Clinical Linguistics: Theory and Applications in Speech Pathology and Therapy (Current Issues in Linguistic Theory)

CLINICAL LINGUISTICS

AMSTERDAM STUDIES IN THE THEORY AND HISTORY OF LINGUISTIC SCIENCE General Editor E. F. KONRAD KOERNER (University of Ottawa) Series IV – CURRENT ISSUES IN LINGUISTIC THEORY

Advisory Editorial Board Raimo Anttila (Los Angeles); Lyle Campbell (Christchurch, N.Z.) Sheila Embleton (Toronto); John E. Joseph (Edinburgh) Manfred Krifka (Berlin); Hans-Heinrich Lieb (Berlin) E. Wyn Roberts (Vancouver, B.C.); Hans-Jürgen Sasse (Köln)

Volume 227

Elisabetta Fava (ed.) Clinical Linguistics Theory and applications in speech pathology and therapy

CLINICAL LINGUISTICS THEORY AND APPLICATIONS IN SPEECH PATHOLOGY AND THERAPY

Edited by

ELISABETTA FAVA Università di Ferrara

JOHN BENJAMINS PUBLISHING COMPANY AMSTERDAM/PHILADELPHIA

To

Alberto Mioni for his 35 years of dedication in the teaching of General Linguistics

8

TM

The paper used in this publication meets the minimum requirements of American National Standard for Information Sciences — Permanence of Paper for Printed Library Materials, ANSI Z39.48-1984.

Library of Congress Cataloging-in-Publication Data Clinical Linguistics: theory and applications in speech pathology and therapy / edited by Elisabetta Fava. p. cm. -- (Amsterdam studies in the theory and history of linguistic science. Series IV, Current issues in Linguistic theory, ISSN 0304-0763; v. 227) Includes bibliographical references and index. 1. Speech disorders. 2. Speech theraphy. 3. Applied linguistics. I. Fava, Elissabetta. II. Series. RC432 .C556 2002 616.85'506--dc21 2002025406 ISBN 90 272 4735 8 (Eur.) / 1 58811 223 3 (US) (Hb; alk. paper) © 2002 – John Benjamins B.V. No part of this book may be reproduced in any form, by print, photoprint, microfilm, or any other means, without written permission from the publisher. John Benjamins Publishing Co. • P.O.Box 63224 • 1020 ME Amsterdam • The Netherlands John Benjamins North America • P.O.Box 27519 • Philadelphia PA 19118-0519 • USA

ACKNOWLEDGMENTS I am indebted to many people for making this volume possible. First of all, to the authors of the articles here included, for having presented and discussed their theses at the conference "Linguistics Theory in Speech and Language Pathology and in Speech Therapy", held in Padova in August 2000. Their help and co-operation with the editorial work has been invaluable. My intellectual debt goes to my colleagues of the Department of Linguistics of the University of Padova, Alberto Mioni and Alberto Zamboni. They helped me constantly with my project to involve the Department of Linguistics in the offering of Clinical Linguistics courses at the Faculty of Medicine. Without their unstinting support, this project would never have materialised. This Padova conference was organised with the support of a grant from Cofinanziamento Murst (1998-9810197024), whose national co-ordinator was Professor Alberto Mioni. It was important for this project to succeed that I was at the time a member of the Psychology Faculty of the University of Padova. This volume is giving me the opportunity to thank my colleagues of the Faculty of Psychology of the University of Padova, Remo Job, Alberto Mazzocco and Sergio Roncato, for having first supported the role of linguistics within the psychological sciences, for having fostered the academic co-operation between researchers from various disciplines, and for having included, already ten years ago, the teaching of General Linguistics within the Psychology curricula. Likewise, I would also like to express my gratitude to my colleagues in the Faculty of Medicine of Padova, Corrado Angelini, Anna Maria Laverda, and Carla Monciotti. To all of them, I extend my gratitude for believing that lin­ guistics can play an important role in cognitive and neurolinguistic research, and for their active interest in humanistic disciplines. The delegate of the Rector of the University of Padova, Filippo Donà dalle Rose, responsible for the Socrates exchange program of the Padova Athe­ naeum, has made incredibly tenacious and generous efforts to involve students and professors in the organisation of the Erasmus-Socrates Projects. The di­ dactic experience with Socrates students and colleagues was an important moment of this project. I owe fruitful discussions on the topics of this volume to Paola Crisma of the University of Trieste, Denis Delfitto of the University of Utrecht, Giuseppe Longobardi of the University of Trieste, Anna Thornton of

VI

ACKNOWLEDGMENTS

the University of L'Aquila, and Nicola Grandi of the University of Milano Bicocca. I also would like to thank my new colleagues of the Humanities Faculty of Ferrara, Claudio Iacobini, Marina Nespor, and Carlotta Sanfilippo. I hope to be able to collaborate with them on future projects. Tullio De Mauro of the University of Rome was generous with information on and discussion of the structuralist tradition. I would also like to express my gratitude to my colleagues of the Master's degree program in Logopedics in Padova, especially Paola Crisma, Mario D'Angelo, Sara Gesuato, Carlo Schirru, Flavia Ursini, and Alessandro Zijno. Their contribution in defending the project of a Clinical Linguistics degree at the Faculty of Medicine in Padova has been invaluable. As well, I owe an expression of thanks to Sara Gesuato for her editing of the first draft of the volume; I also remain grateful to her for the emotional support she offered me during this period. To Nadia Radovich and to Alberto Mello, of the printing desk of Palazzo Maldura of the University of Padova, goes my expression of thanks for helping in the formatting of the volume, for their technical support, for sharing their experience with me, and for their kind co-operation generally. Last but not least, I am grateful to E.F.K. Koerner, the editor of "Current Issues in Linguistic Theory", who first proposed the publication of this vol­ ume, for keeping an active interest in the process and progress in the pre­ paration of the final copy. His work as editor and his role as moderator in the linguistic sciences cannot be overestimated. He provided invaluable criticism, comments and directions in editing this volume. It has been a great privilege for me to benefit from his unequalled experience. I also would like to express my gratitude to Anke de Looper of John Benjamins Publishing Company for her patience and kindness. Ferrara, May 2002

