ to characterize facial asymmetry during emotional and non-emotional expressions. By definition, facial asymmetry refers to the fact that the left and right sides of human face during rest or movement are not identical. The phenomenon of facedness is conceived in terms of the relative intensity of expression and the extent of movement on the left and right sides of face (Borod and Koff, 1990). Darwin (1872) was the first to point out that the two sides of human face are not equally expressive. Numerous attempts were made to examine the phenomenon of facial asymmetry. These studies revealed that emotions are expressed more intensely on the left than on the right side of the face (Asthana, 1992; Asthana and Mandal, 1997, 1998; Borod and Koff, 1990; Borod et al., 1988). Studies on facial expression of infants, however, documented a greater bias on the right side of the face during smiling and distress (Rothbart et al., 1989). Studies that examined asymmetry in hemifacial mobility documented more mobility in the left hemiface than in the right hemiface (Campbell, 1982; Chaurasia and Goswami, 1975). In some studies, greater upper-left hemifacial mobility than upper-right mobility was found (Moscovitch and Olds, 1982).
138 Manas K. Mandal et al. Literature on hemifacial size reported a larger right hemiface (Nelson and Horowitz, 1980). Recently, Keles et al. (1997) reported that about 96 per cent of the right-handers had larger left than right facial areas and about 68 per cent of left-handedness had larger right than left facial areas. The resting left hemiface is judged either more happy (Springer and May, 1981) or miserable (Campbell, 1978) or more emotional than the right hemiface (Mandal et al., 1992a).
Cradling Bias It refers to the bias or preference to cradle a baby in one side of the arms than the other. Evidence showed that women are twice as likely to hold a baby on the left rather than on the right arm. This preference has been found reliable in a wide range of contexts regardless of age or parental status (Bruser, 1981; Saling and Coioke, 1984; Turnbull and Lucas, 2000). The leftward cradling bias has been observed in the real mother–infant interaction (Salk, 1960), when a doll is used to represent a baby (De Chateau and Andersson, 1976) – and even when participants are asked to imagine cradling an infant (Harris et al., 2000; Nakamichi and Takeda, 1995). Research indicated a non-significant relationship of handedness and cradling behaviour, although the magnitude of the cradling bias is marginally affected by handedness. In general, left-handers show almost as leftward cradling preference as do right-handers (Turnbull and Lucas, 2000). Salk (1960) stressed the importance of the lateral position of the heart to explain the leftward cradling bias. She stressed the importance of the heart in pacifying the newborn baby. But Salk’s cardiac account as the basis of lateral cradling bias has been criticized on many grounds, and a hemispheric asymmetry hypothesis has been invoked to explain the phenomenon (Lockard, Daley, and Gunderson, 1979).
Cognitive Bias Cognitive bias is reflected in many forms and may be classified as visual, auditory and attentional biases.
Visual-Field Bias In general, visual-field studies have shown a left visual-field (a function of the RH) advantage in the perception of facial expressions. With cartoon and line drawings of facial stimuli, similar findings are documented (Strauss and Moscovitch, 1981). A left hemispace bias has been documented in the perception of emotional/chimeric faces in free-viewing condition (Asthana and Mandal, 1996; Moreno et al., 1990). However, some investigators have observed a leftvisual-field advantage in the processing of negative emotions and a rightvisual-field advantage in the processing of positive emotions (Borod, 1992; Mandal et al., 1996). In a recent study, Asthana and Mandal (2001) reported no visual-field bias/
Side Bias in Human Behaviour
139
advantage in the perception of positive emotions. When verbal material or cognitive task is presented to the left and right visual-fields, most of the right-handed subjects showed a right visual-field superiority or bias in response to accuracy or reaction due to better right-field connection to the LH.
Auditory Bias It has been observed that the left and right ears are differentially preferred in processing different kinds of verbal material. Using dichotic-listening technique, researchers have shown a left-ear superiority for the judgement of the emotional tone of speech (Bryden et al., 1982; Safer and Leventhal, 1977; Saxby and Bryden, 1984), and non-verbal vocalization and musical process (Borod, 1992). Ley and Bryden (1982) found a left-ear advantage for identifying the emotional tone of voice, and at the same time a right-ear advantage for identifying the content of the voice. A left-ear advantage was apparent when the task difficulty was minimal; it was reversed to a right-ear advantage when the difficulty was increased (Mondor and Bryden, 1992). Music, but not the poetry, was judged more soothing when it came to the left ear (Beaton, 1979).
Attentional Bias It refers to the preference or bias in evaluating and/or scanning objects from one side to the other. Gilbert and Bakan (1973) suggested a left-ward attentional bias hypothesis, which suggests that individuals have a tendency to process information more efficiently if presented to the left than to the right visual-space. Mead and McLaughlin (1992) examined the phenomenon of scanning bias by presenting paintings in original and mirror-reversed orientations. Paintings were perceived to have a high aesthetic quality when presented to the viewers’ left side. The leftward attentional bias has been reported with chimeric facial stimuli also (Mattingley et al., 1994). Nicholls and Roberts (2002) supported a leftward attentional bias in the judgement of relative magnitude between the left and right sides of a stimulus. Recent investigators suggest that the leftward bias is generated by left to right reading habits (Chokron et al., 1997). The effect of scanning habit has been investigated by comparing readers of language with different scanning directions. Sakhuja et al. (1996) compared the reading of Hindi (left to right) and Urdu (right to left) on a chimeric face recognition task. When subjects were asked to select the face that appeared to be happier, readers of Hindi language selected the face with happier expression on the left side, whereas readers of Urdu language selected the face with happier expression on the right. Eviatar (1997) has reported a similar reduction in the leftward bias for face recognition for readers of Hebrew (right to left). In the following sections, two forms of side bias involving face and hand will be discussed in detail with some research evidences and theoretical propositions.
140 Manas K. Mandal et al. BIAS IN FACIAL EXPRESSION Unlike other paired organs such as hand, foot, ear and eye, face is a non-paired organ. Often, people fail to notice differential involvement of the two sides of the face. This behavioural asymmetry, generally referred to as facedness, is conceived in terms of the relative intensity of expression and the extent of movement on the left and right sides of the face (Borod and Koff, 1990). Functionally, facedness differs from other indices of side bias (handedness, footedness, earedness and eyedness); whereas the latter indices provide important cues to understand subjective preference or proficiency in unimanual activities of sensory or motor origin, the former index provides interpersonal cues important to understand social interaction. Facial asymmetry or face bias refers to the fact that the left and right sides of human face during rest or movement are not identical (Asthana et al., 2000). The asymmetry may be produced because of anatomical, physiological, neurological psychological, pathological and socio-cultural factors (Gelder and Borod, 1990). In order to measure facial behaviour, two commonly used methods are utilized: judgemental and anatomical/electrophysiological. Between these two, the judgemental method is used more frequently to study facial asymmetry. In this method, observers’ judgement is considered as the dependent and facial behaviour as the independent measure. Based on observers’ judgements, facial behaviours are calibrated and inferences are drawn. In the judgemental method, a variety of stimuli are used, for example, symmetrical composite faces, hemiregional composite faces (see Asthana and Mandal 1997), videotapes of whole faces and hemifaces. Lynn and Lynn (1938; cited in Borod et al., 1997) conducted the first systematic study of facial asymmetry. These authors used the term ‘facedness’ to characterize facial asymmetry during emotional or non-emotional expressions. Wolff (1943) examined facial asymmetry (cited in Sackeim et al., 1978) from an emotional quality point of view, and reported that the right side of the human face offers social or public expressions, whereas the left side of the face reveals hidden and personalized feelings. In facial bias studies, normal subjects are asked to produce expressions in posed and/ or spontaneous conditions. Posed expressions are those that are expressed when the individual is instructed to do so, whereas spontaneous expressions are as a result of an instinctual reaction to an appropriate situation (Myers, 1976). Experiments with posed and spontaneous emotion expressions were carried out separately because of the evidence that suggested differential anatomical involvement during these conditions. Cortical structures were found to be more involved for posed expressions, while for spontaneous expression it was subcortical structures (DeJong, 1979; Miehlke, 1973). Asymmetry in facial expressions has been studied using still photographs and videotapes. Several researchers have tried to examine the hemifacial asymmetry during posed expression of emotion (Campbell, 1978; Karch and Grant, 1978; Sackeim and Gur, 1978). These studies use composite facial photographs prepared by vertically bisecting a
Side Bias in Human Behaviour
141
photograph and reassembling these with the mirror image of the same hemiface to create left–left and right–right facial composites (Sackeim and Gur, 1980). These researchers found that the left hemifacial composite was judged to have expressed emotion more intensely than the right hemifacial composite. Support for the left hemifacial bias during expression was obtained from other studies as well (Asthana and Mandal, 1997, 1998; Baribique et al., 1987; Mandal et al., 1993; Moreno et al., 1990). In the studies reported by Borod and her associates (Borod and Caron, 1980; Borod et al., 1981), encoders were videotaped while posing expressions; slow motion replay of facial expression was presented before the subjects. The left hemiface was judged as significantly more involved in facial expression than the right hemiface (Borod and Koff, 1983, 1990; Borod et al., 1988). Borod et al. (1997) reviewed 49 studies on facial expression and concluded that ‘the left hemiface is more involved than the right hemiface during facial expression of emotion’. The greater left hemifacial activity has been interpreted as a function of greater RH involvement in emotional expression because fibre connections of each hemiface, especially the lower two-third parts, are predominantly innervated by the contralateral hemisphere (Kuypers, 1958). The phenomenon of facial asymmetry has not been much studied with regard to the spontaneous expression of emotion. Borod and her associates (Borod and Koff, 1983; Borod et al. 1997) examined facial asymmetry in spontaneous conditions. The left hemiface was found to be more involved than the right hemiface during expression of negative emotions, while asymmetries were found equally often on both sides of the faces for positive emotions (Hager and Ekman, 1985; Sackeim and Gur, 1978). Asymmetry of facial action is lost during spontaneous emotional expression because subcortical structures innervate the face with bilateral fibre connections (Borod et al., 1997; DeJong, 1979). A meta-analytic study by Skinner and Mullen (1991) reported that the left hemiface expresses emotion more intensely than the right hemiface. This asymmetry was more pronounced for negative emotion than for positive emotions. Left-sided asymmetries were more frequent for negative (100 per cent) than for positive (76 per cent) emotional expressions. Right-sided asymmetries were more frequent for positive (24 per cent) than for negative (10 per cent) emotional expressions (Borod, 1993: 457) Borod (1992: 341) had two speculations in favour of this finding: 1. Negative emotions are linked to a survival mechanism, thereby requiring a gestalt, synthetic processing (a function of the RH). 2. Positive emotions are more linguistic/communicative than emotional; thus, a relative dominance of the LH is more evident. In recent times, a great deal of attention has been paid to the issues of the hemifacial bias and culture-specificity of facial expressions of emotion. Researchers have argued about the extent to which the two hemifaces differ in the expression of positive or
142 Manas K. Mandal et al. negative emotions as a function of culture. It is proposed that while some components of facial expression of emotion are governed biologically, others are culturally influenced (Mandal and Ambady, 2004). The left side of the face is more expressive of emotions, more inhibited, and displays culture-specific emotional norms. The right side of the face, on the other hand, is less susceptible to cultural display norms, and exhibits more universal emotional signals. Empirical findings on this issue showed that the Japanese had a right hemifacial bias for positive and left hemifacial bias for negative emotions; Indian and North Americans had a left hemifacial bias for all emotions (Mandal et al., 2001).
HANDEDNESS Hand bias or handedness is primarily a function of the contralateral cerebral hemispheres. Most individuals are found to be right-handed because of the physiological organization (speech centre being located in the left) of the brain (Bryden, 1982). Anatomically, the contralateral fibre connection from the LH to the right side of the body crosses at a higher level of the brain than its counterpart, resulting in greater dominance for the right as compared to the left hand. For various reasons (such as genetic, obstetrical, maturational, neuropathological, environmental), some individuals develop either a reverse or reduced cerebral dominance pattern. They are the ones who develop left-handedness, though this does not hold true for all left-handers. Handedness can be explained on a continuum varying from pure right-handers to pure left-handers, with mixed and clumsy handers in between. Although signs of right-hand preference are found early in infancy (Michel and Harkins, 1986; Thompson and Smart, 1993), the mechanism that elicits handedness is evident in the uterus (Hepper et al., 1991). It gradually develops with age, and becomes fully stable by approximately 10 years of age (Gesell and Ames, 1947).
Incidence of Handedness Approximately 10 per cent of the human population is estimated to be left-handed. According to Hecaen and Ajuriaguerra (1964), the reported figure vary from as high as 30 per cent to less than 10 per cent. In a survey of 5,000 years of artwork, Coren and Porac (1977) observed an average of 93 per cent right-handedness irrespective of historical era (for example, Pre 3000 BC: 90 per cent, 1000 BC: 94 per cent, 1 AD: 1950: 89 per cent) or geographical location (for example, Central Europe: 93 per cent, Central Asia: 92 per cent, the Middle East: 96 per cent, Africa: 90 per cent). More recent studies showed a somewhat similar picture (see Table 11.1). Western studies showed that 13.8 per cent of young adults were left-handed (Spiegler and Yeni-Komshair, 1983). Maehara et al.
Side Bias in Human Behaviour
143
(1988) found that about 5 per cent of young people in Japan (aged 14 to 15 yrs.) were left-handed. A large-scale study conducted in Western societies found the prevalence of left-handedness among young people to be over 10 per cent ranging from 13 per cent in US (Gilbert and Wysocki, 1992; Van Strien, 2000) to 11.2 per cent in England (Ellis et al., 1998). McManus (1995) stated that a true incidence of left-handedness is 7.5 per cent, whereas Iwasaki et al. (1995) found the prevalence of left-handedness to be 6 per cent for men and 2.4 per cent for women among the age group of 13–19 years. TABLE 11.1
Incidence of left handedness across countries
Geographic location
Left-handers (%)
Source
Asian and Hispanic
Asian 9.3
Gilbert and Wysocki (1992)
Hispanic 9.1 Orientals and Caucasians
Orientals 4.6
Porac et al. (1990)
Caucasians 8.7 Japan
3.1
Hatta and Nakatsuka (1976)
Taiwan
1.5
Teng et al. (1976)
Korea
1.0
Harris (1990)
North America
12.0
Gilbert and Wysocki (1992)
India
6.8
Mandal et al. (1992b)
Japan, Canada
Japan 1.4
Ida and Bryden (1996)
Canada 8.6 Nigeria
4.9
Perelle and Ehrman (1994)
Belgium
15.7
Perelle and Ehrman (1994)
UK
11.2
Ellis et al. (1998)
Japan
5
Maehara et al. (1998)
Source: Mandal and Dutta, 2002.
Coren and Halpern (1991), in one of their studies, found that the proportion of lefthanders decrease with increasing age. About 11.3 per cent individuals were found to be left-handed below 40 years of age, but it reduced to 4.7 per cent between 40 and 80 years of age, and none above 80 years of age (Porac and Coren, 1981). Coren and Halpern (1991) explained this fact with the help of two hypotheses, ‘the “modification hypothesis”— which suggests that left-handers change their hand preference in order to adapt with the “right-fit world” and the “elimination hypothesis”—that precludes an extinction of left-handers due to a variety of reasons involving genetic, prenatal, neuropathological, immunological or environmental risk factors’.
144 Manas K. Mandal et al. Measurement of Handedness Handedness is generally assessed by three methods: preference, performance and frequency of use. These three measures are relatively independent of each other. It is possible that a person prefers to use his/her left hand, but normally performs that act with the right hand. It is also possible that a person is proficient with his/her left hand, but uses the right hand more frequently. Factors that induce these variations in hand behaviour include genetic predisposition, social learning and cultural pressure or task characteristics. Different patterns of distribution have been found regarding the preference and performance measures of handedness. It has been observed that the distribution of hand preference is mostly ‘J’ shaped; most individuals exhibit strong right preference, a few strong right preference, a few strong left preference and almost none in the centre (Annett, 1970). Unlike the distribution of hand preference, the relative distribution of hand proficiency has been found near normal (Porac and Coren, 1981).
Theories of Left-Handedness Two groups of theories try to explain the development and distribution of handedness. The first theory suggests that there is a physiological (Geschwind and Galaburda, 1987) or genetic (Annett, 1985; Gangstead and Yeo, 1994) predisposition of handedness that is manifested in the preference for one hand over the other. Due to this physiological predisposition, similar patterns of handedness had been found in the families (Annett, 1973; Porac and Coren, 1976). The second theory suggests that right-handedness is a function of social pressure/conformity or environmental factor (Blau, 1946; Collins, 1975). The social pressure is considered ‘overt’ in the sense that authorities in the society teach in favour of dextrality. Environmental factors, such as the design of tools or physical structures, are made to suit the right-handers, enforcing a ‘covert’ pressure to follow a dextral norm (Coren and Halpern, 1991). These overt or covert pressures gradually strengthen the development of right-handedness across the lifespan (Porac et al., 1980). A general outline of the specific theories associated with left-handedness is given in Table 11.2.
CONCLUSIONS Side bias is a unique, characteristic feature of human behaviour. While some researchers suggest that this feature of human behaviour is biologically meaningful, others suggest that side bias akin to effective coping mechanism, functional resource allocation and purposeful social behaviour. Evidence from handedness research, for example, revealed that the left
Side Bias in Human Behaviour TABLE 11.2
145
Theoretical notions behind the incidence of left handedness
General theory
Specific theory
Main contention
The ‘sword and shield’ theory (see Kolb and Whishaw, 1989)
The shield is held in the left hand by the soldier to protect his heart thereby increasing the chance of survival; the right hand could then be effectively used to wield the sword.
The ‘mother and baby’ theory (Salk, 1973)
Holding the child in the left arm allows (a) the child to get the mother’s heartbeat and (b) the mother to keep the right hand free for doing skilful jobs.
Reinforcement (Collins, 1977)
Social punishment against the left hand use is responsible for most people being right-handed.
Enhanced maturation (see Kolb and Whishaw, 1989).
The cause of right-handedness is enhanced maturation leading to a greater development of the LH.
Developmental advantage (Morgan, 1977)
There is an inward left-ward bias or some fundamental asymmetry in the human body. The heart, speech centre are all situated on the left side of the body.
Elevated levels of testosterone (Geschwind and Galaburda, 1987)
Elevated levels of testosterone have an inhibiting effect. It is due to this that the maturation of the LH gets inhibited and as a consequence, the RH develops giving rise to lefthanders in the process.
Evolutionary theory
Corballis (1983)
Early hominids developed fine motor control for manipulation of tools which is a function of LH. Both handedness and language require skilled motor activity.
Genetic theory
Gene for right-handers (Annett, 1970)
There is a specific gene for right-handers, absence of which displays a random pattern of handedness.
Two genes, four allele model (Levy and Nagylaki, 1972)
Unlike the above, this theory states that there are two genes: one for handedness and the other for speech localization in the RH.
Polygenic model (Gangstead and Yeo, 1994)
Developmental instability that gives rise to left-handedness has a polygenetic basis.
Environmental theory
Developmental theory
Source: Mandal and Dutta, 2002.
hand is used for protection, while the right hand is used for attack (by a right-handed person); likewise the information available in the left hemiface is processed holistically and the information available in the right hemispace is processed analytically by most right-handers. Studies on facedness indicated that the right-facial behaviour reflects the prevailing social display rule (intensification, masking and so on) in a given culture, and the left-facial behaviour reflects encoder’s personalized expressions. Put together, such evidence refers to the functional relevance of side bias in human behaviour. It is, however, not clear whether this behavioural feature is governed by a
146 Manas K. Mandal et al. common biological mechanism or not (such as, cerebral hemispheres, left and right). Literature on functional laterality has evidence in favour of handedness and facedness as regulated primarily by contralateral cerebral hemispheres, but not for all forms of side bias. It would be of great importance, therefore, to examine the role of cerebral hemispheres in all forms of side biases and their inter-dependence in future research on side bias.
ACKNOWLEDGEMENTS This research is partly supported by a grant (No. SP/SO/B-15/2001) from the Department of Science and Technology, Ministry of Science and Technology, New Delhi, to Manas K. Mandal. Authors thankfully acknowledge the help extended by Dr Tanusree Dutta in completing this chapter.
REFERENCES Annett, M. 1970. ‘The Classification of Hand Preference by Association Analysis’. British Journal of Psychology, 61: 303–32. ———. 1973. ‘Handedness in Families’. Annals of Human Genetics, 37: 93–105. ———. 1985. Left, Right, Hand and Brain: The Right Shift Theory. London: Erlbaum. Asthana, H.S. 1992. ‘Functional Asymmetry in the Processing of Facial Expression of Emotion in Cortically Operated Subjects’. Unpublished Ph.D. Thesis, Banaras Hindu University, Varanasi, India. Asthana, H.S. and M.K. Mandal. 1996. ‘Mirror-Reversal of a Face is Perceived as Expressing Emotions More Intensely’. Behavioral Neurology, 9: 115–17. ———. 1997. ‘Hemiregional Variations in Face During Emotion Expression’. The British Journal of Psychology, 88: 519–25. ———. 1998. ‘Hemifacial Asymmetry in Emotion Expression. Behavior Modification, 22: 177–83. ———. 2001. ‘Visual Field Bias in the Judgment of Facial Expression of Emotion’. The Journal of General Psychology, 12: 21–29. Asthana, H.S., B. Bhushan and M.K. Mandal. 2000. ‘Side Bias in Facial Expression’. In M.K. Mandal, M.B. Bulman-Fleming and G. Tiwari (eds), Side Bias: A Neuropsychological Perspective (pp. 289–312). The Netherlands: Kluwer Academic Press. Baribique, J., R. Guerett and C. Brann. 1987. ‘Hemifacial Production of Posed Emotional Expression in Normal Men and Women’. Paper presented at the Meeting of the International Neuropsychological Society, Washington, DC. Beaton, A.A. 1979. ‘Hemispheric Emotional Asymmetry in a Dichotic-Listening Task’. Acta Psychologica, 43: 103–09. Blau, A. 1946. The Master Hand. New York: American Orthopsychiatric Association Inc. Borod, J.C. 1992. ‘Interhemispheric and Intrahemispheric Control of Emotion: A Focus on Unilateral Brain Damage’. Journal of Consulting and Clinical Psychology, 60: 339–48. ———. 1993. ‘Cerebral Mechanisms Underlying Facial, Prosodic and Lexical Emotional Expression: A Review of Neuropsychological Studies and Methodological Issues’. Neuropsychology, 7: 445–63.
Side Bias in Human Behaviour
147
Borod, J.C. and H.S. Caron. 1980. ‘Facedness and Emotion Related to Lateral Dominance, Sex and Expression Type’. Neuropsychologia, 18: 237–42. Borod, J.C. and E. Koff. 1983. ‘Hemiface Mobility and Facial Expression Asymmetry’. Cortex, 19: 355– 61. ———. 1990. ‘Lateralization for Facial Emotional Behavior. A Methodological Perspective’. International Journal of Psychology, 25: 157–77. Borod, J.C., H.S. Caron and E. Koff. 1981. ‘Asymmetry of Facial Expression Related to Handedness, Footedness and Eyedness: A Quantitative Study’. Cortex, 17: 381–90. Borod, J.C., J. Kent, E. Koff, C. Martin and M. Alpert. 1988. ‘Facial Asymmetry While Posing Positive and Negative Emotions. Support for the Right Hemisphere Hypothesis’. Neuropsychologia, 26: 759–64. Borod, J.C., C.S. Haywood and E. Koff. 1997. ‘Neuropsychological Aspect of Facial Asymmetry During Emotional Expression: A Review of the Normal Adult Literature’. Neuropsychology Review, 7: 41–60. Bruser, E. 1981. ‘Child Transport in Sri Lanka’. Current Anthropology, 22: 288–90. Bryden, M.P. 1982. Laterality: Hemispheric Specialization in Intact Brain. New York: Academic Press. Bryden, M.P., R.G. Ley and J.H. Sugarman. 1982. ‘A Left Ear Advantage for Identifying Emotional Quality of Tonal Sequences’. Neuropsychologia, 20: 83–87. Campbell, R. 1978. ‘Asymmetries in Interpreting and Expressing a Posed Facial Expression’. Cortex, 14: 327–42. ———. 1982. ‘The Lateralization of Emotion: A Critical Review’. International Journal of Psychology, 17: 211–29. Chaurasia, B.D. and H.K. Goswami. 1975. ‘Functional Asymmetry in the Face’. Acta Anatomica, 91: 154–60. Chokron, S., J.M. Bernard and M. Imbert. 1997. ‘Length Representation in Normal and Neglect Subjects with Opposite Reading Habits Studied Through a Line Extension Task’. Cortex, 33:47–64. Collins, R.L. 1975. ‘When Left-Handed Mice Live in a Right-Handed World’. Science, 181–84. ———. 1977. ‘Toward an Admissible Genetic Model for the Inheritance of the Degree and Direction of Asymmetry’. In S. Harnard, R.W. Doty, L. Goldstein, J. Jaynes and G. Krauthamer (eds), Lateralization of the Nervous System (pp. 137–50). New York: Academic Press. Corballis, M.C. 1983. Human Laterality. New York: Academic Press. Coren, S. and S. Kaplan. 1973. ‘Patterns of Ocular Dominance’. American Journal of Optometry, 50: 283–92. Coren, S. and C. Porac. 1977. ‘Fifty Centuries of Right-Handedness: The Historical Record’. Science, 198: 631–32. Coren, S. and D.F. Halpern. 1991. ‘Left-Handedness: A Marker for Decreased Survival Fitness’. Psychological Bulletin, 109: 90–106. Coren, S., C. Porac and P. Duncan. 1979. ‘A Behaviorally Validated Self-Report Inventory to Assess Four Types of Lateral Preference’. Journal of Clinical Neuropsychology 1, 55–64. Darwin, C. 1872. The Expression of the Emotions in Man and Animals. London: Murray. De Chateau, P. and Y. Andersson. 1976. ‘Left Side Preference for Holding and Carrying Newborn Infants. II: Doll-Holding and Carrying from 2 to 16 Years’. Developmental Medicine and Child Neurology, 18: 738–44. DeJong, R.N. 1979. The Neurological Examination. Hagerstown, MD: Harper & Row. Dittmar, M. 2002. ‘Functional and Postural Lateral Preference in Humans: Inter-Relations and Life Span Age Differences’. Human Biology, 74: 569–85.
148 Manas K. Mandal et al. Ellis, S.J., P.J. Ellis, E. Marshall, C. Windridge and S. Jones. 1998. ‘Is Force Dextrality an Explanation for the Fall in the Prevalence of Sinistrality with Age? A Study in Northern England’. Journal of Epidemiology and Community Health, 52: 41–44. Eviatar, Z. 1997. ‘Language Experience and Right Hemisphere Tasks: The Effects of Scanning Habits and Multilingualism’. Brain and Language, 58: 157–73. Gabbard, C. and M. Iteya. 1996. ‘Foot Laterality in Children, Adolescents and Adults’. Laterality, 1: 199–205. Gangstead, S.W. and R.A. Yeo. 1994. ‘Parental Handedness and Relative Hand Skill: A Test of the Developmental Instability Hypothesis’. Neuropsychology, 8: 572–78. Gelder, R.S. and J.C. Borod. 1990. ‘Neurobiological and Cultural Aspects of Facial Asymmetry’. Journal of Communication Disorders, 23: 273–86. Gentry, V. and C. Gabbard. 1995. ‘Foot Preference Behavior: A Developmental Perspective’. The Journal of General Psychology, 122: 37–45. Geschwind, N. and A.M. Galaburda 1987. Cerebral Laterlization: Biological, Mechanisms, Associations and Pathology. Cambridge, Massachusetts: MIT Press. Gesell, A. and L.B. Ames. 1947. ‘The Development of Handedness’. Journal of Genetic Psychology, 70: 155–75. Gilbert, C. and P. Bakan. 1973. ‘Visual Asymmetry in Perception of Faces’. Neuropsychologia, 11:355–62. Gilbert, A.N. and C.J. Wysocki. 1992. ‘Hand Preference and Age in the United States’. Neuropsychologia, 30: 601–08. Hager, J.C. and P. Ekman. 1985. ‘The Asymmetry of Facial Action in Inconsistent with Model of Hemispheric Specialization’. Psychophysiology, 22: 307–18. Harris, L.J. 1990. Cultural Influences on Handedness: Historical and Contemporary Theory and Evidence. In S. Coren (ed.), Left Handedness: Behavioral Implications and Anomalies (pp.195–258). Amsterdam, NL: North Holland. Harris, L.J., J.B. Almerigi and E.A. Kirsch. 2000. ‘Side Preference in Adults for Holding Infants: Contributions of Sex and Handedness in a Test of Imagination’. Brain and Cognition, 43: 246–52. Hatta, T. and Z. Nakatsuka. 1976. ‘Note on Hand Preference of Japanese People’. Perceptual and Motor Skills, 42:530. Hecaen, H. and J. Ajuriaguerra. 1964. Left-Handedness: Manual Superiority and Cerebral Dominance. New York: Grune & Stratton. Hepper, P.G., S. Shahidullah and R. White. 1991. ‘Handedness in the Human Fetus’. Neuropsychologia, 29: 1107–11. Ida, Y. and M.P. Bryden. 1996. ‘A Comparison of Hand Preference in Japan and Canada’. Canadian Journal of Experimental Psychology, 50: 234–39. Iwasaki, S., T. Kaiho and K. Iseki. 1995. ‘Handedness Trends Across Age Groups in a Japanese Sample of 2316’. Perceptual and Motor Skills, 80: 979–94. Karch, G.R. and C.W. Grant. 1978. ‘Asymmetries in Perception of the Sides of the Human Face’. Perceptual and Motor Skills, 47: 727–34. Keles, P., S. Diyarbakirli, M. Tan and U. Tan. 1997. ‘Facial Asymmetry in Right and Left Handed Men and Women’. International Journal of Neuroscience, 91: 147–60. Kolb, B. and I.Q. Whishaw. 1989. Fundamentals of Human Neuropsychology. New York: W.H. Freeman & Company. Kuypers, H.J.M. 1958. ‘Cortico-Vestibular Connections to the Pons and Lower Brain Stem in Man’. Brain, BI: 364–90. Levy, J. and T. Nagylaki. 1972. ‘A Model for the Genetics of Handedness’. Genetics, 72: 117–28.
Side Bias in Human Behaviour
149
Lockard, J.S., P.C. Daley and V.M. Gunderson. 1979. ‘Maternal and Paternal Differences in Infant Carry: U.S. and African Data’. The American Naturalist, 113: 235–46. Lynn, J.G. and D.R. Lynn. 1938. ‘Face Hand Laterally in Relation to Personality’. Journal of Abnormal and Social Psychology, 38: 250–76. Maehara, K., N. Negishi, A. Tsai, N. Otuki, S. Suzuki, T. Takahashi and Y. Sumiyoshi. 1988. ‘Handedness in the Japanese’. Developmental Neuropsychology, 4: 117–27. Mandal, M.K. and S.K. Singh. 1990. ‘Lateral Asymmetry in Identification and Expression of Facial Emotions’. Cognition and Emotion, 4: 61–70. Mandal, M.K. and T. Dutta. 2002. ‘Left-Handedness: Facts and Figures Across Cultures’. Psychology and Developing Societies, 13: 173–91. Mandal, M.K. and N. Ambady. 2004. ‘Laterality of Facial Expressions of Emotions: Universal and CultureSpecific Influences’. Behavioral Neurology, 15: 23–34. Mandal, M.K., G. Pandey, S.K. Singh and H.S. Asthana. 1992a. ‘Hand Preference in India’. International Journal of Psychology, 27: 433–42. Mandal, M.K., H.S. Asthana, S. Madan and R. Pandey. 1992b. ‘Hemifacial Display of Emotion in the Resting Face’. Behavioral Neurology, 5: 169–71. Mandal, M.K., G. Pandey, S.K. Singh and H.S. Asthana. 1992c. ‘Degree of asymmetry in lateral preferences: eye, foot, ear’. Journal of Psychology, 126: 155–62. Mandal, M.K., H.S. Asthana and S.C. Tandon. 1993. ‘Judgment of Facial Expression of Emotion in Unilateral Brain-Damaged Patients’. Archives of Clinical Neuropsychology, 8: 171–83. Mandal, M.K., H.S. Asthana, R. Pandey and S. Sarbadhikari. 1996. ‘Cerebral Laterality in Affect and Affective Illness: A Review’. The Journal of Psychology, 130(4): 447–59. Mandal, M.K., M.B. Bulman-Flemming and G. Tiwari. 2000. Side Bias: A Neuropsychological Perspective. North Holland: Kluwer Academic Press. Mandal, M.K., S. Harizuka, B. Bhushan and R.C. Mishra. 2001. ‘Cultural Variation in Hemifacial Asymmetry of Emotion Expressions’. British Journal of Social Psychology, 40: 385–98. Mattingley, J.B., J.L. Bradshaw, N.C. Nettleton and J.A. Bradshaw. 1994. ‘Can Task Specific Perceptual Bias be Distinguished from Unilateral Neglect?’ Neuropsychologia, 32: 805–17. McManus, I.C. 1995. ‘Familial Sinistrality: The Utility of Calculating Exact Genotype Probabilities for Individuals’. Cortex, 31: 3–24. Mead, A.M. and J.P. Mclaughlin. 1992. ‘The Roles of Handedness and Stimulus Asymmetry in Aesthetic Preference’. Brain and Cognition, 20: 300–07. Michel, G.F. and D.A. Harkins. 1986. ‘Postural and Lateral Asymmetries in the Ontogeny of Handedness During Infancy’. Developmental Psychobiology, 19: 247–58. Miehlke, A. 1973. Surgery of Facial Nerve. Philadelphia, PA: W.B. Saunder. Mondor, T.A. and M.P. Bryden. 1992. ‘On the Relation Between Auditory Spatial Attention and Auditory Perceptual Asymmetries’. Perception and Psychophysics, 52: 393–402. Moreno, C., J.C. Borod, J. Welkowitz and M. Sepert. 1990. ‘Lateralization for the Expression and Perception of Facial Emotion as a Function of Age’. Neuropsychologia, 28: 199–209. Morgan, M. 1977. ‘Embryology and Inheritance of Asymmetry’. In S. Harnard, R.W. Doty, L. Goldstein, J. Jaynes and G. Krauthamer (eds), Lateralization of the Nervous System. New York: Academic Press. Moscovitch, M. and J. Olds. 1982. ‘Asymmetries in Spontaneous Facial Expressions and their Possible Relation to Hemispheric Specialization’. Neuropsychologia, 20: 71–81. Myers, R.F. 1976. ‘Comparative Neurology of Vocalization and Speech: Proof of a Dichotomy’. Annals of the New York Academy of Science, 280: 745–57.
150 Manas K. Mandal et al. Nakamichi, M. and S. Takeda. 1995. ‘A Child Holding Thought Experiment: Students Prefer to Imagine Holding an Infant on the Left Side of the Body’. Perceptual and Motor Skills, 80: 687–90. Nelson, C.A. and F.D. Horowitz. 1980. ‘Asymmetry in Facial Expression’. Science, 209: 834. Nicholls, M.E. and G.R. Robert. 2002. ‘Can Free Viewing Perceptual Asymmetries be Explained by Scanning, Pre-Motor or Attentional Biases?’ Cortex, 38: 113–36. Perelle, I.B. and L. Ehrman. 1994. ‘An International Study of Human Handedness: The Data’. Behavior Genetics, 24: 217–27. Peters, M. 1981. ‘Attentional Asymmetries During Concurrent Bimanual Performance’. Quarterly Journal of Experimental Psychology, 33A: 95–103. Porac, C. and S. Coren. 1976. ‘The Dominant Eye’. Psychological Bulletin, 83:880–97. ———. 1981. Lateral Preferences and Human Behavior. New York: Springer-Verlag. Porac, C., S. Coren and P. Duncan. 1980. ‘Life Span Age Trends in Laterality’. Journal of Gerontology, 35: 715–21. Porac, C., L. Ress and T. Buller. 1990. ‘Switching Hands: A Place for Left Hand Use in a Right Hand World’. In S. Coren (ed.), Left-handedness: Behavioral Implications and Anomalies (pp. 259–90). The Netherlands: North-Holland/Elsevier. Rothbart, M.K., S.B. Taylor and D.M. Tucker. 1989. ‘Right Sided Facial Asymmetry in Infant Emotional Expression’. Neuropsychologia, 27: 675–87. Sackeim, H.A. and R.C. Gur. 1978. ‘Lateral Asymmetry in Intensity of Emotional Expression. Neuropsychologia, 16: 473–81. ———. 1980. ‘Asymmetry in Facial Expression’. Science, 209: 834–36. Sackeim, H.A., R.C. Gur and M. Saucy. 1978. ‘Emotions are Expressed More Intensely on the Left Side of the Face’. Science, 202: 434–36. Safer, M.A. and H. Leventhal. 1977. ‘Ear Differences in Emotional Tones of Voice and Verbal Content’. Journal of Experimental Psychology: Human Perception and Performance, 3: 75–82. Sakhuja, T., G.C. Gupta, M. Singh and J. Vaid. 1996. ‘Reading Habits Affect Asymmetries in Facial Affect Inducements: A Replication’. Brain and Cognition, 32: 162–65. Saling, M.M. and W. Coioke. 1984. ‘Cradling and Transport of Infants by South African Mothers: A CrossCultural Study’. Current Anthropology, 25: 333–35. Salk, L. 1960. ‘The Effects of the Normal Heart Beat Sound on the Behavior of the New Born Infants: Implications for Mental Health’. World Mental Health, 12: 168–75. ———. 1973. ‘The Role of Heartbeat in the Relation Between Mother and Infant’. Scientific American, 228: 24–29. Saxby, L. and M.P. Bryden. 1984. ‘Left Ear Superiority in Children Fiat Processing Auditory Emotional Material’. Developmental Psychology, 21: 72–80. Skinner, M. and B. Mullen. 1991. ‘Facial Asymmetry in Emotional Expression: A Meta-Analysis of Research’. British Journal of Social Psychology, 30: 113–24. Spiegler, B.J. and G.H. Yeni-Komshian. 1983. ‘Incidence of Left Handed Writing in a College Population with Reference to Family Patterns of Hand Preference’. Neuropsychologia, 21: 651–59. Springer, P. and P. May. 1981. ‘Attributional Asymmetries in the Perception of Moving, Static, Chimeric and Hemisected Faces’. Journal of Nonverbal Behaviour, 5: 238–52. Strauss, E. and M. Moscovitch. 1981. ‘Perception of Facial Expression’. Brain and Language, 13: 308–32. Teng, E.L., P. Lee, K. Yang and R.C. Chang. 1976. ‘Handedness in a Chinese Population: Biological, Social and Pathological Factors’. Science, 193:1148–50. Thompson, A.M. and J.L. Smart. 1993. ‘A Prospective Study of the Development of Laterally Neonatal Laterality in Relation to Perinatal Factors and Maternal Behavior’. Cortex, 29: 649–59.
Side Bias in Human Behaviour
151
Turnbull, O.H. and M.D. Lucas. 2000. ‘“Tell Me, Where is (this) Fancy Bred?”: The Cardiac and Cerebral Accounts of the Lateral Cradling Bias’. In. M.K. Mandal, M.B. Bulman-Fleming and G. Tiwari (eds), Side Bias: A Neuropsychological Perspective (pp. 267–87). The Netherlands: Kluwer Academic Press. Van Strien, J.W. 2000. ‘Genetic, Intrauterine and Cultural Origins of Human Handedness’. In M.K. Mandal, M.B. Bulman-Fleming and G. Tiwari (eds), Side Bias: A Neuropsychological Perspective (pp. 41–61). NorthHolland, NL: Kluwer Academic Press. Wolff, W. 1943. The Expression of Personality. New York: Harper.
SECTION
III
Computational Modelling
A
significant aspect of cognitive science is the development of computational algorithms to understand various cognitive processes. Computational approaches range from techniques concentrating on the neuronal level to symbolic approaches for studying cognition. Symbolic rule-based approaches have been a preferred way to model and simulate cognition. Dissatisfaction with the rule-based approaches in the 1980s lead to alternative approaches to study language, the most notable being the dynamic approaches to cognition. Dynamical approaches provide a significant alternative by not being rule-based and address performance/implementational issues in cognition as well as incorporate the temporal dimension into modelling and explaining cognition. Connectionist models which are a subset of dynamical models for cognition have become very popular with work spanning perception and language processing (Bechtel and Abrahamsen, 2001). For example, connectionist models have been proposed for learning the past tense of verbs and generation of sentences. Researchers trying to incorporate dynamics have based arguments ranging from temporal phenomena and statistical information available in input data for opting for a dynamic approach. While dynamical models show promise, further work is needed on the development of new mathematical techniques in dynamical systems and application of those techniques to understand cognition. One major area where computational models have been developed is computational neuroscience (Koch, 1999). Computational neuroscience aims to develop computational models to understand brain activity. The biophysics of a neuron plays a critical role in the firing of an action potential, and models of neuronal capability capture the dynamics of neural activity. Hodgkin and Huxley proposed their model for the production of action potential in a neuron using ionic channels (Dayan and Abbott, 2001). Neurons communicate with each other through the synapse between two neurons. The chapter by Mishra et al. focuses on non-linear dynamical analyses for analysing the propagation of action potentials in a single neuron, which is important for neuronal communication. Mishra et al. perform non-linear dynamical analysis on a point neuron model to understand the qualitative and chaotic behaviour of neurons. Various models are studied, including the Hodgkin–Huxley, FitzHugh–Nagumo, Wilson–Cowan and Cortical neuron models. They integrate the Morris–Lecar model to the multi-compartment model to study the phenomenon of spatial dependency. Most of the current neural network models for cognitive processes are based on research done on artificial neural networks, and the learning algorithms developed for those networks. The advent of the back propagation algorithm significantly advanced the parallel distributed processing approach to study cognition. Connectionist models are typically based on a network of neurons connected to each other (Rumelhart and McClelland, 1986). The architecture can be feed forward or can contain feedback or recurrent connections. A typical feed forward neural network consists of at least two layers of neurons with one or more hidden layers and one output layer. Various learning algorithms exist for
156 Advances in Cognitive Science changing weights in such a neural network. Learning algorithms differ in terms of whether weights are changed based on explicit knowledge of the expected outputs (supervised), or whether they are changed without any explicit knowledge of the expected outputs (unsupervised). Feedback connections have been added to feed forward neural networks to obtain recurrent neural networks which make use of time in an explicit fashion (Elman, 1990). Other types of connectionist models include fully recurrent networks like Hopfield networks as well as further modifications including Boltzmann machines (Rumelhart and McClelland, 1986). While a significant amount of work on neural networks has been done in the last two decades, there is still some way to go before they can be used for simulating the large-scale functions of the human brain or mind. The chapter by Krebs provides a critique on the use of neural network models so far in simulating cognitive processes. Processes including pattern recognition and language processing have been modelled and simulated using neural networks. Krebs argues that computational models based on neural networks serve as objects of experimentation, and results from these virtual experiments are tacitly included in the framework of empirical science. Some simulations of cognitive functions are even being used to make claims about human cognitive capacities. Krebs raises the question of equivalence between results obtained from experiments that are essentially performed on data structures, and results from ‘real’ experiments. Some of the basic algorithms and architectures for models of cognitive functions using artificial neural nets are discussed. Given the lack of similarities between artificial neural networks and real brains, Krebs argues that the contribution of simple artificial neural network models to cognitive science is diminished. In addition to behavioural and neuroscientific approaches, computational approaches have been used to study most of the cognitive processes. One aspect of cognition that has been explored extensively using computational approaches is language processing. Computational approaches have focussed on various grammar formalisms and generate sentences using some primitive structures and operations performed on these structures. The chapter by Joshi presents a different solution to this problem by proposing complex primitives and operations performed on these complex primitives. Joshi focuses on a framework that he and his colleagues have developed based on lexicalized tree-adjoining grammar (Joshi, 2004). Joshi describes his approach, which starts with complex primitives that directly capture some crucial linguistic properties, and discusses some general operations for composing these complex structures. This approach, pioneered by Joshi, pushes all non-local dependencies to become local, and has led to some new insights into syntactic description, semantic composition, language generation, statistical processing, and some psycholinguistic phenomena, as well as some aspects of discourse structure, all with possible relevance to the cognitive architecture of language. The chapter by Srinivasan and Pariyadath focuses on a unique aspect related to language processing, but it is also closely tied with emotions. The chapter focuses on computational models of humour perception. General theories and approaches to humour
Computational Modelling
157
are discussed, which are in general not quantifiable and cannot account very well for both behavioural and neuroscientific findings. A new computational model for identifying phonological jokes is also introduced. Research on humour perception has been carried out in several disciplines such as psychology, linguistics and artificial intelligence. Various methodologies including computational and neural approaches are being used to study humour perception in cognitive science. This chapter is an attempt at evaluating the existing theories and models for humour perception, and aims to establish a dialogue between cognitive neuroscientists and computational humour researchers. In comparison with research on humour production, theories and models for humour perception are scarce. In this chapter, Srinivasan and Pariyadath discuss and critically evaluate the existing theories and computational models for humour perception. There has been a recent upsurge in the research on cognitive neuroscience on humour perception. Unfortunately, neuropsychological findings are currently unimpressive because of the lack of testable models for humour perception. We also suggest future avenues of research on humour perception.
REFERENCES Bechtel, W. and W. Abrahamsen. 2001. Connectionism and the Mind: Parallel Processing, Dynamics and Evolution in Networks (2nd edition). USA: Blackwell Publishers. Dayan, P. and L.F. Abbott. 2001. Theoretical Neuroscience: Computational and Mathematical Modeling of Neural Systems (2nd edition), Cambridge, Massachusetts, USA: MIT Press. Elman, J.L. 1990. ‘Finding Structure in Time’. Cognitive Science, 14: 179–211. Joshi, A.K. 2004. ‘Starting with Complex Primitives Pays Off: Complicate Locally, Simplify Globally’. Cognitive Science, 28: 637–68. Koch, C. 1999. Biophysics of Computation: Information Processing of Single Neurons. Oxford, UK: Oxford University Press. Rumelhart, D.E. and McClelland, J.L. 1986. Parallel Distributed Processing: Explorations in the Microstructure of Cognition, Vol. 1: Foundations. Cambridge, Massachusetts: MIT Press.
Chapter 12 Non-linear Dynamical Analysis of Point Neuron Models and Signal Propagation along Axon Deepak Mishra, Abhishek Yadav, Sudipta Ray and Prem Kumar Kalra
INTRODUCTION
U
nderstanding dynamical behaviour is primary step in finding intricacies in the neuron models (Dyan and Abbott, 1999; Koch, 1999). It is a common practice to represent a dynamical system by its bifurcation diagram. However, because of the possibilities of the occurrence of both supercritical and subcritical Hopf bifurcations, as well as the occurrence of homo-clinic orbits of the saddle associated with the appearance/disappearance of periodic orbits, the number of possible sequences of bifurcation events is much larger. Here, we present the bifurcation analysis for various single neuron models. These models possess a richness of bifurcation events when the parameter is varied. A more careful numerical investigation uncovers that the stable and unstable periodic orbits appear and disappear via different bifurcations associated with the homo-clinic orbits of the saddle (Freeman and Schneider, 1982; Izhikevich, 1999). The bifurcation diagram determines the transition from a quiescent to oscillatory state (Strogatz, 2000; Zak, 2003). In order to explore the dynamics of single neuron models, we have carried out our analysis on the Hodgkin–Huxley (HH) model, the FitzHugh–Nagumo model (Fitzhugh, 1961; Nagumo et al., 1962), the Wilson–Cowan model, and the cortical neuron model. These neuron models can be used to study the propagation of signal through axons as well. However, we have only considered the dynamics of Morris–Lecar model (Morris and Lecar, 1981) to find out the propagation characteristics. Axons are generally very long in nature and can be of varying cross-section. So, the point neuron models are not suitable to emphasize the axon property. Instead, the long axons are
160 Deepak Mishra et al. broken up into several small compartments and then neuronal dynamics are applied to each compartment separately.
NON-LINEAR DYNAMICAL ANALYSIS OF HODGKIN–HUXLEY NEURON MODEL Hodgkin and Huxley gave their basic neuron model in 1952. This model shows that the electrical current in nerve cells is carried by the flow of ions through membrane proteins called channels (Hodgkin and Rushton, 1948). The concentration of sodium ions is high in the extracellular fluid and low in the intracellular axoplasm. The concentration gradient gives rise to a tendency for sodium ions to flow into the cell. This process gives rise to generation of action potential (Hodgkin and Huxley, 1952). Here, we show the numerical results obtained from bifurcation analysis for the HH neuron model. The synaptic current input to the neuron model is taken as the bifurcation parameter. In this section, behavioural change in the dynamics of the HH model with respect to current is investigated. We plotted time responses and phase portraits for HH model for different values of currents.
Time Response and Phase Portrait Analysis Time responses and phase-portraits for this model at various values of injected currents are drawn in Figure 12.1. It is observed here that for injected current I = 60 nA, response is converging, for I = 50 nA, limit cycle of lower magnitude exists, and for I = 30 nA, periodic behaviour is observed. We have plotted the phase portrait between voltage and gating variables m, n and h. It can be inferred from these results that the behaviour of HH model is dependent on the magnitude of injected current.
NON-LINEAR DYNAMICAL ANALYSIS OF FITZHUGH–NAGUMO NEURON MODEL The FitzHugh–Nagumo (FHN) equations are the simplest equations that have been proposed for spike generation. This model is derived by reducing the dimension of Hodgkin–Huxley neuron model (FitzHugh, 1961, 1969). In this section, behavioural change in the dynamics with respect to current is investigated by calculating the eigenvalues and plotting time responses, phase portraits and bifurcation diagram. The FHN model is represented by Equation 12.1.
Non-linear Dynamical Analysis FIGURE 12.1
161
Time response and phase portraits for HH model
dv 1 = [ v( a - v )( v - 1) - w + I ] dt m
(12.1) dw = (v - g w ) dt Here, v is the membrane potential and w is the recovery variable. The set of equations is linearized at its equilibrium point (v*, w*). We got the following Jacobean matrix J for the linearized model:
162 Deepak Mishra et al. Ê f ¢( v * ) 1ˆ - ˜ J= Á m m Á ˜ -g ¯ Ë 1 Here,
f ¢( v *) = v( a - v ) + v(1 - v ) + ( v - a)(1 - v )
Phase Plane Analysis We studied the dynamics of this model by plotting time responses and phase portraits for different values of bifurcation parameter. Figure 12.2 shows the time response and phase portrait for the FHN model at different values of injected currents. Response of model is converging for I = 0.5, periodic for I = 1.5, and again converging for I = 2.35. Parameters for FHN model used for simulation are µ = 0.5, = 0.25, a = 0.2. The nullclines for FHN model is drawn in Figure 12.3. The intersection point shows the stable points. Any perturbation from the stable points causes oscillations and therefore exhibits bifurcation.
Bifurcation Diagram In this section, the bifurcation diagram is plotted considering the injected current as the bifurcation parameter. The bifurcation diagram for the model is shown in Figure 12.4.
NON-LINEAR DYNAMICAL ANALYSIS OF WILSON–COWAN NEURON MODEL Wilson and Cowan provides a model of neural oscillators (Wilson, 1999). This model describes the coupling of an excitatory and inhibitory neuron via synapses. This model exhibits some bifurcation phenomenon, such as saddle node and Hopf bifurcation. The dynamics of model is studied by considering ry as the bifurcation parameter. Equation (12.2) represents the dynamics of Wilson–Cowan model. dx = - x + f ( rx + ax - by ) dt dy = - x + f ( ry + cx - dy ) dt
(12.2)
For fixed weights, a, b, c, d, rx and ry are the bifurcation parameters. Here, function f is sigmoid function and is given by
Non-linear Dynamical Analysis
163
FIGURE 12.2 Time responses and phase portraits for FitzHugh–Nagumo neuron model at (a) I = 0.5 (b) I = 1.5 (c) I = 2.35
164 Deepak Mishra et al. FIGURE 12.3
Nullclines for FitzHugh–Nagumo neuron model
Note: W-nullcline is shown by the solid line and V-nullcline is shown by the dashed line. FIGURE 12.4
Bifurcation diagram for FitzHugh–Nagumo neuron model
Non-linear Dynamical Analysis
f(u) =
165
1 1 + e -u
The values of rx and ry can be found by equating the equation to zero, about its equilibrium points (x, y). rx = f –1(x) – ax + by ry = f –1(y) – cx + dy The Jacobean matrix for the model at its equilibrium point is - bx(1 - x) ˆ Ê -1 + ax(1 - x) J= Á -1 - dy(1 - y )˜¯ Ë cy(1 - y )
Phase Plane Analysis We studied the dynamics of this model by plotting time responses and phase portraits for different values of bifurcation parameter. It is observed here that the model shows different behaviour at different values of bifurcation parameters. Time responses and phase-portraits for the model are drawn in Figure 12.5. The response of model is converging for ry = −2, periodic for ry = −3 and again converging for ry = −9.5. Parameters used for simulations are a =10, b =10, c =10, d = −2.
Bifurcation Diagram In this section, bifurcation diagram is plotted for the parameter ry. It is observed from Figure 12.6 that the bifurcation takes place at ry = −3.
NON-LINEAR DYNAMICAL ANALYSIS OF CORTICAL NEURON MODEL This neuron model exhibits many properties of a real cortical neuron. A cortical model can be considered a modified FitzHugh–Nagumo neuron model (Gerstner and Kistler, 2002). Here, recovery variable equation consists of quadratic terms. In this section, behavioural change in the dynamics with respect to current is investigated by plotting time responses, phase portraits and bifurcation diagram. The dynamics of cortical neuron model is given by equation set (12.3). dv 1 = ( v( a - v )( v - 1) - w + I ) dt m dw = b ( v - v1 )( v - v2 ) - g w dt
(12.3)
166 Deepak Mishra et al. FIGURE 12.5
Time response and phase portrait for Wilson–Cowan model (a) ry= –2 (b) ry= –3 (c) ry= –9.5
Non-linear Dynamical Analysis FIGURE 12.6
167
Bifurcation diagram for Wilson–Cowan model with ry as the bifurcation parameter
Phase Plane Analysis The time responses and phase portraits for different magnitudes of current are shown in Figurze 12.7. It is observed from the results that introduction of quadratic term in the recovery variable causes chaotic response at I = −3. If we further increase the current, the oscillatory behaviour is observed at I = 1.5. Model shows stable response at I = 6. Parameters for cortical neuron model used for simulation are = 0.5, = 0.5, µ = 0.01, a = 0.5 and b = 0.6.
Bifurcation Diagram In this section, we study the effect of variation in current I to the dynamics of the model. The bifurcation diagram is drawn in Figure 12.8. It is clear from this figure that the model exhibits stable, periodic and chaotic dynamics for various values of the parameter, I.
CONCEPT OF CABLE THEORY Evolution of Idea Single compartment models describe the membrane potential over an entire neuron with a single variable. Membrane potentials can vary considerably over the surface of the cell membrane, especially for the neurons with long and narrow processes. There is a delay
168 Deepak Mishra et al. FIGURE 12.7
Time response and phase portrait for cortical neuron model at (a) I = –3 (b) I = 1.5 and (c) I = 6
Non-linear Dynamical Analysis FIGURE 12.8
169
Bifurcation diagram for cortical neuron model
as well as an attenuation of action potential as it propagates from soma out to the axons. The excitatory post-synaptic potential (EPSP) initiated in the dendrite by synaptic input is also suffered by attenuation and delay as it spreads to the soma. An understanding these features is crucial for determining whether and when a given synaptic input will cause a neuron to fire an action potential. The attenuation and delay within a neuron are very significant when electrical signals travel down the long, narrow, cable like structure of dendritic or axonal branches. For this reason, the mathematical analysis of signal propagation within neurons is called cable theory (Hausser et al., 1995; Holt, 1998).
The Cable Equation The cable equation is one that describes the propagation of a signal along a cable like structure. It is given by Cm
∂Vm 1 ∂ 2 ∂Vm = (a ) - im + I (t ) ∂t 2arL ∂x ∂x
(12.4)
where a is the radius of cable, rL the intracellular resistivity and im represents ionic current. Before solving the cable equation by any method, membrane current im must be specified. Various ion channels contribute to the membrane current. Almost all dynamical
170 Deepak Mishra et al. neuron models produce non-linear expressions that are too complex to allow the analytical solution of cable equation. The membrane current can be linearly approximated as im =
Vm - Vrest Rm
where Vrest is resting potential and Rm is the specific membrane resistance. Let us define two parameters, space constant denoted by l and time constant denoted by tm. We also define v = Vm – Vrest as membrane potential relative to resting potential. aRm The expressions describing the space constants and time constants are l = and 2rL tm = RmCm With the help of these two constants, the linear cable equation can be written as tm
∂v ∂2 v = l 2 2 - v + ie Rm ∂t ∂x
(12.5)
THE MORRIS–LECAR MODEL Morris and Lecar (1981) postulated a neuron model described by a set of coupled ordinary differential equations. This model incorporates two ionic currents, one outward going non-inactivating potassium current, and the other inward going non-inactivating calcium current. Since calcium current responds much faster than potassium current, it is assumed that the former is always in equilibrium for the time scales considered. So the dynamics of calcium currents can be ignored, and the model is reduced to two dimensions. Different variants of the Morris–Lecar model are in usage, but the one considered here is taken from Rinzel and Ermentrout (1998). In the reduced form, the equations can be given by, Cm
dVm = - I ionic (Vm , w ) + I (t ) dt
(12.6)
dw w• (Vm ) - w = dt t w (Vm ) where Vm is the absolute membrane potential (mV), Cm is the membrane capacitance taken as 1 µF/cm2, w is the activation variable for potassium, all currents are in the units of µA/cm2 and time t is measured in milliseconds. The ionic current is given by, I ionic (Vm , w ) = G Ca m• (Vm )(Vm - ECa ) + G K w(Vm - EK ) + Gm (Vm - Vrest )
(12.7)
The calcium current is assumed to be in equilibrium at any time, and its activation curve is given by
Non-linear Dynamical Analysis
171
V + 1ˆ Ê m• (Vm ) = 0.5 ¥ Á 1 + tanh m ˜ Ë 15 ¯ The activation function for potassium current and the time constants follow the equations given by V ˆ Ê w• (Vm ) = 0.5 ¥ Á 1 + tanh m ˜ Ë 30 ¯ 5 t w (Vm ) = ÊV ˆ cosh Á m ˜ Ë 60 ¯ Other parameters are GCa = 1.1, GK = 2.0, Gm = 0.5,
Ca =
100, EK = -70, Vrest = -50
All conductances are in units of mS/cm2 and the reversal potentials are in mV.
ACTION POTENTIAL IN AN AXON Multi-Compartment Model As stated before, the axons are generally long and the point neuron model cannot be applied directly. Cable equation is used to model for spatial dependencies in neuron. The axon diameter is also varying and ionic currents are non-linear in nature. So, the cable equation cannot be solved analytically using the linear cable theory model. Therefore, in such situations, the long axons are broken up into small compartments, and the point neuron model is then applied to each compartment. The length and variation of diameter in a compartment should be small enough to apply the point model. The thumb rule for splitting the cable into compartments is that the length of each compartment should not exceed 0.1 times the space constant and the variation of diameter should be less than 2 µm within the compartment (Arvanitaki, 1942; Clark and Plonsy, 1971; Segev et al., 1989; West, 1996). Each compartment is electrically coupled to the adjacent compartments (Perkel and Mulloni, 1978). So, there is a flow of current between two successive compartments. The magnitude and direction of this current will depend on the potential difference and coupling conductance between the adjacent compartments, and also on the coupling conductance between these compartments. Therefore, the dynamic equation for the membrane potential will change. This equation for the µth compartment can be written as: dVmm (12.8) Cm = - I ionic (Vm , w ) + I (t ) - g m ,m +1 ¥ (Vmm - Vmm +1 ) - g m ,m -1 ¥ (Vmm - Vmm -1 ) dt
172 Deepak Mishra et al. where gµ, µ−1 is the coupling conductance between µth and µ–1th compartment, gµ, µ+1 is coupling conductance between µth and µ+1th compartment and Vmµ is the membrane potential of µth compartment. The coupling conductance does not have any influence on the dynamics of potassium currents. So the other equations remain same. The value of the coupling conductance between µ and µ' compartments can be given by: g m ,m ¢ =
am am2¢ rL Lm ( Lm am2¢ + Lm ¢ am2 )
where L is the length of compartment, a is the radius of the compartment and rL is the intracellular resistivity.
Simulation Results The first simulation has been carried out by injecting a constant current of 18 µA/cm2 only at the first compartment. The acronyms c–1, c–3 and so on in Figures 12.9, 12.10, 12.11 and 12.12 are used for the compartment numbers, that is, c–1 is the acronym for the first compartment and so on. The initial values w(0) and V(0) have been taken as 0.01 and −14.8 mV for which spike was generated for point neuron model. FIGURE 12.9
Response of Morris–Lecar model when current is injected only at first compartment
Note: (a) action potential (b) phase portrait.
Non-linear Dynamical Analysis
173
So, it is observed from Figure 12.9 that current injection at only the first compartment is not sufficient to generate spike. After this, we find out the responses for different types of current injections. Two cases have been considered. Figure 12.10 shows the response, while the current injection uniformly varies inversely as the fourth root of axon length. Figure 12.11 is for uniform variation of current as inverse of square root of the axon length. FIGURE 12.10 of axon length
Response of Morris–Lecar model when current injection varies inversely with fourth root
Note: [w(0) = 0.01] (a) action potential (b) phase portrait. FIGURE 12.11 axon length
Response of Morris–Lecar model when current injection varies inversely with square root of
Note: [w(0) = 0.01] (a) action potential (b) phase portrait.
174 Deepak Mishra et al. FIGURE 12.12 of axon length
Response of Morris–Lecar model when current injection varies inversely with square root
Note: [w(0) = 0.09] (a) action potential (b) phase portrait.
The initial conditions are taken to be same as the earlier case. The action potential is generated when the current is varying inversely as fourth root of the distance along the axon. There is no generation of action potential at the end of the 42nd compartment, even if the variation is of square root instead of fourth root. So, the action potential could not travel up to the synapse for a higher decaying space rate of current. Hence, the current distribution along the axon is also an important factor for a neuron to fire the spike. To study the effect of initial value of activation variable in the multi-compartment model, simulation has been carried out when the current is varying inversely as the fourth root of axonal distance, which was the case when action potential was generated for the earlier initial condition. The new initial condition for the activation variable of the potassium channel was set to 0.09. The simulation again proves the effect of initial condition as there is no action potential generation even at the 11th compartment.
Inference The simulations prescribed the distribution of current along the length of the axon as an important factor for generation of action potential. However, it is assumed that the neuron gets stimuli at the synapse. So, even if there is some current near the synapse, the neuron fires. This type of the distribution can be represented by the current that varies inversely as a square root of axon length because of its high attenuation. But, this type of current distribution also fails to generate action potential at the end of the 42nd compartment. So, we can assume that there is some means by which the axon gets stimuli far from the soma. The means can be Ephaptic Interaction. It is the process where another axon partly
Non-linear Dynamical Analysis
175
stimulates the axon considered. The initial condition is even more prone in long axon rather than a point neuron. This could be because the action potential has to travel a long distance, and hence needs a greater momentum initially.
Logical Justification To justify the conclusion, first let us assume that the current is injected at the first compartment only. The coupling conductances are generally of the order of 3 ms/cm2. The potential difference between first two adjacent compartments after the first time interval of 0.01 ms is found out to be 0.18 mV. Hence, the current injection to the second compartment after elapse of 0.01 ms is only 0.54 µA/cm2. This is not sufficient enough to nullify the hyperpolarizing effect of potassium current. So, the potential of the second compartment will go down. A greater amount of current will flow from the first to the second compartment in the second time span of 0.01 ms. This will hyperpolarize the first compartment. This will aid the hyperpolarization of the second compartment, and also the first. As a result, action potential will not be generated. So, each compartment has to be supplied with a minimum amount of current that will act in conjunction with the current flow from the neighbouring compartment to generate action potential. This strengthens the concept of ephaptic interaction.
CONCLUSIONS We have analysed single neuron models for various values of bifurcation parameters. We have carried out our analysis on the Hodgkin–Huxley, FitzHugh–Nagumo, Wilson–Cowan and the Cortical neuron models. Also, the propagation of action potentials in the neuron has been studied using multi-compartmental modelling, and associated phenomena of attenuation and phase difference have been demonstrated considering the Morris–Lecar neuron model. It has been found from the dynamical analysis that behavioural change in the dynamics takes place with the variation in its bifurcation parameters. The chaotic response is observed in cortical neuron model at some values of the bifurcation parameters. It is also concluded from the simulation results carried out for the axonal dynamics that the generation of action potential is very much dependent on the initial conditions as well as the spatial distribution of current injections. Increased dependency of action potential generation on initial conditions in case of long axons has been studied.
REFERENCES Arvanitaki, A. 1942. ‘Effects Evoked in an Axon by the Activity of a Contiguous One’. Journal of Neurophysiology, 5: 89–108.
176 Deepak Mishra et al. Clark, J.W. and R. Plonsy. 1971. ‘Fiber Interaction in a Nerve Trunk’. Biophysical Journal, 11: 281–94. Dyan, P. and L.F. Abbott. 1999. Theoretical Neuroscience: Computational and Mathematical Modeling of Neural Systems. Massachusetts: The MIT Press. FitzHugh, R. 1961. ‘Impulses and Physiological States in Models of Nerve Membrane’. Biophysical Journal, 1: 445–66. ———. 1969. ‘Mathematical Models for Excitation and Propagation in Nerve’. In H.P. Schawn (ed.), Biological Engineering. New York: McGraw-Hill. Freeman, W.J. and W. Schneider. 1982. ‘Changes in Spatial Patterns of Rabbit Olfactory EEG with Conditioning Odors’. Physchophysiology, 19: 44–56. Gerstner and Kistler. 2002. Spiking Neuron Models. Single Neurons, Populations, Plasticity. Cambridge: Cambridge University Press. Hausser, M., G. Stuart, C. Racca and B. Sakmann. 1995. ‘Axonal Initiation and Active Dendritic Propagation of Action Potential in Substantia Nigra Neurons’. Neuron, 15: 637–47. Hodgkin, A.L. and W.A.H. Rushton. 1948. The Electrical Constants of a Crustacean Nerve Fiber, Proceedings of the Royal Society of London, B133: 444–79. Hodgkin, A. and A. Huxley. 1952. ‘A Quantitative Description of Membrane Current and Its Application to Conduction and Excitation in Nerve’. Journal of Physiology, 117: 500–44. Holt, G.R. 1998. ‘A Critical Reexamination of Some Assumptions and Implications of Cable Theory in Neurobiology’. Doctoral Dissertation, California Institute of Technology. Izhikevich, E.M. 1999. ‘Class 1 Neural Excitability, Conventional Synapses, Weakly Connected Networks and Mathematical Foundations of Pulse Coupled Models’. IEEE Transaction on Neural Networks, 10: 499–507. Koch, C. 1999. Biophysics of Computation. Oxford: Oxford University Press. Morris, C. and H. Lecar. 1981. ‘Voltage Oscillations in the Barnacle Giant Muscle Fiber’. Biophysical Journal, 35: 193–213. Nagumo, J.S., S. Arimoto and S. Yoshizawa. 1962. Proceedings of the IRE 50: 2061–74. Perkel, D.H. and B. Mulloni. 1978. ‘Electronics Properties of Neurons: Steady-State Compartmental Models’. Journal of Neurophysiology, 41: 621–39. Rinzel, J. and B.B. Ermentrout. 1998. ‘A Formal Classification of Bursting Mechanism in Excitable Systems’. In C. Koch and I. Segev (eds), Methods in Neuronal Modeling (pp. 251–92). Cambridge: The MIT Press. Segev, I., J.W. Fleshman, R.W. Burke. 1989. Methods in Neuronal Modeling. In C. Koch, and I. Segev (eds). Cambridge: The MIT Press. Strogatz, S.H. 2000. Nonlinear Dynamics and Chaos. Reading, Massachusetts: Westview Press. West, R.M.E. 1996. ‘On the Development and Interpretation of Parameter Manifolds for Biophysically Robust Compartmental Models of CA3 Hippocampal Neurons’. Doctoral Dissertation, University of Minnesota Dissertation. Wilson, H.R. 1999. ‘Simplified Dynamics of Human and Mammalian Neocortical Neurons’. Journal of Theoretical Biology, 200: 375–88. Zak, S. 2003. Systems and Control. Oxford: Oxford University Press.
Chapter 13 Smoke Without Fire: What Do Virtual Experiments in Cognitive Science Really Tell Us? Peter R. Krebs
INTRODUCTION
A
dvances in computer technology make it possible to deal with computational models of considerable complexity in the field of cognitive science. The ability to deal with complexity should not detract from some of the fundamental issues that even apply to the simplest of models. Some of these issues concern the kind of evidence that models can actually provide as evidence to explain or to support the claims that are made about the real entities that are modelled. Essentially, the question is about what kind of conclusions we can draw from models. A few decades ago, models and simulations were often described as mock-ups, analogies, simplifications or metaphors, and sometimes the term simulation carries connotations of pretence, or even deceit. However, in the world of science and technology, these connotations have largely faded, and the use of computational models and simulations has become common practice. Keller (2003) remarked that during the 1940s, ‘the valence of the term [simulation] changes decisively: now productive rather than merely deceptive, and, in particular, designating a technique for the promotion of scientific understanding’ (Keller, 2003: 198). While models are generally accepted as tools in the empirical sciences, their epistemological status is still to be determined. Ziman (2000) points out that the notion of a model defies formal definition like other metascientific concepts. Without a proper definition of what is understood by a model, it is also difficult to place models into a scientific framework. Models have served as the basis for major shifts in theories in the physical sciences. Bohr’s model of atoms, for example, changed the way in which chemists could predict the properties of substances. Models and simulations can also be in the form
178 Peter R. Krebs of some apparatus, like the model of an aeroplane in a wind tunnel, or the ball and stick models of molecules. Computer models and simulations (CMS)1 are a progression from computational utility. In the early days of modern computing, the range of problems that could be subjected to quantitative analysis had been radically extended (Keller, 2003). Calculations, which had been too complex and tedious to deal with numerically, became trivial in a very short period of time.2 CMS possess properties that extend the utility of previous models in terms of speed and numerical accuracy. They allow for convenient ‘what if’ experiments, where the behaviour of a mathematical model can be examined over a range of changing parameters. On this basis, CMS seem to surpass theoretical models, such as Bohr’s atom model, or physical models that are made of real material. However, there are particular problems in using CMS as objects of scientific experimentation, and also as tools to support claims about theories.
EXPERIMENTATION CMS in engineering disciplines or the physical sciences deal with real-world entities and phenomena. Some mapping from the real entities to the modelled entities usually exists, and such ‘realistic models’ are approximate representations of the real world. How approximate do the models have to be for scientific experimentation? For Hacking, experimentation is not merely about the observation of phenomena and the subsequent inferences about the underlying theories, but about observing and interfering with the objects in question. The ability to manipulate objects is an essential part of the experiment, which is ‘to create, produce, refine and stabilize phenomena’ (Hacking, 1983: 230). The close connection between experiment and some real-world entities is also a key requirement in the definition offered by Harré, who says that’ [an] experiment is the manipulation of [an] apparatus, which is an arrangement of material stuff integrated into the material world in a number of different ways’ (Harré, 2003: 19).
1
I will use the terms computer model and computer simulation interchangeably. In the context of this chapter, models and simulations are mathematical constructs that have been instantiated as executable programs. A simulation is a model that has been designed to illustrate its dynamics, but a clear distinction is neither possible nor necessary in the context of this chapter. 2 Monstrous analog tide calculating machines, high precision developments of Lord Kelvin’s tide predictors, were operating until the mid-1960s. Williams (1997) notes that the machine, constructed and operated by the US Coast and Geodetic Survey, could calculate the height of the tides to the nearest 0.1 ft for each minute of the year for a location in a few minutes. The magnitude of calculations involved to establish a tidal forecast can be gauged by the fact that in a modern computer, the cosine sub-routine is called about 20 million times to predict the tides for a single year for a single location (Williams, 1997). The tide predictors were essentially mechanical models of the cyclic movement of heavenly bodies. Nowadays, these calculations can be solved numerically with higher precision in a few seconds.
Smoke Without Fire
179
The kinds of experiments that fit the criteria, which relate to the discussions by Harré and Hacking, are the activities we often associate with what happens in the laboratory. These are the kinds of experiments that we know from our high school days. However, it has become obvious that the vast majority of experiments are different from this stereotypic view (Morgan, 2003). There are no materials that could possibly be manipulated in experiments with CMS. The material, the apparatus and the process of interference are all replaced by data structures and computational processes.3 The nature of the entities and the phenomena that are the points of interest in the field of cognitive science dictates that CMS are often the only way to do any experimentation at all. The experiment is moved into the realm of the virtual for convenience or necessity.
ASSUMPTIONS AND METHODS In the current Theory of Mind in cognitive science, it is a fundamental assumption that cognitive functions are computational processes, or at least computable processes, performed by brains. This hypothesis is attractive, because, as Sterelny puts it, [i]t is good research strategy to try to model our information processing on something we already know a bit about. And we do know a good deal about computation, both from the theory of formalized systems of reasoning and from the actual implementation of some of those systems on real machines (Sterelny, 1989: 74).
It seems obvious that CMS would be the ideal approach to provide the clues and explanations for the Theory of Mind, as computation is the essence of what happens in brains and in CMS. Viewing the brain as a black box that computes is only one level of description. Another takes into account the fact that brains are amongst other things composed of about 1010 neurons that are interconnected by about 1014 synapses. The neuron has been determined as the smallest building block in terms of computational power.4 CMS can be based on either assumption, resulting in models that treat either the brain or the neuron as a black box. In Artificial Intelligence, these levels of description correspond in many ways to the symbolic and connectionist paradigms. The symbolic approach is primarily concerned with what goes on in the brain, while the connectionists are interested in how things happen in the brain. The connectionist approach adds a
3 The only material part of the experiment is the computer hardware. With the proliferation of the personal computer, it turns out that the vast majority of CMS are implemented on the same platform, that is, more or less identical hardware. 4 Penrose (1990) and others propose that some form of quantum computing is performed within the neuron.
180 Peter R. Krebs kind of neural plausibility, because these modelling techniques are methodologically comparable with what neurons do in brains (Schultz, 2003). At this point, it is important to note that the description of the functionality of a neuronal model in AI is already far removed from the actual observable functions of a neuron, even if we restrict the discussion to the most basic input/output functionality of the biological neuron.5 The theory about the outward computational functionality of the simplified neuron does not take this behaviour into account. The dissimilarities initially concerned hardware-based models, but shifting these models into the ‘virtual’ does not reduce, or eliminate, these discrepancies. For example, the rate of switching is of importance in biological neurons,6 but it has no significance in the neural models used commonly in artificial neural nets (ANN). Also, the number of connections between neurons is much larger than most models take into account. Some neurons may have 106, or 107, synapses, whereas the number of connections in models is usually restricted to less than 102. Modelling techniques at neural level are only loosely aligned with the real biology and the real observable behaviour of neurons.
CONSTRUCTION The earliest functional models of neurons by McCulloch and Pitts (1943) were simple switching devices, and the modelled behaviour was strictly according to the rules of elementary logic. The discreet components, which only operated in switching mode, that is, on/off, were designed to implement the basic logic functions AND, OR and NOT. These early models had little in common with real neurons and the experiments did not yield much in terms of progress towards an artificial intelligence. With hindsight, Copeland points out that [h]alf a century later on, it is clear that people were putting two and two together to make five. Even at that time, there was a certain amount of nagging evidence regarding the dissimilarities between neurons and computing hardware (Copeland, 1993: 185).
Hebb’s discovery, that some neural connections are modified over time by patterns of excitations (Hebb, 1949), led to the development of the perceptron, which represented one of the first learning networks (Rosenblatt, 1958, 1962). A decade later, the connectionist
5
For example, real neurons change their output behaviour in response to the firing rate of neighbouring neurons. The rate at which information is presented to the inputs (synapses) has a real effect on the eventuality and timing of the neuron’s output (spiking). 6 The firing rate changes the behaviour dynamically, so that a neuron is more sensitive after a burst of activity.
Smoke Without Fire
181
programme came almost to a halt after the ‘remorseless analysis of what perceptrons cannot do’ (McLeod et al., 1998: 323). This analysis by Minsky and Papert (1988) showed that perceptrons can only solve tasks that are linearly separable.7 Neural network models that employ primary elements of the all-or-nothing type without sophisticated transition functions offer little, if anything, beyond digital circuitry. Since then, artificial multilayered networks comprising non-linear neurons have been shown to be much more powerful than Minsky and Papert were willing to admit. Churchland writes that: [...] a nonlinear response profile brings the entire range of possible nonlinear transformations within reach of three-layer networks [...] Now there are no transformations beyond the computational power of a large enough and suitably weighted network (Churchland, 1990: 206).
and Elman et al. say that: [f]or some ranges of inputs [...] these units exhibit an all or nothing response (i.e., they output 0.0 or 1.0). This sort of response lets the units act in a categorical, rule-like manner. For other ranges of inputs, however, [...] the nodes are very sensitive and have a more graded response. In such cases, the nodes are able to make subtle distinctions and even categorize along dimensions which may be continuous in nature. The nonlinear response of such units lies at the heart of much of the behaviour which makes networks interesting (Elman et al., 1998: 53).
As Elman et al. point out, the non-linear and continuous transition functions of more ‘nature-like’ neural models seem to offer much more. The question is, of course, how much more? If we examine the mathematics of the models,8 it becomes apparent that there is only some rudimentary, if not superficial, functional similarity on offer.9 This kind of simplicity remains in the assemblies of these units, ANNs. Feed forward and recurrent networks implement functions from the inputs to the outputs—these networks perform linear or non-linear regression. The connections between the nodes in the hidden layer contain information about the mappings of these functions, and these mappings become the source of ‘insights’ about what goes on in the ANNs. It is possible, with appropriate statistical methods, to map this information (activation patterns) onto locations in n-space to make new claims about what the models achieved. Cluster analysis is often used as a method to gain insights into the internal representations of ANNs, but cluster analysis is not without some conceptual problems. 7
The relatively simple XOR-function cannot be handled by a single-layer network of perceptrons. Refer to Haykin (1999), Russel and Norvig (1995), Hoffmann (1998), and many others. 9 Neural models in the field of Computational Neuroscience are much more aligned with the biology and chemistry of neurons, but they are not the kind of model employed in Artificial Intelligence (AI). 8
182 Peter R. Krebs Clark (2001) argues that cluster analysis is an analytic technique to provide answers to the crucial question of what kinds of representations the network has acquired. However, cluster analysis does not reveal anything that is not already contained in the raw data of the model. The relationships and patterns in the input data sets and training data sets become embedded in the structure of the network during training.10 What counts are the mathematical and statistical relations that are contained in the training data sets. In many cases, the relations may just be tacitly accepted. In other models, these relations are purposefully introduced from the outset. Under these conditions, the relations are part of the model’s design. Elman (1990), for example, states that ‘13 classes of nouns and verbs were chosen’ for generating the data sets. Whether the relations in the data are introduced by design, or whether the experimenter is unaware of these statistical artefacts, there should be no surprise that the analysis will reveal these relations later during the analysis of the experiment. The implementation of a model as an ANN and the subsequent extraction of results that are already in the data may have little value in terms of obtaining empirical evidence. The training set of pairs of input and output vectors already contains all there is to the model, and the ANN does not add anything that could not be extracted from the training sets through other mathematical or computational methods. Green (2001) argues that: these results [from the analysis of ANNs] are just as analytic as are the results of a mathematical derivation; indeed they are just mathematical derivation. It is logically not possible that [the results] could have turned out other than they did (Green, 2001: 109).
A trained ANN implements a mapping from the input nodes (I1–k} to the output nodes (O1–i). The power of the ANN is in its ability to implement some function O1–n = f(I1–k) from the training data set. Hoffmann (1998) emphasizes this point and says that [t]he greatest interest in neural nets, from a practical point of view, can be found in engineering, where high-dimensional continuous functions need to be computed and approximated on the basis of a number of data points (Hoffmann, 1998: 157).
The modeller does not need to specify the function ƒ, in fact, the modeller does not even need to know anything about ƒ. Knowledge extraction (KE) from ANNs is concerned with providing a description of the function ƒ that is approximated by the trained ANN. The extraction of the function ‘lies in the desire to have explanatory capabilities besides the pure performance’ (Hoffmann, 1998: 155). The ability to determine ƒ, however, may or may not add to the explanatory value of the model.
10 The patterns and relationships in these data sets can either be carefully designed or might be an unwanted by-product.
Smoke Without Fire
183
INTERPRETING MODELS The assumption that symbols and their semantic contents are distributed throughout the network is part of the connectionist doctrine. However, the interpretation of experimental results in the context of neural nets is not possible without the use of symbols at some level of description. Hoffmann points out that: [...] in more complex systems, the use of symbols for describing abstractions of the functionality at the lowest level in inevitable [...] any description of a sufficiently complex system needs layers of abstraction. Thus, even if a non-symbolic approach uses tokens at its base level which cannot be reasonably interpreted, there still needs to be a more abstract level of description (Hoffmann, 1998: 257).
For a meaningful interpretation of the network and its dynamics, it is necessary to convey content and meaning in terms of non-distributed, or localized, symbols. Elman et al. suggest that: localist representations [...] provide a straightforward mechanism for capturing the possibility that a system may be able to simultaneously entertain multiple propositions, each with different strength, and that the process of resolving uncertainty may be thought of as a constraint satisfaction problem in which different pieces of information interact (Elman et al., 1998: 90).