Elisabetta Fava

CONTENTS Acknowledgments Editor's Introduction I . PHONOLOGY IN CLINICAL APPLICATIONS Yishai TOBIN Phonology as human behavior: Theoretical implications and cognitive and clinical applications

v ix

3

Dirk-Bart D E N OUDEN

Segmental vs syllable markedness: Deletion errors in the paraphasias of fluent and non-fluent aphasics

23

II WORDS IN DEAFNESS AND STUTTERING Roberto AJELLO, Giovanna MAROTTA, Laura MAZZONI & Florida NICOLAI

Morphosyntactic fragility in the spoken and written Italian of the deaf 49 Peter HOWELL & James Au-YEUNG The EXPLAN theory of fluency control applied to the diagnosis of stuttering 75 Peter HOWELL The EXPLAN theory of fluency control applied to the treatment of stuttering 95

III. MORPHOLOGY AND SYNTAX IN CHILD LANGUAGE DISORDERS Roelien BASTIAANSE, Gerard BOL, Sofie VAN MOL & Shalom ZUCKERMAN Verb movement and finiteness in language impairment and language development Stavroula STAVRAKAKI A-bar movement constructions in Greek children with SLI: Evidence for deficits in the syntactic component of language Susan M. SUZMAN Morphological accessibility in Zulu

119

131 155

vin

TABLE OF CONTENTS

Yumiko TANAKA WELTY, Jun WATANABE & Lise MENN Language production in Japanese preschoolers with SLI: Testing theories

175

IV. ISSUES ON GRAMMAR AND COGNITION Leah PALTIEL-GEDALYOVICH Testing linguistic concepts: Are we testing semantics, syntax or pragmatics ? Dusana RYBÁROVÁ SLI and modularity: Linguistic and non-linguistic explanations Vesna STOJANOVIK, Mick PERKINS & Sara HOWARD The language/cognition interface: Lessons from SLI and Williams Syndrome

197 213

229

V. GRAMMATICAL STRUCTURE IN APHASIA Susan EDWARDS Grammar and fluent aphasia Anna GAVARRO Failure to agree in agrammatism Judith RISPENS, Roelien BASTIAANSE & Susan EDWARDS The verb and sentence test: Assessing verb and sentence compre­ hension and production in aphasia Esther RUIGENDIJK Case assignment as an explanation f or determiner omission in German agrammatic speech Kyrana TSAPKINI, Gonia JAREMA & Eva KEHAYIA The role of verbal morphology in aphasia during lexical access: Evidence from Greek

249

Index of Subjects

337

List of Contributors

267

279

299

315

EDITOR'S INTRODUCTION The present volume deals with different aspects of speech and language pathology and identifies and re-examines, from various perspectives, a number of standard assumptions in clinical linguistics and cognitive science. It encom­ passes issues concerning deafness, stuttering, child language acquisition, Spe­ cific Language Impairment (SLI), William's Syndromes deficit, fluent aphasia, and agrammatism. Different levels of linguistic analysis are considered: phonetics, phonology, syntax, semantics, and pragmatics. Some of their prop­ erties, their formal representations and their interfaces with other levels, are examined closely and clarified. Researchers typically summarise their individual results in individual con­ tributions, not only in the format of an article (subjects, materials, languages, scores, etc.), but also through critical discussion of the relevant literature. Their work is located in different lines of research: structuralist tenets, genera­ tive approaches, and frameworks such as Optimality Theory. The topics dis­ cussed are intricate and complex and, at the same time, wide-ranging in scope, but the chapters in the present book offer a fairly comprehensive overview of the complexity and the emerging importance of the field of Clinical Linguistics. The analyses presented here reflect various aspects of the at times fierce theo­ retical and empirical debates currently ranging over almost every issue dis­ cussed, with respect to linguistics and the other cognitive disciplines. Although it is very difficult to do justice to all of the subtleties of argumentation that each of these lines of investigation require, the rather technical issues discussed in this volume have a bearing on questions of considerable interest. They presup­ pose or imply assumptions about the internal architecture of the language fac­ ulty, whose location among other systems of the mind/brain is not at all obvi­ ous. The interdisciplinary complexity of the language/cognition interface is also explored by focusing on empirical data of different languages: among them, Germanic languages (such as Dutch, English, German), Greek, Hebrew, Japa­ nese, Romance languages (such as Catalan, Italian, and Spanish) and Sohsho, a Zulu language. Despite the different approaches and the variety of problems posed in this volume, there is a common aim to the chapters. The authors piece together various fragments of clinical linguistic research, trying to bring them into a

X

ELISABETTAFAVA

more cohesive whole, and offer a sense of some of the technical problems that lie at the forefront of research and suggest the kind of answers that their work may provide. The aim of this volume is to stress the growing importance of the theoreti­ cal and methodological tools developed in Clinical Linguistics; to put under scrutiny assumptions taken for granted in previous analyses, which may not be as obvious as they seem; to investigate how even apparently minimal choices in the description of phenomena may affect the form and complexity of the lan­ guage/cognition interface. What is called 'clinical work' is not separable from the scientific work done by linguists; instead, it should be considered as a com­ ponent part of General Linguistics. It is at the same time concerned with the actual foundation of the study of linguistic pathologies, together with neurobi­ ology, psychology and neurology and the outer reaches of scientific specula­ tion about the nature of the mind/brain complex with regard to language. In this perspective, it is very important to defend the continuity of linguistic speculations beyond different frameworks. Such a continuity, which has often been emphasised and is among the goals of the CELT series, of which this vol­ ume is a part, is even more important in clinical applications. The study of the relationship between language and the brain, or rather between language disor­ ders and brain lesions goes back to the beginning of the 19th century. Begin­ ning with the pioneering, albeit contentious, work of Franz Joseph Gall (17581828), the observations, arguments and opinions advanced in the research of Paul Broca (1824-1880), Carl Wernicke (1848-1905) or Pierre Marie (18531940) are still today crucially important, considering the frequency with which they are cited in current research papers.1 Since then, linguistics has included data and results from speech and language pathology, broadening its empirical basis and reframing data and research in speech and language pathology. The structuralist research, by reconsidering the traditions of this branch of science, has paved the way for the inclusion of language disorders research within linguistic theory. Considerations of the 19th-century discoveries of the French surgeon and anthropologist Paul Broca can be found in Ferdinand de Saussure's (1857-1913) work. We may also refer to some of the key concepts identified and discussed by Saussure himself, such as the methodological and ontological notions of langue and parole, which support and justify a large part of clinical linguistic research. They may be considered part of the implicit or explicit distinctions systematically adopted in any clinical work. Thirty years later, Roman Jakobson's systematic search for what was later described as the 1 See Paul Eling (ed.), Reader in the History of Aphasia: From [Franz] Gall to [Norman] Geschwind (Amsterdam & Philadelphia: John Benjamins, 1994).

INTRODUCTION

XI

elementary quanta of language offered tentative generalisations of acquisition and loss of grammatical systems. His hypotheses on child language acquisition and language breakdown, together with his inquiry into language change, are still part of the research into the structuration, restructuration and loss of lan­ guage. They deal with general tendencies, considered 'lawful' by Jakobson, whose status is still debated in the most recent literature.2 Empirical evidence in favour or against some of his hypotheses is offered and explicitly reconsidered in this volume. With Noam Chomsky's work, the impact of linguistic theory on the study of language in wider context as part of the scientific investigation of the archi­ tecture of the mind became significant. Many analyses proposed in this volume refer to rules and principles elaborated by Chomsky, although in slightly dif­ ferent frameworks, including various 'minimalist 'versions. One of the main phenomena that has motivated Chomsky since Syntactic Structures is the un­ limited possibility of expression in human language, what he now calls the dis­ crete infinity of language.3 In order for speakers of a language to create and understand sentences that they have never heard before, there must be a way of combining some finite number of memorised units into phrases and sentences of arbitrary length. The speaker's knowledge of the language must include a set of principles of combinations that determine which combinations are well formed (and which are ill-formed) and what they mean. The construction of minimal pair paradigms is one of the major features of experimental design concerning different abstract levels of representations. In their search for rigorous stipulations, linguists, like physicists, biologists or chemists, manipulate the environment experimentally with a central concern on issues of acceptability, as a pretheoretical term, and of grammaticality. The manner in which the linguist discusses pairs of grammatical sentences, juxta­ posed with their non-grammatical counterparts, has offered a way to under­ stand sentences, not to reject them. This approach has had a dramatic impact on clinical linguistic applications and in this volume too, this major point, inex­ tricably connected with a cluster of other issues, is reconsidered. Some fre­ quently observed features in the clinical literature, which are registered as 'er­ rors' concerning inflectional properties, determiners, or Wh-features, are not considered merely impressionistically: in this volume, the nature of the various kinds of assertions involved in identifying them as errors is discussed con2

Cf. Roman Jakobson, Studies on Child Language and Aphasia (The Hague: Mouton, 1971 [1941]), p. 51. 3 Cf. Noam Chomsky, New Horizons in the Study of Language and Mind (Cambridge: Cam­ bridge University Press, 2000), pp. 3-4, 184.

Xll

ELISABETTAFAVA

sistently. The way Chomsky sets up problems is an inducement to reflection and research not only when there is a convergence of methodological tools and theoretical apparatus, but also with differing and contrasting frameworks. Continuity in linguistic research has been even stronger in the last few decades, as issues on cognitivism and functionalism have been reconsidered, sometimes radically. It is claimed that cognitive abilities and experientially de­ rived cognitive models have direct and pervasive linguistic manifestations, and, conversely, that language structure furnishes important clues concerning basic language phenomena.4 By stressing that the conceptual setting for the work the linguist is doing is a scientific one, the research focuses on the role of larger contextual frames in describing language disorders: properties of different lan­ guages, principles of conversation, and various language functions. An under­ lying common feature of this approach, which is considered in this volume, concerns the role of semantic and pragmatic factors in assessing clinical data and evaluating therapeutic interventions. Many contributions to this volume show how delicate and difficult the treatment of these data is, involving deci­ sions which do not always find a consensus among researchers, as the debate on the empirical adequacy of many analyses suggests. Elsewhere I have de­ fended the relevance of semantics and pragmatics in controlling linguistic stipu­ lations in non-pathological frameworks: in dealing with descriptive and, more importantly, explanatory adequacy for some well-known phenomena such as interrogative or relative clause distinctions, crucial evidence is offered by re­ considering semantic and pragmatic aspects systematically.5 Attention to se­ mantics and pragmatics becomes even more crucial when dealing at the same time with both sentences and utterances, and when trying to apply distinctions at the boundaries between the grammatical and the pragmatic properties of what is said in contexts such as the linguistic production of a child with SLI or an aphasic patient. In this volume issues of appropriateness are discussed and considered as instruments to clarify the nature of such deficits. Moreover, it is an important merit of this volume that it discusses the role of semantic and pragmatic factors not only in the evaluation of deficits, but also in the perspec4 See Ronald W. Langacker, "Reference-point Constructions", Cognitive Linguistics 4.1-38 (1993). 5 Cf. Elisabetta Fava, "Questioning Interrogative Interpretation in Some Indo-European Lan­ guages", Contrastive Semantics and Pragmatics, vol. I: Meanings and representations ed. by Katarzyna Jaszczolt & Ken Turner (Oxford: Elsevier, 1996), 87-110; "Langue and Parole in Speech Act Theories: Some considerations and a proposal", Langue and Parole in Synchronic and Diachronic Perspective: Selected proceedings of the XXXIst Annual Meeting of the Societas Linguistica Europea (26-30 August 1998, St Andrews) ed. by Christopher Beedham (Oxford: Pergamon, 1999), 263-283.