Localized representations and distributed representations cannot be used together in a single representational system without confusing the semantic meaning of representations. ANNs are described as having distinct and discrete inputs and outputs, each labelled with distinct and discrete meaning. These localized representations are no longer available once the focus shifts on to hidden nodes within the network, and the ‘representations’ are now described in terms of weights, or synaptic strengths, between individual units. The mistake, I believe, is to bring the top–down psychological model and the bottom–up neural environment together, and to treat the result as a coherent and meaningful combination. The difficulties became apparent when the number of neurons in an ANN (feed forward network of simple recurrent networks) is small, and the model is designed to support or explain claims about higher cognitive functions. Elman (1990), for example, presented bit patterns to a small recurrent network, where each individual input node represented a particular word in the human language. The patterns themselves were presented in sequences forming two- and three-word sentences that had been generated according to a set of fixed templates. A cluster analysis of the hidden nodes revealed that the trained network exhibits similar activation patterns for inputs (words) according to their relative position in the sequence (sentence). The analysis of these activation patterns allowed for the classification of inputs (words) into
184 Peter R. Krebs categories like nouns or verbs. It is important, here, to understand that these results are not furnished by the ANN as some output, instead they are interpretations of internal structures at a higher level. The actual role of the ANN is that of a predictor, where the network attempts to predict the next word following the current input.11 If the ANN is meant to be a model of what happens at the neural level, then the question arises: what is the mechanism responsible for the equivalent analysis of activation patterns in the brain? We will have to assume another neural circuit to do an analysis of the hidden nodes. This new network could categorize words into verbs and nouns, but then we need another circuit to categorize words into humans, non-humans, inanimates, or edibles, and another to categorize words into mono-syllabic and multi-syllabic. In fact, we will need an infinite number of neural circuits just for the analysis of word categories. Churchland (1998) describes a small feed forward network that could model more challenging cognitive functions, albeit hypothetically. He suggests that an artificial network may have an appropriate architecture for learning and simulating moral virtues. He considers, given the ‘examples of perceptual or motor categories at issue’, that a network would be able to map concepts like cheating, tormenting, lying or self-sacrifice within a n-space of classes containing dimensions of morally significant, morally bad or morally praiseworthy actions. Churchland says that: [t]his high-dimensional similarity space [...] displays a structured family of categorical ‘hot spots’ or ‘prototype position’, to which actual sensory inputs are assimilated with varying degree of closeness (Churchland, 1998: 83).
I believe that this approach towards a calculus of moral virtues is flawed for two reasons. First, there is the question of what kinds of ‘actual sensory inputs’ could be available to train a network in moral virtues, or to condition a brain in moral virtues. The second problem is whether moral viewpoints can be synthesized from a possibly large number of discrete constituents. For this approach to work, it would have to be possible to define a moral action as the function over a set of discrete inputs. However, a morally bad action like stealing an item of clothing is not simply the result of poverty = true and night time = true or low temperature = true and coat available = true, and so on. If it were so, then our lives would need to be expressed in terms such that a set of mathematical functions could determine our next action, a proposition that has profound philosophical consequences. Regardless of whether one subscribes to Chomsky’s notion of a universal grammar, relationships between syntax and semantics do exist in natural language. It is these relationships that were explored by Elman (1990). It should be clear that a grammar has far less rules than what makes up the moral fabric of a human being. Unlike the formalisms
11
The actual word, which follows the input in the training set, is used as the target to determine the error for back propagation during the training phase.
Smoke Without Fire
185
that are evident in natural language, there are no similar formalisms available for the analysis of moral virtues by means of non-linear regression. A more interesting problem lies in the interpretation of representations that are within the network. First, there is the question of locating suitable representations that could carry any semantics, given that the representations are distributed in the network. Rosenblatt explained that: [i]t is significant that the individual elements, or cells, of a nerve network have never been demonstrated to possess any specifically psychological functions, such as ‘memory’, ‘awareness’, or ‘intelligence’. Such properties, therefore, presumably reside in the organization and functioning of the network as a whole, rather than in its elementary parts (Rosenblatt, 1962: 9).
However in CMS, the input nodes and output nodes are treated as localized representations (symbols). Individual model neurons do have semantics bestowed upon them by Elman (1990), Churchland (1998), and Rogers and McLelland (2004), who map meaningful words and moral concepts to the inputs and outputs of their networks. Treating the activation patterns of the hidden units as a resource for categorical ‘hot spots’, to use Churchland’s term, is an even more contentious exercise. The relationships and patterns in the input data sets and training data sets become embedded in the structure of the network during training.12 The internal representations, which are ‘snapshots of the internal states during the course of processing sequential inputs’ (Elman, 1990), are extracted by means of cluster analysis of the hidden layer in the ANN. Who really does the analysis and interpretation of the distributed representations? The experimenter performs these tasks using a new tool, that is, cluster analysis—the network has no part in this. Moreover, an appropriate analysis, performed on the training data, would yield the same information. The networks merely compute functions, and the activities of the networks do not add any additional information. Despite all the complexities of the mathematics of ANNs, the functions that are performed are relatively trivial. The class of simple ANNs that I have discussed here cannot provide any new ‘insights’ in any meaningful symbolic13 or coded form on some output nodes. This, however, would have to be a crucial function of the model to be considered neurologically plausible. For a model to be neurologically plausible, it would need to deduce new information about itself. More importantly, it would be necessary to signal the newly obtained knowledge to other neurons by changing the state of some nodes. Both cluster analysis and current methods of KE clearly fail to do this, although more recent developments in KE can
12 The patterns and relationships in these data sets can either be carefully designed or might be an unwanted by-product. 13 I do not think that ‘distributed’ or ‘sub-symbolic’ representations are helpful here. Moreover, this alternative approach of dealing with network structures is usually not employed by modellers either.
186 Peter R. Krebs deliver much more accurate description of ƒ. However, the renewed and possibly accurate synthesis of relations that were present in a training data set does not warrant claims that the ANN ‘discovered’, ‘learned’ or ‘recognized’ something or other, even if these relations were not evident to the experimenter before. The ability to determine a function ƒ that is contained in some data set illustrates the power of ANNs as analytical tools. However, it should be clear that a different analytical tool could also have been used to detect the function ƒ. We must conclude then that the model has failed to explain any processes at the neural level. Instead, the network model has only succeeded in offering an alternative method to encode the data, and the cluster analysis provides an alternative method to analyse the data.
CONCLUSIONS A model is a simplification of its real-world counterpart, but models must maintain some plausible connection to real-world objects or real-world phenomena. Green (2001) suggested that some of the ‘apparent’ successes of connectionist modelling may well be based on a rather vague concept of what is actually modelled. The simple functional neurons that are employed in AI, only loosely resemble actual biological neurons, and ANNs exhibit only superficial commonalities with brain structures. The building blocks and tools used in the connectionist paradigm of AI nevertheless offer some plausibility for the bottom–up approach. While many models share by design the connectionist architecture, the processes and functions under investigation seem quite different. The investigations about language, moral virtues and many other topics, belong to the top– down approach, where localized representations are used to convey the semantic contents of sensory and conceptual entities. Merging the two opposing paradigms within models is not without problems. We can easily assign meaning to localized representations, and we can manipulate representations without loss of semantics, provided we maintain appropriate syntactic rules. The processes break down when localized representations are ‘manufactured’ by assigning them to concepts seemingly emerging from ANNs. The danger is that statistical artefacts are presented as novel phenomena of the model. However, there are no novel phenomena emerging from the kinds of neural nets employed in the field of artificial intelligence. There is nothing intelligent happening within these mathematical structures when we compare them to what seems to be going on in the brain—there is no fire, not even a spark.
ACKNOWLEDGEMENT I would like to thank Anthony Corones for his encouragement, valued comments and suggestions on earlier drafts of this chapter.
Smoke Without Fire
187
REFERENCES Churchland, P.M. 1990. Cognitive Activity in Artificial Neural Networks. In Cummins and Delarosa Cummins 2000. ———. 1998. Toward a Cognitive Neuro-biology of the Moral Virtues. In Branquinho 2001. Clark, A. 2001. Mindware: An Introduction to the Philosophy of Cognitive Science. Oxford: Oxford University Press. Copeland, J. 1993. Artificial Intelligence: A Philosophical Introduction. Malden: Blackwell. Cummins, R. and D. Delarosa Cummins (ed.). 2000. Minds, Brains, and Computers: The Foundations of Cognitive Science. Malden: Blackwell. Elman, J.L. 1990. ‘Finding Structure in Time’. Cognitive Science, 14: 179–211. Elman, J.L., E.A. Bates, A. Karmiloff-Smith, D. Parisi and K. Plunkett. 1998. Rethinking Innateness: A Connectionist Perspective on Development. Cambridge, Massachusetts: MIT Press. Green, C.D. 2001. ‘Scientific Models, Connectionist Networks, and Cognitive Science’. Theory & Psychology, 11: 97–117. Hacking, I. 1983. Representing and Intervening. Cambridge: Cambridge University Press. Harré, R. 2003. The Materiality of Instruments in a Metaphysics for Experiments. In H. Radder (ed.) The Philosophy of Scientific Experiment. Pittsburgh: University of Pittsburgh Press. Haykin, S. 1999. Neural Nets: A Comprehensive Foundation. Upper Saddle River, New Jersey: PrenticeHall. Hebb, D.O. 1949. The Organization of Behavior, Chapter 19, In Cummins and Delarosa Cummins 2000. Hoffmann, A. 1998. Paradigms of Artificial Intelligence, Springer. Keller, E.F. 2003. Models, Simulation, and ‘Computer Experiments’, In H. Radder (ed.) The Philosophy of Scientific Experiment. Pittsburgh: University of Pittsburgh Press. McCulloch, P. and W. Pitts. 1943. ‘A Logical Calculus of the Ideas Immanent in Nervous Activity’. Bulletin of Mathematical Biophysics 5: 115–33. McLeod, P., K. Plunkett and E.T. Rolls. 1998. Introduction to Connectionist Modelling of Cognitive Processes, Oxford: Oxford University Press. Minsky, M. and S. Papert. 1988. Perceptrons: An Introduction to Computational Geometry (2nd edition), Cambridge, Mass chusetts: MIT Press. Morgan, M.S. 2003. Experiments Without Material Intervention, In H. Radder (ed.) The Philosophy of Scientific Experiment. Pittsburgh: University of Pittsburgh Press. Penrose, R. 1990. The Emperor’s New Mind, London: Vintage. Rogers, T.T. and J.L. McLelland. 2004. Semantic Cognition: A Parallel Distributed Processing Approach. Cambridge, Massachusetts: MIT Press. Rosenblatt, F. 1958. ‘The Perceptron: A Probabilistic Model for Information Storage and Organization in the Brain’. Psychological Review, 65: 386–408. ——. 1962. Principles of Neurodynamics. Washington: Spartan. Russel, S. and P. Norvig. 1995. Artificial Intelligence: A Modern Approach. Upper Saddle River, NJ: Prentice Hall. Schultz, T.R. 2003. Computational Developmental Psychology. Massachusetts: MIT Press. Sterelny, K. 1989. Computational Functional Psychology: Problems and Prospects. In P. Slezak and W.R. Albury (eds) Computers, Brains and Minds: Essays in Cognitive Science. Dordrecht: Kluwer. Williams, M.R. 1997. A History of Computing Technology. Los Alamitos: IEEE Press. Ziman, J. 2000. Real Science: What It is, and What It Means. Cambridge: Cambridge University Press.
Chapter 14 Complex Primitives and Their Linguistic and Processing Relevance Aravind K. Joshi
INTRODUCTION
T
he conventional (mathematical) wisdom in specifying a grammar formalism is to start with basic primitive structures as simple as possible, and then introduce various operations for constructing more complex structures. These operations can be simple or complex, and the number of operations (although finite) need not be limited. New operations (simple or complex) can be introduced in order to describe more complex structures. An alternate approach is to start with complex (more complicated) primitives, which capture directly some crucial linguistic properties and then introduce some general operations for composing these complex structures (primitive or derived). What is the nature of these complex primitives? In the conventional approach, the primitive structures (or rules) are kept as simple as possible. This has the consequence that information (for example, syntactic and semantic) about a lexical item (word) is distributed over more than one primitive structure. Therefore, the information associated with a lexical item is not captured locally, that is, within the domain of a primitive structure (Joshi, 2004).
DOMAIN OF LOCALITY In a context-free grammar (CFG), the domain of locality (Figure 14.1) is the one-level tree corresponding to a rule in a CFG, for example, rules such as S → NP VP, VP → V NP and V → likes. It is easily seen that the arguments of a predicate (for example, the two arguments of likes) are not in the same local domain. The two arguments are distributed over the
Complex Primitives and Their Linguistic and Processing Relevance
189
two rules (two domains of locality)—S → NP VP and VP → V NP. They can be brought together by introducing a rule S → NP V NP. However, then the structure provided by the VP node is lost. We should also note here that not every rule (domain) is associated with a lexical item. FIGURE 14.1
Domain of locality of a context-free grammar
Can a CFG be lexicalized by another CFG? It can be shown that this is not possible in general. However, it can also be shown that extending the domain of locality and adding two combining operations, substitution and adjoining lexicalization can be achieved. By extended domain, we mean, for example, for a verb, likes, we set up a larger structure (more than a one-level tree) that includes slots for both the arguments of likes. The operation of substitution (Figure 14.2) corresponds to attaching a structure to a frontier node of another structure, and adjoining (Figure 14.3) corresponds to inserting (splicing in) a structure at an internal node of another structure. The resulting framework is called the lexicalized tree-adjoining grammar (LTAG). We will illustrate it by means of a simple example, and then later, discuss some processing implications of this architecture (Joshi and Schabes, 1997). FIGURE 14.2
Substitution b
a X
X
X*
g X
X
b
190 Aravind K. Joshi FIGURE 14.3
Adjoining
By setting up appropriate structures (Figures 14.4 and 14.5) for the lexical items likes, Harry, Bill, think, and does, and substitution and adjoining as the two combining operations, it can be shown that (Figure 14.5) who(i) does Bill think Harry likes e(i) can be built up by starting with a structures for likes that includes slots for the arguments who and Harry, we can build a structure corresponding to who(i) Harry likes e(i) e is an empty element, which represents the complement of likes and the index i denotes the dependency between who and the empty element e. Similarly, starting with a structure for think with slots for its arguments, Bill and a clausal argument, we obtain Bill thinks S The structure associated with does is then adjoined to the root node of the think structure to obtain does Bill thinks S which is then adjoined into an internal node of the structure associated with who(i) Harry likes e(i) resulting in (Figures 14.6 and 14.7) who(i) does Bill think Harry likes e(i) Note that the dependency between who and the complement of likes, which is empty, local to the likes tree, has been stretched. It has become long distance. However, it started
Complex Primitives and Their Linguistic and Processing Relevance
191
out as a local dependency. A key property of LTAGs is that all dependencies are local, that is, they are specified in the elementary trees. They can become long distance as a result of the composition operations. All properties of LTAG (mathematical, linguistic, and even psycholinguistic) follow from the localization of dependencies to the elementary structures associated with the lexical items, thus directly from the architecture of the elementary trees. FIGURE 14.4
An LTAG example
FIGURE 14.5
An LTAG derivation
192 Aravind K. Joshi FIGURE 14.6
An LTAG derived tree
FIGURE 14.7
An LTAG derivation tree
PROCESSING ISSUES In this section, we will very briefly discuss some of the implications of the LTAG architecture for certain processing issues, in particular, (1) how the fine-grained distinctions between lexical and structural ambiguities that LTAG makes and their relevance to processing, and (2) the relative processing complexities of certain embedding constructions.
Complex Primitives and Their Linguistic and Processing Relevance
193
LEXICAL AND STRUCTURAL AMBIGUITIES Recently, there has been an increasing convergence of perspectives in the fields of linguistics, computational linguistics and psycholinguistics, especially with respect to the representation and processing of lexical and grammatical information. More specifically, this convergence is due to a shift to lexical and statistical approaches to sentence parsing. The particular integration of lexical and statistical information proposed in Kim et al. (2002) is highly relevant from the perspective of the LTAG architecture. LTAG associates with each lexical item one or more elementary structures (often called, supertags), which encapsulate the syntactic and associated semantic dependencies. The computational process based on supertags as if they are more complex parts of speech is called supertagging. It shows that much of the computational work of linguistic analysis, which is traditionally viewed as the result of structure building operations, can be viewed as lexical disambiguation, in the sense of supertag disambiguation. If the supertagging model is integrated in a psycholinguistics framework, then one would predict that many of the initial processing commitments of syntactic analysis are made at the lexical level, in the sense of supertagging. The model proposed in Kim et al. (2002) is an integration of the Constraint-Based Lexicalist Theory (CBL) (MacDonald et al., 1994), where the lexicon is represented as supertags with their distributions estimated from corpora. For example, in this model, there is a distinction between the prepositional phrase attachment ambiguity (PP ambiguity) as in (1) below (1) I saw the man in the park with a telescope and the PP attachment ambiguity as in (2) below. (2) The secretary of the general with red hair. In the first case, the PP (with a telescope) either modifies an NP, the man or a verb phrase VP, headed by saw (Figure 14.8). There are two supertags associated with the preposition with, one that attaches to an NP node and another that attaches to a VP node. That is, the ambiguity is resolved if we pick the correct supertag for with anchored on the preposition with. Thus, this PP attachment ambiguity will be resolved at the lexical level. However, in the second case, in both readings of (2), the supertag associated with with is the one whose root and foot nodes are both NP. Thus, in this case, the ambiguity will not be resolved at the lexical level. It can only be resolved at the level when the attachment is computed. The first PP attachment ambiguity is not really an attachment ambiguity. It should be resolved at an earlier stage of processing. In the second case, it will be resolved at a later stage. Similarly, the ambiguity associated with a verb such as forgot, because it can take either an NP complement as in (3) below
194 Aravind K. Joshi FIGURE 14.8
Two supertags for with
(3) The student forgot her name or a VP complement as in (4) below (4) The student forgot that the homework was due today is a lexical (supertag) ambiguity and need not be viewed as a structural ambiguity.
PROCESSING OF CROSSED AND NESTED DEPENDENCIES Context-free grammars (CFG) and the associated automata, pushdown automata (PDA) have been extensively used in modelling several aspects of sentence processing. Similarly, LTAGs are associated with embedded pushdown automata (EPDA) (Vijay-Shanker, 1987), an extension of PDA, that is, for every LTAG, G, there is an EPDA, M that recognizes the language of G, and vice versa. TAGs and EPDAs provide a new perspective on the relative ease or difficulty of processing crossed and nested dependencies, which arise in centre embedding of complement constructions. Consider the following examples, (5) for German, (6) for Dutch, and (7) for English. (5) Hans1 Peter2 Marie3 schwimmen3 lassen2 sah1 (6) Jan1 P iet2 M arie3 zag1 laten2 zwemmen3 (7) Jan saw Piet make Marie swim In (5), we have the nouns and the corresponding verbs in a nested order in a German construction, and in (6), we have a crossed order in a related construction in Dutch. The indices on the nouns and verbs show these dependencies. The English word order is shown in (7). These are called complement embedding constructions because each verb is embedded in a higher verb of which it is a complement, except, of course, the matrix
Complex Primitives and Their Linguistic and Processing Relevance
195
(top level and tensed) verb, that is, make is embedded in saw and swim is embedded in make, saw is the matrix verb. In (5) and (6), we have centre embedding of the complements. In the corresponding English construction as in (7), we do not have centre embedding, the complements are just iterated. The main observation for our purpose is that Standard German prefers the nested order and Standard Dutch prefers the crossed order. In the well-known study by Bach et al. (1986), the authors investigated the consequences of these differences between German and Dutch for the processing complexity of sentences, containing either nested or crossed dependencies. Stated very simply, their results show that Dutch is easier than German. More specifically, in their study, ‘German and Dutch subjects performed two tasks—ratings of comprehensibility and a test of successful comprehension on matched sets of sentences which varied in complexity from a simple sentence to one containing three levels of embedding’, three levels means three verbs and three nouns, as in our examples above. Their results show no difference between Dutch and German for sentences within the normal range (up to one level), but with a significant preference emerging for the Dutchcrossed order. These results in Bach et al. (1986) show that PDA cannot be the universal basis for human parsing mechanism (as the authors themselves point out). It can be shown that EPDA (corresponding to LTAG) can correctly model these differences qualitatively and quatitatively, both matching the results reported in (Bach et al., 1986). For further details, see Joshi (1990).
IMPLICATION FOR DISCOURSE STRUCTURE Insights from the LTAG theory have been extended to the domain of discourse structure, primarily by treating discourse connectives as lexical anchors of LTAG like trees at the discourse level, thereby blurring the line between sentence level and discourse level descriptions. For further details, see Forbes et al. (2001) and Webber et al. (2003).
ACKNOWLEDGEMENTS This work was partially supported by the National Science Foundation Grant STC-SBR 8920230.
REFERENCES Bach, E., C. Brown and W. Marslen-Wilson. 1986. ‘Crossed and Nested Dependencies in German and Dutch: A Psycholinguistic Study’. Language and Cognitive Processes, 1: 249–62.
196 Aravind K. Joshi Forbes, K., E. Miltsakaki, R. Prasad, A. Sarkar, A. K. Joshi and B. Webber. 2001. ‘D-LTAG System-Discourse Parsing with a Lexicalized Tree-Adjoining Grammar’. In ESSLI’2001 Workshop on Information Structure, Discourse Structure and Discourse Semantics, Helsinki. Joshi, A.K. 1990. ‘Processing Crossed and Nested Dependencies: An Automaton Perspective on the Psycholinguistic Results’. Language and Cognitive Processes, 5: 1–27. ———. 2004. ‘Starting with Complex Primitives Pays Off: Complicate Locally, Simplify Globally’. Cognitive Science, 28: 637–68. Joshi, A.K. and Y. Schabes. 1997. ‘Tree-Adjoining Grammars’. In G. Rosenberg and A. Salomaa (eds), Handbook of Formal Languages (pp. 69–123). Berlin: Springer. Kim, A., B. Srinivas and J.C. Trueswell. 2002. ‘The Convergence of Lexicalist Perspectives in Psycholinguistic and Computational Linguistics’. In P. Merlo and S. Stevenson (eds), Sentence Processing and the Lexicon: Formal, Computational and Experimental Perspectives (pp. 15–25). Philadelphia: John Benjamin Publishing. MacDonald, M.C., N.J. Pearlmutter, M.S. Seidenberg. 1994. ‘Lexical Nature of Syntactic Ambiguity Resolution’. Psychological Review, 101: 676–703. Vijay-Shanker, K. 1987. A Study of Tree-Adjoining Grammars, Ph.D. Dissertation, Philadelphia, University of Pennsylvania. Webber, B., M. Stone, A.K. Joshi and A. Knott. 2003. ‘Anaphora and Discourse Structure’. Computational Linguistics, 29: 545–88.
Chapter 15 Dissecting the Frog: Computational Approaches to Humour Perception Narayanan Srinivasan and Vani Pariyadath
Humor can be dissected, as a frog can, but the thing dies in the process and the innards are discouraging to any but the scientific mind. E.B. White
INTRODUCTION Research on humour perception has been carried out extensively under domains such as psychology, linguistics and artificial intelligence. One of the main approaches to study humour is the computational approach, which includes writing programs that process humour. While numerous models exist for humour generation (Binsted and Ritchie, 1994; Binsted and Takizawa, 1998; Loehr, 1996; McDonough, 2001; Stock and Strapparava, 2002), computational models for humour perception are relatively scarce (Taylor and Mazlack, 2004; Yokogawa, 2002). Commercially, models for humour generation are perhaps more viable making them more attractive for research than those for humour perception. It has been suggested that computational models for humour generation can be useful in advertising (Ritchie, 1998). In the human computer interface too, it is desirable to incorporate humour production or a machine sense of humour (Binsted, 1996; Stock, 2003). However, humour perception could be equally advantageous in the human computer interface. What better way to please the user than to show appreciation at all the appropriate places. Of more significance here is the fact that humour perception is an overlooked topic in cognitive science. This disregard is surprising, considering that humour perception is undoubtedly one of the most intriguing aspects of human cognition. What makes humour perception even more interesting is its pervasiveness. Although the usefulness of humour and laughter is debatable, one cannot deny their impact and the frequency
198 Narayanan Srinivasan and Vani Pariyadath with which they permeate into our daily life. There is no doubting that the average person perceives humour more frequently than produces humour. Of course, here we allude only to creative production of humour. Most jokes and humorous anecdotes that we narrate are not original, and are those heard or read elsewhere. Funny events that we relate are funny because we perceive them to be so and are not innately humorous. One-liners and wit too are not frequent productions for the average person. On the other hand, the number of times we perceive humour on a daily basis is clearly not a small number. It is important to understand some of the terms and classifications involved in humour research for understanding the research on humour perception. Based on the cortical areas activated during the processing of different types of humorous instances, humour can be classified into three categories—phonological jokes, semantic jokes and non-verbal humour (Goel and Dolan, 2001; Shammi and Stuss, 1999). Phonological jokes or jokes that involve wordplay such as, A bicycle cannot stand on its own because it is too tired. (Punoftheday.com)
(15.1)
activate the right temporal lobe. Semantic jokes or those that involve more complex processing, for example, All those who believe in psychokinesis, raise my hand.
(15.2)
activate the posterior temporal lobe. The third category is formed by non-verbal humour such as slapstick and cartoons. Of course, this categorization does not include other related forms of humour such as wit, irony and sarcasm, but we will follow this broad classification throughout the course of this chapter. Humour has also been classified by means of factor analytic studies, which show that both structural properties and the content of the jokes are essential in producing individual differences (Ruch, 2001). These studies divide humour into two main categories—incongruity-resolution and nonsense humour based on structure. At the same time, few contents were prominent enough to override the structural variance and form independent factors. In typical conversations, perception of humour is typically associated with explicit facial responses or laughter. Needless to say, perceived humour does not necessarily result in laughter. Also, a large proportion of laughter is more commonly aimed at nonhumorous utterances in social gatherings than at potential howlers (Provine, 1996). Bearing these points in mind, one can quite easily separate the study of humour and laughter. Some earlier studies have investigated humour perception using the degree of laughter as a measurement (McGhee, 1983), which we do not discuss in this chapter since the relationship between humour and laughter has yet to be quantified. To understand humour perception, some behavioural experiments have been performed to understand the processes and mechanisms involved in humour processing. These
Dissecting the Frog: Computational Approaches to Humour Perception
199
experiments have focussed on different response measures. Researchers have focussed on studying facial responses to humorous instances (Harris and Alvarado, 2005), reaction times (RTs) to identify humorous sentences (Vaid et al., 2003) and the degree of funniness of a joke (Ruch et al., 1993). Certain studies involve choosing the correct punch line to the given setup of a joke (Shammi and Stuss, 1999). This paradigm has generally been followed with humour-impaired subjects and will be discussed in detail later. Humour perception has been given some attention in behavioural studies, but why is such a peculiar but commonly used faculty ignored when it comes to computational models? A quick examination of the existing theories and models for humour perception is sufficient to realize that quantitative models are the need of the hour. As a first step in this direction, we review the existing literature on computational approaches of studying humour perception. As already mentioned, our primary interest is in models for humour perception and not humour production. While reviews of existing computational models exist in the literature (Mulder and Nijholt, 2002; Ritchie, 2001), the focus here is different. We critically evaluate the models based on the way they address significant issues in humour research and their plausibility as cognitive models for humour processing.
THEORIES OF HUMOUR PERCEPTION The various theories of humour perception may be broadly classified as relief, incongruity and linguistic theories. We have not discussed older theories such as the superiority theory in detail. The superiority theory considers laughter as an expression of sudden glory when faced with the misfortunes or defects of others, and dates as far back as Aristotle. Naturally, such a theory accounts for only a small subset of humorous instances and has been left out for this reason.
Incongruity Theories All incongruity theories rest on the notion of an initial incongruity/violation and eventual resolution of that incongruity/violation. The earliest incongruity theory was proposed by Abhinavagupta (Visuvalingam, 1983), who argued that humour is produced by incongruities in opposing emotions elicited by stimuli. In the recent incongruity– resolution (IR) model, humour perception is caused by a multistage process in which an initial incongruity is created, and then some further information causes that incongruity to be resolved (Ritchie, 1999). There are two variations of this model—Shultz’s Sudden Disambiguation model and Suls’ two stage model. Veatch’s violation theory and Paulos’ catastrophe theory models have also been grouped under the incongruity resolution domain since they also focus on incongruity/violation, but differ from other IR models.
200 Narayanan Srinivasan and Vani Pariyadath According to Shultz’ model, there are two or more interpretations possible as the audience listens to the set-up of the joke, where one interpretation is more obvious than the others. But, when the punch line is revealed, there is a conflict with the more obvious explanation, and the other interpretation is evoked. Here again, it is difficult to gauge why jokes can be perceived as funny when they have been heard once before, if surprise is essential. In Suls’ two stage model, an algorithm is used where predictions are made as the joke is listened to, but the punch line triggers an incongruity, which can only be resolved with a cognitive rule (Suls, 1983). The algorithm is as follows: • As text is read, make predictions • While no conflict with predictions, keep going • If input conflicts with predictions: If not ending—PUZZLEMENT If it is the eznding, try to resolve: No rule found—PUZZLEMENT Cognitive rule found—HUMOUR Suls’ model, however, does not elaborate on what such a cognitive rule might be. Veatch has proposed a related theory with three necessary and sufficient conditions for humour perception (Veatch, 1998). These conditions according to the Violation theory are: • V—The perceiver has in mind a view of the situation as constituting a violation of some affective commitment of the perceiver to the way something in the situation ought to be. That is, a ‘subjective moral principle’ of the perceiver is violated. • N—The perceiver has in mind a predominating view of the situation as being normal. • Simultaneity—The N and V understandings are present in the mind of the perceiver at the same instant in time. Veatch also points out that humour is doubly subjective, because it is a psychological event in a subjective perceiver and different subjects may differ in their perceptions. Simultaneity is also vital, because according to Veatch, if one interpretation follows another, it may lead to feelings of dismay or relief, but not humour. Veatch holds this to be the reason for surprise and ambiguity playing a major role in humour perception. He also stresses that the N interpretation must predominate over the V interpretation, in other words, the perceiver must feel the situation really is normal in spite of the violation. This theory has the advantage that a unique interpretation is not required for perceiving humour.
Dissecting the Frog: Computational Approaches to Humour Perception
201
Paulos has elaborated on a catastrophe theory model of jokes and humour, based on the assumption that humour contains ambiguity of some form, which causes one interpretation to be more apparent than another (Paulos, 1980). With the punch line, a switch between the two interpretations occurs, which can be modelled using a cusp catastrophe. The catastrophe theory model, it would seem, combines both the incongruity theory and prevailing psychological theories. Nevertheless, it is not clear how it accounts for puns of the type: What do you get if you do not pay your exorcist promptly? ... Repossessed.
(15.3)
In the first place, there is no obvious interpretation to the question. The act of paying an exorcist promptly is not the one most people would be familiar with, or the one that most people would have imagined even once before hearing this joke. The answer to the riddle does not supply another less-apparent solution either because no strong predictions have been made in the first place. The only expectation may have been of hearing a bizarre answer, and this type of jokes needs to be handled by other types of theories as well.
Linguistic Theories The second prominent category of theories on humour perception is made up by the linguistic theories. These consist of the Semantic Script Theory of Humor (SSTH) and a more general version of this theory. The SSTH applies script theory to verbal jokes and proposes a five-level joke model for joke representation. The SSTH has been further developed into the Global Theory for Verbal Humor or GTVH (Attardo and Raskin, 1991). The five levels in the model are: (a) (b) (c) (d) (e)
Surface Language Target and situation Template Basic
The GTVH describes five Knowledge Resources (KR) that are the concerned parameters of joke difference. These are: language, narrative strategy, target, situation, logical mechanism and script opposition. The main claim of the SSTH is that the text of a joke is always fully or in part compatible with two distinct scripts which are opposed to each other. The hierarchical organization of KRs is: SO, LM, SI, TA, LA, NS and LA, which results in production of joke text. This hierarchy is organized from less similar and less determined to more similar and more determined.
202 Narayanan Srinivasan and Vani Pariyadath Although the GTVH provides an elaborate description of joke texts, it neither explains humour perception nor puts forward any concrete predictions regarding what texts would be deemed as funny and unfunny. Attempts have been made to empirically test the claims made by the GTVH by introducing variations in any one of the six KRs in a joke and obtaining perceived changes in the degree of funniness (Ruch et al., 1993). But unless the GTVH deals with the actual mechanisms involved in perceiving humour, testing the claims made by it could prove futile.
Relief Theories The relief theories are based on the idea that laughter arises from suppressed thinking. These include Freud’s psychoanalytic theory (Freud, 1905) and Minsky’s frame theory (Minsky, 1980) for humour perception. We have also included Ramachandran’s false alarm theory (Ramachandran and Blakeslee, 1999) in this category since it bears a strong resemblance to the general theme involved in relief theories. Freud proposed censors which act to suppress thoughts on taboo subjects such as sex and aggression (Freud, 1905). According to Freud, this led to psychic energy being pent up, and laughter caused the release of this energy. The flaw in this theory was that a large number of jokes bear no sexual or aggressive content. Moreover, one may come up with several references to sex or aggression, which are not considered humorous. Minsky’s theory is based on Freud’s psychoanalytic theory. Minsky suggested that to a great extent, we process information using ‘common sense logic’, which sometimes proves useless (Minsky, 1980). To compensate for these lapses in applying logic, our system builds up a collection of ‘cognitive censors’, which quite naturally differs from person to person. This would assist in removing inconsistencies in logic, and prevent us from repeating such ‘mistakes’ in future. These censors can work in one of two ways—either by suppressing a thought when it occurs, or by preventing the occurrence of the thought altogether. The most important aspect of Minsky’s theory is his concept of ‘frames’. According to Minsky, perceptions are ordinarily interpreted by the mind in terms of previously acquired description-structures called Frames. These frames help us to process stereotyped situations, and Minsky suggests that this transition between frames is where the humour lies. While one frame seems more obvious at the onset, the ending of the joke calls for another less obvious frame to replace the earlier one. Various attempts have been made to understand how evolution paved the way for our ability to perceive humour. Minsky suggests that we developed this ability when we were learning to plan (Minsky, 1980). The ethological significance of the vocal component of laughter brings our attention to the possibility that laughter could be a signal to other members of the group to ‘stop whatever he was doing whether because dangerous, pointless, objectionable, ridiculous or otherwise forbidden’. Ramachandran has suggested that laughter is a means of communicating to other members of a social group that a false
Dissecting the Frog: Computational Approaches to Humour Perception
203
alarm has been detected (Ramachandran and Blakeslee, 1999), implying that humour perception is the detection of misplaced logic. While such a theory could explain why laughter is a social phenomenon, and why we tend to laugh at others’ mistakes, it does little to explain the mechanisms behind humour perception.
COMPUTATIONAL MODELS Computational models aim at writing programs to simulate how humour is perceived and generated by humans. These models can be divided into two main groups: those for humour generation and those for humour perception. As mentioned earlier, a large proportion of work on computational humour has been dedicated to humour generation owing to their usefulness in the human computer interface as well as potential commercial benefits. These models, however, do not depend on a theory of humour, and are mostly template-based models. The notable models include Joke Analysis and Production Engine or JAPE that generated punning riddles (Binsted, 1996; Binsted and Ritchie, 1994). JAPE was integrated with Elmo, a Natural Language Robot, where Elmo would produce a punning riddle generated by JAPE on demand or based on present context (Loehr, 1996). Other well-known models include the Mnemonic Sentence Generator, a program that converts any alphanumeric password into a humorous sentence (McDonough, 2001), LIBJOG or the Light Bulb Joke Generator built by Attardo and Raskin, which uses a template for creating light bulb jokes from a lexicon of stereotyped groups (Mulder and Nijholt, 2002). The HAHAcronym project is another recent experiment that generates humorous acronyms in order to make them easier to remember (Stock and Strapparava, 2003). Clearly, the HAHAcronym project and other models of humour generation do not deal with the cognitive processes involved in producing humour and do not merit a more detailed discussion here (for such a discussion, see Ritchie, 2001 or Mulder and Nijholt, 2002). The models for humour perception form a more interesting group. Of these, we will discuss in detail here the Knock-Knock Joke Recognizer (Taylor and Mazlack, 2004), Yokogawa’s (2000) Pun Analyser, Katz’s neural model (1996), Takizawa’s model (Ritchie, 2001) and GraPHIA (Pariyadath and Srinivasan, 1994). Unlike models for humour generation, these models make less use of templates. Taylor and Mazlack have designed a Knock-Knock Joke Recognizer which uses N-grams to detect wordplay and consequently punch lines in jokes involving verbal play (Taylor and Mazlack, 2004). An N-gram is basically a model that uses conditional probability to predict the Nth word based on N−1 previous words. The KK Joke Recognizer first validates the joke format, and then generates wordplay sequences, which are placed in a heap, in decreasing order of similarity. Each word in this heap is then retrieved, and an attempt is made to decompose it into a meaningful utterance, with multiple alterations, until a satisfactory wordplay sequence is obtained. The sequence is judged meaningful by using a bigram (an N-gram with N = 2). Once
204 Narayanan Srinivasan and Vani Pariyadath wordplay has been detected, the Joke Recognizer tries to detect a punch line involving this wordplay, using a trigram (an N-gram with N = 3). The KK Joke Recognizer was found to successfully detect wordplay in KK jokes, but was not successful in recognizing most punch lines in jokes. The authors themselves have suggested that a simple N-gram would not be able to ‘understand’ jokes, and that a more sophisticated tool may be needed to improve the Joke Recognizer. A trigram proves unsuccessful since it only checks two words in the punch line that follow Line 3. Also, they point out that the wordplay generator itself requires modifications since it presently generates wordplay sequences only by letter substitutions. Phonemic substitutions also need to be incorporated. Wordplay detection is obviously not equivalent to joke detection. Of course, the latter may be considered to be arising from the former. Aside from this obvious shortcoming in the model, the KK Joke Recognizer does not indicate why these jokes are funny in the first place. If mere wordplay merits humour, all individuals would perceive these jokes as equally funny. This is obviously not the case. Does the model suggest that individual differences stem only from differences in linguistic abilities? Yokogawa’s Pun Analyser uses articulation similarities to generate possible pun candidates in place of ungrammatical parts in sentences (Yokogawa, 2002). The Pun Analyser uses the similarity of sound and sound deformation rules (based on the Japanese Mora System) to identify pun candidates in ungrammatical sentences. This is carried out in four steps: morphological analysis, connection check between the analysed morphemes, similar expression generation for selecting pun candidates, and a pun candidate checker. The Japanese Pun Analyser Taylor is similar to Mazlack’s Joke Recognizer and suffers from some of the same drawbacks. The Pun Analyser does not rely on a strong theoretical framework, and is too narrow to be generalized to other subtypes of humour perception. Katz has suggested a neural account for humour, where the set-up leads to predictions being made (Katz, 1996). This results in one part of the network being activated. But the actual ending of the joke results in another part of the network being activated. When both units are active, there is a high level of overall activation, and this transient surge leads to humour perception. This theory accounts for such situations causing stimulation in the brain, but not why these situations should be perceived as being funny. One can list several situations where an inconsistency between the prediction and actuality does not lead to humour perception. GraPHIA or the Graphical Phonological Humor Identification Algorithm is a model based on graph theoretical ideas, as well as relevant findings from cognitive neuroscience for phonological joke identification (Pariyadath and Srinivasan, 2004). GraPHIA could accurately classify, given input sentences as homophonic puns or puns involving words that are spelt differently but sound the same such as, Atheism is a non-prophet organization.
(15.4)
Dissecting the Frog: Computational Approaches to Humour Perception
205
aside from normal sentences and ambiguous nonsense sentences. These results suggest that it could be considered a plausible model for perception of phonological jokes. Moreover, GraPHIA provides a way of computing quantifiable features that could be linked to the funniness of a phonological joke or RTs to a phonological joke. GraPHIA involved three stages: detection of violation, resolving this violation, and establishing two valid interpretations to the pun. These computations were theorized as being performed in the prefrontal networks associated with working memory (WM), as suggested by neuroimaging findings (Shammi and Stuss, 1999). The model was designed to improve upon the violation theory (Veatch, 1998). It should be pointed out that Veatch refers to moral violations, and GraPHIA only accounts for semantic violations. While phonological jokes were chosen for study because of their apparent minimal affect, further modifications are needed in the model to understand the role of emotions in the perception of humour. It should be noted that GraPHIA proposes novel computable features that could be correlated with funniness ratings and RTs for responding to jokes. While other models for humour perception or generation have incorporated some module for wordplay detection, GraPHIA skips such a step. The model solely focuses on possible extra processing needed for the identification of phonological jokes and not the initial language processing. So far, we have discussed the important theories and models in detail, and it is worth evaluating them based on certain norms in order to extract attributes in them that would go into making a more complete theory for humour perception. It is also important for modelling efforts to be closely linked with findings from cognitive neuroscience, and some studies relevant to humour perception from cognitive neuroscience are discussed in the next section.
COGNITIVE NEUROSCIENCE OF HUMOUR PERCEPTION Early neuroscientific work on humour focussed significantly on the relation between humour and arousal, which has been very well reviewed in McGhee (1983). The first landmark was Schachter and Wheeler’s classic study in 1962, which showed increased laughter while watching a humorous film after being injected with epinephrine. McGhee (1983) cites studies reporting increased heart rate, increased skin conductance, increased muscle tension, altered respiratory patterns, and characteristic electroencephalogram (EEG) changes while experiencing humour. Although, it is not clear how this association was separated from the act of laughter itself. The trouble with the studies discussed by McGhee (1983) is that most of them use laughter as the measure, which is not a direct measurement of perceived humour. In addition, the arousal levels may stem from other qualities of the joke text. Indeed, McGhee mentions studies that disagree over the time course of changes in arousal levels. These factors strongly suggest that arousal by itself is not sufficient for humour perception. A lot more work is definitely required in terms of event related potentials (ERPs) and neuroimaging studies on the relation between arousal and humour.
206 Narayanan Srinivasan and Vani Pariyadath Recent work has focussed on the cognitive and emotional aspects of humour perception. A significant proportion of neuropsychological studies on humour perception involve patients with right hemisphere damage (RHD), and associated impairments in humour perception or appreciation (Brownell et al., 1983; Gardner et al., 1975; Wapner et al., 1981). Recently, studies have been conducted with normals to obtain the neural correlates of humour and laughter (Goel and Dolan, 2001; Shammi and Stuss, 1999). Attempting to separate humour from laughter and vice versa can be problematic, which may be one reason for conflicting results arising from different studies over the areas involved in humour perception. It has also been difficult to establish control over confounds such as attention and facial expression (Wild et al., 2003). This section will focus primarily on cognitive neuroscience studies on humour perception and not on laughter. The right hemisphere (RH) has long been hypothesized as playing a principal role in humour perception stemming from it being purported to be the seat of emotion. One of the earliest studies to test this hypothesis did not find a significant difference between the performance of patients with right and left hemisphere damage in distinguishing between funny and non-funny cartoons (Gardner et al., 1975). On the other hand, Wapner et al. (1981) found that the non-dominant hemisphere plays a vital role in humour perception. Again, it was reported that RHD patients could detect the element of surprise in the punch line of a verbal joke, but could not determine the endings that could also resolve the incongruity in the joke (Brownell et al., 1983). Shammi and Stuss (1999) conducted a study on patients with right frontal lesions. They found that even when humorous stimuli were occasionally quantitatively rated by these patients as being funny, and understanding of these items appeared to be reasonably adequate based on their explanations, they did not respond to these items with smiles or laughter, unlike normal participants and those with lesions in other brain regions. That is, while they may have grasped the cognitive basis of humour, they did not affectively respond to the humorous stimuli (Shammi and Stuss, 1999). Neuropsychological studies on humour perception are not restricted to RHD patients. One study with adolescents with high-functioning autism or Asperger’s syndome found that they performed poorly on a task requiring them to choose the correct last frame given a series of two cartoons (Emerich et al., 2003). The autistic patients were found to choose the straightforward ending to the text most frequently. The patients were also found to choose the humorous non-sequitur endings or slapstick more often for a given joke text, although this finding was not significant. While the authors of the study considered both these error types to be consistent with impairments in cognitive flexibility associated with autism, why then were the error types different on both tasks? Intuitively, one would expect the patients to either choose the non-sequitur ending on both tasks, or choose only the straightforward ending. Problems such as these could have risen from the small number of subjects (n = 8). The authors themselves have acknowledged that a great deal of variability exists among adolescents with autism. Parkinson’s syndrome has also been associated with humour impairments (Benke et al., 1998), but this impairment is more
Dissecting the Frog: Computational Approaches to Humour Perception
207
of a problem involving spontaneous emotional expressions. Patients apparently are able to detect the surprise element in humorous sketches. Recently, several neuroimaging studies have come up which investigate the brain areas activated during humour perception. One such study, which gained prominence, indicated that areas in the left hemisphere (LH) responsible for speech production were activated by phonological jokes, while a bilateral temporal lobe network was activated by semantic jokes (Goel and Dolan, 2001). Activation was also found in the medial ventral prefrontal cortex independent of the above-mentioned networks, regardless of the modality, which was attributed as processing the affective component. Another recent functional magnetic resonance imaging (fMRI) study showed that humour activates several cortical and subcortical regions, including the mesolimbic reward centres, and that the degree of humour intensity could be positively correlated with the BOLD signals in these areas (Mobbs et al., 2003). The study involved an event-related design where subjects were required to categorize cartoons as funny and non-funny, and was the first to report activations in the mesolimbic areas. One view is that this mesolimbic activation could be considered an index of humour (Berns, 2004). Berns maintains that this finding could provide the ‘neurobiological bridge to psychological theories of humour’, since the mesolimbic dopamine system plays an important role in incongruity detection (Schultz, 1997).
A CRITICAL EVALUATION Before we discuss the pros and cons of each theory in detail, it is worthwhile to first establish norms for comparing these theories. Any theory that claims to explain humour perception must account for psychological phenomena as well as findings from cognitive neuroscience. As already mentioned, humour is a generic term and refers to a large number of instances such as puns, anecdotal jokes, slapstick, sarcasm and caricature. Aside from these, infants also tend to find the game of peek-a-boo immensely amusing. While a single theory may be impossible to explain humour perception (Veatch, 1998) or indeed unnecessary, a theory must at least be suggestive of why these seemingly different instances bring about a common response. While most humorous instances are perceived as funny only on the first occasion, some continue to be appreciated on repeated narrations. A good theory must account for this selective occurrence of repeated humour perception. A large number of humorous instances deal with objects of prohibition such as sex and aggression. No theory is complete without explaining why such taboo subjects merit jokes in greater numbers. Also, a universal fact of humour is that humorous instances are short pieces. What role does this brevity play in perceiving humour? Although not directly related, it is worth investigating why laughter is primarily a social phenomenon. What implications do this have for the role of humour perception? The various theories and models were compared keeping these aspects in mind.
208 Narayanan Srinivasan and Vani Pariyadath The computational models that have been developed so far for humour perception throw little light on the cognitive processes involved in perceiving humour. Most of the theories themselves are not quantifiable, and the need of the day is for the convergence of computational modelling and empirical data for the development of a complete theory of humour. Table 15.1 shows critical evaluations of various theories and models for humour perception and generation. As mentioned before, the theories are evaluated based on whether they apply to all subtypes and whether they discuss the role of prohibition, affect, context and knowledge in humour perception. The table also indicates whether the theory explains why certain humorous instances are repeatedly funny. The most important criterion is how quantitative the theory or model is ‘Y’ (for Yes) and ‘N’ (for No) denote whether or not the theory accounts for a phenomenon in a plausible manner. If the theory does not address a particular issue at all, it is indicated by a ‘–’. TABLE 15.1
A comparative study of theories of humour perception
Issue
Applies to all subtypes
Effect of prohibition
Role of affect
Role of context and knowledge
Selective occurrence of repeated humour perception
Quantitative
Superiority Theory
N
–
Y
–
N
N
Relief Theory
N
Y
Y
–
–
N
Incongruity Resolution Theory
N
–
N
–
N
N
Violation Theory
Y
–
Y
–
–
N
Semantic Script Theory
N
N
N
N
N
N
KK Joke Analyser
N
N
N
Y
N
Y
Yokogawa’s model
N
N
N
Y
N
Y
Katz’s model
Y
N
N
–
–
N
Notes: Y = Yes N = No – = Not Applicable
Dissecting the Frog: Computational Approaches to Humour Perception
209
As is evident from Table 15.1, almost all theories for humour are not applicable to all types of humour. For example, the superiority theory cannot explain why phonological jokes are funny. Of course, one can naturally question the need for such a unified theory of humour perception. A good theory nevertheless should attempt to address a particular class of humour, such as semantic or phonological. The GTVH, in spite of its ambitious moniker, does not explain the mechanisms behind any single type of humour. A common observation is that most humorous texts deal with topics of prohibition such as sex and violence, which is another issue that a complete theory should take up. In this respect, the relief theory does attempt an explanation for the production of taboo topics in humour perception. Freud and Minsky have pointed out that a large portion of jokes is aimed at sex and aggression, and to some extent stupidity (Freud, 1905; Minsky, 1980). The cognitive censors are apparently more selective in this regard. While Freud suggested that the prohibition of these entities makes them more ‘funny’, there seems no obvious reason for this. Does this imply that humour perception lies at a decision stage rather than in sensitivity? As far as the role of affect is concerned, the relief and superiority theories do tackle the issue, as does the violation theory to some extent. Relief appears to be a sequence of positive emotions following negative ones. Veatch speaks of scenarios where the joke is offensive due to excessive violation. The term violation itself carries a strong emotional connotation as opposed to the word incongruity. The IR and linguistic theories do not specifically address the part of emotions in humour perception. All the theories run into trouble when dealing with the issue of context and with the selective occurrence of repeated humour perception. While some theories like the Minsky’s frame theory can explain why the same joke does not appear funny on a second hearing, but not why some jokes seem funny on multiple hearings. We come finally to the most important question: Is the theory quantitative? This factor plays the biggest role in making a theory a complete one, or at least a plausible one. Sadly, most of the existing theories suffer deeply in this regard, possibly one reason why computational modelling has proved of little use in humour research. Paulos’ model based on catastrophe theory does not furnish the actual computations to be made or a mode of making predictions given a joke. Unlike the usual non-linear dynamics models, this model does not offer any specific differential equations to model the processes involved in perceiving humour. In this sense, it suffers like other theories in not being quantitative enough to be tested. GraPHIA on the other hand is a quantitative model. It is an attempt in the direction of computational modelling for better cognitive theories for humour perception. Given the numerous inadequacies of computational models for humour perception, it is important that these models rely on behavioural data and findings from cognitive neuroscience, which have especially focussed on humour perception as opposed to humour generation. Having perused the literature on theories and models for humour perception and the cognitive neuroscience of humour perception, we shall now attempt to bridge the gap
210 Narayanan Srinivasan and Vani Pariyadath between the two approaches. Katz’s model considers humour a consequence of high levels of activation, the activation arising from simultaneous predictions and reality opposing these predictions. In other words, humour is strongly coupled with arousal levels. While this goes well with findings from arousal studies discussed earlier, it still does not form a complete model. High levels of activation or arousal are found in other scenarios, too, and are certainly not sufficient to perceive humour. The other position dating as far back as Darwin (1872) has been that a sudden drop in stimulation is responsible for laughter and therefore humour perception. What still works in Katz’s favour are studies which associate laughter with the function of general arousal-reduction (McGhee, 1983). So, humour perception could still be related to increase in the level of arousal and not the drop. Ramachandran has attempted to furnish neuropsychological evidence for the false alarm theory (Ramachandran, 1998). He cites patients with pain asymbolia, a condition in which patients with damage to the insular cortex do not feel pain. These patients are also sometimes found to laugh in response to pinpricks. The insular cortex receives information about pain from the skin and viscera and passes this on to the cingulate gyrus and other parts of the limbic system. Ramachandran suggests that a disconnection between these two centres caused by a lesion could result in the insula detecting a potential threat, but the cingulate gyrus informing that there is no threat. One implication is that some other areas compare the information from these two areas and of course decide that there is no threat. It is not clear how this happens and which area is involved in this decision. It is also not clear that this explains verbal or other types of humour. A closer look at some of the recent studies on cognitive neuroscience of humour point out to the evaluation of the Suls’ two-stage model for humour. Unfortunately, the findings contradict each other on this aspect. Shammi and Stuss (1999) interpret their findings as support to the model since the frontal lobes, which was identified as the key player in humour appreciation in their study, is associated with problem solving and Suls’ model treats humour detection as a case of problem solving. However, Coulson and Kutas (2001) report that no direct mapping could be made between their findings in an event related potential (ERP) study, and the surprise and coherence stages of Suls’ model. One difficulty in testing this model may be that there is no clear behavioural marker of the transition between these two stages (Moran et al., 2004). GraPHIA describes the so-called cognitive component of phonological joke perception. The feature value computations that are carried out by GraPHIA are assumed to be done in WM, which is associated with the prefrontal cortex (Courtney et al., 1998). Also, this model mentions an interaction between the semantic and phonological networks, which again conforms to neuropsychological findings. Having examined the current state of the art, we consider possible directions for future research. While so far, models and theories for humour perception have been built based mostly based on experimental or (more commonly) experiential evidence, the time has come to instead focus on evidence from neuropsychological studies as well. Presently,
Dissecting the Frog: Computational Approaches to Humour Perception
211
computational models for humour perception as well as humour generation seem to be intended solely for the machine interface. A change in approach needs to be brought about here too. For example, none of the existing models incorporates the time course of humour perception in them, but no doubt timing is an important aspect of perceiving and producing humour. A similar paradigm such as that used by Vaid et al. (2003) could be attempted using fMRI if possible. As far as cognitive neuroscience is concerned, we need to move on from localizing the areas involved in humour perception to exploring the plausible mechanisms involved in the act of perceiving humour.
CONCLUSIONS From our scrutiny of the existing models for humour perception, it is evident that a great deal of work is required before any major advances can be made in the field of humour perception. While some models are plausible cognitive models, others are evidently in a primitive state to explain or generate testable findings using behavioural experiments or neuroimaging paradigms. Efforts like GRAPHIA are initial attempts to bridge that gap and it is to be hoped that further modelling efforts and interactive research from multiple disciplines including psychology and neuroscience will enable us to understand such a fascinating phenomenon like humour perception.
REFERENCES Attardo, S. and V. Raskin. 1991. ‘Script Theory Revis(it)ed: Joke Similarity and Joke Representation Model’. Humor: International Journal of Humor Research, 4: 293–347. Benke, T., S. Bösch and B. Andree. 1998. A Study of Emotional Processing in Parkinson’s Disease. Brain & Cognition, 38(1): 36–52. Berns, G. S. 2004. Something Funny Happened to Rewards. Trends in Cognitive Sciences, 8: 193–94. Binsted, K. 1996. Machine Humor: An Implemented Model of Puns, Doctoral Dissertation, Edinburgh, Scotland: University of Edinburgh. Binsted, K. and G. Ritchie. 1994. An Implemented Model of Punning Riddles. Proceedings of the Twelfth National Conference on Artificial Intelligence (Vol. 1) (pp. 633–38), Seattle, Washington, United States. Binsted, K. and T. Takizawa. 1998. ‘BOKE: A Japanese Punning Riddle Generator’. Journal of Japanese Society for Artificial Intelligence, 13: 920–27. Brownell, H.H., D. Michel, J. Powelson and H. Gardner. 1983. ‘Surprise but not Coherence: Sensitivity to Verbal Humor in Right Hemisphere Patients’. Brain and Language, 18: 20–27. Coulson, S. and M. Kutas. 2001. ‘Getting it: Human Event-Related Brain Response to Jokes in Good and Poor Comprehenders’. Neuroscience Letters, 316: 71–74. Courtney, S.M., L. Petit, J.H. Haxby, L.G. Ungerleider. 1998. ‘The Role of Prefrontal Cortex in Working Memory: Examining the Contents of Consciousness’. Philosophical Transactions of the Royal Society, B 353: 1819–28.
212 Narayanan Srinivasan and Vani Pariyadath Darwin, C. 1872. Chapter 8: Joy, High Spirits, Love, Tender Feelings, Devotion. The Expression of the Emotions in Man and Animals (pp. 196–219). New York: D. Appleton & Company. Emerich, D.M., N.A. Creaghead, S.M. Grether, D. Murray and C. Grasha. 2003. ‘The Comprehension of Humorous Materials by Adolescents with High-Functioning Autism and Asperger’s Syndrome’. Journal of Autism and Developmental Disorders, 23: 253–57. Freud, S. 1905. Jokes and Their Relation to the Unconscious. Vintage, UK Random House. Gardner, H., P.K. Ling, L. Flamm and J. Silverman. 1975. ‘Comprehension and Appreciation of Humorous Material following Brain Damage’. Brain, 98: 399–412. Goel, V. and R.J. Dolan. 2001. ‘The Functional Anatomy of Humor: Segregating Cognitive and Affective Components’. Nature Neuroscience, 4: 237–38. Harris, C.R. and N. Alvarado. 2005. ‘Facial Expressions, Smile Types, and Self-Report During Humor, Tickle, and Pain’. Cognition and Emotion, 19(5): 655–69. Katz, B. 1996. A Neural Invariant of Humour, Proceedings of the International Workshop on Computational Humor. Twente Workshops on Language Technology, University of Twente, The Netherlands. Loehr, D. 1996. An Integration of a Pun Generator with a Natural Language Robot. Proceedings of the International Workshop on Computational Humor, Enschede, University of Twente, The Netherlands. McDonough, C.J. 2001. ‘Mnemonic String Generator: Software to Aid Memory of Random Passwords’. Technical Report. CERIAS, Purdue University. McGhee, P.E. 1983. ‘The Role of Arousal and Hemispheric Lateralization in Humor’. In P.E. McGhee and J.E. Goldstein (eds), Handbook of Humor Research, Vol. 1 (pp. 13–37). New York: Springer. Minsky, M. 1980. Jokes and the Logic of the Cognitive Unconscious. Technical report, Massachusetts Institute of Technology, Artificial Intelligence Laboratory. Mobbs, D., M.D. Greicius, E. Abdel-Azim, V. Menon and A.L. Reiss. 2003. ‘Humor Modulates the Mesolimbic Reward Centers’. Neuron, 40: 1041–48. Moran, J.M., G.S. Wig, R.B. Adams, P. Janata, Jr and W.M. Kelley. 2004. ‘Neural Correlates of Humor Detection and Appreciation’. NeuroImage, 3: 1055–60. Mulder, M.P. and A.A. Nijholt. 2002. Humour Research: State of the Art, Technical Report. No. 02–34, University of Twente, Netherlands. Pariyadath, V. and N. Srinivasan. 2004. ‘A Computational Model for Identifying Phonological Jokes. Proceedings of the International Conference on Cognitive Science, Allahabad, pp. 247–51. Paulos, J. 1980. Mathematics and Humour. Chicago: University of Chicago Press. Provine, R.R. 1996. Laughter. American Scientist, New York. Available from http://www.sigmaxi.org/ amsci/Articles/96articles/Provine-R.html. Pun of the Day. Available from http://www.punoftheday.com. Ramachandran, V.S. 1998. ‘Consciousness and Body Image: Lessons from Phantom Limbs, Capgras Syndrome and Pain Asymbolia’. Philosophical Transactions of Royal Society, London, B 353: 1851–59. Ramachandran, V.S. and S. Blakeslee. 1999. Phantoms in the Brain. London: Fourth Estate. Ritchie, G. 1998. Prospects for Computational Humour, Proceedings of 7th IEEE International Workshop on Robot and Human Communication, Japan: Takamatsu, pp. 283–91. ———. 1999. Developing the Inconguity-Resolution Theory, Proceedings of the AISB ‘99 Symposium on Creative Language: Humor and Stories (pp. 78–85), Edinburgh. ———. 2001. ‘Current Directions in Computational Humour’. Artificial Intelligence Review, 16(2): 119–35. Ruch, W. 2001. ‘The Perception of Humor’. In A.W. Kaszniak (ed.), Emotion, Qualia, and Consciousness (pp. 410–25). Tokyo: Word Scientific Publisher.
Dissecting the Frog: Computational Approaches to Humour Perception
213
Ruch, W., S. Attardo and V. Raskin. 1993. ‘Towards an Empirical Verification of the General Theory of Verbal Humor’. Humor: International Journal of Humor Research, 6: 123–36. Schultz, W., P. Dayan and P.R. Montague. 1997. ‘A Neural Substrate of Prediction and Reward’. Science, 275: 1593–99. Shammi, P. and D. Stuss. 1999. ‘Humor Appreciation: A Role of the Right Frontal Lobe’. Brain, 122(4): 657–66. Stock, O. 2003. ‘Password Swordfish: Verbal Humour in the Interface’. Humor: International Journal of Humor Research, 16: 281–95. Stock, O. and C. Strapparava. 2003. ‘Humorous Agent for Humorous Acronyms: The HAHAcronym Project’. Humor: International Journal of Humor Research, 16(3): 297–314. Suls, J. 1983. ‘Cognitive Processes in Humor Appreciation’. In McGhee, P.E. and Goldstein, J.E. (eds), Handbook of Humor Research Vol. 1 (pp. 39–57). New York: Springer. Taylor, J. and L.J. Mazlack. 2004. Computationally Recognizing Wordplay in Jokes, Proceedings of the Annual Meeting of the Cognitive Science Society (pp. 1315–20), Chicago. Vaid, J., R. Hull, R. Heredia, D. Gerkens and F. Martinez. 2003. ‘Getting a Joke: The Time Course of Meaning Activation in Verbal Humor’. Journal of Pragmatics, 35: 1431–49. Veatch, T.C. 1998. ‘A Theory of Humor’. Humor: International Journal of Humor Research, 11: 161–215. Visuvalingam, S. 1983. ‘Abhinavagupta’s Conception of Humor (Hasya): The Aesthetics and Semiotics of the Indian Clown’. Available from http://www.svabhinava.org/HumorPhd/ default.htm. Wapner, W., S. Hamby and H. Gardner. 1981. ‘The Role of the Right Hemisphere in the Apprehension of Complex Linguistic Materials’. Brain and Language, 14: 15–33. Wild, B., F.A. Rodden, W. Grodd and W. Ruch. 2003. ‘Neural Correlates of Laughter and Humor’. Brain, 126: 2121–38. Yokogawa, T. 2002. Japanese Pun Analyser Using Pun Articulation Similarities. Proceedings of the IEEE International Conference on Fuzzy Systems (pp. 1114–19), Honolulu.
SECTION
IV
Culture and Cognition
C
ulture, often invisible and unnoticed, influences various cognitive, affective as well as cognative processes. Vygotsky’s (1978), through his original work of the 1930s, convincingly asserted that cognitive development is shaped by the cultural context of individuals. As discussed by Berry and Dasen (1974), historically, in the study of culture, behavioural scientists followed two traditions called ethos and eidos. With a longer history (since the 1930s), the ethos tradition represents the study of socio-emotional aspects and, more broadly, the ‘Culture and Personality’ research approach. The eidos research tradition, beginning in the 1960s, covered cognitive domains like sensation, perception, attention and problem solving. There are, however, some examples of the century-old, seminal, cross-cultural work widely known as the Cambridge anthropological expedition, demonstrating the influence of culture on cognition, covering a sample of natives of Torres Strait Islands and Toda of Southern India (Rivers, 1901, 1905). Both cross-cultural and cultural approaches have extensively studied various aspects of cognition, and demonstrated how culture and environment shape, mould and transform cognitive processes (Berry et al., 1997; Triandis and Lonner, 1980). The first chapter in this section by Poortinga is on the methodological issues related to culture and cognition. Poortinga critically discusses research designs and their validity for cross-cultural research on cognition. The next four chapters discuss large-scale studies done across continents in various countries, focussing on specific aspects of the development of spatial cognition in various cultures. They have reported investigation on the three frames of reference, namely, intrinsic, egocentric and geocentric used to describe space. Poortinga has analysed various comparative (cross-cultural) methodological issues related to culture-and-cognition research. The debate related to universalism versus relativism is discussed and the methodological aspects (for example, equivalence, generalization, validity) of universalism elaborated to understand the common terms of cultural variations and human psychological functioning. Poortinga suggests that cognitive scientists should ‘lean towards quasi-experimental designs and low level inferences in research involving cross-cultural variation’. The chapter by Mishra and Dasen explains the general framework and methodology used in these studies conducted in various cultural settings. In a revived form of the linguistic relativity hypothesis, Levinson (2003) and other researchers in comparative cognitive linguistics have demonstrated that languages may use one or more of the three frames of reference (FoR) to describe space: intrinsic, egocentric and geocentric. The dominant choice of frame in the language is shown to determine the non-linguistic modes of spatial encoding. European languages all use an egocentric FoR, while Balinese, Newari, Nepali and Hindi are among the languages where a geocentric frame is predominant. The study examines how the choice of linguistic FoR and spatial encoding are linked to ecological, cultural, social and neuropsychological features. It takes the perspective of cross-cultural human development (Dasen, 2003; Dasen and Mishra, 2000). The results of previous studies by our team in Bali, Nepal and India are summarized. They show that the particular
218 Advances in Cognitive Science form that the geocentric FoR takes (and the language to express it) is linked to spatial orientation systems, themselves congruent with ecology and cosmology. Our research assessed the impact of rural and urban residence and that of schooling. This introduction to the symposium also presents the methodology that is common to all the studies, in particular the tasks used for language elicitation and for assessing spatial encoding. Vajpayee, Dasen and Mishra discuss the differences in spatial encoding among Sanskrit school children compared with Hindi school children. In this part of the larger study, we compare 155 students attending Sanskrit schools (125 boys and 30 girls) with 221 pupils (172 girls and 49 boys) of a Hindi-medium school. It was expected that the socialization into a Sanskrit cosmology would foster a better knowledge of the spatial orientation system, the use of geocentric spatial language and geocentric spatial encoding. This hypothesis is fully confirmed. In addition, the Sanskrit school children were also more correct in their knowledge of right, left, front and back (RLFB), and they were more field-independent. Some gender differences existed, but were not very important; for example, Sanskrit school boys use slightly more geocentric encoding than girls. Dasen and Wassmann, in their chapter, discuss spatial language and encoding in Bali and Geneva. The study shows differences in the reference system with the geocentric system being more prevalent in Bali, and the egocentric system being more prevalent in Geneva. The chapter presents some comparative aspects of the ongoing study, in particular some results from a replication study in Bali, which showed a strong impact of socio-economic and cultural variables (that can be subsumed under the concept of acculturation). Results from a study in Geneva with monolingual and bilingual children, demonstrate the consistent use of egocentric language and encoding in French and other European languages. The chapter by Mishra and Dasen discusses relationships between spatial encoding and psychological differentiation. The chapter points to the presence of hemispheric differences related to geocentric spatial encoding. A study was conducted in Varanasi with 376 boys and girls, aged between 10 to 15 years, attending Hindi-medium or Sanskrit schools. In addition to language elicitation and spatial encoding tasks, Block Design and Story–Pictorial Embedded Figure Test (SPEFT) were used as measures of psychological differentiation (field-dependence/independence). A structural link is found between these three areas of psychological functioning even when controlling for age, gender and schooling. A further study of a sub-sample of 80 children using neuro-psychological indicators of brain lateralization revealed geocentric encoding to be linked to indicators of central brain lateralization, but not to peripheral measures (hand, foot, eye and ear dominance). The chapter by Mishra and Berry presents the results from their study examining the development of cognitive processes in the context of cultural adaptations of tribal children in Chotanagpur. This chapter reports an empirical study that examines the development of cognitive processes in the context of cultural adaptations of individuals. Children (N = 400, age 9–12 years) representing ‘hunting gathering’, ‘rudimentary
Culture and Cognition
219
agriculture’, ‘irrigation agriculture’ and ‘wage earning’ groups were studied in Chotanagpur region of undivided Bihar (now Jharkhand). Cultural dimensions of ‘societal size’ and ‘social conformity’ were assessed with the help of rating scales, both at the group and the individual levels. Cognitive dimensions of ‘intraunit distinctiveness’ and ‘extraunit connectedness’ were assessed with the help of several tests that attempted to measure differentiation and contextualization in cognition. Findings revealed the existence of ‘societal size’ and ‘social conformity’ as two independent dimensions, which showed considerable variation according to economic activities of the groups. On the other hand, differentiation and contextualization did not emerge as the neat cognitive dimensions that had been emphasized in previous research. The expectation about the relationship between subsistence strategies and cognitive dimensions was fulfilled only for differentiation. A ‘task-specific’ relationship of subsistence strategies with contextualization in cognition demands more research in this area. The chapters so far on culture and cognition raised some of the fundamental issues related to the universality of cognitive processes, and the differences in behaviour across cultures. The chapter by Berry provides an overarching framework to understand the findings from cross-cultural studies on cognition. John W. Berry presents a general eco-cultural framework which, according to him, is both context-sensitive and comparative with a scope of an evaluation of the universalist assumption. The eco-cultural approach is outlined to understand the developmental process as well as performance of cognitive competence.
REFERENCES Berry, J.W. and P.R. Dasen. 1974. Culture and Cognition: Readings in Cross-Cultural Psychology. London: Methuen. Berry, J.W., P.R. Dasen and T.S. Saraswathi. 1997. Handbook of Cross-cultural Psychology: Basic-processes and Human Development. Boston: Allyn and Bacan Inc. Dasen, P. R. 2003. ‘Theoretical Frameworks in Cross-cultural Developmental Psychology: An Attempt at Integration’. In T. S. Saraswathi (ed.), Cross-cultural Perspectives in Human Development: Theory, Research, and Applications (pp. 128–65). New Delhi: Sage India. Dasen, P. R., and Mishra, R. C. 2000. Cross-cultural Views on Human Development in the Third Millennium. International Journal of Behavioral Development, 24: 428–34. Reprinted in R. K. Silbereisen and W. Hartup (eds), Growing Points in Developmental Science (pp. 266–86). London: Psychology Press, 2002. Levinson, S. 2003. Space in Language and Cognition. Cambridge: Cambridge University Press. Rivers, W.H.R. 1901. ‘Introduction to Vision’. In A.C. Haddon (ed.), Reports of the Cambridge Anthropological Expedition to the Torres Straits, Vol. II, Part I. Cambridge: The University Press. ———. 1905. ‘Observation on the Senses of the Todas’. British Journal of Psychology, 1: 321–96. Triandis, H.C. and W. Lonner. 1980. Handbook of Cross-cultural Psychology: Basic Processes, Vol. 3. Boston: Allyn and Bacan Inc. Vygotsky, L.S. 1978. Mind and Society: The Development of Higher Psychological Process. Cambridge: Harvard University Press.
Chapter 16 Sources of Evidence and Levels of Interpretation in Culture-andCognition Research Ype H. Poortinga
INTRODUCTION
C
ross-cultural research, both the cultural and culture-comparative variety, are mainly about differences in behaviour between groups of people; far less attention is given to the question of how much humans all over the world behave in similar ways (Brouwers et al., 2004). The domain of cognition is no exception. There is a substantial body of research geared towards finding and subsequently explaining differences in cognitive performance. Many explanations are in terms of broad differences in cognitive functioning. They tend to be of two kinds. The first refers to internal psychological dispositions, either biologically primed or culturally induced. The second emphasizes general cognitive consequences of external antecedents, such as little and low quality school education, or adverse social and economic conditions. A smaller part of the literature is on research designed to analyse narrowly described relationships between cultural antecedents and behavioural consequents. Some of these relationships pertain to a well defined aspect of cognitive behaviour, others may cover a wider array, but it is always more or less clear which behaviours are included and which not. I will first discuss four levels of equivalence of cross-cultural data in combination with three levels of interpretations reaching from low-inclusive to high-inclusive generalizations. Thereafter, I will present a scheme that is based on the same three levels of generalization, but now considering different methods of design and data collection. The analysis leads to the paradoxical situation that inclusive interpretations of the kind that most researchers of culture and behaviour prefer for explaining cross-cultural differences escape virtually all empirical control on validity. I end with some conclusions about crosscultural psychology’s contribution to cognitive science.
222 Ype H. Poortinga EQUIVALENCE OF DATA AS A CONDITION FOR CROSS-CULTURAL COMPARISON Much of the early interest of psychologists in cognitive performance of other people had strong racial overtones. Within decades after the invention of the intelligence test data were collected with a view to compare the IQ of culturally and ethnically defined groups, mostly the latter (Porteus, 1937; Shuey, 1966). The results confirmed the popular stereotypes of psychologists of that time: peoples of European descent were consistently found to be smarter than other groups. Soon a reaction started, both in USA (for example, Klineberg, 1935) and in Africa (Biesheuvel, 1943). The arguments raised by Biesheuvel have basically remained the same in more recent literature with an equalitarian perspective. The single most important point is that measures of intelligence, or for that matter of other aspects of cognition, are influenced by cultural context in ways that go well beyond what these measures are supposed to assess. Views on the implications of the context boundedness of cognitive measures vary greatly. Some authors argue that cognitive processes basically are the same in humans of any cultural group; cross-cultural differences are mainly in the manifestation of these processes (Berry et al., 2002). For example, major differences may have to do with the readiness with which certain cognitive algorithms are solicited. Other authors go beyond this and argue that cognitive functioning as such is inherently cultural (Cole, 1996; Rogoff, 2003; Vygotsky, 1978). These two positions of universalism and relativism can be defined as a dichotomy, reflecting two incommensurable paradigms. They can also be seen as clusters of opinions that vary along a dimension from extreme relativism to extreme universalism (the latter position has been labelled as ‘absolutism’ by Berry et al., 2002).
THE LOGIC OF COMPARISON The central concern of relativism is to understand people ‘in their own terms’. This has various methodological implications. In particular, the cross-cultural use of the same methods and instruments across cultures is questioned. Instruments are developed within a certain context; their use elsewhere amounts to an imposition of extraneous cultural elements. Culture-comparative studies are often avoided (though not by all, cf. Fiske et al., 1998) because a valid comparison is said to be questionable or even meaningless. Psychological assessment should use instruments and procedures that have been designed within, and are suitable for, the cultural population in which they are to be used. Two points may be noted. First, a global statement to the effect that two entities of any complexity, such as biological organisms or social communities, are different is a truism. It is uninformative in empirical science. I can expect agreement when I say that all humans differ from each other, but also when I say that all humans are the same. Second,
Sources of Evidence and Levels of Interpretation
223
a more precise statement indicating in which respect two entities differ requires at least some implicit standard in terms of which a distinction can be made between sameness and difference. There are two possibilities, the relativistic position that behaviour across cultures is incomparable being either an axiomatic viewpoint that is unassailable, or an empirical position that should be open to empirical scrutiny. In this chapter, I will take the latter viewpoint, in this way providing a bridge to universalism. The central concern of universalism is to understand in common terms the cultural variations in human psychological functioning. For example, school children in China obtain better scores on tests of arithmetic than in USA. In culture-comparative research that starts from a presumption of basic human psychological identity (psychic unity), the main goal is to identify antecedent factors that can explain such a difference. In fact, for the difference mentioned, one important factor has been identified repeatedly. In Chinese, words for numerals are shorter than in English, and this facilitates cognitive processing (Hoosain and Salili, 1987; Stigler et al., 1986). The main methodological problem of universalism is that some measurements, such as arithmetic test scores, apparently provide a reasonably invariable standard of comparison, at least across some cultural populations, while other measurements do not, such as assessments of intellectual capacity. Since the 1960s, the question how to differentiate between valid and invalid comparisons has led to the development of psychometric and data analytic distinctions, some of which I will briefly review.
LEVELS OF EQUIVALENCE Inequivalence or incomparability refers to unequal representation across cultural groups in a measure of a domain of behaviour or a hypothetical trait.1 Inequivalence can occur for various reasons. For example, a trait may be absent in a cultural population (for example, abstract reasoning was seen by Luria [1971, 1976] as a faculty absent in some cultures). It is also possible that the trait is present, but that the measurement procedure does not address it (for example, syllogistic reasoning items do not always solicit formal logical reasoning [Scribner, 1979]). A third possibility is that there is some factor not related to the trait targeted by a measure, which leads to systematic distortion or bias in the score level of a cultural group (for example, in an educational system with emphasis on accuracy, a speeded cognitive test may suppress score levels [Van de Vijver and Tanzer, 1997]). Finally, there may be incidental difficulties with specific items of a measure (for example, an item has an ambivalent meaning when translated [Ellis, 1989; Hambleton, 1994]). 1 The terms ‘domain’ and ‘trait’ are used in this chapter rather indiscriminately. Domain refers more to a perspective in which there is emphasis on situations that solicit similar reactions or presumably belong together for some other reason. Trait refers more to a presumed characteristic of the person that leads to similar reactions across a cluster of situations.
224 Ype H. Poortinga Such difficulties can be dealt with more easily when different levels of equivalence are distinguished. A first level addresses the question as to whether a domain of trait makes sense in all of the cultural populations to be compared in a study, and is referred to as conceptual equivalence (Fontaine, 2005; Poortinga, 1989). For example, the notion of arithmetical skill may be said to not apply in isolated illiterate groups, if the domain of arithmetic is defined in terms of decimal numbers and arithmetical operations. Conceptual equivalence can be said to be either assumed or rejected a priori. If rejected, there is no point in cross-cultural comparison; if accepted, the onus is on the researcher to supply empirical evidence on the assumption. Three further levels of equivalence have been distinguished by Van de Vijver and Leung (1997). These distinctions can be illustrated with reference to measurements of temperature. A direct comparison of temperature does not make sense if some of the measurements have been made on a Celsius scale and others on a Fahrenheit scale, or if a Celsius and a Kelvin scale have been used. The three hirerarchical levels have been defined as follows (cf. Berry et al., 2002): 1. Structural or functional equivalence implies that the same trait or domain is measured cross-culturally, but not necessarily on the same quantitative scale (cf. measurements in Fahrenheit and Celsius). 2. Metric or measurement unit equivalence implies that a difference between two scores has the same meaning, independent of the culture in which it was found (cf. measurements in Celsius and Kelvin, where the distance in temperature between two values is the same on each scale, but the zero-point differs). 3. Scale equivalence or full score comparability implies that scores of a given value have in all respects the same meaning cross-culturally, and can be interpreted in the same way (cf. for example, measurements in Celsius and Celsius). The relevance of these distinctions derives from statistically testable conditions for each level that presumably are satisfied by equivalent scores, but not by inequivalent or culturally biased scores. Structural equivalence and conceptual equivalence are mainly examined by means of correlational techniques like factor analysis. Statistical procedures are available to estimate the degree of factorial similarity in data sets collected in different societies. Most familiar is a congruence coefficient called Tucker’s phi (Tucker, 1951; Van de Vijver and Leung, 1997). If factor structures of the items in an instrument are similar across cultures, this is an indication that traits or domains correspond measured; if the factor structures show substantial differences, meaningful comparison is ruled out. Metric equivalence can be examined, for example, by means of an analysis of variance; the interaction between items and cultures (or in the case of a repeated measurements design, the interaction between measurement occasions and culture) should not deviate significantly from zero. Full score equivalence can be examined by means of an analysis of covariance structures applied to the items in a test (Holland and Wainer, 1993; Van de Vijver and Leung, 1997).
Sources of Evidence and Levels of Interpretation
225
LEVELS OF GENERALIZATION Lack of equivalence is often seen as reflecting differences in the scaling properties of the test score variables in different cultural groups, that is, as a property of the instrument. However, equivalence should be analysed in function of the inferences that a researcher wants to derive from the data, rather than as a characteristic of an instrument (Poortinga and Malpass, 1986). As already mentioned, if between-group differences in score distributions are found for an arithmetic test, these are likely to reflect valid differences in arithmetical achievements. Inferences to the effect that respondents in the two groups differ in arithmetical aptitude (that is, the ability to learn arithmetic) would be questionable, and very few psychologists nowadays would see the finding as support for group differences in inborn intellectual capacity. These examples illustrate how interpretations of results can be said to amount to generalizations from data to domains of behaviour. This idea forms the basis of Generalizability Theory developed by Cronbach and colleagues (Cronbach et al., 1973). According to this theory, a measurement forms a sample from the entire set of possible behaviours that might have been included in that measurement. Or, in other words, a measurement is of interest insofar as it represents the set of behaviours to which it is being generalized. Cronbach and colleagues call this set a universe, which corresponds quite closely to the terms ‘domain of behaviour’ and ‘trait’ used here. Poortinga and Malpass (1986; see also Berry et al., 2002; Poortinga, 2003) made a distinction between three levels of generalization that differ in inclusiveness or broadness (see Table 16.1). TABLE 16.1 An overview of four levels of psychometric equivalence of data and three levels of inclusiveness of interpretations Low level inferences (measures are samples from domain)
Medium level inferences (measures contain core elements)
High level inferences (unconstrained domain)
Conceptual equivalence
assumed (or denied) by definition
assumed (or denied) by definition
assumed (or denied) by definition
Structural equivalence
varies across domains and cultures
empirical evidence mostly positive
beyond empirical analysis?
Metric equivalence
varies across domains and cultures
empirical evidence limited
beyond empirical analysis
Full score equivalence
varies across domains and cultures
beyond empirical analysis?
beyond empirical analysis
Source: Adapted from Poortinga and Malpass, 1986.