INTRODUCTION

Xlll

tive of the therapy intervention with an attempt to take a larger perspective, opening up a discussion about the consequences and fruitfulness of different approaches. Another general remark should be made here. In volume contains some currently used terms which may cover empirically and conceptually distinct notions. A notion for which there may be no congruence of criteria is that of 'word': the same term may cover empirically and conceptually distinct concepts such as phonological word, content word, syntactic atom, morphological ob­ ject, lexical item, etc. Here phonological, syntactic and morphological criteria may be in conflict. Other crucial notions are 'subject' and 'object': although such terms appear frequently in linguistic argumentation (including in this vol­ ume), the notions underlying them may be radically different. They may be re­ garded as relating to semantic roles or considered pre-theoretically. The Rela­ tional Grammar model, instead, treats grammatical relations like Subject and Direct Object as primitive and central to grammar, while in the Government and Binding framework, X-bar type constituent structure representations are con­ sidered as basic.6 As a result, the interactions of the properties of Subject and Object, considered in rather different frameworks, may be analysed in rather different ways in this volume. After this lengthy introductory statement, let me now offer comments on the individual contributions to this volume. As the table of contents indicates, the volume is organised into five sections. The first section considers clinical applications in phonology. The data analysed are taken from different populations: child acquisition, deafness, flu­ ent and non-fluent aphasia, and other pathologies. The first essay, by Yishai Tobin (Ben Gurion University), is on Phonology as human behavior: Theoret­ ical implications and cognitive and clinical applications. By defending the con­ tribution of Structural Phonology to different clinical data applied to a large va­ riety of languages, he discusses the conflict between the communication and the human factors in language users' search for maximum communication with minimal effort both in the diachronic development and the synchronic state of a language. This conflict is even more keenly felt both in language acquisition, 6

Cf. Elisabetta Fava, "Contextualising Corpora in Testing Grammatical Hypotheses: Searching for preposed and postposed syntactic subjects in North Italian and Veneto childadult conversations", Lingua, Discourso, Texto: I Simposio internacional de analisis del discourso, vol. I, ed. by José Jesús de Bustos Tovar, Patrick Charaudeau, José Luis Giron Alconchel, Silvia Iglesias Recuero & Alonso Covadonga Lopez (Madrid: Visor Libros, 2000), 263-280; cf. David Perlmutter & Carol Rosen (eds.), Studies in Relational Grammar, vol. II (Chicago: University of Chicago Press, 1984).

XiV

ELISABETTAFAVA

where functional errors and processes may be observed, and in the clinical set­ ting, where developmental and pathological errors and processes become ap­ parent. The theory of phonology as human behaviour can explain, in a princi­ pled way, the connection and interrelationship between the phylogeny, the on­ togeny, and the pathology of the development of sound systems in human lan­ guages. Phonotactic skewing in language (diachronic, synchronic, and in de­ velopmental and pathological data) reflect the learning process of speakers. Tobin includes data of a hearing child of deaf parents and other series of func­ tional processes from different pathologies. The second contribution, by Dirk-Bart Den Ouden (Groningen University), is on Segmental vs. syllable markedness: Deletion errors in the paraphasias of fluent and non-fluent aphasics. It applies Optimality Theory to the analysis of deletion errors in the paraphasias of fluent and non-fluent aphasics, by compar­ ing the literal paraphasias of fluent and non-fluent aphasie speakers on a repeti­ tion task, aimed at determining the influence of syllable structure on error pat­ terns. Markedness is one of the key concepts of this chapter. Where phonolog­ ical theory has been applied to clinical data, the concept of markedness has of­ ten played a significant role. Not often taken into account, however, is the fact that the markedness value of linguistic structures may not be the same at all psycholinguistic levels of processing. What is marked at some linguistic level of representation may well be unmarked at another. The influence of different types of markedness on literal paraphasias may be related to specific psycholinguistic levels of processing and a difference between the pre-phonetic and the phonetic level of processing is hypothesised. The second section discusses problems of words in deafness and stutter­ ing. Although deafness is a pathology, which is widely discussed in the litera­ ture, the phenomena involved are not so often considered with sophisticated linguistic tools, especially with regard to writing systems where some linguis­ tic generalisations seem to be missed. The two papers on stuttering, instead, provide insight into a major area of speech phenomena that have traditionally been treated outside of linguistics proper. Stuttering is typically thought to re­ flect aspects of an individual's speech output that suggest problems with what are vaguely referred to as the performance systems. As well as other speech disorders demanding an explanation outside of the realm of formal grammar is not too much studied in clinical linguistics. Roberto Ajello, Giovanna Marotta, Laura Mazzoni & Florida Nicolai (University of Pisa) deal with The Morphosyntactic fragility in the spoken and written words of the deaf They analyse the linguistic production of profoundly deaf Italian people who received an oralist instruction in specialised institutions

INTRODUCTION

XV

and who never wore a prothesis. Their results reveal a discrepancy between a fairly good lexical competence and a poor morphological competence heavily dependent on the input, and a similarly poor syntactic competence, which relies fundamentally on pragmatic communication principles. The not sufficiently mastered morphology, and free rather than bound morphology, is explained on the basis of the process of learning, based mainly on general, not specifically linguistic, cognitive mechanisms. Peter Howell and his research group at Uni­ versity College London have elaborated a new perspective on stuttering, the EXPLAN theory, based on some linguistic insights, and offer an account of how differences between fluent and stuttered speech arise. Although the main focus is on fluent speech control, it is also relevant to the diagnosis and treat­ ment of stuttering. In the first paper, by Peter Howell & James Au-Young, The EXPLAN the­ ory of fluency control applied to the diagnosis of stuttering, the use of 'phono­ logical word' to predict the development of stuttering from childhood to adult­ hood is discussed. According to EXPLAN, the distinguishing characteristics of stuttering emerge in late childhood; a contrasting view is that people who stutter have incipiently different problems in controlling speech from the onset of language. Diagnosis of the disorder involves measuring and specifying the types of fluency failures that occur in late adolescence in order to see if they have changed in a way commensurate with adult stuttering. Early intervention may not be advisable in such suspected cases, not as long as the child is ex­ hibiting the types of fluency failure common to all children. The second paper, The EXPLAN theory of fluency control applied to the treatment of stuttering, Peter Howell poses the question of how speech can be manipulated, that is, how fluent speech can be made to contain fluency failures and how stuttered speech can be changed to make it more fluent. He proposes a linguistically motivated operant procedure for treating stuttering. A treatment for stuttering is considered successful if it decreases the incidence of fluency failures. EXPLAN theory is more specific as it requires a decrease in the inci­ dence of fluency failures that involve production of parts of words. Howell's results show the reduction in stalling fluency failures may be a result of the op­ erant procedures reducing overall speech rate, thereby cutting down all types of fluency failure. In the course of the treatment sessions, content word (advanc­ ing) fluency failures decreased significantly. The third section concerns morphology and syntax in child language disor­ ders. Four contributions discuss data from Dutch, Greek, Bantu, and Japanese in relation to a major issue, Specific Language Impairment (SLI). This term, applied to a child whose language development is substantially below age level

XVi

ELISABETTAFAVA

for no apparent cause, describes delays and/or disorders in the procedure of the acquisition of grammar in the absence of any hearing loss, mental retardation (performance IQ is within normal range), articulator motor impairment or psy­ cho-emotional disorders. The language development of SLI children is char­ acterised by severe problems in the acquisition of morphology and syntax, while their cognitive, motor and social development is considered to fall within the normal range. In attempts to identify the nature of the linguistic deficit in SLI, diverse linguistic models have been proposed. Although most authorities agree that SLI is heterogeneous, there is little consensus about how it should be classified. Based on cross-linguistic research findings indicating that inflec­ tional and derivational morphology is severely impaired in SLI children, several explanatory models of the linguistic deficit in SLI have been developed. They advance different hypotheses on the locus of the deficit in SLI grammar as well as on the nature of developmental patterns followed in SLI, i.e., whether lan­ guage development in SLI is a normal but delayed procedure or an abnormal process. The chapter by Roelien Bastiaanse, Gerard Boll, Sofia van Mol & Shalom Zuckerman (Groningen University) on Verb Movement and finiteness in lan­ guage impairment and language development poses two central questions. First, what is the origin of the problems with finite verbs in three populations, normally developing children, SLI children, and agrammatic aphasics? Second, whether is it true that SLI children deal with the problems of the production of finite verbs in the same way as normal children do? Although there are clear similarities between normally developing children and SLI children with respect to the production of finite verbs, the SLI children resemble agrammatic aphasics when it comes to strategies in circumventing their problems. According to the authors, the three populations have the same underlying problem, i.e., verb movement, but different ways to solve it. The normally developing children have problems with Verb Second position and circumvent these problems by inserting dummy auxiliaries, either from their dialect, or from a construction that in adult grammar has a different meaning. There is a clear dichotomy be­ tween the errors produced by the normally developing children and the aro­ matic speakers. The results of the Bastiaanse research group do not confirm the hypothesis that language decay is simply the opposite of language learning, as suggested by Jakobson sixty years ago. Agrammatic aphasies have knowledge of their language that young children have not yet acquired. Although there are certain similarities, given that both populations have problems with verb move­ ment, the solutions are different.