226 Ype H. Poortinga First, generalization can be to a domain of which all the elements can be listed, or for which at least it can be decided unambiguously whether or not a given element belongs to it. In principle, a (random) sample of elements can form the basis for constructing a scale representing such a domain. A test of arithmetic with items on operations, such as addition and subtraction, is an obvious example. With slightly less well-defined domains a certain error may be introduced, but cross-cultural comparison of data at face value can still be quite meaningful. This would seem to have been the case in a study by Willemsen and Van de Vijver (1997) of parental beliefs on mastery of age-related tasks with samples of mothers from Zambia, Turkey and the Netherlands. All children (in these societies as well as elsewhere) face most of these tasks (walking, talking, and so on) independent of cultural context. In respect of other domains, such as beliefs in supernatural influences (for example, Nsamenang, 1992), or skills such as reading and writing in literate as opposed to illiterate societies, postulating cross-cultural identity of domains does not seem to make much sense. Thus, at this first level of generalization it is easy to think of identical domains across all cultures, but also of non-identical domains. The second level consists of generalizations to domains defined in terms of hypothetical constructs, like cognitive abilities or emotions. Such domains are more inclusive, and tend to be unsharp; there is lack of clarity as to which elements belong or do not belong to it. Psychometric tests or other procedures (for example, observations) to assess such a domain are targeted at core aspects rather than at obtaining a representative sample of all possible elements. A measurement instrument may have to be carefully screened for a culture-specific content, but the omission of some small parts (‘cultural decentering’, Werner and Campbell, 1970) does not necessarily lead to a serious misrepresentation of the domain as a whole. For the construction of cross-culturally valid instruments, it is a minimum requirement that between the cultures concerned most elements relevant for a domain are shared. At this level of generalization, most culture-comparative researchers find it difficult to think of a domain that makes psychological sense in one culture but not in another, even when they will admit that current western definitions and operationalizations may be incomplete and biased. Thus, if ‘spatial orientation’ is a dimension in terms of which valid differences can be found between individuals in western societies, can we assume that there are other societies where such individual differences are absent or totally invalid? The third and final category of generalizations refers to highly inclusive and fuzzy domains that can be qualified as ‘unconstrained’. For a domain in this category, it is impossible to demarcate what does and what does not belong to it; even to think of a measure (for example, a questionnaire) that covers most of the core aspects cross-culturally may be difficult. Poortinga and Malpass (1986) argued that in studies in which observed cross-cultural differences are explained post hoc, the choice of a particular universe of interpretation is also ‘unconstrained’. Given the many ways in which cultures differ from each other, such interpretations are fairly gratuitous. Examples include intelligence
Sources of Evidence and Levels of Interpretation
227
(in the sense of intellectual capacity) and individualism-collectivism (Berry et al., 2002). In Table 16.1, the four levels of equivalence are crossed with the three levels of generalization. In the first column of the table, low level generalizations are portrayed as ‘variable across domains and cultures’, that is, some such domains are invariant across all cultures, while others do not occur in at least some cultural populations. A test of colour blindness like that of Ishihara, based on items where testees have to trace a curvy line that is invisible to colour blind testees, is likely to meet almost any condition of equivalence. Perhaps there is some uncertainty about full-score equivalence (factors such as illumination during the test could differentially affect score distributions from various populations), but in general scores should mean the same thing, independent of the cultural background of the testee. Moreover, colour blindness could be examined with laboratory standards if there should be doubt. Hence, it seems reasonable that group differences in rates of colour blindness (for example, Post, 1962) tend to be accepted at face value. On the other hand, writing or reading skills are not even examined in illiterate societies because the outcome is clear and of no interest. Greeting procedures are found everywhere (Eibl-Eibesfeldt, 1979, 1989), but placing them on common quantative scales (for example, in a comparative study of level of respect or friendliness) would seem a difficult task. Low-level generalizations tend to be given limited attention in cross-cultural psychology. Most authors search for relationships between culture and behaviour at more inclusive levels. Interpretations of cross-cultural differences in terms of highly inclusive concepts in the last column of Table 16.1 are marked as ‘beyond empirical analysis’. In cognition, the outstanding examples are notions such as general intelligence or ‘g’ (Jensen, 1985, 1998; Rushton, 1995) that presume a strong inborn component. According to some, the validity of cross-cultural differences on these concepts has been demonstrated. According to others, cultural factors better explain cross-cultural variance (Helms et al., 2003). It is unclear how the construct validity of quantitative cross-cultural differences can be demonstrated unambiguously. In any case, as long as less encompassing inferences can reasonably account for cross-cultural variance, it is scientifically and ethically unacceptable to make further reaching inferences, especially if these are likely to be discriminatory in respect of certain groups. The column in the middle of Table 16.1 refers to cognitive abilities, such a spatial orientation, numerical ability, abstract thinking and verbal fluency, which require generalizations of a medium level. Inferences at this level of inclusiveness, on the one hand, are subject to overgeneralization, while, on the other, they address larger domains or traits. If sufficiently validated, concepts at this level perhaps can show broader influences on cognition of eco-cultural factors or socialization patterns, which are often continuously present over the lifespan of individuals belonging to a cultural group. In summary, equivalence of domains in the first column of Table 16.1 is mainly a matter of common sense or face validity. In instances where there is doubt, empirical
228 Ype H. Poortinga analysis should lead to unambiguous results. In contrast, interpretations belonging in the last column are largely beyond the empirical control on equivalence. Medium-level inferences are literally and figuratively in the middle.
VALIDITY CONSTRAINTS ON GENERALIZATION In the previous section, levels of equivalence were crossed with levels of generalization. In the present section, I have used the same three levels of generalization, but now crossed with common kinds of data as found in culture-comparative and cultural research.
QUASI-EXPERIMENTS Broadly speaking, culture-comparative research is modelled on experimental and psychometric approaches in psychology. The need to provide evidence for the validity of interpretations in ways that are independent of the person of the researcher is a central concern. Such research tries to relate cognitive performance (dependent variables) to cultural conditions (independent variables or antecedents). The most powerful design is the experiment in which the researcher has control over the various treatments or conditions, and over the assignment of subjects to conditions. However, interpretations of culturecomparative studies modelled on the experiment tend to be post hoc because cultural factors, especially those that exert long-time influences, can rarely be manipulated as in a laboratory experiment. Moreover, the other major requirement for a true experiment, namely, random assignment of persons to treatment conditions (that is, different cultures) is incompatible with their embeddedness in only one cultural context. On the other hand, various quality controls are available that can help to strenghten the validity of interpretations, such as actual checks on the presumed presence of antecedent factors, statistical elimination of confounding effects and the use of multiple methods, implying that culture-comparative studies in principle can follow a quasi-experimental design, that is, a design in which existing groups are examined under different treatments (cf. Berry et al., 2002; Poortinga and Malpass, 1986; Shadish et al., 2001). In the last row of Table 16.2 it is suggested that quasi-experimental approaches are limited to studies which are geared to drawing low level inferences, that is, the domain or trait about which conclusions are drawn should be defined quite precisely. Medium-level and high-level inferences are described as being beyond manipulation. This seems rather evident, as manipulation, including quasi-manipulation, of conditions is only possible when the researcher has a precise idea of what the domain of interest entails and a fair deal of control on manipulations.
Sources of Evidence and Levels of Interpretation
229
PSYCHOMETRIC VALIDITY More often we find that in culture-comparative research, antecedent conditions are not varied systematically. Data are collected in various cultural populations on some dependent variable and the relevant antecedents, which have supposedly caused or facilitated observed cross-cultural differences in score distributions, are inferred. The validity of inferences is examined by relating a dependent variable to other supposedly relevant variables in a nomological network. The impossibility to control through manipulation, which distinguishes the second last row in Table 16.2 from the bottom row, implies that there usually will remain a gap of uncertainty. For example, the rather low cognitive performance of children in the majority world countries is often related to variables in the educational system, and for good reasons. However, psychologists rarely consider nutritional variables, such as a lack of certain minerals or vitamins (Bleichrodt et al., 1980; Louwman et al., 2000). It is highly likely that cognitive performance across a wide range of countries, including some with low GNP, will correlate with both educational and nutritional variables.
PLAUSIBILITY The main methodological orientation in relativism is towards qualitative research, employing a range of strategies of inquiry and methods of data analysis (Creswell, 1998; Silverman, 1993). In general, assessment methods are interpretive, with scoring methods that are not rule-bound and the insight of the psychologist in the meaning of the respondent’s reactions as a central feature. In a review of methods, Greenfield (1997) has emphasized the monitoring of events in context. She stressed the importance of an analysis of culture as an ongoing process, drawing on information from members of the community being studied. There is an extensive literature in psychology describing the procedures for analysing validity, and mentioning numerous threats to valid interpretation of data. Insofar as this literature is being ignored or even rejected as irrelevant (see several chapters in Denzin and Lincoln, 2000), relativism and universalism can be seen as contrasting paradigms that have little in common. However, there are also authors with a qualitative orientation to research, for whom validity is essential in scientific inquiry (Creswell, 2002). Once such an orientation is taken, differences become a matter of degree. Van de Vijver and Poortinga (2002) have pointed out how three forms of validity suggested by Greenfield (1997), namely, interpretive, ecological and theoretical validity, correspond to distinctions in quantitative methodology. Interpretive validity refers to the comparison of meanings of objects, events and behaviours to people engaged in them. In a quantitative framework this is known
230 Ype H. Poortinga as structural equivalence (Van de Vijver and Leung, 1997). Similar parallels can be drawn between ecological validity as defined by Greenfield and content validity in a quantitative framework, and between Greenfield’s theoretical validity and generalizability. In Table 16.2, the notion of plausibility is used. The reconstruction post hoc of how a particular state of affairs has come about or why a relationship exists between certain cultural variables can be less convincing or more convincing. Familiar examples of reconstructive accounts are found in jurisdiction. Court trials illustrate how judges or juries can be convinced by post hoc evidence to accept a particular reconstruction as ‘beyond reasonable doubt’. In cross-cultural psychology, as in cultural anthropology, such plausibilty is most often derived from systematic description. For low-level inferences, it is usually possible to make checks on plausibility. For medium-level domains, it is far less clear what such checks should entail. An example is the ethnographic description by Lutz (1988) of specific emotional states among the Ifaluk in the Pacific which, in her view, do not occur in USA. She presents various observations in support of her argument, but it is unclear which kinds of observations would have provided convincing evidence to the sceptic. Around domains at a high level of inclusiveness, there is even more fuzziness; notions such as discriminatory validity (Campbell and Fiske, 1959) tend to be shining in absence when high-level inferences are made. TABLE 16.2 Extent to which validity of cross-cultural differences in score levels is open to empirical control (ruling out alternative explanations) Low level inferences (measures are samples from domain)
Medium level inferences (measures contain core elements)
High level inferences (unconstrained domain)
Ad hoc observation
Controls absent
Controls absent
Controls absent
Systematic description
Plausibility checks
Plausibility arguments
Plausibility arguments
Psychometric assessment
Checks on obvious alternatives
Checks on some alternatives
Range of alternatives poorly defined
Quasi-experiments
Quasi-experimental manipulation
Usually beyond manipulation
Beyond manipulation
The top row in Table 16.2 refers to ad hoc observation or hearsay about a state of affairs for which an ad hoc explanation is given. Cultural stereotyping to ‘explain’ some publicized event by media commentators is a good example of inferences that belong to this row. Here a differentiation between levels of generalization is hardly meaningful, as evidence on the validity of any generalization is missing (that is, all generalizations are with regard to ‘unlimited domains’). This row is of virtually no interest to cross-cultural psychology.
Sources of Evidence and Levels of Interpretation
231
IMPLICATIONS FOR CROSS-CULTURAL RESEARCH ON COGNITION Controls on generalizations gradually weaken as one moves from lower left to upper right in Table 16.2. On the other hand, generalizations to more inclusive domains are more interesting as they provide broader explanations. In my opinion, the literature as a whole tends towards an overgeneralization of cross-cultural differences (Poortinga, 2003). In this section, I will discuss cross-cultural research on cognition with two questions in mind. The first is what the empirical evidence tells us about the equivalence of measures of cognition across cultures. The second question is what we appear to know, and not know, about the validity of such differences (Poortinga and Van de Vijver, 2004).
EQUIVALENCE ON COGNITIVE ABILITIES There are numerous studies with intelligence batteries for populations with limited or no formal education, but they are less recent and lack precise psychometric analyses on universality. Irvine (1979) listed 91 factor analyses based on studies in non-western countries, many of them with illiterate testees in Africa. He found numerous similarities in factor labels, but higher educated samples tended to reflect a larger number of factors. At the time there were few psychometric techniques for comparison of factor structures, and Irvine’s analysis was mainly impressionistic. Unfortunately, there do not appear to be more recent reviews of similar evidence comparing literate and illiterate samples. The translation and transfer of intelligence batteries has shown high congruence of factor structures, at least across literate societies. One recent example is based on 12 national standardization studies of the WISC-III, including North American, European and East Asian countries (Georgas et al., 2003). Unfortunately, in view of the limited range of educational variation between these countries, this does not constitute substantial evidence for cross-cultural invariance at a more global scale. An explicit attempt to examine quantitative forms of equivalence was reported by Van de Vijver (2002) for a battery of eight tests, based on either sets of figures or sets of letters, for various aspects of inductive reasoning administered in Zambia, Turkey, and the Netherlands among groups of young adolescents. Each of the eight tasks was based on a facet design (Cantor, 1985), allowing systematic combinations of elements with varying difficulty. Facets for the letter tasks included, for example, the presence (versus absence) of vowels, identical letters and sequential position in the alphabet. Van de Vijver found that all facets played a role in all samples, and that they contributed to a similar extent to the difficulty of the tasks in each sample. The findings supported structural equivalence of the tests, implying that inductive reasoning and its various component processes are present in the cultural populations examined. However, quantitative differences in mean score levels could not be interpreted at face value because statistical conditions for metric
232 Ype H. Poortinga and full score equivalence were not fully satisfied. Van de Vijver concluded that the strong findings on structural equivalence made it realistic to assume that inductive reasoning is an aspect of cognition with largely identical components in all schooled populations. Since the 1970s, there has been extensive research on the conditons for equivalence, or comparability, of cognitive tests for ethnic groups in USA. Most of these studies were aimed at identifying the bias of individual items, or Differential Item Functioning (DIF) (Holland and Wainer, 1993). In these analyses, the performance on all other items of a test is taken as a standard for evaluating whether or not a particular item is biased. However, many threats to equivalence are probably similar for all items in an instrument, such as poor quality of school education, differences in speed-accuracy trade off, low motivation, and so on. Such threats are unlikely to be detected at the item level (Poortinga, 1989; Van de Vijver and Leung, 1997). The examples given in this subsection illustrate two trends in findings, namely: (a) tests of cognitive abilities usually satisfy psychometric conditions of structural equivalence at least in literate populations and (b) full-score equivalence is largely beyond our reach. Moreover, we usually just cannot attribute quantitative test score differences on ability to specific cultural factors, with the exclusion of other factors.
VALIDITY OF CROSS-CULTURAL DIFFERENCES ON COGNITIVE VARIABLES There have been extensive discussions on the organization of cognition, especially intelligence (for example, Detterman; 1994). Even if it is accepted that intelligence is best represented by a hierarchical model with a central factor, such as ‘g’ or cognitive capacity (Carroll, 1993), this by no means implies that cross-cultural differences are also organized in a hierarchical fashion. An alternative view is that differences in scores on IQ tests reflect a (perhaps extensive) set of algorithms and skills, each of which forms a domain of low inclusiveness. The apparent coherence in these differences, as reflected in correlations between tasks across cultural samples, is then a function of factors that are not primarily of a cognitive nature, such as school education or motivation. Tasks that one finds in intelligence batteries are typically of the kind trained in western school curricula. This idea is supported, for example, by research in the tradition of Piaget, where early studies showed average differences of several years between illiterate and western groups in the onset of concrete operational thinking (Dasen, 1972). Subsequently, Dasen (1975; see also Segall et al., 1999) demonstrated that a small amount of training could be sufficient to elicit this mode of thinking in illiterate children, suggesting that generalizations of differences to broader domains were questionable. In another early tradition mentioned before, Vygotsky’s cultural mediation theory, there has been a similar trend. The context-specific orientation of Cole and his colleagues
Sources of Evidence and Levels of Interpretation
233
(Cole, 1996; Cole et al., 1971; Scribner and Cole, 1981) provides clear examples that lowlevel generalizations can account for differences on cognitive tasks. The best-known study in this tradition is on the cognitive effects of literacy among the Vai in Liberia (Scribner and Cole, 1981). The research question was whether literate and illiterate groups differed in terms of broad cognitive processes, or whether the effects of literacy were task-specific and limited to what may be called cognitive algorithms. In most societies literacy is confounded with western-style formal schooling, but among the Vai one finds next to illiteracy and a western form of literacy two other forms of literacy, namely in Vai, a local syllabic script, and in Arabic among those who have attended a Qu’ranic school. Scribner and Cole showed that Vai literacy and Arabic literacy had some limited effects on performance, with tasks that matched specific requirements of phoneme organization in Vai and memorization in Arabic (rote learning is prominent in Qu’ran schools). However, major differences in performance levels on a wider range of ability tests were found only between western style literates and the other three samples. Thus, a western-type school curricula can be said to teach a diverse array of cognitive algorithms found in westernstyle ability tests. A recent line of research that clearly goes against the trend towards low-level generalizations of cross-cultural differences in cognition ascribes a dialectical form of reasoning to Chinese, as opposed to a more logical way of reasoning among Americans (Peng and Nisbett, 1999). This is a fascinating development, but a note of caution may be appropriate. First, there is an historical argument; previous contrasts, for example, in counter-factual thinking due to grammatical differences between Chinese languages and English, did not stand up to later more precise empirical scrutiny (see Berry et al., 2002). Second, most of the tasks that are used have a certain ambivalence, perhaps inviting less outspoken answers from respondents in a society where modesty of expression is a social convention more valued than in USA. Third, the presumed links to cultural systems and history going back to classical Greek and Chinese sources, drawn by Nisbett et al. (2001) without much consideration of historical discontinuities (for example, the Middle Ages in Europe), and the social elitist character of learning in peasant societies clearly point to the dangers of overgeneralization. Segall et al. (1990) have referred to past dichtomies in modes of thinking as ‘Great Divide’ theories. In view of the rapid decline of many earlier ideas after initial success, it may be premature to embrace this new Great Divide too readily. Evidence about cross-cultural invariance is also obtained when observed cross-cultural differences can be explained in terms of specific variables that influence the results, but that are not part of the target domain in a study. Such sources of cultural bias can be examined when they are included in a quasi-experimental design. An example is a recent study by Shebani et al. (2005b) on the short-term memory (STM) span of Libyan and Dutch school children. The design was based on two well-known findings, namely, that memory span is dependent on the length of the words to be remembered and that cross-cultural
234 Ype H. Poortinga differences in memory span are linked to differences in word length in various languages, operationalized as differences in reading speed (Baddeley, 1997; Naveh-Benjamin and Ayres, 1986). The Dutch children had a somewhat higher span, but this could be accounted for completely in terms of the longer reading time for Arabic numerals. The findings were further validated by experimental manipulation of a feature of Arabic, namely, the existence of a longer and a shorter form of the numerals. When the longer form was used, memory span was lower in the Libyan children. Thus, by including a relevant explanatory variable (reading speed) in a quasi-experimental design, results were obtained supporting a view of memory span scores as meeting conditions of full-score equivalence (Table 16.1), and of a specific antecedent variable requiring only a low-level generalization (Table 16.2) that could fully account for the observed differences in memory span. Shebani et al. (2005b) also conducted a study on rehearsal training with Libyan and Dutch school children. Libyan children compared with Dutch children on pretest score levels showed somewhat higher gains than their Dutch counterparts, both in an experimental and in a control group. The correspondence in patterns of change supported metric equivalence of the scores (Table 16.1), but no variable was found that could explain the cross-cultural difference. Hence, it is not clear whether the changes were score equivalent, and thus valid, or reflected some source of cultural bias. Another study in the same project (Shebani 2001) involved four tests of metamemory. These tests showed far larger cross-cultural variation in scores, and most of this variation could not be accounted for. Psychometric analyses showed that the tests did not even meet conditions for structural equivalence, suggesting that perhaps non-corresponding aspects of behaviour were tapped. This may be due to poor test construction, but the findings are also compatible with the view that cultural specificity is a real feature of metamemory (see Table 16.1 top row). The tradition of everyday cognition (for example, Schliemann et al., 1997) is based on descriptive accounts of the cognitive demands and problem-solving strategies found in a particular group. Fascinating skills have been described including, for example, the navigation of boats by the Pulawat across large distances in the Pacific without a compass (Gladwin, 1970), counting among the Oksapmin who have a number system based on parts of the body (Saxe, 1981, 1982), and weaving in various societies (for example, Childs and Greenfield, 1980; Rogoff and Gauvain, 1984; Tanon, 1994). Studies of everyday cognition have generally shown limited transfer and generalization of learning from one category of situations (domain) to another, including from school to non-school situations (Segall et al., 1999). Still, authors tend to see culture-specific knowledge and skills as the outcome of more general modes of teaching and learning that differ from those prominent in the western school setting. Examples include scaffolding (for example, Greenfield and Lave, 1982) and apprenticeship (Rogoff, 1990, 2003). If there were less emphasis on broad cultural mechanisms and meaning, and more on external situational demands and opportunities, descriptions in the everyday cognition literature would fit low-level generalizations.
Sources of Evidence and Levels of Interpretation
235
In summary, I would like to argue that so far there is hardly any strong empirical evidence that causally links broader patterns in cognition to socio-cultural or ecocultural conditions. Evidence linking phenomena that tend to be wide apart is essentially correlational, and such evidence does not seem sufficient to convince researchers of cognitive science.
CONCLUSIONS Cross-cultural research on cognitive variables should be an important source of information for cognitive science. As in other fields of psychology, this kind of research serves ‘to extend the range of variation’ (Whiting, 1954: 524) beyond what is found in a single group or society. I have tried to illustrate how researchers need to pay attention not only to crosscultural differences, but also to invariance. Two arguments derive from this chapter. First, so far structural equivalence has been found for common western intelligence tasks in nonwestern societies, although for lack of empirical data it is not clear how far this finding will extend to non-literate societies. Second, it can be a fruitful strategy for culture-comparative research to examine specific antecedent variables against a background of invariance. A third comment is perhaps less a conclusion than a somewhat provocative suggestion: the history of cross-cultural studies has shown so many times that broad or inclusive differences in cognitive functioning did not hold up under close empirical scrutiny, that claims for the existence of such differences should be looked at with suspicion (Poortinga, 2003). If this latter statement has any merit, cognition scientists should lean towards quasi-experimental designs and low-level inferences in research involving cross-cultural variation, rather than to the inclusive and fuzzy concepts that have been dominant for such a long time in cross-cultural research.
ACKNOWLEDGEMENTS This chapter was prepared during a stay as Fellow-in-Residence at the Netherlands Institute for Advanced Study in the Humanities and Social Sciences (NIAS).
REFERENCES Baddeley, A. 1997. Human Memory: Theory and Practice (revised edition). Mahwah, NJ: Erlbaum. Berry, J.W., Y.H. Poortinga, M.H. Segall and P.R. Dasen. 2002. Cross-Cultural Psychology: Research and Applications. Cambridge: Cambridge University Press. Biesheuvel, S. 1943. African Intelligence. Johannesburg: South African Institute of Race Relations. Bleichrodt, N., P.J.D. Drenth and A. Querido. 1980. ‘Effects of Iodine Deficiency on Mental and Psychomotor Abilities’. American Journal of Physical Anthropology, 53: 55–67.
236 Ype H. Poortinga Brouwers, S.A., D.A. Van Hemert, S.M. Breugelmans and F.J.R. Van de Vijver. 2004. ‘A Historical Analysis of Empirical Studies Published in the Journal of Cross-Cultural Psychology, 1970–2004’. Journal of Cross-Cultural Psychology, 35: 251–62. Campbell, D.T. and D.W. Fiske. 1959. ‘Convergent and Discriminant Validation by the Multitrait– Multimethod Matrix’. Psychological Bulletin, 56: 81–105. Cantor, D. (ed.). 1985. Facet Theory: Approaches to Social Research. New York: Springer. Carroll, J.B. 1993. Human Cognitive Abilities: A Survey of Factor-Analytic Studies. Cambridge: Cambridge University Press. Childs, C.P. and P.M. Greenfield. 1980. ‘Informal Modes of Learning. The Case of Zinacanteco Weaving’. In N. Warren (ed.), Studies in Cross-Cultural Psychology, Vol. 2 (pp. 269–316). London: Academic Press. Cole, M. 1996. Cultural Psychology: A Once and Future Discipline. Cambridge, MA: Belknap. Cole, M., J. Gay, J. Glick and D.W. Sharp. 1971. The Cultural Context of Learning and Thinking. London: Basic Books. Creswell, J.W. 1998. Qualitative Inquiry and Research Design: Choosing Among Five Traditions. Thousand Oaks, California: Sage Publications. ———. 2002. Research Design: Qualitative, Quantitative, and Mixed Methods Approaches (2nd edition). Thousand Oaks, California: Sage Publications. Cronbach, L.J., G.C. Gleser, H. Nanda and N. Rajaratnam. 1973. The Dependability of Behavioral Measurements. New York: Wiley. Dasen, P.R. 1972. ‘Cross-Cultural Piagetian Research: A Summary’. Journal of Cross-Cultural Psychology, 7: 75–85. ———. 1975. ‘Concrete Operational Development in Three Cultures’. Journal of Cross-Cultural Psychology, 6: 156–72. Denzin, N.K. and Y.S. Lincoln (eds). 2000. Handbook of Qualitative Research (2nd edition). Thousand Oaks, California: Sage Publications. Detterman, D.K. 1994. ‘The Case for the Prosecution: Transfer as an Epiphenomenon’. In D.K. Detterman and R.J. Sternberg (eds), Transfer on Trial: Intelligence, Cognition and Instruction (pp. 1–24). Norwood, NJ: Ablex. Eibl-Eibesfeldt, I. 1979. ‘Human Ethology: Concepts and Implications for the Sciences of Man’. The Behavioral and Brain Sciences, 2: 1–57. ———. 1989. Human Ethology. New York: Aldine de Gruyter. Ellis, B.B. 1989. ‘Differential Item Functioning: Implications for Test Translations’. Journal of Applied Psychology, 74: 912–21. Fiske, A., S. Kitayama, H. Markus and R. Nisbett. 1998. ‘The Cultural Matrix of Social Psychology’. In D. Gilbert, S. Fiske and G. Lindzey (eds), Handbook of Social Psychology (pp. 915–81). New York: McGraw-Hill. Fontaine, J.R.J. 2005. ‘Equivalence’. In K. Kempf-Leonard (ed.), Encyclopedia of Social Measurement, Vol. 1 (pp. 803–13). San Diego, California: Academic Press. Georgas, J., F.J.R. Van de Vijver, L. Weiss, D. Saklofske. 2003. ‘A Cross-Cultural Analysis of the WISC-III’. In J. Georgas, L. Weiss, F.J.R. Van de Vijver and D. Saklofske (eds), Cultures and Children’s Intelligence: A Cross-Cultural Analysis of the WISC-III (pp. 278–314). New York: Academic Press. Gladwin, T. 1970. East is a Big Bird: Navigation and Logic on Puluwat Atoll. Cambridge, MA: Harvard University Press. Greenfield, P.M. 1997. ‘Culture as Process: Empirical Methods for Cultural Psychology’. In J.W. Berry, Y.H. Poortinga and J. Pandey (eds), Handbook of Cross-Cultural Psychology, Vol. 1, Theory and Method (2nd edition) (pp. 301–46). Boston: Allyn and Bacon.
Sources of Evidence and Levels of Interpretation
237
Greenfield, P.M. and J. Lave. 1982. ‘Cognitive Aspects of Informal Education’. In D.A. Wagner and H.W. Stevenson (eds), Cultural Perspectives on Child Development (pp. 181–207). San Francisco: Freeman. Hambleton, R.K. 1994. ‘Guidelines for Adapting Educational and Psychological Tests: A Progress Report’. European Journal of Psychological Assessment, 10: 229–44. Helms-Lorenz, M., F.J.R. Van de Vijver, Y.H. Poortinga. 2003. ‘Cross-Cultural Differences in Cognitive Performance and Spearman’s Hypothesis’. Intelligence, 31: 9–29. Holland, P.W. and H. Wainer (eds). 1993. Differential Item Functioning. Hillsdale, NJ: Erlbaum. Hoosain, R. and F. Salili. 1987. ‘Language Differences in Pronunciation Speed for Numbers, Digit Span, and Mathematical Ability’. Psychologia: An International Journal of Psychology in the Orient, 30: 34–38. Irvine, S.H. 1979. ‘The Place of Factor Analysis in Cross-Cultural Methodology, and Its Contribution to Cognitive Theory’. In L. Eckensberger, W. Lonner and Y.H. Poortinga (eds), Cross-Cultural Contributions to Psychology. (pp. 300–41). Lisse: Swets & Zeitlinger. Jensen, A.R. 1985. ‘The Nature of Black–White Difference on Various Psychometric Tests: Spearman’s Hypothesis’. Behavioral and Brain Sciences, 8: 193–263. ———. 1998. The g Factor. The Science of Mental Ability. Westport, CT: Praeger. Klineberg, O. 1935. Race Differences. New York: Harper & Row. Louwman, W.J., M. van Dusseldorp, F.J.R. van de Vijver, C.M.G. Thomas, J. Schneede and W.A. van Staveren. 2000. ‘The Effect of a Marginal Vitamin B12 Status on Cognitive and Psychomotor Development’. American Journal of Clinical Nutrition, 72: 762–69. Luria, A.R. 1971. ‘Towards the Problem of the Historical Nature of Psychological Processes’. International Journal of Psychology, 6: 259–72. ———. 1976. Cognitive Development: Its Cultural and Social Foundations. Cambridge, Massachusetts: Harvard University Press. Lutz, C. 1988. Unnatural Emotions: Everyday Sentiments on a Micronesian Atoll and Their Challenge to Western Theory. Chicago: University of Chicago Press. Naveh-Benjamin, M. and T.J. Ayres. 1986. ‘Digit Span, Reading Rate, and Linguistic Relativity’. Quarterly Journal of Experimental Psychology, 38 A: 739–51. Nisbett, R.E., K. Peng, I. Choi and A. Norenzyan. 2001. ‘Culture and Systems of Thought: Holistic Versus Analytic Cognition’. Psychological Review, 108: 291–310. Nsamenang, A.B. 1992. Human Development in Cultural Context: A Third World Perspective. Newbury Park, California: Sage. Peng, K. and R. Nisbett. 1999. ‘Culture, Dialectics and Reasoning About Contradiction’. American Psychologist, 54: 741–54. Poortinga, Y.H. 1989. ‘Equivalence of Cross-Cultural Data: An Overview of Basic Issues’. International Journal of Psychology, 24: 737–56. ———. 2003. ‘Coherence of Culture and Generalizability of Data: Two Questionable Assumptions in CrossCultural Psychology’. In J. Berman and J. Berman (eds), Cross-Cultural Differences in Perspectives on the Self, Vol. 49 of the Nebraska Symposium on Motivation (pp. 257–305). Lincoln, Nebraska: University of Nebraska Press. Poortinga, Y.H. and R.S. Malpass. 1986. ‘Making Inferences from Cross-Cultural Data’. In W.J. Lonner and J.W. Berry (eds), Field Methods in Cross-Cultural Research (pp. 17–46 ). Beverly Hills, California: Sage. Poortinga, Y.H. and F.J.R. Van de Vijver. 2004. ‘Culture and Cognition: Performance Differences and Invariant Structures’. In R.J. Sternberg and E. Grigorenko (eds), Culture and Competence (pp. 139–62). Washington, DC: American Psychological Association. Porteus, S.D. 1937. Primitive Intelligence and Environment. New York: McMillan.
238 Ype H. Poortinga Post, R.H. 1962. ‘Population Differences in Red and Green Color Vision Deficiency: A Review, and a Query on Selection Relaxation’. Eugenics Quarterly, 9: 131–46. Rogoff, B. 1990. Apprenticeship in Thinking: Cognitive Development in Social Context. New York: Oxford University Press. ———. 2003. The Cultural Nature of Human Development. Oxford: Oxford University Press. Rogoff, B. and M. Gauvain. 1984. ‘The Cognitive Consequences of Specific Experiences, Weaving Versus Schooling Among the Navajo’. Journal of Cross-Cultural Psychology, 15: 453–75. Rushton, P. 1995. Race, Evolution and Behavior. New Brunswick: Transaction. Saxe, G.B. 1981. ‘Body Parts as Numerals: A Developmental Analysis of Numeration Among Remote Oksapmin Village Populations in Papua New Guinea’. Child Development, 52: 306–16. ———. 1982. ‘Developing Forms of Arithmetical Thought Among the Oksapmin of Papua New Guinea’. Child Developmental Psychology, 18: 583–94. Schliemann, A., D. Carraher and S. Ceci. 1997. ‘Everyday Cognition’. In J.W. Berry, P.R. Dasen and T.S. Saraswathi (eds), Handbook of Cross-Cultural Psychology, Vol. 2, Basic Processes and Human Development (2nd edition) (pp. 177–216). Boston: Allyn and Bacon. Scribner, S. 1979. ‘Modes of Thinking and Ways of Speaking: Culture and Logic Reconsidered’. In R.O. Freedle (ed.), New Directions in Discourse Processing (pp. 223–43). Norwood, New Jersey: Ablex. Scribner, S. and M. Cole. 1981. The Psychology of Literacy. Cambridge, Massachusetts: Harvard University Press. Segall, M.H., P.R. Dasen, J.W. Berry and Y.H. Poortinga. 1990. Human Behavior in Global Perspective: An Introduction to Cross-Cultural Psychology. New York: Pergamon. ———. 1999. Human Behavior in Global Perspective: An Introduction to Cross-Cultural Psychology (2nd edition). Boston: Allyn & Bacon. Shadish, W.R., T.D. Cook and T.D. Campbell. 2001. Experimental and Quasi-Experimental Designs for Generalized Causal Inference. Boston: Houghton Mifflin. Shebani, M. 2001. Memory Development of Libyan and Dutch Children. PhD thesis. Tilburg, The Netherlands: Tilburg University. Shebani, M.F.A., F.J.R. Van de Vijver and Y.H. Poortinga. 2005. ‘A Strict Test of the Phonological Loop Hypothesis With Libyan Data’. Memory and Cognition, 33: 96–102. Shuey, A.M. 1966. The Testing of Negro Intelligence (2nd edition). New York: Social Science Press. Silverman, D. 1993. Interpreting Qualitative Data: Methods for Analysing Talk, Text, and Interaction. London: Sage Publications. Stigler, J.W., S.W. Lee and H.W. Stevenson. 1986. ‘Digit Memory Span in Chinese and English: Evidence for a Temporary Limited Store’. Cognition, 23: 1–20. Tanon, F. 1994. A Cultural View on Planning: The Case of Weaving in Ivory Coast. Tilburg: Tilburg University Press. Tucker, L.R. 1951. A Method for Synthesis of Factor Analytic Studies, (Personnel Research Section Report No. 984). Washington, DC: Department of the Army. Van de Vijver, F.J.R. 2002. ‘Inductive Reasoning in Zambia, Turkey, and the Netherlands Establishing Cross-Cultural Equivalence’. Intelligence, 30: 313–51. Van de Vijver, F.J.R. and K. Leung. 1997. Methods and Data Analysis for Cross-Cultural Research. Newbury Park, California: Sage Publications. Van de Vijver, F.J.R. and N.K. Tanzer. 1997. ‘Bias and Equivalence in Cross-Cultural Assessment: An Overview’. European Review of Applied Psychology, 47: 263–79. Van de Vijver, F.J.R. and Y.H. Poortinga. 2002. ‘On the Study of Culture in Developmental Science’. Human Development, 45: 246–56.
Sources of Evidence and Levels of Interpretation
239
Vygotsky, L.S. 1978. Mind in Society: The Development of Higher Psychological Processes. Cambridge, Massachusetts: Harvard University Press. Werner, O. and D.T. Campbell. 1970. ‘Translating, Working Through Interpreters and the Problem of Decentering’. In R. Naroll and R. Cohen (eds), A Handbook of Method in Cultural Anthropology (pp. 398–419). New York: Natural History Press. Whiting, J.W.M. 1954. ‘The Cross-Cultural Method’. In G. Lindzey (ed.), The Handbook of Social Psychology, Vol. 1 (pp. 523–31). Cambridge, Massachusetts: Addison-Wesley. Willemsen, M.E. and F.J.R. Van de Vijver. 1997. ‘Developmental Expectations of Dutch, Turkish–Dutch, and Zambian Mothers: Towards an Explanation of Cross-Cultural Differences’. International Journal of Behavioral Development, 21: 837–54.
Chapter 17 Spatial Language and Concept Development: Theoretical Background and Overview R.C. Mishra and Pierre R. Dasen
SPATIAL FRAMES OF REFERENCE (FoR): THEORETICAL BACKGROUND Developmental studies carried out in western societies suggest that children first build up spatial concepts in relation to their own body, following the sequence of topological, projective and Euclidean space (Piaget and Inhelder, 1956). More recent theories of human development propose much the same scheme (for example, Taylor and Tversky, 1996). This is illustrated in Table 17.1. The theory that spatial representation is basically built up from the point of view of the human body is still current (Grabowski, 1999; Taylor and Tversky, 1996; Werner and Hubel, 1999). While the ability to use geocentric landmarks varies with age depending on the experimental conditions, in the existing (western) literature, it never occurs before the child has built up body-related spatial representations. Occasionally, there is the suggestion that both these abilities might develop together, and the choice of reference frame is very much situation dependent rather than a developmental feature, but there never seems to be any complete reversal from the sequence originally described by Piaget. A major problem with the whole area of spatial concept development is that it has completely relied on research with western samples (Mishra, 1997). Whether the same sequence of stages of development would hold up in other cultures is not much known. The very centration on the construction of space on the basis of the body could be a bias due to western individualism. The distinction between these spatial concepts is akin to the distinction at the linguistic level between intrinsic or object-centred, deictic or viewer-centred (relative or egocentric), and extrinsic or environment-centred (absolute or geocentric) spatial terms. In an intrinsic
Spatial Language and Concept Development
241
frame, the location of objects is described one in relation to the others. In the relative frame, the description is in relation to a viewer’s front, back, left and right, that is, it is viewer-centred, and requires knowledge of the viewer’s position and orientation in space. In the absolute frame, objects are located according to a coordinate system that is external to the scene. Different language communities preferentially use different reference frames (Levinson, 1996, 2003). TABLE 17.1
Spatial frames of reference in developmental psychology and in linguistics
Piaget and Inhelder (1956)
Topological
Projective
Euclidean
Taylor and Tversky (1996)
Intrinsic: objectcentred
Deictic: viewercentred
Extrinsic: environmentcentred
Levinson (1996, 2003)
Intrinsic
Relative, egocentric
Absolute, geocentric
Cross-cultural studies of spatial cognition provide a contrast between the emic and etic approaches (Segall et al., 1999). On the emic side we find anthropological descriptions of how space is organized in different cultures (for example, Gladwin, 1970; Hutchins, 1995; Pinxten et al., 1983) that speak little about specific psychological processes and developmental aspects. On the (derived) etic side is the cross-cultural replication of Piaget’s theory, using classical ‘Piagetian’ tasks (Dasen, 1993). However, this research neither suggests any reversals in the sequence of stages, nor indeed any culturally specific cognitive processes. The possibility of different developmental pathways and different developmental end stages has been suggested (for example Greenfield, 1976; Troadec, 1999), but never convincingly demonstrated.
LINGUISTICS AND LINGUISTIC RELATIVISM Does language determine the way one thinks? The issue of linguistic relativity has been revived recently (Gumperz and Levinson, 1996; Levinson, 2003). Cross-cultural research shows that only a weak form of linguistic relativism finds empirical support (Berry et al., 2002) and that basic cognitive processes are universal (Mishra, 1997; Segall et al., 1999). It is widely assumed that the coding of spatial arrays for memory will be determined by general properties of visual perception, and that it is natural and, thus, universal to conceptualize space from an egocentric or ‘relative’ point of view. Research also indicates that speakers of European languages are used to egocentric encoding; other forms of encoding appear impossible to them. The egocentric conception of space has been considered universal and ‘more natural and primitive’ (Miller and Johnson-Laird, 1976: 34).
242 R.C. Mishra and Pierre R. Dasen However, there are growing doubts about these basic assumptions (Wassmann, 1994). If we have to describe the position of an object or person with respect to another, we achieve this in English by utilizing the projective notions of right and left with reference to the body. Some languages do not use the body-centred spatial notions of right and left, front and back. Instead they use fixed, environment-centred or geocentric FoR. While in a relative frame the description of an object or person changes depending on the speaker’s body position, in the geocentric frame the description does not necessarily change with the viewer’s change of position. Such a non-egocentric linguistic coding of a spatial array seems to be incongruent with the perceptual information in fundamental ways, and the question is whether these linguistic differences correspond to conceptual differences. We may assume that spatial representations are influenced either by sensory information (which is egocentric) or by language (which may or may not be egocentric). In European languages, which are egocentric, the two are confounded, but there are other languages that use exclusively intrinsic or absolute or mixed FoR. It is possible to dissociate these influences by carrying out studies with speakers of these languages. Coordinated research in several locations has been carried out by members of the Cognitive Anthropology Research Group (CARG) at the Max Planck Institute for Psycholinguistics, Nijmegen. Levinson (2003) summarizes all this research in a single volume. Two studies by de León (1994, 1995) in Mexico and among australian Aborigines show that the overall developmental trend in these two communities, where a geocentric FoR characterizes adult language, seems to go from the egocentric, intrinsic terms to locally geocentric, and in some cases to abstract geocentric terms. This seems to confirm the classical developmental trend even in populations that use an absolute, geocentric system only. However, since relative terms (left/right) are not used at all, the research cannot say much about the relationship between the relative and the geocentric systems. This shows that research is needed in locations where a relative system exists in the language, but is not the predominant one. The studies by de León deal only with the development of language. From these, it cannot be directly concluded that cognitive development has to follow the same sequence. Further research is therefore needed, combining the study of spatial language development and the study of cognitive development, which is what we are doing in this project.
KAJA-KELOD: OUR FIRST STUDY IN BALI While the CARG group consists mainly of linguists, they have associated other social scientists to their research, in particular anthropologists, and among them a cognitive anthropologist, Jürg Wassmann, who is himself interested in the collaboration between
Spatial Language and Concept Development
243
anthropologists and psychologists (Wassmann, 1995, 1997; Wassmann and Dasen, 1994a, 1994b). Dasen joined Wassmann in Bali, Indonesia, in 1994 to study the Balinese orientation system and its importance in Balinese culture (Wassmann and Dasen, 1998). Orientation is geared to the island’s central volcano, the dwelling place of the Hindu gods of Bali. Kaja (towards the mountain) is the sacred and pure direction, opposed to kelod (towards the sea). The axis kaja–kelod is in effect a variable direction as one turns around the island. In principle, the axis kangin–kauh is orthogonal to it. The entire Balinese cosmology is related to this orientation system: from the human body to the whole universe, from the architecture of temples and villages to the social structure. Children learn the use of the orientation system very early in life, although they also learn the ritually important distinctions between the left and the right hand. Wassmann and Dasen (1998) carried out a linguistic survey of the use of spatial terms in Balinese, and examined in detail how the inhabitants of various sites on the eastern peninsula of Bali use the system, and documented the local adaptations of the system. They discovered that geocentric terms were applied not only to macro space, but also to micro space whenever an object had to be located or a direction indicated. Left and right were applied only to objects in contact with the body, while all other objects were located with the geocentric orientation system. While the Balinese could use two coding systems, the preference for the absolute system was clear. In the Balinese language, the system of geocentric orientation is so strong that it determines not only the manner of speaking, but also a mode of spatial representation and its commitment to memory. Children started with a ‘geocentric’ notion, which was represented in an ‘up to the mountain—down to the sea’ orientation to space. While the very young children (four to five years) used exclusively the geocentric system in their language and in their way of memorizing a spatial device, there seemed to be a developmental change towards relative encoding. This study indicates that in some cultural and linguistic contexts, the sequence of acquisition of spatial knowledge could be reversed, and that different orientation systems and their linguistic encoding might have a significant impact on spatial cognitive development.
STUDIES IN INDIA AND NEPAL The development of spatial orientation and its relationship with cognitive performance has been studied in the Indian and Nepalese cultural contexts also. Niraula and Mishra (2001) analysed the development of the spatial orientation system among the Newar children of Nepal, aged 5–12 years, and attending primary schools in the city of Kathmandu and neighbouring villages. They used a pictorial display, and asked children to explore various objects along a path, recall the objects and their spatial locations, and tell the spatial position
244 R.C. Mishra and Pierre R. Dasen of various objects relative to other objects depicted in the picture. An analysis was made of the language that children at different age levels used to describe the spatial position of objects. The findings revealed that at the age of five to six, children could not use geocentric terms (North-South-East-West [NSEW]) to describe the position of objects; by the age of eight to nine, almost 40 per cent of children used geocentric terms; and by the age of 11–12, almost 85 per cent of children did so. There was no evidence of difference between boys and girls or rural and urban children in the use of these terms at different age levels. An analysis of children’s performance on cognitive tasks revealed that geocentric language use was positively correlated with the memory of objects and memory of spatial location of objects along the path, as well as performance on the Story–Pictorial Embedded Figures Test that measures the level of psychological differentiation. These measures generally appeared to be negatively correlated with the use of relative language (Left-Right-Front-Back [LRFB]). In a previous project of our team, we (Mishra et al., 2003) worked with 4–14-year-old, schooled and unschooled village and city children in India, and mainly Newar mountain children in Nepal (Niraula et al., 2004). The impact of schooling was the focus of a chapter by Dasen et al. (2004) and the urban/rural comparison is presented in Mishra and Dasen (2005). In all these locations we attempted to understand (a) the language that children use to describe space, and developmental changes in these; (b) absolute and relative encoding of spatial arrays following the paradigm developed by the CARG group; and (c) the relationship of language use with performance on spatial cognitive tasks. One of the assumptions was that cultural features of the groups, including the spatial orientation system and the language that goes with it, would be adaptive to respective ecological settings. Thus, we predicted and found that in the mountains of Nepal, where the obvious feature of the terrain is the slope, people would refer to ‘up’ and ‘down’, and occasionally to local landmarks to describe objects in space. In a village in India, which was located in the flat plains of the Ganges characterized by a complete absence of hills and very few obvious landmarks, use of cardinal directions (NSEW) was the norm. In the city (Varanasi), that provides limited space and a highly congested setting for the organization of activities with narrow lanes requiring several left–right movements in walking, the use of relative (LRFB) terms was expected to predominate in language, and indeed the results showed that both systems were used. Children’s language use was studied with a Route Description Task, and the Piagetian Perspectives Task. ‘Animals in a Row’, ‘Chips’ and ‘Steve’s Maze’ tasks of the Nijmegen series were used for elicitation of language and the study of spatial encoding. These tasks are described below. Route Memory, Rotation of Landscapes, Horizontality and Perspectives Tasks were used for the assessment of spatial cognitive development. This research design had a definite advantage over those used in other studies. For example, we were able to record the language used by each child and to relate it to his or her encoding measured task by task. We were also able to examine the particular language used on particular
Spatial Language and Concept Development
245
items in relation to the encoding on the same items. In the previous research, the CARG team (and Wassmann and Dasen, 1998 as well) had examined the relationship between language and spatial encoding only at the group level. A linguistic analysis shows that in many societies that have been studied by that team, a geocentric orientation system exists in the culture, is accompanied by a predominance of geocentric language, and goes with a strong trend towards absolute encoding, providing a seemingly strong confirmation of linguistic relativism. In terms of language use, the development trend showed a beginning from the use of ‘this way/that way’ (deictic, often accompanied by gestures or hand movement), through situationally specific landmarks inside of the room and conventional landmarks further removed, to the clearly geocentric ones (that is, Up/Down in Nepal or the cardinal directions, NSEW). The results seem to indicate that the children use references outside of the display starting at a very early age, but that there is a developmental trend from the use of more concrete landmarks to more abstract dimensions of space. The analysis of the encoding of simple spatial arrays revealed that the children in the plains of India and mountains of Nepal generally do it in an absolute rather than ‘relative’ frame, which is congruent with the geocentric orientation systems that are prevalent in these locations. On the other hand, there was also evidence for ‘task specificity’ in spatial encoding. More ‘absolute’ encoding was observed if the task could be easily coded in linguistic terms (for example, Animals), whereas more relative encoding was evident if the task was more difficult to encode linguistically and easier to encode iconically (for example, Steve’s Maze). The results are illustrated in Figure 17.1. While in Bali, in the study carried out in 1994 (Wassmann and Dasen, 1998) all of the 4 and 5-year-old children used an absolute encoding, followed by a slight decrease until the age of 10–11 years, there is an opposite trend of a steady increase in the proportion of items with an absolute encoding in all of the samples in India and Nepal. Given the failure to find the same results in India and Nepal as in Bali, one of the first questions that arise is whether the Balinese data were reliable and could be replicated in Bali itself. Does the geocentric and egocentric language use, and an absolute or relative encoding have any relationship with, the cognitive performance of children? This question was addressed by analysing the relationships of modal language and encoding with performance on a series of Piagetian tasks that assess spatial cognitive development at a more general level. The findings revealed no relationship between absolute or relative spatial encoding and broader aspects of cognitive development. On the other hand, the data did show some correlation between geocentric language use and some of the cognitive development tasks, even when the effect of age and schooling were partialled out. The overall developmental trend indicated slightly more relative encoding in young children, which was replaced by absolute encoding in older children. This trend, however, was not reflected in the modal language used. Children almost never used relative language except in the city, where it increased with age, while relative encoding decreased with age. This presents another contradiction to the linguistic relativity hypothesis.
246 R.C. Mishra and Pierre R. Dasen FIGURE 17.1 Results of previous studies (Bali, India, Nepal), absolute encoding for Animals task (3 animals only) and Steve’s Maze
Note: For the Bali, age group 13 includes 12–14 year olds and adults.
Compared to previous research on spatial orientation, our studies in India and Nepal add several features that were not evident in earlier research: 1. Developmental data are systematically collected (whereas most previous research was carried out only with adults). 2. Data on several spatial encoding tasks are obtained with the same subjects, and various formats of task instructions are being tried out, showing that spatial encoding is task dependent. 3. Language production data is obtained with the same individuals as data on spatial encoding, so that the relationship between the two can be tested at the individual rather than only at the group level. 4. We include other spatial concept development tasks in order to study the generalization to the larger domain of spatial cognition. Cross-cultural psychology as a method is using cultural diversity as a ‘natural laboratory’. When working in one single cultural context (such as is typical of mainstream western developmental psychology), several variables of interest are often confounded (such as ontogenetic development and schooling, as mentioned earlier). Using carefully selected samples allows us to ‘unconfound’ these variables (Segall et al., 1999). The Indian subcontinent, while sharing the same macro-culture, is so diverse in terms of ecological
Spatial Language and Concept Development
247
settings, cultural practices and languages spoken that it allows for perfect cross-cultural research conditions. While these previous studies answered some questions, many more emerged. We therefore designed a new, still ongoing, multi-site, partly intra-cultural and partly comparative study, the first results of which are presented in this volume. A summary of the various locations is presented here, and the tasks used in every location are presented. Additional tasks specific to some projects and sample characteristics will be presented in each individual chapter.
RESEARCH DESIGN Locations The research is carried out in several locations in India and Nepal, the island of Bali in Indonesia, and in Geneva, Switzerland. The research design involves choosing several locations on the main criterion of the prevalence of an egocentric or a geocentric orientation system in language use.
Bali (Indonesia) In our previous study in India and Nepal, despite the geocentric language used predominantly by children and adults alike, the developmental trend of spatial encoding was found to be from relative to absolute, contrary to what we had found in Bali, where it seemed to be from absolute to relative. It was therefore very important for us to establish whether the Balinese results can be replicated with a larger sample.
Varanasi (India) As we have established in our previous project, children in Varanasi use a mixed frame of reference. However, some city children use predominantly geocentric language and some predominantly relative language. This is of interest for our study, since we will be able to compare these two sub-groups. In Varanasi, we also compare pupils from a Hindimedium school to students in Sanskrit schools.
Geneva (Switzerland): French-Speaking and Bilingual Children The predominant use of relative terms is characteristic of European languages, such as French, English and Dutch. In Europe, the spatial encoding tasks have been used only with
248 R.C. Mishra and Pierre R. Dasen Dutch adults (Brown and Levinson, 1993; Levinson, 2003), and developmental data with a European language is still needed for our study (but see Troadec and Martinot, 2001, for some data on French children). The techniques adapted in India in our previous study need to be used in the same format for cross-cultural comparisons, as well as for linking the study to mainstream developmental psychology. In Geneva, monolingual Frenchspeaking children are included as well as bilingual children (43 per cent of the pupils in Geneva schools have a language other than French as their first language, usually another European language).
Kathmandu (Nepal): Monolingual (Nepali) and Bilingual (English/Nepali) Children Nepali children who are fully educated in English (although they are always bilingual with Nepali), and attending English medium schools, could be expected to use a relative frame. These are compared to monolingual children in Nepali-medium schools.
Panditpur near Gorakhpur (India): Village Sample Using Relative Language The possibility seems to exist to find rural samples in India characterized by the predominant use of relative language. This would provide a challenge to the ecological hypothesis supported in our previous research. In our study, this is attempted in a village near Gorakhpur, but results are not reported in this volume.
TASKS The tasks that are used in all parts of our research are described below, while other tasks specific to various parts are described in the separate contributions.
Language Production (Elicitation) We record the spontaneous language used in three different situations, the first one involving movement through space, and the two others static displays. When egocentric or geocentric locators are produced, we record whether these are used correctly or not.
Route Description In our previous study, we obtained route descriptions from children by asking them to guide one experimenter who is blindfolded to move along a pathway laid out on the
Spatial Language and Concept Development
249
ground. While we have kept this procedure for the current research in Bali, we have adapted it for the other locations to using a model with the same outline (consisting of eight segments) on a board, some toys being placed along the route; the child is asked to verbally guide the experimenter who moves a car along the route.
Perspectives Task Three non-fronted objects are displayed on a board. The child is asked to describe the location of the objects from one position, then, moving around the display, from the opposite side, and then when the display is rotated by 180°.
Language on Spatial Encoding Tasks These are described subsequently. On the last two items of each task, the child is asked to tell the reason for his answer, and the language used is recorded.
Spatial Encoding Tasks These are tasks initially devised by CARG at the Max Planck Institute of Psycholinguistics in Nijmegen, and are described in Levinson (2003).
Animals in a Row This task presents the child with four animal models, three of them facing in one direction, and the fourth one placed at right angle to the three others. The child is asked to remember this alignment, and move on to another table with 180° rotation to align another set of the same animals the way they were shown before. Five trials, with animals oriented to right or left are given in a standard sequence, and then two trials with a 90° rotation. The way animals are aligned by the child is coded as indicating an absolute or relative encoding of the display. An innovation compared to our previous research is that we now use four animals instead of only three, which allows for a more valid distinction between intrinsic and absolute encoding (Levinson et al., 2002).
The Chips Task This task intends to measure visual recognition memory of two-dimensional shapes (small or large, red or blue squares) drawn on cards, two at a time. The child is shown five cards of a series, all with the same orientation, and asked to notice that all of them are similar. Then, one of the cards is rotated by 90°, and the child is asked to tell how it is different from other cards. Following this exercise, the child is presented with a card oriented in a particular direction by the configuration of the two squares, and is asked to remember this orientation. Then the child moves on to another table to choose from a set of four cards
250 R.C. Mishra and Pierre R. Dasen set out as a cross, one of them displaying the same spatial orientation as seen before. A series of practice trials are given before moving on to actual testing, which includes five trials with a 180° rotation and two at 90°.
Steve’s Maze This task consists of six pictures of landscapes that depict a house, rice fields, trees and an incomplete pathway. The child is presented with a picture and is told a story, showing the route that can take the child from the end of the drawn path back to the house. The child is asked to remember the path while moving on to another table (with 180° rotation) where three cards are given that show three different path segments. One of these represents a relative solution, another an absolute solution and the third one an irrelevant choice. Five trials are used.
RESEARCH QUESTIONS The research questions, being specific to each component of the project, will be presented in each individual chapter. Overall, we can refer to a recent paper by Majid et al. (2004) reviewing research about spatial FoR, in which, quite independently of our research venture, the authors conclude with the following list of ‘questions for future research’: 1. What are the neurocognitive underpinnings for linguistic frames of reference? How much plasticity is there? 2. How do children learn linguistic frames of reference? And when do linguistic frames of reference begin to influence spatial cognition? 3. What are the cognitive consequences of being a bilingual in languages that rely on different frames of reference? 4. Not all rural societies use an Absolute frame of reference, but urban languages appear to use a Relative frame of reference. Why is this? 5. What mechanisms do speakers of Absolute languages use to keep track of directions in the Absolute frame of reference? 6. Are speakers of Absolute languages better than speakers of Relative languages at viewindependent object recognition? (Majid et al., 2004: 113) It is rewarding to find that our project is designed to contribute partial answers to precisely five of the above questions formulated by experts other than ourselves. The following three chapters in this volume provide the initial data for some of the studies that were carried out since 2002. Mishra and Dasen (this volume) examine the possibility of a biological base to frames of spatial encoding by studying the link with hemispheric lateralization. Vajpayee, Dasen and Mishra (this volume) report a comparison
Spatial Language and Concept Development
251
between two samples with respectively. Sanskrit and Hindi medium schooling, that is, it deals with the cultural variable of religious practices that may foster geocentric encoding. Dasen and Wassmann (this volume) report the results from Bali and Geneva, examining in particular ecological (urban/rural) and socio-economic variables. All together, the current research demonstrates that spatial encoding and spatial language occur together in a complex set of biological and environmental variables that need to be studied within an eco-cultural theoretical framework (Dasen, 2003).
REFERENCES Berry, J., Y.P. Poortinga, M.H. Segall and P.R. Dasen. 2002. Cross-Cultural Psychology. Research and Applications (2nd revised edition). Cambridge: Cambridge University Press. Brown, P. and S.C. Levinson. 1993. ‘‘Uphill’ and ‘Downhill’ in Tzeltal’. Journal of Linguistic Anthropology, 3(1): 46–74. Dasen, P.R. 1993. ‘Schlusswort. Les Sciences Cognitive: Do They Shake Hands in the Middle?’ In J.Wassmann and P.R. Dasen (eds), Alltagswissen – Les Savoirs Quotidiens – Everyday Cognition (pp. 331–49). Fribourg: Presses De l’Université De Fribourg. ———. 2003. ‘Theoretical Frameworks in Cross-Cultural Developmental Psychology: An Attempt at Integration’. In T.S. Saraswathi (ed.), Cross-Cultural Perspectives in Human Development: Theory, Research, and Applications (pp. 128–65). New Delhi: Sage Publications. Dasen, P.R., R.C. Mishra and S. Niraula. 2004. ‘The Influence of Schooling on Cognitive Development: Spatial Language, Encoding and Concept Development in India and Nepal’. In B.N. Setiadi, A. Supratiknya, W.J. Lonner and Y.H. Poortinga (eds), Ongoing Themes in Psychology and Culture (pp. 223–37). Yogjakarta: Kanisius. de León, L. 1994. ‘Exploration in the Acquisition of Geocentric Location by Tzotzil Children’. Linguistics, 32: 857–84. ———. 1995. ‘The Development of Geocentric Location in Young Speakers of Guugu Yimithirr’. Nijmegen: CARG Working Paper No. 32. Gladwin, T. 1970. East is a Big Bird: Navigation and Logic on Puluwat Atoll. Cambridge, Massachusetts: Harvard University Press. Grabowski, J. 1999. ‘A Uniform Anthropomorphical Approach to the Human Conception of Dimensional Relations’. Spatial Cognition and Computation, 1(4): 349–63. Greenfield, P.M. 1976. ‘Cross-Cultural Research and Piagetian Theory: Paradox and Progress’. In K.F. Riegel and J.A. Meacham (eds), The Developing Individual in a Changing World, Vol. 1 (pp. 322–33). The Hague: Mouton. Gumperz, J.J. and S.C. Levinson (eds). 1996. Rethinking Linguistic Relativity. Cambridge: Cambridge University Press. Hutchins, E. 1995. Cognition in the Wild. Cambridge, Massachusetts: MIT Press. Levinson, S. 1996. ‘Frames of Reference and Molyneux’s Question: Cross-Linguistic Evidence’. In P. Bloom, M. Peterson, L. Nadel and M. Garrett (eds), Language and Space (pp. 109–69). Cambridge, Massachusetts: MIT Press. ———. 2003. Space in Language and Cognition. Cambridge: Cambridge University Press. Levinson, S., S. Kita, D. Haun and B. Rasch. 2002. ‘Returning the Tables: Language Affects Spatial Reasoning’. Cognition, 84: 155–88.
252 R.C. Mishra and Pierre R. Dasen Majid, A., M. Bowerman, S. Kita, D.B.M. Haun and S. Levinson. 2004. ‘Can Language Restructure Cognition? The Case for Space’. Trends in Cognitive Sciences, 8(3): 108–14. Miller, G. and P. Johnson-Laird. 1976. Language and Perception. Cambridge: Cambridge University Press. Mishra, R. 1997. ‘Cognition and Cognitive Development’. In J.W. Berry, P.R. Dasen, and T.S. Saraswathi (eds), Handbook of Cross-Cultural Psychology (2nd edition), Vol. 2, Basic Processes and Human Development (pp. 143–76). Boston: Allyn & Bacon. Mishra, R.C. and P.R. Dasen. 2005. ‘Spatial Language and Cognitive Development in India: An Urban/Rural Comparison’. In W. Friedlmeier, P. Chakkarath, and B. Schwarz (eds), Culture and Human Development: The Importance of Cross-Cultural Research to the Social Sciences (In Honour of Gisela Trommsdorff’s 60th Birthday) (pp. 153–79). Hove, UK: Psychology Press. Mishra, R.C., P.R. Dasen and S. Niraula. 2003. ‘Ecology, Language, and Performance on Spatial Cognitive Tasks’. International Journal of Psychology, 38: 366–83. Niraula, S. and R.C. Mishra. 2001. ‘Spatial Orientation of the Newar Children in Nepal’. Social Science International, 17: 36–48. Niraula, S., R.C. Mishra and P.R. Dasen. 2004. ‘Linguistic Relativity and Spatial Concept Development in Nepal’. Psychology and Developing Societies, 17: 99–124. Piaget, J. and B. Inhelder. 1956. The Child’s Conception of Space. London: Routledge and Kegan Paul. Pinxten, R., I. Van Dooren and F. Harvey. 1983. Anthropology of Space: Explorations into the Natural Philosophy and Semantics of the Navajo. Philadelphia: University of Pennsylvania Press. Segall, M.H., P.R. Dasen, J.W. Berry and Y.H. Poortinga. 1999. Human Behavior in Global Perspective: An Introduction to Cross-Cultural Psychology (2nd revised edition). Boston: Allyn and Bacon. Taylor, H.A. and B. Tversky. 1996. ‘Perspective in Spatial Descriptions’. Journal of Memory and Language, 35: 371–91. Troadec, B. 1999. Le Développement De La Pensée Chez L’enfant. Catégorisation Et Cultures. Toulouse: Presses Universitaires du Mirail. Troadec, B. and C. Martinot. 2001. De La Variabilité Interindividuelle À La Variabilité Interculturelle: L’exemple Du Développement De La Représentation De L’espace Chez L’enfant De Tahiti (Océanie), Paper presented at the 8ème Congrès International De l’ARIC, 24–28 September, Genève. Available from http://www.unige.ch/fapse/SSE/groups/aric/Textes/Troadec%20Martinot.pdf. Wassmann, J. 1994. ‘The Yupno as Post-Newtonian Scientists. The Question of What is ‘Natural’ In Spatial Description’. Man, 29: 1–24. ———. 1995. ‘The Final Requiem for the Omniscient Informant? An Interdisciplinary Approach to Everyday Cognition’. Culture and Psychology, 1: 167–201. ———. 1997. ‘Finding the Right Path. The Route Knowledge of the Yupno of Papua New Guinea’. In G. Senft (ed.), Referring to Space (pp. 143–74). Oxford: Oxford University Press (Oxford Studies in Anthropological Linguistics). Wassmann, J. and P.R. Dasen. 1994a. ‘‘Hot’ and ‘Cold’: Classification and Sorting Among the Yupno of Papua New Guinea’. International Journal of Psychology, 29: 19–38. ———. 1994b. ‘Yupno Number System and Counting’. Journal of Cross-Cultural Psychology, 25: 78–94. ———. 1998. ‘Balinese Spatial Orientation: Some Empirical Evidence for Moderate Linguistic Relativity’. The Journal of the Royal Anthropological Institute, Incorporating Man (N.S.), 4: 689–711. Werner, S. and C. Hubel. 1999. ‘Spatial Reference Systems’. Spatial Cognition and Computation (Special Issue), 1(4): iii–vii.
Chapter 18 Spatial Encoding: A Comparison of Sanskrit- and Hindi-Medium Schools Aparna Vajpayee, Pieree R. Dasen and R.C. Mishra
INTRODUCTION
T
his study examines spatial frames of encoding, spatial language use and knowledge of the spatial orientation system in students attending either Sanskrit or Hindi-medium schools. While the former are bilingual (in Hindi and Sanskrit), this is not our main issue; rather, we focus on the different socialization patterns that go with these two types of school systems (see Broyon, 2004; Mishra and Vajpayee, 2004). Anthropologists and linguists, such as Levinson (2003), attribute a fundamental role to language in the construction of human culture. According to Levinson: We are a species lacking many wonderful modular endowments like echo-location or innate fixed-bearing dead-reckoning systems, but with one spectacular specialization, namely language, which has come to play a dominant role in our psyche. Language has an interstitial status—it is a public, shared, cultural representation system at the same time that it is a private, internal representation system. And some choices made at the cultural, external, variable level come to ramify right through our inner representational systems. … We normally do not have a choice between the systems—the linguistic traditions in our local communities have long ago opted for one system or the other. (ibid.: 290–1)
While language has without any doubt played a major role in hominization, the word ‘language’ in the above quote could easily be replaced by ‘religion’. The purpose of this chapter is to examine the role a particular cultural feature, such as religious education, can play in influencing the complex structure of the spatial orientation system, spatial language and spatial encoding. We take advantage of the existence in Varanasi of Sanskrit schools for both girls and boys to compare 155 students of these schools to the 221 pupils following the more common Hindi-medium stream of schooling. Even though the latter
254 Aparna Vajpayee et al. are also of Hindu faith, Sanskrit education entails both a cosmology and practices that are expected to be relevant for geocentric spatial orientation. The history and current status of Sanskrit schooling in India have been described by Mishra and Vajpayee (2004), while Broyon (2004) provides a case study of the particular schools used in this research.
SANSKRIT SPATIAL COSMOLOGY The most ancient description of spatial language in Sanskrit literature is found in the Rig Veda, which is the oldest religious scripture in India. The hymns of the Rig Veda are composed in praise of the deities that were originally conceived as presiding over natural phenomena. These include such deities as fire, wind, water, rain, sun, moon, and so on. Among these, the sun (surya) is regarded as the soul of all living beings, whether they are a part of the moving or stationary world. The rising and setting of the sun provide us with a measure of time, the day and night, which together form a day. There are approximately 30 days in a month, 12 months in a year with 360 days. In the Rig Veda hymns, the time frame is described as ‘the wheel of the sun with its 360 spokes’. According to the hymns of the Rig Veda, the gods also produce an order in space. The sun is the lord of heavens; its movement from the northern to the southern corner gives us a year. Thus, in the Indian setting, all divisions of time and space are completely based on the position of the sun. Anchored by the sun, other gods are also manifested in different directions (mandals). Sanskrit literature presents us with a description of 10 directions: uttar (north), dakshin (south), purva (east), pashchim (west), ishan (northeast), agneya (southeast), vayavya (northwest), nairitya (southwest), urdhava (up) and adhar (down). Each direction is identified with a particular god. Mental representations or cognitive maps of the space guided by this particular picture constitute the viewer’s perspective to understand the surrounding world. The concept of space and the person’s relative position in it is also guided by the philosophical questions of ‘who am I?‘, and ‘where do I stand?’. Order in the perception of the world starts from omkara, that represents all gods together, or the highest God (the Brahman), or the divine soul in man. This orients persons towards more ‘global cues’ and allows to place themselves in the world wherever they go. Space can be perceived as surrounding or engulfing people with objects such as buildings or furniture, or space can be perceived as separate from the individual with reference to the outside space such as the sun, or scenic views of mountains and rivers, and so on. These notions imbibe the concept of cosmos in any given culture. Thus, the viewers’ perspective to resolve the spatial world can represent the same world in different manners with multiple modalities and frames. The experience with different sets of mind or expertise in philosophy can affect the meaning and the ways to understand the world. Thus, a child can experience space in various ways, and the environment can be viewed from different perspectives depending on the child’s experiences in a particular culture.
Spatial Encoding
255
Knowledge about space can be obtained directly through actual navigation (for example, walking, running, driving), or indirectly through depictions and descriptions. Indirect experiences also include the speech of caretakers (parents, teachers) in the form of specific social and communicative practices, and artefacts. To what extent does spatial language depend on paralinguistic spatial schematization (Mandler, 1996)? Do children learning languages with widely differing spatial semantic systems display different patterns of acquisition (Bowerman, 1996)? What role as a spatial ‘source domain’ does the representation of the human body (as a person engulfed in nearby surroundings or as a person using more global cues) play in the acquisition of spatial concepts and language? These are some questions that lead us to the present study involving samples of Hindi- and Sanskrit-school children.
HINDI- AND SANSKRIT-SCHOOL CHILDREN The ability to acquire and use spatial knowledge varies over the lifespan, and studies show a definite trend in spatial abilities in relation to age and gender. It has been shown that women have a better memory for object names and locations, landmarks on navigational routes, and memory for details in drawings. Males, in contrast, have better mental rotation ability and a better memory for overall spatial layouts in route navigation (Humphrey, 1997). There is evidence in other studies that females tend to use cognitive strategies paying attention to local detail, while males demonstrate cognitive strategies towards the overall spatial structure (Humphrey, 1997). If we observe the life of both Hindi- and Sanskrit-medium girls, a clear difference in socialization is obvious. Boys of both schools are allowed to go out frequently as compared to girls. Sanskrit schoolboys are allowed to go out to buy things and to take baths in the river Ganges. They go to temples on Mondays, Tuesdays and Saturdays. On the other hand, the girls in Sanskrit school are not allowed to go out. Interaction with males is also generally prohibited. They depend entirely on teachers for the necessities of everyday life. This is true for girls of Hindi-medium schools as well. By the age of 11–12 they are also not allowed to go out freely. After grade 5, the schools generally make separate arrangements for teaching of boys and girls.
METHODS The background of this particular study, its general theoretical framework, and a description of the tasks and tests that have been used in this cross-cultural, collaborative project have been presented earlier. The sample characteristics are presented in Table 18.1 (see also Mishra and Dasen, this volume).
256 Aparna Vajpayee et al. TABLE 18.1
Sample characteristics Gender Age
Hindi-medium schools
Sanskrit-medium schools
Total
Boys
Girls
11
9
30
39
12
26
59
85
13
9
46
55
14
5
31
36
15
0
6
6
Total
49
172
221
10
2
0
2
11
6
3
9
12
35
5
40
13
49
6
55
14
30
7
37
15
3
9
12
Total
125
30
155
376
The present study was carried out in one Hindi-medium school and three Sanskritmedium schools. These two types of schools widely differ with respect to the philosophy and pedagogy of education. A brief descriptions of these schools is given below.
Sanskrit-Medium Schools for Boys The basic philosophy of Sanskrit education is the simultaneous development of physical, mental, psychic and spiritual capacities of children. This can be achieved by knowledge not only of subjects like history, geography and mathematics, but also by physical and mental exercises, practice of rituals, meditation and yoga. Satsang (contact with noble people) is also considered an important means of attainment of knowledge and development of mind, to distinguish not only between good and bad or desirable and undesirable education, but also between truth and non-truth, as well as learn what to accept or reject in life. Thus, the aim of these institutions is to provide children with an ideal education of Gurukul (an ancient institution of education), which allows them to fulfill various responsibilities of living the life of a true Brahmin. The message is that worldly duties are not bondages that must be severed; the world is the best teacher, and one should not renounce the world in search of a supernatural life or the realization of the supreme God. Several virtues of life such as respect for teachers and parents, mercy, love, generosity, patience, tolerance, purity, prudence and right judgement are deliberately transmitted.
Spatial Encoding
257
On the other hand, if someone has an orientation towards renunciation, he can opt for that life with the permission of teachers and parents. Thus, one can find in these schools that students study together, but that some live a life detached from worldly affairs. They learn Sanskrit to study Vedas, religious scriptures, and other books for self analysis and self correction. By observing sadachara (moral behaviour), and self-imposed restrictions on physical, mental and verbal behaviours, they divert their energies to the attainment of an eternal state of bliss, called sachchidanand. Most of the parents today want their children to attend modern schools because they provide several options for livelihood. Those who cannot afford the cost of education in modern schools send children to Sanskrit schools, in which the attainment of knowledge is now a less valued goal than the search for employment (see Mishra and Vajpayee, 2004).
Mumukshu Bhawan This school is run by Kashi Mumukshu Bhawan Sabha, which was established in 1920. Spread over five acres of land, it comprises a Veda Vedang Mahavidyalaya (for higher level education) and a Prathama Vidyalaya (primary school for younger boys), besides a math (residence for aged saints). The school premise also houses charitable Ayurvedic and Homeopathic dispensaries, a guest-house and a couple of temples. The primary school enrols 50 boys as students whose board and lodging is fully taken care of by the trust. The boys live in a hostel in care of the family of a school teacher. The students generally come from lower middle-class Brahmin families living in rural areas near Varanasi. The school imparts education in subjects like history, geography, literature, philosophy, religion, including a variety of rituals, meditation and yoga.
Nand Lal Bajoria Sanskrit Mahavidyalaya This school was established by a lady in memory of her husband, Late Nand Lal Bajoria, and is operated by a trust. The trust includes a Sanskrit school, a school for blind, a hospital for the poor and houses for teachers. The trust fully supports the board and lodging of 80 Brahmin children for the learning of Sanskrit for two years. Children mostly come from village areas of eastern Uttar Pradesh and Bihar. The school teaches not only Sanskrit language and a variety of rituals and religious practices, but also subjects like history, geography and literature. In comparison to Mumukshu Bhawan, children of this school live in physically and economically less privileged conditions.
Sanskrit-Medium Schools for Girls These schools are meant exclusively for the education of girls. They are different from boys’ schools with respect to the degree of freedom allowed to students for outside movement.
258 Aparna Vajpayee et al. Panini Kanya Mahavidyaliya This school was founded by the late Prajna Devi with the help of the king of Varanasi. It was started as a small school that was expanded over the last three decades. The school campus includes a school building, a library, a hostel, residence of teachers, a guest-house, a Yagyashala (place for worship), an armoury, a cow yard, and some fields for Ayurvedic plants and a kitchen garden. The school admits students from all parts of India regardless of caste, class or social background. The students are adopted as a children of Gurukul family, and are renamed there. The school comprises 70 students and five full-time teachers. Education in this school is based on Vedas, with great emphasis on learning of grammar (developed by an ancient scholar, Panini), which includes 4,000 formulae of Sanskrit grammar. The curriculum includes Sanskrit language, Sanskrit grammar, Vedas, Ayurveda, history, geography, English, marshal arts, physical education, yoga, classical music and dance, cooking, a bit of animal husbandry, agriculture and horticulture, along with rituals of different ceremonies of life (for example, birth ceremony, sacred thread ceremony, and so on). The primary aim of this school is to save the cultural heritage of Sanskrit language, and make women aware of their strengths. While the school combines job-oriented with self-oriented education, the ultimate aim is to help a student discover her inner spirit so that the hidden mystery of the eternal soul could be revealed.
Hindi-Medium School: Malaviya Shiksha Niketan This imparts education through the medium of Hindi language; it is a semi-government co-educational school. Established some 15 years ago, the school provides education to about 2,000 day scholars. Since the school is a profit making organization, the students pay a good amount of money as fees. The syllabi of the school are according to the norms prescribed by the state government. Hence, all subjects including the subjects of science are taught in the school. The school has no choice in this respect. The philosophy of education is to make students independent members of the society, so that as adults they are able to manage life on their own. The focus is on the overall personality development of children in order to ensure their success in life. Unlike the Sanskrit school, the education does not focus on the development of hidden potentials.
RESULTS On the Animals task (seven items), the Sanskrit school group produces an average of 4.7 geocentric encoding items compared to only about three in the Hindi-medium school. The contrast is much the same, or possibly even more striking, for Chips (seven items, including two with a 90-degree rotation) where the respective averages are 5.6 and 4.
Spatial Encoding
259
(Figures 18.1 and 18.2). There is not much of an age trend, and indeed, for this study, we are not focussing particularly on developmental aspects. There is the now commonly found effect of task specificity, namely, a predominantly egocentric encoding on Steve’s Maze produced by the more iconic nature of the task, and less difference between the schools (Figure 18.3), although the means are statistically significant for all three tasks (see Table 18.2). Sanskrit school children have, as expected from their training, a much better knowledge of the North-South-East-West (NSEW) orientation system, both outside and inside a room. In fact, their performance is almost at ceiling level, and no doubt we could also have asked for the specifically named intermediate spatial directions. However, they are also more correct in their knowledge of right, left, front and back (RLFB). On the three language elicitations tasks, Sanskrit school children spontaneously use more (correct) geocentric language and less egocentric language. In fact, on the Route display, more than half of the Hindi-medium pupils do not use geocentric language at all, but use RLFB (as expected from our previous results in Varanasi city [see Mishra et al., 2003]). This task, being close to the practice of moving about in city streets, is expected to elicit more egocentric descriptions. However, only 10 per cent of Sanskrit pupils use egocentric language for the Route. On the Perspective display, which is much more static, and on the six items of the encoding tasks for which language was recorded, all Sanskrit pupils except five use the geocentric language systematically, and it is also more common with the Hindi-medium children, 66 of whom nevertheless use egocentric language at least occasionally. Note that on these tasks, virtually no other spatial references (such as intrinsic, or landmarks) are used. FIGURE 18.1 four animals
Animals: number of items with completely geocentric encoding, out of seven items, using
260 Aparna Vajpayee et al. FIGURE 18.2
Chips: number of items with completely geocentric encoding, out of seven items
7.0 6.0
Mean ChipSG7
5.0 4.0 3.0 2.0
Sample Hindi-medium schools
1.0
Sanskrit schools
0.0 10
FIGURE 18.3
11
12
Age
13
14
15
Steve’s Maze: number of items with completely geocentric encoding, out of five items
Spatial Encoding TABLE 18.2
261
One-way ANOVA comparing Hindi-medium (H) and Sanskrit schools (S) School type
Mean
SD
H
2.94
1.80
S
4.71
1.83
H
4.07
2.06
S
5.57
1.73
H
1.84
1.16
S
2.22
1.25
H
4.30
3.44
S
7.26
2.02
H
6.93
1.71
S
7.65
1.21
H
1.43
2.50
S
5.50
2.59
H
2.19
2.21
S
0.72
1.91
G+ language on Perspectives
H
3.54
3.73
S
7.94
2.49
E+ language on Perspectives
H
1.34
2.58
S
0.22
1.26
G+ language on encoding tasks
H
1.49
1.73
S
2.22
1.88
E+ language on encoding tasks
H
0.59
1.08
S
0.005
0.26
F
Sig.
86.401
0.000
55.148
0.000
9.311
0.002
92.200
0.000
19.970
0.000
234.446
0.000
44.998
0.000
164.621
0.000
25.221
0.000
14.949
0.000
37.467
0.000
Animals GG 7 items
Chips
Steve’s Maze
Knowledge of NSEW
Knowledge of RLFB
G+ language on Route
E+ language on Route
Note: N = 221 Hindi-medium schools, N = 155 S.
Gender Since we were able to include girls and boys in both types of schools, gender is one variable that can be studied in more detail; however, because of gender bias in the sampling in
262 Aparna Vajpayee et al. the two schools, analyses have to be carried out separately for each school type. In our previous research with children aged 4 to 12 years (sometimes 14) in Bali, India, Nepal and Switzerland, we never found any significant gender effect. Here, on the other hand, gender does show up as an effect, although only a slight one. This could be expected, since we are now dealing with adolescents, for whom gender introduces markedly different socialization practices (as well as, of course, biological sex differences). On spatial performance tests, it is well-known that gender becomes a significant factor only with age. In the Sanskrit school, a significant correlation is found between gender and encoding (boys produce more geocentric encoding). That boys use more geocentric encoding than girls could be due to the fact that, although the curriculum is the same in the school, the boys’ experience of the outside world is quite different. Most of them have come from rural areas, where they have attended regular government schools, while the girls have joined their school at an early age, and are then separated from their families. The girls are also not allowed to leave the compound of the school, while the boys can move about more freely, at least to visit temples. In the Hindi-medium school, boys use more (correct) egocentric language. Generally speaking, we conclude from these findings that gender effects are in fact not very important compared to the effects of school type, for which our hypothesis is entirely confirmed.
DISCUSSION AND CONCLUSIONS Sanskrit school children get more training with geocentric cues to orient themselves in their environment, but all Indian children also get training of egocentric cues. From the very beginning they are asked to use the right hand for auspicious activities (such as eating, writing, rituals, and so on), and they get training not to use the left hand. The use of the left hand is considered a bad habit for this kind of tasks. In the same way, they are asked to use the left hand for inauspicious activities (for example, they use only the left hand in the toilet). While they perform rituals, Sanskrit school children change their left and right hands several times according to instructions or the principles of a particular ritual. In our day-to-day life, we go out and find ourselves in previously visited places of interest. For that, we need to maintain a mental map and the skill to navigate. This requires spatial comprehension with an ability to perceive, understand, remember and recall spatial features for future use. In our route task, as in real-life situations, one has to use many salient cues to follow the path. Cues might be a landmark, left and right turns, or can be geocentric. How people relate themselves to the physical world depends upon their previous experience and understanding of the environment. They are likely to find spatial information most useful in navigation in a particular context, and are likely to avoid irrelevant or less important information. All the children of Sanskrit schools have come from village surroundings, where geocentric cues are more useful. In cities,
Spatial Encoding
263
there are rules to follow for pedestrians based on the left and right sides of roads. Thus, to execute a route task one can extract information according to one’s preference and previous experience. Therefore, when we compare the performance of Hindi and Sanskrit school children, we need to focus not only on what is encoded, but also how it is encoded. Knowledge about space can be obtained directly through actual navigation, which includes locomotion through the environment (for example, walking, running, driving and so on) or through viewing stationary scenic views. There are also indirect ways to acquire spatial information through depictions and descriptions (for example, maps, pictures, photographs and written descriptions). We presume that the Sanskrit school children acquired geocentric knowledge through their village background and with direct use in their daily life in school, while the children in the Hindi-medium school may have acquired this information more from depiction or description. Spatial language and conceptualization has been claimed to be of fundamental importance in the development of cognition. According to Levinson (2003), linguistic variation gives rise to differences in non-linguistic cognitive processes of people speaking different languages. If we compare Hindi and Sanskrit School children, this is illustrated by the fact that Sanskrit school children are more geocentric in linguistic as well as nonlinguistic cognitive tasks.
REFERENCES Bowerman, M. 1996. ‘Learning How to Structure Space for Language: A Cross-Linguistic Perspective’. In P. Bloom, M. Peterson, L. Nadel and M. Garrrett (eds), Language and Space (pp. 385–436). Cambridge, Massachusetts: MIT Press. Broyon, M.A. 2004. ‘L’éducation Sanskrite À Bénarès, Enjeu D’une Société Qui Oscille Entre Tradition Et Transition’. In A. Akkari and P.R. Dasen (eds), Pédagogues Et Pédagogies Du Sud (pp. 231–50). Paris: L’Harmattan. Humphrey, D. 1997. ‘Preferences in Symmetries and Symmetries in Drawings: Asymmetries Between Ages and Sexes’. Empirical Studies of the Arts, 15: 41–60. Levinson, S. 2003. Space in Language and Cognition: Explorations in Cognitive Diversity. Cambridge: Cambridge University Press. Mandler, J. 1996. ‘Preverbal Presentation and Language’. In P. Bloom, M. Peterson, L. Nadel and M. Garrrett (eds), Language and Space, (pp. 365–83). Cambridge, Massachusetts: MIT Press. Mishra, R.C. and A. Vajpayee. 2004. ‘Les Écoles Sanskrites en Inde [Sanskrit Schools in India]’. In A. Akkari and P.R. Dasen (eds), Pédagogies Et Pédagogues Du Sud (pp. 207–30). Paris: L’Harmattan. Mishra, R.C., P.R. Dasen and S. Niraula. 2003. Ecology, Language, and Performance on Spatial Cognitive Tasks. International Journal of Psychology, 38(6): 366–83.
Chapter 19 A Cross-Cultural Comparison of Spatial Language and Encoding in Bali and Geneva Pierre R. Dasen and Jürg Wassmann
INTRODUCTION
O
ur first study in Bali carried out in 1994 (Wassmann and Dasen, 1998), and subsequent research in India and Nepal (Mishra et al., 2003), have been reviewed briefly in the introduction to this volume (Mishra and Dasen). Overall, we found a strong tendency towards the use of a geocentric frame of spatial reference (FoR) in all of our samples, taken from societies in which the language allows a choice between all three frames, but in which a geocentric orientation system is generally used (North-South-East-West [NSEW] in India and Nepal, and also Up–Down in the latter). Large differences between encoding tasks showed that the choice of a FoR is strongly task dependent, and differences between rural and urban groups showed the influence of ecological settings (Mishra and Dasen, 2005). While differences between schooled and unschooled children were also explored (Dasen et al., 2004), we did not take socio-economic indicators or other family background variables into account. We also did not find the same results in India and Nepal as in Bali in terms of age trends: while absolute encoding seemed to decrease with age in Bali, it increased in all three samples in India and Nepal. One of the first endeavours in the current project was therefore to replicate the study in Bali itself.
BALI 2002, A REPLICATION STUDY One of the limitations of our earlier study in Bali was the small sample size. This was due to the fact that we carried out the research in a fairly remote area of Bali, in the small
265
A Cross-Cultural Comparison of Spatial Language and Encoding
village of Bunutan, and we did not have access to the local school, which would have provided us with a larger sample of children. In 2002, we were able to work with much larger samples, in schools of two different locations, as well as testing again a small sample of young children in the Bunutan area. The number of children tested is given in Table 19.1. The city sample was obtained in a private, experimental school attached to the only teachers’ training college of Bali, in the city of Singaraja, on the North coast of Bali. The pupils attending this school were mainly from middle-class families. TABLE 19.1
Sample characteristics of studies in Bali 2002 and 1994
Age groups
4–5
6–8
9–11
12
Total 72
Singaraja (city)
15
24
28
5
Sambangan (village)
16
35
37
10
Bunutan (remote village)
15
18
Total
46
77
65
15
203
9
8
7
14
38
Bunutan (remote village), 1994
98 33
The village of Sambangan was situated in a rural area about 15 km from Singaraja, on a north facing slope of the mountain, with a clear view of the sea in the distance. The parents of the children were mainly farmers, except for the 4- and 5-year-olds who attended a preschool in the same locality, but closer to town. Information on the family background of each child was obtained through interviews with teachers, or a questionnaire sent home to the parents. For the background variables of mother and father’s level of education, occupation, and access to media such as newspapers, radio and TV, all t-tests comparing the two locations were statistically significant at the 0.01 level. The language of instruction is Indonesian in both schools, so all children were bilingual; however, Balinese is the common language used in the community in Sambangan, while Indonesian is spoken more in the city.
METHODS The tasks used in this study were divided into two main categories, language elicitation and spatial encoding. They have been described by Mishra and Dasen (this volume). All instructions were given in Balinese by two local research assistants. The testing was carried out in various rooms in the school or, for the youngest age group, outside in front of the pre-schools. Towards the end of our fieldwork period in Bali, we also returned to the Bunutan area where we had carried out our initial study in 1994, and tested a few children aged four to six years with the Animals task only; this was done at the children’s home, with the help of local interpreters and often with a large group of onlookers.
266 Pierre R. Dasen and Jürg Wassmann We had requested the teachers to select only children whom they knew to speak Balinese in their homes, but a later analysis of home background data and the language they spontaneously used on some of our tasks proved that this was partly unsuccessful. In the village, 70 per cent of the children use Balinese only, while this is the case of only 30 per cent of the children in the city. Teachers were also asked to provide some background information, for which they sometimes consulted the parents. Information was collected on mother’s and father’s education and occupation, media in the home (newspapers, radio, TV), language spoken in the home, and contact with the city in rural areas and contact with village in the city.
RESULTS Knowledge of Orientation System Do the children actually know the Balinese orientation system? And can they point to the directions correctly? For those who systematically speak Balinese, this is indeed the case even for the youngest children, while those who mix the two languages know it perfectly only by age nine or 10 and those who speak Indonesian do not perform above chance level. More detailed analyses show that for the Balinese speakers, all the children but three, if they know the kaja–kelod axis, they also know the kangin–kauh axis, and if they know it outside, they also know it inside a room. The correlation between knowledge outside and inside is 0.86, which is significant at the 0.01 level. This capacity to carry the orientation system with them inside a room is indeed one of the most striking features of a geocentric orientation system, and warrants a further study of how exactly it is done (such a study is in progress in Varanasi). Levinson (2003) mentions a constant dead-reckoning process that becomes an unconscious routine, and functions even in unfamiliar surroundings. In the present case, the rooms we used were familiar to the children and there were open windows; it would have to be established whether they could also carry the orientation system into an unfamiliar room without any possible visual cues. How do the children learn the Indonesian cardinal directions (that is, NSEW as opposed to kaja–kelod)? Contrary to what might have been expected, it is not the Indonesianspeaking children who learn these first, but the Balinese-speaking children. This is no doubt because they are socialized early to use the Balinese orientation system, and therefore find it easy to transfer this knowledge to the Indonesian cardinal directions. Their performance, however, is far from perfect, and the NSEW system is completely mastered by all children only by the age of 12. Again, further analyses show that if the children know the NSEW system outside, they can also use it inside; the correlation between performance outside and inside is 0.83,
A Cross-Cultural Comparison of Spatial Language and Encoding
267
which is significant at the 0.01 level, and this is the case even if age and language groups are controlled. Balinese children are also taught very early on to distinguish their right and left hands, because the appropriate use of each has to be acquired. Right, left, front and back, in Balinese and in Indonesian, are known by all the children by age nine in both languages, but before that age, the Indonesian-speaking children still have some problems in Balinese, and the Balinese speaking 4–5-year-olds the same in Indonesian.
Spontaneous Use of Language While describing motion along a road, the younger children of our study, up to age 11, use mainly geocentric terms if they are Balinese speaking (or mix the two languages), and they never use relative terms; they do to some extent (on the average on half of the segments of the road) if they speak Indonesian. The 12-year-olds of all groups use geocentric terms almost exclusively. In other words, there is a trend for the Indonesian-speaking children to change from relative to geocentric terms, while the Balinese-speaking children do not show any age trend. The same is true for language use on perspectives, except that the use of relative terms is less frequent even for the Indonesian-speaking children. The Indonesian-speaking children up to age eight also use some intrinsic locatives, which is not the case for the Balinese-speaking children. Landmarks (whether situation specific or conventional) are never used. As to language used on items four and five of each of the three encoding tasks, it is generally geocentric; in the Balinese-speaking group, this is almost exclusive, while in the Indonesian speaking and mixed language groups, some relative language occurs up to age-group nine to 11, as well as some intrinsic and deictic terms. In sum, the various language tasks complement each other to show a systematic difference between the Indonesian and the Balinese-speaking groups. In the former, while geocentric language is present even in the youngest age group, some of the children start using some intrinsic, deictic and relative language, and do this up to about age 10. In contrast, in the Balinese-speaking group, geocentric language is an almost exclusive usage from 4 to 12 years. To obtain summary measures for egocentric and geocentric language over the four language tasks, we computed two variables adding up the correct knowledge of the respective terms with their correct use on the Road and Perspectives, and their use on the Nijmegen tasks. An age effect is seen in these summary measures in the Indonesianspeaking group, and in the city, with a sharp increase of geocentric language with age, while in the Balinese-speaking group and the village, geocentric language is already established early and has less room for change.
268 Pierre R. Dasen and Jürg Wassmann Spatial Encoding Let us start with the simplest task, Animals, the only one that was used systematically in all of our previous research and in all three samples in the present one, and look at the results with three animals only (since that was the format used in previous studies). The results are presented in Figure 19.1. As we can see, the results in the three samples are all marked by a very high RelativeAbsolute (RA)-gradient, all between 50 and 100 per cent. Similarly to our previous results in India and Nepal, the rural samples have on the average a higher RA-gradient over the whole age span than the city sample. Bunutan, the remote village location where we used only this single task, has the highest proportion of absolute encoding in the comparison; however, compared to the 1994 research, the age trend is different: instead of a 100 per cent absolute encoding in the 4–5-year-old age group followed by a slight increase of relative encoding, we now have the same age trends we found in India and Nepal, namely, a steady increase of absolute encoding with age. FIGURE 19.1
Mean items with absolute encoding for four animals (seven items) in three samples of 2002 study
Compared to our previous study in 1994, the proportion of absolute encoding (RAgradient) is lower, but we are now using a more stringent test of absolute encoding, namely, taking four animals into account (the fourth one at right angle with the three others) and seven items, including two with a rotation of 90° (as described in Mishra and Dasen,
A Cross-Cultural Comparison of Spatial Language and Encoding
269
this volume). The correlation between the simple RA-gradient (three animals, five items) and the more stringent measure is 0.85 (which, with an N of 203, is highly significant). This confirmation with more stringent conditions comforts us in the affirmation that even our previous measures with three animals were reflecting mainly absolute and not intrinsic encoding (see Levinson et al., 2002; Li and Gleitman, 2002 for the controversy over this issue). On Chips and on Steve’s Maze, we find the same difference between the village and the city samples, with a higher RA gradient in the village. Steve’s Maze produces, as it did in all our previous research, a much lower RA-gradient; in fact, there is on this task more relative than absolute encoding. As we have hypothesized before (Mishra et al., 2003; Wassmann and Dasen, 1998), this is no doubt due to the fact that it is easy to verbalize the Animals and Chips task in a geocentric form (‘They all look kangin’, ‘The red square is kaja’), while it would take several sentences to do this for the segment of path in Steve’s Maze, and it is therefore easier to remember it as a shape (that is inherently egocentric). Another of our previous findings that is therefore strongly confirmed with these results is the task specificity of frames of reference used for encoding: individuals do not necessarily use the same frame under all conditions, but ‘choose’ one frame over another, this choice being of course quite unconscious. The mean RA-gradients in each age group may of course hide the fact that some children indeed use an absolute frame systematically (for all or at least most items), while others do not; in other words, that there may be two different pathways or cognitive styles. This question is related to the fact that we found, in Bunutan in 1994, 100 per cent absolute encoding by the younger children. Despite lower mean values, do we also find such a sub-group within our present sample? In Bunutan, if we look at the youngest children (four to five years), we find that half of them are indeed in the systematically absolute group, the other giving some mixture of encodings (but never systematically relative). The same conclusion would be reached if we looked at the Animals task in the way it was presented in 1994, that is with three animals only. If we also include slightly older children, up to age eight, the percentage of children in the systematic A group is somewhat lower (41 per cent) but still quite important. We can therefore conclude that very young children who start with a completely absolute frame of reference (FoR) indeed do exist. The difference with the 1994 results is that they do not represent the complete cohort, but only about half of it. How could this be explained? Possibly, it could be due to sampling, since we are dealing with very small numbers; out of nine, it is not impossible that we sampled by chance nine from that group, instead of five or four. We cannot rule this out, but it is also possible that social change in the eight years separating the two studies could have introduced more of the urban characteristics that seem to go with more relative encoding. Indeed, the Bunutan region has become an attractive site for tourists, with many lodges, dive-shops and even an Internet café, while it was very isolated in 1994.
270 Pierre R. Dasen and Jürg Wassmann If we look at the other two locations, in the village of Sambangan we also find a systematic absolute group in the younger ages, representing approximately 1/4 of the children, but there is also another group of the same size that is using a systematically relative encoding. In the city, there is no systematically absolute group, only a relative one. Of 27 children in the absolute group (ages four to eight), none speaks Indonesian and only one lives in the city; on the other hand, the systematically relative encoding group of 16 children is rather mixed, coming from both the city and Sambangan, and from the three language groups. This pattern of results seems to show that we indeed have a socio-historical process going on, where the more traditional Balinese cognitive style (or developmental path?) is being replaced slowly through the processes of acculturation and globalization. If that is so, and since we have collected other background information on the children’s families in the main two samples, there may be other social indicators that we could link to the two styles of encoding. One difference with the results we obtained previously in India and Nepal is that in our present results, the average RA-gradient for Chips is the same as that for Animals, or even higher, particularly if we take the more stringent measure for Animals, while in our previous research in India and Nepal, Chips was always intermediate between Animals and Steve’s Maze. Why this should be so is not immediately obvious, but may be due to the fact that we increased the complexity (and hence possibly the memory load) of the Animals task (even if coding only for the three animals in a row, the task did comprise a fourth animal) and decreased that of Chips (using only squares instead of squares and circles). As acknowledged by Levinson (2003), memory load does influence the encoding on these tasks.
Language and Encoding In our first study in Bali, we had collected spatial language data, but were not able to relate each individual’s language use to the FoR on the encoding tasks. When we did this in India and Nepal, we found, surprisingly, that there was hardly any relationship between the two (while geocentric language was found to be related to performance on spatial cognitive tasks). So we concluded that the hypothesis of linguistic relativism received support at the group level (when geocentric language is used as the adult group norm, encoding also tends to be absolute), but not at the individual level. We can now explore this issue with the new Bali data. Correlation coefficients between the two domains of interest, language and encoding, and some of the background measures are presented in Table 19.2. Absolute encoding is positively linked to geocentric language on Animals and Chips (but not Steve’s Maze), and negatively with egocentric language. However, it is also related
A Cross-Cultural Comparison of Spatial Language and Encoding
271
to background measures such as age and schooling (but not gender), as well as location (more absolute in rural samples) and preferred language during testing (more absolute with Balinese rather than Indonesian). Partial correlations controlling for age, schooling and location are 0.35 and 0.23, and remain statistically significant (p < 0.001 and 0.01 respectively). These relationships between language and encoding are not very strong, but nevertheless present at the structural level (that is cannot be explained by common background variables such as increase with age). This is indeed an interesting finding, since it contradicts the absence of relationships we found in previous research in India and Nepal, and will comfort adherents to the linguistic relativity theory. TABLE 19.2
Animals
Chips
Steve’s Maze
G language
E language
Pearson correlation coefficients between language, encoding and background variables
R p< N r p< N r p< N r p< N r p< N
Gender
Age
School
Location
Balinese vs Indonesian
G language
−0.053 0.228 200 −0.062 0.216 160 −0.035 0.348 130 −0.028 0.365 149 0.038 0.320 150
0.143 0.022 200 0.236 0.001 160 −0.069 0.216 130 0.424 0.000 149 0.047 0.284 150
0.119 0.046 200 0.206 0.005 160 −0.107 0.114 130 0.386 0.000 149 0.101 0.109 150
0.284 0.000 200 0.406 0.000 160 0.272 0.001 130 0.484 0.000 149 −0.551 0.000 150
0.186 0.004 200 0.350 0.000 160 0.113 0.100 130 0.677 0.000 149 −0.581 0.000 150
0.403 0.000 149 0.436 0.000 144 0.105 0.129 119 1.000 149 −0.594 0.000 142
E language −0.177 0.015 150 −0.412 0.000 145 −0.159 0.042 119 −0.594 0.000 142 1.000 150
Family Background and Other Social Indicators When language and encoding are correlated with the various family background variables, a coherent picture emerges: The children who use more geocentric (and less egocentric) language have parents who are less educated and have more modest occupations, less contact with the media (in particular newspapers), speak Balinese in the home and have less contact with the city. The Pearson correlations are to the order of 0.4 to 0.5 with geocentric language and –0.3 with egocentric language, all significant beyond the 0.01 level. Absolute encoding is similarly related to these variables, in particular for Chips. Using principal component factor analysis as a data reduction technique, the family
272 Pierre R. Dasen and Jürg Wassmann background variables were combined into a single factor score, indicating what we might call social change or ‘acculturation’. The principal component explained 66 per cent of the variance. Similarly, the three encoding tasks were combined into a factor score explaining 54 per cent of the variance. The correlations between these summary variables are shown in Table 19.3. TABLE 19.3
Pearson correlations between acculturation, language and encoding
r Acculturation
p< N
G language
E language
−0.486
0.438
0.000 125
r G language
p< N r
E language
p< N
0.000 125
Encoding −0.331 0.000 100
−0.649
0.474
0.000
0.000
156
124 −0.388 0.000 123
The correlations remain the same when age and schooling are controlled. These results point to a strong link between the choice of FoR in both language and encoding and social family background variables that combine to show that in Bali, the geocentric system is linked to a more rural, traditional lifestyle, and that contact with the city, media and Indonesian brings about the use of more egocentric language and encoding.
BALI 2002: DISCUSSION AND CONCLUSION Already in the first fieldwork, Wassmann and Dasen (1998) had noted that there was somewhat more use of egocentric language in the (more urban) south of Bali than in our remote location on the North coast. We also considered the possibility that the developmental trend from more absolute to more relative encoding could have been produced by acculturative factors such as schooling, and the Indonesian language that is linked to it. The present results strongly point to the importance of social change or acculturation. The Balinese orientation system, and the Balinese language (in particular in its complex form with several levels of address depending on social position), together with their religious and symbolic connotations, are all part of the traditional Balinese culture, and so is the geocentric FoR and the process of absolute spatial encoding that go with it. We
A Cross-Cultural Comparison of Spatial Language and Encoding
273
therefore find more of it (geocentric language, systematically absolute encoding) in rural locations than in the city (and more in a more remote rural location such as Bunutan), more in Balinese speakers than in those who prefer to use Indonesian, more in traditional, farmer families with little contact with the city and limited access to the media than in families where parents have a higher level of education, paid employment and access to media, in particular to newspapers. In all three samples, the age trends on the language elicitation tasks indicate an increase of geocentric language with age, and an increase of absolute encoding (except for Steve’s Maze). It seems that the influence of acculturation does not increase with age, as one might have expected (with the number of years of schooling, greater proficiency in Indonesian, possibly more wider geographical mobility with age), but the children seem, on the contrary, to be socialized progressively into the Balinese cultural norm irrespective of their various social and linguistic backgrounds. In other words, Balinese traditional culture shows its strength!
GENEVA, SWITZERLAND We will report only briefly on the study carried out in Geneva, which was done mainly to ascertain that children speaking European languages do indeed use exclusively egocentric language and relative encoding. In Geneva, Switzerland, we tested 70 children aged four to 12, with an equal number of boys and girls; 27 of these were monolingual French speakers, the others were bilingual (mainly with Portugese and Spanish, Albanian, Arabic, English and a few other languages). Of the 70 children, only two (aged nine and 10) knew about cardinal directions, both being boy scouts. This is despite of the fact that a huge compass was drawn in the middle of the play ground. On the language elicitation tasks, no geocentric language was used whatsoever. Until age six, deictic (‘this way’) and intrinsic words are used, almost entirely replaced from age seven onwards by egocentric language. The encoding on the Animals and Chips tasks is presented in Figure 19.2, comparing Geneva and Bali (2002) results, the latter being calculated on the combined city and village samples. It is not really surprising that there is a very low rate of absolute encoding in Geneva (the 10 to 20 per cent rate is no doubt spurious, due to guessing or distraction mainly for the younger children). The results on Steve’s Maze, on the other hand, show no difference between Bali and Geneva, which again confirms task specificity. No differences in encoding were found between monolingual and bilingual children.
274 Pierre R. Dasen and Jürg Wassmann FIGURE 19.2
Absolute encoding on Animals and Chips, in Bali (2002 study) and in Geneva
CONCLUSION In cross-cultural developmental psychology (see Berry et al., 1997; Segall et al., 1999), it is not often that we find attempts at a replication of field studies. In this case, pursuing research first in India and Nepal was, in a way, a first attempt at replication, since in the locations used, the people also use geocentric orientation systems. Similarly, they also have at their disposal, as do the Balinese, two systems of reference to describe and encode space, the geocentric and the egocentric one (discounting the intrinsic that is universal), so that we can study when and how they give preference to one or to the other. We are therefore in a case that is somewhat different from the two populations on which Levinson (2003) concentrates most of his discussion, the Tzeltal speakers of Tenejapa in Mexico, and the Guugu Yimithirr Australian Aborigines, who use only the geocentric system. That some relative encoding occurs at all in these two populations is therefore more surprising than if it occurs in all of the samples we have been studying, including Bali. The research in India and Nepal allowed us to show that the proportion of absolute and relative encoding, and the development of geocentric and egocentric language, are to a large part linked to the eco-cultural context, combining the appropriateness of the orientation system to the local topography and symbolic and religious aspects, as well as language, the latter being, as we see it, a part of the overall culture and not a single
A Cross-Cultural Comparison of Spatial Language and Encoding
275
determinant of cognition in itself. The fact that the choice of a FoR for encoding is also task specific had been borne out by our first research in Bali, and was confirmed throughout further research. On the other hand, the developmental trend of an early absolute encoding on the Animals task, which then changes to some extent to relative with age, was not confirmed in any of the locations in India and Nepal, and it therefore seemed imperious to replicate the study also in Bali itself. While this particular age trend was not found again in this renewed study, not even in the same area of Bali (although Bunutan itself, as we have mentioned, has also changed), we now have what appears to be a very coherent body of results, all of which make good sense. Working in several locations, and with relatively large samples, gives us a chance to study several variables beyond age, such as urban and rural contexts, the use of Balinese or Indonesian, and other social indicators. Levinson (2003) reviews carefully the evidence collected by the Cognitive Anthropology Research Group (CARG) at the Max Planck Institute of Psycholinguistics (MPI) on various socio-cultural factors, such as urban/rural location, literacy and age, schooling and ‘conservatism’ as indices of social change. He ends up discarding all these variables, except for language, and he concludes that ‘these variables show no substantial effect’ (p. 196) and ‘all further tests provide no evidence for the effect of cultural change’ (p. 197). However, Levinson also realizes that some of the measures of social variables (and even age) in these studies were rather crude and the sample sizes quite small, so that ‘there remains a good chance that a positive effect might be found with better data’ (Levinson, 2003: 196). We believe that we now have better data, and contrary to Levinson’s affirmation, we think that acculturation is one of the main variables to explain our results in Bali, of both 1994 and 2002.
ACKNOWLEDGEMENTS The research in Bali was carried out under the authority of Professor I. Gde Pitana, Udayana University, Denpasar, Bali, and Professor Wayan Nurkancana and Dr I. Nyoman Adil, IKIP, Singaraja, Bali. The data were collected by Mr Made (Kadek) Aryawan Adijaya and Mr I. Nyoman Pasek Hadisaputra, whom we thank sincerely. We also thank Marie Anne Broyon, Anahy Gajardo and Yvan Leanza who helped collect the data in Geneva. The funding was provided by grant 113-67178.01 of the SNF to the first author.
REFERENCES Berry, J.W., P.R. Dasen and T.S. Saraswathi. 1997. ‘Basic Processes and Human Development’. In J.W. Berry, P.R. Dasen and T.S. Saraswathi (eds), Handbook of Cross-Cultural Psychology, Vol. 2 (2nd edition). Boston: Allyn & Bacon.
276 Pierre R. Dasen and Jürg Wassmann Dasen, P.R., R.C. Mishra and S. Niraula. 2004. ‘The Influence of Schooling on Cognitive Development: Spatial Language, Encoding and Concept Development in India and Nepal’. In B.N. Setiadi, A. Supratiknya, W.J. Lonner and Y.H. Poortinga (eds), Ongoing Themes in Psychology and Culture (pp. 223–37). Yogjakarta: Kanisius. Levinson, S. 2003. Space in Language and Cognition: Explorations in Cognitive Diversity. Cambridge: Cambridge University Press. Levinson, S., S. Kita, D. Haun and B. Rasch. 2002. ‘Returning the Tables: Language Affects Spatial Reasoning’. Cognition, 84: 155–88. Li, P. and L. Gleitman. 2002. ‘Turning the Tables: Language and Spatial Reasoning’. Cognition, 83: 265–94. Mishra, R.C. and P.R. Dasen. 2005. ‘Spatial Language and Cognitive Development in India: An Urban/Rural Comparison’. In W. Friedlmeier, P. Chakkarath, and B. Schwarz (eds), Culture and Human Development: The Importance of Cross-Cultural Research to the Social Sciences (in Honor of Gisela Trommsdorff’s 60th Birthday) (pp. 153–79). Hove, UK: Psychology Press. Mishra, R.C., P.R. Dasen and S. Niraula. 2003. ‘Ecology, Language, and Performance on Spatial Cognitive Tasks’. International Journal of Psychology, 38: 366–83. Segall, M.H., P.R. Dasen, J.W. Berry and Y.H. Poortinga. 1999. Human Behavior in Global Perspective: An Introduction to Cross-Cultural Psychology (Revised 2nd edition). Boston: Allyn & Bacon. Wassmann, J. and P.R. Dasen. 1998. ‘Balinese Spatial Orientation: Some Empirical Evidence for Moderate Linguistic Relativity’. The Journal of the Royal Anthropological Institute, Incorporating Man (N.S.) 4: 689–711.
Chapter 20 Culture, Language, Spatial Frames of Reference and Hemispheric Dominance R.C. Mishra and Pierre R. Dasen
INTRODUCTION
T
his chapter reports a part of a larger research study carried out in Varanasi city. The purpose is to examine the linkage between language, spatial frames of reference (FoR) and hemispheric dominance. It also explores the relationship of these measures with psychological differentiation (field-independence/fielddependence). The theoretical background and the general intent of the project have been described in Mishra and Dasen (this volume). The focus of this chapter is on structural dimensions of these measures and their linkages with hemispheric lateralization. Literature on language socialization indicates that language is largely a product of parent–child interaction particularly in early years (Bates et al., 1995; Harris, 1992; Snow, 1995). In later years, other influences (for example, peers, teachers) may also shape language development. This indicates that there may be different routes into language (and cognition, if it is linked to language) for different individuals, with different patterns of causation along the way (Bates et al., 1995). Studies distinguish between ‘expressive’ and ‘referential’ uses of language. While almost all individuals seem to function adequately in both, the use of referential language seems to be highly correlated with cultural features on the one hand and cognitive functioning on the other (Gumperz and Levinson, 1996; Levinson, 2003; Werner and Hubel, 1999). A striking feature of the work on spatial cognition is the diversity of tasks and mental processes subsumed under this term (Linn and Peterson, 1985). Involvement of different brain structures makes it even more complex. As a result, many researchers define spatial cognition in terms of the tasks that are processed predominantly by the right hemisphere (RH). Neuropsychological studies of spatial cognition generally hold on to the hemispheric lateralization hypothesis. Witelson and Swallow (1988) have presented an overview of
278 R.C. Mishra and Pierre R. Dasen the tasks that clearly involve a strong spatial component and are processed by the RH of the brain. They have also made reference to some tasks that are characterized by a seemingly strong spatial component, but are found to be more dependent on the left than RH. This makes the prediction of spatial cognition difficult in terms of hemispheric lateralization theory. Recent advances in neuroimaging have made it clear that both hemispheres of the brain are active in almost all tasks (for example, Grimshaw, 1998; Sergent et al., 1992), but there is some sort of division of labour between these coordinated hemispheres. This is accomplished by dividing stimulus inputs: the RH deals with information in the left visual field, and the left hemisphere deals with information in the right visual field. It is also indicated that the same type of processing may occur in each hemisphere with distinctive qualitative differences, and that the corpus callosum aids parallel processing by shielding each hemisphere from the other until some late integration stage (Chiarello and Maxfield, 1996). Task characteristics may play an important role in determining the process of integration and the efficiency of one hemisphere over another in dealing with information. Thus, tasks that require processing in terms of global features (for example, recognition of male or female faces) may not reveal a distinctive superiority of one hemisphere over another. On the other hand, tasks that require processing in terms of analytic features (for example, words related to different semantic categories) may reveal a distinctive superiority of the right over the left hemisphere in terms of efficiency (quicker processing) in dealing with stimuli. Research on spatial cognition in Nepal (Niraula, 1998; Niraula and Mishra, 2001) suggests that children who are psychologically more differentiated use predominantly an absolute frame North-South-East-West (NSEW) to describe spatial information, whereas those who are less differentiated describe spatial information predominantly by using a relative frame Left-Right-Front-Back (LRFB). Witkin and Goodenough (1981) cite studies that tend to support the hemispheric lateralization hypothesis of psychological differentiation. Children who perform cognitive tasks in a more differentiated manner demonstrate a right hemispheric dominance, whereas those who perform cognitive tasks in a less differentiated (global) manner demonstrate a left hemispheric dominance. Taylor and Tversky (1996) suggest a probable linkage between the predominant spatial reference system (encoding) and hemispheric lateralization, but convincing data in this respect are lacking. In the present chapter, we have attempted to examine the relationship between language, spatial FoR, psychological differentiation (field-independence–field-dependence) and hemispheric lateralization. It was hypothesized that there will be a structural link between geocentric language, absolute spatial FoR, psychological differentiation and hemispheric lateralization. It was also expected that this relationship would stand even when controlling for age, gender and schooling. It was further posited that superiority of the right over the left hemisphere would be evident particularly in processing categorical information on a verbal task more efficiently than in processing distinctive features of male or female faces.
279
Culture, Language, Spatial Frames of Reference and Hemispheric Dominance
SAMPLE The study was carried out with 376 boys and girls, aged 10–15 years, attending Hindimedium and Sanskrit schools. Sample characteristics are given in Table 20.1 along with the total number of children tested in each type of school. The schools are described in Vajpayee et al. (this volume). TABLE 20.1
Sample characteristics Gender Age
Boys
Girls
Total
11
9
30
39
12
26
59
85
13 14
9 5
46 31
55 36
15 Total
0 49
6 172
6 221
10
2
0
2
11
6
3
9
12
35
5
40
13
49
6
55
14
30
7
37
15
3
9
12
Total
125
30
155
Grand total
221
Hindi-medium schools
Sanskrit schools
+
155
=
376
TASKS AND TESTS In addition to language elicitation and spatial encoding tasks described in Mishra and Dasen (this volume), the following tests were also used.
Psychological Differentiation Psychological differentiation (field-dependence/independence) was measured with the help of Block Designs Test and Story–Pictorial Embedded Figures Test (SPEFT). Block Designs Test (Koh’s Blocks) involved construction of pictorially presented designs of increasing difficulty with the help of four, nine and 16 blocks within specified periods of time. Both time and accuracy of performance were recorded. A short (10 designs) version of the test (Mishra et al., 1996) was used.
280 R.C. Mishra and Pierre R. Dasen The SPEFT (Sinha, 1984) comprised seven sets of pictures. Each set consisted of a simple and a complex card. In the simple card some objects and animals were depicted, which were embedded in a larger situation depicted in the complex card (for example, squirrels on a tree). The child had to locate within a maximum of 90 seconds the objects and animals of the simple card in the complex card in the background of a story, which was narrated with each card to encourage the child to locate the embedded items. Time taken and the number of objects correctly located by the child were recorded.
Hemispheric Lateralization This included a peripheral and a central measure. The peripheral measure, called handedness, was adapted from Mandal et al. (1992). It consisted of a number of tasks that the child could do with the hand, foot, eye or ear. Children were first asked about their preference to do the task with the left or right limb/organ, and whether they would choose right or left ‘always’ or ‘sometimes’. Then they were asked to perform the tasks (there were separate tasks for hand, foot, eye and ear). The use of the right or left limb/organ was recorded. The central measure consisted of a brain lateralization task administered with the help of a laptop using a programme developed by Mandal (personal communication). It used the ‘split-field’ technique in which the child was asked to concentrate on a black spot that appeared in the centre of the computer screen. Then appeared an arrow at the fixation point that pointed either to the left or right in a random order. The child was asked to look at the stimuli that appeared in the direction the arrow pointed to. The arrow and the stimuli appeared simultaneously. The distance between the fixation point and the presented stimuli on the computer screen was 12 cm. The child’s had to respond by pressing a key (of the keyboard) as soon as the stimulus pointed by the arrow was correctly recognized. Three practice trials were given. In the first trial, human faces and words were presented for 800 ms to either side in a random order. The child responded to the items, but the key was pressed by the experimenter to demonstrate to the child what was to be done in terms of response. In the second trial, the same stimuli were presented for 800 ms, but this time, the child had to respond to them by pressing the key. In the third trial, the same stimuli were presented for 180 ms to which the child had to respond by pressing the key. The child was allowed to proceed with the main experiment only when the presented stimuli were correctly responded to on eight out of 12 presentations. If the child did not meet this criterion, further practice trials were given. The main experiment was carried out with a stimulus presentation time of 180 ms. The child’s reaction time (RT) to stimuli presented to left and right was assessed with the help of the computer programme. Accuracy of response was recorded manually on a sheet developed for that purpose. In one sequence of trials, the child responded to words for
Culture, Language, Spatial Frames of Reference and Hemispheric Dominance
281
objects or animals. In another sequence, responses for male or female faces were obtained. In each sequence 12 trials were given.
Procedure The study was carried out in two phases. In the first phase, all tests, except the brain lateralization measure, were administered to children, generally in two or more sessions. Based on their performance on language elicitation and language encoding tasks, we attempted to classify them as using either a ‘geocentric’ (G) or an ‘egocentric’ (E) frame of reference. In this classification, consistency in correct language use and encoding were considered separately. Of the 376 children tested in the first phase of the study, only 86 were found to fulfil these criteria. For the G language group, there had to be at least five out of six G items on Encoding tasks, at least six out of seven on Road, and six out of nine on Perspectives. For E language group, there had to be more than 2/6 E items on Encoding, 3/7 on Road and 3/9 on Perspectives. There were 28 subjects who demonstrated consistent use of G language and 16 who used E language consistently. In terms of encoding, 13 subjects showed completely G encoding on Animals (on the seven items), completely on Chips (also seven items), and three or more G on Steve’s Maze. There were 14, called ‘mainly G’, the criteria for whom were less stringent (five G items out of seven on Animals, six out of seven on Chips, and two or more on Steve’s Maze). Another 25, called ‘mainly E’, had more than three completely E encoding on Animals and on Chips, and two or more E on the Steve’s Maze. Using the above stated language and encoding criteria, 86 children were selected for administration of the brain lateralization measure. Unfortunately, six of them were no longer available for testing; this reduced the sample size to 80 in the second phase of the study. In addition to the brain lateralization measure, a child questionnaire and a home questionnaire were administered to children and parents to assess a number of socioenvironmental variables that might be linked to G or E FoR, but this part is not reported in this chapter.
RESULTS The language and encoding scores were derived from several measures that were reduced to factor scores by principal component factor analyses, and coefficients of correlation were computed across different factors (Table 20.2). The analysis revealed that not only were G encoding and G language positively and significantly related, but they were also correlated positively and significantly with field-dependence–independence (FDI) scores. These relationships were found to stand even after controlling for age, gender, preschooling, grade, years of schooling and school type (Table 20.3).
282 R.C. Mishra and Pierre R. Dasen TABLE 20.2
Pearson correlations between language, encoding and FDI Regression factor scores
G encoding
G+ language
E+ language
FDI
0.455
−0.334
0.324
R Sig. (2-tailed)
0.000
N G+ language
0.000
375
R Sig. (2-tailed) N
E+ language
0.000
375
375
−0.625
0.191
0.000
0.000
376
376
R
−0.034
Sig. (2-tailed)
0.510
N
TABLE 20.3 school type
376
Partial correlations controlling for age, gender, preschooling, grade, years of schooling and
G+ language
E+ language
0.27**
−0.21**
0.23**
−0.55**
0.05
G encoding G+ language E+ language
FDI
0.07
Note: ** = p < 0.01.
Hemispheric Lateralization Separate analysis of variance (ANOVA) were performed on preference and performance measures of peripheral lateralization (hand, foot, eye, ear); they revealed no significant differences either between encoding or language groups. Analysis of brain laterality measures revealed no significant differences on accuracy scores indicating the pattern of performance for words and faces presented either to the left or right visual field to be almost similar across groups. ANOVA on Total Reaction Time computed for words and faces presented to right and left also revealed no significant difference for the language groups. For encoding, ANOVA (Table 20.4) revealed one significant difference for Word left (F2, 49 = 3.86, p < 0.05) in favour of G encoders. Other differences were not significant.
Culture, Language, Spatial Frames of Reference and Hemispheric Dominance TABLE 20.4
283
ANOVA outcomes on brain lateralization measures, G and E encoding groups
Words right RT
Word left RT
Face right RT
Face left RT
df
Mean square
F
Between Groups
2
85,919,772.315
2.131
0.130
Within Groups
49
40,319,844.657
Total
51 3.855
0.028
0.051
0.951
0.917
0.407
Between Groups
2
159,207,329.359
Within Groups
49
41,301,112.122
Total
51
Between Groups
2
Within Groups
49
Total
51
397,177.324
Sig.
7,817,764.327
Between Groups
2
3,950,106.181
Within Groups
49
4,309,852.387
Total
51
DISCUSSION The findings of the study suggest that although the relationship between language and encoding is not perfect, G and E languages often go with G and E encodings of spatial arrays respectively. The relationship between G and E language is negative, and so is the relationship between G and E encoding. This finding may be taken to offer support to a moderate form of the linguistic relativity hypothesis (Niraula et al., 2004). Analyses further suggest a structural link between G language, G encoding and FDI. This means that those who use G language and encoding also tend to be psychologically more differentiated. This supports the previous findings of Niraula and Mishra (2001), which suggested a significantly positive correlation of G language and encoding with SPEFT scores. While Niraula and Mishra (2001) also reported a significantly negative correlation of E language with SPEFT scores, the present findings suggested a very marginal negative correlation. It indicates that field-independence (higher level of differentiation) is seemingly governed by the RH with which G encoding and G language use are also supposedly linked. With respect to hemispheric dominance, the findings consistently show that in Varanasi, peripheral measures of lateralization are related neither to G language nor to G encoding. The pattern of results is the same for preference and performance measures. This suggests
284 R.C. Mishra and Pierre R. Dasen that users of G language and encoding do not differ significantly in terms of peripheral aspects of brain functioning. On the other hand, the central measure of brain lateralization does provide us with some evidence of difference between G and E in terms of the functioning of their RH. The difference in RT of G and E groups for words presented to the left was significant, indicating a more efficient (quicker) processing of words by the G than the E group in the RH. This finding suggests that the RH of G encoders is slightly more specialized for word processing than that of E encoders. With regard to face processing, although studies indicate that it requires both componential (a left hemispheric function) and configurational (a right hemispheric function) processes (Moscovitch, 1979) to operate in parallel (Sergent, 1984), no substantial support is found for right hemispheric bias in processing of facial information (Mandal and Asthana, 1999; Sergent et al., 1992). The findings of the present study reveal no significant difference between G and E groups in terms of processing of facial information by either left or RH of the brain. With regard to the processing of words, on the other hand, they indicate that G encoders perform significantly more efficiently than E encoders when the words are presented to the left than to the right visual field. It may be noted that processing of words in terms of conceptual categories like ‘objects’ or ‘animals’ requires a high level of abstraction, and the RH of G encoders, compared to that of E encoders, appears to be quicker or more efficient in performing this function than their left hemisphere. Accuracy remains high in the RH possibly because it does not engage in interpretive or elaborative processes, which often take place in the left hemisphere with verbal materials and interfere with its processing efficiency (Gazzaniga et al., 2002; Wolford et al., 2000). It may be mentioned here that the present study has attempted the analysis of brain lateralization only at the functional level. Hence, we cannot claim that the G and E encoders really differ with respect to structural or neural organization of their brain. As Witelson and Swallow (1988) indicate, brain lateralization studies with normal subjects demand that before attempting to determine brain lateralization in a new group with a new task, certain conditions must be fulfilled. For example, it is necessary to ensure that the new task has been validated with individuals who can be claimed to have structural brain lateralization. If this condition is not fulfilled, it would not be possible to specify the mechanisms underlying differences in performance of subjects. Efficiency of performance on a cognitive task can be explained in terms of physiological mechanisms (for example, differential neural organizations), as well as psychological mechanisms (for example, use of differential cognitive strategies). The laterality task used in this study still requires validation with individuals who may be claimed to have well-established lateralization of the brain at the structural level. We expect to discover some of these linkages in a study in which brain damaged patients have been examined not only for the size, site and severity of the damage with the help of computational topography (CT) and magnetic resonance imaging (MRI) scans, but also for the use of language and encoding on the same tasks that
Culture, Language, Spatial Frames of Reference and Hemispheric Dominance
285
were used in the present study. Until these results are worked out, our claim regarding right brain dominance in the G group is confined to the functional aspects of the brain on the particular tasks we have worked with in this study. In fact, there is also need to focus more on other behavioural characteristics of right brain dominant people in order to cross-validate the findings with respect to differences in the functional organization of the brain in G and E encoders.
REFERENCES Bates, E., P.S. Dale and D. Thal. 1995. ‘Individual Differences and Their Implications for Theories of Language Development’. In P. Fletcher, and Mac Whinny (eds), The Handbook of Child Language (pp. 96–151). Oxford: Blackwell. Chiarello, C. and L. Maxfield. 1996. ‘Varieties of Interhemispheric Inhibition, or How to Keep a Good Hemisphere Down’. Brain and Cognition, 30: 81–108. Gazzaniga, M.S., R.B. Ivry and G.R. Mangum. 2002. Cognitive Neuroscience: The Biology of the Mind. New York: W.W. Norton. Grimshaw, G.M. 1998. ‘Integration and Interference in the Cerebral Hemispheres: Relations with Hemispheric Specialization’. Brain and Cognition, 38: 108–27. Gumperz, J.J. and S.C. Levinson (eds). 1996. Rethinking Linguistic Relativity. Cambridge: Cambridge University Press. Harris, M. 1992. Language Experience and Early Language Development: From Input to Uptake. Hove: Lawrence Erlbaum Associates. Levinson, S. 2003. Space in Language and Cognition, Cambridge: Cambridge University Press. Linn, M.C. and A.C. Peterson. 1985. ‘Emergence and Characterization of Sex Differences in Spatial Ability. A Meta-Analysis’. Child Development, 56: 1479–98. Mandal, M.K. and H.S. Asthana. 1999. ‘Hemispheric Regulation of Emotion’. Indian Psychological Abstracts and Reviews, 6: 3–28. Mandal, M.K., G. Pandey, S.K. Singh and H.S. Asthana. 1992. ‘Hand Preference in India’. International Journal of Psychology, 27: 433–42. Mishra, R.C., D. Sinha and J.W. Berry. 1996. Ecology, Acculturation and Psychological Adaptation: A Study of Adivasis in Bihar. New Delhi: Sage Publications. Moscovitch, M. 1979. ‘Information Processing and the Cerebral Hemispheres’. In Gazzaniga, M.S. (ed.), Handbook of Behavioral Neurology, Vol. 2 (pp. 379–446). New York: Plenum,. Niraula, S. 1998. Development of Spatial Cognition in Rural and Urban Nepalese Children. Unpublished Doctoral Thesis. India: Banaras Hindu University. Niraula, S. and R.C. Mishra. 2001. ‘Spatial Orientation of the Newar Children in Nepal’. Social Science International, 17: 36–48. Niraula, S., R.C. Mishra and P. Dasen. 2004. ‘Linguistic Relativity and Spatial Concept Development In Nepal’. Psychology and Developing Societies, 17: 99–124. Sergent, J. 1984. ‘An Investigation into Component and Configurational Processes Underlying Face Perception’. British Journal of Psychology, 75: 221–42. Sergent, J., S. Ohta and B. MacDonald. 1992. ‘Functional Neuroanatomy of Face and Object Processing: A PET Study’. Brain, 115: 15–29.
286 R.C. Mishra and Pierre R. Dasen Sinha, D. 1984. Manual for Story–Pictorial EFT and Indo-African EFT. Varanasi: Rupa. Snow, C.E. 1995. ‘Issues in the Study of Input: Finetuning, Universality, Individual and Developmental Differences, and Necessary Causes’. In P. Fletcher and B. MacWhinney (eds), The Handbook of Child Language (pp. 180–93). Oxford: Blackwell. Taylor, H.A. and B. Tversky. 1996. ‘Perspective in Spatial Descriptions’. Journal of Memory and Language, 35: 371–91. Werner, S. and C. Hubel. 1999. ‘Spatial Reference Systems (Special Issue)’. Spatial Cognition and Computation, 1(4): iii–vii. Witelson, S.F. and J.A. Swallow. 1988. ‘Neuropsychological Study of the Development of Spatial Cognition’. In J. Stiles-Davis, M. Kritchevsky and U. Bellugi (eds), Spatial Cognition: Brain Bases and Development. (pp. 373–409). Hillsdale, NJ: Lawrence Erlbaum Associates. Witkin, H.A. and D.R. Goodenough. 1981. Cognitive Styles: Essence and Origin. New York: International University Press. Wolford, G., M.B. Miller and M.S. Gazzaniga. 2000. ‘The Left Hemisphere’s Role in Hypothesis Formation’. Journal of Neuroscience, (Online), 20: RC64.
Chapter 21 Cultural Adaptations and Cognitive Processes of Tribal Children in Chotanagpur R.C. Mishra and John W. Berry
INTRODUCTION
O
ne of the recent developments in psychology is the emergence of an ‘ecological perspective’ in which the culture of the group and behaviour of individuals are comprehended in terms of their adaptations to differing ecological settings. A major emphasis of the ecological school of thought is on resource utilization. This approach has been used in three different ways to understand human adaptations to habitat. First is the analysis of the relationship between the physical environment and the economic system; the second is the examination of shared behaviour patterns (customs) associated with a particular subsistence activity; and the third is the search for any broader effects of such customs on other aspects of the culture (Berry et al., 1995). In the earlier ecological approaches, cultural systems have been viewed as relatively stable or even permanent adaptations. A current view is that cultures evolve over time in response to changing ecological circumstances, or due to contact with other cultures. This viewpoint has led to a conception of ecological adaptation as a continuous interactive process between ecological, cultural and behavioural variables. The dynamic relationship among these processes is represented in an ‘eco-cultural’ model (Berry, 1976, 1987). The model examines similarities and differences in human psychological functioning (both at individual and group levels) by taking into account two fundamental sources of influence (ecological and socio-political), and a set of variables that link these influences to psychological characteristics. These include cultural and biological adaptations at the population level, and certain ‘transmission variables’ (enculturation, socialization, genetic and acculturation) between the population and the individual levels. Thus, the model
288 R.C. Mishra and John W. Berry considers human diversity (both cultural and psychological) to be a set of collective and individual adaptations to the context. Within this general perspective, it views cultures as evolving adaptations to ecological and socio-political influences and psychological characteristics in a population as adaptive to their cultural context as well as to the broader ecological and socio-political influences (see Berry, this volume). Support for the predictions of the eco-cultural model has been obtained in a number of research studies (Berry, 1976; Berry et al., 1986; Mishra et al., 1996). The basic ecological component of the model is rooted in the interaction patterns of human organisms with their physical environment (habitat) for the satisfaction of their primary needs. The interaction patterns result in certain economic and demographic characteristics of the groups. Based on subsistence activities, societies have been classified as gathering, hunting, agricultural, irrigation and industrial (Murdock, 1969). This classification has been used to sample societal variations for cognitive research (Berry, 1976; Berry et al., 1986; Mishra et al., 1996). The ecological dimension in Berry’s framework places hunting–gathering (HG), nomadic and small-scale groups at one end of the dimension, whereas agricultural, sedentary and large-scale groups are placed at the other end. The cultural dimension involves four variables: degree of political stratification, degree of social stratification, type of family, and socialization emphases on assertion or compliance. The cultural index, thus obtained, has been combined with an ecological index to produce an eco-cultural index (Berry, 1976), which is a unidimensional and bipolar index of ecological and cultural adaptation. Later research (Boldt, 1976; Boldt and Roberts, 1979; Gamble and Ginsberg, 1981) has suggested the possibility of the existence of two independent dimensions similar to the ‘differentiation–integration’ distinction proposed by Lomax and Berkowitz (1972). In view of these evidences, Berry et al. (1995) have proposed and operationalized ‘societal size’ and ‘social conformity’ as two cultural dimensions, which tend to vary considerably as a function of subsistence strategies of the groups. While societal size seems to be a linear function of subsistence strategy, social conformity seems to present a curvilinear reationship (relatively low in gathering, hunting and industrial societies, but higher in rudimentary and irrigation agricultural societies). A similar distinction has been proposed with respect to cognition. A cognition consists of certain units and parts. For example, a block design (unit) consists of several blocks (parts). The units and parts may have two basic relations. One, called ‘distinctiveness’, refers to the recognition of parts and units as distinct from one another. The second, called ‘connectedness’, refers to the recognition of relationship among parts and units. These relationships may be either intra-unit (among the parts within an unit) or extra-unit (among different units). If these notions are combined, we get four cognitive functions. Two of these, intra-unit distinctiveness (ID) and extra-unit connectedness (EC), have been of main interest in research on cognition in a cross-cultural perspective. In some previous research, these dimensions have been referred to as ‘differentiation’ and ‘contextualization’ respectively.
Cultural Adaptations and Cognitive Processes of Tribal Children in Chotanagpur
289
The findings of various studies provide evidence for a curvilinear relationship of ID with subsistence economy. The tendency to emphasize distinction among parts of a cognitive unit appears to be low among gatherers, high among hunters, low among rudimentary agriculturists, medium among irrigation agriculturists and high among urban industrial societies (Berry, 1976; Iwawaki, 1986; Iwawaki and Vernon, 1988; McIntyre, 1976; Mishra et al., 1996). On the other hand, the relationship of EC with subsistence economy appears to be linear, showing a systematic decrease from gatherers and hunters to industrial society. This conclusion is largely based on studies of syllogistic reasoning (Denny and Davis, 1989; Luria, 1976; Scribner, 1977) and mathematical thinking (Denny, 1986; Goody, 1977). Anthropological research suggests the existence of ‘societal size’ and ‘social conformity’ dimensions. Cognitive research provides evidence for the existence of ID and EC dimensions. Some cross-cultural research on cognition does suggest the possibility of a predictable realtionship between the two cultural and cognitive dimensions. However, these relationships have not been systematically examined by employing measures of differentiation and contextualization. This chapter reports a study that examines (a) the distribution of cultural dimensions of ‘societal size’ and ‘social conformity’ in different subsistence level groups; (b) the development of cognitive differentiation and contextualization in relation to subsistence strategies of groups; and (c) the relationship between the two cultural and two cognitive dimensions. In the light of the findings of studies, it was hypothesized that: 1. Social conformity would be low in HG and industrial samples and higher in rudimentary and irrigation agricultural samples. 2. Societal size would be low in HG societies and increase over agricultural societies to a high in industrial societies. 3. ID would be relatively higher in HG and industrial samples than in the agricultural samples. 4. There would be a decrease in the level of EC from HG to agricultural to industrial samples.
DESIGN AND SAMPLE The study was carried out with children belonging to four groups that varied in subsistence activities. These included hunting–gathering (HG), dry agriculture (DA), irrigation agriculture (IA) and industrial wage earning groups. At each level of subsistence strategy, boys and girls of 9–12 years of age were studied. This led to a 4 (subsistence levels) × 2 (gender) distribution of the sample, with equal number of children represented in each cell.
290 R.C. Mishra and John W. Berry Overall 400 children were studied. They were drawn from the Birhor and Oraon tribal cultural groups living in Ranchi, Gumla and Hazaribagh districts of the Chotanagpur region of Bihar (now Jharkhand), India. These districts are characterized by heavy concentration of tribal populations. The sampling was carried out randomly at the family level. The HG sample was drawn from the Birhor cultural group. The other three samples were selected from the Oraon cultural group. The mean age of children ranged from 10.14 to 10.88, and the mean years of schooling from 1.97 to 5.51. The Birhor largely represent a nomadic tribal population. They move about from one forest to another in small bands in search of game, fruit and roots, fibres, honey and bees wax. Women and children also join them in this activity. Children often engage in fishing and in trapping wild rats. They also gather fruits, roots, yams, tubers, caterpillars and firewood. Certain changes have taken place in the life of Birhor during the last few decades. Consequently, some of them have gradually taken to a fixed habitation and learnt a rudimentary form of agriculture. However, their dependence on the forest continues. The Oraon represent a tribal cultural group that settled down some 1,800 years ago. They have acquired highly a developed knowledge of agriculture. Depending on the terrain they occupy and the facilities available there, they engage in either dry or irrigation agriculture, which constitutes the main source of their livelihood. Cattle-rearing is another major source of their economy. The Oraon have highly developed rules regarding authority structure, land rights, marriage and transfer of property. School education has been quite popular among the Oraon. As a result, many of them have taken wage employment in schools, offices, hospitals as well as government and private industries.
Measures A number of measures were employed for the assessment of cultural and cognitive dimensions.
Societal Size This was determined on the basis of an index derived from the ratings of the following: 1. Population size: It referred to the number of persons living in a settlement (less than 100 to more than 500 persons, rated on a 5-point scale). 2. Habitation: It referred to the lifestyle of people (nomadic to sedentary, rated on a 4-point scale). 3. Political stratification: It referred to the political hierarchy prevailing in the group (local unit to three or more units, rated on a 4-point scale). 4. Religion: It was conceived in terms of the belief and faith of the members of a group (animism to Hinduism, Islam or Christianity, rated on a 3-point scale).
Cultural Adaptations and Cognitive Processes of Tribal Children in Chotanagpur
291
The total score range on this cultural dimension was 4–16, with a high score indicating high societal size.
Social Conformity This consisted of five elements: 1. Hereditary distinctions: It referred to permanent distinctions based on heredity, wealth, caste or class that prevailed in the society (rated as low and high on a 2-point scale). 2. Socialization for compliance: It referred to the degree to which children were expected to be obedient, nurturant and responsible by the members of a community (very little to very much, rated on a 5-point scale). 3. Role obligation: It referred to the degree to which age or gender specific norms for various roles were well-defined and strongly emphasized in a community (little to high, rated on a 4-point scale). 4. Freedom from parents: It referred to the degree to which children were allowed freedom by their parents (very much to very little, rated on a 5-point scale). 5. The score: The score range on this dimension was 4–16, with a high score indicating high social conformity.
Social Connectedness This measure consisted of eight items that involved rating of the community on a 5-point scale on issues like the selection of spouse, decision for marriage, choice of child’s work, naming of a child, matters related to residence, raising of children, kinship orientation and social support. The score range was 8–40, with a high score indicating high social connectedness.
Individual Connectedness This was assessed with a six-item scale (5-point rating), which probed into the extent to which an individual participated in akhara (collective dance), birth, marriage and death ceremonies, and frequency of visit to relatives and ailing neighbours. The score range was 6–30, with a high score indicating high individual connectedness.
Measures of Cognitive Dimensions The cognitive tests were intended to assess the processes of differentiation, contextualization and integration.
292 R.C. Mishra and John W. Berry Differentiation This was assessed by using the Story–Pictorial EFT (SPEFT) and Hidden Words Test (HWT). The SPEFT (Sinha, 1984) requires a child to disembed and locate certain meaningful stimuli (for example, birds, snakes, butterflies) in a complex picture. A story is also told to the child to describe the circumstances that led to the hiding of objects in the complex picture. The number of stimuli correctly disembedded by the child are noted. A high score indicates high differentiation. The HWT follows closely the test developed by Berry et al. (1995). The subjects listen to a series of words and non-words, but are asked to repeat only the words. This is considered to be a verbal counterpart of the visual EFT. The number of words correctly repeated by the child is noted. A high score indicates high differentiation.
Contextualization This cognitive process was assessed by using Locating Objects Test (LOT), Syllogistic Reasoning Test (SRT), and Unfamiliar Words Test (UWT). The LOT was developed by Berry et al. (1995). It consists of 13 ‘snoopy’ pictures. The child is presented with the name of an object, and is required to point to that object in the picture as quickly as possible. The success/failure and time taken are recorded. In one set of pictures, the objects are placed in appropriate context; in the other, they are placed in an inappropriate context. Discrepancy in performance (time on inappropriate minus appropriate) indicates high contextualization. The SRT is modelled on Luria’s test. It assesses contextualization in verbal reasoning. The test consists of syllogisms that pose familiar and unfamiliar problems for reasoning. The answers along with the requests for repeating the syllogisms are recorded. Greater discrepancy in the reasoning of familiar and unfamiliar problems (familiar minus unfamiliar) and a greater number of requests indicate more contextualization. The UWT consists of stories developed around themes familiar to children. In each story, an unknown (meaningless) word is used to describe a person, object or activity. The child listens to the story, and is asked at the end of the story to tell the meaning of the unknown word. These meanings are evaluated for appropriateness in the context of stories. More correct meanings and more requests for repetition of stories indicate greater contextualization.
Integration This was assessed by using Visual Closure Test (VCT) and Object Enumeration Test (OET). The VCT consisted of two practice and 10 test pictures drawn from the Street Closure Test. Each picture represents an object, animal or a human profile through visually disintegrated lines or patches of shades. The child is asked to tell or guess the stimuli depicted in the picture. A child can make more than one response to the same picture,
Cultural Adaptations and Cognitive Processes of Tribal Children in Chotanagpur
293
which are evaluated as correct or incorrect. Higher score and lesser time on the test indicates greater integration. In the OET, children are asked to tell the name of various objects present in their environment. Greater naming of items on this test indicates greater integration.
ANALYSES AND RESULTS The findings of the study are summarized according to the cultural and cognitive dimensions.
Cultural Dimensions Table 21.1 presents the mean score of groups on measures of societal size and social conformity. There is a progressive increase from HG to WE through the two agricultural samples on the measure of societal size. The relationship of social conformity with subsistence strategies is curvilinear, that is, low in HG and WE, but high in DA and IA groups. TABLE 21.1 Mean score of groups on the measures of cultural dimensions (high scores indicate higher placement on the dimension) Groups
Societal size
Social conformity
Social connectedness
Individual connectedness
Hunting–gathering Mean
6.11
6.51
10.47
12.33
S.D.
0.74
0.98
1.49
2.16
Mean
8.82
13.10
30.12
22.85
S.D.
0.74
1.34
4.25
3.43
11.79
12.14
25.85
21.63
0.81
1.95
4.10
3.74
13.92
7.08
14.65
14.28
0.69
1.32
2.31
2.39
Dry agriculture
Irrigation agriculture Mean S.D. Wage earning Mean S.D.
Mean scores of social and individual connectedness measures (Table 21.1) reveal a greater connectedness in DA and IA than in HG and WE samples. However, a similar
294 R.C. Mishra and John W. Berry trend of increase and decrease in individual and social connectedness scores suggests their covariation in all the samples. TABLE 21.2
Mean score of groups on differentiation measures (high scores indicate greater differentiation) Groups
SPEFT disembedding
HWT words
HWT non-words
Hunting–gathering Mean S.D.
22.19
7.33
2.12
1.60
2.00
2.12
21.34
8.01
3.21
2.08
1.48
1.86
21.45
8.32
2.56
2.48
1.50
1.15
22.66
9.94
2.06
1.53
1.32
1.98
Dry agriculture Mean S.D. Irrigation agriculture Mean S.D. Wage earning Mean S.D.
Cognitive Dimensions Differentiation Mean scores obtained by different groups on the measures of differentiation are summarized in Table 21.2. On the SPEFT, ANOVA revealed significant main effects of economy (F3, 392 = 13.48, p < 0.01) and gender (F1, 392 = 4.12, p < 0.05). All mean comparisons were significant (Newman Keuls test), except one for HG and IA samples. On the Hidden Words Test, the findings revealed a lesser repetition of words by HG than other groups. A progressive increase in word repetition from HG to WE groups was in evidence, and the main effect of economy was significant (F3,392 = 48.46, p < 0.01). On the other hand, WE and HG groups made fewer non-word repetitions than IA and DA groups did. The main effect of economy was again significant (F3,392 = 9.92, p < 0.01). Girls made non-word repetitions more than boys did (F1,392 = 6.15, p < 0.01).
Contextualization The mean scores of groups on LOT, SRT and OET are summarized in Table 21.3.
Cultural Adaptations and Cognitive Processes of Tribal Children in Chotanagpur
295
TABLE 21.3 Mean score of groups on contextualization measures (high scores indicate greater contextualization) Groups
LOT discrepancy
SRT reasoning
SRT request
UWT meaning
UWT request
Hunting–gathering Mean
1.05
2.60
0.10
1.27
9.27
S.D.
0.65
0.57
0.33
1.04
3.67
Mean
3.02
2.71
0.06
1.63
8.43
S.D.
0.96
0.56
0.24
0.96
2.54
Dry agriculture
Irrigation agriculture Mean
3.93
2.76
0.12
2.02
7.01
S.D.
1.26
0.45
0.38
1.13
2.15
Mean
2.70
2.84
0.13
2.46
4.70
S.D.
0.51
0.37
0.36
1.03
2.75
Wage earning
Boys Mean
2.06
2.69
0.10
1.73
7.30
S.D.
0.68
0.54
0.32
1.12
3.03
Mean
3.30
2.77
0.10
1.96
7.41
S.D.
1.05
0.46
0.46
1.13
3.59
Girls
The discrepancy scores on the LOT revealed a main effect of economy (F3,392 = 6.05, p < 0.01), with evidence for greater contextualization in DA and IA than in WE and HG samples. None of the mean comparisons was significant on the measures of SRT. On the other hand, economy had significant effect on meaning (F3,392 = 24.22, p < 0.01) and request (F3,392 = 49.31, p < 0.05) measures of the UWT. Meaning scores showed an increase from HG to WE levels, whereas on the request measure, the WE sample scored lower than IA, HG and DA samples respectively.
Integration Mean scores of the groups on VCT and OET are given in Table 21.4. On VCT, the effect of economy was significant on naming (F3,392 = 23.74, p < 0.01) and time (F3,392 = 6.76, p < 0.01) measures, with a significant economy × gender effect on naming as well. The WE sample identified more pictures and made more requests than other
296 R.C. Mishra and John W. Berry groups did. Girls from DA and IA samples identified less number of objects than boys. The WE sample took lesser time than DA, HG and IA samples. TABLE 21.4 Mean score of groups on integration measure (higher scores on naming and lower on time indicate greater integration) Groups
VCT naming
VCT time
OET naming
Mean
1.59
105.35
10.41
S.D.
1.20
85.62
3.04
Mean
1.36
106.63
13.12
S.D.
1.15
60.56
4.04
Mean
1.82
115.67
14.49
S.D.
1.27
65.53
3.99
Mean
2.66
82.86
17.42
S.D.
1.06
67.01
3.96
Mean
1.93
102.99
13.64
S.D.
1.21
62.77
4.46
Mean
1.73
112.27
14.09
S.D.
1.33
79.74
4.62
Hunting–gathering
Dry agriculture
Irrigation agriculture
Wage earning
Boys
Girls
On the OET, the economy effect was significant on the measures of object naming (F3,392 = 59.43, p < 0.01). HG sample enumerated lesser number of objects than did DA, IA and WE samples respectively. Girls of the DA and IA samples used lesser categories than boys. Inter-correlations among the core measures of the cognitive tests were computed, and factor analysis (principal components with varimax rotation) was carried out to examine the structure of cognitive dimensions (Table 21.5). A three-factor interpretation appeared to be the best outcome. Factor I had high positive loadings on SPEFT, HWT, VCT and OET; it accounted for approximately 46 per cent of the variance. Factor II was characterized by high positive loading on the LOT (discrepancy score); it accounted for approximately 15 per cent of the variance. Factor III was characterized by high positive loadings on SRT and UWT; it accounted for approximately 10 per cent of the variance.
Cultural Adaptations and Cognitive Processes of Tribal Children in Chotanagpur TABLE 21.5
297
Factor analysis outcomes on core cognitive measures
Cognitive measures
Factor loadings I
SPEFT
0.74
HWT
0.58
LOT
Factor loadings II
Factor loadings III
0.67
SRT
0.50
UWT
0.63
VCT
0.65
OET
0.78
Percentage of variance explained
45.7
14.7
9.6
DISCUSSION The findings bring out a relatively complex set of results. It is clear that peoples’ subsistence activities do relate in important ways with their cultural features and cognitive characteristics. Patterning of gender differences in cognitive functioning seems to be linked with cultural demands placed on boys and girls in different subsistence level societies. The results generally support the hypothesis regarding the existence of cultural dimensions of societal size and social conformity, and their linkages with the subsistence economy of groups. The level of individual and social connectedness did not differ in HG and WE samples, but it did in the DA and IA ones, showing significantly greater social than individual connectedness. The pattern of their relationship with subsistence strategies was identical across the samples. In cognitive research, differentiation and integration have been regarded as two poles of the same dimension, the former emphasizing breaking apart the various parts of a unit; the latter laying emphasis on putting them together. This dimension stands in sharp contrast to contextualization. Thus, one would expect a negative relationship among differentiation, integration and contextualization. This expectation was not fully borne out of the findings. The outcomes of the factor analysis indicate neither differentiation nor contextualization as such neat dimensions as they have been conceptualized in cognitive research. The findings do broaden the scope of differentiation by suggesting positive correlation of integration measures with differentiation measures. A similar broader factor was evident in a previous work with these samples (Mishra et al., 1996). On the other hand, the scope of contextualization was limited by the organization of its measures under two separate factors, one relating to visual domain (LOT) and another to verbal domain (SRT and UWT).
298 R.C. Mishra and John W. Berry The evolutionary scale of Lomax and Berkowitz (1972) can be used to account for coherence of differentiation and integration measures. They had found these two as independent cultural dimensions in the samples of exclusive gatherers and of European and ‘old high cultures’. Over the middle range of subsistence strategies (hunting and agricultural), the two dimensions were found to be positively related. Our samples, including even the wage earners, broadly represented this middle range of cultures. A second explanation may be given in terms of the features of the test that was used for the study of differentiation. While in the EFT, the larger figure is well-integrated, and no background is provided by it, the SPEFT provides a context to the larger picture by using a highly familiar setting and a culturally salient story. Such contextual variations do create ambiguities concerning the definition of cognitive unit. Results obtained with such stimuli (for example, Kojima, 1978) have presented evidence that generally does not conform to the usual expectations made on the EFT. It appears that the use of meaningful objects as stimuli in both the SPEFT and VCT generated a common process, which resulted in the coherence of these measures. With respect to contextualization, the findings tentatively suggest that it is not a unitary process. Since the present factor analysis combines individual and group differences, a factor analysis within and across groups is needed in order to validate this claim. On the other hand, there was evidence for considerable variation according to the tasks. The findings were in the predicted direction on the visual task, but they deviated from expectations on the verbal task. For example, there was no difference among groups in terms of requests for extra information on familiar and unfamiliar problems of the SRT. The tendency to process syllogisms in a decontextualized manner and stories (UWT) in a contextualized manner presented us with evidence of ‘task-specific’ outcomes. It appears that participation of children in solving riddles (a popular practice in tribal cultures) helped the processing of syllogisms in a decontextualized manner. Thus, the expectation about the relationship between subsistence strategies and cognitive dimensions is fulfilled only for differentiation. For contextualization, the findings do not allow us to draw a general conclusion. At best, we can say that children from agricultural societies engage in more contextual processing than those of other samples. The conceptualization of integration, as well as its relationship with subsistence economy, also does not appear to be as neat as it is generally hypothesized in cognitive research. The task-specific relationship between subsistence strategies of groups with contextualization and integration in cognition draws attention to the need for more extensive research.
ACKNOWLEDGEMENTS The study was funded by grant ERIC/1224/31.3.1993 of the Educational Innovations and Research Committee, National Council of Educational Research and Training, New Delhi,
Cultural Adaptations and Cognitive Processes of Tribal Children in Chotanagpur
299
attributed to R.C. Mishra. The assistance of Sri Yugant Kumar in fieldwork is gratefully acknowledged.
REFERENCES Berry, J.W. 1976. Human Ecology and Cognitive Style: Comparative Studies in Cultural and Psychological Adaptation. New York: Sage/Halsted. ———. 1987. ‘The Comparative Study of Cognitive Abilities’. In S.H. Irvine and S. Newstead (eds), Intelligence and Cognition: Contemporary Frames of Reference (pp. 393–420). Dordrecht: Nijhoff. Berry, J.W., J.M.H. van de Koppel, C. Senechal, R.C. Annis, S. Bahuchet, L.L. Cavalli-Sforza and H.A. Witkin. 1986. On the Edge of the Forest: Cultural Adaptation and Cognitive Development in Central Africa. Lisse: Swets & Zeitlinger. Berry, J.W., J.A. Bennett and J.P. Denny. 1995. Ecological and Cultural Adaptation. Unpublished manuscript. Boldt, E.D. 1976. ‘Structural Tightness and Cross-Cultural Research’. Journal of Cross-Cultural Psychology, 7: 21–36. Boldt, E.D. and L.W. Roberts. 1979. ‘Structural Tightness and Social Conformity’. Journal of Cross-Cultural Psychology, 10: 221–30. Denny, J.P. 1986. ‘Cultural Ecology of Mathematics: Ojibway and Inuit Hunters’. In M. Closs (ed.), Native American Mathematics (pp. 129–80). Austin: University of Texas Press. Denny, J.P. and L. Davis. 1989. Contextualization During Reasoning. An Experimental Study of Amerindians. Unpublished manuscript. Gamble, J.J. and P.E. Ginsberg. 1981. ‘Differentiation, Cognition and Social Evolution’. Journal of CrossCultural Psychology, 12: 445–49. Goody, J. 1977. The Domestication of Savage Mind. Cambridge: Cambridge University Press. Iwawaki, S. 1986. ‘Achievement Motivation and Socialization’. In S.E. Newstead (ed.), Human Assessment: Cognition and Motivation (pp. 341–50). Dordrecht: Nijhoff. Iwawaki, S. and P.E. Vernon. 1988. ‘Japanese Abilities and Achievement’. In S.H. Irvine and J.W. Berry (eds), Human Abilities in Cultural Context (pp. 358–82). New York: Cambridge University Press. Kojima, H. 1978. ‘Assessment of Field Dependence in Young Children’. Perceptual and Motor Skills, 46: 479–92. Lomax, A. and W. Berkowitz. 1972. ‘The Evolutionary Taxonomy of Culture’. Science, 177: 228–39. Luria, A.R. 1976. Cognitive Development: Its Cultural and Social Foundations. Cambridge: Harvard University Press. McIntyre, L.A. 1976. ‘An Investigation of the Effect of Culture and Urbanization on Three Cognitive Styles and Their Relation to School Performance’. In G.E. Kearney and D.W. Mc Elwain (eds), Aboriginal Cognition (pp. 231–56). Canberra: Australian Institute of Aboriginal Studies. Mishra, R.C., D. Sinha and J.W. Berry. 1996. Ecology, Acculturation, and Psychological Adaptation: A Study of Adivasis in Bihar. New Delhi: Sage Publications. Murdock, G.P. 1969. ‘Correlation of Exploitation and Settlement Patterns’. In D. Damas (ed.), Contribution to Anthropology: Ecological Essays (pp. 129–46). Ottawa: National Museum of Canada. Scribner, S. 1977. ‘Mode of Thinking and Ways of Speaking.: Culture and Logic Reconsidered’. In P.N. Johnson-Laird and P.C. Watson (eds), Thinking: Readings in Cognitive Science (pp. 223–43). Cambridge: Cambridge University Press. Sinha, D. 1984. Manual for Story–Pictorial E.F.T. Varanasi: Rupa.
Chapter 22 An Eco-cultural Perspective on Cognitive Competence John W. Berry
CROSS-CULTURAL PSYCHOLOGY This chapter begins by outlining a general framework for linking the development and display of human behaviour to the contexts (ecological and cultural) in which an individual lives. It adopts a universalist perspective, within which the assumption is made that basic psychological processes are shared, species-wide phenomena upon which cultural experiences create behavioural variations. It further proposes that these variations are adaptive to habitat, in that they permit effective operation in the environment in which a person develops. The chapter then focuses on competence as one psychological feature that is both universal and adaptive, drawing upon concepts and empirical research from the field of cross-cultural psychology. It ends with a proposal for designing studies of competence in various cultural contexts that are both context-sensitive and comparative, allowing for an evaluation of the universalist assumption.
AN ECO-CULTURAL PERSPECTIVE For many years, I have advocated an eco-cultural perspective (Berry, 1966). It has evolved through a series of research studies devoted to understanding similarities and differences in cognition and social behaviour (Berry, 1976; Berry et al., 1986, 2000; Mishra et al., 1996) to a broad approach to understanding human diversity. The core ideas have a long history (Jahoda, 1995), and have become assembled into conceptual frameworks (Berry, 1975, 1995) used in empirical research and in coordinating textbooks in cross-cultural psychology (Berry et al., 1992/2002). Similar ideas and frameworks have been advanced both by anthropologists (for example, Whiting, 1974) and psychologists (for example,
An Eco-cultural Perspective on Cognitive Competence
301
Bronfenbrenner, 1979), who share the view that human activity can only be understood within the context in which it develops and takes place. The eco-cultural perspective is rooted in two basic assumptions, both deriving from Darwinian thought. The first (the universalist assumption) is that all human societies exhibit commonalities (‘cultural universals’), and that basic psychological processes are shared, species common characteristics of all human beings on which culture plays infinite variations during the course of development and daily activity. The second (the ‘adaptation’ assumption) is that behaviour is differentially developed and expressed in response to ecological and cultural contexts. This view allows for comparisons across cultures (on the basis of the common underlying process), but makes comparison worthwhile (using the surface variation as basic evidence). Whether derived from anthropology (for example, Murdock, 1975) or sociology (for example, Aberle et al., 1950), there is substantial evidence that groups everywhere possess shared socio-cultural attributes. For example, all peoples have language, tools, social structures (for example, norms, roles) and social institutions (for example, marriage, justice). It is also evident that such underlying commonalities are expressed by groups in vastly different ways from one time and place to another. Similarly, there is parallel evidence at the psychological level for both underlying similarity and surface variation (Berry et al., 1997). For example, all individuals have the competence to develop, learn and perform speech, technology, role playing and norm observance. At the same time, there are obviously vast group and individual differences in the extent and style of expression of these shared underlying processes. This combination of underlying similarity with surface expressive variation has been given the name universal by Berry et al. (1992/2002) to distinguish it from two other theoretical views: absolutism denies cultural influence on behavioural development and expression; while relativism denies the existence of common underlying psychological processes. Of course, while variations in behavioural expression can be directly observed, underlying commonalities are a theoretical construction and cannot be observed directly (Troadec, 2001). Paradoxically, this search for our common humanity can only be pursued by observing our diversity. And this dual task is the essence of cross-cultural psychology (Berry, 1969, 2000). Two basic assumptions of the eco-cultural approach were articulated at the outset: universalism and adaptation. While no claim can be made that these two assumptions have been verified, they have served as a useful and important heuristic in the field (see Troadec, 2001). One other theoretical issue has not yet been addressed. This is the question: Is culture conceptualized as an independent or as an organismic variable in the framework? My answer (Berry, 2000) is that it is both. To justify this view, it is helpful to recall the argument (Kroeber, 1917) that culture is superorganic, super meaning above and beyond, and organic referring to its individual biological and psychological bases. Two arguments were presented by Kroeber for the independent existence of culture at its own level. First, particular individuals come and
302 John W. Berry go, but cultures remain more or less stable. This is a remarkable phenomenon; despite a large turnover in membership with each new generation, cultures and their institutions remain relatively unchanged. Thus, a culture does not depend on particular individuals for its existence, but has a life of its own at the collective level of the group. The second argument is that no single individual ‘possesses’ all of the culture of the group to which one belongs; the culture as a whole is carried by the collectivity, and indeed is likely to be beyond the biological or psychological capacity (to know or to do) of any single person in the group. For example, no single person knows all the laws, political institutions and economic structures that constitute even this limited sector of one’s culture. For both these reasons, Kroeber considered that cultural phenomena are collective phenomena, above and beyond the individual person, and hence his term superorganic. This position is an important one for cross-cultural psychology since it permits us to employ the group–individual distinction in attempting to link the two, and possibly to trace the influence of cultural factors on individual psychological development. From the superorganic perspective, which also proposes that culture exists prior to any particular individual, we can consider culture as ‘lying in wait’ to pounce on newcomers (be they infants or immigrants), and to draw them into its fold by the processes of cultural transmission and acculturation (see Figure 22.1). Hence, we can claim that culture is, in important ways, an independent variable (or more accurately, a complex set of inter-related independent variables). However, these same two transmission processes lead to the incorporation of culture into the individual; hence culture also becomes an organismic variable. It is simultaneously outside and inside the individual. Being both ‘out there’ and ‘in here’ (Berry, 2000), the interactive, mutually influencing, character of culturebehaviour relationships becomes manifest. This view is indicated by the feedback loop shown in Figure 22.1, where individuals are in a position to influence, change and even destroy their ecosystem and cultural accomplishments. One continuing theme in cultural anthropology is that cultural variations may be understood as adaptations to differing ecological settings or contexts (Boyd and Richerson, 1983). This line of thinking, usually known as cultural ecology (Vayda and Rappoport, 1968), ecological anthropology (Moran, 1982; Vayda and McKay, 1975), or the ecosystem approach (Moran, 1990) to anthropology, has a long history in the discipline (see Feldman, 1975). Its roots go back to Forde’s (1934) classic analysis of relationships between physical habitat and societal features in Africa, and Kroeber’s (1939) early demonstration that cultural areas and natural areas co-vary in Aboriginal North America. Unlike earlier simplistic assertions by the school of ‘environmental determinism’ (for example, Huntington, 1945), the ecological school of thought has ranged from possiblism (where the environment provides opportunities, and sets some constraints or limits on the range of possible cultural forms that may emerge) to an emphasis on resource utilization (where active and interactive relationships between human populations and their habitat are analysed).
An Eco-cultural Perspective on Cognitive Competence
303
The earlier ecological approaches have tended to view cultural systems as relatively stable (even permanent) adaptations (as a state), largely ignoring adaptation (as a process), or adaptability (as a system characteristic) of cultural populations (Bennett, 1976). However, it is clear that cultures evolve over time, sometimes in response to changing ecological circumstances, and sometimes due to contact with other cultures. This fact has required the addition of a more dynamic conception of ecological adaptation as a continuous, as well as an interactive, process (between ecological, cultural and psychological variables). It is from the most recent position that we approach the topic. It is a view that is consistent with more recent general changes in anthropology, away from a ‘museum’ orientation to culture (collecting and organizing static artifacts) to one that emphasizes cultures as constantly changing, and being concerned with creation, metamorphosis and recreation. Over the years ecological thinking has influenced not only anthropology, but also psychology. The fields of ecological and environmental psychology have become fully elaborated (see Werner et al., 1997), with substantial theoretical and empirical foundations. In essence, individual human behaviour has come to be seen in its natural setting or habitat, both in terms of its development and its contemporary display. The parallel development of cross-cultural psychology (see Berry et al., 1997) has also ‘naturalized’ the study of human behaviour and its development. In this field, individual behaviour is accounted for to a large extent by considering the role of cultural influences on it. In my own approach, ecological as well as cultural influences are considered as operating in tandem, hence the term eco-cultural approach.
AN ECO-CULTURAL FRAMEWORK The current version of the eco-cultural framework (Figure 22.1) proposes to account for human psychological diversity (both individual and group similarities and differences) by taking into account two fundamental sources of influence (ecological and socio-political), and two features of human populations that are adapted to them: cultural and biological characteristics. These population variables are transmitted to individuals by various ‘transmission variables’ such as enculturation, socialization, genetics and acculturation. Our understanding of both cultural and genetic transmission have been greatly advanced by recent work on culture learning (for example, Tomasello et al., 1993) and on the human genome project (for example, Paabo, 2001). The essence of both these domains is the fundamental similarity of all human beings (at a deep level), combined with variation in the expression of these shared attributes (at the surface level). Work on the process and outcomes of acculturation has also been advancing (for example, Marin et al., 2001), necessitated by the dramatic increase in intercultural contact and change. To summarize, the eco-cultural framework considers human diversity (both cultural and psychological) to be set of collective and individual adaptations to context. Within
304 John W. Berry this general perspective, it views cultures as evolving adaptations to ecological and sociopolitical influences, and views individual psychological characteristics in a population as adaptive to their cultural context, as well as to the broader ecological and socio-political influences. It also views (group) culture and (individual) behaviour as distinct phenomena at their own levels, that need to be examined independently. FIGURE 22.1
Eco-cultural framework linking ecology, cultural adaptation and individual behaviour
Within psychology, the findings of the burgeoning field of environmental psychology have attempted to specify the links between ecological context and individual human development and behaviour. Cross-cultural psychology has tended to view cultures (both one’s own, and others one is in contact with) as differential contexts for development, and view behaviour as adaptive to these different contexts. The eco-cultural approach offers a ‘value neutral’ framework for describing and interpreting similarities and differences in human behaviour across cultures (Berry, 1994). As adaptive to context, psychological phenomena can be understood ‘in their own terms’ (as Malinowski insisted), and external evaluations can usually be avoided. This is a critical point, since it allows for the conceptualization, assessment and interpretation of culture and behaviour in non-ethnocentric ways. It explicity rejects the idea that some cultures or behaviours are more advanced or more developed than others (Berry et al., 1983; Dasen et al., 1979). Any argument about cultural or behavioural differences being
An Eco-cultural Perspective on Cognitive Competence
305
ordered hierarchically requires the adoption of some absolute (usually external) standard. But who is so bold, or so wise, to assert and verify such a standard? Finally, the socio-political context brings about contact among cultures, so that individuals have to adapt to more than one context. When many cultural contexts are involved (as in situations of culture contact and acculturation), psychological phenomena can be viewed as attempts to deal simultaneously with two (sometimes inconsistent, sometimes conflicting) cultural contexts. These attempts at understanding people in their multiple contexts is an important alternative to the more usual pathologizing of colonized or immigrant cultures and peoples. Of course, these intercultural settings need to be approached with the same non-ethnocentric perspective as cross-cultural ones (Berry, 1985). A core feature of the eco-cultural framework is the process of cultural transmission: how do features (ecological and cultural) of one’s groups become transmitted and incorporated into the day-to-day behaviour of individuals who develop in a particular habitat? One proposal (Berry, 1980) is rooted in the ecological thinking of Egon Brunswik’s ‘probabilistic functionalism’ (1955a, b), which employs an ‘arc model’ that links contexts to behaviour, through features of populations and individuals. The goals of the scheme (see Figure 22.2) are to link psychological effects or outcomes (on the right) to their contexts (on the left) across the scheme; to do so at four distinct levels (down the scheme) ranging from naturalistic (at the top) to controlled (at the bottom) forms of research; and to link the four levels of contexts by ‘nesting’ each one in the level above it. FIGURE 22.2
Cultural transmission: Linkages between contexts and outcomes
306 John W. Berry Traditionally, much of the discipline of psychology has attempted to comprehend behaviour as a function of stimuli impinging upon an individual. The approach of ecological psychology (for example, Barker, 1969; Brunswik, 1957) has noted that the stimuli usually employed in psychology represent only a very narrow range of all possible stimuli and that they are excessively artificial in character. As a result, ecological psychology has emphasized the need to study behaviour in more naturalistic contexts. Similarly, as we have mentioned frequently, cross-cultural psychology proposes that we should be attending to broad ranges of situations drawn from a cross-section of cultures. Sampling from new cultures also means sampling from the new physical environmental contexts in which the cultures are situated. Thus, it is essential that the extension of research crossculturally be accompanied by increased attention to the natural environmental settings of the cultures studied, a position similar to that espoused by ecological psychology and presented in our general framework (Figure 22.1). Figure 22.2 represents four relationships (perhaps sometimes causal linkages) between environmental contexts and behavioural outcomes. Towards the top of the model are natural and holistic contexts and outcomes, while at the bottom they are more experimental (controlled and reductionistic). Looking in more detail at the environmental contexts, the ecological context is the ‘natural–cultural habitat’ of Brunswik (1957) or the ‘preperceptual world’ of Barker (1969). It consists of all the relatively stable and permanent characteristics of the habitat that provide the context for human action, and includes the population-level variables identified in Figure 22.1: the ecological context, the socio-political context, and the general cultural and biological adaptations made by the group. Nested in this ecological context are two levels of the ‘life space’ or ‘psychological world’ of Lewin (1936). The first, the experiential context, is that pattern of recurrent experiences that provides a basis for individual learning and development; it is essentially the set of independent variables present in a particular habitat during the development of behavioural characteristics. These variables include such day-to-day experiences as child-rearing practices, occupational training and education. The other, the situational context, is the limited set of environmental circumstances (the ‘setting’ of Barker, 1969), which may be observed to account for particular behaviours at a given time and place. They include features such as specific roles or social interactions that can influence how a person will respond to that setting. The fourth context, the assessment context, represents those environmental characteristics, such as test items or stimulus conditions, that area designed by the psychologist to elicit a particular response or test score. The assessment context may or may not be nested in the first three contexts; the degree to which it is nested represents the ecological validity of the task. Paralleling these four contexts are four behavioural outcomes. The first, customs, refers to the complex, long-standing and developed behaviour patterns in the population or
An Eco-cultural Perspective on Cognitive Competence
307
culture that are in place as a traditional response to the ecological context. Customs include established, collective and shared patterns of behaviour exhibited by a cultural group. The second, repertoire, is the relatively stable complex of behaviours that have been learned over time in the recurrent experiential or learning context. Included are the skills, traits and attitudes that have been nurtured in particular roles or acquired by specific training or education whether formal or informal. The third effect, actions, connotes those behaviours that appear in response to immediate stimulation or experience. In contrast to repertoire, they are not so much a function of role experience or long-term training, but appear in reply to immediate situational experiences. The fourth effect, scores, is comprised of those behaviours that are observed, measured and recorded during psychological assessment (experiments, interviews or testing). If the assessment context is nested in the other contexts, the scores will be representative of the repertoire of the organism and the customs of the population. Relationships can be traced between the environmental contexts and the behavioural outcomes across the scheme (dotted lines in Figure 22.2). The first level is concerned with the life situation (in physical, environmental and cultural terms) and its relationship to daily customs and practices of the population. It is here that other disciplines (such as anthropology and ecology) can supply valuable information to cross-cultural psychology. The second level is concerned with tying together recurrent experiences of individuals with their characteristic repertoire of behaviours. The third level is interested in more specific acts as a function of immediate and current experience; and the fourth level is devoted to the systematic study of relationships between stimuli and the scores obtained by individuals. A recurrent problem for general psychology, in terms of this scheme, has been the difficulty of contributing to an understanding of relationships at the three higher levels while collecting data almost exclusively at the lowest level. The problem facing crosscultural psychology is more complex. Rather than ascend the reductionistic–holistic dimension to achieve ecological validity, cross-cultural psychology has typically failed to work systematically at all levels to achieve a specification of context variables that are responsible for task performance and behavioural variation across natural habitats. In our view, only when tests and experimental tasks and outcomes (scores) are understood in terms of their relationships with variables in the upper levels of the scheme, will crosscultural psychologists be able to grasp the meaning of their data.