INTRODUCTION

XVii

In her contribution A-bar movement constructions in Greek children with SLI: Evidence for deficits in the syntactic component of language, Stavroula Stavrakaki (University of Thessaloniki) presents experimental data concerning the production of structures involving A-bar movement, that is, relative clauses and wh-questions, by Greek SLI children and their normally developing peers. Her results show that the performance of SLI children is qualitatively different from that of the control group. She discusses the status of the underlying lin­ guistic mechanism in SLI and normal grammar, making an attempt to evaluate the implications of her research findings for the core of the theoretical issues surrounding SLI, that is, the locus of the deficit and the way language devel­ opment takes place in SLI children. Assuming that linguistic development in SLI is an explicit rather than an implicit procedure, she argues that SLI children have problems with purely syntactic operations, such as A-bar movement. Within the minimalist framework, this is seen as due to a severe deficit in the in-interpretable features of grammar, that is, features with no semantic interpre­ tation. Susan M. Suzman (University of the Witwatersrand) analyses Morpho­ logical accessibility in Zulu. She explores a well-known problem, the vulner­ ability of morphology in language impairment. Such vulnerability is well at­ tested and varies cross-linguistically according to language type. Such a differ­ ential access to morphology, observed by several researchers, is discussed in two case studies of language impairment in Zulu. She investigates the morpho­ logical development by considering the multiple and diverse rules of morphol­ ogy: the range of noun classes, the agreement, and the agglutinative mor­ phemes used by normal and language-impaired Zulu-speaking children. The data considered reflects differential access to morphology depending in part on the obligatory or optional status of morphemes in the language. Considerations on the access to core grammar mediated by language-specific organisation and representation of basic concepts are offered. The study on Language production in Japanese preschoolers with SLI: Testing theories is a joint research project by Yumiko Tanaka Welty (Univer­ sity of Tochigi), Jun Watanabe (Arts Junior College, Osaka) and Lise Menn (University of Colorado). In their investigation they reconsider two claims: the etiological claim, according to which SLI is a unitary disorder with a single cause and the cross-linguistic claim, independent of the etiological claim, ac­ cording to which SLI (or different SLI's, if there are several) will be underlyingly comparable across languages. That is, if both of these claims are true, then a theory, which fails to explain SLI in even one language, cannot be an ad­ equate theory of SLI in general. If a general explanation offered for SLI cannot

XViii

ELISABETTAFAVA

apply to a particular language, but children speaking that language in fact have SLI, then either they have a different type of SLI, violating the etiological claim, or the explanation is inadequate. From this theoretical standpoint, Japanese SLI is discussed. The authors hypothesise that children with JSLI, and by implica­ tion, children with SLI in general, have a general language-processing deficit, which means that they are under 'overloaded' conditions. In the fourth section, issues of grammar and cognition are discussed. The linguistic and non-linguistic accounts, the interface of grammar and pragmatics in the characterisation of some deficits are reconsidered. Among the assump­ tions going back to the early 1950s is that there is a component of the human brain dedicated to the language faculty, which interacts with other systems. After the publication of Jerry Fodor's Modularity of the Mind in 1983, issues of modularity have made a profound impact on linguistic research. One major issue is whether pragmatics should be considered a module of the grammar or not.7 The evidence for the nature of genetic deficits is offered mainly by anal­ ysing SLI and Williams Syndrome (WS). WS is a rare genetic disorder, which presents a variety of cardiovascular difficulties, failure to thrive in infancy, etc. The neurolinguistic profile is characterised by relative strengths in language, facial processing and social cognition, and profound impairment in spatial cog­ nition, planning, and problem solving. Testing linguistic concepts: Are we test­ ing semantics, syntax or pragmatics? is the question raised by Leah PaltielGedalyovich (Ben-Gurion University). She investigates what kind of knowl­ edge is being assessed in the variety of standardised tests of language acqui­ sition available to speech/language clinicians, which are purported to assess children's linguistic, i.e., syntactic or semantic, knowledge. In particular, she considers the linguistic knowledge necessary to comprehend and produce coordinated sentences, and the interaction of this knowledge with pragmatic knowledge. She argues that in some cases it is pragmatic, rather than syntactic or semantic knowledge, which is required to complete test items successfully. Test items may reflect the interaction between the semantics of co-ordinators and the influence of the Grice's pragmatic principle of quantity. It appears that the failure of children to complete the task successfully reflects impaired prag­ matic and not impaired syntactic ability. An analysis of a failure as a pragmatic failure as opposed to a linguistic failure will affect the direction of a treatment program. In order to assess children's communicative abilities accurately, and 7

On the two opposite claims and their different underlying assumptions, cf. Asa Kasher. "Pragmatics and the Modularity of Mind". Pragmatics: A reader ed. by Steven Davis (Oxford: Oxford University Press, 1991), 567-582; and Deirdre Wilson & Dan Sperber. "Pragmatics and Modularity", ibid., 583-595.

INTRODUCTION

XiX

to design appropriate remediation programs for them, it is crucial to analyse ac­ curately the tools of assessment. The need for pragmatic knowledge to com­ plete test items, which purport to test purely linguistic knowledge, challenges standard interpretation of these tests. However, the usefulness of these tests is not denied. Rather, it is necessary to recognise the various skills needed in or­ der to use test results appropriately in diagnosis and in the planning of the treatment of language delayed and disordered children. The contribution by Dusana Rybárová (Slovak Academy of Sciences) on SLI and modularity: Linguistic and non-linguistic explanations argues that the current status of the debate between the proponents of linguistic and non-lin­ guistic approaches to explanation of SLI is characterised by a severe divide between these two approaches. Both are outlined from the point of view of one of their proponents, so as to show their premises and weak points in relation to the modularity issue. She argues that, although claiming to be completely dif­ ferent, both the linguistic and non-linguistic accounts are based on the same idea of 'strong modularity', characterised as nativist, static and non-interactionist. Such understanding of modularity causes them to appear as mutually ex­ clusive. To her, the concept of 'strong modularity' seems to be an unproduc­ tive and inappropriate view in accounting for the broad array of SLI symptoms, symptoms that probably result from different primary deficits. Rybárová ar­ gues in favour of a possible way to bridge the gap between those two posi­ tions, by suggesting the abandonment of the 'strong modularity' thesis in fa­ vour of a finer-grained view on modularity, 'weak modularity'. Vesna Stojanovik, Mick Perkins & Sara Howard (University of Sheffield) debate The language/cognition interface: Lessons from SLI and Williams Syn­ drome. Despite years of research in this area, the issue of language/cognition dissociation as evidenced in cases of WS and SLI is far from being resolved. According to Stojanovik, Perkins & Howard, there have been hardly any stud­ ies which consider in detail the linguistic and non-linguistic functioning in the same subjects with WS or SLI, respectively, or which provide linguistic and non-linguistic data from both populations with WS and with SLI. They investi­ gate the complex patterns of impairment in WS and SLI, highlighting aspects of the language/cognition interface. The question addressed is what aspects of language might develop independently, and whether pragmatics belongs to the linguistic or the cognitive system or whether it is a domain where both linguis­ tic and cognitive modules interact. When detailed information about linguistic, cognitive and communicative functioning is obtained for the same subjects, the two profiles may not be as opposed to each other as has been claimed in the lit­ erature. Their results suggest that the language/cognition interface is still terra