CULTURAL DIFFERENTIATION OF COMPETENCE Since there is no culture-free behaviour, there can be no-culture-free competence. This view is central to the proposal made by Ferguson (1956), who argued that ‘Cultural
308 John W. Berry factors prescribe what shall be learned and at what age; consequently different cultural environments lead to the development of different patterns of ability’ (p. 121). Embedded in this ‘Law of Cultural Differentiation’ (as it was named by Irvine and Berry, 1988) is the distinction between a species-shared process (learning) and cultural-variable outcomes (abilities). As we have argued, the existence of process commonality and outcome variability characterizes the universalist approach. In this section, we examine some implications of this distinction for understanding human cognition in context. There is an adage that states that ‘You can’t tell how far a person has progressed unless you know where he is going’ (Irvine and Berry, 1988). In this saying lies the core of the issue addressed in this section. In the general developmental literature, work on parental etho-theories has revealed that all societies espouse developmental goals for their children, which then become socialized by parents and others responsible for nurturing (see Segall et al., 1999, Ch. 3). The concept of developmental niche (Super and Harkness, 1986) incorporates this basic idea as one of its core components, as well the early work by Barry et al. (1959) on emphases in socialization. Since societies vary in their ecological demands, and in their cultural adaptations to these demands (including their socialization emphases), we have a solid basis for studying variations in developmental goals. Societies do not all share the same adaptive challenges, and hence it is unlikely that they would all direct development towards the same goals. In the Eco-cultural Framework, there is a second context (the socio-political) that imposes new challenges for developing individuals and their societies. The process of acculturation following culture contact may interfere with those developmental goals that have been adaptive to the ecological context. If so, two sets of demands may conflict; and if the new set dominates (for example, in schooling or employment settings), previously adaptive competencies may no longer be the most appropriate ones. However, in acculturation theory (Berry, 2003) such domination is not the only way to deal with contact and change, nor is it the most functional. Assimilation (loss of heritage culture, combined with the imposition of the dominant culture) is less functional than integration (selected maintenance of heritage culture, combined with learning useful features of the dominant culture). In essence, assimilation replaces heritage developmental goals with those of the dominant society, while integration values both, and seeks to balance them so that competence in both societies is permitted, sought and developed. Thus the study of developmental goals during acculturation requires concepts that provide alternatives to the common assumption that all groups in a society should share the same competencies. In this way, we can avoid the inevitable labelling of cognitive differences as cognitive deficiencies.
An Eco-cultural Perspective on Cognitive Competence
309
CONCLUSIONS In this chapter, I have outlined an eco-cultural approach to understanding the development and display of cognitive competence. It takes the position that since all human behaviour is nurtured in specific cultural contexts, it is only possible to discover competence when these contexts are understood. Much research on cognition has ignored this fact, with the result that much of the assessment of culturally diverse populations, and the consequent placement and training (in schools and elsewhere) of individuals, has misjudged human competence. In my view, only when the ecological settings, cultural adaptations and the demands of living in culture-contact situations are taken into account will a valid appraisal of cognitive competence be achieved.
REFERENCES Aberle, D.F., A. K. Cohen, A. Davis, and F. X. Sutton. 1950. ‘Functional Prerequisites of Society’. Ethics, 60: 100–11. Barry, H., I. Child, and M. Bacon. 1959. ‘Relations of Child Training to Subsistence Economy’. American Anthropologist, 61: 51–63. Bennett, J. 1976. The Ecological Transition. London: Pergamon. Berry, J.W. 1966. ‘Temne and Eskimo Perceptual Skills’. International Journal of Psychology, 1: 207–29. ———. 1967. ‘Independence and Conformity in Subsistence-Level Societies’. Journal of Personality and Social Psychology, 7: 415–18. ———. 1969. ‘On Cross-Cultural Comparability’. International Journal of Psychology, 4: 119–28. ———. 1975. ‘An Ecological Approach to Cross-Cultural Psychology’. Nederlands Tijdschrift voor de Psychologie, 30: 51–84. ———. 1976. Human Ecology and Cognitive Style: Comparative Studies in Cultural and Psychological Adaptation. New York: Sage/Halsted. ———. 1980. ‘Ecological Anlayses for Cross-Cultural Psychology’. In N. Warren (ed.), Studies in CrossCultural Psychology (pp. 157–89). London: Academic Press. ———. 1985. ‘Cultural Psychology and Ethnic Psychology’. In I. Reyes, Lagunes and Y. Poortinga (eds), From a Different Perspective (pp. 3–15). Lisse: Swets & Zeitlinger. ———. 1994. ‘An Ecological Approach to Cultural and Ethnic Psychology’. In E. Trickett (ed.), Human Diversity (pp. 115–41). Jossey-Bass. ———. 1995. ‘The Descendants of a Model’. Culture & Psychology, 1: 373–80. ———. 2000. ‘Cross-Cultural Psychology: A Symbiosis of Cultural and Comparative Approaches’. Asian Journal of Social Psychology, 3: 197–205. ———. 2003. ‘Conceptual Approaches to Acculturation’. In K. Chun, P. Balls-Organista, and G. Marin (eds), Acculturation: Advances in Theory, Measurement and Applied Research (pp. 3–21). Washington: APA Books. Berry, J.W., Y.H. Poortinga, J. Pandey, P.R. Dasen, T.S. Saraswathi, M.S. Segall and C. Kagitcibasi. 1997. Handbook of Cross-Cultural Psychology, 3 volumes. Boston: Allyn & Bacon.
310 John W. Berry Berry, J.W., P.R. Dasen and H.A. Witkin. 1983. ‘Developmental Theories in Cross-Cultural Perspective’. In L. Alder (ed.), Cross-Cultural Research At Issue (pp. 13–21). New York: Academic Press. Berry, J.W., J.M.H. van de Koppel, C. Sénéchal, R.C. Annis, S. Bahuchet, L.L. Cavalli-Sforza and H.A. Witkin. 1986. On the Edge of the Forest: Cultural Adaptation and Cognitive Development in Central Africa. Lisse: Swets and Zeitlinger. Berry, J.W., J.A. Bennett and J.P. Denny. 2000. ‘Ecology, Culture and Cognitive Processing’. Paper presented at IACCP Congress. Pultusk, Poland. Berry, J.W., H.H. Poortinga, M.H. Segall and P.R. Dasen. 1992/2002. Cross-Cultural Psychology: Research and Applications. New York: Cambridge University Press. Boyd, R. and P. Richerson. 1983. ‘Why is Culture Adaptive?’ Quarterly Review of Biology, 58: 209–14. Bronfenbrenner, U. 1979. The Ecology of Human Development. Cambridge: Harvard University Press. Brunswik, E. 1957. ‘Scope and Aspects of the Cognition Problem’. In A. Gruber (ed.), Cognition: The Colorado Symposium (pp. 63–92). Cambridge: Harvard University Press. Barker, R. 1969. Ecological Psychology. Stanford: Stanford University Press. Dasen, P.R., J.W. Berry and H.A. Witkin. 1979. ‘The Use of Developmental Theories Cross-Culturally’. In L. Eckensberger, W. Lonner, and Y. Poortinga (eds), Cross-Cultural Contributions to Psychology (pp. 69–82). Lisse: Swets & Zeitlinger. Feldman. 1975. ‘The History of the Relationship Between Environment and Culture in Ethnological Thought’. Journal of the History of the Behavioural Sciences, 110: 67–81. Ferguson, G. 1956. ‘On Transfer and the Abilities of Man’. Canadian Journal of Psychology, 10: 121–31. Forde, D. 1934. Habitat, Economy and Society. New York: Dutton. Huntington, E. 1945. Mainsprings of Civilization. New York: John Wiley. Irvine, S.H. and J.W. Berry. 1988. ‘The Abilities of Mankind’. In S.H. Irvine and J.W. Berry (eds), Human Abilities in Cultural Context (pp. 3–59). New York: Cambridge University Press. Jahoda, G. 1995. ‘The Ancestry of a Model’. Culture & Psychology, 1: 11–24. Kroeber, A. 1917. ‘The Superorganic’. American Anthropologist, 19 : 163–213. ———. 1939. Cultural and Natural Areas of Native North America. Berkely: University of California Press. Lewin, K. 1936. Principles of Topological Psychology. New York: McGraw-Hill. Marin, G., P. Balls-Organista and K. Chung (eds). 2001. Acculturation. Washington: APA Books. Mishra, R.C., D. Sinha and J.W. Berry. 1996. Ecology, Acculturation and Psychological Adaptation: A Study of Adivasi in Bihar. New Delhi: Sage Publications. Moran, E. 1982. Human Adaptability: An Introduction to Ecological Anthropology. Boulder: Westview Press. ———. (ed.). 1990. The Ecosystem Approach in Anthropology. Ann Arbor: University of Michigan Press. Murdock, G.P. 1975. Outline of Cultural Materials. New Haven: Human Relations Area Files. Paabo, S. 2001. ‘The Human Genome and Our View of Ourselves’. Science, 291: 1219–20. Segall, M.H., P.R. Dasen, J.W. Berry and Y.H. Poortinga. 1999. Human Behaviour in Global Perspective: Introduction to Cross-Cultural Psychology (Second edition). Boston : Allyn & Bacon. Super, C. and S. Harkness. 1986. ‘The Developmental Niche’. International Journal of Behavioural Development, 9: 545–969. Tomasello, M., A. Kruger and H. Ratner. 1993. ‘Culture Learning’. Behavioral and Brain Sciences, 16: 495–552. Troadec, B. 2001. ‘Le modèle écoculturel: un cadre pour la psychologie culturelle comparative’. International Journal of Psychology, 36: 53–64. Vayda, A.P. and B. McKay. 1975. ‘New Directions in Ecology and Ecological Anthropology’. Annual Review of Anthropology, 4: 293–306.
An Eco-cultural Perspective on Cognitive Competence
311
Vayda, A.P. and R. Rappoport. 1968. ‘Ecology, Cultural and Non-Cultural’. In J. Clifton (ed.), Cultural Anthropology (pp. 641–62). Boston: Houghton Mifflin. Werner, C., B. Brown and I. Altman. 1997. ‘Environmental Psychology’. In J.W. Berry, M.H. Segall and Kagitcibasi (eds), Handbook of Cross-Cultural Psychology, Vol. 3, Social Behaviour and Applications (pp. 253–90). Boston: Allyn & Bacon. Whiting, J.W. M. 1974. A Model for Psychocultural Research, Annual Report. Washington: American Anthropological Association.
SECTION
V
Cognitive Development and Intervention
D
evelopment can be defined as systematic changes and continuities in the individual that occur between conception and death, or from ‘womb to tomb’. Development entails many changes, and these changes are described as systematic, orderly, patterned and relatively enduring (Sigelman and Rider, 2003). The human development falls into three broad domains: physical development, psycho-social development and cognitive development. Here, the focus is on cognitive development, which consists of changes and continuities in perception, language, learning, memory, problem solving, and other cognitive processes. It was once believed that infants lacked the ability to think or form complex ideas and remained without cognition until they learned language. At present, the general view on infant cognition is that babies are aware of their surroundings and interested in exploring them, and actively learning about the environment even during infancy. They gather, sort and process information from around them, using the data to develop various cognitive skills that enables them to adjust and cope with environmental demands. All cognitive processes including attention, working memory (WM), language comprehension and problem solving demonstrate growth during childhood. The patterns of growth of these cognitive processes provide the basis on which neuropsychological tests can be developed for children. Normal development of the cognitive processes is very necessary, though there are very few studies on normative cognitive development. Sometimes children display developmental discrepancies, with delays in one domain of development in relation to others. For many children, delays are modest and temporary. For other children, however, delays are extensive and pervasive and may lead to developmental disorders. Developmental disorders of cognition are those where a cognitive system or systems have never attained a normal level over the course of cognitive development. There have been attempts to study normal and atypical cognitive development and also on effective methods of intervention. Any cognitive intervention is based on the assumption of neural plasticity, which is higher in children. Interventions can be planned to remediate children with specific cognitive disorders like learning disabilities, and also to enhance cognitive performance on tasks of specific skills like calculation, using novel behavioural methods. Developmental changes in cognitive functions underlie the dynamic interaction between the brain and environment. Changes in cognitive performance are also a result of experience and learning over the years from childhood to adolescence, which cannot be ignored (Johnson and Munakata, 2005). Other than specific developmental disorders caused by neurobiological factors, normal cognitive development is also affected by understimulation and adverse environmental conditions. Such factors also influence cognitive development, particularly observed in underprivileged societies such as the rural population in India. Efforts to conduct carefully drafted cognitive stimulation programmes for such populations have been an important contribution to the field. Among the various developmental disorders, research on dyslexia has ranged from epidemiological studies to exploring the cognitive mechanisms of dyslexia using behavioural and imaging methods.
316 Advances in Cognitive Science Developmental dyslexia is a major language-related disorder that affects children. Dyslexia is the most common of the learning disorders that interferes with the child’s ability to acquire speech reading despite average intellectual functions. Dyslexia is supposed to be a failure in learning to optimize the coordination of sub-processes involved in reading, leading to errors in integrating reading-related information represented in WM. Reading involves visual and semantic decoding, temporal processing, phonological processing, orthographic, syntactic and contextual analysis, and comprehension. An inefficient synchrony of these underlying mechanisms results in reading disability (Lachmann, 2002). Research on dyslexia in India has been just a decade old, and epidemiological studies have been particularly difficult because of the complexities determined by the varied socio-cultural context. Dyslexia research has primarily focussed on explaining the cognitive mechanisms of difficulties in English language, but very few studies have looked at reading impairments in Indian languages. Lachmann discusses some of the experiments he and his colleagues have performed on possible mechanisms that may underlie by looking at symmetry generalization among dyslexic children. Reversal errors are common in developmental dyslexia, and Lachmann and his colleagues proposed that the failure to suppress symmetry generalization results in reversal errors. Lachmann extensively reviews the literature and relevant findings on reversal errors. Lachmann argues that dyslexics who make reversal errors have problems in coordinating the visual and phonological processes due to symmetry generalization. Lachmann discusses results from an experiment with the same—different tasks performed with letters and pattern (shape) stimuli. The experiment has two conditions: physical and categorical. They found that children with disabilities made a large number of errors in the physical condition, especially with letters. Lachmann also discusses results from a mental rotation task in which disabled readers show a similar pattern with respect to the amount of rotation but with larger reaction times (RTs). Lachmann discusses the implications of these experiments and other studies, and also points to the importance of the presence of sub-groups among dyslexics. Lachmann’s chapter highlights one of the cognitive mechanisms operating in dyslexia pertaining to the English language, but there is very little research to explore similar mechanisms related to reading difficulties in Indian languages such as Hindi. There are significant differences among languages like English and Hindi, and these differences are important in understanding dyslexia in the Hindi-speaking population. It is also important to understand the cognitive profiles of dyslexic children, especially in an Indian setting. The chapter by Kar and Tripathi provides such information, which will be very important in developing intervention techniques for dyslexic children. They assessed children for Specific Learning Disability (SLD) from two schools using the National Institute of Mental Health and Neurosciences (NIMHANS) index of specific learning disability, and identified 18 children with dyslexia. The dyslexic children had average/above average scores on the test of intellectual functions. Some of these children
Cognitive Development and Intervention
317
had behavioural problems. The performance on reading tests indicated errors pertaining to deficits in phonological decoding in Hindi and English language, poor reading comprehension, poor short-term memory, and deficits in visuo-spatial analysis and perceptual organization. The children’s performance in Hindi and English writing tests indicated deficits in orthographic processing. In order to study the cognitive processes involved in reading and writing, Kar and Tripathi examine the cognitive profiles of these children on tests of planning, attention, and simultaneous and successive processing using the Cognitive Assessment System (CAS). They evaluate the performance of each child with reference to the norms of CAS. In general, the children had lower scores on all the components of CAS. In addition, Kar and Tripathi find lower performance with simultaneous processing than successive processing. This differs from most reports, which suggest that there may be greater successive processing deficits in dyslexics. Children with dyslexia seem to have specific difficulty in establishing a logical inter-relationship between the perceptual and conceptual components of language, as indicated by deficient performance on simultaneous and successive processing tests. They also seem to have difficulty in serial temporal processing of letter strings, as observed in their performance on successive processing tests. These results need further verification with a larger sample, but they certainly give an insight into the underlying cognitive processes that need to be worked upon while planning a remediation programme for children with dyslexia. Over the past 25 years, extensive research has been done on many aspects of rehabilitation, including the effectiveness of treatments aimed to reduce the problems arising from cognitive deficits and disorders. Effective treatment of cognitive deficits is very important because of several reasons. First, there is the growing increase in patients recognized as suffering from cognitive deficits. Rehabilitation professionals increasingly encounter such patients, and the consequences of these deficits are not trivial. Over the past decade, there has been a strong interest in the cognitive neuropsychological approach as a model for planning rehabilitation interventions. The cognitive-neuropsychological approach to treatment is not confined to acquired disorders, but is also intended to be applicable to developmental disorders such as specific language impairment or developmental dyslexia. Cognitive-neuropsychological rehabilitation refers to the use of models of normal processing as an aid to rehabilitation (Coltheart, 1994). Such model-based assessment allows the clinician to determine which modules of the relevant informationprocessing system are functioning abnormally and which have been spared. The findings from basic cognitive neuroscience research have already led to the development of various cognitive intervention approaches. The next two chapters focus on these developments. The chapter by Aalsvoort focuses on cognitive intervention in a western setting. Aalsvoort takes a socio-cultural perspective to examine the relationship between social play and emergent numeracy of at-risk Grade 2 students in primary schools. The study utilizes a longitudinal experimental design and follows students until Grade 3. After careful selection, two groups of children were placed in the control and
318 Advances in Cognitive Science experimental conditions. The experimental condition consisted of six 20-minute sessions the children participated in over a period of three weeks, followed by a seventh session two months after the sixth one. However, the control condition had only play sessions one and seven. Each play session included an invitation to build a zoo using toy-animals and wooden blocks, and it was videotaped, transcribed and analysed. Aalsvoort evaluated academic performance at the end of Grade 2 and in the middle of Grade 3, and these results, along with the results related to mathematical knowledge during the play sessions (one and seven), are presented. Aaalsvoort discusses the findings and its implications for school curriculum. The chapter by Kapur focuses on cognitive intervention in an Indian setting. Kapur evaluates an intervention package for cognitive stimulation of rural school children. The children were assessed before and after intervention using various cognitive tests on attention, intelligence, language, arithmetic and creative functions. The study was performed with a large sample of children from both genders ranging from class 1 to 9. Developmentally appropriate, child-friendly activities such as play, art, craft, dance, music and story telling were used with children for more than three years. Significant improvement was obtained with all the parameters, but there were variations across age, gender and the tests. The encouraging results point to the utility of this intervention programme, especially in developing countries like India. Further work on cognitive intervention is needed to make full use of the advances in cognitive science and enhance the quality of cognitive life of children and adults throughout the world.
REFERENCES Coltheart, M., A. Bates and A. Castles. 1994. ‘Cognitive Neuropsychology and Rehabilitation’. In G.W. Humphreys and M.J. Riddoch (eds), Cognitive Neuropsychology and Cognitive Rehabilitation (pp. 17–38). London: Lawrence Erlbaum Associates. Johnson, M.H. and Y. Munakata. 2005. ‘Cognitive Development: At the Crossroads?’. Trends in Cognitive Science, 9: 91. Lachmann, T. 2002. ‘Reading Disability as Deficit in Functional Coordination’. In E. Witruk, A.D. Friederici and T. Lachmann (eds), Basic Functions of Language, Reading, and Reading Disability (pp. 165–98). Netherlands: Kluver Academic Publishers. Sigelman, C.K. and E.A. Rider. 2003. Life-Span Human Development. Australia: Thomas Wadsworth.
Chapter 23 Experimental Approaches to Specific Disabilities in Learning to Read: The Case of Symmetry Generalization in Developmental Dyslexia Thomas Lachmann
INTRODUCTION
D
evelopmental dyslexia is defined as a specific disability in learning to read and to spell adequately despite at least normal intelligence, adequate instruction, socio-cultural opportunities, and the absence of sensory defects in vision and hearing. By this definition, instead of aetiologically grounded criteria, the diagnosis rests upon a criterion of discrepancy between the reading performance expected from measures of general intelligence and the observed reading performance—or, in other words, the discrepancy between how a child is expected to learn to read and how, in fact, she/he does. This definition is poorly descriptive because more than 100 years of scientific research have failed to provide a consistent account of the aetiology of developmental dyslexia (Miles and Miles, 1999). Therefore, the diagnosis remains on the level of symptoms, instead of causing deficits. One of the major symptoms of developmental dyslexia are reversal errors. Children confuse, for instance, letters which are horizontally or vertically symmetrical to each other or rotated (for example, p/q; b/d; p/d). These errors (static reversals) are normal in beginning readers (Hicks, 1981), but become very rare after developing reading practice. In dyslexics, however, they still occur even after years of reading experience. Orton (1925, 1928) postulated that these problems reflect the cardinal symptom of developmental dyslexia and that these confusion errors (‘strephosymbolia’) are caused by a faulty development of cerebral dominance and sub-optimal inter-hemispheric communication.
320 Thomas Lachmann In the last decades, phonological skills were considered most relevant in explaining dyslexia (for example, Bradley and Bryant, 1978; Liberman et al., 1971). In this context, it was often argued that Orton wrongly believed that dyslexics have a ‘perceptual deficit’ and actually see letters in a wrong orientation (Vellutino, 1977). However, as discussed at great length elsewhere (Corballis and Beale, 1993; Lachmann, 2002), Orton assumed reversals to be the result of labelling problems in memory, and not of problems in visual perception. It is obvious that many of Orton’s particular assumptions, such as a serial order of processing steps in word processing, the existence of a separate word centre, and his assumption about the inter-hemispheric communication are antiquated or simply wrong. However, as we will argue, his basic ideas may still be useful in understanding the origin of one group of symptoms of dyslexia, the orientation errors. There are a couple of alternative models which try to connect these ideas with more recent neurological theories (Corballis and Beale, 1993), and with more recent experimental results (Brendler and Lachmann, 2001; Lachmann and Geyer, 2003; Lachmann and van Leeuwen, 2007). One is the multi-causal Functional Coordination Deficit (FCD) model (Lachmann, 2002). Within this model, reversal errors are explained as a result of a failure in suppressing symmetry generalization in reading. Symmetry generalization, as a result of evolutionary and individual development, is a mechanism assumed to be learned as an infant to warrant behavioural advantages such as object constancy. In reading, however, this mechanism may be a hindrance because graphemes are visual symbols, and as such they have to have a non-ambiguous relation to the respective phonological information they represent. It was argued that learning to read is comprised of learning to treat graphemes as symbols instead of objects. This skill is achieved very early, during the first stage of reading acquisition. A failure in the complete suppression of symmetrical information in the representation of visual symbols during reading produces ambiguous relations between visual and phonological information, and thus disturbs the functional coordination and may cause problems in learning to read. In this respect, reversal errors do not reflect a visual perceptual deficit (as announced by Vellutino, 1977) or an incomplete hemispheric dominance (as postulated Orton, 1925), and they do not reflect the one and only cardinal symptom of dyslexia. Instead, the FCD model assumes that this kind of problems occurs in a subgroup but not in all dyslexics (Lachmann, 2002; Lachmann and Geyer, 2003). In our own studies (Lachmann and van Leeuwen, 2004a; van Leeuwen and Lachmann, 2004) we have shown that objects (pseudo-letters and shapes) are perceived differently from letters even at a very early stage of information processing (early feature integration). In recognition tasks, using irrelevant form-congruent versus form-incongruent surroundings, letters evoked negative congruence effects on reaction times (RTs), which contrasted with the positive congruence effects for objects (Lachmann and van Leeuwen, 2004a). In contrast to these very early effects, reversal errors are assumed to occur relatively late
Experimental Approaches to Specific Disabilities in Learning to Read
321
in information processing, that is, they represent a representational deficit of grapheme– phoneme correspondence. This assumption indicates that reversals are not necessarily the result of problems in the processing of spatial information per se, neither for objects nor for symbols. They are rather an effect of a disability to represent symbols, such as graphemes, strictly view-point dependent. In particular, problems occur when different orientations have to be represented and performed distinctly to be connected unambiguously with distinct phoneme categories.
STUDIES ON REVERSAL ERRORS In the last three decades, the relevance of reversal errors was questioned by those who believed them to merely be yet another illustration of phonological deficits (Bigsby, 1985). While dyslexics in general may have no problems in distinguishing non-letter objects, it was believed that the confusion of reversed letters was based on their phonological similarity (‘bee’, ‘dee’), and it was argued, as mentioned earlier, that dyslexics can ‘see’ (Vellutino, 1987) the distinction between ‘d’ and ‘b’ very well. However, as we have argued before, Orton never said they could not. He believed that the problem resides at a level of abstract labelling of the visual representation by a phonological code (Corballis and Beale, 1993; Lachmann and Geyer, 2003). In the following discussion, we will introduce some of the classical studies that doubted the importance of reversal errors. The classical study by Liberman et al. (1971) was one of the most influential early studies that explicitly questioned Orton’s theoretical approaches to reading disability (but see also Hildreth, 1934, as an example for earlier critique). The study of Liberman et al. (1971) was aimed directly at testing Orton’s (1925) theory of anomalous cerebral dominance as the universal cause for developmental dyslexia. In this context, the investigation concentrated on the analysis of the proportion of reversal errors to other reading errors (Orton’s secondary symptoms), especially of the relation between errors based on optical (reversals) and on linguistic properties (vowel and consonant errors), and of the relationship between reversals of sequence (kinetic reversals) and reversals of orientation (static reversals). The second graders of a primary school were asked to read a word list consisting of 60 monosyllabic words, including primer-level sight words, nonsight words, and words where reversals (both types) were possible versus words where reversals were not possible (for example get, was, bed, bet, not, pin), respectively. Those 18 students who performed within the lower third, measured by the rate of errors made while reading this word list, were defined as ‘poor readers’. The other 36 students were considered as control group of normal readers. In addition to the word list, all students had to perform the Gray Oral Reading Test and a matching task, where a given letter was to be matched to one out of a group of five, four of which were reversible. The results were analysed for the whole population, as well as for poor and normal readers separately.
322 Thomas Lachmann Liberman and her colleagues found a significant correlation between the total performance in the reading test and the more artificial task of reading monosyllabic words in isolation. This may reflect a stage in reading development which is based on scanning syllables rather than larger chunks of text. The main results the authors found were as follows. First, nearly all of the reversals were done by the poor reader group. Second, vertical b–d confusion (which they called horizontal transformation) occurred with as great a frequency (10.2 in average for both directions) as b–p confusion (11.4 in average for both directions), while both confusions obviously occurred more often than d–p confusion (1.1) as well as confusions of b, d, and p with g (0.9 in average for both directions). Furthermore, the authors found that for static reversals, the presence of equivalent shapes within the alphabet is important, and the reversibility of a letter is not by itself a sufficient condition for confusion. Another important finding was that normal as well as poor readers made less reversal errors than other reading errors (error percentages according to the opportunities of occurrence in poor readers: 6.3 kinetic reversals; 12.7 static reversals; 16.3 consonant confusion; 26.8 vowel confusion; error percentages according to the total number of errors in the same sequence: 10, 15, 32, 43). Within the group of poor readers, individual differences were larger and test–retest reliability was lower for reversals in comparison to vowel and consonant errors, indicating that reversals do not reflect a constant proportion of all errors as would be expected from assuming them to be cardinal symptoms. Finally, static and kinetic reversals in poor readers were found to be uncorrelated to each other, but both reversal types were correlated with other measures such as the reading test, but in a different way. These findings were counted as evidence for the importance of linguistic determinants in reading problems and against the theory of strephosymbolia as a consequence of hemispheric imbalance expressed by reversals. However, these results may be representative for the lower third of a regular second grade class only, but not for developmental dyslexics, not for strephosymbolics, and not even for poor readers. The results by Liberman and her colleagues are not sufficient to disprove Orton’s concept. The merit of their work does consist much more of a stronger emphasis on linguistic aspects in order to understand reading disability. Their publication is in fact a milestone in reading disability research, but it delivers no evidence against the significance of reversals. In a follow-up study, published by Fisher et al. (1978), the performance of the poor reading students described by Liberman and her colleagues was compared to that of dyslexic children, defined by the discrepancy criterion (normal intelligence with a retardation of at least 18 months in Gray Reading Test performance), of the same age and intelligence. The authors showed that while both groups performed almost equally in the reading test, the dyslexic group (smaller data set) made significantly more errors in the Word List Test. However, these errors were relatively made in the same proportion as by the poor readers group (error percentages according to the opportunities of occurrence: 8.3 kinetic reversals, 11.4 static reversals, 32.9 consonant confusion, 40.3 vowel confusion),
Experimental Approaches to Specific Disabilities in Learning to Read
323
underlining the importance of linguistic aspects. In contrast to the poor readers, the dyslexic children performed more consistently in their reversal errors, and both kinds were significantly correlated. Lyle and Goyen (1968) presented letters and shapes tachistoscopically within different task settings in order to study the performance of disabled and normal reading primary school children of different ages. The authors found disabled readers to perform worse than normal readers on all conditions. Interestingly, there was an Age (second versus third graders) × Group interaction in some of their conditions; the group effect was greater at the younger age level. A further question was whether or not the higher number of errors made by retarded readers could be characterized by reversal and rotation tendencies. The authors analysed static reversals (simultaneous and delayed condition), kinetic reversals (sequenced condition) and the remaining ‘miscellaneous’ errors separately. None of the analyses, however, revealed effects. For letters and shapes, the proportion between reversals and other errors did not differ between the groups. This was interpreted as evidence against Orton’s theory (1925) of reversals as the cardinal symptom of reading disability and as result of problems in binding visual and phonological information adequately (to speak in nowadays words). Instead, problems in the speed of visual decoding were assumed to be the reason for reading problems, independent from phonological information possibly inherent to be the stimulus. On the other hand, the results of a reading test applied by these authors showed indeed a higher number of reversals in disabled readers, and a higher proportion of reversals in relation to the total error rate. Lyle and Goyen suggested problems in verbal labelling as one of the reasons for these phenomena. This, however, does not separate them from Orton (1925); what it is in contrast to Orton is the assumption about the underlying mechanism producing these problems. Lyle and Goyen argued that the speed of perceptual encoding may be a reason for problems in labelling, since it reduces the load of visual information in a given time unit. Grosser and Trzeciak (1981) studied the significance of reversal errors for reading disability by using a threshold and a masking paradigm in which the exposure time needed to name reversible (b, d, p, q) or non-reversible letters (w, x, u, o) correctly was used as a dependent variable. The performance of 29 disabled readers, aged between six and 14 years, and 15 normal readers of about the same age was compared. The authors found significant group effects for all conditions; disabled readers needed longer exposure times to name the letters correctly. However, there was no interaction between the group and the letter sets. In particular, all participants needed longer exposure times to name reversible letters; the proportion between reversible letters and non-reversible letters, however, was about the same in both groups. This result, coupled with the finding that normal reading younger children perform worse than the older ones, whereas such a correlation was not found within the group of reading disabled children, was interpreted as evidence for a disturbance of the development of visual perception in disabled readers. Within the scope of this chapter—the question whether or not reversals can be thought as characteristic for
324 Thomas Lachmann dyslexia—Grosser and Trzeciak’s empirical findings led to the assumption that reversals have no relation to specific reading disability. The naming speed of disoriented letters presented in the left or the right visual field of dyslexic and normal reading children (11 to 13 years old) was studied by Corballis et al. (1985b) in order to compare inter-hemispheric differences (Bradshaw et al., 1976) between the groups. According to Orton, both hemispheres initially store information, and problems in reading and writing can be attributed to an inadequate cerebral dominance. Therefore, Corballis and his colleagues (1985b) hypothesized that inter-hemispheric effects should be diminished or absent in disabled readers. However, no group effect for latency or error rate was found. As expected, they found a clear effect of speed of naming normal over backward letters (RT to normal letters = 797 ms, RT to backward letters = 908 ms). However, the interaction between visual field and letter orientation failed to reach significance as did their interaction with the group factor. Corballis et al. (1985b) concluded that such a pattern of results does not support the assumption that reading-disabled children show an abnormally high degree of left–right equivalence, as proposed by Orton’s (1925) original theory. In a later study, Wolff and Melngailis (1996) found differences in accuracy between dyslexics and normal reading children for repeated naming of confusable letters. Extending the work of Corballis et al. (1985a, 1985b), Rusiak et al. (2007) investigated mental rotation of letters (Experiment 1) and of letters and shapes (Experiment 2) in normal readers and developmental dyslexics. Whereas the overall RTs were equal for shapes in both groups, for letters they were about 100 ms slower in dyslexics. For letters as well as for shapes, however, the same mental rotation effects were obtained between the groups. These results were interpreted as support for the symmetry generalization account, that is, the notion of developmental dyslexia as a deficit in functional coordination between graphemic and phonological letter representations. In contrast to the finding that dyslexics and controls did not differ in mental transformation skills, Rüsseler et al. (2005) found dyslexic readers impaired in mental rotation for letters as well as for shapes and pictures. Another relevant study is that of Hicks (1981), who studied the failure in integrating visual and phonological information in four different groups of readers; beginning readers, skilled readers, retarded readers (that is, disabled readers according to the discrepancy definition but without showing typical dyslexia errors, for example left/right confusion), and dyslexic readers (that is, disabled readers according to the discrepancy definition showing typical dyslexia errors, for example left/right confusion or neurological defects). All groups, except that of beginning readers, were matched by age and intelligence, and all groups, except that of normal readers, were matched by intelligence and literacy level. All participants had to perform a search task. A target had to be identified within a list of reversible (b, p, q, d) and non-reversible (g, h, k, l, t, y) letters. The target and the list were either presented visually or auditory. Consequently, there were four conditions, two testing
Experimental Approaches to Specific Disabilities in Learning to Read
325
the intra-modality functions and two the inter-modality functions. The main results were as follows: normal readers made fewer errors than all of the non-fluent reading groups in all conditions. When a visually presented target item had to be identified within a visually presented list, most errors were made by beginning readers. When both the target and the list were presented auditory, most errors were made by dyslexics and retarded readers. In the inter-modality conditions most errors were made by dyslexics, whereas beginning and retarded readers performed about equally. All together, the results indicate two major conclusions. First, normal reading children perform better than all of the non-fluent reading children. Second, the underlying cause of reading problems may differ between beginning readers and disabled readers and between disabled readers depending on the pattern of their reading performance. A more recent study is that of Patton et al. (2000). The authors investigated the rate of static and kinetic reversals in 201 children in a four-year longitudinal study from kindergarten to Grade 3. After three years, they found that there was no correlation between the two kinds of reversals, which may be interpreted as evidence against Orton’s (1925) theory of reversals. Furthermore, the rate of reversal errors in their study did not discriminate between children with and without problems in learning to read. Finally, reversals did not contribute to the prediction of the performance in a reading test. However, when the authors analysed the data of the fourth year of this longitudinal study (when children are in Grade 3) kinetic reversals proved to be an excellent predictor for the performance in a reading test. Unfortunately, no interpretation of this effect was given. Terepocki et al. (2002) used computer-based reversal detection tasks for numbers, letters, letter strings and words in order to compare 10-year-old normal readers and dyslexics in their performance accuracy. Dyslexics made more orientation errors than normal readers, whereas the two groups did not differ on attention control tasks. The authors suggested that the problems of dyslexics in discriminating similar looking items is due to poorly specified representations of letters. This interpretation is in some accordance with our view, as well as the assumption that such problems disrupt the process of learning to read and that, however, reversal errors are likely to disappear with reading experience because alternative strategies become habitual in dyslexics. An interesting extension of Terepocki et al.’s work was published by Badian (2005). She argued that only a subgroup of dyslexics have problems in orthographic memory for the orientation of letters and numbers. A study, often cited as evidence for the absence of a visual deficit in dyslexics was conducted by Bigsby (1985). She used a paradigm introduced by Posner and Mitchell (1967). These authors used a same–different task, in which participants were asked to decide whether two letters are the same or different. In one condition, a pair of letters had to be responded to as same only when the letters were physically identical (for example, B–B); in another condition, when they had the same name (‘value’) [a–A]; and in a third condition, only when they belonged to the same abstract category, such as vowels [a–e]. Posner and Mitchell (1967) found a temporal hierarchy of decision times depending on the instruction
326 Thomas Lachmann which they interpreted as evidence for a theory of serial processing stages. According to this theory, at the lowest level, patterns are encoded in terms of their elementary physical characteristics (‘... the visual pattern ‘A’ is coded as a set of lines forming a unified but unfamiliar figure, which is not different from an infinite number of line combinations of similar complexity that are not letters’; Posner and Mitchell, 1967). If a decision cannot be made at this level, memory information has to be used. In a terminology characteristic of the time, memory information was taken to consist of ‘templates’, which are searched serially (cf., Sternberg, 1966). If no match is found, comparison operations proceed to the next stage of ‘conceptual’ processing, at which isolated concepts are used to tag the perceived objects. If comparison fails in this stage, semantic coding is used, which evokes rules and abstract categories to make identity decisions. Following this logic, Bigsby (1985) argued that if dyslexics, as compared to normal reading children, show higher error rates and/or increased RTs for visual matches, then misperceptions would be present in dyslexics. In contrast, if the majority of errors and longer response latencies would occur for name matches, a linguistic dysfunction is to be assumed. Accordingly, in her experimental design, Bigsby differentiated between visual code pairs (lowercase letters, for example, ss versus rz) and a name code pairs (Dd versus Rs), whose letter elements were either reversible (bb) or not (Hs). For error rates, in case of the visual code pairs, no group effect and no effect of reversibility were found. Simple tests for the name code condition again showed no differences between disabled readers and normal reading children, but varieties between reversible and non-reversible letters were found (reversible >non-reversible). The results indicate that all of the conditions were easy for both groups. An analysis of reversibility confusion (the RT difference between reversible and corresponding non-reversible letter pairs) showed effects in all three groups for the name code pairs; dyslexic children showed more confusions than normal reading children; in particular, they had problems in differentiating lower and upper-case letters if they were reversible (for example P–q). In contrast, for the visual code pairs, no such group differences were found. Bigsby concluded from her findings that ‘Dyslexic children function well in the visual code and this code does not appear to be the locus of their reversible letter confusions’, and later she suggested that reversibility confusions seem ‘…to be mainly occurring somewhere during the translation from the visual code to either the name code or its referent, and to be more pernicious in dyslexic readers’. Bigsby (1985) introduced reversal errors as to be introduced by Orton (1925) as misperceptions based on a visual deficit. By doing so, she lumped together Orton’s model with models of a visual perceptual deficit in dyslexics (Stanley and Hall, 1973). By citing studies which revealed null-findings between dyslexics and control in visual object recognition (Ellis, 1981), she argued that visual models of dyslexia may be wrong. However, from our point of view, neither the results cited by her nor her own data are sufficient to reject the importance of reversal errors. In fact, the data could even be used in advance of the Symmetry Generalization model to explain reversals. Her interpretation is that
Experimental Approaches to Specific Disabilities in Learning to Read
327
reversals are an effect of a labelling deficit—this is exactly what symmetry generalization in letter perception would provoke. Even Orton’s theory does not contradict her at all. Altogether, we actually should expect a faulty interaction between the verbal and the visual code only in her verbal condition. Furthermore, a theoretical shortcoming may also be seen in her understanding of Posner and Mitchel’s (1967) model, whose approach she adopted for her study. The idea of distinct codes (physical and the name code) acting in serial stages of information processing, was rejected by Posner himself (1978; the serial stage approach ‘…is simply too restrictive to use as a complete description of the processes involved’, and ‘The temporal hierarchy ... does not imply that the processes involved at the different levels represent a strict series’, p. 35). With the beginning of the 1970s, serial models of information processing were found to be restricted to only some information processes. As a possible alternative, Posner considered a redundancy model in which all comparisons are made on the basis of a single memory code. This approach entails that comparison processes are facilitated by shared features between the items at different levels of abstraction. A second suggested alternative, the pathway activation model, states that each comparison reaches the highest possible level, but reaches this level the sooner the more lower-level features are being shared.
SYMMETRY GENERALIZATION IN LETTER AND SHAPE RECOGNITION IN DEVELOPMENTAL DYSLEXIA One of our own experiments (Brendler and Lachmann, 2001) was conducted in order to test specific hypotheses that arise from the model of Symmetry Generalization, that is, the assumption that those reading disabled children who show more reversal errors in reading, as measured by a reading test (Zuericher Lesetest, ZLT, Grissemann, 2000), have problems in coordinating the visual and phonological subfunction due to an abnormal tendency of symmetry generalization in the representation of visual symbols, such as letters. A same–different task was applied to test this hypothesis. Children with and without reading problems were asked to compare two stimuli, either two letters or two patterns, and to answer as fast and accurate as possible whether they are the same of different. Since symmetry generalization is assumed to operate in memory, the stimuli were presented successively, that is, one item in memory had to be compared with a visible one. The orientation of the stimuli was varied systematically, and thus in some pairs items were reflected or rotated version of each other. The task was varied between blocks. In the categorical condition, the orientation (rotation or reflection) had to be ignored, the children had to press same if letters or patterns had the same shape but of different orientation. In the physical condition, the children had to press different in these cases. The difference between these conditions is crucial because in the physical condition symmetrical relations between items must be suppressed in order to respond correctly.
328 Thomas Lachmann Experiment The authors tested 66 children from Grades 3 and 4, half of which were diagnosed as dyslexics according to the discrepancy definition, and the others served as Controls. The authors used lexical versus non-lexical material (see Figure 23.1), and a physical versus a categorical instruction in a blocked design. The lexical material consisted of the letters ‘b, e, f, h, n, r’. As non-lexical material five-dot patterns were used (constructed on a 3×3 grid, leaving no row or column empty (Garner and Clement, 1963). Pairs of letters and pairs of patterns were shown successively. The items were presented in normal and in mirrored orientation. Therefore, a pair consisted of two items which were either identical in shape and orientation, identical in shape but not in orientation, or different in shape (for methodological details, see Lachmann and Geissler, 2002; Lachmann and van Leeuwen, 2004b, 2005a, 2005b). FIGURE 23.1 (2001)
Letter and shape stimuli used in different blocks in the experiment by Brendler and Lachmann
In the physical condition the children had to respond to two items as to same only when they were physically identical, that is, the same in shape and orientation. Items of different orientation had to be judged as different—just as those items which are different in shape. In contrast, in the categorical condition the children were instructed to ignore the orientation of the stimuli. Items same in shape but different in orientation had to be responded as to same (Lachmann and van Leeuwen, 2005b). It is important to note that this same–different experiment differs from that of Bigsby (see above) in a rigorous way; we do not test any name categories (same responses are always required to the same stimulus) and in half of our pairs the stimuli are transformationally related—not only vertically and independent of phonological similarities.
Main Results and Discussion The main result was that disabled readers made more errors in discriminating stimuli under the physical condition relative to controls. This effect was strongest when letters
Experimental Approaches to Specific Disabilities in Learning to Read
329
had to be compared. In other words, children with problems in learning to read have special problems to give a different response when the same two letters were shown in different orientation. This was interpreted as evidence for symmetry generalization in the representation of visual symbols. The results of the reading test (ZLT) exhibit a general overweight of the total number of reading errors within the disabled readers in contrast to normal readers. The children with reading disability had the greatest deficit in reading letters that were connected via left–right or vertical symmetry (for example reading ‘b’ instead of ‘d’). However, most important with respect to the Functional Coordination Deficit account is the fact that the authors found a significant correlation between the errors observed in the experiment and the reversal errors in the reading test. Those children who had more problems in discriminating orientationally related letters or patterns showed more reversal errors in text reading (see Figure 23.2). This result, however, was restricted to the b–d reversals. For other reversals the number of errors was too small, and an analysis could not be conducted. This is not less interesting for the symmetry generalization account; note that the errors in the experiment were made to all orientational axis. FIGURE 23.2 Relationship between ‘b–d’ reversals in word reading and errors in the same–different task with physical instruction and lexical material
Source: Brendler and Lachmann, 2001.
The results were counted as evidence that difficulties in learning to read are related to difficulties in suppressing mirror-generalization in the representation of visual symbols, and thus to difficulties in the coordination of reading related subfunctions. Such difficulties are typical for beginning readers. In this respect, however, the questions remain whether reading disabled children have generally a greater tendency of symmetry generalization
330 Thomas Lachmann and therefore have more problems to suppress this mechanism in reading, or whether the degree of symmetry generalization is equal to that of normal readers, but there is a problem in learning to suppress this mechanism when confronted with visual symbols.
Mental Rotation of Letters in Developmental Dyslexia Mental rotation is a task which requires visuo-spatial processing of images in mind. Usually, people are asked to decide whether two items, presented either simultaneously or in succession, are the same or different regardless of their orientation. The first study on mental rotation was published by Shepard and Metzler (1971). They showed that the RT increased linearly with the three-dimensionally angular difference of the objects presented, suggesting that the shown image is rotated in the individuals’ minds just as they would physically manipulate the object. The mental rotation effect has been replicated in many studies irrespective of the type of stimuli used in these experiments and even for other tasks than a same–different task (Cohen et al., 1996; Harris et al., 2000; Jordan et al., 2001; Kosslyn et al., 1998). Cooper and Shepard (1975), for instance, displayed isolated letters of different orientation, and the task was to decide whether the letter was shown in normal or mirrored parity. Ruthruff et al. (1995) used a mental rotation task within a dual-task setting and concluded that this operation requires central capacity.
Mental Rotation Studied in Dyslexics Mental rotation studied in dyslexics was the subject of some experimental studies (Corballis et al., 1985b; Karadi et al., 2001; Rusiak et al., 2003; Rusiak et al., 2007), some of which have been described earlier in this chapter. In the second study, Karadi et al. (2001) used pictures of right and left hands as stimuli. They were presented at angle of 0°, 50°, 90° and 180° clockwise from the normal upright. The participants’ task was to decide whether a left or a right hand was shown on the screen and to press the corresponding key. The results indicated that dyslexics perform this task with a higher speed but with lower accuracy. Moreover mental rotation occurred in non-dyslexics, but not in dyslexic children. The authors interpreted the results in terms of a dysfunction of the parietal cortex. They concluded that dyslexics have general problems in the processing of spatial information (and thus in mental rotation tasks), which may result in a variety of reading and writing difficulties. From our point of view, a deficit in spatial information processing may be one possible problem in dyslexics, but it is unlikely to produce reversal errors. In one of our own studies (Rusiak et al., 2003) we asked 15 dyslexic and 14 non-dyslexic teenagers to take part in a classical mental rotation experiment. The participants’ Intelligent Quotients (IQs) were estimated with the Wechsler Adult Intelligence Scales. Moreover, a
Experimental Approaches to Specific Disabilities in Learning to Read
331
reading test using chain-words and chain-sentences (Ober, 1998) was performed. The task and the main results are described in the following paragraphs in some more detail.
The Experiment The stimuli were black letters: R, F, G, e, k and the exact mirror-images of those letters. They were presented against a white background in the centre of a computer monitor. The letters were displayed in different angular orientations: 0º, 45°, 90°, 135°, 180°, 225°, 270°, and 315º clockwise from the normal upright. The participants’ heads were fixated. The participants were instructed to look at the fixation cross and to decide whether the subsequent stimulus was a normal or mirrored letter regardless of its orientation by pressing corresponding keys. All participants were asked to perform the task as quickly and as accurately as possible.
Main Results and Discussion To statistically evaluate the results of mental rotation an analysis of variance (ANOVA) was performed for RTs and error rate with one between-subject (group) and one within-subject factor (rotation angle). For RT, a significant group main effect was found; dyslexics were slower than normal readers, and for both groups, RT and error rate increased significantly with increasing angle of rotation from the upright orientation. This finding reflects a typical mental rotation effect (Cooper and Shepard, 1975). Dyslexics needed 300 ms more to decide whether the letter presented was normal or mirrored. However, as can be seen in Figure 23.3, the slope is the same in both groups, that is, the effect is additive with rotation angle between the groups. The null finding of an interaction between the factors confirmed the interpretation that dyslexics show no difference in the operation of mental rotation. Instead, the reason for the dramatic increase in RT is either before or after the mental operation under investigation. The analysis of accuracy rate revealed no difference between the groups (see Figure 23.4). According to the Symmetry Generalization model, which assumes that orientation errors in dyslexics occur from a failure in suppressing symmetry generalization in the representation of graphemes, the difference in RT between the groups may be caused by a decision problem, that is, after the mental rotation process has been completed. In simple words, after the dyslexics have finished the rotation process, they still have a problem to decide whether the letter (the grapheme) is in normal orientation or mirrored because to decide this requires a mapping between the input and the representation of the grapheme in memory. This problem produces higher decision times and more errors which, of course, are then independent from the degree of rotation.
332 Thomas Lachmann FIGURE 23.3
Mean RT (ms) as a function of angle of rotation for dyslexics and controls
Source: Rusiak, Lachmann and Jaskowski, 2003.
The present results are somehow in opposite to what was obtained by Karadi et al. (2001). This, however, is expected by the Symmetry Generalization model, because they used non-linguistic stimuli. Also the null-finding of group differences in the study of Corballis et al. (1985a) is actually in good accordance with our model. Dyslexics who make a lot of reversal errors and fail in the present task may not have a problem in naming the letter but in differentiating the normal from the mirrored because both versions represent the grapheme; whereas in non-dyslexics, only the normal letter is connected with the phoneme representation (but not the mirrored). Of course, this interpretation is not exclusive. It could also be argued that this particular group of dyslexics tested here has a problem in encoding the stimuli or in other processing stages prior mental rotation. Therefore, a number of other studies are planed to test this interpretation directly and indirectly. A direct way could be to use electrophysiological methods. Heil and Rolke (2002) introduced a method in which Event Related Potentials (ERPs) can be used to localize mental rotation in space and time. If the deficit is indeed located after mental rotation, there should be no difference in this ERP component
Experimental Approaches to Specific Disabilities in Learning to Read
333
between the groups. However, if the relevant peak will appear in dyslexics later than in non-dyslexics it could be concluded that the deficit is located before the process of mental rotation. FIGURE 23.4 group
Accuracy rate (%) as a function of angle of rotation in the dyslexic and normal reader
Source: Rusiak, Lachmann and Jaskowski, 2003.
An indirect way has already been delivered with the results of Brendler and Lachmann (2001), described before and of a more recent study (Lachmann and van Leeuwen, 2007). In these studies, the authors showed that dyslexics have indeed problems in suppressing symmetry in the representation of graphemes, indicated by increased response latencies and error rates for the physical and categorical match under the physical instruction, but not under the categorical instruction, as well as by the fact that the difficulty is especially prominent for letters. Since for both instructions the same stimuli and procedure were used, but a group effect was found for the physical instructions only, it was concluded that the problem of dyslexics in the mental rotation task may not be caused by an early processing deficit. Instead the interpretation of a decision deficit was supported.
334 Thomas Lachmann GENERAL DISCUSSION On the basis of the studies reviewed, the question about the significance of reversal errors for explaining the nature of reading disability cannot clearly be answered because, at the first glance, the results seem to be very heterogeneous. However, spoken in simple words, the reason for the heterogeneity of the results is the heterogeneity of the studies, that is, different methodological approaches were applied and therefore the results are not comparable. We reviewed a number of reversal studies. Some of these studies investigate the relation between reading and the processing of reversible visual shapes (Goins, 1958; Lyle and Goyen, 1968). Using non-verbal material, however, they may have tested hypotheses of certain deficits in the visual subfunction as to be responsible for the problems in learning to read. In itself, this is absolutely legitimate. However, such studies can neither be used to test the theory of Orton (1925) nor any other theory that assumes reversals to result from a failure in coordinating visual and phonological subfunctions. In fact, from some reversal models (Corballis and Beale, 1993; Lachmann, 2002; Orton, 1925), a null-finding in these studies would even be expected. Yet, studies using verbal material may also differ significantly, as for in the instance of the degree of functional fragmentation (Lachmann, 2002). Whereas Orton (1925) analysed the writing performance of disabled readers and a reading test measures the reading performance per se, experimental studies test the performance of normal versus disabled readers on an experimental task which requires only certain subfunction(s) of the reading process (functional fragmentation). In some experiments, reversible and not reversible letters are presented in the context of words (Lyle, 1969; Seidenberg et al., 1985), and in others in the context of non-words (Seidenberg et al., 1985), which may affect subgroups of disabled readers quite differently (for example, Boder, 1973a, 1973b). The involved cognitive functions in both conditions, however, may differ from those functions involved when isolated letters are presented (Brendler and Lachmann, 2001; Corballis et al., 1985a; Liberman et al., 1971). Experimental studies on reversal errors do not only differ in the used material, but also in the procedure. For instance, presenting reversible shapes or letters very briefly, followed by a mask (Grosser and Trzeciak, 1981) to measure the recognition threshold, may test the pre-representational processing (albeit not exclusively), whereas presenting one of the items for a sufficient time to create a representation may test a different kind of processing. The required response must be considered as well. The response can be speeded or not and may require naming, reproducing, recalling or simply recognizing; the response may require the manipulation of the representation (as in our mental rotation experiment), or not, and the decision may be based on the same, or a higher (categorical) level of processing (for example, Bigsby, 1985; cf., Posner and Mitchell, 1967). Furthermore, the modality of the input and the response may differ, verbal material may be presented
Experimental Approaches to Specific Disabilities in Learning to Read
335
visually or auditively, and the response may be visual or auditive as well (Fuchs and Lachmann, 2003; Hicks, 1981). Thus, the material and the procedure should be chosen carefully depending on the hypothesis. Unfortunately, some review articles and introduction sections put all reversal studies together, and conclude undifferentially that the majority of them show that reversals do not play an exceeding role in the Error rate of reading disabled children. Moreover, some articles characterize reversal studies as testing the visual deficit theory, and conclude on the basis of some null-findings that visual problems are generally unlikely to be present in disabled readers (for example, Bigsby, 1985; Vellutino and Scanlon, 1998). A further problem in comparing studies on reversal errors is the definition of reversals. Some experimenters only consider static reversals (for example, Brendler and Lachmann, 2001), while others are especially interested in testing the relation between static and kinetic reversals (Liberman et al., 1971; Patton et al., 2000). But even within the two kinds of reversals, there is no consistency about what is to be counted as a reversal. Lyle and Goyen (1968), for instance, assume their sequential condition as measuring mechanisms equivalent to those responsible for kinetic reversals in writing and reading. Grosser and Trzeciak (1981), for instance, define ‘u’ as a non-reversible letter; but since they used ‘b’ and ‘q’ as reversible letters, they should have considered ‘u’ as reversible to ‘n’ (even though ‘n’ was not presented). In Liberman et al. (1971) the letter ‘g’ counted as reversible, which is questionable. In the previous section, we argued that experiments on reading have to test certain component sub-functions of reading. Thus, in studies on reversal errors the experimenter aims on testing functions which are assumed to produce reversal confusion. When the performance in the used task proved to be significantly different in students with reading disability and normal reading students, it is concluded that the tested function is responsible for reversal confusion and problems in learning to read. Consequently, not only does the selection of sub-functions have to be tested, but the definition of the reading disability group is also crucial. We already introduced the discrepancy definition of reading disability and the problems related to that definition. We mentioned that there is a critical discussion going on about this definition. Consequently, not all experimenters use it (for example, Liberman et al., 1971), and those who do are not uniform about the criterions for a discrepancy. Hicks (1981), for instance, differentiated between retarded readers and dyslexics as two different groups. The retarded readers are defined as having a reading retardation of 2.1 years from chronological age. Most studies, however, would define this group as dyslexic. The dyslexic group in Hicks’ study was defined as having 1.5 years retardation in reading and showing typical patterns of errors in reading and writing. Of course, this may influence the experimental results and their interpretation. There is also inconsistency about the terms used to describe the samples. The terms poor readers, dyslexics, disabled readers, retarded readers, strephosymbolics, and so on, are
336 Thomas Lachmann either used synonymously or to distinguish between different groups. As a consequence, the same term may be used in different studies; the definition, however, may differ. Lyle and Goyen (1968), for instance, used the term reading retarded children, just as Hicks (1981) did, but in contrast to her, Lyle and Goyen defined retarded readers as showing a reading retardation of nine months in Grade 2 and 18 months in Grade 3. In Corballis et al. (1985a, 1985b), as a further example, the term disabled readers was used and two years retardation in reading was required to meet the criterion. The discrepancy criterion depends not only on reading performance, but also on measures of the general cognitive abilities. Whereas, for instance, Lyle and Goyen define an IQ value of at least 90 to meet the criterion, most authors just report the average of the samples, and in Grosser and Trzeciak we learn nothing about the IQ of the students at all. Not only the discrepancy definition, but the age of the children the sample consists of may also differ between studies. Whereas the children in Corballis et al. (1985b) were 12 years old, Patton et al. (2000), and Lyle and Goyen (1968) asked younger children (Grades 1–3) to participate in their experiments. The participants in the experiments of Grosser and Trzeciak (1981) were aged between 7 and 14 years, and the authors revealed a strong correlation between age and performance, but only in normal reading students. From this result, we may expect greater differences in samples of higher average age, but the results in Lyle and Goyen suggest bigger differences at younger age level. In any case, the age is a crucial factor; as in all developmental disorders, the older the person, the less sure we can be that what we have measured reflects the primary or a secondary lag. A problem arises when reading disabled children receive a special training, as for instance those in Brendler and Lachmann (2001). In some German Federal States a diagnostic procedure takes place in Grade 2, and those diagnosed as dyslexics attend the Grade 3 level over two years instead of one in order to have the chance for an extensive training in reading and writing. This results in methodological problems for researchers: testing the reading disabled children means that the controls are either younger or have a higher grade level, which may be important for the interpretation of reading tests. Moreover, children who already attended the two-year dyslexia training programme may show more similar performance than Controls in reading, including reversal errors, while their underlying deficit may still exist.
SUBGROUPS Last but not least, we shall emphasize the importance of subgroups of dyslexics (Becker et al., 2005; Boder, 1970, 1973a; Flynn and Deering, 1989). The FCD model is a multi-causal one. Developmental dyslexia is not a homogeneous syndrome. Children differ in their reading problems (Boder, 1973b), and it was shown that different deficits are responsible for these problems. For instance, Lachmann et al. (2005) studied auditory processing
Experimental Approaches to Specific Disabilities in Learning to Read
337
in 8–11-year-old children with developmental dyslexia by means of event-related brain potentials. Cortical sound discrimination was evaluated by analysing the mismatch negativity (Näätänen, 1992; Schröger, 1997) to syllable and tone changes. The analysis of the data obtained from two dyslexic subgroups, Dyslexics-1 being impaired in non-word reading (or both non-word and frequent word reading) and Dyslexics-2 in frequent word reading but not in non-word reading, revealed that the mismatch negativity was specifically diminished in the latter group, whereas it was normal-like in Dyslexics-1. These results showed that different diagnostic subgroups of dyslexics have different patterns of auditory processing deficits as suggested by sound-discrimination impairment specific to one of the groups. We may assume that reversal errors are a symptom of Dyslexics-1, those with problems in grapheme-to-phoneme conversion (cf. Lachmann et al., 2005).
REFERENCES Badian, N.A. 2005. ‘Does a Visual-Orthographic Deficit Contribute to Reading Disability?’ Annals of Dyslexia, 55: 28–52. Becker, C., M. Elliott and T. Lachmann. 2005. ‘Evidence for Impaired Visuo-Perceptual Organization in Developmental Dyslexia and Its Relation to Temporal Processes’. Cognitive Neuropsychology, 22: 499–522. Bigsby, P. 1985. ‘The Nature of Reversible Letter Confusion in Dyslexic and Normal Readers: Misperception or Mislabelling?’. British Journal of Educational Psychology, 55: 264–72. Boder, E. 1970. ‘Developmental Dyslexia: A New Diagnostic Approach Based on The Identification of Three Subtypes’. Journal of School Health, 40: 289–90. ———. 1973a. ‘Developmental Dyslexia: Prevailing Diagnostic Concepts and a New Diagnostic Approach’. Bulletin of the Orton Society, 23: 106–18. ———. 1973b. ‘Developmental Dyslexia: A Diagnostic Approach Based on Three Atypical Reading-Spelling Patterns’. Developmental Medicine and Child Neurology, 15: 663–87. Bradley, L. and P.E. Bryant. 1978. ‘Difficulties in Auditory Organization as A Possible Cause of Reading Backwardness’. Nature, 271: 746–47. Bradshaw, J., D. Bradley and K. Patterson. 1976. ‘The Perception and Identification or Mirror-Reversed Patterns’. Quarterly Journal of Experimental Psychology, 28: 221–46. Brendler, K., and T. Lachmann. 2001. ‘Letter Reversals in the Context of the Functional Coordination Deficit Model of Dyslexia’. In E. Sommerfeld, R. Kompass, and T. Lachmann (eds), Proceedings of the International Society for Psychophysics (pp. 308–13). Lengerich, Berlin: Pabst. Cohen, M.S., S.M. Kosslyn, H.C. Breiter, G.J. DiGirolamo, W.L. Thompson, A.K. Anderson, S.Y. Brookheimer, B.R. Rosen and J.W. Belliveau. 1996. ‘Changes in Cortical Activity During Mental Rotation: A Mapping Study Using Functional MRI’. Brain, 119: 89–100. Cooper, L.A. and R.N. Shepard. 1975. ‘Mental Transformation in the Identification of Left and Right Hands’. Journal of Experimental Psychology: Human Perception & Performance, 1: 48–56. Corballis, M.C. and I.L. Beale. 1993. ‘Orton Revisited: Dyslexia, Laterality, and Left-Right Confusion’. In D. Willows, R.S. Kruk and E. Corcos (eds), Visual Processes in Reading and Reading Disabilities (pp. 57–73). Hillsdale, NJ: Erlbaum.
338 Thomas Lachmann Corballis, M.C., L. Macadie and I.L. Beale. 1985a. ‘Mental Rotation and Visual Laterality in Normal and Reading Disabled Children’. Cortex, 21: 225–36. Corballis, M.C., L. Macadie, A. Crotty and I.L. Beale. 1985b. ‘The Naming of Disoriented Letters by Normal and Reading-Disabled Children’. Journal of Child Psychology and Psychiatry and Allied Disciplines, 26: 929–38. Ellis, E. 1981. ‘Visual and Name Coding in Dyslexic Children’. Psychological Research, 43: 201–18. Fisher, F.W., I.Y. Liberman and D. Shankweiler. 1978. ‘Reading Reversals and Developmental Dyslexia: A Further Study’. Cortex, 14: 496–510. Flynn, J.M and W.M. Deering. 1989. ‘Subtypes of Dyslexia: Investigation of Boder’s System Using Quantitative Neurophysiology’. Developmental Medicine and Child Neurology, 31: 215–23. Fuchs, S. and T. Lachmann. 2003. ‘Functional Coordination Deficit in Children with Developmental Dyslexia’. In B. Berglund and E. Borg (eds), Proceedings of the International Society for Psychophysics (pp. 87–90). Stockholm: The International Society for Psychophysics. Garner, W.R. and D.E. Clement. 1963. ‘Goodness of Pattern and Pattern Uncertainty’. Journal of Verbal Learning and Verbal Behavior, 2: 446–52. Goins, J.T. 1958. Visual Perceptual Abilities and Early Reading Process. Chicago: University of Chicago Press. Grissemann, H. 2000. Zuercher Lesetest (ZLT) (6th edition). Goettingen: Hogrefe. Grosser, G.S. and G.M. Trzeciak. 1981. ‘Durations of Recognition for Single Letters, With and Without Visual Masking, by Dyslexics and Normal Readers’. Perceptual and Motor Skills, 53: 991–95. Harris, I.M., G.F. Egan, C. Sonkkila, H.J. Tochon-Danguy, G. Paxinos and J.D. Watson. 2000. ‘Selective Right Parietal Lobe Activation During Mental Rotation: A Parametric PET Study’. Brain, 123: 65–73. Heil, M. and B. Rolke. 2002. ‘Towards a Chronopsychophysiology of Mental Rotation’. Psychophysiology, 39: 414–22. Hicks, C. 1981. ‘Reversal Errors in Reading and Their Relationship to Inter- and Intra-modality Functioning’. Educational Psychology, 1: 67–79. Hildreth, G. 1934. ‘Reversals in Reading and Writing’. Journal of Educational Psychology, 25: 1–20. Jordan, K., H.J. Heinze, K. Lutz, M. Kanowski and L. Jancke. 2001. ‘Cortical Activations During the Mental Rotation of Different Visual Objects’. Neuroimage, 13: 143–52. Karadi, K., B. Kovacs, T. Szepesi, I. Szabo and J. Kallai. 2001. ‘Egocentric Mental Rotation in Hungarian Dyslexic Children’. Dyslexia, 7(1): 3–11. Kosslyn, S.M., G.J. DiGirolamo, W.L. Thompson and N.M. Alpert. 1998. ‘Mental Rotation of Objects Versus Hands: Neural Mechanisms Revealed by Positron Emission Tomography’. Psychophysiology, 35: 151–61. Lachmann, T. 2002. ‘Reading Disability as a Deficit in Functional Coordination and Information Integration’. In E. Witruk A.D. Friederici and T. Lachmann (eds), Basic Functions of Language, Reading, and Reading Disability. Neuropsychology and Cognition series (pp. 165–98). Boston: Kluwer/Springer. Lachmann, T., S. Berti, T. Kujala and E. Schröger. 2005. ‘Diagnostic Subgroups of Developmental Dyslexia have Different Deficits in Neural Processing of Tones and Phonemes’. International Journal of Psychophysiology, 56: 105–20. Lachmann, T. and H.G. Geissler. 2002. ‘Memory Search Instead of Template Matching? RepresentationGuided Inference in Same–Different Performance’. Acta Psychologica, 111: 19–43. Lachmann, T. and T. Geyer. 2003. ‘Letter Reversals in Developmental Dyslexia: Is the Case Really Closed? A Critical Review and Conclusions’. Psychology Science, 45: 53–75.
Experimental Approaches to Specific Disabilities in Learning to Read
339
Lachmann, T. and C. van Leeuwen. 2004a. ‘Negative Congruence Effects in Letter and Pseudo-letter Recognition: The Role of Similarity and Response Conflict’. Cognitive Processing, 5: 239–48. ———. 2004b. ‘Memory-Guided Inference in Same–Different Comparison Tasks’. In C. Kaernbach, E. Schroeger, and H.J. Mueller (eds), Psychophysics Beyond Sensation. Scientific Psychology Series (pp. 199–225). Hillsdale, NJ: Erlbaum. ———. 2005a. ‘Individual Pattern Representations are Context-Independent, But Their Collective Representation is Context-Dependent’. Quarterly Journal of Experimental Psychology: Human Experimental Psychology, 58: 1265–94. ———. 2005a. ‘Task-Invariant Aspects of Goodness in Perceptual Representation’. Quarterly Journal of Experimental Psychology: Human Experimental Psychology, 58: 1295–310. ———. 2007. ‘Paradoxical Enhancement of Letter Recognition in Developmental Dyslexia’. Developmental Neuropsychology, 31: 61–77. Liberman, I.Y., D. Shankweiler, C. Orlando, K.S. Harris and F. Bell Berti. 1971. ‘Letter Confusions and Reversals of Sequence in the Beginning Reader: Implications for Orton’s Theory of Developmental Dyslexia’. Cortex, 7: 127–42. Lyle, J. 1969. ‘Reading Retardation and Reversion Tendency: A Factorial Study’. Child Development, 40: 833–43. Lyle, J.G. and J. Goyen. 1968. ‘Visual Recognition, Developmental Lag, and Strephosymbolia in Reading Retardation’. Journal of Abnormal Psychology, 73: 25–9. Miles, T.R. and E. Miles. 1999. Dyslexia: A Hundred Years On. Buckingham: Open University Press. Näätänen, R. 1992. Attention and Brain Function. Hillsdale, New Jersey: Erlbaum. Ober, J. 1998. ‘Presentation of the New Method of Testing Development–Reading Function–Word Test and Chain Sentences Test’. Logopedia, 25: 81–99. Orton, S.T. 1925. ‘“Word-blindness” in School Children’. Archives of Neurology and Psychiatry, 14(5): 581–615. ———. 1928. ‘Specific Reading Disability—Strephosymbolia’. Journal of the American Medical Association, 90: 1095–99. Patton, J.E., D.Y. Yarbrough and D. Thursby. 2000. ‘Another Look at Children’s Symbol Reversals’. Perceptual and Motor Skills, 90: 577–78. Posner, M.I. 1978. Chronometric Explorations of Mind. Hillsdale, New Jersey: Erlbaum. Posner, M.I. and R. F. Mitchell. 1967. ‘Chronometric Analysis of Classification’. Psychological Review, 74: 392–409. Rusiak, P., T. Lachmann and P. Jaskowski. 2003. ‘Mental Rotation of Letters in Dyslexics and Implications for Diagnosis and Educational Treatment’. In B. Berglund and E. Borg (eds), Proceedings of the International Society for Psychophysics (pp. 263–68). Stockholm: The International Society for Psychophysics. ———. 2007. ‘Mental Rotation of Letters and Shapes in Developmental Dyslexia’. Perception, 36: 617–31. Rüsseler, J., J. Scholz, K. Jordan and C. Quaiser-Pohl. 2005. ‘Mental Rotation of Letters, Pictures, and ThreeDimensional Objects in German Dyslexic Children’. Journal of Child Neuropsychology, 11: 497–512. Ruthruff, E., J.O. Miller and T. Lachmann. 1995. ‘Does Mental Rotation Require Central Mechanisms’. Journal of Experimental Psychology: Human Perception and Performance, 21: 552–70. Schröger, E. 1997. ‘On the Detection of Auditory Deviations: A Pre-Attentive Activation Model’. Psychophysiology, 34: 245–57. Seidenberg, M., M. Bruck, G. Fornarolo and J. Backman. 1985. ‘Word Recognition Processes of Poor and Disabled Readers: Do They Necessarily Differ?’. Applied Psycholinguistics, 6: 161–80.
340 Thomas Lachmann Shepard, R.N. and J. Metzler. 1971. ‘Mental Rotation of Three-Dimensional Objects’. Science, 171: 701–03. Stanley, G. and R. Hall. 1973. ‘Short-Term Information Processing in Dyslexics’. Child Development, 44: 841–44. Sternberg, S. 1966. ‘High-Speed Scanning in Human Memory’. Science, 153: 652–54. Terepocki, M., R.S. Kruk and D.M. Willows. 2002. ‘The Incidence and Nature of Letter Orientation Errors in Reading Disability’. Journal of Learning Disabilities, 35: 214–33. van Leeuwen, C. and T. Lachmann. 2004. ‘Negative and Positive Congruence Effects in Letters and Shapes’. Perception and Psychophysics, 6: 908–25. Vellutino, F.R. 1977. ‘Alternative Conceptualizations of Dyslexia: Evidence in Support of a Verbal-Deficit Hypotheses’. Harvard Educational Review, 47: 334–54. ———. 1987. ‘Dyslexia’. Scientific American, 256: 20–27. Vellutino, F.R. and D.M. Scanlon. 1998. ‘Research in the Study of Reading Disability: What Have We Learned in the Past Four Decades?’. Paper presented at the Annual Conference of The American Educational Research Association, San Diego. Wolff, P.H. and J. Melngailis. 1996. ‘Reversing Letters and Reading Transformed Text in Dyslexia: A Reassignment’. Reading and Writing: An Interdisciplinary Journal, 8: 341–55.
Chapter 24 Cognitive Profiles of Children with Dyslexia Bhoomika R. Kar and Nishi Tripathi
INTRODUCTION
D
yslexia is supposed to be a failure in learning to optimize the coordination of sub-processes involved in reading with the consequence of errors in integrating reading related information represented in working memory (WM) (Lachman, 2002). Dyslexia is the most common learning disorder observed in 10 per cent of the school going population (Suresh and Sebastian, 2003). It interferes with the child’s ability to acquire speech reading despite average intellectual functions. In the Indian context, research on dyslexia has received attention only during the last one decade. In the context of the complexities associated with defining and identifying children with learning disabilities and those with dyslexia in particular, it has been more difficult in India due to socio-cultural diversities (Karanth, 2003). Reading is one of the most complex cognitive processes for humans, involving visual and semantic decoding, temporal and phonological processing, orthographic, syntactic and contextual analysis and comprehension. An inefficient synchrony of these underlying mechanisms results in reading disability. Cognitive processes such as attention, motor control, WM, word recognition and visual integration also have been a topic of research in dyslexia. Reading is not a single process rather a coordination of many processes. There is a reciprocal influence of the development of cognitive processes and early reading acquisition on reading skill and reading comprehension (Share and Stanovich, 1995). Reading is not automatic, but must be learned (Shaywitz, 1996). Word identification and comprehension are two separate processes. In reading at first the visual input serves as a key to trigger a bottom up process. The word is first decoded into its phonological form and identified. Once it is identified, higher level cognitive functions such as intelligence and vocabulary are applied to understand the word’s meaning. In children with dyslexia, a phonological deficit impairs decoding thus preventing the reader from using his/her intelligence and vocabulary to get the word’s meaning (Shaywitz, 1996).
342 Bhoomika R. Kar and Nishi Tripathi This generates hypotheses in terms of interacting bottom-up–top-down information integration processes (Masaro and Cohen, 1991). Word reading is related to this dominant interaction between the simultaneous (top-down) and successive (bottom-up) processes. Simultaneous processing is dominant in visual coding, whereas successive processing plays a major role in phonological coding (Das et al., 1994b). Visual information can trigger hypotheses on the semantic level (bottom-up), and activated semantic information can influence the speed and quality of the creation of a visual representation (top-down). In reading, the bottom-up information from visual analyses plays an important role in the decoding process. In this context, the present study examined the underlying deficient cognitive processes resulting in reading disability. These cognitive processes have to be understood within a theoretical context, which in this case is a neurocognitive framework. This work is a part of an ongoing project on learning disability, which started with the assessment of teachers’ perception of learning related problems in school children in Allahabad to provide an initial screening. Children once perceived with learning related problems by the teachers were formally assessed to identify children with Specific Learning Disability (SLD). The need for a cognitive assessment was realized while looking at the complex nature of reading errors committed by the children identified as dyslexics. The cognitive processes, such as planning, attention, simultaneous and successive processing, were examined in children identified as dyslexics. It was assumed that an assessment of SLD using reading writing tests and that of cognitive processes would help narrow down the search for the basic underlying processes involved in dyslexia in order to carefully plan the remediation programme for these children.
METHODS The present study is a part of an ongoing project on teachers’ perception of learning-related problems in school children in Allahabad in the context of lack of established services for this group of children. We started with four different medium level schools in Allahabad and examined the teachers’ perception of learning related problems in reading, writing and mathematics in children from third to seventh standards using a problem checklist. Teachers perceived language related problems concerned with phonetics, syntactical errors, reversals, reading speed, writing errors, poor concentration, and behaviour problems as the most prevalent problems in school children, but had misconceptions regarding learning disability, equating it to underachievement and low IQ (Tripathi and Kar, in press). Hence, the second phase aimed at validating teachers’ identification of children with learning-related problems to identify children with specific learning disabilities and also to examine the cognitive profiles of these children. The cognitive profiles of children with dyslexia would aid in understanding the underlying cognitive processes affected, thus giving an insight into the remedial strategies which would work on the underlying cognitive processes first and then on the specific reading skills.
Cognitive Profiles of Children with Dyslexia
343
Objectives • To identify children with dyslexia from the third to the seventh standards using the NIMHANS index of SLD. • To examine the cognitive processes underlying dyslexia.
Participants
43 children from 3rd to 7th std taken for assessment on the NIMAHNS index of SLD to identify children with reading, writing and arithmetical disability
14 children left school
29 children assessed on NIMHANS index
18 children from 3rd to 7th std with dyslexia assessed on CAS to evaluate their cognitive profiles
Forty-three children from the third to the seventh standards identified with learning related problems by the teachers were taken for assessment on the NIMAHNS index of SLD to identify children with reading, writing and arithmetical disability. Out of these 43 children, 14 had left school at the time of assessment. Hence, 29 children were assessed on the NIMHANS index of SLD. Out of these 29, 18 children (8–13 years of age; mean age: 10.7; SD: 1.67) were identified as dyslexics. None of these children had arithmetical disability. These children were assessed on cognitive assessment system (CAS) to evaluate their cognitive profiles. Following tools were used for the study.
Tools Coloured Progressive Matrices Coloured Progressive Matrices (Raven, 2000) were used to assess the intellectual functions of children with dyslexia in the age range of 7–11 years and Standard Progressive Matrices (Raven, 2000) for the children above the age of 11 years.
344 Bhoomika R. Kar and Nishi Tripathi NIMHANS Index of Specific Learning Disability This index (Hirisave, Oommen and Kapur, 2002) (SLD) was used to identify children with reading, writing and arithmetical disability. The children taken for this assessment were those perceived as having learning related problems by the school teachers teaching language and mathematics from the third to the seventh standards. NIMHANS index of SLD consists of the following subtests: number cancellation for attention; reading a passage (English); reading comprehension; spelling; copying a passage (English); arithmetic test—simple and graded addition, subtraction, multiplication, division, fractions); Bender Gestalt Test; Benton visual retention test; developmental test for visuo-motor integration and auditory memory test (paired associate learning). Children performing two grade lower than the current grade at the time of assessment are considered learning disabled on the NIMHANS index. Reading and writing tests were also developed and administered for Hindi language for children from the third to the seventh standards. These tests included passages to read and write and test of reading comprehension.
Cognitive Assessment System CAS (Naglieri and Das, 1997) was used to examine the cognitive profiles of children with dyslexia. It is based on the PASS model of cognitive processing, which was initially described as an information-processing model by Luria (1973). Das et al. (1994b) conceptualized the PASS model as a new approach to the assessment of cognitive processes, that is, planning, attention, and simultaneous and successive processing. CAS is organized into four different scales with subtests in each of the four scales. Each scale has a normative mean of 100 and SD of 15. The planning scale consists of tests like matching numbers, planned code, and planned connections. The attention scale consists of expressive attention, number detection and receptive attention. The simultaneous scale consists of non-verbal matrices, verbal–spatial relations and figure memory; and the successive processing scale word series, sentence repetition and sentence questions.
Scoring of the CAS Scales Each subtest of each of the four scales obtains a raw score. These raw scores are converted to scaled scores and further a total scaled score is obtained for each of the four scales. Standard scores are further obtained for each scale and the percentile rank at which the standard score falls. These percentile ranks provide the level of an individual’s cognitive functioning. Psychometric properties: CAS was standardized on 2,200 children and adolescents, 5–11 years of age. Reliability coefficients for the basic battery were high. The full-scale values
Cognitive Profiles of Children with Dyslexia
345
ranged from 0.85 to 0.90 (mean = 0.87) and the average reliability for the basic battery scales was 0.85 (planning), 0.84 (attention), 0.90 (simultaneous) and 0.90 (successive). Evidence for CAS criterion related validity had been obtained from investigations of (a) relationships between PASS processes to tests of achievement and intelligence; (b) the use of strategies and the relationship of strategy use to planning sub-test scores; (c) the PASS performance of the exceptional students (for example, those with attention deficit, mental retardation, learning disabilities, or traumatic brain injury).
Procedure Twenty-nine children from the third to the seventh standards from an English medium, coeducation, medium-level school in Allahabad city, perceived with learning related problems of reading, writing and arithmetic by the teachers, were taken for a formal assessment of SLD using the NIMHANS index of SLD in two sessions per child. Each child was also assessed on Coloured Progressive Matrices/Standard Progressive Matrices (CPM/SPM) to evaluate the intellectual functions. Eighteen children who performed at two grades below the grade in which they were studying on reading, writing and mathematics tests, were identified as learning disabled. These 18 children were further taken for an assessment of cognitive functions using the CAS based on the PASS model.
RESULTS AND DISCUSSION Out of the 18 children identified as dyslexic, eight had writing difficulty, and three were also reported to have mild behaviour problems. Children with dyslexia were found to be average or above average on intellectual functions, which supports the fact that reading disability is independent of the status of intellectual functions. Low intelligence does not predict dyslexia. As we have found that children with dyslexia were in the range of 25th to 75th percentile (mean: 53.2; SD: 21.8) on tests of intelligence (coloured/standard progressive matrices). It has been observed that reading disabled children with average or higher intelligence experience difficulty with both articulatory and phonological coding processes. Children with dyslexia were also found to be low achievers with respect to their academic performance as they scored in the range of 2 to 50 per cent (mean: 25.5; SD: 9.4). The profiles of the 18 children identified as dyslexics on the NIMAHNS Index of SLD revealed specific difficulties in reading ability and reading comprehension. These children performed two grades lower than their current grade on reading tests. Figures 24.1 and 24.2 present the frequency of errors committed while reading in English and Hindi on grade appropriate reading tests.
346 Bhoomika R. Kar and Nishi Tripathi FIGURE 24.1
Errors in reading English Language
Note: Types of errors: 1 = Cannot use phonetic cues; 2 = Ignores punctuation; 3 = Reads word-by-word; 4 = Guesses at words; 5 = Adds words; 6 = Spells out words; 7 = Omits words; 8 = Reversals; 9 = Adds letters; 10 = Difficulty in letter recognition; 11 = Omits/substitutes letters; 12 = Substitutes words; 13 = Unable to read. FIGURE 24.2
Errors in reading Hindi
Note: Types of errors: 1 = Cannot use phonetic cues; 2 = Guesses at words; 3 = Reads word-by-word; 4 = Spells out words; 5 = Adds words; 6 = Ignores punctuation; 7 = Substitutes letters; 8 = Difficulty in letter recognition; 9 = Substitutes matras; 10 = Omits matras; 11 = Adds letters; 12 = Substitutes words; 13 = Omits letters; 14 = Adds matras; 15 = Omits words.
Cognitive Profiles of Children with Dyslexia
347
We found that children having difficulty with English language also had difficulty with Hindi, though to a lesser extent. The second language learning effect on reading does not seem to be very significant as children who committed errors like the inability to use phonetic cues while reading an English passage have also made similar errors while reading a Hindi passage. This highlights the need to reconsider the overestimated effect of second language acquisition on reading ability. It appears that reading disability pertains to a specific difficulty in language processing, irrespective of whether it is the first or second language. The nature of errors committed while reading suggests that phonological processing appears to be one of the most affected process involved in reading English language. The inability to engage in phonological coding has been suggested as the major cause of reading disability (Torgesen et al., 1994). Children with dyslexia showed difficulty in word recognition, resulting in an error like joining letters and reading. Inability to use phonetic cues affects their ability to discriminate between the different sounds associated with different phonemes, ability to join letters/phonemes and pronounce a word. It is accepted that impairment of temporal processing mechanisms results in deficits in phoneme discrimination abilities that affect reading (Barth et al., 1999). Reading wordby-word is suggestive of difficulty with WM and temporal processing. Children have frequently guessed at words, which could be due to deficit in the short-term memory. With some exceptions, Torgesen recognized that the majority of reading disabled children had shorter memory spans than those who are not reading disabled (Torgesen, 1995; Torgesen and Houck, 1980). Though WM is directly related to phonological processing and syntactic awareness, some reports such as one by Das and Mishra (1991) reported weak correlations between memory span and naming time, as well as between these two variables and speech rate, which is a measure of phonological coding and articulation. This, coupled with the results of such studies as those by Bowey et al. (1992), which observed rather weak connections between WM and reading disability, suggests that many poor readers have average or even good memory spans. We also observed mixed results pertaining to the paired associate learning and memory test in the NIMHANS index of SLD. Few children showed deficits in memory span on the dissimilar pair series of the paired associate learning test. Error of guessing at words is closely linked to the difficulty in decoding the words phonologically. The inability to engage in phonological coding has been suggested as the major cause of reading disability (Torgesen et al., 1994). Due to the importance of phonological processing in word decoding, successive processes are naturally expected to be more important at this level. Similarly, Share (1994) suggested a ‘domain-general temporal processing dysfunction’ in reading disabled children. Phonological processing seems to be the most affected component of reading process in the Hindi language also. Errors in reading Hindi, such as an inability to use phonetic cues, guessing at words, reading word-by-word and spelling out words, appear to be
348 Bhoomika R. Kar and Nishi Tripathi similar to those in reading English, along with errors related to the use of matras, which is specific to Hindi language. However, the frequency of such errors in Hindi language is less as compared to those in English, which could be partially due to the first language advantage. Reading comprehension was poor for both Hindi (hit rate: 26.6 per cent) and English (hit rate: 14.4 per cent), but it was better for Hindi. Episodic context and easier search for related words in one’s first language could be one reason, which helps the child comprehend the text better in Hindi language despite of the reading difficulties. Poor reading comprehension could be because the child takes time to decode the words phonologically and does not retain the entire sentence as each word is processed at a time and not as a logical sequence of words. Children with dyslexia were found to be poor on reading comprehension, which is partially dependent on rapid decoding skills, sometimes called as bottom-up processing skills as well as on top-down higher level processing. Higher-level cognitive processes, such as attention and WM, are also important for reading comprehension (Montgomery et al., 1991), which is supported by the below average performance of children with reading difficulties on the attention scale of CAS. On tests of writing in English and Hindi, 50 per cent of the children with reading disability showed writing difficulties also, with higher frequency of errors in English than Hindi. Their performance was replete with spelling, punctuation and capitalization errors (Figures 24.3 and 24.4). Their writing was short, poorly organized and impoverished in terms of ideation. FIGURE 24.3
Writing errors—English
Note: Types of errors: 1 = Adding letters; 2 = Missed out a letter; 3 = Wrong capitals; 4 = No space between words; 5 = Substitutes letters.
Cognitive Profiles of Children with Dyslexia FIGURE 24.4
349
Writing errors—Hindi
Note: Errors: 1= Omits matra; 2= Missed out a letter; 3= Substitutes letters; 4= Adds letters; 5= Omits a sentence.
It has been reported that orthographic processing is not as impaired in dyslexics as phonological processing in English (Siegel, 1999). However, we have observed that orthographic processing is also affected along with phonological processing in Hindi language. The specific deficits in phonological coding and visuo-spatial analysis observed on reading and writing tests need to be understood in terms of the status of cognitive processes mediating the interaction between language processing and visual processing. Children with dyslexia showed deficits on the CAS, which is based on the planning attention, simultaneous and successive processing model of information processing. Figures 24.5 and 24.6 present the cognitive profiles of children with dyslexia on CAS scales with reference to the norms. Children with dyslexia were found to be well below average on PASS scales as indicated by the low Full scale standard score with reference to the norms (Figure 24.5). It has been reported that most of the poor readers are to a certain extent low on all the PASS scales (Das et al., 1994b; Perez-Alvarez and Timoneda-Gallart, 2000). Children with dyslexia were found to be most deficient on simultaneous processing (well below average: standard score= 61.1) tasks, followed by the successive processing tasks (below average: standard score= 71.3). They were also found to be below average on tests of attention (standard score= 71) and planning (standard score= 73.7). We found that the performance of children with dyslexia was below average on tests of receptive and expressive attention, where the test of expressive attention was based on colour word interference effect, which required selection and response inhibition, and the tests of receptive attention such as the speed test of number detection and receptive attention that required rapid visual search, spatial attention, selective attention and response inhibition. Certain studies have reported impairment in rapid visual search, selective attention and change detection in dyslexic children. These studies have reported impairment on the tests of rapid automatized naming which was due to deficits in the speed of access to the lexicon as well as deficits in the
350 Bhoomika R. Kar and Nishi Tripathi speed of visual object recognition (Rutkowski et al., 2003). Below average performance on the tests of planning such as matching numbers, planned codes and planned connections suggested mild deficits in the use of strategies and conceptual tracking. Deficits in planning and attention are also related to the problems in phonological decoding and reading comprehension difficulties observed in dyslexic children. Deficits in planning account for the inefficient utilization of strategies for coding and comprehension (Das and Cummins, 1982; Ramey, 1985). FIGURE 24.5
Mean standard scores on PASS scales of CAS
Note: 1= Planning scale: 73.8 (below average); 2= Simultaneous scale: 61.1 (well below average); 3= Attention scale: 71 (below average); 4= Successive scale: 71.3 (below average); 5= Full scale: 61.7 (well below average). FIGURE 24.6
Mean performance of children with dyslexia on subtests of CAS
Note: 1= Matching numbers; 2= Planned codes; 3= Planned connections; 4= Non-verbal matrices; 5= Verbal spatial relations; 6= Figure memory; 7= Expressive attention; 8= Number detection; 9= Receptive attention, 10= Word series; 11= Sentence repetition; 12= Sentence question.
Cognitive Profiles of Children with Dyslexia
351
The pair-wise comparisons among the four scales of CAS showed significant difference in the performance on simultaneous and successive scales, which further verifies that the simultaneous scale was found more impaired in children with dyslexia (mean: 10.2; SD: 11.7; ‘t’ value: 3.7; p < 0.05). Difficulty in phonological decoding observed in children with reading disability could be linked to the deficits observed on tests of successive processing, whereas deficits on tests of simultaneous processing relate to deficits in visual coding required in reading. Pronunciation of the word is assembled by organizing speech sounds corresponding to the printed word, and this is primarily a successive process requiring the motor programme for oral reading (Figure 24.7). FIGURE 24.7
Cognitive processes in decoding
Source: Das, Naglieri and Kirby, 1994.
Figure 24.7 explains very well the simultaneous contribution of successive and simultaneous processing in dyslexia. Although most of the researches on simultaneous and successive processing in dyslexia have primarily shown deficits in successive processing, as reading is predominantly a temporal process (Perez-Alvarez and Timoneda-Gallart, 2000). South African children with dyslexia have also been reported to have more deficits in successive processing associated with their reading difficulties (Churches et al., 2002). Cognitive tasks that differentiate better between dyslexics and non-dyslexics have been
352 Bhoomika R. Kar and Nishi Tripathi the successive tasks and tasks of attention, which require phonological coding (Das et al., 1994a). Other than these deficits in successive processing, we have found very strong indications of deficits in simultaneous processing, particularly on tests which required both visuo-spatial analysis and linguistic processing. Simultaneous processing has strong spatial and logical-grammatical components and it requires perception of parts into a gestalt (Naglieri and Das, 1997). Below average performance on successive processing could be closely linked with phonological decoding deficit. Poor performance on tests like verbal spatial relations particularly indicates difficulty in visuo-spatial analysis and logical linguistic processing involved in reading difficulties (Figure 24.6). Cognitive processes underlying reading involve both visuo-spatial analysis and linguistic processing. Few evidences have indicated that dyslexics also suffer a low level visual information processing deficit (Slaghuis et al., 1993). Little evidence is available to indicate the extent to which dyslexics simultaneously show a visual and language processing dysfunction, though concurrence of such deficits has been reported to continue even in adult dyslexics (Slaghuis et al., 1996). We have also found greater impairment on simultaneous tasks such as the test of verbal spatial relations, which required visuo-spatial as well as phonological processes to perform the task. Though the dyslexic children were found to be deficient on tests like this, they performed well of tests on visuo-spatial perception alone, which required only visuo-perceptual abilities. We would like to hypothesize that children with dyslexia have difficulties when they need to manipulate both visuo-spatial and phonological information simultaneously, rather than when they need to process visuo-spatial stimuli alone. Recent work on auditory and spatial temporal processing deficits also has shown independent deficits in both. (Becker et al., 2005; Nittrourer, 1999). They have reported elevated target detection latencies, both for auditory and visual stimuli in children with dyslexia. The present study indicates that when one has to manipulate and process visuo-spatial and linguistic stimuli simultaneously to perform tests like the verbal spatial relations test, the dyslexic child’s performance suffers. This could underlie the faulty and unusual ways of reading we observed in dyslexic children. The kind of errors children have shown, such as spelling out the word, adding or missing out letters and words, guessing words, and so on, could be related to the deficits observed on tests of simultaneous and successive processing. Thus, other than the serial or temporal processing involved in reading, it is also the contribution of visuo-spatial analysis towards phonological decoding, which needs to be understood as both are needed to perform a task like reading.
CONCLUSION Our preliminary observations on the nature of reading and writing errors and the status of cognitive processes of simultaneous and successive processing implied in dyslexia need further verification with a larger sample in each grade, and a developmental analysis of
Cognitive Profiles of Children with Dyslexia
353
cognitive components of dyslexia. We have found that phonological decoding, reading comprehension and visuo-spatial analysis are primarily affected in dyslexia in both English and Hindi. The specific cognitive processes, that is, simultaneous processing, was found most affected in children with dyslexia and successive processing was also found impaired. Simultaneous processing appears to mediate the interaction between visuo-spatial perception and language processing involved in reading, whereas successive processing is involved in the serial order processing involved in phonological decoding (Das, 1988). Mechanisms underlying the specific skills and cognitive components involved in reading Hindi in particular need to be explored. This would help in analysing and explaining the nature of errors unique to English and Hindi, particularly with reference to ways of learning in an Indian context.
ACKNOWLEDGEMENTS The participation of the children and the cooperation extended by the school authorities is deeply acknowledged.
REFERENCES Barth, K., N. Steinbuchel, M. von Wittman, H. Kappert and C. Leyendecker. 1999. ‘Temporal Processing and its Correlation to Phonological Awareness and to Reading and Writing Efficiency’. In Evelin Witruk and Thomas Lachmann (eds), Abstracts of the International Conference on Basic Mechanisms of Language and Language Disorders (pp. 85–98), 26–30 September, Leipziger Universitatsverlag. Becker, C., M.A. Elliot and T. Lachmann. 2005. ‘Evidence for Impaired Visuoperceptual Organization in Developmental Dyslexics and its Relation to Temporal Processes’. Cognitive Neuropsychology, 22: 499–522. Bowey, J., M. Cain and S. Ryan. 1992. ‘A Reading-Level Design Study of Phonological Skills Underlying Fourth Grade Children’s Word Reading Difficulties’. Child Development, 63: 999–1011. Churches, M., M. Skuy and J.P. Das. 2002. ‘Identification and Remediation of Reading Difficulties Based on Successive Processing Deficits and Delay in General Reading’. Psychological Reports, 91: 813–24. Das J.P. 1988. ‘Simultaneous and Successive Processing and Planning’. In R.R. Schmeck (ed.), Learning Strategies and Learning Styles (pp. 101–29). New York: Plenum Press. Das, J. P., and J. Cummins 1982. ‘Language Processing and Reading Disability’. Advances in Learning and Behavioral Difficulties, 1: 3–24. Das, J.P. and R.K. Mishra. 1991. ‘The Relationship Between Memory Span, Naming Time, Speech Rate, and Reading Competence’. Journal of Experimental Education, 59: 129–39. Das, J.P., R.K. Mishra and J.R. Kirby. 1994a. ‘Cognitive Patterns of Children with Dyslexia: A Comparison Between Groups with High and Average Nonverbal Intelligence’. Journal of Learning Disabilities, 27: 235–42. Das, J.P., J.A. Naglieri and J.R. Kirby. 1994b. Assessment of Cognitive Processes. Boston: Allyn and Bacon. Hirisave, U., A. Oommen, and M. Kapur (eds). 2002. Psychological Assessment of Children in Clinical Setting. Bangalore: NIMHANS.
354 Bhoomika R. Kar and Nishi Tripathi Karanth, P. 2003. ‘Introduction’. In Karanth, P. and Joe Rozario (eds), Learning Disabilities in India: Willing the Mind to Learn (pp. 17–29). New Delhi: Sage Publications. Lachman, T. 2002. ‘Reading Disability as Deficit in Functional Coordination’. In E Witruk., A.D. Friederici, and T. Lachman, (eds), Basic Functions of Language, Reading, and Reading Disability (pp. 165–98). The Netherlands: Kluver Academic Publishers. Luria, A.R. 1973. The Working Brain: An Introduction to Neuropsychology. New York: Basic Books. Masaro, D.W. and M.M. Cohen. 1991. ‘Integration versus Interactive Activation: The Joint Influence of Stimulus and Context in Perception’. Cognitive Psychology, 23: 558–614. Montgomery, J.W., J. Windsor and R.E. Stark. 1991. ‘Specific Speech and Language Disorders’. In J.E. Obrzut and G.W. Hynd (eds), Neuropsychological Foundations of Learning Disabilities (pp. 573–601). New York: Academic Press. Naglieri, J.A. and J.P. Das. 1997. Cognitive Assessment System. Illinois: Riverside Publishing. Nittrouer, S. 1999. ‘Do Temporal Processing Deficits Cause Phonological Processing Problems?’ Journal of Speech, Language and Hearing Research, 42: 925–42. Perez-Alvarez, F. and C. Timoneda-Gallart. 2000. ‘Dyslexia as a Dysfunction in Successive Processing’. Review of Neurology, 30: 614–19. Ramey, G.W. 1985. The Role of Planning in Reading Comprehension. Unpublished Ph.D. Thesis, University of Alberta, Edmonton. Raven, J.C. 1998. Coloured Progressive Matrices. London: H.K. Lewis and Co. Ltd. ———. 2000. Standard Progressive Matrices. London: H.K. Lewis and Co. Ltd. Rutkowski, J.S., D.P. Crewther and S.G. Crewther. 2003. ‘Change Detection is Impaired in Children with Dyslexia’. Journal of Vision, 3: 95–105. Share, D. L. 1994. ‘Deficient Phonological Processing in Disabled Readers Implicates Processing Deficits Beyond the Phonological Module. In K. P. van den Bos, L. S. Siegel, D. J. Bakker, and D. L. Share (eds), Current Directions in Dyslexia Research (pp. 149–67). Lisse, Netherlands: Swets & Zeitlinger. Share, D.L. and K.E. Stanovich. 1995. ‘Cognitive Processes in Early Reading Development: Accommodating Individual Differences into a Model of Acquisition’. Issues in Education, 1: 1–57. Shaywitz, S.E. 1996. ‘Dyslexia’. Scientific American, November: 98–104. Siegel, L.S. 1999. ‘The Cognitive Components of Dyslexia’. In Evelin Witruk and Thomas Lachmann (eds), Abstracts of the International Conference on Basic Mechanisms of Language and Language Disorders (pp. 7–14), 26–30 September, Leipziger Universitatsverlag. Slaghuis, W.L., W.J. Lovegrove and J.A. Davidson. 1993. ‘Visual and Language Processing Deficits are Concurrent in Dyslexia’. Cortex, 29: 601–15. Slaghuis, W.L., A.J. Twell and K.R. Kingston. 1996. ‘Visual and Language Processing Disorders are Concurrent in Dyslexia and Continue into Adulthood’, Cortex, 32: 413–38. Suresh, P.A. and S. Sebastian. 2003. ‘Epidemiological and Neurological Aspects of Learning Disabilities’. In P. Karanth, and J. Rozario (eds), Learning Disabilities in India: Willing the Mind to Learn (pp. 30–43). New Delhi: Sage Publications. Torgesen, J. 1995. ‘Instructions for Reading Disabled Children: Questions About Knowledge and Practice’. Issues in Education: Contributions from Educational Psychology, 1: 91–6. Torgesen, J. and G. Houck. 1980. ‘Processing Deficiencies in Learning Disabled Children Who Perform Poorly on the Digit Span Task’. Journal of Educational Psychology, 72: 141–60. Torgesen J.K., R.K. Wagner and C.A. Rashotte. 1994. ‘Longitudinal Studies of Phonological Processing and Reading’. Journal of Learning Disabilities, 27: 276–86. Tripathi, N. and B.R. Kar. In press. ‘Teachers’ Perception of Learning Related Problems in Children in School Settings in Allahabad City’. In K. Thapa, G. Aalsvoort and J. Pandey (eds), Perspective on Learning Disabilities in India: Current Practices and Prospects. New Delhi: Sage Publications.
Chapter 25 Emergence of Social Play and Numeracy: A Related Development with Young At-Risk Students? Geerdina M. van der Aalsvoort, Arjette M. Karemaker and Mieke P. Ketelaars
INTRODUCTION
T
he process of observing changes in reciprocal interaction as a proof of changes over time is addressed in this chapter by using an experimental design including a microgenetic study to study social play. First, sociocultural theory will be described to frame the general research question, namely, whether social play and numeracy develop differently by students at-risk visiting regular primary education than by students at-risk visiting special primary education. Next, the clinical group that was involved in the study is described. The results of the intervention are then presented.
SOCIO-CULTURAL THEORY AND THE MEANING OF PLAY Within socio-cultural theory, culture is understood as the created environment of people. It contains views, meanings and ways of collaborating, rules and values as well as visible ‘products‘, for example, signs and symbols in the material world. According to Vygotsky (1978), the child needs to interact with this created environment in order to learn. Results of learning thus comprise knowledge that is socially constructed. Next to cultural meaning, situations in which tasks are presented are related to specific learning processes. There is a dynamic exchange between individual meaning making and the task at hand. The result of this exchange is displayed in increasing stability of performance related to individual variability (Rogoff, 1998).
356 Geerdina M. van der Aalsvoort et al. Within this framework, play is a form of imitation. As a culturally meaningful practice, the child creates zones of proximal development in play by visualizing processes of interiorization. Play is meaningful for development. The child experiments with his or her environment, materials and language. He or she experiences social processes and tries to influence them. He can take initiatives and try to resolve problems to get acquainted with the cultural customs and the role that he is allowed to take (Van Oers and Wardekker, 1997). It seems logical that play also reveals how children experience the classroom in that they play in order to understand the social practices they are experiencing. The quality of play, as seen in the skills children use to solve social and cognitive problems during play, have been found to be related to future success in school as well as to the ability to work collaboratively together with colleagues later in life (Hännikäinen, 1998; Lloyd and Nowe, 2003; Van Oers, 1997; Verba, 1998; Wyver and Spence, 1999). Does quality of play also include social play? Göncü (1993) suggests, on the basis of work of Piaget, Vygotsky and Parten (Göncü, 1993: 100), that social play requires intersubjectivity. This emerges in the third year of life when solitary play gives way to social play. Children achieve intersubjectivity in social play by negotiating their ideas with one another. Social play is thus a major developmental task, as children become part of a classroom and a school community. In social play, there are different kinds of cooperative formats related to joint role-play, which reveal that cooperation develops from asymmetrical interactions into symmetrical co-elaboration. Various social, motivational and cognitive factors become interwined and balanced during cooperation as opinions and intentions related to goal orientation, meanings and management are shared (Verba, 1993, 1994, 1998). The children’s play and learning are not cognitive activities in themselves, but they come about related to situational factors, such as activity, time and actors. Göncü (1993) studied how social play evolves with 3-year-olds compared to 4-yearolds by investigating how play interactions are expanded. His study included pairs of children of the same sex from one regular primary school who were friends according to their teachers. He showed that there was a significant effect of age with respect to length of sequences of social play. His study was an interesting start of studying social play with youngsters. As children become more aware of cultural tools used by the adults surrounding them, it can be expected that emerging interest in mathematics is revealed during social play. The interactions displayed can show emerging number sense when concepts related to role play are discussed as well acted out during counting and discussing numbers of toys (Malofeeva et al., 2001). We expect that social play in itself reinforces the appearance of utterances that are related to emergent mathematics and rises above the influence of instruction on academic assignments. As Göncü (1993) did not address the topic of situatedness of social play in his investigation, we will include environmental characteristics, such as the quality of the
Emergence of Social Play and Numeracy
357
classroom that are related to the effectivity of interventions (Kontos and Keyes, 1999; Nolen, 2001). Moreover, we will control for the socio-economic background of students when we assign students to intervention conditions to elicit social play.
YOUNG CHILDREN AT-RISK FOR LEARNING DIFFICULTIES IN THE NETHERLANDS One per cent of all young children in the Netherlands are in special primary schools by the time formal academic learning of reading and writing starts at the age of six (van der Aalsvoort et al., 2002). Although these children at-risk for learning difficulties appear to have special educational needs, it is not clear whether placement in special primary education is the solution to their problems. We refer to these children as ‘at-risk’ because they: ‘… manifest some or all of the following behavioural characteristics: difficulty in using language fluently and effectively in a range of situations, inability to attend to and persevere with tasks and activities, lack of purposefulness, imagination and variety in play, lack of initiative, lack of ‘normal’ social and emotional maturity’. (Elliott and Hall, 1997: 198) Accurate placement in special education of these children is difficult for several reasons. First, reliable prediction of success or failure to become readers and writers is difficult. Children develop at their individual pace and seemingly slow or quick progress can be caused both by endogenous and environmental factors (Hauser-Cram et al., 1993; Kontos and Keyes, 1999; Pianta and Stuhlman, 2004). In other words, some of the so-called students at-risk will indeed become poor readers and writers; others, however, will not. Second, placement recommendation for a special primary school at a young age suggests a clear-cut advantage for the child being among comparable performing peers. Results of teaching homogeneous groups of ‘slow developers’, as opposed to heterogeneous groups of students, however, are mixed. Some studies have revealed that attending school with comparable performing peers will make it easier for the child to feel at ease (Chandler et al., 1992; Guralnick et al., 1996). Other findings suggest that education in the early years is the result of a challenging environment. Thus, on the one hand a child may feel challenged by this environment irrespective of adult involvement (Kontos et al., 2002; Stagnitti and Unsworth, 2000). On the other, the school as an environment can play a negative role in that children at-risk may be perceived as insensitive to teaching efforts (Keogh, 1982 in Coplan et al., 1999). The arguments listed above are at the centre of discussions in the Netherlands about the content of diagnostic decision-making related to referral of young children to special education. The question is, what type of information is helpful in deciding whether a special primary school suits the needs of a child at-risk for learning difficulties? Our study aims at clarifying the role of schooltype with young children at-risk since the Dutch school system provides two types of academic environment for young children
358 Geerdina M. van der Aalsvoort et al. developing at-risk. Children enter primary school at the age of four in Dutch schools, both in regular primary and special primary education. Teachers in Grades 1 and 2 offer a comparable programme during a school day. Each day involves whole-group sessions. These sessions include storybook reading, singing songs and children telling everyday experiences. Birthday celebrations and festivities can also be part of these sessions. The rest of the day includes tasks that elicit experiences of emergent literacy and mathematics, alternated by snack time, playtime in the classroom based on schedules and playtime outside. Thus the importance of play is fully understood. Why would we expect differences between school types? We found that differential influence may occur because of the time to interact with peers and teachers differs between regular and special primary education (van der Aalsvoort et al., 2002). Although the mean group size of classes in regular primary schools is about 25 compared to a maximum of 15 in those of special primary schools, several characteristics of special education are different. First, teachers in special primary schools tend to interact with each child individually more often in their wish to respond to the child’s special needs. At the same time, the child loses opportunities to interact with his peers. Second, children placed in special schools are often included in individual therapy sessions such as speech therapy and locomotor therapy, which also decrease chances to be involved with peers. We do not know whether these differences have an impact upon the child’s developmental progress.
RESEARCH QUESTION The general research question to be answered was to do with whether an intervention that includes opportunities to play improves mathematical knowledge as assessed by standardized achievement tests. We expected that the intervention would have a differential effect upon the students involved as two types of education were included.
Design The design aimed at describing development in social play. Therefore, we planned a longitudinal experimental design. The design involved dyads of young at-risk children from regular and special education in two conditions. The dyads of the experimental condition took part in seven play sessions and those of the control condition participated in two sessions only. To allow statistical group comparison the subjects were selected based on intelligence score, passive language development and temperament. Whole classrooms were tested with respect to passive vocabulary and intelligence to include children developing at-risk. Moreover, the teachers rated the children’s temperament to ensure that children who were either too introvert and at-risk for communication disorder or too extrovert and at-risk for behavioural disorders were excluded from participating in
Emergence of Social Play and Numeracy
359
the study. The children selected were then matched within classrooms according to age and sex to achieve dyads. Within schools, several classrooms could take part. Randomly dyads were chosen to be part of the experimental or control condition. At least one of the dyads was part of the control condition in the case of more than three dyads within a classroom. To control for quality of home and school environment, the home environment was assessed by collecting data on education and occupation of both parents. Moreover, pedagogical quality within the classroom and within the school was rated. Table 25.1 shows that the intervention took three weeks, in which the experimental conditions played six sessions, followed by a seventh session eight weeks after the sixth session to control for maturity. The control conditions took place in sessions one and seven only. At the end of Grade 2 (June 2004) and in the middle of Grade 3 (March 2005), mathematical knowledge was assessed. TABLE 25.1
Design of the longitudinal study
RPS
regular primary school
SPS
special primary school
E-C
E = experimental group; C = control group
C
Co-variables: ECERS, SES
S
Selection of subjects: Raven CPM; Passive vocabulary; Temperament
P1–P6
Play session 1 to 6
P7
Play session 7
AP1
Tests to assess mathematical knowledge: Ordering (CITO), Number sense
AP2
Tests to assess mathematical knowledge (CITO)
Play-Session A play-session typically took place in a room separate from the classroom and the room included a video camera. Before the first session, each dyad was made familiar with the camera. Next the task was offered to trigger the children’s interest and materials that aimed at eliciting social play. The researcher said: ‘The two of you can play. Do you see the blocks and toy-animals? Can you build a zoo for me? It’s up to you to think about how to build the zoo and which animals are going to live in it. I will tape your play.’ The researcher was present during the sessions but she did not participate or interfere in the unfolding activities. A session was stopped after 20 minutes or when one of the children asked permission to leave the room. The researcher then said: ‘You built a lot of things! I will make a photo of your zoo so that you can play with it again next time.’ Then she escorted the dyad back to the classroom. The same procedure was followed for the experimental and control condition. The experimental condition then had five
360 Geerdina M. van der Aalsvoort et al. more sessions. In the second session, the configuration built at the end of the first session was built again. The researcher asked the dyad whether she had done this correctly, and invited the dyad to start playing again. Thus, sessions three to six was carried out. When the dyads were invited to play once more after two months, the instruction was the same one as in session one.
Instruments Passive Vocabulary: Subtest of Language of Children in Kindergarten The norm-referenced test is administered individually. The child is offered two practice items before test items are presented. The test includes four subtests: two for expressive language development and two for passive language development. The passive vocabulary test was used to select participants. In the test, the child is offered four pictures, and the tester gives one word. The child should point at the picture that represents the word. Scores ranged from 1 (high) to 5 (low). All children from one classroom were tested to identify who would participate in the study. The cut-off score was lower than 3.
Raven Coloured Progressive Matrices This is a non-verbal intelligence test that can be administered to children aged between four to 10. The test contains 36 items. In each item the child is asked to point at a picture that would complete a series of three pictures. The choice made requires analogical reasoning. The raw score of each child was compared to Van Bon’s standard norms of the Dutch population (Van Bon, 1986). Scores ranged from 9.9 (high) to 0.9 (low). All children from one classroom were tested to identify who would participate in the study. The cut-off range was higher than 2 and lower than 6.
The School Behaviour Evaluation List The instrument contains 52 rating scales that refer to four factors: extraversion, task behaviour, emotional stability and agreeableness. Teachers rated each child separately and then the factors were traced and related to norm tables. Factor scores could range from 1 to 19. Children who rated between 4 and 16 for each factor were selected.
Pedagogical Quality The schools were rated using the Early Childhood Environment Rating Scales (ECERS) (Harms et al., 1998). This rating scale includes seven categories that can be rated from 1 (poor quality-level) to 7 (high quality level). The categories are: space and furnishings, personal care and routines, language-reasoning, activities, interaction, programme structure and parents and staff. The ECERS was rated in each classroom of the school
Emergence of Social Play and Numeracy
361
which took part in the study. The categories parents and staff, however, were rated on school level only. Reliability was checked by rating four classrooms from different schools twice and the ratings were highly comparable: SPS-1: r = 0.95; SPS-2: r = 0.99; RPS-1: r = 0.93; RPS-2: r = 0.98; RPS-3: r = 0.98.
Socio-economic status (SES) The SES of each parent was rated using the Standard International Socio-Economic Index of Occupational Status (Ganzeboom et al., 1992). The index contains four levels of education and six levels of occupation. The levels of education ranged from 1 (low education level) to 4 (high quality level), whereas level of occupation ranged from 1 (low occupation level) to 6 (high occupation level). The results of both indexes were added for the fathers and the mothers separately.
Mathematical Knowledge Ordering The test is administered individually. The child is offered two practice items before test items are presented. Then page 1 of the testbooklet of Grade 2 test is presented. When a child failed on more than two items of the 4 on the first page, the child was presented the testbooklet of Grade 1. The booklet contains 42 items. Each item requires that the child points at the picture that fits the question. Although the booklet can be used as a classroom assessment, the children were tested individually. The norm tables for Grade 2 students were used to derive the normscore. A score was given to reveal skill and performance related to comparable age groups. Scores ranged from 1 (high) to 5 (low).
Number Sense The task includes the principles of number sense as suggested by Piaget. They are counting to 10 and back: six items; sense of measure: eight items; conservation: six items; correspondence: six items; classification: four items; seriation: seven items. The results on the items solved correctly on Piagetian tasks were used for the study as well as the score on counting.
Mathematical Test (CITO) The test is administered individually, and contains 24 items related to mathematical knowledge included in problem solving tasks. The norm tables for Grade 3 students were used to derive the normscore. A score was given to reveal performance related to comparable age groups. Scores ranged from 1 (high) to 5 (low).
362 Geerdina M. van der Aalsvoort et al. Analyses of the Transcripts The videotaped sessions were transcribed on verbal utterances and non-verbal behaviours displayed. Next the transcripts were analysed with respect to quality of social play and utterances of emergent mathematics. Collaboration during play was determined using an adjustment of the procedure of Verba (1994). All play-sessions were transcribed followed by a three-step procedure. Step 1: Assessing episodes in the transcripts. An episode takes place between the moment that one child talks to the other child of the dyad or elicits response by materials used and the moment that the interactions stop. The duration of the episodes is expressed in seconds. The percentage of the total amount of seconds in episodes related to the total time of the session is computed to express playtime. Step 2: Assessing cognitive level per episode as realism (playing with blocks and toyanimals as they are) or role-play. The percentage of the total number of episodes related to realism and role-play are computed to express cognitive level. Step 3: Level of collaboration. Each episode is analysed to categorize simple and deep social play. Simple: Seeking contact: saying something and being heard; reciprocal eye-contact; verbal communication may be part of the exchange but adds little information to the interaction. Deep: Seeking contact: saying something and being heard; reciprocal eyecontact followed by verbal communication followed by planning (proposing ideas, and/or giving directions and or carrying them out) and/or evaluating (role-) play. Emerging mathematical knowledge was identified by identifying utterances referring to mathematical meaning making, such as numerical relationships (comparing, counting, utterances related to time and to space) during episodes. The numbers of utterances found were used to compute mathematical knowledge. The percentage of episodes that contained these utterances was computed and used for analysis.
Participants The study was carried out with 56 children from five regular and six special primary schools from different regions in Holland. The scores for the two conditions at each school condition are listed in Table 25.2. The school conditions were comparable with respect to all variables except for age, with F (3, 52) = 4.904, p = 0.004. These results were taken into account by using analyses of covariance using age as covariate with respect to the variables compared.
Emergence of Social Play and Numeracy
363
The selection procedure included that scores on factors related to temperament could range from four to 16. Therefore we tested whether the mean scores for the four factors were comparable between conditions. The findings revealed that task behaviour differed between schooltypes, F (3, 50)=21, 325, p=0.006, and within schooltypes: regular primary schools, t (36)=2,853, p=0.005, and special primary schools, t (14)=2,359, p=0.033. Moreover, emotional stability differed between schooltypes, F (3, 50)=18,423, p=0.019, and within special primary schools, t (14)=3,787, p=0.002. These findings will be used when we analyse the correlations of findings related to quality of social play. TABLE 25.2 The means and standard deviations of the instruments used to select the subjects listed per school and per research condition RPS: e-condition
RPS: c-condition
SPS: e-condition
SPS: c-condition
24
14
12
6
N M Age
SD
M
SD
M
SD
M
SD
67.0
2.83
69.0
4.81
71.0
7.45
74.0
2.34
Raven score
4.1
0.88
4.3
0.85
4.4
1.10
3.8
1.38
Language score
3.1
1.53
3.2
1.37
3.8
1.55
4.0
0.82
10.6
2.36
10.8
2.23
10.9
3.20
11.3
1.50
9.0
2.33
11.1
1.92
9.9
1.93
7.3
2.06
Agreableness
8.8
2.77
9.9
2.84
9.5
3.06
7.8
2.22
Emotional st.
10.0
2.40
9.7
2.43
8.1
1.78
12.0
1.83
Temperament Extraversion Task Behaviour
We compared the pedagogical quality of the research conditions, and SES. The results are listed in Table 25.3. When the conditions were compared related to school type, remarkable differences were found with respect to activities, F (3, 52)=5,301, p=0.003, and programme structure, F (3, 52)=3,863, p=0.000. When the experimental conditions were compared within school condition, differences showed related to programme structure within the regular primary schools, t (14,831)=3,500, p=0.003. Language-reasoning, t (13,470)=2,714, p=0.017, and activities, t (16)=2,344, p=0.032 differed within the intervention conditions of the special primary schools. These findings will be included when we discuss the results. Regarding SES, no significant differences were found between schools and between conditions.
364 Geerdina M. van der Aalsvoort et al. TABLE 25.3 The means and standard deviations of the ECERS ratings per category and the SES per school and per research condition RPS: e-condition
RPS: c-condition
24
N M
SPS: e-condition
14 SD
M
SPS: c-condition
12 SD
M
6 SD
M
SD
ECERS Space and furnishings Personal care routines Languagereasoning
6.2
0.48
6.0
0.44
5.7
0.58
5.8
0.52
6.0
0.77
5.9
0.09
5.4
1.45
5.2
1.33
6.1
0.94
5.5
1.21
5.5
0.89
6.3
0.22
Activities
5.0
0.69
4.7
0.45
4.2
0.56
4.7
0.34
Interaction
6.1
1.22
6.4
0.51
5.6
1.09
6.3
0.85
6.7
0.40
5.6
1.14
6.3
0.57
6.7
0.30
5.5
0.61
5.6
0.61
5.6
1.01
5.3
0.78
Father
6.0
2.39
5.7
1.98
6.3
2.63
4.3
0.58
Mother
5.6
2.12
5.6
2.19
5.0
0.00
4.7
0.58
Programme structure Parents and staff SES index
Procedure The principals of schools for regular primary and special primary education were asked to participate in the study. When the principals agreed, the teachers from Grade 1 were addressed. As soon as they gave their consent the study’s procedure was explained to each teacher. Then a letter that was composed by the research team informed the parents and they were invited to give their consent. After having received the written consent, the assessment of child characteristics began. Due to the strict selection procedure only five regular primary schools and six special primary schools could take part. Then the ECERS was carried out including an inter-rater reliability check. Next the play sessions were planned and carried out accounting for school holidays, school schedules, and so on. No child was absent for sessions one to six. One dyad from the special school condition was missing from session seven as one student had left the school. After the last session, the children were thanked for their cooperation. The teacher of the children received books as a present for her class. In June 2004, the subjects were tested with respect to emergent mathematics. Each participating teacher received the results of the tests of the subjects that had been collected for selection and the results on the tests on emergent mathematics. In
Emergence of Social Play and Numeracy
365
March 2005, mathematical knowledge was tested again. This was possible in the regular primary schools only as the teacher from the special primary schools had not offered mathematics instruction to the students yet. There were two dyads missing from the regular primary schools as the parents of those children had moved to another city. One dyad of the special primary schools was missing as one of the students had been referred to another school. The videotapes were transcribed and coded. The data of all the tests were coded and the data were analysed with a statistical package (SPSS).
RESULTS We expected that an intervention including opportunities for social play would improve mathematical knowledge. The findings related to this hypothesis are presented in Table 25.4. TABLE 25.4 The means and standard deviations on number sense, counting and ordering per school and per research condition RPS: e-condition N
RPS: c-condition
24
SPS: e-condition
14
SPS: c-condition
12
6
M
SD
M
SD
M
SD
M
SD
24.7
18.54
21.9
17.01
39.5
18.95
16.3
6.35
Mathematical utt: 2
52.7
26 .97
59.7
30.32
59.0
28.43
27.3
22.22
Number sense
23.0
2.70
25.4
2.10
21.4
5.18
14.8
3.49
2.6
1.10
2.5
1.09
3.4
0.98
4.8
0.41
2.5
0.94
1.2
0.44
Grade 2 Mathematical utt: 1
Ordering Grade 3 Mathematic test
An analysis of covariance (ANCOVA) with age as covariable revealed that performance on number sense tasks was higher in the experimental condition than in the control condition (F (3)=15, 354, p=0.000). This finding was related to a significant difference between the intervention conditions of the regular and special primary schools. The same results were found with respect to ordering, F (3)=10,258, p=0.000, Λ =, 19. This finding was related to a significant difference between the intervention conditions of the special primary schools only. No remarkable differences were found with respect to frequency of utterances related to mathematical knowledge. As to the results on the mathematic test, a significant difference was found in favour of the control condition, F (1)=17,685, p=0.000, Λ =, 42) in the experimental condition of the regular primary schools.
366 Geerdina M. van der Aalsvoort et al. DISCUSSION AND CONCLUSION Our study aimed at clarifying the role of schooltype with young children at-risk since the Dutch school system provides two types of academic environment for young children developing at-risk. The study was undertaken to study whether opportunity at social play would improve mathematical knowledge. The comparison of intervention conditions on a group level suggests that mathematical knowledge came forward stronger in the experimental condition than in the control condition. This finding was mainly due to the improvement of scores of the subjects in the special primary schools. No remarkable differences were found with respect to the frequency of utterances related to mathematical knowledge, although the number of the utterances improved remarkably in all the conditions. Moreover, the results of the mathematical test in Grade 2 suggest that the experimental condition performed better, although this could not be confirmed with the findings of Grade 3 as the students in the special primary schools could not be tested: the instruction of mathematics had not started yet. Thus, although the findings reported here suggest that profiting from collaboration may be more significant for children attending special primary education than for those from regular primary education, we cannot be sure as the context of the students differs from those attending regular primary education. The study presented offers insight into how young at-risk children intertwine social play with developing concepts of academic knowledge within the school as a context. We reported the case of dyads involved in the study elsewhere (van der Aalsvoort et al., 2005) to clarify the kind of processes that are involved in social play. As the study reported here is a study-in-progress, we will investigate further whether temperament as a child-related factor is related to the progress of mathematical knowledge. It may be that task behaviour and emotional stability are related to the dyad’s possibilities to profit from social play (Cole et al., 2004; Lutz et al., 2002; McClelland et al., 2000; Ogden, 2000). We stress the finding that the social play emerging did not seem to be constrained by the poor language level of the participants. Moreover, even within a very short period of time, the participants showed their potential to collaborate and exchange their emerging knowledge on numbers and use of mathematical knowledge, thus inviting teachers to make use of their potential in the classroom. The design of the study suggests that there are many opportunities to observe the child’s development-in-progress.
REFERENCES Chandler, L.K., R.C. Lubeck and S.A. Fowler. 1992. ‘Generalization and Maintenance of Preschool Children’s Social Skill: A Critical Review and Analysis’. Journal of Applied Behavior Analysis, 25: 415–28.
Emergence of Social Play and Numeracy
367
Cole, P.M., S.E. Martin and T.A. Dennis. 2004. ‘Emotion Regulation as a Scientific Construct: Methodological Challenges and Directions for Child Development Research’. Child Development, 75: 317–33. Coplan, R.J., A.M. Barber and D.G. Lagacé-Séguin. 1999. ‘The Role of Child Temperament as a Predictor of Early Literacy and Numeracy Skills in Preschoolers’. Early Childhood Research Quarterly, 14: 537–53. Elliott, A. and N. Hall. 1997. ‘The Impact of Self-Regulatory Teaching Strategies on ‘At-Risk’ Preschoolers’ Mathematical Learning in a Computer-Mediated Environment’. Journal of Computing in Childhood Education, 8: 187–98. Ganzeboom, H.B.G., P.M. De Graaf and D.J. Treiman. 1992. ‘A Standard International Socio-Economic Index of Occupational Status’. Social Science Research, 21: 1–56. Göncü, A. 1993. ‘Development of Intersubjectivity in the Dyadic Play of Preschoolers’. Early Childhood Research Quarterly, 8: 99–116. Guralnick, M.J., R.T. Connor, M.A. Hammond, J.M. Gottman and K. Kinnish. 1996. ‘The Peer Relations of Preschool Children with Communication Disorders’. Child Development, 67: 471–89. Hännikäinen, M. 1998. ‘From Togetherness to Equal Partnership in Role-Play’. Early Child Development and Care, 142: 123–32. Harms, T., R.M. Clifford and D. Cryers. 1998. Early Childhood Environment Rating Scale. New York: Teachers College Press. Hauser-Cram, P., M.B. Bronson and C.C. Upshur. 1993. ‘The Effects of the Classroom Environment on the Social and Mastery Behavior of Preschool Children with Disabilities’. Early Childhood Research Quarterly, 8: 479–97. Kontos, S. and L. Keyes. 1999. ‘An Ecobehavioral Analysis of Early Childhood Classrooms’. Early Childhood Research Quarterly, 14: 35–50. Kontos, S., M. Burchinal, C. Howes, S. Wisseh and E. Galinsky. 2002. ‘An Eco-Behavioral Approach to Examining the Contextual Effects of Early Childhood Classrooms’. Early Childhood Research Quarterly, 17: 239–58. Lloyd, B. and N. Nowe. 2003. ‘Solitary Play and Convergent and Divergent Thinking Skills in Preschool Children’. Early Childhood Research Quarterly, 18: 22–41. Lutz, M.N., J. Fantuzzo and P. McDermott. 2002. ‘Multidimensional Assessment of Emotional and Behavioral Adjustment Problems of Low-Income Preschool Children: Development and Initial Validation’. Early Childhood Research Quarterly, 17: 338–55. Malofeeva, E., D. Cianco and D. Jeanne. 2001. ‘Development of Emergent Math and Literacy Skills’. Paper presented at the Annual Meeting of the American Psychological Association, 109th, 24–28 August, San Francisco, California. McClelland, M.M., F.J. Morison and D.L. Holmes. 2000. ‘Children at Risk for Early Academic Problems: The Role of Learning-Related Social Skills’. Early Childhood Research Quarterly, 15: 307–29. Nolen, S.B. 2001. ‘Constructing Literacy in the Kindergarten: Task Structure, Collaboration and Behavioral Problems. Cognition and Instruction, 19: 95–142. Ogden, L. 2000. ‘Collaborative Tasks, Collaborative Children: An Analysis of Reciprocity During Peer Interaction at Key Stage 1’. British Educational Research Journal, 26: 211–26. Pianta, R.C. and M.W. Stuhlman. 2004. ‘Conceptualizing Risk in Relational Terms: Associations Among the Quality of Child–Adult Relationships Prior to School Entry and Children’s Developmental Outcomes in First Grade’. Educational and Child Psychology, 21: 32–46. Rogoff, B. 1998. ‘Cognition as a Collaborative Process’. In W. Damon (ed.), Handbook of Child Psychology (pp. 679–744). New York: John Wiley. Stagnitti, K. and C. Unsworth. 2000. ‘The Importance of Pretend Play in Child Development: An Occupational Therapy Perspective’. British Journal of Occupational Therapy, 63: 121–27.
368 Geerdina M. van der Aalsvoort et al. Van Bon, W.H.J. 1986. Raven’s Colored Progressive Matrices: Norm tables. Lisse, The Netherlands: Swets & Zeitlinger. van der Aalsvoort, G.M., A.M. van Tol and M. Thomeer-Bouwens. 2002. Zorg bij jonge risicokinderen: professionele toewijding gevraagd [Care for Young at-risk Children: Professional Care Required]. Leiden, The Netherlands: Universiteitsdrukkerij. van der Aalsvoort, G.M., M.P. Ketelaars and A.M. Karemaker. 2005. ‘Social Play by Young At-risk Students: A Microgenetic Approach to Study Emergent Collaboration and Numeracy’. Journal of Education, 35: 159–80. Van Oers, B. 1997. ‘On the Narrative Nature of Young Children’s Iconic Representations: Some Evidence and Implications’. International Journal of Early Years Education, 5: 237–45. Van Oers, B. and W. Wardekker. 1997. ‘De cultuurhistorische school in de pedagogiek [The Culture-Historic School in Pedagogy]’. In S. Miedeman (ed.), Pedagogiek in meervoud [Plural pedagogy] (pp. 171–213). Houten/Dieghem, The Netherlands: Bohn Stafleu Van Loghum. Verba, M. 1993. ‘Cooperative Formats in Pretend Play Among Young Children’. Cognition and Instruction, 11: 265–80. ———. 1994. ‘The Beginnings of Collaboration in Peer Interaction’. Human Development, 37: 125–39. ———. 1998. ‘Tutoring Interactions Between Young Children: How Symmetry Can Modify Symmetrical Interactions’. International Journal of Behavioral Development, 22: 195–216. Vygotsky, L.S. 1978. Mind and Society: The Development of Higher Mental Processes. Cambridge, Massachusetts: Harvard University Press. Wyver, S.R. and S.H. Spence. 1999. ‘Play and Divergent Problem Solving: Evidence Supporting a Reciprocal Relationship’. Early Education and Development, 10: 419–44.
Chapter 26 Cognitive Stimulation of Rural School Children in India: An Evaluative Study Malavika Kapur
INTRODUCTION
I
t has been widely acknowledged in India that a significant proportion of the children, especially those from under-privileged backgrounds and girls, drop out before they reach Class V or, even if they continue to attend school, learn very little. It is common knowledge that there is a wide gap in the learning achievements between government schools (urban and rural) and private/aided schools (Ramachandran, 2003). There have been several initiatives in the various states to improve the quality of education in the government-run schools. The government-run DPEP (District Primary Education Project) initiated its Nali–Kali (Rejoice–Learn) approach in H.D. Kote in Karnataka in 1995. The Pratham initiative started in Maharastra has been attempted elsewhere in the country, and is an example of the bold efforts made across the country. However, such initiatives are rare exceptions. The mainstream government schools have been mostly left untouched. A review of over 50 such innovative approaches by Kumar and Sarangapani (2005) reveals that these efforts aim at improving subjects of the school curriculum, and involves teacher training to promote the school performance especially in language and maths. These efforts rely heavily on teacher motivation and involvement. Most importantly, the evaluation of these programmes has not been carried out in a systematic manner, and the sustainability of these initiatives has not been examined. (Ramachandran, 2003). All the programmes have provided us with better models to choose from. The emphasis is on improving the school infrastructure and teacher contribution. The role of the children in improving the quality of their environment has been paid scant attention. Enabling children to be skilled learners even in the schools with poor infrastructure is important, especially among children who come from rural and tribal backgrounds.
370 Malavika Kapur PRESENT STUDY The present study attempts to evolve and evaluate an innovative programme that can become part of daily school routine without further burdening the teachers and using the available infrastructure. The objective of the study was to evolve and evaluate a programme to promote cognitive development of rural school children. More specifically the objectives were: • To develop an intervention programme that focuses on play and other child friendly approaches. • To adopt a child-to-child rather than a teacher centred approach. • To assess the short-term effects of the programmes through psychological tests before and after the intervention. • To assess the outcome of the intervention on attention, intelligence, language, arithmetic skills and creativity of the children. • To target the entire pupil population from Class I to Class IX for a universal intervention.
Sample A group of 15 schools was taken up for the project from one of the 19 cluster of schools in H.D. Kote, in rural Mysore district, 250 km from Bengaluru in the state of Karnataka. The children in the government schools mostly belonged to lower castes and the tribal population. Most of them were first generation literates. The area, though economically backward, is rich in the abundant flora and fauna, and also in culture, especially among the tribals. The total number of pupils was approximately 1,200, and the number of teachers was 41. Most of the schools, especially the primary ones, had single teachers and lacked basic amenities such as classroom furniture, toilets, drinking water and play/sports equipment. More often than not, the school had only one or two classrooms and the classes had to be held for the children in three or four grades in the same classroom. The classrooms had poor illumination and ventilation and were indifferently maintained, bearing the brunt of heavy monsoon rains. Teacher and pupil absenteeism was equally common. The teachers had the problems of commuting great distances each day with scarce public transport and dealing with the poor infrastructure, all of which resulted in poor motivation and commitment. Pupils’ absences were common during the seasons of agricultural demand of the farming communities. For example, during the cotton picking season, the children stayed from the schools for even a month or more.
Cognitive Stimulation of Rural School Children in India: An Evaluative Study
371
Tools (i) Tests of attention (a) Colour Cancellation Test (Kapur, 1975) for children of Classes I and II to assess attention for simple and complex tasks. (b) Number Cancellation Test (Kapoor, 1972) for children of Classes III and IX to assess attention for simple and complex tasks. In both the colour and number cancellation tests, cancellation of a single colour or number and the complex task requires the cancellation of two colours or numbers. (ii) Seguin Form Board (SFB) (Cattel, 1948) for children in Classes I and II to assess intelligence. (iii) Raven’s Progressive Matrices: Standard and coloured versions (Raven, 1956, 1965), Coloured Matrices for children in Classes III to VII and standard matrices for children in Classes VIII and IX. (iv) Vocabulary Test (NIMHANS 1982) to assess vocabulary as an index of language development. (v) Arithmetic Test (Modified Schoenell Test, NIMHANS 1982) for children in Classes III to IX to assess arithmetic skills in base computations of simple and complex nature, for example, addition, subtraction, division, multiplication, and the same for fractions. (vi) Non-Verbal Test of Creative Thinking (Mehdi’s Indian modification of Kagan and Wallach’s test of creativity 1985) for children in Class III–IX using 10 of the 20 designs.
PROCEDURE The permission of the Block Education Officer was taken, the teachers met and the programme was explained to as their cooperation was needed to ensure that an hour a day was allotted for the programme. In each of the 15 schools, the two field workers who had backgrounds in education/psychology conducted the intervention programme in a staggered manner. The programme was conducted for groups of 10–12 children from Class I to Class IX over a period between August 2000 and April 2003. The process of conducting the programme was interrupted by unscheduled school holidays, teacher absences and various logistic problems related to travelling long distances in rural areas from the schools. However, the fact that children enjoyed the programme immensely made up for all the frustrations faced. Children in Class I and II were administered only the Color Cancellation Tests and the Seguin Form Board, as they could not comprehend instructions for the other tests.
372 Malavika Kapur The Number Cancellation Test, Raven’s Progressive Matrices, vocabulary, arithmetic and creativity tests were administered to the children in Class III to IX. All the children were administered these tests before and after the intervention. In addition, a sub-sample was taken as a control group to examine whether the outcome was related to the intervention or due to the practice effect or normal course of development. These children had two ‘pre’ assessments with a gap of three months between the two, without any intervention. A third assessment was carried out after they were exposed to the intervention similar to the rest of the sample. This was carried out of the ethical consideration of not denying any child the advantage of the programme.
Description of the Stimulation Package Day 1: Activities to promote gross and fine motor skills, eye-hand coordination and form perception through play or games, indigenous or otherwise. Several local games were identified and encouraged. These were video recorded for future use. Day 2: Activities to promote linguistic skills, through word games, story telling, story building and reading, adopting the ‘top down’ approach of starting from the story and pictures, rather than the ‘bottom up’ approach of starting with letters. For example, a simple illustrated story book describing a village, using words of graded difficulty. The children were expected to track the sentences as the reader (project staff or older children) read the text aloud using enlarged pages which could be seen by the entire class. They were asked to repeat aloud what they had heard. The first goal was to see, then listen and to learn by rote after 10–15 repeated sessions of listening to the entire story. They were also asked questions about the content of the story to test their comprehension and also to enact it as a play. Once they learnt the story by rote, even a non-reader could pretend to read. This gradually led to letter recognition and grasping the basics of reading the whole sentence in about two months. The children would demand new story books and started reading by themselves. This was termed the ‘top down’ method. The conventional teaching consisted of teaching to first write and copy the numerous letters in Kannada, a phonetic language, and eventually go on to words and sentences. This is termed as the ‘bottom up’ approach. Even the fifth graders were very poor readers with the use of this conventional method. Day 3: Activities to promote language and creativity, including word and song games, story enactment, skits, dramas and dances, creative or traditional. Day 4: To promote number concepts by number recognition, identification through games, board games and Chinese abascus. Day 5: Activities to promote intellectual functioning through attention enhancement, memory games, problem solving, analysis, synthesis and planning were incorporated into play.
Cognitive Stimulation of Rural School Children in India: An Evaluative Study
373
Day 6: To promote creativity, child initiated free art and craft activities were carried out using inexpensive/indigenous material. The emphasis was to counter the normal teaching practice of the entire group, which was usually asked to copy what was drawn or made by the teacher. The children were encouraged to independently create the art or craft from their imagination. Though a well-recognized practice in the ideal educational practice, it is hardly ever practised in most Indian schools. The aforementioned programme was carried out with all the children in small groups in a staggered manner, over a period of three years. The programme was carried out over a day for six days a week for two–three months, consisting of approximately 20–25 sessions per group.
RESULTS AND DISCUSSION Twelve hundred children were exposed to the programme, and 1,088 had both the preand the post-assessment. The children belonged to the lower strata of society, as the more affluent families sent their children to private/convent schools nearby or in the town. The data were analysed separately for classes I and II, III to VII, and VIII and IX, as recommended for the various tests. The impact of the intervention was examined through the statistical tests of ‘t’ statistic and ‘effect size’ (Cohen, 1988).’Effect size’ is especially recommended to examine the significance of differences as a result of intervention (BPS 2003). Even when the differences are statistically significant on ‘t’ statistic, ‘small’, ‘medium’ and’ large’ effect sizes lend stronger support to the claim of efficacy of the intervention. The results will highlight the following: 1. 2. 3. 4.
Sample description Impact of the programme Age trends Gender trends
Sample Description Logistic difficulties caused the dropouts either before or after the intervention.The number of the schools in the study was 15. The children were studying in Classes I–IX. The children in Class X were not permitted to participate in the programme as they were due to face the Class X public examination. In these schools, the medium of instruction was Kannada with English and Hindi as other languages in the higher classes. There were 517 boys and 571 girls. The gender distribution was somewhat similar in the population studied. The distribution by class in percentages was as follows: Class I (7.44 per cent); Class II (9.83 per cent); Classes III to VIII (between 10 and 17 per cent); and Class 9 had only 3.22 per cent of students.
374 Malavika Kapur Impact of the Programme Comparison of results ‘before’ and ‘after’ the intervention. The results are being described separately for the three groups. • Classes I and II • Classes III to VII • Classes VIII and IX Classes I and II (age: six to eight years; N = 188) The results indicate that the programme has been successful in enhancing attention as shown by the measures of speed and accuracy. Classes III to VII (age: eight to 12 years; N = 753) The number of children for each of the tests ranged between 606 and 632 with the exception of creativity test, where it was 355. On the tests of attention, intellectual functioning, arithmetic, vocabulary and creativity, their performance improved significantly after the intervention. Classes VIII and IX (age: 12 to 16 years; N = 147, with 112 in Class 8 and 35 in Class 9) Children in eighth standard showed significant improvement on attention, intellectual functioning, arithmetic, vocabulary and creativity. Performance of Class IX with only 35 children revealed significant improvement on the test of intelligence and creativity. They showed a decline in attention (simple tasks). It is difficult to comment on the performance of children in ninth grade as they were not cooperative and sincere enough while performing the tests.
Age Trends The children in Class II showed more improvement than Class I children. Compared to the children from third to the seventh standard for the children in the eighth standard, the improvement occurred to a larger extent with the exception of arithmetic and vocabulary. This may be due to the fact that the tests were too easy for the eighth and ninth standard children. The age trends show that the baseline performance gradually improved as seen on the pre-assessment on all the tests. This also supports the contention that the tests were sensitive to age related changes. When each class was examined separately, Class V showed less improvement on intellectual functioning than the others. Class VII showed maximum improvement in vocabulary and creativity compared to the others.
Cognitive Stimulation of Rural School Children in India: An Evaluative Study
375
A small sub-sample of 200 children with children in Classes I to VIII were taken to examine 1. 2. 3. 4.
Whether improvement was due to the programme Whether it was due to the practice effect of taking the tests twice Whether it was due to the time lapse of two to three months Whether it was due to the normal teaching practices in these schools.
This was done by repeating the tests after three months without the intervention. It was found that there was no difference in the test scores. This clearly demonstrates that the improvement in the children was entirely due to our intervention. For ethical reasons, these children were given the intervention, and the third set of test scores showed improvement that was similar to the original sample. This successfully countered the argument the teachers offered that they were already familiar with the methods of joyful learning.
Gender Trends Boys and girls in Classes III to VII showed a similar rate of improvement with an exception on creativity scores. Boys scored higher on the base line assessment of creativity, while girls showed greater improvement on the post-assessment following the intervention, indicating that the girls benefited more by the activities. Boys in the eighth standard improved on all the tests, but to a lesser extent on attention (complex) arithmetic and vocabulary, compared to creativity and simple attentional tasks. However, girls showed significant improvement to a larger extent on intellectual functioning and attention (simple and complex). Older girls gained on tests of intelligence, attention and creativity after the intervention, but not on academic skills. The analysis by gender was not carried out for the ninth standard as it was a very small group. To sum up, the study indicates that the intervention was generally effective, though in varying degrees across tests, age and gender. The study demonstrates that promotion of psychosocial development can be effectively carried out in the rural schools. There is, however, a scope to improve the programmes by attending to the specific needs of very young children, and adolescents and school age children. The tasks should be sensitive to genderrelated differences as they respond differently to the effects of cognitive intervention.
Intellectual Functioning The baseline and post-intervention percentile scores of the children in Classes III to VII were compared to the percentile norms developed by Barnabas et al. (1995) for the urban
376 Malavika Kapur population. School children in the age range of 7 to 11 years studying in English medium were taken from schools in Bengaluru city. The following profiles emerged: Class III Class IV Class V Class VI Class VII
changed from 5th to 10th percentile changed from 5th to 10th percentile changed from 10th to 15th percentile changed from 30th to 40th percentile changed from 40th to 55th percentile
The above results indicate that the performance of the children on Raven's Progressive Matrices (RPM) before and after the intervention is far below the average. In Class V, it is marginally better. However, in Classes VI and VII, it lies in the average range. It is, of course, not fair to compare the performance of children of urban schools to that of rural ones. If RPM is culture-free, no child with average milestones of development can fall in the defective range. If intelligence is stable, no intervention can produce positive changes over such a short period of time using simple child-friendly activities, which move the performance to a higher percentile range. However, the fact remains that the performance changed for the better in the region of 5 to 15 percentiles, with the extent being largest in the oldest children in the group. The boys have higher RPM scores, show more improvement and also are better in arithmetic, while the girls are better in vocabulary. The boys have a better baseline in creativity, while the girls show more improvement. The girls in Class VIII have a lower baseline in RPM, but eventually catch up with the boys. The performance on attentional tasks is superior in the girls, but they lag on arithmetic, vocabulary and creativity. The reasons for these variations need to be speculated on and validated in future studies. The school atmosphere and differential child-rearing practices may lead to the observed gender differences. Attention improves with age and stabilizes earlier in the girls when compared to boys. Older girls show less improvement in academic skills. This may be due to differential attitude to the academic performance of the girls by the parents, teachers and the community. Age effects are predictable, but gender effects are not, and this needs further exploration.
CONCLUSIONS The intervention programme has been found to be effective in providing cognitive stimulation, leading to enhanced performance on attention, intelligence, arithmetic, vocabulary and creativity. There has been a normal age trend of an increase in the baseline scores on all the measured parameters. Children in higher classes than those in Classes I and II benefitted more from the intervention. There are gender variations which elude explanations in the present stage of inquiry.
Cognitive Stimulation of Rural School Children in India: An Evaluative Study
377
Implications The intervention programme is the first ever evaluative study of an effective universal intervention at the school level in the rural areas of India to enhance cognitive functions like attention, intelligence, arithmetic, vocabulary and creativity. The present programme can be used across the nation and in other developing countries as it is a low-cost, childcentred programme which can be integrated into a school system with poor infrastructure, and can be easily adapted to fit into the regular school routine. Child-friendly methods of the programme enhance child participation. It places lesser burden on the teachers and renders their cooperation.
ACKNOWLEDGEMENTS The author is grateful to the National Council of Rural Institutes (Human Resource Development Ministry, Government of India) for funding the project. Help and support rendered by Dr Uma, Head of the Department of Clinical Psychology, and Mr Suresh K.P. of the Department of Bio-statistics at the National Institute of Mental Health and Neurosciences, Bengaluru is gratefully acknowledged. A deep debt of gratitude is owed to Ms A. Geetha, a doctoral scholar at the Bangalore University, who bore the burden of carrying out the project with immense commitment.
REFERENCES Barnabas, I.P., M. Kapur and S. Rao. 1995. ‘Norm Development and Reliability of Colored Progressive Matrices’. Journal of Personality and Clinical Studies, 11: 1–2, 17–22. British Psychological Society. 2003. ‘In Psychologist’. Special Issue on Statistics 16: 631–46. Cattel, R.B. 1948. A Guide to Mental Testing. London: London University Press. Cohen, J. 1988. Statistical Power Analysis for the Behavioral Sciences (2nd edition). Hillsdale, New Jersey: Lawrence Erlebaum. Kapoor, S.D. 1972. Speed and Accuracy Test (SAT, LCT, MFT). New Delhi: Psycho-Centre. Kapur, M. 1975. Measurement of Organic Dysfunction: Color Cancellation Test. Unpublished Doctoral Thesis, Bangalore: Bangalore University. Kumar, M and P.M. Sarangapani (eds). 2005. Improving the Government Schools Books for Change (Action Aid). Bangalore: Books for Change. Mehdi, B. 1985. Non-Verbal Test of Creative Thinking (2nd edition), Agra Psychological Corporation. Agra: National Psychological Corporation. Ramachandran, V. (ed.). 2003. Getting Children Back to School. Case Studies in Primary Education. New Delhi: Sage Publications. Raven, J.C. 1956. The Standard Progressive Matrices Test and Manual. London: H.K. Lewis and Co. Ltd. ———. 1965. Colored Progressive Matrices, Test and Manual. London: H.K. Lewis and Co. Ltd.
SECTION
VI
Consciousness
S
o far, the sections have focussed on empirical and theoretical aspects of cognitive science and cognitive processes like attention, language and cognitive development. An important goal of cognitive science is to understand one of the deepest mysteries of human life and cognition, namely, consciousness. There has been a great surge of books on consciousness in the past decade, denoting the increasing significance of the issues related to consciousness and its increasing acceptance as a valid topic for scientific study (Blackmore, 2004). What makes the problem of consciousness such a mystery? The difficulty of explaining consciousness is to bridge the ‘explanatory gap’ between subjective experience and the objective world. The problem of consciousness has been phrased in terms of the ‘hard problem’ and the ‘soft problem’ (Chalmers, 1995). The ‘hard problem’ is to understand ‘what is it to like to be something’, and we are still not close to understanding it. Responses to the hard problem range from concluding that the problem is unsolvable, to arguing that it does not exist. Most scientists focus on the easy problems of attention, perception, and so on, and hope that this will result in eventually solving the ‘hard problem’ of consciousness. Does consciousness cause actions or events? Most research on skilled movements shows that we can make movements without being fully conscious of them. Dissociation between visuomotor control and visual perception has been linked to the presence of two different pathways in the visual system (Milner and Goodale, 1995). Theories of consciousness differ in terms of attributing causal roles of consciousness. Baar’s global workspace theory is an example of a theory where consciousness plays a causal role in global broadcasting, coordination and control (Baars, 1993). William James talked about consciousness as a ‘stream’. Questions have been raised whether the stream of consciousness, which appears rich and continuous, is an illusion. Change blindness and inattentional blindness studies question the idea that visual consciousness is a continuous stream of rich representations (O’Regan, 2003; Rensink, 2002). Separate from academic psychology, work has also been done on paranormal phenomena related to consciousness. The field of parapsychology has investigated phenomena like extrasensory perception and psychokinesis. At this point in time, doubts exist regarding the validity of these phenomena, and there is no clear-cut objective scientific evidence for such phenomena. In addition to the normal consciousness as we experience it, people have experienced and have reported other altered states of consciousness. These include altered states induced by drugs, sleep and dreaming, and meditation. In studying consciousness, there is a need to distinguish between first-person and third-person approaches. It is important to distinguish between first-person versus third-person science, as well as firstperson versus third-person methodology. While it is not clear that there is a first-person science, first-person methodology will be useful in studying consciousness. Some first person approaches include phenomenology, neurophenomenology (Varela, 1996) and heterophenomenology. Another first-person approach is to look inward using techniques like meditation. It may also be important to explore the implications of Buddhist ideas of illusion and no-self for psychology and consciousness.
382 Advances in Cognitive Science The contribution in the final section addresses the important topic of consciousness from an Indian perspective, and provides an important addition to the already existing body of work on consciousness. The chapter by K. Ramakrishna Rao discusses in detail all the varieties and aspects of conscious experience. Consciousness is a much used and no less abused word. Professor Rao discusses the various meanings of consciousness and the multitude ways in which the word is used in western studies of consciousness. The purpose of this is to arrive at a descriptive phenomenology of states of consciousness under various possible categories. Indian approaches to consciousness include those associated with Vedanta as well as Buddhism. Professor Rao discusses the classification of states of consciousness in Vedanta tradition and the extensive taxonomy of consciousness found in Buddhist literature, especially in Abhidhamma. He concludes in favour of eastern approaches, which appear to approach consciousness from a broad perspective. The approaches based on Indian traditions are proposed to have advantages in providing alternate models. The eastern approach may solve some of the issues that have reached a stalemate and provide a better alternative to reductionist attempts to solve the problem of consciousness.
REFERENCES Baars, B.J. 1993. ‘How Does A Serial, Integrated, and Very Limited Stream of Consciousness Emerge From A Nervous System that is Mostly Unconscious, Distributed, Parallel, and of Enormous Capacity?’ In G.R. Bock and J. March (eds), Experimental and Theoretical Studies of Consciousness (pp. 282–303). Chicester, UK: CIBA Foundation Symposium 174. Blackmore, S. 2004. Consciousness: An Introduction. New York: Oxford University Press. Chalmers, D. 1995. The Conscious Mind: in Search of A Fundamental Theory. Oxford: Oxford University Press. Milner, M.A. and A.M. Goodale. 1995. The Visual Brain in Action. Oxford: Oxford University Press. O’Regan, J.K. 2003. ‘Change Blindness’. In L. Nadel (ed.), Encyclopedia of Cognitive Science, Vol. 1 (pp. 486–90). London: Nature Publishing Group. Rensink, R.A. 2002. ‘Change Detection’. Annual Review of Psychology, 53: 245–77. Varela, F.J. 1996. ‘Neurophenomenology: A Methodological Remedy for the Hard Problem’. Journal of Consciousness Studies, 3: 330–49.
Chapter 27 Taxonomy of Consciousness K. Ramakrishna Rao
INTRODUCTION
W
hat is consciousness? How may we study it? These are questions that engaged the minds of many scholars/scientists for millennia without yielding unequivocal answers. The debate on and discussion of consciousness is often clouded by the conflation of a variety of its meanings, leading to multiple ontologies, diverse epistemological and sometimes conflicting axiological implications. ‘Consciousness’ has been a theoretical quagmire, intellectually challenging, but scientifically unsettling and often considered irrelevant. In psychology it has had a checkered history, postulated sometimes as the primary and essential feature of human condition, and considered at other times as a useless appendage that is better denied or ignored rather than studied. It would seem, therefore, that a systematic study of consciousness should begin with the clarification of the multiple connotations it enjoys, and their relevance to a given discussion of consciousness. A meaningful taxonomy of consciousness is therefore a necessary prelude to productive inquiry into consciousness. This, in turn, requires an exercise not only in phenomenology, but also in the semantics of consciousness. This chapter is thus an exercise in the semantics of consciousness, and an attempt at arriving at a plausible taxonomy of conscious states. Despite the multiple meanings consciousness enjoys in psychological discourse, awareness is ubiquitous in most senses it is used (Natsoulas, 1978). However, consciousness as awareness falls into two broad avenues or categories. First, consciousness denotes a state of awareness; second awareness is noetic; it is of or about something, real or imaginary. As a state, awareness is a subjective condition, which may or may not have identifiable physiological factors associated with it. In its noetic aspect awareness refers to knowledge, awareness of something. In ordinary states of awareness, when one thinks, remembers, plans, perceives, and so on, the person has something to which thoughts, memories, and so on, are directed. This aspect of consciousness, inasmuch as it contains an objective
384 K. Ramakrishna Rao informational component, may be considered to be the objective aspect of consciousness, whereas consciousness as an experienced state refers to its subjective aspect. Thus, we may ask of consciousness, ‘what is it about?’ (objective aspect) or ‘what is it like?’ (subjective aspect) to be conscious (Figure 27.1). FIGURE 27.1
Consciousness as awareness
STATES OF AWARENESS Psychological literature and physiological data suggest the existence of multiple states of consciousness. These include normal waking states and others like drug or hypnosis induced or dream-driven consciousness, each with its own distinctive phenomenological, and in some cases unique physiological, characteristics. A state of consciousness appears to make a difference to one’s experience. While it is arguable whether a state of pure consciousness, that is, consciousness-as-such without any content, is possible, it is entirely unproblematic to assume that consciousness manifests in different forms and that awareness in these states has distinguishable features (Figure 27.2). FIGURE 27.2
States of awareness
WAKING STATES Waking states of consciousness may be further distinguished into five categories (Figure 27.3). During the normal waking condition, there are events in one’s focal attention of which the person is subjectively aware, and to which one has introspective access. We may call this kind of awareness primary awareness. Then, there are things in the fringe of awareness to which the person is not attending at the time, but are in the periphery
Taxonomy of Consciousness
385
influencing one’s awareness, which may be referred to as peripheral awareness. At times a person may not be subjectively aware of things and events that she in a significant sense knows. These are cases of not knowing that one knows. This is paradoxical awareness. Then there are the occasions when what one perceives to be the case is not really so. For example, one may hallucinate and see things that are not actually there. Such dysfunctional awareness may be called pathological awareness. Also, on rare occasions, people report that they have veridical awareness of something taking place at a distance or in the future, information to which they have no sensory access, as, for example, in the case of an ESP experience. This is considered paranormal awareness, in that its manifestation in one’s experience has no normal explanation. Apparently, there is no conceivable sensory flow connecting the target event or object with the subject, as is generally required for one to have awareness. Thus in the waking state we may distinguish between five forms of awareness. All these forms may manifest in waking states, as well as in some other states of consciousness such as dreams. FIGURE 27.3
Waking states of consciousness
PRIMARY AWARENESS Primary awareness is involved in one’s acts of perception, feeling and volition. Included in this category are percepts, images, inner speech and feelings such as pain, all of which have sensory qualities, but also abstract concepts, beliefs, intentions and expectations. The defining characteristic of primary awareness is that the information is in the focus of the person, and that the person has subjective (introspective) awareness of it. William James (1890/1952) attributes five primary characteristics to consciousness, which seem to belong appropriately to primary awareness in the waking state. According to James: 1. Consciousness is subjective; it has ‘warmth and intimacy’. As James puts it: ‘Absolute insulation, irreducible pluralism, is the law. It seems as if the elementary psychic fact were not thought or this thought or that thought, but my thought, every thought being owned’ (1890/1952, p. 147, emphasis original). 2. Consciousness is in constant change. No conscious ‘state once gone can recur and be identical with what it was before’ (James, 1890/1952: 149).
386 K. Ramakrishna Rao 3. Consciousness is sensibly continuous; it flows like a stream. It is ‘for itself unbroken. It feels unbroken’(p. 154). 4. Consciousness has the function of knowing; it is noetic. 5. Consciousness involves selective attention. It is the result of constant choosing, of discrimination and organization. Thus, at the primary level, consciousness appears to have two components, subjective and objective. The subjective character of primary awareness makes it person-centred. Only the experiencing person knows what it is like to have that experience. That knowledge may be communicated through verbal reports or other behavioural manifestations. Such communications, however, are at best descriptive translations, and are never the same as the actual experience itself. Whether the information obtained through such translations is sufficient to understand and explain the nature of consciousness is arguable. The answer rests on the weight one is prepared to accord to subjectivity as an essential ingredient of consciousness, whether it is regarded as central or extraneous to consciousness. Unlike the subjective factor, the objective information involved in the cognitive function of consciousness and the process of attention are available to third-person observation. Consequently, they can be studied behaviourally and neurologically. Again, whether such studies, which do not take into account the subjectivity factors of the experience, would ever tell the whole story about consciousness is debatable. Also, in traditions that believe in the possibility of pure conscious states, one may raise the question of whether the content of consciousness is indeed central consciousness.
PERIPHERAL AWARENESS There is a vast surround, a backdrop that is embedded in primary awareness, which gives meaning to the objects and events of awareness that are in focus. William James, for example, speaks of a fringe and focus of awareness. While a person’s attention is usually focussed on a centre and around a theme, there is also a vast surround of impressions and sensations at the periphery. The fringe is the transitive, floating backdrop of substantive states that are the focus of attention and the subject of primary awareness. In the metaphor of the stream, every definite image in the mind, to quote James again, ‘is steeped and dyed in the free water that flows round it. With it goes the sense of its relations, near and remote, the dying echo of when it came to us, the dawning sense of whither it is to lead. The significance, the value, of the image is all in this halo or penumbra that surrounds and escorts it’ (1890/1952: 165–66). The fringe is not the same as the unconscious. It is what provides the context and the network of past experiences and future expectations that give meaning to the contents of present awareness. It often accounts for individual differences in perceptions. Illustrative
Taxonomy of Consciousness
387
of the role of fringe or peripheral awareness are feelings of familiarity and affinity and of being right or wrong.
PARADOXICAL AWARENESS Awareness that lacks subjective feel and is not accessible to introspection is paradoxical awareness. It is paradoxical inasmuch as it is ‘knowing’ without knowing or ‘awareness’ without awareness. There are compelling reasons and massive data for inferring awareness when, for example, a subject in a subliminal perception experiment systematically responds to stimuli while denying them any awareness, or when a hypnotized person carries out a post-hypnotic suggestion apparently without any subjective awareness of the suggestion while carrying it out. Paradoxical awareness may occur in normal and healthy subjects under certain circumstances as in the case of subliminal perception, or it may be the result of injury or insult to the brain as in the case of blindsight and prosopagnosia. Subliminal perception refers to the possibility that people might be influenced by objects and events of which they report to have no perceptual awareness. Research in the area of subliminal perception has produced impressive evidence that stimuli presented at low energy levels, or in masked forms so that they cannot be recalled, recognized, or discriminated from other similarly presented stimuli, can nevertheless produce detectable cognitive, affective and other behavioural effects (Bornstein and Pittman, 1992). It has been observed that some cortically blind people, because of damage to primary visual areas, can respond discriminatively to different visual stimuli, while reporting no subjective awareness of the stimuli. For example, in one of the cases studied by Weiskrantz (1986), the patient D.B. was able to neither name nor describe any object that was presented to his damaged left visual field. However, when asked to guess in an experiment whether the stimulus was ‘X’ or ‘O’ when ‘X’ and ‘O’ were presented randomly, he guessed correctly 27 times in a total of 30 trials. Similarly, patients suffering from prosopagnosia, a kind of memory disorder following brain damage, who report no sense of familiarity for faces, including those of their own family members, provide experimental evidence for covert evidence of familiarity (Bauer, 1984).
PATHOLOGICAL AWARENESS Pathological awareness is dysfunctional awareness. Awareness primarily serves the function of relating us to the world around us by faithfully representing the world in concrete or abstract forms, in actual or imaginary situations. However, awareness may on occasion misrepresent the world, and lead us to indulge in behaviour inappropriate to the occasion. Hallucinatory awareness and other forms of cognitive abnormalities are
388 K. Ramakrishna Rao examples of pathological awareness. It is common to find among schizophrenic patients hallucinations (usually in the form of hearing voices), delusions (often involving claims that their thoughts are controlled by others) and thought disorders (seen in fluent but abnormal speech that makes little sense). Similarly, in cases of clouding of consciousness such as amentia, a subacute delirious state in which the patient engages in haphazard associations and random performance, there is evidence of pathological awareness. Unlike primary awareness, in which consciousness appears to be unified, coherent and continuous, has self-reference and is adaptive, in cases of pathological awareness, it tends to lack such unity and continuity, and becomes incoherent, fragmented and clouded. It may even be dissociated as in the case of multiple personality disorder, and become severely maladaptive as in cases of schizophrenia.
PARANORMAL AWARENESS There is substantial experimental evidence to suggest that it is possible to have information shielded from the senses such as Extra-Sensory Perception (ESP) (Rao and Palmer, 1987). In the case of paranormal awareness, knowing is non-conventional. Unlike in other forms of awareness, there does not appear to be an information flow from the object to the subject. Paranormal awareness in a sense extends the range of awareness beyond sensory and representational forms. It may manifest itself in a form similar to primary, paradoxical or peripheral awareness. On occasions it may even be pathological. Veridical intuitions are very much like ordinary thoughts, except that they contain information that is in principle inaccessible to normal cognitive activity. Often paranormal awareness is implicit, but it is different from paradoxical awareness, because there is no identifiable channel or link, connecting the person and the object of awareness. Again, a paranormal experience may occur in the form of a hallucination or in a dream. However, unlike pathological hallucination, paranormal awareness conveys valid information. In cases of psi-missing, it would seem that the awareness could be dysfunctional and pathological. Paranormal awareness may extend from simple ESP kind of experiences to profound transformation of persons following such awareness. At the extreme end we have pure conscious experiences, where there appears to be awareness without any content. If consciousness is both a state of awareness and an informational state, and if information and awareness can be dissociated as in the case of paradoxical awareness, it should be theoretically possible to have a state of awareness without content. Such states tend to be ineffable inasmuch as they are transcognitive. It is more appropriate perhaps to include them in the category of pure conscious states rather than treat them as paranormal awareness. Paranormal awareness may be limited to ESP and other intuitive information states.