XX

ELISABETTAFAVA

incognita and may have been approached too simplistic ally up until now. This has important implications both for linguistic theory and for speech and lan­ guage therapy. It is extremely important for clinicians to be aware of the degree of individual variation possible in conditions such as WS and SLI. Given the current state of our knowledge about these language deficits, intervention pro­ grammes should be sensitive to the unique interplay of linguistic, cognitive, and pragmatic abilities in each individual, and should not assume that all cases of WS and SLI conform to an even relatively homogeneous profile. The final section is devoted to acquired language problems subsequent to cerebral damage, to be precise non-fluent aphasia, associated with lesions in the frontal part of the left cerebral cortex, and fluent aphasia, associated with lesions in the posterior areas of the left cerebral cortex grammatical structure. The data discussed derive mainly from Catalan, Dutch, Spanish, English, Ger­ man, and Greek. In Grammar and fluent aphasia Susan Edwards (University of Reading) offers some evidence regarding the many frequently spurious generalisations on fluent aphasia. These generalisations are due to the fact that the much of the research is not sufficiently linguistically oriented. Fluent apha­ sia is usually characterised as a disorder of lexical-semantic access. Any prob­ lem in sentence structure is seen as arising from difficulties in accessing lexical items which in turn arise from either semantically based problems or from problems in phonological representation. Although these lexical problems are characteristic of all fluent aphasic speakers, lexical deficits cannot account for all the errors found in fluent aphasia. The view that grammar is preserved in fluent aphasia and that any errors observed result from faulty lexical retrieval has gone largely unchallenged, perhaps because the unequivocal lexical errors in fluent aphasia are more obvious, often more 'flamboyant' and more frequent than the subtle, less frequent, grammatical deficits. On the contrary, the evi­ dence offered by the fluent aphasic subjects described by Susan Edwards is not so neat. They were better at sentence construction than typical agrammatic pa­ tients and they produced utterances that were considerably longer than agram­ matic speakers can produce. However, the errors made by the fluent aphasie speakers were of the same kind if not the same magnitude as those made by agrammatic speakers. The data discussed support the conclusions that, in the same way as non-fluent aphasie patients, fluent aphasies have faulty access to their grammar. The assumption that syntactic abilities are intact in this particu­ lar aphasie population can no longer be assumed. Anna Gavarro (Universitat Autònoma de Barcelona), in Failure to agree in agrammatism, focuses on the linguistic evidence provided by agrammatism, which has a bearing on the evaluation of competing linguistic theories. She

INTRODUCTION

xxi

explores how linguistic theory informs the study of agrammatism, especially the empirical domain of inflection, within competing Minimalist theoretical constructs. While the deficit of inflectional markers, including case markers, and function words associated with aphasia was recorded long ago in the sci­ entific literature, the first approach to inflectional disorders in agrammatism within generative grammar occurred in the 1970s. Results considering phono­ logical clitics, phonological words, morphological objects and syntactic atoms are debated: the question is whether errors derive from syntactic impairment rather than a morphological or phonological impairment. The categories af­ fected by this syntactic impairment are tense, aspect and person on verbs, gen­ der or number on Determiner/Noun categories, or are related to displacement of constituents within a structure. Gavarro points out that asymmetries may arise between production and comprehension. Grammaticality judgments in agrammatic subjects are not central to the thesis she defends. She attempts to charac­ terise the disruption that affects grammatical derivations; the reasons why this disruption is not apparent in all cases remain a topic for future research. According to her analysis, inflectional and word-order deviations constitute a natural class, and agrammatism results in part from inability to apply the op­ eration Agree. This hypothesis lends support to the latest version of Chom­ sky's minimalist position over previous ones. The Verb and Sentence Test: Assessing verb and sentence comprehension and production in aphasia by Judith Rispens & Roelien Bastiaanse (University of Groningen) and by Susan Edwards (University of Reading) offers some crucial reasons for constructing a new test for the diagnosis and treatment of aphasics, the Verb and Sentence Test (VAST), assessing verb and sentence com­ prehension and production in aphasia. From clinical experience and theoretical investigations, it has become clear that disorders at the level of verbs and sen­ tences are frequent in aphasic patients. Verbs play an important role in sentence comprehension and production; a disorder in processing verbs therefore has a great effect on linguistic processing and, very importantly, on communicative ability. There are currently few standardised assessment materials, which sys­ tematically investigate disorders at the level of verbs and sentences. It seems that a gap exists between neurolinguistic findings and clinical application of this knowledge. The test battery for verb and sentence processing developed uses linguistic insights and is theoretically motivated; it is regarded as suitable for different types of aphasie patients and it is clinically relevant. The only tasks which have been included are those which reveal impairments, which have been described adequtely in the literature and for which therapy pro­ grams or methods exist. These principles have led to the inclusion of ten sub-