Taxonomy of Consciousness
389
ALTERED STATES OF CONSCIOUSNESS A pure state of consciousness in which the person has awareness with no content is usually one that is believed to be associated with a state of consciousness that is qualitatively different from the normal waking state, such as the one obtained in a deep state of meditation. Also, hypnosis appears to induce an altered state. Thus some of the phenomena of consciousness referred to as occurring in waking states may indeed be contained in an altered state, even though the person may be in a physiologically wakeful state. An altered state of consciousness is one that is significantly different from the normal waking state in several respects (Tart, 1969). The first and foremost difference is indicated by alterations in one’s perception of self and reality consequent to changes in attention and processes of thought. In an altered state of consciousness, perceptual distortions may occur; one’s sense of time may be altered; and changes in body image and emotional expressions are possible. One may have a transcognitive experience, which is essentially ineffable and beyond words. Some of the altered states are naturally occurring. Others may be induced by psychological or spiritual techniques as in hypnosis and meditation, ingestion of drugs as in the use of marijuana, or by physiological changes as in sensory deprivation. Again, altered states of consciousness may be maladaptive and lead to pathological states, or they can be profoundly transformative and beneficial in therapeutic and learning situations. Here are some of the well-recognized altered states of consciousness (Figure 27.4). FIGURE 27.4
Altered states of consciousness Altered states of consciousness
Dreaming
Hypnagogic imagery
Hypnotic states
Drug induced imagery
Sensory isolation
Meditation
Multiple personality
DREAMING Dreaming is a naturally occurring and universally experienced nocturnal state. It is a rhythmic, cyclic and physiologically identifiable activity that occurs intermittently throughout the night, beginning with brief episodes during the early part of the night and culminating in prolonged periods lasting for 30 minutes or more towards the end of the sleep cycle. The physiological indices of dreaming include rapid eye movements, increased muscle tonus, irregular breathing, increased blood flow in the brain and electroencephalogram (EEG) activity dissimilar to deep sleep states (Dement, 1958). Dreaming is considered an altered state of consciousness because in dreams the way we think and experience appears to be markedly different from awareness in waking states.
390 K. Ramakrishna Rao When we dream in sleep, we are conscious in the sense of being subjectively aware. As in waking awareness, in dreams also we subjectively experience events and objects in the form of vivid images. The imagery can be as complex and organized as in the waking state. However, a dream may be very different from waking experience, bizarre and incoherent. Dreams often tell a story. However, contradictory things may appear in quick succession without creating any dissonance. Scenes and characters may shift and change in unnatural ways. In a sense, dream experience is translogical. Rules that govern rational thought are often relaxed in dream mentation. Dreams are quasi-perceptual phenomena in that the dreaming person has subjective experience of perceiving an object or event when in fact there is no such object or event other than the image itself. Dream experience fades out rather quickly unless rehearsed soon after its occurrence. Dreams are regarded as representing a distinct, naturally occurring altered state of consciousness, because the dream experience is organized differently from the normal waking awareness. Perceptual distortions are not uncommon. The sense of time is often altered. Logic and reason, coherence and continuity that generally mark waking awareness tend to be missing or distorted.
DAYDREAMING AND HYPNAGOGIC IMAGERY Between waking and dream are daydreaming and hypnagogic states, which appear to be distinct awareness states. One’s thoughts may be stimulus-related and oriented to the task on hand. Or they may be spontaneous, involuntary, directionless and stimulus independent. Daydreaming is defined as stimulus independent mentation (Singer, 1975), as distinguished from normal waking mental activity in which attention is directed at stimuli in the environment. The Freudian distinction between primary and secondary process thinking points to a somewhat similar categorization. Primary process mentation is considered irrational and not reality oriented. It occurs in the id and is therefore unconscious and directed towards gratification of instinctual urges and drives. The secondary processes, however, are ego-driven, conscious and volitionally guided, generally rational and reality oriented. Daydreaming, like dreaming, is a normal activity. Personal preoccupations and stress-related factors are known to influence the content of day dreams, which, apart from being wish-fulfilling fantasies, serve a variety of purposes such as goal-setting and strategy planning. Similar to dreams is the physiologically identifiable state called the hypnagogic state. It occurs just before a person falls asleep. It is a state characterized by vivid and sometimes distorted imagery of hallucinatory quality in which the experiencing person is usually a participant. There is, however, one important difference between dreams and hypnagogic hallucinations. The images in the hypnagogic state tend to be like still pictures, unlike the dream imagery, which is dynamic (Foulkes and Vogel, 1965). If the person in the
Taxonomy of Consciousness
391
hypnagogic state is awakened and asked about the depth of her sleep, she is likely to say that she had been merely drowsy, whereas a subject awakened while dreaming tends to report that she had been in deep sleep (Foulkes et al., 1966).
HYPNOSIS Among the psychologically induced alterations of consciousness, hypnosis is prominent. In a deeply hypnotic state, a person’s self-perception and experience of reality may be significantly altered following a hypnotic suggestion. Through hypnotic suggestion it is possible to induce perceptual distortions, positive and negative hallucinations, and analgesia. Susceptibility to hypnosis appears to be highly correlated with one’s ability to involve in fantasies. Highly susceptible subjects express great interest in the ‘savoring of sensory experiences, such as esthetic enjoyment of nature … an involvement in either reading or drama or both’ (Hilgard, 1977: 172). It is estimated that about 10 to 15 per cent of the population who are highly hypnotizable may be rendered insensitive to painful stimulus, such as an electric shock by hypnotic suggestion (Hilgard and Hilgard, 1983). Also striking perceptual distortions such as ‘seeing’ things that are not visually present and not seeing objects that are actually there are also observed with hypnotized subjects (Kihlstrom, 1979, 1984). Such ‘cognitive control’ of reality is sharply in contrast to normal waking state.
MEDITATION Today, there are many forms of meditation practised around the world. Thanks to the efforts of Maharishi Mahesh Yogi and his organizing network, Transcendental Meditation (TM) is a worldwide phenomenon practised by millions of people. The roots of most of the meditational practices are in the classical Hindu and Buddhist writings such as Patanjali’s Yoga–Sutra and Buddhaghosa’s Visuddhimagga. Daniel Goleman (1978) surveyed 14 different techniques of meditation currently in vogue, and argued that all of them employ essentially one of the two methods described in Visuddhimagga. These are concentration and mindfulness. Concentration involves focussing attention on an object, whereas in mindfulness attention is focussed on bare sense impressions. Both the techniques involve, however, redeployment of attention. ‘The need for the meditator to retrain his attention, whether through concentration or mindfulness’, according to Goleman, ‘is the single invariant ingredient in the recipe for altering consciousness of every meditation system’ (1978, p. 111). A state of deep meditation is believed to have profound effects on the practising person. These include achieving transpersonal and transcognitive states, experience of
392 K. Ramakrishna Rao oneness and unity with the universe, and disappearance of subject–object duality. Also, meditation appears to result in far reaching transformation of the person. It is arguable, however, whether occasional and even daily practice of meditation like TM or relaxation response (Benson, 1975) for a few minutes would lead to an altered state of consciousness. Traditionally, meditation is a highly specialized, disciplined practice that stretches over many years under the supervision of acknowledged experts. The instant meditation of the TM type may have several beneficial effects. It is doubtful, however, that it results in the fourth state, which is qualitatively different from waking consciousness as claimed by the TM practitioners.
MULTIPLE PERSONALITY OR DISSOCIATIVE IDENTITY DISORDER (DID) In the normal waking condition, consciousness appears continuous and unified. It is subjective and has a single self-reference. In cases of multiple personality disorder or, its more recent rendering, dissociative identity disorder (DID), there are startling discontinuities and disunity in the stream of consciousness. Discontinuities and breaks in consciousness imply dissociation. Dissociative states are not necessarily pathological. They may be in some cases quite adaptive. In extreme forms, however, dissociation is clearly worrisome and dysfunctional. Janet (1907/1920) among others attributed personality disorders to dissociation and deep divisions of consciousness. According to Diagnostic and Statistical Manual of Mental Disorders (fourth edition), the following are the diagnostic criteria of DID: (a) two or more distinct identities exist in the person; (b) they recurrently control the behaviour; (c) the person is unable to recall important personal information too extensive to be attributed to forgetting; (d) the condition is not due to physiological effects of substance use or general medication. DID patients are often the victims of child abuse. In 75 per cent of cases, child personalities under the age of 12 manifest as alternate personalities. According to surveys, the average number of multiples among DID patients ranges between six and 16. In some instances, alternate personalities claim continued existence even when they are not in control of the subject’s body. There is evidence suggesting that an alternate personality may influence the patient’s behaviour when not in executive control.
PURE CONSCIOUS STATES As mentioned earlier, at the extreme end of paranormal awareness we have mystical and contentless awareness. In his Gifford lectures, William James speaks of ‘non-rational forms’ of consciousness that belong to this category. He wrote that
Taxonomy of Consciousness
393
Our normal waking consciousness, rational consciousness as we call it, is but one special type of consciousness, whilst all about it, parted from it from the flimsiest of screens, there lie potential forms of consciousness entirely different … No account of the universe in its totality can be final which leaves these other forms of consciousness quite disregarded (1902/1914, p. 298).
Notwithstanding such a strong endorsement from one of the greatest American thinkers and psychologists, these forms of consciousness are routinely ignored or denied in the western scientific tradition. To label anything mystical is to denigrate it as unworthy of scientific attention (Figure 27.5) FIGURE 27.5
Pure conscious states
MYSTIC AWARENESS James mentions four marks or criteria for distinguishing a mystical experience. First, mystical experiences are ineffable. They defy adequate expression in words. They are more like experiences of feeling than of intellect. Second, they possess a noetic quality. Unlike feelings, they are experiences of knowledge. They carry with them a conviction of certitude. ‘They are illuminations, revelations, full of significance and importance’ (James, 1902/1914: 330). Transience and passivity are the other two less sharply marked criteria of mystical states. In describing consciousness in the mystic state, James uses the metaphor of ‘mother-sea’ instead of the stream metaphor he employed while discussing the waking consciousness in his Principles of Psychology. Whereas the stream metaphor reflects the flow of waking consciousness, the ocean metaphor emphasizes the interconnectedness of individual consciousness at a deeper level with ‘cosmic consciousness.’ It is interesting to note, however, that James considers the mystic awareness noetic, as he does the waking consciousness. In other words, mystic consciousness also contains content even though it is in the form of illumination and revelations.
394 K. Ramakrishna Rao According to W.C. Stace (1960), an important scholar of mysticism, the central characteristic in which all fully developed mystical experiences agree, and which in the last analysis is definitive of them and serves to mark them off from other kinds of experience, is that they involve the apprehension of an ultimate non-sensuous unity in all things, a oneness or a one to which neither the sense nor the reason can penetrate. In other words, it entirely transcends our sensory-intellectual consciousness (emphasis original, p. 14).
Stace distinguishes between two forms of mysticism. He calls them ‘introvertive’ mysticism and ‘extrovertive’ mysticism. In the latter, the person experiences unity between the self and the world. In introvertive mysticism, the unity is contained in the experience of the self, with no awareness of the external world.
CONTENTLESS (NON-INTENTIONAL) CONSCIOUS STATES Another category of pure conscious states is the experience of awareness-as-such, a pure conscious experience. Robert Forman (1990) defines pure conscious experience as ‘a wakeful though contentless (non-intentional) consciousness’ (p. 8). A pure conscious or contentless state of consciousness is non-relational. It is objectless and non-referential awareness. The possibility of such non-intentional consciousness presupposes a dissociation between awareness-as-such and objects of awareness as in the case of paradoxical awareness, where a person has access to the information but not awareness of it. In the case of contentless or non-intentional consciousness, there appears to be no information but just awareness. As an example of a pure conscious event, Robert Forman (1990) quotes the account of a TM practitioner who describes how the boundary that ordinarily separates ‘individuality from unbounded pure consciousness began to dissolve’. Once he was able to let go of the veil of individuality, there is no longer ‘I perceiving’ or ‘I aware’. There is only that, there is nothing else there. In this state, the experiencer is not experiencing as he/she normally does. It is there ready to experience, but the function has ceased. There is no thought, there is no activity, and there is no experiencer, but the physiology after that state is incredible. It is like a power surge of complete purity (p. 28).
In the Yoga tradition, two types of samadhi are distinguished. They are samprjnata samadhi and asamprjnata samadhi. The latter is the final point in the journey of a yogin to achieve liberation (kaivalya). In asamprjnata samadhi one has the experience of pure consciousness. It is described as a state of non-relational consciousness that has no describable content. Such an objectless state of consciousness is believed to be the highest state in which self-realization takes place. It is a state of perfection, liberated from all the existential constraints that distort truth, condition our being and bring suffering. The Indian tradition boasts of a succession of saints who were able to achieve such states. That
Taxonomy of Consciousness
395
such states of pure consciousness or awareness-as-such exist is attested by the testimony of trustworthy persons. Also, it is an empirical statement that can be falsified because the claim of achieving pure states of consciousness is said to be associated with the behaviour of the person believed to achieve such states. While pure conscious experience is a subjective state, according to traditional accounts, it has important behavioural consequences involving profound personal transformation of the person having such experience. Abraham Maslow (1973) observed analogous, though not identical, transformations among what he called self-actualized persons, who had peak experiences. According to Maslow, self-actualized persons have ‘feelings of limitless horizons opening up to the vision, of the feeling of … great ecstasy and wonder and awe, the loss of placing in time and space with, finally, the conviction that something extremely important has happened, so that the subject to some extent is transformed and strengthened in his daily life by such experience’ (1973, p. 190, emphasis added). The loss of self or transcendence, intense concentration and self-forgetfulness also appear to characterize self-actualized persons.
VARIETIES OF AWARENESS As mentioned, awareness in normal conditions almost always has content. Awareness is an information state. In its most general form, it is information concerning the environment or the person, one’s bodily or mental states. Thus, we are aware when we experience pain, perceive an object, visualize an image, recall a memory or engage in an imaginary thought. Thus, the content of one’s awareness may refer to an external, physical object, or an inner feeling, image, thought or an act of volition. The content of consciousness may be real or imaginary. Again, what is taken to be real may turn out to be a false representation. The content of awareness may not relate to an object or image, but to the identity of the person, one’s sense of self-awareness. Awareness may also be referred simply to the sense of being aware, which generally accompanies awareness in a wakeful state. Also, an object of one’s awareness may be a distinct perception or a defused sensation with no clear cut identification (Figure 27.6). FIGURE 27.6
Content of awareness
396 K. Ramakrishna Rao The content of awareness may be seen to fall into three broad categories. It may be a cognition, as in perceptual awareness, conation as in a desire to act, or feeling as in an experience of pain or pleasure. In addition to perception, cognition may refer to memory, imagination, reasoning, doubt, and so on. These can be further sub-classified. For example, perception may be divided into visual, tactile, auditory, and so on. Also, further division of these modalities is possible. The Indian physician of antiquity, Caraka recognized 63 different ‘tastes’ and classified them under six categories. Conation is what is behind action. It may be voluntary or autonomous. Conation includes desires, aversion, resolution and volition. Pain and pleasure are the primary categories of feelings, which could be further classified.
AWARENESS OF AWARENESS Awareness of awareness is a common experience in normal waking states. Whether such awareness is inherent in awareness itself or whether it is a distinctive awareness is arguable. According to G.E. Moore (1922), there are two elements in every sensation. For example, when a person has the experience of blue, one element of his experience is the sensation of the blue object and the other element is the awareness of that object. While it is difficult to discern the element of awareness in one’s experience of sensation, Moore believed that if one looks ‘attentively enough’ she can ‘see’ the consciousness of the blue object. In other words, in this view, not only can the object be conceptually distinguished from the consciousness of it, but the latter can be directly observed as a distinctive element. The awareness of awareness then is not merely an inference derived from self-consciousness, but can be an object of awareness as well. Awareness of awareness is generally understood to mean that a person knows that he/she knows. Sometimes, such knowledge may not be available to introspection as in the case of paradoxical awareness as mentioned earlier. If dissociation between awareness and its awareness is conceived, then it implies that awareness may be a two-stage process. D.M. Armstrong (1984), for example, calls awareness of awareness ‘introspective awareness’, as distinguished from ‘perceptual awareness’. Introspective awareness, according to him, arises from scanning the neural representations in the brain following perception. Without such scanning, Armstrong contends that one can have no awareness of awareness. Sleepwalkers, for example, are unaware of what they are doing.
SELF-AWARENESS Whether self-awareness is implied in all awareness is debatable. For thinkers like Descartes (1911/1969), who identify consciousness with the subject, self-awareness accompanies all
Taxonomy of Consciousness
397
awareness because one is reflexive of the other. If one believes in an enduring self/mind that underlies all conscious experience, then it will be completely unproblematic to assume such reflexivity between awareness and self-awareness. Even if one does not subscribe to a view that posits an enduring and abiding self, it is possible to relate self-awareness to one’s self-concept. There is substantial evidence to suggest that the self-concept has a powerful influence on behaviour. In psychological literature, self-awareness is generally considered to be part of the basic structure of consciousness itself. In the developmental process, self-consciousness arises with the infant gaining the necessary linguistic skills to meaningfully use the first-person pronoun. In this view, infants lacking the necessary competence to use meaningfully the words ‘I’ and ‘me’ also do not have self-consciousness. J.L. Bermúdez (1998) argues, however, that such a notion involves circularity and is a paradox. He points out that one cannot have mastery of the first-person pronoun without having first-person thoughts. The latter are not possible without the use of first-person pronoun. Bermúdez presents some supporting empirical evidence and argues that the first-person content exists in the pre-linguistic infants, albeit in a non-conceptual form.
TAXONOMY OF CONSCIOUSNESS From the preceding brief description of the varieties of awareness, we can understand why the semantics of consciousness is so complex and complicated. Consciousness has many forms. It involves an intrinsic and unique subject–object dimension. It manifests in various shades and degrees. There are qualitative differences in what it conveys and how it is conveyed. Regrettably, in the western scientific tradition, consciousness is generally ignored, I believe, because of its complexity. In the few cases when consciousness is considered, no attention is usually paid to these fundamental features of consciousness narrated, and a very narrow and restricted meaning is attributed to it. The result is that a true and comprehensive conception of consciousness has eluded us. Many aspects of consciousness are studied and discussed without any reference to consciousness as such. Since the different aspects of consciousness are studied under different notations, it would seem that nothing is left to be studied as consciousness. When confronted with ‘hard problems’ like subjective, phenomenal experience or extraordinary human abilities, attempts are made either to deny or reduce them to other forms. In recent years, there is renewed interest in consciousness studies in the West. However, these discussions of consciousness are marked more by disagreements than any consensual agreements on even basic things such as its definition. Writers and researchers like Crick and Koch (1998) and Susan Greenfield (1998) felt it better to avoid a definition of consciousness because it is impossible to precisely define it. Even the widely-used stream metaphor and the continuity aspect of consciousness are severely censored and rejected by
398 K. Ramakrishna Rao people like Daniel Dennet (1991) and Susan Blackmore (2002). It would seem that these differences and difficulties cropped up largely because of the narrow and circumscribed conceptions of consciousness, and the failure to focus on the multiple forms in which consciousness manifests. There are of course exceptions and lone voices like those of F.W.H. Myers (1903/1915) and William James (1902). Myers is unequivocal in asserting that waking consciousness or supraliminal consciousness, as he called it, ‘does not comprise the whole of consciousness or of the faculty within us’. There exists a more comprehensive consciousness, a profounder faculty (1903/1915, p. 12), which he called subliminal or ultra-marginal consciousness. According to Myers, consciousness may be represented ‘as a linear spectrum whose red rays begin where voluntary muscular control and organic sensation begin, and whose violet rays fade away at the point at which man’s highest strain of thought or imagination merges in reverie and ecstasy’ (p. 18). As mentioned earlier, William James, while discussing religious and mystic experiences, did speak about non-rational forms of consciousness, their importance and role in our being. However, in his later writings such as Essays in Radical Empiricism, James began having second thoughts. ‘For twenty years past’, he wrote, ‘I have mistrusted “consciousness” as an entity; for seven or eight years past I have suggested its non-existence to my students and tried to give them its pragmatic equivalent in realities of experience. It seems to me that the hour is ripe for it to be openly and universally discarded’ (1912/1947, p. 3). Clearly, there was a conflict and dissonance between the openness that marked James’ attempts to deal with varieties of human experience, and the constraints placed on him by his embrace of western empiricism in his philosophical postulations. The story is altogether different in traditions that accord primacy and centrality to consciousness in human condition. In the Indian tradition, for example, we find not only a pervasive phenomenology of consciousness, but also systematic attempts to study them from different perspectives. These date back to the Vedic times, and are seen not only in orthodox Hindu systems, but found also in Buddhism as well as Jainism. The result is the emergence of a consummate discipline of the science of consciousness and the development of wholesome technologies to ‘cultivate’ consciousness.
STATES OF CONSCIOUSNESS IN HINDUISM The first division of consciousness in the Indian tradition is to categorize it into (a) the ordinary, normal and common states of the mind; and (b) the extraordinary, unusual and non-mundane states of higher consciousness. The ordinary, mundane states universally present in the human condition are (1) the waking state (jagrit), (2) dreaming state (svapna), and (3) state of deep sleep (susupti). As the Brihadaranyaka Upanishad says, the dream state is the intermediary between wakeful consciousness and unconsciousness in deep sleep.
Taxonomy of Consciousness
399
However, we are told, it would be incorrect to consider the deep sleep state as totally devoid of awareness. If it were so, waking up from sleep one cannot say ‘I had a sound sleep’. What the person is unconscious of in a state of deep sleep is implied in the unawareness of the external world mediated by the senses or the internally generated images in one’s mind. This classification of the ordinary, natural states of the mind is generally accepted, even though a few have added swoon and death to the list as separate state (Figure 27.7). FIGURE 27.7
States of consciousness in Hinduism
Some of the Upanishads refer to an extraordinary fourth state in addition to the three natural states. Mandukya Upanishad refers to a fourth state, a state of samadhi, which is called turiya, literally the fourth state. Among others, the Brahmopanishad and the Atharvasikhopanishad refer to the fourth state (Sinha, 1961). Samkara makes extensive use of this classification in his writings on consciousness. In some of the minor Upanishads we find further divisions of the Samadhi state. For example, the Brahmabindu Upanishad, Nandabindu Upanishad and the Yogasikhopanishad refer to unmanibhava as the highest state (Sinha, 1961). The Prapancasara Tantra speaks of a state beyond turiya, and calls it turiyatita in which the yogin experiences consciousness-as-such. Yoga as well as Advaita Vedanta also refer to two kinds of samadhi—determinate (samprajnata) and indeterminate (asamprajnata).
THE WAKING STATE (JAGRIT) Waking consciousness refers to what we earlier called primary awareness. It is the state in which the mind processes information received through the senses. The mind is of course not merely an entity that has the ability to process sensory information, but is also a storehouse of past experiences, and contains stored images and dispositions. Therefore, what is perceived in the waking state is not entirely a faithful reproduction of the external world. However, in the waking condition, what one perceives is preponderantly drawn from the external world. In other words, a person’s cognitions in the waking state are
400 K. Ramakrishna Rao largely determined by the external conditions. It is awareness produced by the senses. Such awareness may be considered objective in that it is shared and is common to all. For this reason, Mandukya Upanishad (1.3) calls it vaisvanara. It is consciousness of external objects. In this state the person experiences gross objects. According to Advaita thinkers, the human body is considered to exist in three forms. The first and the obvious one is the gross body. It is the vehicle by which one makes contact with the external world. There is also the subtle body called suksma deha or linga sarira. It consists of sensory-motor system and the mind. The mind includes the buddhi, ahamkara and manas. The subtle body, unlike the gross body, does not cease to exist after death. In addition to these two kinds of bodies, Samkara postulates a third, the causal body. The causal body is a manifestation of avidya, which is blissful ignorance, and the main source and substance of the other two bodies. In the waking state all the three bodies function and limit the cognitions.
DREAMING STATE (SVAPNA) The state of dreaming is one in which the functions of the gross body are in abeyance, but the other two bodies continue to limit one’s cognitions. The dream state is different from the waking state in that the cognitions are not brought about by the sensory inputs, but are generated internally by the mind. Mandukya Upanishad (1.4) calls it taijasa, because it is brilliant. It is the consciousness of the internal objects/states and involves the experience of subtle objects. According to Samkara, dreams are cognitions generated by the subconscious impressions produced by previous perceptions in the waking state. There appears to be a general consensus among Indian thinkers that dreams are the work of subconscious mind, and occur during the periods of light sleep. There are a number of theories to account for dreams. Some of them are surprisingly similar to current dream theories. Also, a somewhat comprehensive classification of dreams was entertained by people like Caraka in his Caraka Samhita. According to Caraka, there are seven different kinds of dreams. These include (a) dreams that contain previously seen, heard, or otherwise perceived events and objects; (b) dreams that incorporate one’s desires; (c) dreams generated by imagination; (d) paranormal dreams; (e) dreams instigated by external stimulation; and (f) dreams provoked by psychophysiological conditions in the person asleep. The dreams of the first category are mere reproductions of previously experienced events with some possible embellishments brought about by the sleep state. However, they are not experienced as memories, but as actual perceptions. The category of dreams of desire includes the wish-fulfilling ones. As his commentator Chakrapani says in his Chakrapanivyakhya, these are dreams that are gratifications of unfulfilled wishes. Caraka thus appears to have anticipated the wish-fulfillment aspect of Freudian theory many
Taxonomy of Consciousness
401
centuries before Freud. It is to the credit of Caraka that he did not restrict ‘desire’ to just sexual or infantile desires, and that he considered wish-fulfilling dreams as just one category among several others. Imaginative dreams are the creative ones, a product of imagination (kalpita), even though they may involve previously experienced objects and events. Paranormal dreams are those that correctly foretell future, the things to come. It is only recently that science is beginning to come to grips with the dreams of this kind (Ehrenwald, 1942; Ullman, 1969). Even more important is the category of dreams with diagnostic content. These are prompted by the physiological state of the person, and are suggestive of morbid or pathological conditions. While psychoanalysts and Jungian analysts made some use of dreams for psycho-diagnostic purposes, this is largely an unexplored area in the contemporary study of dreams, especially in relation to somatic disease. Against the background of Caraka’s system that postulates three kinds of humours balancing health, there are all kinds of possibilities for research in this area. In Mahabharata we find attempts to relate dreams to the mind’s sattva, rajas and tamas aspects, and to one’s karma. In the Indian classification of dreams there is also a recognition of lucid dreams, called ‘dreamend cognitions, svapnatika jnana’ (Sinha, 1958). These are dreams in which the dreamer has the awareness that she is dreaming. There is considerable phenomenologically meaningful discussion on the question as to whether dream imagery is simply embellished memory or essentially perceptual in character, whether it is representative or presentative. Among the proponents of the representative theory are Samkara and Prabhakara of the Mimamsa school. Samkara points out that dream imagery is of the nature of memory (smriti) (S.B., 2.2.29). Expressing a similar view, Prabhakara argues that dream imagery is memory image without memory. In dreams, past memories appear as perceptions because of a lapse of memory (smritipramosa) due to the sleep condition, which leaves the bare perceptions of the past without the associated memory. Thus, according to Prabhakara, dream images appear as perceptions even though they are essentially memories. In defense of his contention, Prabhakara points out that during the dream period, the sensory system is quiescent and, therefore, without senses operating, there can be no perceptual awareness (Sinha, 1958: 310). Interestingly, one of the proponents of Samkara’s Advaita school, Dharmaraja in his Vedantaparibhasa, summarily refutes the representative theory and argues that dream images are basically perceptual, though illusory. First, he points out, that no one in a dream state thinks that she is recollecting something. Second, in dreams we sometimes become aware of things that were never seen before. Third, things may appear differently at different times, but seldom will we mistake memory for perception and vice versa. When one perceives a pot, for example, it is ‘this’ pot that is immediately present, and, when one recollects a pot, it is ‘that’ pot to which a reference is made. The reference in one case is to a thing in the present and in the other case it is in the past. ‘Thisness is the special characteristic of perception alone, while thatness is of memory’ (Sinha, 1958, vol. I, 312).
402 K. Ramakrishna Rao Udayana of Nyaya school also considers dream images as perceptual. He disagrees with Prabhakara that the sensory system is altogether inoperative in sleep and the dream states. He suggests that peripheral stimuli continue to excite the system, and that on occasion dream awareness may incorporate external stimuli (Sinha, 1958, vol. I, 311). Again, this observation is entirely consistent with some of the modern dream studies. For example, Dement (1958) reported that water was incorporated into the dreams when a fine spray of cold water was applied to the subject while he was in the REM period, indicating that he was dreaming.
DEEP SLEEP STATE (SUSUPTI) A state of deep sleep is one in which not merely the sensory system, but the mind itself is quiescent. As the Mandukya Upanishad says, in the state of deep sleep, one ‘desires no desires whatsoever, sees no dream whatsoever’. It is called prajna, a mass of cognition, unified, consisting of bliss and enjoying bliss (1.5). In sleep the mind ‘retires’ to a region dissociated from sense-object contact and is devoid of any imagery of its own making. Sleep is a necessary condition for recouping from fatigue. There appears to be a general agreement up to this point among the various schools. There are, however, deep disagreements on detail as to what exactly happens in the sleep state, especially whether the dream state is a cognitive state. In the Advaita view, the state of deep sleep is not a cognitive state. Unlike the dream state where the mind is active, the mind in a state of deep sleep is dissolved in nescience (avidya), which is its cause. In other words, the functions of the gross body as well as the subtle body are in abeyance, only the causal body (avidya) functions. The mind undergoes no modifications, there are no vrittis, and the mind abides in avidya. As a result during sleep, one has no determinate or indeterminate perceptions, feelings of pain or pleasure. In short, there are no cognitions in the sleep state. The awareness of sleep is a mode of avidya and not of the mind. Therefore, it is not a cognitive state. Avidya and maya in Advaita Vedanta refer to the physical side of reality, which is termed prakriti in the Samkhya system. Literally avidya means ignorance and maya illusion. It seems to me that these two concepts are merely pejorative characterization, at the epistemic and ontic levels, of prakriti (matter), which in the idealistic system of the Advaita has no legitimate claim for reality. Avidya consists in mistaking matter for consciousness, mind for the self. Maya indicates the ontic status of matter as unreal and illusory. In all this the principal player in the Advaita view is the mind. As Samkara says in his Vivekachudamani, ‘there is no ignorance (avidya) outside the mind. The mind alone is avidya …’(169). If this interpretation is correct, it is difficult to see how awareness in the mental mode is different from awareness in the avidya mode. It would seem, therefore, that in the sleep state the mind is quiescent but its essential clouding of consciousness persists.
Taxonomy of Consciousness
403
In Patanjali’s Yoga, however, the sleep state is considered to be a cognitive state. The mind is not dissolved in sleep. It is not without experience. The cognition here is the experience of the absence of cognition. The person in a deep sleep state has the awareness of not having any cognition, whether in the form of waking experience or a dream. Sleep state is a felt and experienced state because on waking one recollects that she had good sleep. Thus, sleep is a state of the mind. If the mind is not involved, there would be no recollection of sleep. Vijnanabhiksu, for example, disputes the Advaita contention that the awareness of sleep is a mode of avidya and not that of the mind. If that were so, he argues, then there is no need for the mind because avidya can give us cognitions in the waking and dream states as well. It would seem that the mind dissolving itself or abiding in avidya may mean no more than that the usual functions of the mind are in abeyance. The differences between Yoga and Advaita on this point are more at the metaphysical level rather than at the psychological level.
THE FOURTH STATE (TURIYA) Mandukya Upanishad calls it ‘the lord of all. This is the all-knowing. This is the source of all, for this is the beginning and the end of beings’ (1.6). The fourth state, according to Mandukya, is not inwardly cognitive, not outwardly cognitive, not both-wise cognitive, not a cognition mass, not cognitive, not non-cognitive, unseen, with which there can be no dealing, ungraspable, having no distinctive mark, non-thinkable, that cannot be designated, the essence of the assurance of which is the state of being one with the Self, the cessation of development, tranquil, benign without a second (1.7, Hume’s translation).
In essence then turiya is a transcognitive state of pure consciousness. In the Advaita terminology, turiya is a state in which the person (jiva) finds his/her identity with Brahman or pure consciousness. It is a state in which the subject–object dichotomy ceases. It involves the transcendence of the threefold body. As mentioned, in the waking state all the three bodies function. In the dream state the gross body is in abeyance. In deep sleep state, while the gross body and the subtle body are in abeyance, the causal body persists to limit the person. In the fourth state, however, the person overcomes the limitations of all the three bodies. Having transcended avidya, the jiva is able to partake in and access consciousness-as-such. When in the state of turiya the person realizes the illusoriness of waking experience as he does the illusoriness of dreams on waking. The state of turiya is different from moksha (liberation) and jivanmukti (embodied liberation). It is different from the state of embodied liberation in that the transcendence of avidya in the turiya state lasts only as long as the person is in that state, and the person
404 K. Ramakrishna Rao will revert back into empirical consciousness at the end of that state, whereas in jivanmukti (embodied liberation) avidya is overcome once and for all. In other words, avidya persists in turiya, even though it is in a suspended state; but it is destroyed in the liberated person. Jivanmukti is different from moksha in that the latter is a disembodied state attained only after the person’s physical death. Unlike the other three states of consciousness, turiya is a state that is not natural, but one that can be cultivated and developed. It does not seem to involve any phenomenal awareness. Turiya appears to be essentially a non-intentional state of consciousness. In this state, the person is not cognitively aware of any object, except consciousness itself. Turiya is a state of samadhi as it is understood in Yoga. Vedanta thinkers generally accept the classification of samadhi into savikalapa (determinate) and nirvikalpa (indeterminate) forms. Sadananda, in his Vedantasara, distinguishes between the two in the following way. In savikalpa samadhi, the identity of individual consciousness with Brahman (pure consciousness) is experienced through the agency of the mind. Here the vritti has Brahman as the object of awareness. In nirvikalpa samadhi there is no awareness of the mental mode (vritti). The mind is completely dissolved.
BUDDHIST PHENOMENOLOGY OF CONSCIOUSNESS Buddhist writers present possibly the most comprehensive phenomenology of consciousness that we find in any system of thought, whether in the East or in the West. This is all the more interesting because Buddha did not attribute any substantiality to consciousness, mind or the self. Yet Buddhists looked at consciousness from different perspectives, taking into account the multiple forms it takes and the different functions it serves. In Buddhist psychology the basic unit of consciousness is thought. Thoughts arise as a relation between the subject and the object. Buddhaghosa (1920) defines consciousness in Althasalini as ‘that which thinks of its object’ (p. 148). Thoughts are intentional, to use Bretano’s expression. They are of or about something, real or imaginative. Thoughts function to guide, discriminate and inform. Each thought manifests in a series of continuous moments. It is caused by the psychical and endosomatic excitations in the person as well as by physical and external conditions. Thoughts have cognitive, volitional and feeling components. They also have ethical consequences to the person holding them. All these aspects are taken into consideration by Buddhist scholars in classifying consciousness into various categories. Buddhism is of course not a monolithic tradition. It comprises of several schools. Though they all claim to adhere to Buddha’s teachings, we find fundamental differences in their primary philosophical positions. Whereas the Theravada thinkers are naive realists and uphold the reality of the world outside and its independence from our thoughts and perceptions of it, the Yogacara school advocates unabashed idealism and the Madhyamika
Taxonomy of Consciousness
405
school ends up in undisguised nihilism. This situation is not very different from the Hindu tradition in which different systems of thought that champion radically different metaphysical doctrines flourished. However, as in the case of Hindu psychology, it is possible to extrapolate the significant contributions and powerful ideas and doctrines from the different strands of Buddhist thought and weave them into a cohesive framework towards an understanding of consciousness in the human condition. A major portion of Buddhist psychology is in Abhidamma (Davids,1910/1995). Theravadins developed the Abhidamma ideas more than others. Therefore, in the following account, we draw primarily from the Buddhist Theravada tradition, as interpreted by Buddhaghosa. Buddhaghosa was familiar with Buddhist literature in Pali as well as in Sanskrit. He travelled beyond India and studied Buddhism as it flourished in countries like Sri Lanka. We hope to present in this short summary the best Buddhism has to offer to our understanding of consciousness. Buddha focussed mostly on matters that were existentially demanding. He avoided answering some fundamental questions beyond what were relevant to his primary diagnosis of human predicament, which is suffering, and its eradication. However, as Buddhism grew in prominence, the need arose to systematically present Buddha’s views and to face the recurring onslaughts of opposing doctrines. The challenges were met not merely by enlarging on Buddha’s teachings, but by making creative additions to them, out of which much of Buddhist psychology grew. The philosophy of momentariness and impermanence is central to Buddha’s teachings. It provides the theoretical basis for his doctrine of suffering as the pervasive existential human condition. Buddhism avoids falling into the inevitable solipsism, following Buddha’s non-substantial view of nature, by subscribing to the law of universal causation in the doctrine of dependent origination (pratitya-samutpada). According to this doctrine, there is reason why a thing is what it is and not otherwise and that reason in principle is knowable. The law of dependent origination is ‘that being present, this becomes; from the arising of that, this arises; that being absent, this does not become; from the cessation of that, this ceases’ (Davids, 1923: 89). In its crude form, this law could be misconstrued as a kind of deterministic reductionism. Inasmuch as each thought is conceived as a consequence of a preceding condition, and thus each state of consciousness is dependent on a cause, the removal of that cause results in the dissolution of that consciousness. The removal itself is precedent on another cause. The wheel of existence is thus a web of cause–effect sequences. This notion applied to consciousness would be hardly different from epiphenomenalism and philosophical behaviourism except for the fact that Buddhist thinkers believed in the continuity of life beyond the present, and in the moral responsibility of persons for their actions in this birth and thereafter. Therefore, there is much more to consciousness in Buddhism than what meets the eye on the first reading of it. The first classification of consciousness is based on the division of awareness into its subconscious and conscious levels. Another classification follows the recognition
406 K. Ramakrishna Rao of different states of consciousness as we have seen in the Hindu classification. Further classifications include divisions on the basis of the objects of awareness and the affective, conative and ethical aspects of thoughts. Also, consciousness is analysed into what are considered to be the basic elements constituting it.
CONSCIOUSNESS AND THE UNCONSCIOUS Consciousness in Buddhism, as mentioned, is equivalent to thought, a notion somewhat similar to James. He used the word thinking ‘for every form of consciousness indiscriminately’ (James, 1890/1952: 146). Thought is a relation between subject and object. However, ‘subject’ and ‘object’ are to be seen as relative and mutually dependent terms in that one cannot have any meaning without reference to the other. ‘Both the subject and the object are alike transitory, the relation alone between the two impermanent correlates remaining constant. This constancy of relation, which … is consciousness itself, gives rise to the erroneous ideas of personal identity’ (Aung, 1929: 11). Consciousness is not a static state but a dynamic process. It is like a flowing stream. It has subliminal and supraliminal forms. In Theravada tradition, the subliminal stream called bhavanga is the ground condition for all conscious experience. As Aung (1929) puts it, ‘it is the subconscious state of mind—“below the threshold” of consciousness—by which we conceive continuous subjective existence as possible’ (p. 266). Bhavanga, the continuous subconscious medium subsisting in our subjective existence, is like the state of dreamless sleep. It is, however, endowed with the potentiality to guide, influence and cause our thoughts, passions and actions. It is the carrier of karma, the merits and demerits of one’s actions. When an external stimulus is applied or when an internal sensation arises, the thought process starts, which always has reference to an object. Then bhavanga, the unconscious stream is momentarily interrupted. When the flow of bhavanga stops momentarily, the mind acts in the form of attention (manodhatu). Then sensory awareness arises. This followed by a succession of cognitive acts (manovinnana) resulting in the cognition of the perceptual object. Thus the cognitive process involves six steps. First, the sensory stimuli, in perception, for example, impinge on our peripheral system, which is the sense-object contact. Second, the excitation or perturbation of bhavanga. Third, the momentary arrest of the unconscious stream. Fourth, the apperception of the object of perception. Fifth the registering and retention of the perceptual image. Finally, the cessation of awareness and its submergence into the unconscious stream (bhavanga). E.R. Sarachchandra (1958/1994) describes the process of cognition as given in Abhidamma thus: The process of perception in the Abhidhammatthasangaha begins with the vibration of the unconscious for two moments, in the second moment of which unconscious mind is cut off. These succeeding moments are those of attention (pancadvaravajjana), sensation
Taxonomy of Consciousness
407
(cakkhuvinnana), assimilation (smapaticchana), discrimination (santirana), determination (votthapana), seven moments of cognition (javana) and two of registration (tadarammana). The number is made up to seventeen by adding one moment of thought at the beginning of the perceptive process. This thought-moment occurs before the material object makes an impression on the sense-organ, and is technically termed past consciousness (atitabhavanga). (p. 46).
According to Buddhist thinkers, there are four kinds of perception. First is the registering of the sensation resulting from the processing of the sensory inputs, which gives the bare perception (sensation) of the perceptual object. The second involves the mental apprehension of perception (mano-vijnana). The third is self-consciousness (sva-samvedana). The fourth kind is paranormal awareness (yogi-pratyaksa), which, it is believed, can be obtained by intense practice of meditation. The stream metaphor is more appropriate to bhavanga, the unconscious stream, than to cognitive consciousness. Bhavanga may appear like the ‘fringe,’ as distinguished from the ‘focus’ of consciousness in William James. The fringe, according to James (1890/1952) is what surrounds every conscious thought and gives it ‘psychic overtures,’ its ‘halos’. It is the one that provides the context and the network of past experiences and future expectations that render the contents of present awareness semantically sensible. Bhavanga, the unconscious stream, however, has a different and much enlarged role and function in Buddhist psychology. Possibly, it is somewhat closer to the ‘transitive’ states distinguished from the ‘substantive’ states of consciousness in James. The latter are the ‘resting places’, whereas the transitive states are those that give the unbroken continuity between thoughts, despite the intervening periods of no awareness, such as in sleep. In Buddhist literature, the concept bhavanga as a psychological concept first appears in Milinda Panha (The Questions of Milinda) (Sarachchandra, 1958/1994). While explaining to King Milinda the nature of sleep and dreams, Nagasena says that when a man is in deep sleep, the mind ‘goes into bhavanga’. This was the first attempt in Buddhist psychology to face the problem of the discontinuity between different momentary, fleeting states of the mind. In early Buddhism there is a big unexplained jump from normal empirical consciousness and the higher levels of consciousness achieved by such practices as meditation, culminating in nirvana. Further, all schools of Buddhism believe in survival of the person after death and the possibility of reincarnation, and at the same time uncompromisingly advocate the doctrine of no soul (anatmavada). In the absence of a soul that survives death, what is it that survives death and is reborn in the next life? This question needed a convincing answer. Nagasena did not, however, develop the concept bhavanga any further beyond the notion that it is one where the mind takes recluse during states of sleep. It was left to Theravadins like Buddhaghosa to make use of the bhavanga concept to explain not only sleep states, but also higher forms of consciousness and the continuing cycle of births in the person’s march towards achieving nirvana.
408 K. Ramakrishna Rao In the absence of the notion of self and substantive mind in Buddhism, bhavanga came in handy to account for continuity of being and self-identity in the person. The mind itself is seen as no more than an aggregate of the five skandhas. The skandhas are (1) rupa (form) (2) vedana (feeling) (3) sanna (perception) (4) sankhara (volition) and (5) vinnana (consciousness). Rupa appears to include not only gross matter, but also the senses. Vedana is of three kinds—pleasurable, painful and indifferent. Sanna includes sensory as well as conceptual forms. Sankhara ‘really means the group of volitions and other associated factors’ (Aung, 1929: 274). Vinnana refers to the stage at which the cognitive process starts as well as the resulting awareness. Milinda Panha describes the person as a series of mental states (dhamma-santati). As far as the waking state is concerned, this explanation appears plausible. However, when it comes to sleep and dream states, there is a clear discontinuity between these states. Therefore, Nagasena had to take refuge in bhavanga as the sub-terrain of consciousness into which the mind retreats during sleep. Now, seen as the ‘life-continuum’, bhavanga becomes the much needed principle that not only gives continuity to the different states of the mind and identity to the person in this life, but also as the cause of the person’s unbroken continuity in other lives, past and future. As mentioned, Buddhists subscribe to the doctrine of survival and believe in moral responsibility of the person after death. Nagasena has left the problem of the continuity of the person in various incarnations vague. He clearly sounds unconvincing when pressed by the King for an answer to the question of whether the person in another birth is the same person who died. All that Nagasena could say was that a reborn person is neither the same nor different. It was a major problem in Buddhism to reconcile its theory that postulates no permanent substratum of consciousness with the belief in moral responsibility and survival of the person after death. With the help of the concept bhavanga, Buddhaghosa (1923) attempts to account for the continuity of the same person across different births. According to Buddhaghosa (1920), mind in its unconscious passive condition is bhavanga. Though free from the usual thought processes, bhavanga is a natural state of the mind. In this state, the mind is free from all its impurities because all impurities are caused by the external influences. Without these stains, the mind shines by its own radiance and luster. The thoughts of the dying person, which connect him to the next birth, and the ‘first dawn’ of consciousness in the new born are related to and are the species of the bhavanga of that person. The bhavanga consciousness in the new birth is caused by the bhavanga thoughts of the person at the time of death. There is thus continuity from birth to birth. As Sarachchandra (1958/1994) puts it: ‘When a man is to be reborn, his consciousness at birth takes for its object one of these dying thoughts, and begins to function in the new life. The consciousness of the newborn child is, therefore, the result of the consciousness of the dead man. Its moral character is also determined by the moral character of the dying thought’ (p. 83).
Taxonomy of Consciousness
409
Bhavanga is of course not meant to be a substitute to the soul or atman as used in Hindu thought. Unlike atman, bhavanga is not permanent. Like other things in Buddhism, it changes from moment to moment. The continuity itself consists in the causal connection between one momentary state to another. ‘Bhavanga is not a self-conscious soul. Consciousness always involves reference to an object (arammanam cinteti ti cittam) and this object (of consciousness in the new life) is the last thought at death’ (ibid, p. 88). There is some problem here unless we assume that every thought carries with it the unconscious stream. The identity of bhavanga with conscious thought is not explicitly recognized. Buddhaghosa is silent on the relation between the two (Sarachchandra, 1958/1994). If bhavanga is considered to be the pure mind radiating its own brilliant light unstained by any accretions of impurities from cognitive activities, its uniqueness for the person, as well as its association with any given thought becomes problematic. In order to overcome the problem, Buddhaghosa had to attribute cognizing function to bhavanga to cause the transition of the person’s consciousness to the next birth. However, what is cognized here is not an external object, but the last thought of the dying person. The most plausible interpretation of bhavanga, it seems to me, is to regard it as a nonsubstantial ground condition of consciousness, a condition without which no cognition is possible. At the same time, it is not cognition itself. If cognition is noetic consciousness, bhavanga may be considered as non-noetic consciousness. In fact, Buddhist scholars like Sarachchandra (1958/1994) use expressions like ‘anoetic consciousness’ and ‘potential consciousness’ to describe bhavanga. It would not be incorrect to say, it seems to me, that bhavanga is being consciousness as distinguished from cognitive consciousness, which is knowing consciousness. Cognitive consciousness is intentional whereas bhavanga consciousness is non-intentional. In a sense it is pure consciousness, consciousness-assuch. Bhavanga in some respects appears similar to alayavijnana in Yogacara school of Buddhism, which itself reminds us of the concept purusha in Samkhya-Yoga or atman in Vedanta. The Yogacaras are idealists, vijnanavadins. Alayavijnana is a crucial concept in this school. It is the ground condition of all reality. Alaya consciousness is a continually changing stream. It is the foundational, ground condition for all our thoughts and feelings. As Radhakrishnan (1989) says, it ‘is the mother-sea of consciousness, out of which things arise and into which they again return’ (vol. I, p. 631). This vast and deep sea of consciousness is embedded in our psyche, even though we are not aware of it, except what we see at the surface. It is pure and unblemished, self-manifesting consciousness. It is described in Lankavatara Sutra in the following verse: Like unto ocean waves Which by a raging storm maddened Against the rugged precipice strike Without interruption,
410 K. Ramakrishna Rao Even so in the Alaya-sea, Stirred by the objectivity-wind All kinds of mentation waves Arise a-dancing, a-rolling (translated by Suzuki (2003) in Awakening of Faith)
Thus, the alaya is the ocean of consciousness. The sensory stimuli are the winds that ruffle the surface of the ocean and create the waves, our thoughts. Like bhavanga, alaya is a continually changing. It is pure consciousness in of itself, but blemished in human minds, which make subject–object distinctions. It is mind in its non-conscious but potentially conscious state. Manas (individual consciousness) evolves out of alaya, which is universal consciousness. These are significant similarities; there are, however, some important differences between alayavijnana and bhavanga consciousness. In the Yogacara system alaya is also the memory bank, a storehouse of a person’s thoughts, deeds and their consequences. In Buddhaghosa’s writings, bhavanga is not credited with such a memory function. Another significant difference is that in Yogacara theory, cognitive consciousness is regarded as a manifestation of the alaya, whereas in Buddhaghosa’s view there is a sharp contrast between cognitive consciousness and bhavanga consciousness. Bhavanga is arrested when cognitions are apprehended. ‘The Yogacara theory of alayavijnana was’, writes Sarachchandra, ‘to all intents and purposes, a disguised soul’ (1958/1994, p. 95). Indeed, alayavijnana looks very much like the soul, a permanent entity carrying with it memories and vasana impressions. However, the Buddhist roots of Yogacara thinkers prevent them from treating it as such an entity with a semblance of permanence. ‘The philosophical impulse’, says Radhakrishanan (1989) ‘led the Yoagacaras to the Upanishad theory, while the Buddhist presuppositions made them halting in the acceptance’ (vol. I, p. 635).
ELEMENTS OF CONSCIOUSNESS (CETASIKAS) In Buddhism, a conscious state, as we noted, is one in which we have thoughts. A thought is a complex amalgamation. It is a compound consisting of various elements. These include the physical elements relating to the object and the mental factors called cetasikas. Some of these factors are common to all thoughts, whereas some others are specific to certain kinds of thoughts. In Abhidhammasangaha, we find reference to 52 different elements of consciousness. These are grouped into four classes. There are seven in the first group, which are universal and common to all states of consciousness. These are (a) contact (phassa), (b) feeling (vedana), (c) volition (cetana), (d) perception (sanna), (e) individuation or one-pointed concentration (ekaggata), (f ) psychic life (jivitindriya), and (g) attention (manasikara). The second group consists of six particular elements, which are neither moral nor immoral. Included in the third group are 14 elements involved in
Taxonomy of Consciousness
411
immoral thoughts. The fourth is the group of 25 elements accompanying moral thoughts. Whereas the seven universal mental elements ought to be found in all the 89 states of consciousness, the six particular elements are present in 55 states of consciousness. The universal elements seem to relate to the basic psychological processes.
Universal Elements The universal elements seem to relate to the basic psychological processes. The first universal element is contact (phassa), which brings the object into the field of consciousness, is described in the Atthasalini as the pillar that supports the structure of consciousness (p. 143). Contact consists in consciousness coming into touch with the object and then producing an impact. Contact is described both as a cause and an effect. As the sense and the object come together and are in contact, there arises the potential for awareness. The second universal element of consciousness is feeling (vedana). Feeling is not considered a quality of experience, but an agency of experience: ‘Feeling is that which feels. It has (a) experiencing as characteristic; (b) enjoying as function, or possessing the desirable portion of an object as function; (c) taste of the mental properties as manifestation; and (d) tranquility as proximate cause’ (Buddhaghosa, 1920: 145). Although feeling has enjoyment as function, it is not confined to pleasurable feelings alone. It includes painful as well as neutral ones. The third element vedana approximates what we characterized as subjectivity, ‘what is it like’, aspect of consciousness. The characteristic of volition (cetana) is coordination. It coordinates all the associated states of an object and, in so doing, it binds together the various states related to an object. The function of volition is disposition to act, which is present in moral as well as immoral states, but not in morally inoperative states. Volition is the source of a good deal of energy, and it manifests in the form of directing the associated states. Some writers on Abhidhamma criticize the translation of cetana as ‘volition’. For example, Guenther writes that ‘cetana not only arouses mass activity, but also sustains it so that certain definite results appear. This shows beyond doubt that the translation of cetana by volition is against all evidence… Cetana, to state it plainly, is something that corresponds to our idea of stimulus, motive or drive’ (1976: 43–4). Although cetana implies something more than what is normally meant by volition, neither ‘stimulus’, ‘motive’, nor ‘drive’ convey its precise intent either. Sanna is what gives us distinct cognitions. It enables us to recognize general relations between objects, and to have perceptions of all kinds—sensuous and mental. ‘The perception, the perceiving, the state of having perceived which on that occasion is born of contact with the appropriate element of representative intellection—this is the perception that there then is’ (Davids, 1923: 7). Attention (manasikara) is what brings the mind and its object together. ‘Attention is like a charioteer harnessing two horses (mind and object) into a pair’ (Aung, 1929: 282). ‘Mind indeed always gets at its object, its constant companion being attention (manasikara), without which it would be like a
412 K. Ramakrishna Rao rudderless ship, drifting on to any object. With this rudder the senses arrive at their proper destination’ (Aung, 1929: 283). Jivitindriya is described as the faculty of life. Dhammasangani defines jivitindriya thus: ‘The persistence of these incorporeal states, their subsistence, going on, their being kept going on, their progress, continuance, preservation, life, life as faculty—this is the faculty of life that there then is’ (Davids, 1923: 16–7). Ekaggata is ‘the stability, solidity, absorbed steadfastness of thought which on that occasion is the absence of distraction, balance, unperturbed mental procedure, quiet, the faculty and the power of concentration, right concentration’ (Davids, 1923: 11–2). Thus, we find that the above mentioned seven universal mental properties are the basic psychological processes involved in all states of consciousness. First of all, there is the life process without which no psychic activity is possible. Then, there is the contact between the sense organ and the object, which is made possible by attention. An amount of concentration that would enable the emergence of the object into a specific spacetime setting is also required. The emergence of the object into the conscious field causes both perceptions as well as feelings, which are also influenced by one’s volition. We may note how similar the universal mental elements described in Buddhism are to the characteristics attributed to consciousness by William James in the Principles. Vedana is the subjectivity, the personal aspect. Cetana is the process that gives us the unity and coherence in conscious experience. Sanna refers to the noetic aspect and intentionality, whereas manasikara is selective attention.
Particular Elements The six particular elements (pakinnaka cetasika) are (a) vitakka (initial application of the mind) (b) vicara (continued application of the mind) (c) adhimokkha (decision) (d) viriya (effort) (e) piti (joy) and (f ) chanda (intention/resolution). These are termed particular because they may not accompany all thoughts, whether the thoughts are moral or immoral. Unlike the universal elements, all of these are not found in every thought. For example, according to Buddhists, joy, which arises when one anticipates an agreeable object, is present in meditative states of consciousness. Resolution is absent in thoughts where there is doubt. The particular elements seem to refer essentially to the psychological factors involved in initiating and carrying out action following thoughts. They refer to the action mode of consciousness.
Immoral Elements The 14 immoral elements (akusala cetasika) are the motivating factors leading to evil acts. Of them, attachment or greed, ill-will or jealousy, and delusion (moha) are the
Taxonomy of Consciousness
413
basic and foundational sources of evil acts. Attachment is rooted in desires, cravings, greed, liking, acquisitiveness, immoderation in eating, and lust for power and praise. Ill-will is rooted in aversion. It involves hatred, envy, jealousy and remorse. It is a state of agitation accompanied by repulsion to the object of cognition. It leads to aggressive acts, slighting others and destroying things. Delusion is essentially ignorance, inability to discriminate between the right and the wrong. Delusional thoughts are sometimes marked by doubt, restlessness and tension. The element of delusion is present in all immoral thoughts.
Moral Elements The moral or wholesome elements (sobhana or kusala cetasika) are 25 in number. Nineteen of them are common to all moral thoughts and six are present only in some of them. The common elements in moral thoughts include confidence, mindfulness, equanimity and non-attachment. Among the particular elements in this category are right speech, right action and right livelihood, pity, rejoicing in the happiness of others and wisdom. Just as immoral thoughts lead to evil acts, moral thoughts lead to moral actions, which according to Buddhists are of 10 kinds. Thoughts with moral elements in them lead to agreeable effects and happy consequences. They ward off evil acts. The moral thoughts with elements of wisdom are motivated by non-attachment, right knowledge and no ill-will to others.
FOUR FORMS OF CONSCIOUSNESS Buddhists describe consciousness as belonging to four different realms depending on the state of the person in his psychological development (Figure 27.8). They are (a) the empirical world known through the senses and driven by desire (kama loka); (b) the meditative world with form (rupa loka); (c) the meditative world without form (arupa loka); and (d) the world of transcendence (lokottara). This division is very similar to the fourfold division found in the Upanishads and elaborated in Advaita. The difference between the two consists in that the single state of turiya; the samadhi state of consciousness in Advaita is divided into three separate states in Buddhism and the empirical consciousness of kama loka appears as three different states of waking, dream and deep sleep in Vedanta. In addition to the four planes mentioned, there appear to be two other distinct principles employed by Buddhist scholars in the classification of states of consciousness. One principle refers to the ethical nature of the thoughts. The second relates to the underlying sources of motivation.
414 K. Ramakrishna Rao FIGURE 27.8
Buddhism: four planes of consciousness
Buddhist thinkers have gone far beyond their Hindu counterparts in their classification of consciousness into myriad forms. They looked at the different elements that are involved in every thought. They paid attention to the cognitive, conative and emotive factors, as well as the ethical aspects that enter into the determination of consciousness at any given time. A given conscious state, a thought, can be something that arises automatically on the presentation of a stimulus or something that is desired by the person and determined by volition. Whether a thought is voluntary and self-willed or a spontaneous occurrence has important implications to the person in Buddhist psychology. Voluntary actions have ethical imputation, leading to harm and pleasant or unpleasant consequences (Figure 27.9). FIGURE 27.9
Ethical classification of consciousness in Buddhism Ethical classification Immoral
Moral
Resultant
With roots
Indifferent and ineffective
Inoperative
Without roots
Consciousness of the object may be accompanied by feeling/emotion that is wholesome or unwholesome or it may be devoid of any feeling. Cognition may be true or erroneous and thus give rise to true or false knowledge. Permutation and combination of these and other associated factors give Buddhists a variety of forms of consciousness, a variety broader and comprehensive than anything we find in the Hindu or any other psychological literature. Buddhist thinkers identify 54 forms of consciousness at the level of kama loka, 15 at rupa, 12 at arupa, and eight at the lokottara (transcendental) planes, a total of 89.
Taxonomy of Consciousness
415
According to another count, which subdivides the transcendental state into five classes, there are 40 distinguishable forms of consciousness in the lokottara plane, thus raising the total from 89 to 121.
Forms of Consciousness in Kama loka The empirical is the world common to all humans. As mentioned, Buddhism distinguishes between 54 forms of consciousness in the empirical world. This is the world operated by the senses, cognitively known, and driven by desire. The sensory awareness constitutes the objective aspect and the accompanying desires and cravings are the subjective aspect of a given thought in the empirical domain. Kama loka forms are divided into eight moral states, 12 immoral states and 34 morally indifferent states. The indifferent states are further classified into 11 inoperative states, and 23 resultant states. Some of these states are with roots and others are without roots. The moral states are those that lead to ethically desirable actions and result in agreeable and pleasant consequences. The thoughts in them may arise spontaneously or they may be willed by the person. Wisdom may or not be present in them. They arise as a consequence of correct knowledge of the object of thought and result in a mental state of purity. Moral states of consciousness are accompanied by confidence, mindfulness and joy. They lead the person to avoid evil and to engage in right action, right speech and right livelihood. As mentioned, moral thoughts lead to 10 kinds of moral actions such as charity, meditation and ethical behaviour. Of the eight moral states, four are accompanied by non-attachment and non-bias or ill-will, but without wisdom. Immoral thoughts lead to disagreeable effects, unethical behaviour and suppression of moral judgement. Some of these are prompted by attachment; some are rooted in illwill, hatred, envy and remorse; and the others are caused by delusion born of ignorance. Immoral thoughts are accompanied in the person’s mind as tension, restlessness and doubt. Eight of the immoral states are rooted in greed and attachment, two of them in ill-will and aversion, and two in ignorance and delusion. Immoral thoughts lead to immoral actions numbering 10, which include killing, stealing, lying, abusing, adultery and idle talk. In the empirical realm, there are 11 inoperative states and 23 resultant states. Unlike moral and immoral states which have effects, that is, accumulated karma, inoperative and resultant states have no future effects. The resultant states (vipaka citta) are the effects of past actions, whereas the inoperative states are not due to past deeds. The resultant thoughts are those over which the person has no voluntary control. They arise automatically. Unlike the moral and immoral thoughts, which are voluntary, resultant and inoperative states are not willed. For example, the occurrence of sensation following sensory contact with an object is something over which one has no voluntary control. Unlike the inoperative states, resultant states arise as the effects of past-life actions.
416 K. Ramakrishna Rao They are the causal and explanatory link between the present and past states in this and previous lives. The mental states are further classified into those with roots and those without roots. The roots are basic motives (Figure 27.10). They are six such motives, three leading to moral thoughts and three resulting in immoral thoughts. The motives associated with immoral states are (a) attachment or greed (lobha), (b) ill-will and hatred (dosa), (c) delusion or ignorance (moha). Non-attachment (alobha), absence of ill-will (adosa), and non-delusion (amoha) accompany moral thoughts. FIGURE 27.10
Roots of mental states
Higher States of Consciousness In Buddhism, as we have seen, empirical consciousness is mundane cognitive consciousness. It is possible for humans to achieve higher states of consciousness beyond the sense-mediated and mind-processed cognitions by disciplined training of the mind, accompanied by ethical living and followed by right knowledge, wisdom. The mind is normally hindered in its functioning by five factors. These hindrances are desires, hatred, indolence, restlessness and doubt. These need to be overcome in order to achieve higher states of consciousness. Meditation, coupled with moral habits, helps to suppress and overcome the hindrances. Buddhists trace five steps in the progress of meditation. They are called (a) vitakka (discursive and relational thought, also labelled as initial application), (b) vicara (inquiry, also called continued application), (c) piti (thrill and enjoyable interest), (d) sukha (psychological state of ease, comfort or equanimity), and (e) ekaggata (concentration). De Silva (1937/1988: 38) describes the functions of the five constituents of meditation in the following way: Initial application causes consciousness to apply itself on an object. Sustained application causes consciousness to apply itself on the object continuously and in a sustained manner. Pleasurable interest causes consciousness to derive pleasure from the object. Pleasure causes consciousness to partake the taste of the object. One pointedness causes consciousness to be concentrated or one-pointed without being distracted.
Taxonomy of Consciousness
417
Initial application inhibits sloth and torpor; sustained application inhibits doubt; pleasurable interest inhibits hate; pleasure inhibits restlessness and worry; one pointedness inhibits sense desires. In this manner when consciousness is accompanied by these five properties, it is called moral consciousness of the first stage of Jhana.
It is easy to see that initial application, continued application and concentration are very similar to dharana, dhyana and samadhi in Patanjali’s yoga. Dharana is the initial attempt to concentrate. Dhyana is sustained concentration; and samadhi is a state of absorption in the object of concentration, resulting from prolonged concentration. Piti and sukha are just the resultant effects on the mind following sustained concentration on an object. Thus what we seem to have here is a meditative state that helps to achieve higher states of consciousness by sustained concentration. In the Theravada tradition, there are basically three higher states of consciousness. One is said to belong to the world of rupa (form), another to the world without form (arupa). The third is an altogether different state beyond these worlds (lokottara), belonging to a different realm of being. Consciousness in the lokottara realm appears to be qualitatively different from the rest. It is the state Buddha attained; and it is the goal of all aspirants of nirvana. Rupavacara citta is the mind of the person reaching the rupa loka realm. When the practitioner progresses step by step to the state of concentration from initial application, the mind enters the meditative world of rupa. After reaching the level of concentration and continued practice, one reaches the first stage of meditation. The meditating person then by continued practice sheds one by one the other four constituents preceding concentration and thus is totally absorbed in a state of concentration. As the meditator sheds them progressively, he advances into different stages of jhana (meditation). By the time he reaches the fifth jhana state, he is just left only with one-pointed concentration. In the rupa plane they are 15 thought forms. All of them are moral. The thoughts in this realm are intentional in that they refer to objects, even though the thoughts become progressively more and more subtle. In the urupavacara citta, when the mind is in the formless realm, the thoughts are devoid of form. The objects of thought are too abstract and subtle. The meditator, who achieves this level of concentration, meditates on non-material things—(a) infinite space, (b) infinite thought, (c) nothingness, and (d) awareness which is neither perception nor non-perception. There are 12 thought forms in this state of the mind. As in the empirical realm of kama loka, the mind functions with distinctions of the subject and object in the rupa and arupa planes. Even though the objects of thoughts, whether physical or conceptual, are now more subtle, abstruse and abstract, they continue to be incorporated in the thoughts. It is, however, difficult to conceive of thoughts without having any form as the concept urupavacara citta implies. It would seem that urupa is the intermediate stage before reaching a state where awareness becomes completely non-intentional, devoid of any reference to an object.
418 K. Ramakrishna Rao Lokottara citta, consciousness in the fourth realm, is possibly just that non-intentional pure consciousness state. The person who achieves this state is completely transformed and regarded as a saint. It is said that a person may reach this state and attain nibbana without going through the rupa and arupa realms by practising insight meditation. Lokottara citta appears very much like what we have said about pure consciousness. The concept loka clearly implies the subject–object duality. The lokottara, which literally means beyond the world, implies that in this realm the subject–object duality is overcome. In fact, in Buddhism, the mind is no more than a succession of thoughts; and a thought is a relation between a subject and an object. What is the nature of thoughts in the formless and transcendental realms? Here, some important epistemological questions arise. They are too abstruse and complicated to discuss here. All that can be said at this point is that in the lokottara realm there can be no thoughts in the normal sense. If there are any thoughts, they must be considered to be beyond the categories of understanding we use in the empirical world. There is sufficient justification to consider the state of nibbana (nirvana) unconditioned, non-relational consciousness-as-such. It does not seem to have any proximate cause. It is non-intentional and without symbol. It is not a representation. Knowledge arising in that state is called realization (pativedha nana) rather than knowing. Buddhist scholar W.F. Jayasuriya (1963/1988) categorically asserts: ‘Unlike the cognition element and the mental factors the Nibbana element does not take or grasp objects (Anarammana), nor engage in any creative action having time relations’ (p. 96). It would seem that knowledge in the empirical realm (kama loka) is in the form of sensing. The knowledge in rupa and arupa realms is by way of understanding, and it is realization in lokottara citta. Realization may be seen as knowing by being, whereas the forms of consciousness in the empirical domain involve knowing by sensing. The forms of consciousness in the rupa and arupa states are the intermediary processes that progressively move the mind from sensory and rational modes of cognitive activity to transcendental consciousness. From this brief account of the Buddhist psychology of consciousness, we see how Buddhist thinkers paid a great deal of attention to psychological aspects in the human condition. Clearly there are significant similarities between the Hindu and Buddhist ideas bearing on consciousness, showing their reciprocal influence. However, there are also major differences between them, which we may not ignore. Buddhism is more empirical and analytical than Indian psychology in general. Buddhist psychology does not make an explicit distinction between consciousness and mind; and intentionality appears on the surface as the defining characteristic of mind and consciousness. In all the other Indian systems, a clear-cut distinction is made between mind and consciousness. Systems like Samkhya-Yoga and Advaita Vedanta explicitly attribute intentionality to mind alone and not to consciousness. It is not surprising, therefore, that among those scientists in the West who pay attention to eastern ideas, Buddhism appears more appealing because in the western tradition mind and consciousness are used interchangeably and intentionality is considered to be the defining feature of consciousness (Rao, 2002).
Taxonomy of Consciousness
419
It is hazardous to hasten to make inter-cultural comparisons. However, the similarity between the views of William James and Buddhism on consciousness is too striking to go unobserved. For example, the shift in focus and perspective and the consequent understanding of consciousness from the Principles of Psychology (1890) to the Varieties of Religious Experience (1902) in James has an uncanny resemblance to the shift in Buddhist description and discussion of consciousness from the kama loka plane to lokottara state. One wonders whether, had Buddha been a systematic thinker instead of a religious reformer, he would have said as James did when he wrote in his Essays on Radical Empiricism (1912) that ‘the hour is ripe for it (consciousness) to be openly and universally discarded’ (p. 3). On the surface, it might appear that James was inconsistent in his discussion of consciousness in the Principle and the Varieties and his rejection of it in the Varieties. I think it would be a misunderstanding of James. In the Principles James was dealing with empirical, or as Buddhists would say, kama loka consciousness. Here he uses, again like the Buddhists, the fitting metaphor of the stream. His description of consciousness as personal, changing and yet continuous, selective, noetic and unified is not different from the Buddhist accounts. When dealing with what he called ‘non-rational forms’ of consciousness, the stream metaphor in James gives away to the metaphor ‘mother sea’. The description moves on to the subliminal, the multiple sub-terrain streams that lack unity and cohesion, and to the experiences that transcend reason and defy logic. Yet James recognizes them as real facts of experience we must deal with as psychologists. In his Essays, where James was inclined to do away with the concept of ‘consciousness’, what he was attempting to discard was not consciousness per se, but consciousness as an entity or substance, a notion which is repugnant to his radical empiricism. In a similar fashion, Buddhism is opposed to all attempts to hypostatize consciousness as an entity, such as soul, self, atman or Brahman. Again, the concept bhavanga in Buddhism befits the metaphor ‘mother sea’. In fact, the Yogacara thinkers used the metaphor of sea and waves to describe alaya consciousness in relation to empirical consciousness at the level of the individual. We do not find Buddhist thinkers making any clear-cut distinction between mind and consciousness in the kama loka domain. Also, contentless or non-intentional consciousness in this domain appears to be an absurdity. Here, the mind rules and consciousness follows. However, at the rupa and arupa planes, we find the distinction between consciousness and the mind making its way. The mind in these lokas is restrained in a state of samadhi, and consciousness takes an altogether different form. In the lokottara plane the mind is not merely restrained, it is completely annihilated in the state of nirvana and transformed, manifesting consciousness in its most sublime form. Thus, our reading of Buddhist accounts of consciousness at different levels of being shows that they are not inconsistent in the same way the discussions of consciousness by James in different contexts are not contradictory when we take into consideration the different forms consciousness takes at different levels of its manifestation.
420 K. Ramakrishna Rao CONCLUSION As we have seen, consciousness is classed into four basic categories in Vedanta as well as in Buddhism. The categories at both ends are nearly identical. They begin with the existential, natural condition of empirical consciousness and move on to describe the highest possible state of consciousness as turiya or nirvana. The middle categories vary, however, because their focus varies. Even though the ultimate goal advocated by them is freedom from ignorance and suffering via liberation of consciousness from its existential constraints in human condition, the focus in Buddhism is on the ground reality and empirical consciousness, whereas in Advaita it is on the ultimate principle of consciousnessas-such. The problem in both cases is, however, to connect the two realms, which are apparently disparate. In a sense, the approach of Buddhism is bottom up whereas in Vedanta it is top down. Buddhism, with its primary practical concerns, looks into higher forms of consciousness to provide the connecting links and finds them in the rupa and arupa forms, which facilitate the transition from the empirical to the transcendental. From an experiential perspective, when one moves from intentional consciousness of the empirical realm to states of pure consciousness without content, there is a progressive shift from the concrete to the abstract and finally to emptying of consciousness of all its cognitive content. In concentrative forms of meditation, the subject first focuses on an object in its physical form. As the meditation progresses, the object in the focus becomes more and more abstract, subtle and formless. In a state of samadhi, the object ‘merges’ with the subject and the distinction between the two vanishes. Following this, the person reaches a state of pure consciousness. In Advaita, with theoretical interests dominant, the focus is on consciousness-assuch. Consequently, it looks at consciousness as it manifests in the empirical realm and attempts to find in it the necessary links with pure consciousness. State of dreaming and sleep provide the necessary links, as they suggest how the mind gets disconnected from normal information processing modes. In dream states the mind’s connection with the sensory system is temporarily in abeyance. In deep sleep state, the mind itself retires, but awareness of the self persists. While Buddhism is pretty much content with exploring the ways of reaching nirvana, Advaita goes further to understand consciousness itself as a metaphysical principle. According to Samkara, consciousness is self-luminous and self-revealing. Self-luminosity implies that it is known without being an object of knowing. It is self-disclosing. It manifests without being mediated. Its experience is immediate and direct. In other words, consciousness is always present in all its acts without in anyway being an object. It follows therefore that consciousness is self-revealing subjectivity (svaprakasa). Such self-revealing is beyond doubt, error and contradiction in the person. The rationale behind the Vedanta classification of consciousness into four states, as explained, lends support to the notion that consciousness is undifferentiated subjectivity or non-objective consciousness.
Taxonomy of Consciousness
421
In dreams, what one has usually are copies of waking percepts that do not follow any definite rules of formation. Here imaginative construction has greater freedom than in waking condition. In the waking state, consciousness is so deeply intertwined with objective determinations that it can hardly move of its own accord, as it does in the dreaming state. So long as the person is awake, there is scarcely any freedom for consciousness from the objects with which it is associated. In dreams, however, the association with external objects is reduced to a point that they do not limit consciousness. Thus in dream states consciousness seems free from objective determinations. Indeed objectivity itself appears to disappear. Dream consciousness is not bound to any sense datum arising at the time. No object manifests it. As K.C. Bhattacharya (1909), one of the brilliant Indian philosophers in the 20th century, puts it, ‘dreams may be described as perceptions without sensations’ (p. 2). The self-revealing nature of consciousness is more clear and unmistakable in the state of deep sleep (susupti). Samkara points out that the state of deep sleep is not without consciousness because after waking up, the person says that she had blissful sleep. Even assuming that this awareness is a result of memory and not of any specific perception in that state, Samkara argues, that memory itself is always a recall of something that was actually presented to the person in that state. Therefore, the consciousness of sleep and bliss must have been experienced by the person in sleep. Since the empirical consciousness determined by sensory objects lapses during the state of sleep, what is left is the subjectivity itself. In the state of deep sleep, consciousness consists in the absence of awareness. It is ‘direct cognition of the absence of specific cognition’ (Bhattacharya, 1909: 14). Thus, in a state of deep sleep, consciousness continues to have an object as a reference. The determining object in this case is the absence and not the presence of the object. So what is experienced in deep sleep is not consciousness-as-such. Consciousness free from all objective determination, pure consciousness, is experienced only in the fourth state of turiya. Consciousness in the state of turiya is undetermined and free from all objective limitations. It represents, according to Advaita, a higher level of being. It signifies realization of reality transcending all determinateness. It is a state where there is no duality; the veil of ignorance (avidya) is lifted so that the being itself is installed in consciousness itself. As Bhattacharya (1909) points out, the four different states of consciousness form a gradation of existence. At one extreme there is empirical awareness, the waking experience of consciousness, where there is relatively little freedom for the subject from objective determination of consciousness. Consciousness at this stage completely identifies itself with the objective world. The other extreme is the state of undifferentiated subjectivity in a state of pure consciousness. Between these two come the states of dream and sleep. They reveal the ever increasing freedom from objectivity and a deepening of the installation of the person in non-dual consciousness. It does not take much imaginativeness to see how similar the notions of nirvana in Buddhism and the concept of pure consciousness in Advaita are. Like pure consciousness,
422 K. Ramakrishna Rao nirvana is beyond relativity and duality. Nirvana and Brahman appear to be undetermined and unconditioned pure consciousness. They suggest manifestation of being without symbols. They represent direct and unmediated awareness. Both are considered states of bliss. Attributions of good and evil, right and wrong are inapplicable to them. They are non-relational states without subject–object distinctions. They are states in which the veil of ignorance is removed; the person is installed in complete undifferentiated subjectivity. Advaitins as well as Buddhists believe that the states of pure consciousness can be achieved by embodied persons. To conclude, consciousness has many meanings because it manifests itself in different states of the mind and in different forms. The failure to recognize the complexity of consciousness, the tendency to confuse one state for another, or attempt to reduce one form to another, and the conflation of mind and consciousness are some of the reasons why it came to be so problematic to study consciousness in the western tradition. That there is such close resemblance between the widely different systems of thought, as represented by Advaita Vedanta and Theravada Buddhism on the nature of consciousness at different levels is a matter of considerable significance in the context of developing a science of consciousness.
REFERENCES Armstrong, D.M. and N. Malcom. 1984. Consciousness and Causality: A Debate on the Nature of Mind. Oxford: Basil Blackwell. Aung, S.Z. (trans.). 1929. Compendium of Philosophy. London: Oxford University Press. Bauer, R.M. 1984. ‘Autonomic Recognition of Names and Faces in Prosopagnosia: A Neuropsychological Application of the Guilty Knowledge Test’. Neuropsychologia, 22: 457–69. Benson, H. 1975. The Relaxation Response. New York: William Marrow. Bermúdez, J.L. 1998. The Paradox of Self-Consciousness. Cambridge, Massachusetts: MIT Press. Bhattacharya, K.C. 1909. Studies in Vedantism. Calcutta: Calcutta University Press. Blackmore, S. 2002. ‘There is No Stream of Consciousness’. Journal of Consciousness Studies, 9(5–6): 17–28. Bornstein, F. and T.S. Pittman (eds). 1992. Perception Without Awareness. New York: Guilford Press. Buddhaghosa. 1920. Althasalini (The Expositor). 2 Vols. (Trans. M. Tin). London: Oxford University Press. ———. 1923. Visuddhimagga (The path of purity). 3 Vols. (Trans. M. Tin). London: Oxford University Press. Crick, F. and C. Koch. 1998. ‘Consciousness and Neuroscience’. Cerebral Cortex, 8: 97–107. Davids, C.A.F.R. (ed. and Trans.). 1923. A Buddhist Manual of Psychological Ethics. London: Royal Asiatic Society. ———. 1910/1995. Compendium of Philosophy: Translation of Abhidhammattha Sangaha, Pali Text Society. Dement, W. 1958. The Physiology of Dreaming. Doctoral Dissertation. University of Chicago. Dennett, D. 1991. Consciousness Explained. New York: Little Brown. Descartes, R. 1911/1969. The Philosophical Works of Descartes. (Trans. E.S. Haldane and G.R.T. Ross). Cambridge: Cambridge University Press.
Taxonomy of Consciousness
423
De Silva, C.L.A. 1937/1988. A Treatise on Buddhist Philosophy Abhidhamma. Delhi: Sri Sadguru Publications. Ehrenwald, J. 1942. ‘Telepathy in Dreams’. British Journal of Medical Psychology, 19: 313–23. Forman, R.K.C. (ed.). 1990. The Problem of Pure Consciousness. New York: Oxford University Press. Foulkes, D. and G. Vogel. 1965. ‘Mental Activity at Sleep Onset’. Journal of Abnormal Psychology, 70: 231–43. Foulkes, D., P. Spear and J. Symonds. 1966 ‘Individual Differences in Mental Activity at Sleep Onset’. Journal of Abnormal Psychology, 71: 280–86. Goleman, D. 1978. The Varieties of the Meditative Experience, New York: Irvington publishers. Greenfield, S. 1998. ‘How Might the Brain Generate Consciousness?’ In S. Rose (ed.), From Brains to Consciousness. London: Allen Lane. Guenther, H. 1976. Philosophy and Psychology in the Abhidharma. London: Routledge. Hilgard, J.R. 1977. ‘Imaginative Involvement: Fantasy and Susceptibility’. In P. Zimbardo and C. Maslach (eds), Psychology for Our Times. Glenview, Ill: Scott, Foresman. Hilgard, E.R. and J.R. Hilgard. 1983. Hypnosis in the Relief of Pain. Los Altos: William Kaufmann, Inc. James, W. 1902/1914. Varieties of Religious Experience. New York: Longmans, Green & Co. ———. 1912/1947. Essay in Radical Empiricism. New York: Longmans, Green & Co. ———. 1890/1952. The Principles of Psychology. New York: Henry Holt & Co. Janet, P. 1907/1920. The Major Symptoms of Hysteria. Oxford, England: Macmillan. Jayasuriya, F. 1963/1988. Philosophy of Buddhism. Colombo: YMBA Press. Kihlstrom, J.F. 1979. ‘Hypnosis and Psychopathology: Retrospect and Prospect’. Journal of Abnormal Psychology, 88: 459–73. ———. 1984. ‘Conscious, Subconscious, Unconscious: A Cognitive Perspective’. In K.S. Bowers and D. Meichenbaum (eds), The Unconscious Reconsidered (pp. 149–211). New York: Wiley-Interscience. Maslow, A.H. 1973. The Farther Reaches of Human Nature. Harmondsworth, UK: Penguin. Moore G.E. 1922. A Refutation of Idealism: Philosophical Studies. London: Routledge & Kegan Paul. Myers, F.W.H. 1903/1915. Human Personality and Its Survival of Bodily Death, Vol.1. New York: Longmans, Green. Natsoulas, T. 1978. ‘Consciousness’. American Psychologist, 33(10): 906–14. Radhakrishnan, S. 1989. Indian Philosophy. (Centenary Edition) 2 Volumes. New Delhi: Oxford University Press. Rao, K.R. 2002. Consciousness Studies: Cross-Cultural Perspectives. Jefferson, North Carolina: McFarland & Co. Rao, K. R. and J. Palmer. 1987. ‘The Anomaly called Psi: Recent Research and Criticism’. Behavioral and Brain Sciences, 10, 539–555. Sarachchandra E.R. 1958/1994. Buddhist Psychology of Perception. Colombo: Cylon University Press. Singer, J.L. 1975. ‘Navigating the Stream of Consciousness: Research in Daydreaming and Related Inner Experience’. American Psychologist, 30: 727–38. Sinha, J. 1958. Indian Psychology: Cognition, Vol. I. Calcutta: Sinha Publishing House. ———. 1961. Indian Psychology: Emotion and Will, Vol. 2. Calcutta: Sinha Publishing House. Stace, W.T. 1960. Mysticism and Philosophy. Los Angeles: Tarcher. Suzuki, T. 2003. The Awakening Faith. New York: Dover Publications. Tart, C.T. (ed.). 1969. Altered States of Consciousness. New York: Wiley. Ullman, M. 1969. ‘Telepathy in Dreams’. Experimental Medicine and Surgery, 27: 19–38. Weiskrantz, L. 1986. Blindsight: A Case Study and Implications. Oxford: Clarendon Press.
About the Editors and Contributors
The Editors Narayanan Srinivasan is currently a senior faculty at the Centre for Behavioural and Cognitive Sciences, University of Allahabad, India. He obtained his Master’s degree in Electrical Engineering from Indian Institute of Science and his Ph.D. in Psychology from University of Georgia. His main interests are in visual perception, attention, consciousness, and cognitive modelling. A.K. Gupta is a Professor in Computer Science (former Head of the Department) at J.K. Institute of Applied Physics and Technology, University of Allahabad, India. He has held various important positions at other institutions. He was a visiting Assistant Professor at the Department of Physics and Astronomy, Northwestern University, Evanston (Illinois) from 1974 to 1977. He is interested in Signal and Image processing, Evolutionary computation, Chaotic dynamics, Cognitive neuroscience, Neural networks, Artificial life, Number theory, Statistical properties of networks, and Computational biology. He teaches various courses in Bioinformatics, Computer Science and Cognitive Science. Prof. Gupta has a large number of publications and has supervised a large number of theses and dissertations. Janak Pandey has served as Professor of Psychology, University of Allahabad, India since 1978. He is at present Head of the Department of Psychology and Co-ordinator of the Centre of Advanced Study in Psychology and for the Centre of Behavioural and Cognitive Sciences, University of Allahabad, India. He earned his Ph.D. degree as a Fulbright Scholar from Kansas State University and later he served as Assistant Professor of Psychology at the Indian Institute of Technology, Kanpur. He has also been a scholar-in-residence and Visiting Professor at the Wake Forest University, Professional Associate at the East-West Centre, Hawaii, Visiting Senior Commonwealth Fellow at the University of Manitoba, Visiting Professor, University of Leiden, and the Director of G.B. Pant Social Science Institute at Allahabad. His contributions to the discipline have received recognition in the form of several awards, including the ICSSR Professor V.K.R.V. Rao Award in Psychology in 1989 and National Fellowship in 1998. He has held significant positions, such as President of the International Association for Cross-Cultural Psychology, and Associate Editor of
About the Editors and Contributors
425
The Journal of Cross-Cultural Psychology. His tireless efforts as the editor of the Third and Fourth Surveys of Psychology in India over two decades have resulted in publication of six volumes of Psychology in India: State of the Art, providing international recognition to psychology research and profession. In September 2006, Prof. Pandey was recognised by the Chancellor of the University of Allahabad for his contribution to the discipline of psychology.
The Contributors Geerdina M. van der Aalsvoort, Ph.D., is Associate Professor at the Faculty of Social Science, Utrecht University, and Professor at the Faculty of Education, Utrecht School of Applied Sciences, The Netherlands. Her expertise is theory building on development of children at-risk for learning difficulties, aged 4 to 8 years. Her current research activities include studies related to social interaction during academic tasks and learning potential tests, and the relationship between emergent collaboration play and future academic performance. Ahmed received M.Tech. (AI) degree from University of Hyderabad, Hyderabad, India. He is currently a graduate student in the Computational Intelligence Laboratory, Department of Computer and Information Sciences, University of Hyderabad, Hyderabad, India. His research interests include Reinforcement Learning, Computational Neuroscience, Machine Learning and fMRI. Hari S. Asthana is currently the Reader in Department of Psychology, Kumaon University, Uttaranchal. His area of research interest includes Cross-cultural Neuropsychology and Health Psychology. He has to his credit two book chapters and 25 research papers published in both national and international journals. John W. Berry, Ph.D., is Professor Emeritus of Psychology, Queen’s University, Canada. He obtained his B.A. at Sir George Williams University (Montreal), his Ph.D. at the University of Edinburgh (Scotland), and has received Honorary Doctorates from the University of Athens and University of Geneva. He has published over 30 books in the areas of crosscultural, social and cognitive psychology. His main research interests are in the areas of ecology of cognition, acculturation and intercultural relations. Ramakrishna Biswal is currently a Junior Research Fellow at Defence Institute of Psychological Research Defence R&D Organisation, India and is pursuing his Doctoral Degree from University of Delhi, Delhi. He earned his M.Phil in Psychology from the Centre of Advanced Study in Psychology, Utkal University, Bhubaneswar. His area of research interest includes Social-cognitive Neuropsychology and Advanced Psychological Testing.
426 Advances in Cognitive Science Avi Chaudhuri was raised and educated in Canada and USA. He took his B.Sc. and M.Sc. degrees from the University of Toronto and Ph.D. from the University of California at Berkeley. Following his Ph.D., Dr Chaudhuri spent three years at the Salk Institute in San Diego as a research scientist. In 1993, Dr Chaudhuri set up his own research laboratory at McGill University where he conducts neuroscience research at both molecular and systemic levels. He is currently Full Professor of Psychology at McGill. In 2002, he received the James McGill Chair. Dr Chaudhuri has served as an expert consultant for the US Federal Government in his capacity as Program Director for Neurosciences at the National Science Foundation, Arlington, VA during his sabbatical leave of 1999–2000. He has received numerous awards including the Medical Research Council Fellowship of Canada and the prestigious Alfred P. Sloan research fellowship. Dr Chaudhuri currently runs a highly active and successful research lab with over 15 researchers at McGill University. H. Branch Coslett is Chief of Cognitive Neurology at the University of Pennsylvania. He trained with Dr Ken Heilman at the University of Florida before moving to Temple University where he collaborated extensively with Dr Eleanor Saffran. His research interests include motor control, sensory-motor integration, temporal processing, acquired dyslexia and amnesia. Research methodologies employed in his lab include investigation of patients with brain dysfunction (focal lesions and degenerative diseases), transcranial magnetic stimulation and fMRI. Matthew W. Crocker received his Doctorate in Philosophy from the Department of Artificial Intelligence at the University of Edinburgh, Scotland in 1992. In January 2000, he was appointed to the Chair in Psycholinguistics in the Department of Computational Linguistics at Saarland University, Germany. His research is concerned with the experimental and computational investigation of the adaptive mechanisms that underlie human language processing. Pierre R. Dasen is Professor of Anthropology of Education at the Faculty of Psychology and Education of the University of Geneva. He studied developmental psychology in Geneva, was an assistant to J. Piaget, and received a Ph.D. from the Australian National University. He studied the cognitive development of Aboriginal children in Australia, Inuit in Canada, Baoulé in Côte d’Ivoire and Kikuyu in Kenya; he has also contributed to research in cognitive anthropology among the Yupno of Papua-New-Guinea and in Bali. His research topics have included visual perception, the development of sensory-motor intelligence, the causes and effects of malnutrition, the development of concrete operations as a function of eco-cultural variables and daily activities, definitions of intelligence, number systems and spatial orientation. His interests are in everyday cognition, informal education and parental ethnotheories. He is currently completing a study of spatial cognitive development in Bali, India and Nepal. Dasen is the co-author or co-editor of several volumes and textbooks on cross-cultural psychology and intercultural education.
About the Editors and Contributors
427
Kenji Doya is currently Head, Department of Computational Neurobiology, Computational Neuroscience Laboratories, ATR International, Kyoto, Japan, and is also heading the Neural Computation Unit, Okinawa Institute of Science and Technology, Gushikawa, Okinawa, Japan. He obtained his Ph.D. from University of Tokyo. His research interests include Reinforcement Learning, Neuromodulators, Computational Neuroscience, Cyber Rodents and Recurrent Neural Networks. He is action editor for several journals including Neural Networks, Neural Computation and Network: Computation in Neural Systems. He has been conducting the Okinawa Computational Neuroscience Course since 2004. Pamela M. Greenwood is Research Professor, George Mason University, Virginia, USA. She has completed her Ph.D. on psychophysical scaling in split-brain patients. Her major area of study is changes in cognition that occur in the course of healthy ageing and in Alzheimer’s disease. She has published more than 100 papers and contributed several chapters in various volumes. Her research has been funded by NASA, NIH, Alzheimer’s Association and by the National Institute on Aging. Aravind K. Joshi is the Henry K. Salvatore Professor of Computer and Cognitive Science at the University of Pennsylvania. Prof. Joshi has received various awards including the David Rumelhart Prize for the year 2003 by the Cognitive Science Society and the Research Excellence award of the International Joint Conference of Artificial Intelligence (highest honour in the field of Artificial Intelligence). He was appointed to the National Academy of Engineering, the only researcher in Natural Language Processing to have ever received this distinction. In 2002, he was the first recipient of the Lifetime Achievement Award given by the Association for Computational Linguistics. Prof. Joshi is best known for his contributions to the formal science of language especially his work on Tree Adjoining Grammar. Prof. Joshi had edited two books and has published more than 100 articles on computational linguistics. Prem Kumar Kalra obtained his Ph.D. in Electrical Engineering from University of Manitoba, Canada. He is currently Professor and Head of Electrical Engineering Department at IIT Kanpur, India. His research interests are Power systems, Expert systems applications, HVDC transmission, Fuzzy logic and Neural networks application, and KARMAA (Knowledge Acquisition, Retention, Management, Assimilation and Application). Malavika Kapur is an honorary Professor at the National Institute of Advanced Studies, Bengaluru. She was earlier the Professor and Head of the Department of Clinical Psychology at the National Institute of Mental Health and Neurosciences, Bengaluru. She has a Ph.D. in Clinical Psychology from Bangalore University and has seven books and over 100 publications to her credit. She is a Fellow of the British Psychological Society, the Indian Association of Clinical Psychologists, the Indian Association of Child and Adolescent Mental Health and the National Academy of Psychology. She has been awarded the Scholar-in-Residency in the Bellagio Study and Conference Centre in Italy, two times by
428 Advances in Cognitive Science the Rockefeller Foundation. Her areas of interest are Developmental Psychology, developing Community Mental Health programmes for children and adolescents in urban, rural and tribal schools, primary health care and Anganawadi workers. She has been involved in the development of assessment tools and intervention packages for children and adolescents in the Indian context. Her main contribution is her work of developing integrated models of mental health service delivery for children and adolescents. Bhoomika R. Kar is a faculty at the Centre for Behavioural and Cognitive Sciences, University of Allahabad, India. Her areas of interest are developmental neuropsychology and cognitive neuroscience. She has done her Ph.D. in Clinical Psychology from the National Institute of Mental Health and Neurosciences, Bengaluru, India. She has developed and standardized a neuropsychological battery for children and looked at the age related differences in cognitive functions in children (5–15 years) as a part of her doctoral work. Her current research work focuses on the cognitive mechanisms in dyslexia in English as well as Hindi language. She is also looking at the behavioural and electrophysiological correlates of the effects of remediation in dyslexia. Arjette M. Karemaker is a Ph.D. student at Nottingham University, Department of Developmental Psychology, England since 2004. She wrote her M.A. thesis on metaplay by young at-risk children, and she graduated from Leiden University in 2004. She assisted as researcher in the study on play that is reported in this book. Her research activities are concerned with testing hypotheses about the advantages of learning to read by electronic books versus traditional books. Takahiro Kawabe is Lecturer of User Science Institute at Kyushu University, Japan. His research focuses on perceptual and cognitive aspects of visual processing in the brain. He has authored several articles on human perception in international journals such as Vision Research, Experimental Brain Research, Visual Cognition, Perception, and so on. Mieke P. Ketelaars is a Ph.D. student at the Department of Learning Difficulties, Radboud University, Nijmegen, The Netherlands since 2004. She wrote her Master’s thesis on autistic syndrome disorders, and she graduated from Leiden University in 2003. She assisted as researcher with the study on play that was reported in this book. Her research activities are related to language development of children with Autistic Syndrome Disorder. She hopes to finish her Ph.D. by 2009. Pia Knoeferle has completed her Ph.D. in Computational Linguistics at Saarland University, Germany in 2005. Her research focuses on the importance of scene information such as depicted events for online language comprehension and has begun to outline a processing account of the interplay between scene information, utterance and world knowledge. Peter R. Krebs recently completed his Ph.D. in Cognitive Science at University of New South Wales (UNSW), Australia. He holds a B.Sc. from the University of Sydney and M.A.
About the Editors and Contributors
429
(Cognitive Science) from UNSW. He has taught Computer Science and Professional Ethics at UNSW for 3 years, and lectures currently at CQU (Central Queensland University) in Software Engineering, both undergraduate and postgraduate subjects. He also teaches History and Philosophy of Science at UNSW. His research interests are in the Philosophy of AI and the Philosophy (Empistemology) of Computer models and experiments. Thomas Lachmann is Professor of Psychology, Dean of the Faculty and Head of the Department of Psychology II, University of Kaiserslautern, Germany. He is also a Visiting Researcher at the University of Leipzig, Germany, and a Guest Scientist at the Brain Science Institute in Wako, Japan. He studied Psychology at the University of Leipzig and the University of California at San Diego and received his Ph.D. from the University of Leipzig. Until 2006 he was Professor of Psychology at the University of Bamberg, Germany. Lachmann’s research interests are on basic functions of reading and dyslexia, procedural learning and structures of visual and phonological representations. Gary Libben is Professor of Linguistics and Director of the Centre for Comparative Psycholinguistics at the University of Alberta. His research focuses on the representation and processing of morphologically complex words in the mind across populations and languages. He is Director of the project ‘Words in the Mind, Words in the Brain’ and, together with Gonia Jarema, he is Editor of the journal The Mental Lexicon. Manas K. Mandal is currently the Director, Defence Institute of Psychological Research, Defence R&D Organization, India. Earlier he was a Professor at Indian Institute of Technology, Kharagpur. His area of research interest includes Experimental Neuropsychology and Hemispheric Lateralization. He has to his credit 4 books, 11 chapters for various books, and 81 research papers published in both national and international journals. Deepak Mishra is pursuing his Ph.D. in the Department of Electrical Engineering in Indian Institute of Technology Kanpur, India. He obtained Masters in Technology in Instrumentation from Devi Ahilya University, Indore in 2003. His major field of study is Neural Networks and Computational Neuroscience. (Home page: http://home.iitk. ac.in/~dkmishra) R.C. Mishra is Professor of Psychology at Banaras Hindu University. He is a D.Phil. from University of Allahabad, India. He has been Research Fellow and Visiting Professor at Queen’s University (Canada), University of Geneva and Jean Piaget Archives (Switzerland). His principal interest is in cultural influence on human development, and he has contributed numerous articles to professional journals and books, both in India and abroad, in the fields of cognition, acculturation, schooling, and cross-cultural studies. He is the co-author (with J.W. Berry and D. Sinha) of Ecology, Acculturation and Psychological Adaptation: A Study of Adivasis in Bihar, and co-editor (with J.W. Berry and R.C. Tripathi) of Psychology in Human and Social Development: Lessons from Diverse Cultures.
430 Advances in Cognitive Science K.P. Miyapuram received M.Tech. (AI) and M.Sc. (Electronics) degrees from the University of Hyderabad, Hyderabad, India. He is now a Graduate student in the Department of Physiology, Development and Neuroscience, University of Cambridge, UK. His research interests include fMRI, Reward Systems in the Brain, Computational Neuroscience and Cognitive Science. Sachio Nakamizo is Ph.D. and honorary Professor of the Department of Psychology at Kyushu University. His research fields include visual perception and motor behaviour such as eye and head movements. He has published papers in journals such as Vision Research, Perception, Perception and Psychophysics, and so on. Prakash Padakannaya is Professor of Psychology, University of Mysore, Mysore. He received his M.Phil. and Ph.D. degrees from Utkal University, Bhubaneswar, Orissa for his work on development of reading, during which time he also got trained at MRC Cognitive Development Unit, London (UK). He is a recipient of Mombusho Post Doctoral Fellowship (Japan), Fulbright Post Doctoral Fellowship (US) and AIEJ Fellowship (Japan). Presently he has numerous national and international research collaborations with him. His current areas of specialization include Reading and Writing Processes, Orthography and Literacy Acquisition, Cognition, Neurolinguistics, Kanji Processing, Developmental Dyslexia and Child Development. V.S. Chandrasekhar Pammi graduated with a Ph.D. in Computer Science from the University of Hyderabad, India. He has M.Sc. (Physics) and M.Phil. (Computational Techniques) degrees from University of Hyderabad. He is currently a Post Doctoral Fellow in the Computation and Cognitive Neuroscience Laboratory, Department of Psychiatry and Behavioural Sciences, Emory University School of Medicine, Atlanta, GA, USA. His research interests include fMRI, Reinforcement Learning and Neuroeconomics. Raja Parasuraman is Professor of Psychology, George Mason University, Virginia, USA. He holds a B.Sc. (Hons) in electrical engineering from London and a Ph.D. in Psychology from Birmingham. He has written extensively and is the author of several books. He has been involved in research on attention, ageing, automation, cognitive neuroscience and workload. Vani Pariyadath is currently a research scholar at Baylor College of Medicine, US. She is a former student of the Centre for Behavioural and Cognitive Sciences, University of Allahabad, India. She has worked on humor perception and is currently working in the area of time perception. Ype H. Poortinga is Emeritus Professor of Cross-cultural Psychology at Tilburg University in the Netherlands, and at the Catholic University of Leuven in Belgium. His most consistent research interest has been in the conditions under which data obtained in different cultural populations can be meaningfully compared. He has been president of the International
About the Editors and Contributors
431
Association for Cross-Cultural Psychology (IACCP), the Dutch Psychological Association (NIP) and the European Federation of Professional Psychologists Associations (EFPA). Suparna Rajaram is Professor of Psychology, Stony Brook University. Rajaram’s research focuses on human memory and amnesia. She is a Fellow of the American Psychological Association and the American Psychological Society, and currently an elected member of the Governing Board and Chair of the Publications Committee of the Psychonomic Society. She has served as Associate Editor of Memory and Cognition (1998–2001) and Psychological Bulletin (2002–04). Her research and professional activities have been funded by the National Institute of Mental Health and the National Science Foundation. S. Bapi Raju is Reader, Department of Computer and Information Sciences, University of Hyderabad, India. He received B.E. (Electrical Engg) degree from Osmania University, and M.S. (BME) and Ph.D. (MSCS) degrees from University of Texas at Arlington. He was an EPSRC Post Doctoral Fellow at University of Plymouth, UK, and worked as Researcher in the Kawato Dynamic Brain Project, ATR Labs, Kyoto Japan. His research interests include Cognitive Science, Machine Learning, Bioinformatics and Neural Networks. K. Ramakrishna Rao is a philosopher, psychologist and educationist with vast experience in national and international arena as a teacher, researcher and administrator. Prof. Ramakrishna Rao studied philosophy at Andhra University and with Richard McKeon at the University of Chicago and received Ph.D. and D. Lit. degrees. He worked with Dr J.B. Rhine at Duke University and later headed his famous Foundation for Research on the Nature of Man as its Executive Director. In a career spanning over several decades, Professor Rao made significant contributions to Gandhian thought, philosophy of mind and cross-cultural studies. He has published nearly 200 research papers and 12 books. Prof. Rao is currently an Editorial Fellow in the Project History of Indian Science, Philosophy and Culture. He has held over a period of 30 years various teaching and administrative positions and started a number of new programmes there. He has also undertaken various important academic assignments abroad and taught at several universities in US, including Duke University, University of North Carolina at Chapel Hill, University of Tennessee at Chattanooga and California Institute for Human Science. In addition to teaching and research, Prof. Ramakrishna Rao served in several top level administrative, executive and advisory positions. He is currently President of the Institute for Human Science and Service, an experimental institution in Gandhian education to link learning and classrooms with community centres of service. He has received numerous national and international honours and recognition for his work. Sudipta Ray works as Faculty Associate in the Indian Institute of Information Technology, Allahabad, India. He obtained Masters of Technology in Electrical Engineering from Indian Institute of Technology Kanpur, India in 2005. His major field of study is Computational Neuroscience.
432 Advances in Cognitive Science Indramani L. Singh is a senior faculty at the Department of Psychology, Banaras Hindu University. He was the country’s first UGC Research Scientist ‘B’ (equivalent to Reader) and Research Scientist ‘C’ (equivalent to Professor) in Psychology in 1988 and 1993, respectively. He was a Post Doctoral Fellow and visiting scientist at the Cognitive Science Laboratory, Washington, DC, USA from 1990 to 2000. He has published two books and over 100 research papers. Anindya Sinha is a senior Fellow at the National Institute of Advanced Studies, Bengaluru, India. His current research interests are in the areas of behavioural ecology and cognitive psychology of primates, population genetics, evolutionary biology, conservation biology and the philosophy of biology. He has a doctorate in molecular biology, and has earlier worked on the biochemical genetics of yeast, the social biology of wasps and the classical genetics of human disease. He is also interested in biology education and popularization of science, and has lectured extensively in a variety of educational and research institutions. Richard Sproat received his Ph.D. in Linguistics from the Massachusetts Institute of Technology in 1985. Since then he has worked at AT&T Bell Labs, at Lucent’s Bell Labs and at AT&T Labs–Research, before joining the faculty of the University of Illinois. Sproat has worked in numerous areas relating to language and computational linguistics, including syntax, morphology, computational morphology, articulatory and acoustic phonetics, text processing, text-to-speech synthesis, writing systems and text-to-scene conversion, and has published widely in these areas. His most recent work includes work on the prediction of emotion from text for TTS, multilingual named entity transliteration, and the effects of script layout on readers’ phonological awareness. Nishi Tripathi is interested to study the area of disability and its impact on cognitive and affective styles. She is also interested to work in the area of cognitive disorders, particularly learning disabilities. She completed her Ph.D. in Psychology from Allahabad University. She has worked on the cognitive and affective styles of females with physical disability as a part of her doctoral thesis. She is currently working as a counsellor in the Department of Psychology, University of Allahabad, India. Muhammad Kamal Uddin is Assistant Professor of Psychology, University of Dhaka, Bangladesh. He did his Ph.D. in Psychology at Kyushu University, Japan. He has authored several scientific papers published in Psychologia, Vision Research, Spatial Vision, and so on. He is Director of the Bangladesh Institute of Psychological Services and Human Resource Development (BIPSHRD). He aims to continue research in the area of visual perception. Jyotsna Vaid is Professor of Psychology, Texas A&M University. She has authored several publications on bilingualism, including a 1986 edited book, Language Processing in Bilinguals (Erlbaum Press), a 2006 book chapter on bilingual humour in Bilingual Minds (Multilingual Matters Press), and a forthcoming meta-analysis of bilingual language lateralization in
About the Editors and Contributors
433
Laterality. Her current research examines writing system effects on word recognition in different languages (Hindi, Urdu, Kannada, Chinese, Korean, Japanese, Spanish, French and English). Aparna Vajpayee is Assistant Professor, Department of Psychology, AMITY University, Noida. She obtained M.A. from Banaras Hindu University, Varanasi and Ph.D. from Chaudhary Charan Singh University, Meerut. She is interested in the study of psychological development of children of the disadvantaged and weaker sections of the Indian society and has done significant work with respect to intervention for their cognitive development. Her area of interest also includes the study of physical and mental health of tribal women. She has contributed several articles to edited volumes and research journal in the fields of cognition, health and community development. Her current interest is focussed on the study of organizational culture of national and multinational companies, stress due to isolation, emotional intelligence, and so on. She also practices HRD as part time activity. Along with this, she is also co-editor of Amity Journal of Psychology and Forensic Science. Jürg Wassmann is Professor and Director, Institute of Ethnology, University of Heidelberg. He obtained his Ph.D. in Social Anthropology from the University of Basel. He has worked as a faculty member in University of Tuebingen (Germany), University of Zuerich (Switzerland), and University of Basel. He is interested to work in the fields of Cognitive Anthropology, Ethnopsychology, Space and Time, Anthropology and Cognitive Sciences. He has worked on various research projects and has extensive field experience. He has numerous publications to his credit including monographs, anthologies and journal publications. Abhishek Yadav works as Assistant Professor in the Department of Electrical Engineering, College of Technology, G.B. Pant University of Agriculture and Technology, Pantnagar, Uttaranchal, India. He obtained Masters of Technology in Electrical Engineering from Indian Institute of Technology Kanpur, India in 2005. His major field of study is Computational Neuroscience.
Subject Index Abhidhamma, 384, 407, 408 Abhidammasangha, 412, 413 Aborginal North America, 304 absolute encoding, 270, 272, 273, 275–77 absolutism, 303 acculturation, 274, 275, 305 action potential generation, in Morris–Lecar model, 174–77 activating-protein I (AP-I), 114 adaptation, 303, 305 , 418 adult language, 61 Advaita Vedanta, on avidya and maya, 404 on consciousness, 422, 425 on deep sleep, 404–05 age/ageing, 23 effect on cued sustained attention, 29 and proficiency effects on bilingual brain, 131 aksara, 66 and vowels, 67 akusala cetasika, see immoral elements alaya, 412, 421 alayaijnana, concept in Yogacara School, 411, 412 algorithms, learning, 157–58 use in humour 202 alobha, 418 ambiguous novel compounds, 77 processing of, 78–79 amnesia(ic), anterograde, 45 associative learning in, 44–49 population, study of, 46 syndrome, 45 amoha, 418 amygdala, 118 analgesia, 393 analysis of variance (ANOVA), 10, 14, 27, 263, 296, 333, 367
brain lateralization, 285, 289 for landmark conditions, 39 and measurement of peripheral lateralization, 284 of sensitivity index, 25 animals, as intentional systems, 90 social cognition among, 6 anthropology, 305 symbol, in Hindi and Kannada, 68 aphasia, in bilinguals, 123–24, 125 crossed, 124 polyglot, 124 Arithmetic Test, 373, 374, 378 artificial intelligence, 181, 188 Artificial Neural Nets (ANN), 182–84 arousal theory, 21 arupa forms, 421, 422 asamprijnata samadhi, 396 Asian languages, 5 Asperger’s syndrome, 208 associated automation, 196 associative learning, 4 asymbolia, 212 at-risk children, and social play, 368 Atharbasikhopanishad, 401 atman, in Vedanta, 411 attention, 3–4, 413 bias, 141 in cognitive psychology, 19 divided, 19 selective, 19, 21–22, 28, 29 sustained, 19, 20, 21–22, 23, 28, 29 covert, 22 eye movements and, 2 focused, 28 orienting, and cued sustained, 19–29 ‘shift’, and memory averaging, 4, 33–34 and reduced foveal bias, 39, 41 theories of, 4 and WM, 4 attribution, intentionality and, 90–92 auditory, bias, 141
Subject Index and temporal processing in dyslexic children, 354 Australian aborigines, 244 autism, and humour, 208 Automatic Progressive Parsing and Lexical Excitation (APPLE) model, 78, 85 II model, 81 averaging technique, 129 avidya, 404, 405, 423 awareness, context of, 397, 398 paradoxical, 387, 389 paranormal, 387, 390 pathological, 387, 389–90 peripheral, 387, 388–89 primary, 386, 388, 387–88 varieties of, 397–88 axon(s), 161, 171 action potential in, 173–76 multicompartment model, 173–74 Bali Indonesia, 249 egocentric language use in, 275 Kaja-Kelod study in, 244–45 spatial language and encoding in, 266 traditional culture in, 274, 275 Balinese cognitive style, 272 Balinese cosmology, 245 Balinese orientation system, 244–45, 268, 274 basal forebrain region, 46 basal ganglia, 8 behaviour(al), analysis, 93, 94 asymmetry in function, 111 change in Hodgkin–Huxley model, 162 development and expression of, 303 disorder, extrovert children and, 360 domains of, 227 experiments, 3, 118 laterality, 125–26 and imaging methods, 131 measures, 10 observations, 89 patterns, 289 reading, 93 results, 16 training, 119 bhavanga, 408–11, 421 bifurcation analysis, 161
435
bilingual population, 110–11, 122–23, 128 code-switching behaviour of, 132 ERP research, 127 language processing in, 122 and monolinguals convergence, 130–32 studies, on brain-injured, 123–25 on brain-intact, 125–30 use of language by, 132 biomorphemic activities, 75, 81 Birhor tribals, 292 Block Designs Test, 282 Bohr’s model, 179 bonnet macaques, 6, 94–95 all grooming supplants among, 95–98 detective action by, 98–101 female, dominant, 94–97 mental representation of attributes, 96–97, 103, 104 motives and a belief system, 97 predator alarm calls by, 91–92, 102 social interaction among, 98–101, 103 social knowledge, 94, 95–96 tactical deception, 98, 102–04 triadic interaction, 96, 97 visual perspective-taking, 101, 103 ‘bottom up’ approach, 374 Brahmabindu Upanishad, 401 Brahmopanishad, 401 Brahmi-derived script, in India, 64, vowels in, 66 brain, area, 5 as black box, 181 damage, resistance to, 45 disorder patients, 110 function, 16 injured bilinguals, 123–25 lateralization measure of, 283, 284, 286 left hemisphere use of, 280 lesions, 109 and nervous system, 109 neurons in, 181–82 plasticity, 123 processing in left hemisphere, 137 right dominance of, 287 right hemisphere use of, 280 brain glucose metabolism, 123 brevity, role in humour, 209
436 Advances in Cognitive Science Brihadaranyaka Upanishad, 400 Broca’s area, 110 Buddhism, and consciousness, 400, 406–21 ideas of illusion, 383 prominence of, 407 schools of, 406 Bunutan region, Bali, 271 cable equation, 171–73 cable theory, 169, 171–72 Cambridge anthropological expedition, 219 Caraka Samhita, 402 cardinal directions, 246 cartoons, 200 catenation, direction, 68, 71 types of, 64, 65, 67 cellular homeostasis, 114 cellular resolution mapping, 114–15 cercopithecine, 103 cerebellum, lateral, 8 cerebral cortex, 137 cerebral dominance pattern, 144 cerebral functional asymmetry, 125 cerebral hemisphere, left, association with language, 122 cetana, 413, 414 cetasikas, 412 Chakrapanviyakhya, 402 child(ren), abuse, and dissociative identity disorder (DID), 394 characteristics, assessment of, 366 with dyslexia problem, 318, 319 with learning-related problems, 344 chimeric facial stimuli, 141 chimpanzees, attribution of beliefs and desires by, 94 mastering new roles by, 94 Chinese abacus, 374 Chinese language, catenators in, 64, 65 characters, 65 and English, 235 left-branching preference in, 84 triconstituent characters in, 84 Chomsky’s notion, 186 chunks/chunking, 8, 14, 16 notion of, 7 process, 9
classroom experience, and play for children, 358 quality of 358–59 cluster analysis, 16, 183–87 ‘cognition’, 3, 137, 186, 290 cross-cultural research on, 233–37 process, 126, 200 and social behaviour, 302 tests, 298 variables, 234–37 cognitive abilities, 228, 229, 233–34 in non-human primates, 89 cognitive abnormalities, 389–90 Cognitive Assessment System (CAS), 319, 346–47 cognitive bias, 140–41 attentional, 141 auditory, 141 visual-field, 140–41 cognitive competence, 302–11 cognitive consciousness, 411 cognitive deficits, and disorders, 319 cognitive dimensions, 293–94, 296–99 cognitive functions, 291 cognitive neuroscience, 107, 109–11 domain of, 3 experiments in, 179–88 of humour perception, 207–09 cognitive neuropsychological rehabilitation, 319 cognitive profiles, of children with dyslexia, 343–55 cognitive process, 354 development of, 317 cognitive stimulation programmes, 317 of rural children, 371–79 Colour Cancellation Test, for children, 373 colour blindness, 229 communication disorder, among children, 360 competitive-integration model, 53, 54 complex primitive linguistic, and processing, relevance of, 190–97 compound words, 5–6 constituents of, 74, 76–79 embedded, 85 left-branching interpretation of, 79–82, 84
Subject Index as morphological structures, 86 parsing of, 76–87 comprehension system, depicted events and, 60 computation(al), algorithms, 157 models, 3, 157–58, 179 of humour, 205–07 work, in linguistic analysis, 195 computer models and simulation(CMS), 180, 181, 187 connectionist approach, to brain, 181, 185 models, 157 conscious state, countless, 396–97 consciousness, 89, 90, 383 altered state of, 391 Buddhist phenomenology of, 406–21 categories of, 422 cognitive function of, 387, 388 ‘cosmic’, 395 elements of, 412–15 forms of, 415–17 higher states of, 418–21 perceptual, 90 pure, 386 reflective, 90 studies in the West, 399 taxonomy of, 399–400 and unconsciousness, 408–12 consonant(s), confusion, 324 in Devanagari script, 64 sequence, 66, 67 constraint-based models, 54 Lexicalist Theory (CBL), 195 context-free grammar (CFG), 190, 191, 196 contextualization, 294, 296–97, 299, 300 coordinated processing, 59–61 corpus callosum regions, 133 cortical electrical stimulation, 122, 124–25 cortical maturation, 118 cortical neuron model, 157, 167, 169, 170, 177 cortical sound discrimination, 339 cosmos, 256 cradling bias, 140 Creativity Test, 373, 374, 378 cross-cultural identity/differences, 228, 229, 232, 235 cross-cultural psychology, 248, 302, 303, 305, 306
437
cross-cultural research, 219, 223, 237 cross-cultural studies, emic approach, 243 etic approach, 243 crossed and nested dependencies, processing of, 196–97 cue/ cuing, 4, 264 facilitation, 28 validity effects, 27, 28 vigilance, 23 culture/cultural, and behavioural difference, 306 bias, 235–36 and cognition, 219, 220, 223 comparative research, 225, 230 diversity, 305, 306 ecology, 304 and their institutions, 304 invariant domains in, 229 language, and hemispheric dominance, 279–87 mediation theory, 234 specific knowledge, 236 as superorganic, 303 stereotyping, 232 study of, 219 transmission, 307 cultural dimension, 295–96 political stratification, 290 social stratification, 290 socialization, 290 type of family, 290 daydreaming, and hypnagogic imagery, 392–93 deception, interaction, analysis of, 93 tactical, 89 decrement function, in vigilance performance, 19, 20 deep sleep state, 404–05, 423 delusions, 390 depicted events, coordinated influence of, 56 double head activation of, 81 importance of, 57, 58 verb-mediated influence on, 57 detection accuracy, in reaction time, 22 Devanagari script, 5 consonants in, 64
438 Advances in Cognitive Science vowels in, 64 writing of anusvara in, 68 development(al), definition of, 317 disorders, 317 Dhammasangani, 414 Dharma, 419 diacriticity, 68 dichotic-listening technique, 141 Differential Item Functioning (DIF), 234 diencephalons, 46 dissociative identity disorder (DID), 394 District Primary Education Programme, 371 dosa, 418 dream(ing), 391–92, 422 consciousness, 423 Dutch language, cross order in, 196, 197 triconstituent compounds in, 80 Dutch school system, 359–60, 368 dynamical models, 157 dyslexia in children, 319 cognitive profile of, 343–55 definition of, 321 developmental, 317–18 intellectual functions of, 347 and reading comprehension, 350 research on, 317–18 symptoms of, 321 earedness, 139 Early Childhood Environmental Rating Scales (ECERS), 362, 366 eco-cultural framework, 305–09 ‘eco-cultural’ model, 289–90 ecological anthropology, 304 ecological environmental psychology, 305 ‘ecological perspective’, in culture system, 289 ecological psychology, 308 ecosystem approach, 304 egocentric encoding, 243 egocentric language system, 247, 269 in Geneva, 220 use by boys in Hindi-medium schools, 264 eidos research tradition, 219 Ekaggata, 414 Elmo, JAPE and, 205 electroencephalogram (EEG), 109
electroencephalographic research, 127 electrophysiological measures, 126, 131 embedded pushdown automata (EPDA), 196, 197 Embodied Construction Grammar, 54, 60 emotion(al), 137 domain, 228 and facial expression, 143 processing, 110 role in humour perception, 207 spontaneous expression of, 143 enculturation, 305 endocrine responses, 116 English triconstituent compounds, 79, 80 English reading, errors in, by dyslexic children, 348, 349 English writing, errors in, by dyslexic children, 350 environment(al), and behavioural outcomes, 308 ‘determinism’, 304 Ephaptic Interaction, 176 epileptic individuals, exposed cortex in, 124 equivalence, analysis ,of 227 conception, 227 full-score, 226, 227, 229, 234 metric, 226, 227 structural, 226, 227, 232, 233, 237 ERP studies, 110, 207 ‘ethos’, 219 etiologies, 47 European languages, egocentric, 244 event, -related brain potential, 126–27 -related brain wave components, 127 rate, 20 excitatory post-synaptic potential (EPSP), 171 expectancy theory, 21 explicit memory function, in brain, 44–46 eye. -movements, 5, 42 and attention, 22 involuntary, 42 utterance and, 55 tracking, 5, 51 in depicted events, 56–59 European languages, 5 eyedness, 139
Subject Index face/ facial, asymmetry, 142 behaviour, 142, 148 expression, 139, 140, 142–44 recognition, in sheep, 118 responses, to humour, 201 facedness, 139–40 and facial expression, 139 false alarm theory, 204 filter theory, 3, 21 family background, and language and encoding, 273–74 FitzHugh–Nagumo model, 157 analysis of, 161, 162–67, 177 bifurcation diagram for, 164, 166 fixations, and utterances, 5 ‘flashed landmark’ condition, 36, 37 foveal bias in, 38 Fodor’s model, of mind, 52 footedness, 138 right-foot bias, 138 fourth state, 405–06 foveal bias, 4, 32–42 in ‘disappeared landmark’ conditions, 39 in landmark conditions, 36, 37 visual attention and, 34–38 frames of reference (FoR), in study of language, 219, 220, 274 ‘frames’, concept of, in humour perception, 204 frontal lobe function, 132 functional coordination deficit (FCD), 331, 332 functional imaging methods, 131 functional magnetic resonance imaging (fMRI), see magnetic resonance imaging functional transcranial Doppler sonography (fTCD), 127, 128 G language, and G encoding, 285, 286 gender, bias in Hindi and Sanskrit schools, 263–64 generalizability theory, 227 generalization, levels of, 227–30 validity constraints on, 230–32 genetic transmission, 305 genetics, 305
439
Geneva, spatial language and encoding in, 266 study on egocentric language in, 275 use of relative terms in, 249–50 geocentric language system, 247, 249 in Bali, 220, 245 use by Sanskrit school children, 261, 262, 264 German language, nested order in, 196, 197 triconstituent words in, 80, 81 left-branching, 82 right-branching, 83 Global Theory of Verbal Humor (GTVH), 203–04 global workplace theory, 383 glyphs, arrangements of, 64, 66 grapheme-phoneme correspondence, deficit, 323 Graphical Phonological Humor Identification Algorithm (GraPHIA) model, 205–07, 211–13 Gray Oral Reading Test, 323 ‘Great Divide’ theories, 235 grey matter in bilinguals, 133 in skilled musicians, 123 Gurukul education, 258, 260 Guugu Yimithirr Australian Aborigines, 276 HAHAcronym project, 205 haemodynamic, and electrophysiological studies, 122 imaging techniques, for neural activity, 128 measures, 127 studies, 129 hallucination, 389, 390, 393 hand behaviour, 146 handedness, 138, 146, 282 bias in, 144 culture, 111 incidence of left-, 138, 144–45, 147 right-hand preference, 144, 146 theories of, 146 hemifacial bias, 143 hemispheric dominance, 110, 285 hemispheric lateralization, 252, 282–85 Hidden Words Test, 294, 296
440 Advances in Cognitive Science Hindi reading, errors in, by dyslexic children, 348–50 Hindi-speaking population, dyslexia among, 318 Hindi writing, errors in, by dyslexic children, 351 Hinduism, state of consciousness in, 400–01 hippocampus, 46 in sheep, 118 Hodgkin–Huxley (HH) model, 157, 161, 162, 177 analysis of, 163 human development, theories of, 242 humour, behaviour, side bias in, 137–48 classification of, 200 generation, 207 and laughter, 200 nonsense, 200 non-verbal, 200 processing, 200–01 taboo, topics in, 210, 211 humour perception, 158–59 cognitive neuroscience of, 207–09 computational approach to, 199–213 theories of, 201–05, 210 incongruity, 201–03, 210, 211 KK Joke analyzer, 210 Katz’s model, 210 linguistic, 203–04 relief, 204–05, 210, 211 semantic script 210 superiority, 210, 211 violation, 210, 211 Yokogawa’s model, 210 hyperset, 9, 14 hypnagogic state, 392–93 hypnosis, 391, 393 identity information, 32 immediate early genes (IEGs), 113-114, 116 c-fos, 113, 114, 115, 117, 119 maps, 117 in visual cortex, 118 zif268, 113, 114, 115, 117, 118, 119 immoral elements, 414–15 immunopositive neurons, cellular morphology of, 117
immunostained neurons, 115 impairment, post-morbid pattern of, 124 impermanence, 407 implicit measures, of memory, 44–45 implicit memory, 46 inclusive and funny domains, 228 incongruity theory, of humour perception, 201–13 Paulos catastrophe models, 201, 203 Shultz’s Sudden Disambiguation models, 201, 202 Suls’s two-stage models, 201, 202 Veatch’s violation model, 201, 202 incremental thematic role assignment, 57, 59 incremental utterance comprehension, 59 Indic writing systems, 67 individual connectedness, 293 Indonesian-speaking children, geocentric term use by, 269 knowledge of cardinal directions by, 268 inducible transcription factors (ITFs), 113–14 induction experiments, 118 infant cognition, 317 inferences, 231 inferotemporal cortex, 118 information, and action, 3 biases, 53 encoded, for memory, 45 processing, 110, 329 inhibition, 21 facilitation and, 22–28 theory, 4 integration, 294–95, 297–98, 300 intellectual functioning, 377–78 intelligence, 229, 234 batteries, 233 measure of, 224 quotient, low, 344 tests (IQ), 224, 234 intentional stance, 90 intentional system, of human beings, 92 first-order, 91 higher-order, 92 second-order, 91 third-order, 91, 92 zero-order, 91, 92 interference, role of, 46
Subject Index intervention conditions, in schools in Netherlands, 368 introvert children, 360 jagrit (waking state), 400, 401–02 Jainism, consciousness in, 400 jayarita, 401–02 jivitindriya, 414 jokes, 20–22 Joke Analysis and Production Engine (JAPE), 207 Jun phosphoprotein family, 114 Jvanamukti, 405, 406 Kaja-Kelod, study in Bali, 244–45, 268 kama loka, 415, 416, 421 forms of consciousness in, 415, 416, 417–18 kangin-kauh, 245, 268 Kannada script, 5, 64, 66–67, 70–71 Karnataka, Nali-Kali approach to education in, 371 Kashi Mummulshu Bhawan Sabha, 259 Kathmandu, Nepal, monolingual and bilinguals in, 2650 Katz’s model, on humour, 212 key-press RT, 10, 12, 13–16 in complex sequence, 14 KK Joke Recognizer, 205–06 knowledge, about others’ mind, 6 resources, and joke difference, 203 verb-based thematic role of, 58 landmark, attraction, 33 bias, 33, 34, 38 conditions, 4, 33–36 and foveal bias, 36 ‘repulsion’, 33 language, 3, 5, 110, 267, 279 acquisition theories of, 60–61 comprehension, 5 disorder, 110, 318 and encoding, 283 in Bali, 272–73 impairment, 124 lateralization, in brain-intact bilinguals, 122 localization, within brain, 124
441
processing, 5, 158 in bilinguals, 122–33 procedural account of, 53–55 psycholinguistic accounts of, 52 scripts, 5 socialization, 279 and spatial encoding, 247 spontaneous of, 269 status of, 255 system, and visions, 52–53 and vision integration, 51–52 Lankavatara Sutra, 411 Latin alphabet, 64 laughter, 204, 209 humour and, 200 learning, 3, 10, 48, 236 disability, 4, 317, 321–39, 343 among children at-risk in Netherlands, 359–67 left hand use, as inauspicious, 264 lexical item, 190, 191 lexical processing system, 81, 86, 79 and cognition, 75 lexical and structural ambiguities, 185–86, 195 lexicalized tree-adjoining grammar (LTAG), 191, 193–97 processing issues in, 194 Light Bulb Jokes Generator (LIBJOG), 205 linguistic coding, non-egocentric, 244 linguistic communication, among humans, 92 linguistic formalism, 54 linguistic information, obtaining, 5 linguistic relativism/relativity, 243–44, 272, 285 hypothesis, 219, 247 and linguists, 243–44 linguistic theories, in humour perception, 203–04 linguistic and visual processes, 54, 354 literacy, and illiterate groups, 235 and western form of schooling, 235 lobha, 418 Locating Object Test (LOT), 294 location, domain of, 190–94 sense of, 32
442 Advances in Cognitive Science locomotive therapy, 360 Lokottara citta, 420 lokottara state, 421 Macaca radiata, 6 see also bonnet macaques machine learning technique, 16 Magnetic Resonance Images (MRI) scanner, 9, 127 functional magnetic resonance imaging (fMRI), 110, 127–30 magnetic source imaging (MSI), 128 magneto-encephalography (MEG), 127–28 magnetocephalographic measures, 127 Mahabharata, 403 Malaviya Shiksha Niketan, 260 Malayalam script, 64 manas, 412 Mandukya Upanishad, 402, 404, 405 manipulation condition, 230 mathematics(al), children’s interest in, 358 knowledge, 363, 364, 367, 368 Max Planck Institute for Psycholinguistics, CARG at, 244, 247, 251, 277 maya, 404 Mazlack’s Joke Recognizer, 206 Madhyamika School, of Buddhism, 406–07 medial temporal lobes, 46 amnesia, 46, 47 functions of, 48 role of, in learning, 47 meditation, 301, 393–94 Buddhism on, 418 memory, 3, 4, 44, 243 averaging, 4, 33, 34 bias, 33 buffer, 7 explicit, 4–5 factors operating, 44 functions, loss of, 44 implicit, 5 intact population, 46 for past experience and future plans, 44 span, 235–36 tasks, 45 mental phenomena, 109 mental rotation, 318, 326, 332
mental state, 93, 418 re-evaluation, of 94 mentalism, 93, 100 Milinda Panha, 409, 410 Mimamsa School, of Buddhism, 403 mind, other, 89–104 reading, 93 Mnemonic Sentence Generator, 205 Moha, 418 moksha (liberation), 406 and turiya, 405 molecular mapping, in visual system, 113–19 strength and caveats of, 114–16 momentariness, philosophy of, in Buddhism, 407 monolinguals, and bilinguals congruence, 130–32 canonical pattern among, 125 moral elements, 415 moral virtues, 186, 187 morphemes, in compound structures, 74, 79–86 morphological parsing, 82–84, 86 morphological word forming system, 74, 75, 77 Morris–Lecar model, 157, 161, 172–77 motor, action, 137 bias, 111, 138–40 cradling, 140 earedness, 139 eyedness, 139 facedness, 139 footedness, 138 handedness, 138 multimorphemic lexical processing, 79 multiple personality disorder, 390 multiple stimulation condition, 110, 116 Mumukshu Bhawan, 259 musical experience, and neural sensitivity, 123 mystical awareness, 395–96 mystical state, 395 naming impairment, 124 Nand Lal Bajoria Sanskrit Mahavidyalaya, 259 Nandabindu Upanishad, 401 near infrared spectrometry (NIRS), 127
Subject Index Nepal, study of Newell children in, 245–46 study of spatial orientation and cognitive performance in 245–49 Netherlands, children and play-session, 361–63 learning difficulties of at-risk children, 359–67 neural activity, 157 neural attention units, 20 neural damage, 45 neural expression, 115–16 neural network, 158, 184, 188 Neural Network Damage, 16 neural plasticity, 317 neural process in bilinguals, 109, 110 neural substrates, of complex visual function, 118–19 of language processing, 122–33 neurobehavioural studies, 126 neuroimaging studies, 126–29, 207, 209, 280 neurological deficits, 326 neurons, in brain, 157, 161, 181–82 neuropsychological studies, on humour perception, 208 neuronal electrical activity, 127, 128 neuropsychology, 131 nervous system, 109 sensory function of central, 116 Newman Keuls test, 296 Nijmegan tasks, 246, 264 NIMHANS, 318, 345–47, 349, 373 nirvana, 420–22, 409 and Brahman, 424 nonsense word parsing, 81 noradrenergic input, role of, 117 norm-referred teat, for children, 362 novel compounds, 75, 76 NSEW system, 261, 268 Number Cancellation Test, 374 Nyaya School, 404 Object Enumeration Test (OET), 294, 295 ocular dominance columns, 117 oral reading, motor programme for, 353 Oraon tribals, in Chotanagpur, 292 Oriya script, 64 orthographic memory, 355 problems in, 327
443
orthographic syllables, 66 orthography, 79 Panini Kanya Mahavidyalaya, Veda-based education in, 260 parapsychology, 383 Parkinson’s syndrome, 208 parsing approach, 77–78 particular elements, six, in Buddhism, 414 passive vocabulary test, for children, 362 Pearson correlation, 273, 274, 284 pedagogical quality, among children in Netherlands’ school, 361–63 perception, as learning network, 182–83 encoding, 325 kinds of, 409 process, 3 sensitivity, 21 stimulus-response model of, 32 Phase Plane analysis, 164, 167, 169 phoneme, 68, 70, 71 phenomenology states, of consciousness, 384 phonetic cues, 349 phonological coding, 319 323 344 349–51 355 phonological deficits, 323, 343, 351, 352 phonological information for dyslexic children, 354 phonological joke perception, 200, 206–07, 212 ‘Piagetian’ tasks, 243 plausibility, notion of, 232 point neuron models, analysis of, 161–77 polyglot aphasia case, 122 positron emission tomography (PET), 109, 110, 127, 129–30 postvocalic nasals, 68 prajna, 404 prakriti, 404 Prapancasara Tantra, 401 Pratham, initiative in Maharashtra, 371 ‘primary vigilance’, 21 priming, 47 in amnesia, 45 primitive mind, 89–90 ‘probabilistic functionalism’, 307 Progressive Matrices, 374 protein synthesis, 113
444 Advances in Cognitive Science psychic energy, and laughter, 204 psychoanalytic theory, 204 psycholinguistic phenomena/ theories, 51, 52, 158 psycholinguistics, 195 psychological differentiation, 281–82, 305, 306 psychological phenomena, 306 psychology, 3, 237 and behaviour, 308 ecological thinking and, 305 psychometric approaches to, 230 psychometric tests, 228, 231, 236 psychosocial development, 377 Pun Analyser model, for humour, 206 Pure consciousness, 394–96 purusha, concept of, 411 pushdown automata (PDA), 196 quantitative analysis, 115 Qu’ranic school, among Arabs, 235 Rapid automatized naming, tests, 351 rate hypothesis theory, 21 Raven’s Progressive Matrices, 363, 373, 378 reading, comprehension, 355 disability, 337, 338 problems in, 352 retarded, 338 skill, 343 test for, 318, 330–31 recall, memory and, 44 receptive language function, 125 recognition, memory and, 44 reinforcement learning, 17 relief theories, of humour perception, 204–05 relativism, 224, 231, 303 repetition, role of, 46 repetitive transcranial magnetic stimulation, 128 representative theory, 403 response times (RTs), 8, 22, 23, 27 retinae, transplanted, 117–18 reversals, errors, 321–22, 328 kinetic, 324, 327, 337 static, 324, 325, 327, 337 studies, 323–29
Rhesus macaques, lack of empathy among, 94 Rig Veda, 256 right hemisphere (RH), role in humour perception, 208 role in language learning, 125 rote, learning by, 374 rupa, 419, 420, 421, 422 Rupavacara citta, 419 sanadhi sate, 396, 401, 419, 422 and turiya, 406 Samkhya system, prakriti in, 404 Samkhya-Yoga, 411 Samprijnata samadhi, 396 sanna, 413, 414 Sanskrit-medium schools, 258–60 Sanskrit spatial cosmology, 256–57 scene, information for sentence structuring, 5, 52 and utterances, 51–61 schizophrenia, 390 script indices, 5, 64–71 semantic(s), 64, 77, 81, 89, 158 analysis of information, 3–4 script of humour (SSTH), 203 syntax and, 186 seizure-inducing brain regions, 122, 124 seizures, in brain, 125 selective basal forebrain damage, 47 self-awareness, 398–99 Senguin Forum Board, 373 sense modality, 20 sensation, experience of, 398 sensitivity index performance, 256 sensory cortex, 118 sentence comprehension, 5, 53, 54 sequence learning, 7 sequential behaviour, 8 serial order, problem of understanding, 7, 8 processing, 328, 355 short-term memory (STM), 7 simultaneous process, in dyslexia, 351, 353 simulation, 3, 179, 180 skandhas, kinds of, 410 skill, 3, 7 slapstick, 200 sobhana/ kusala cetasika see moral elements
Subject Index social behaviour, 302 social cognition, 3 among animals, 6 in wild bonnet macaques, 89–104 ‘social conformity’, 290, 291, 293 social play, 319 and numeracy, 357–68 socialization, 257, 305 ‘societal size’, cultural dimension of, 290, 291 socio-cultural theory, and meaning of play, 357–59 South African children, children with dyslexia, 353 ‘space-based attention’, 4 spatial cognitive development, 246, 279 spatial encoding, 32, 42, 220, 246, 247, 249, 251, 267, 270–72 between Sanskrit- and Hindi-medium schools, 255–65 spatial knowledge, 257, 265 spatial language, and concept development, 242–53 and encoding in Bali and Geneva, 266–77 ‘spatial orientation’, 228 and cognitive performance, 245–49 spatial representation, theory of, 242 special schools, in Netherlands, 360 Specific Learning Disability (SLD), 318, 344 NIMHANS index of, 345, 346, 347 speech therapy, 360 stability, and learning, 4 Standard Progressive Matrices, 345 stereotypes, 6 Steve’s Maze, 247, 252, 261, 262, 271, 275 stimulus, analysis, 3 novelty, in selective cortical activities, 119 response links, 93 selection of, 4 stimulus onset asynchrony (SOA), 4, 27, 40, 41 Story–Pictorial Embedded Figures Test (SPEFT), 281–83, 285, 294, 296, 300 Street Closure Test, 294 Subliminal perception, 389 successive processing in dyslexia, 353, 354 suffixation, in English, 73 Sul’s two-stage model, of humour, 212
445
supertagging, 195 196 susupti (deep sleep state), 400, 404–05 svapana (dreaming state), 400, 402–04 Syllogistic Reasoning Test (SRT), 294 symmetry generalization, in dyslexia 322, 329–36 synaptic density, study of, 123 syntax, semantics and, 186 Takizawa’s model, 205 Tamil script, 64, 67 Tanenhaus, 59 task processing, in the brain, 280 taxonomy, of consciousness, 385–424 teachers’ perception, of learning-related problems among children, 344 temporal-spatial uncertainty, 20 thalamic reticular nucleus (TRN), activation of, 119 ‘theory of mind’, 6, 92, 181 Theravada Buddhism, 406, 408, 419, 425 thought disorders, 390 time, and space, based on sun, 256 response and phase portrait analysis, 162 trait, 225, 227 Transcendental Meditation, 393, 394, 396 transmission variables, 289, 305 tribal children, in Chotanagpur, cultural adaptation by, 220 and cognitive process, 289–301 triconstituent compounds, 79–80 transcranial magnetic stimulation (TMS) technique, 127, 128 Tucker’s phi, 226 turiya, 423 Tzeltal speakers, in Mexico, geocentric system use by, 276 Unfamiliar Words Test (UWT), 294 universalism, 224, 231, 303 and cultural variation, 225 versus Relativism, 219 universality, analysis of, 233 Upanishads, 401, 415 universal causation, law of, 407 universal elements, 413–14 urupavacara citta, 419
446 Advances in Cognitive Science utterances, 5, 53 interplay between scene and scene, 51, 55–56, 61 Vai literacy, in Liberia, 235 vaisvanara, 402 validity, analysis of, 231 discrimination, 232 ecological, 231, 232 interpretive, 231 theoretical, 231, 232 Van Bon’s standard norm, 362 Varanasi, India, geocentric language use in, 249 research study in, 322 Sanskrit schools in, 255 vedana, 410, 413 Vedanta, classification of consciousness in, 422 Vedantaparibhasa, 403 Vedic times, and consciousness, 400 Vedas, study of, 259, 260 verbal learning, 5, 46, 49, 59, 325 verbal spatial relations tests, 354 veridical intuitions, 390 vigilance, decrement, theories of, 20–21, 23 problem of, 19–20 task performance, 29 vinnana, 410 vision/ visual, attention, experiment on shift of, 34–38 code and name code pairs, 328 coding, 344 consciousness, 383 cortex, 118 cues, 24 decoding problem, 325 deprivation studies, 116–17 disruption, 116, 117 embedding language system with, 52–53 field bias, 140–41 function, neural substrates of complex, 118–19 identification, 32
information, reading disability and, 344 and linguistic process, 5 and memory, 41–42 molecular maps of, 113–19 neurons, 117 perception, 110, 325 research, 32 stimuli, 9, 10 in rats, 117 studies on, 116 system, 3, 5 Visual Closure Test (VCT), 294, 297, 300 visuo-motor sequence, 32 organization of, 7–17 visuo-spatial process, for dyslexic children, 351, 354, 355 Vivekachudamani, 404 vocabulary, of adults, 73 Tests, 373, 374, 378 vowels, confusion, 324 in Devanagari script, 64 voxel-based morphometry study, on plasticity, 133 Wad test, for language lateralization, 128–29 waking state, 386–87, 423 Wechsler Adult Intelligence Scale, 332 Wernicke’s area, 110 Wilson–Cowen model, 157, 161, 164, 167, 168, 177 words, centre, 322 processing, 74 working memory (WM), 3, 4, 7, 16, 318 and reading disability, 349 World List Test, 324 writing system, Indian, 5 Yisuddhimagga, 393 Yoga, of Patanjali, 405, 419 Tradition, 396 Yoga-Sutra, 393 Yogacara school, of Buddhism, 406, 411, 421 Yogasikhopanishad, 401
Name Index Aalsvoort, 317, 318 Abbot, L.F., 155, 159 Aberle, D.F., 301 Abhinavagupta, 199 Abrahamsen, W., 155 Adam, J.J., 41 Ahmed, 17 Ajuriaguerra, J., 142 Albert, M., 120 Alvarado, N., 199 Ambady, N., 142 Ames, L.B., 142 Amir, S., 115 Andersson, Y., 138 Annett, M., 144 Aristotle, 199 Armstrong, D.M., 396 Arvanitaki, A., 171 Asthana, H. S., 135, 137, 138, 140, 141, 284 Attardo, S., 201, 203 Aung, S., 406, 408, 411 Ayres, T.J., 234 Baars, B.J., 381 Bach, E., 195 Bacharach, V.R., 27 Baddeley, A., 7, 234 Badian, N.A., 325 Bahri, T., 18, 22 Bajoria, N., 257 Bakan, P., 139 Baker, C.H., 20 Bapi, K.S., 16 Bapi, R.S., 9 Baribique, J., 141 Barker, R., 306 Barnabas, J.P., 375 Barry, H., 308 Barth, 347 Bashinski, H.S., 27 Bates, E., 277 Bavelier, D., 121
Beale, I.L., 320, 321 Beaton, A.A., 139 Bechtel, W., 3, 155 Becker, C., 336, 352 Benke, T., 206 Bennett, J., 92, 303 Benson, H., 392 Bergen, B., 53, 55, 58, 60 Berkowitz, W., 288, 298 Bermu–dez, J.L., 397 Berns, 207 Berry, J., 241, 302, 303, 304 Berry, J.W., 217, 218, 219, 222, 224, 225, 227, 228, 233, 274, 287, 292, 300, 301, 304 Bhattacharya, K.C., 421 Bialystok, E., 130 Biesheuvel, S., 222 Bigsby, P., 321, 325, 326, 328, 335 Binsted, K., 197, 203 Biswal, R., 135 Blackmore, S., 381, 398 Blakeslee, S., 202, 203 Blau, A., 144 Bleichrodt, N., 229 Boder, E., 334, 336 Boldt, E., 288 Bornstein, F., 387 Borod, J.C., 135, 137, 138, 139, 140, 141 Bowey, J., 347 Boyd, R., 302 Bowerman, M., 255 Bradley, L., 320 Bradshaw, J., 324 Brendler, K., 320, 327, 333, 336 Bright, W., 62 Broad, K.D., 116 Broadbent, D.E., 20 Bronfenbrenner, U., 301 Brouwers, S.A., 221 Brown, P., 248 Brownell, H.H., 206 Broyon, M.A., 253, 255
448
Advances in Cognitive Science
Brunswik, E., 305, 306 Bruser, E., 138 Bryant, P.E., 320 Bryden, M.P., 135, 137, 139, 142, 143 Buddhaghosa, 391, 404, 409, 411 Butter, C.M., 31 Byrne, R.W., 96, 99 Call, J., 87 Campbell, D.T., 226, 230 Campbell, R., 137, 138, 140 Cantor, D., 231 Caraka, 396, 400, 401 Caron, H.S., 141 Carroll, J.B., 232 Cattel, R.B., 371 Chakrapani, 400 Chalfonte, B.L., 45 Chalmers, D., 381 Chance, M.R.A., 88 Chandler, L.K., 357 Chang, N., 53, 55, 58, 60 Chaudhuri, A., 111, 116 Chaurasia, B.D., 137 Chee, M., 129, 130 Chen, H.C., 127 Cheney, D.L., 87, 88, 89, 90 Chiarello, C., 278 Childs, C.P., 234 Chokron, S., 139 Christiansen, M.H., 8 Churches, M., 351 Churchland, P.M., 181, 184, 185 Clark, A., 182 Clark, J.W., 171 Clegg, B.A., 7 Clement, D.E., 328 Coates, G.D., 21 Coggins, P., 130 Cohen, J., 373 Cohen, N.J., 45 Cohen, M.M., 342 Cohen, M.S., 330 Cohen, Y., 21 Coioke, W., 138 Cole, M., 222, 232, 233 Cole, P.M., 366
Collins, R.L., 144 Coltheart, M., 317 Conway, C.M., 8 Cooper, L.A., 330, 331 Copeland, J., 180 Coplan, R.J., 357 Corballis, M.C., 320, 321, 324, 330, 332, 334, 336 Coren, S., 137, 142, 143, 144 Correa-Lacarcel, J., 115 Coslett, H. B., 43, 45 Coulson, S., 210 Courtney, S.M., 210 Craner, S.L., 115 Creswell, J.W., 229 Crick, F., 397 Crocker, M.W., 50, 56, 58, 59 Cronbach, L.J., 225 Cummins, J., 350 Cynader, M.S., 115 Daley, P.C., 138 Daniels, P., 62 Darwin, C., 137, 210 Das, J.P., 342, 344, 349, 350, 352, 353 Dasen, P.R., 217, 218, 232, 240, 241, 243, 244, 245, 250, 251, 253, 263, 279, 304 Dasen, R., 277 Davids, C.A.F.R., 405, 411, 412 Davies, D.R., 19 Davis, L., 289 Dayan, P., 155, 157 de Almeida, R.G., 77 De Chateau, P., 138 de Léon, L., 242 Deering, W.M., 336 Dehaene, S., 128, 129 DeJong, R.N., 140, 141 Dember, W.N., 19 Dement, W., 389, 402 Deubel, H., 37 Dennett, D.C., 6, 88, 422 Denny, J.P., 289 Denzin, N.K., 229 Descartes, R., 396 De Silva, C.L.A., 416 Detterman, D.K., 232
Name Index Dharmaraja, 401 Dick, A.O., 31 Dick, S.O., 31 Diedrichsen, J., 32, 41 DiGirolamo, G.J., 107 Dittmar, M., 136 Dolan, R.J., 198, 206, 207 Doya, K., 7 Dressler, W.U., 72 Duffy, E., 20 Dyan, 159 Ehrenwald, J., 401 Ehrman, L., 143 Eibl-Eibesfeldt, I., 227 Ekman, P., 141, 148 Elbert, T., 121 Elliot, A., 357 Ellis, B.B., 223 Ellis, E., 326 Ellis, S.J., 143 Elman, J.L., 156, 181, 182, 183, 184 Emerich, D.M., 206 Eriksen, C.W., 21 Ermentrout, B.B., 170 Eviatar, Z., 139 Fabbro, F., 122 Feldman, 302 Ferguson, G., 307 Finkel, D.L., 31 Fisher, F.W., 322 Fiske, D.W., 230 Fitts, P.M., 7 FitzHugh, R., 159, 160 Flynn, J.M., 336 Fodor, J., 50, 51, 91 Fodor, J.D., 51 Fontaine, J.R.J., 224 Forbes, K., 195 Forde, D., 302 Forman, R., 394 Forster, K., 51 Forster, K.I., 74 Foulkes, D., 390, 391 Franceschini, R., 130 Frazier, L., 51
449
Freeman, W.J., 159 Frenck-Mestre, C., 127, 128, 129 Freud, S., 202, 209 Fuchs, S., 335 Gabbard, C., 136 Gabrieli, J.D.E., 45 Galaburda, A.M., 144 Gamble, J.J., 288 Gangstead, S.W., 144, 145 Ganzeboom, H.B.G., 361 Gardner, H., 206 Garner, W.R., 328 Gaser, C., 121 Gauvain, M., 234 Gazzaniga, M.S., 107, 284 Geissler, H.G., 328 Gelder, R.S., 140 Genesee, F., 120, 123 Gentry, V., 136 Georgas, J., 231 Gerstner, 165 Geschwind, N., 144 Gesell, A., 142 Geyer, T., 320, 321 Gilbert, A.N., 143 Gilbert, C., 139 Ginsberg, P.E., 288 Galaburda, A.M., 144 Gladwin, T., 234, 241 Gleitman, L., 59, 269 Goel, V., 198, 206, 207 Goins, J.T., 334 Goleman, D., 391 Golestani, N., 130 Göncü, 356 Goodale, A.M., 381 Goodenough, D.R., 278 Goody, J., 289 Goswami, H.K., 137 Gourevich, A., 32 Goyen, J., 323, 334, 335, 336 Grabowski, J., 240 Graf, P., 43, 45 Graham, G., 3 Grant, C.W., 140 Green, C.D., 182, 186
450
Advances in Cognitive Science
Green, D., 122, 124 Greenfield, P.M., 229, 230, 234, 241 Greenfield, S., 397 Greenwood, P.M., 18, 22 Gregory, M., 20 Griffin, D.R., 87, 88 Grimshaw, G.M., 278 Grissemann, H., 327 Grosjean, F., 128 Grosser, G.S., 323, 324, 334, 335, 336 Guenther, H., 411 Gumperz, J.J., 241, 277 Gunderson, V.M., 138 Gupta, A., 66 Gur, R.C., 140, 141 Guralnick, M.J., 357 Hall, N., 357 Hacking, I., 178 Hager, J.C., 141 Hall, D.G., 120, 124 Hall, R., 121, 326 Halpern, D.F., 143, 144 Hamann, S.B., 47 Hambleton, R.K., 223 Hännikäinen, M., 356 Harkins, D.A., 142 Harkness, S., 308 Harms, T., 360 Harré, R., 178, 179 Harris, C.R., 199 Harris, I.M., 330 Harris, L.J., 138 Harris, M., 277 Hart, S.G., 19 Hasegawa, M., 128 Hatta, T., 143 Hauser-Cram, P., 357 Hausmann, M., 130 Hausser, M., 169 Haykin, S., 181 Hayman, C.A.G., 47 Hebb, D.O., 180 Hecaen, H., 142 Heijden, 32 Heil, M., 332 Held, R., 31
Helms-Lorenz, M., 237 Hepper, P.G., 142 Hess, 117 Hicks, C., 324, 335, 336 Hikosaka, O., 9 Hildreth, G., 321 Hilgard, E.R., 391 Hilgard, J.R., 391 Hirisave, 344 Hodgkin, A., 155, 160 Hoffman, J.E., 21 Hoffmann, A., 181, 182, 183 Hoffmann, J., 8 Holland, J.G., 20 Holland, P.W., 224, 232 Holt, G.R., 169 Holtzheimer, P., 126 Hoosain, R., 223 Horowitz, F.D., 138 Houck, G., 347 Hubbard, T.L., 32 Hubel, C., 240, 277 Hull, R., 120, 124, 125, 127 Humphrey, D., 255 Humphrey, N.K., 88 Huntington, E., 302 Hutchins, E., 241 Huxley, A., 155, 160 Ida, Y., 143 Ijalba, E., 121 Illes, J., 127, 129 Inhelder, B., 240 Irvine, S.H., 231 Ishihara, 227 Iteya, M., 136 Iwawaki, S., 289 Izhikevich, E.M., 159 Jackendoff, R., 51, 55, 58, 60, 72 Jahoda, G., 300 James, W., 381, 385, 386, 392, 393, 398, 406, 407, 412, 419 Janet, P., 392 Jaskowski, P., 332, 333 Jayasuriya, W.F., 418 Jensen, A.R., 227
Name Index Jerison, H.J., 219 Jian, S., 115 Johnson, M.H., 315 Johnson, M.K., 45 Johnson-Laird, P., 241 Jolly, A., 88 Jordan, K., 330 Joshi, A.K., 158, 188, 189 Jyotsna, V., 66 Kaczmarek, L., 111, 114, 115, 116 Kalra, Prem K., 159 Kamal Uddin, M., 31 Kamide, Y., 56 Kaminska, B., 114 Kaplan, I.V., 115, 116, 137 Kapoor, S.D., 371 Kapur, M., 318, 344, 369, 371 Kar, B., 316, 341 Karadi, K., 330, 332 Karnath, H., 123 Karanth, P., 341 Karch, G.R., 140 Karemaker, A., 355 Katz, B., 203, 204, 210 Kawabe, T., 31 Keles, P., 138 Keller, E.F., 177, 178 Kerzel, D., 32, 40 Ketelaars, M.P., 355 Keyes, L., 357 Kihlstrom, J.F., 391 Kim, A., 193 Kim, K., 128 Kistler, 165 Klineberg, O., 222 Knoeferle, P., 50, 55, 56, 58, 59, 60 Koch, C., 155, 159, 397 Koch, I., 8 Koff, E., 137, 140, 141 Kojima, H., 298 Kolb, B., 145 Kontos, S., 357 Kosnik, W., 21 Kosslyn, S.M., 330 Krebs, P. R., 156, 177 Kroeber, A., 301, 302
451
Krott, A., 78, 80, 82 Kubota, M., 126 Kunda, Z., 6 Kutas, M., 210 Kuypers, H.J.M., 141 Lachmann, T., 316, 319, 327 Lashley, K., 7 Lave, J., 234 Lecar, H., 159, 170 Leung, K., 224, 230, 232 Leventhal, H., 139 Levinson, S., 217, 241, 248, 266, 269 Levinson, S.C., 241, 242, 248, 270, 274, 275, 277 Levy, J., 145 Lewin, K., 306 Ley, J., 139 Li, P., 269 Libben, G., 71, 72, 75, 77, 85 Liberman, I.Y., 320, 321, 322, 334, 335 Lincoln, Y.S., 229 Linn, M.C., 277 Lipski, S.C., 126 Lloyd, B., 356 Lockard, J.S., 138 Loehr, D., 197, 203 Lomax, A., 288, 298 Lonner, W., 217 Louwman, W.J., 229 Lubrano, V., 122 Lucas, M.D., 138 Lucas, T., 122 Luria, A.R., 223, 289, 344 Lutz, C., 230 Lutz, M.N., 366 Lyle, J.G., 323, 334, 335, 336 Lynn, D.R. ,137, 140 Lynn, J.G., 137, 140 MacDonald, M.C., 51, 193 Mackeben, M., 40 Mackworth, N.H., 18 Mackworth, J.F., 20 Mackworth, N.H., 20 Maehara, K., 142 Mahendra, N., 128, 129
452
Advances in Cognitive Science
Mahesh Yogi, Maharishi, 391 Majid, A., 250 Malofeeva, E., 356 Malpass, R.S., 225, 226, 228 Mandal, M.K., 109, 135, 142, 280, 284 Mandelis, L., 74 Mandler, J., 255 Marin, G., 303 Markstahler, U., 115, 116 Marois, R., 16 Martinot, C., 248 Masaro, D.W., 342 Maslow, A., 395 Mateeff, S., 32 Mathiak, K.,126 Matthews, G., 20 Mattingley, J.B., 139 Maxfield, L., 278 May, P., 139 Maylor, E.A., 21, 22 Mazlack, L.J., 197, 203 McClelland, J.L., 45, 155, 156 McClelland, M.M., 366 McCulloch, P., 180 McDermott, K.B., 44 McDonough, C.J., 197, 203 McGhee, P.E., 198, 205, 210 McIntyre, L.A., 289 McKay, B., 302 McLaughlin, J.P., 139 McLeod, P., 181 McManus, I.C., 143 Mead, A.M., 139 Mead, A.P., 88 Mechelli, A., 131 Mehdi, B., 371 Melngailis, J., 324 Metzler, J., 330 Michel, G.F., 142 Miehlke, A., 140 Miles, E., 319 Miles, T.R., 319 Miller, G., 241 Miller, G.A., 3, 7 Milner, M.A., 380 Minsky, M., 181, 202, 209 Mishra, D., 155, 159
Mishra, R., 240 Mishra, R.C., 217, 218, 240, 250, 264, 277, 287, 297, 300 Mishra, R.K., 347 Mitchell, R.F., 325, 326, 334 Miyapuram, K.P., 7 Miyashita, Y., 116 Mobbs, D., 207 Mondor, T.A., 139 Montero, V.M., 115, 117 Montgomery, J.W., 348 Moore, G.E., 396 Moran, E., 302 Moran, J.M., 210 Moreno, C., 138, 141 Moreno, E., 130 Morgan, M., 145 Morgan, M.S., 179 Morris, C., 159, 170 Moscovitch, M., 44, 137, 138, 284 Mower, G.D., 116 Mulder, M.P., 199 Mullen, B., 141 Mulloni, B., 171 Munakata, Y., 315 Munte, T., 121 Murdock, G.P., 288, 301 Murphy, G.L., 55 Musseler, J., 32 Myers, F.W.H., 398 Myers, J., 82 Myers, R.F., 140 Naatanen, R., 337 Nagasena, 407, 408 Naglieri, J.A., 344, 352 Nagumo, J.S., 159 Nagylaki, T., 145 Nakamizo, S., 31 Nakamichi, M., 138 Nakatsuka, 143 Nakayama, K., 40 Natsoulas, T., 383 Naveh-Benjamin, M., 234 Nedivi, E., 116 Nelson, C.A., 138 Neville, H., 121
Name Index Nijholt, A.A., 199, 203 Niraula, S., 278, 283 Nisbett, R., 233 Nicholls, M.E., 139 Niraula, S., 243, 244 Nolen, S.B., 357 Norvig, P., 181 Nothdurft, H.C., 40 Nowe, N., 356 Nsamenang, A.B., 226 Nuechterlein, K., 19 Ober, J., 331 Obler, L.K., 120 Ogden, L., 366 Ojemann, G., 120 Olds, J., 137 Olivers, C.N.L., 32 Olton, D.S., 47 Oommen, 344 O’Regan, J.K., 381 Orton, S.T., 319, 322, 323, 325, 326, 327, 334 Owen, A.M., 16 Paabo, S., 303 Padakannaya, P., 62 Palmer, J., 388 Pammi, V.S.C., 7, 8, 16 Papert, S., 181 Paradis, M., 120, 121, 122 Parasuraman, R., 18, 19, 20, 21, 22, 27, 28 Pariyadath, V., 156, 157, 197 Parten, 356 Patanjali, 391, 403, 417 Pataraia, E., 123 Patton, J.E., 325, 335 Paulos, J., 201, 209 Peng, K., 233 Penrose, R., 179, 187 Pentland, A., 60 Perani, D., 128 Perelle, I.B., 143 Perez-Alvarez, F., 349, 351 Perkel, D.H., 171 Peters, M., 136 Peterson, A.C., 277 Piaget, J., 240, 356
453
Pianta, R.C., 357 Pickett, R.M., 19 Pierce, L., 22 Pinxten, R., 241 Pittman, T.S., 387 Pitts, W., 180 Plonsy, R., 171 Poortinga, Y.H., 217, 221 Porac, C., 136, 137, 142, 143, 144 Porteus, S.D., 222 Posner, M.I., 4, 20, 21, 22, 27, 28, 37, 40, 107, 325, 326, 327, 334 Post, R.H., 227 Povinelli, D.J., 91, 92 Prabhakara, 401 Prajna Devi, 258 Prakash, P., 66 Premack, D., 90 Prinzemetal, W., 21 Provine, R.R., 198 Radhakrishnan, S., 409, 423 Rajaram, S., 43, 45 Ramachandran, V., 369 Ramachandran, V.S., 202, 203, 210 Ramey, G.W., 350 Rao, K.R., 382 Rappoport, R., 302 Raskin, V., 201, 203 Raven, J.C., 343, 371 Ray, S., 159 Remington, R., 22 Rensink, R.A., 381 Richerson, P., 302 Rinzel, J., 170 Ritchie, G., 197, 199, 203 Rivers, W.H.R., 217 Robert, L.W., 288 Roberts, G.R., 139 Robinson, B., 115 Roediger, H.L., 44 Rogers, T.T., 185 Rogoff, B., 222, 234, 355 Rolke, B., 332 Rorden, C., 123 Rosen, K.M., 116 Rosenbaum, D.A., 7
454
Advances in Cognitive Science
Rosenblatt, F., 180, 185 Rothbart, M.K., 137 Roux, F.E., 122 Roy, D., 60 Ruch, W., 198, 199 Rumelhart, D.E., 155, 156 Ruppel, S.E., 32 Rushton, P., 227 Rushton, W.A.H., 160 Rusiak, P., 330, 332, 333 Russel, S., 181 Rüsseler, J., 324 Ruthruff, E., 330 Rutkowski, J.S., 350 Sackeim, H.A., 140, 141 Safer, M.A., 139 Sakai, K., 8 Sakhuja, T., 139 Salili, F., 223 Saling, M.M., 138 Salk, L., 138 Samkara, 400, 401, 402, 420 Sarachchandra, E.R., 406, 408, 409 Sarangapani, P.M., 369 Saxby, L., 139 Saxe, G.B., 234 Scanlon, D.M., 335 Schabes, Y., 189 Schacter, D.L., 43, 45 Schlaug, G., 121 Schliemann, A., 234 Schneider, W., 159 Schröger, E., 337 Schroeder, S.R., 20 Schultz, T.R., 180 Schultz, W., 207 Scribner, S., 223, 233, 289 Sebastian, S., 341 See, J.E., 19 Segall, M.H., 232, 233, 234, 241, 246, 274 Segev, I., 171 Seidenberg, M., 334 Sergent, J., 278, 284 Seyfarth, R.M., 87, 88, 89, 90 Shadish, W.R., 228 Shammi, P., 198, 199, 205, 206, 210
Shapiro, M.L., 47 Share, D.L., 341 Shaw, M.L., 21 Shaywitz, S.E., 341 Shebani, M., 233, 234 Shebani, M.F.A., 233–34, 238 Shepard, R.N., 330, 331 Sheth, B.R., 32, 39, 40 Shimojo, S., 32, 39, 40 Shuey, A.M., 222 Siegel, L.S., 349 Silverman, D., 229 Simos, P., 126 Singer, J.L., 390 Singh, I.L., 18, 22 Singh, S.K., 137 Sinha, A., 87, 90, 92–95, 97, 98, 101 Sinha, D., 280, 292 Sinha, J., 399, 401, 402 Skinner, M., 141 Slaghuis, W.L., 352 Smart, J.L., 142 Smuts, B., 88 Smythe, L., 31 Snow, C.E., 277 Spence, S.H., 356 Spiegler, B.J., 142 Spivey-Knowlton, M.J., 52 Springer, P., 138 Sproat, R., 62, 65, 67, 68 Squire, L.R., 47 Srinivasan, N., 156, 157, 197 Stace, W.C., 394 Stagnitti, K., 357 Stanley, G., 326 Stanovich, K.E., 341 Staveland, L.E., 19 Sternberg, S., 326 Sterelny, K., 179 Stigler, J.W., 223 Stock, O., 197, 203 Strapparava, C., 197, 203 Strauss, E., 138 Strogatz, S.H., 159 Stuhlman, M.W., 357 Stuss, D., 198, 199, 205, 206, 210 Suls, J., 200
Name Index Super, C., 308 Suresh, P.A., 341 Suzuki, T., 410 Swallow, J.A., 277, 284 Swets, J.A., 20 Taft, M., 74 Takeda, S., 138 Takizawa, T., 197 Tanenhaus, M.K., 50, 51, 53, 54, 58, 59 Tanon, F., 234 Tanzer, N.K., 223 Tart, C.T., 389 Taylor, H.A., 240, 278 Taylor, J., 197, 203 Teichner, W.H., 19 Teng, E.L., 143 Terepocki, M., 325 Terrace, H., 8 Tham, W-P., 128 Tharp, D.A., 74 Thompson, A.M., 142 Timoneda-Gallart, C., 349, 351 Todd, J.J., 16 Tomasello, M., 87, 303 Torgesen, J., 347 Tremoulet, M., 122 Triandis, H.C., 217 Tripathi, N., 316, 341 Troadec, B., 248, 301 Trzeciak, G.M., 323, 324, 334, 335, 336 Tse, P.U., 40 Tucker, L.R., 224 Tulving, E., 45 Turnbull, O.H., 138 Tversky, B., 240, 278 Udayana, 402 Uddin, M.K., 41 Ullman, M., 401 Unsworth, C., 357 Vaid, Jyotsna, 108, 120, 199, 211 Vajpayee, A., 250, 253 Vajpayee, A.P., 279 Van Bon, W.H.J., 360 Van de Vijver, F.J.R., 223, 224, 226, 229, 230, 231, 232
455
van der Aalsvoort, G.M., 355 van der Heijden, A.H.C., 32, 40 van Leeuwen, C., 320, 328, 333, 339–40 Van Oers, B., 356 Van Strien, J.W., 136, 143 Varela, F.J., 381 Vayda, A.P., 302 Veatch, T.C., 199, 205, 207 Vellutino, F.R., 320, 321, 335 Verba, M., 356, 362 Vernon, P.E., 289 Verwey, W.B., 8 Vihla, M., 126 Vijay-Shanker, K., 194 Visuvalingam, S., 199 Vogel, G., 390 Vygotsky, L.S., 217, 222, 232, 355 Wainer, H., 224, 232 Wallace, C.S., 116 Wapner, W., 206 Wardekker, W., 356 Warm, J.S., 19 Warrington, E.K., 44 Wartenburger, I., 129 Wassmann, J., 218, 242, 243, 245, 251, 264 Webber, B., 195 Weiskrantz, L., 44 Werner, O., 226 Werner, C., 303 Werner, S., 32, 41, 240, 277 West, R.M.E., 171 Whishaw, I.Q., 145 Whitaker, H.A., 120 White, E.B., 197 Whiten, A., 91, 96, 99, 100 Whiting, J.W.M., 235, 300 Wild, B., 206 Willemsen, M.E., 226 Williams, M.R., 178, 187 Wilson, H.R., 162 Witelson, S.F., 277, 284 Witkin, H.A., 278 Wolff, P.H., 324 Wolff, W., 140 Wolford, G., 284 Woodruff, G., 90
456
Advances in Cognitive Science
Wysocki, C.J., 143 Wyver, S.R., 356 Yadav, A., 159 Yamada, Y., 115 Yeni-Komshair, G.H., 142 Yeo, R.A., 144, 145 Yin, H., 82
Yokogawa, T., 197, 203, 204 Zak, S., 159 Zangenehpour, S., 116 Zatorre, R., 130 Zelinsky, G.J., 54 Zhang, F., 115 Ziman, J., 177