XXii

ELISABETTAFAVA

tests which assess the processes involved in comprehending and producing verbs and sentences: verb retrieval, processing information regarding the gram­ matical and thematic roles, 'mapping' of thematic roles onto grammatical roles and, for verb and sentence production, retrieving morpho-syntactic features such as subject-verb agreement inflection. This evaluation of sentence compre­ hension and sentence production deficits is constructed to apply to many different languages. The second part of the study analyses the first two case studies of the application of the VAST test to Dutch and English patients. Direc­ tions for therapy are also proposed. Esther Ruigendijk (Utrecht University) focuses on Case assignment as an explanation f or determiner omission in German agrammatic speech. She offers a linguistic, psycholinguistic, and aphasiological background to agrammatic aphasics. Her thesis is that the production of determiners and pronouns is not impaired because they are grammatical morphemes, but because their realisa­ tion is dependent on the production of finite verbs. The basic problem is the production of finite verbs; poor determiner and pronoun production can then be considered a side effect. The distinction between functional categories is at the heart of present-day grammatical theory, but plays an equally central role in, among other topics, theories on language acquisition and aphasia. In the course of time, various diagnostic criteria have been identified which distinguish one class from another: e.g., productivity, distribution, and absence versus pres­ ence of semantic content. Her results on German determiner production and case assignment demonstrate that once the case-assigning verb is realised, the production of determiners is possible in agrammatic aphasics. This means that determiner production as such is not impaired, but rather that the problems with determiners are closely related to those of the production of verbs. When no case-assigning verb is realised, no determiners can be produced. Incorrect or incomplete retrieval of the lemma information of the verb can account for case substitution errors that are made with object noun phrases. Ruigendijk's results demonstrate that once the verb is realised, the production of determiners is less impaired in agrammatic aphasies who omit determiners in their sponta­ neous speech production. This has interesting implications for the treatment of agrammatic patients. Training in the production of isolated determiner phrases cannot be successful, since these are related to verbs. It is better to train apha­ sies in verb production, as the results of the present study suggest that this will increase the production of complete noun phrases. Kyrana Tsapkini, Gonia Jarema & Eva Kehayia (University of Montreal) discuss The role of verbal morphology in aphasia during lexical access: Evi­ dence from Greek. They explore the role of regularity in language breakdown

INTRODUCTION

XXiii

by investigating how these phenomena are manifested in the performance of Greek-speaking aphasics. The dissociation between regular and irregular mor­ phology in aphasia has been documented in different languages. Broca's apha­ sics were found to be impaired in their application of the past-tense inflection while having no difficulty with irregular stored forms. Conflicting results have been obtained in different languages on the question of whether regular and ir­ regular inflection can be described as two qualitatively different operations, or a single mechanism based on statistical probability or associative memory. The issue addressed is whether morphological regularity in languages allows the dissociation of the effects of morphological regularity from those of form and semantic similarity. This would allow the maintenance of the claim of univer­ sality of the distinction between rule-based and storage mechanisms in lan­ guage processing and breakdown. Greek is a language that offers the oppor­ tunity to investigate morphological regularity while keeping form and meaning overlap between base and past tense forms constant across regular and irregu­ lar forms. First, the study of morphologically regular and irregular verbs in Greek allows for the addressing of the issue of morphological regularity per se, independently of form and semantic considerations. Second, it clarifies whether the difficulties that Greek-speaking aphasies have with the past tense are due to the different morphological operations postulated, or whether they reflect task-specific deficits related to the lexical access procedures involved in each task. It is the very presence of a stem-allomorph that marks the most 'ir­ regular', in the sense of idiosyncratic and unpredictable, past tense formation, the authors note. In conclusion, this volume investigates relevant intersections among differ­ ent linguistic frameworks, languages, clinical approaches, and pathologies. There is unity, however, in the great effort made to provide some key aspects of the framework adopted and to discuss general problems posed by it. The theoretical arguments advanced and the empirical evidence proffered are bound to offer deeper insights into the factors that shape the nature of language. In this perspective, theoretical linguistics has been and will continue to be of par­ ticular significance for the study of speech pathology and speech therapy, giv­ ing it new life and scope. At the same time, it can be said that clinical research has offered new perspectives on investigative techniques and suggestions as to how to revise theoretical and methodological tools.

I. PHONOLOGY IN CLINICAL APPLICATIONS

PHONOLOGY AS HUMAN BEHAVIOR THEORETICAL IMPLICATIONS AND COGNITIVE AND CLINICAL APPLICATIONS

YISHAI TOBIN Ben-Gurion University of the Negev Phonology, then, is for us an attempt to under­ stand the non-random distribution within the signal, and grammar is the attempt to under­ stand the non-random distribution of the signals themselves. (William Diver 1975:13)

1. Introduction The theory of Phonology as Human Behavior (PHB) was developed by William Diver (1979) in an analysis of the non-random distribution of certain classes of initial consonant clusters in English, which he later expanded to explain the non-random combinations of vowels and consonants in English (Diver 1993) and in language in general (Diver 1995). PHB has been extended further to explain the combinatory phonology of consonant and vowel pho­ nemes in other languages such as Italian (Davis 1987 [1984]), Hebrew (Tobin 1990b,c), Urdu (Azim 1989, 1993, 1995, 1997; Hameed 1999; Jabeen 1993), Mewati (Fatihi 1987) and Spanish (Flores 1997). PHB has also been employed to explain the non-random distribution of initial consonant clusters in fortytwo different languages representing diverse language families as well in arti­ ficial languages such as Esperanto and Klingon (Tobin 2000a); it has been ap­ plied to the areas of developmental and clinical phonology and prosody in languages such as Hebrew, Finnish and Japanese (Moore 1991a,b, 1993; Moore & Korpijaakko-Huuhka 1996; Moore & Rosenberg-Wolf 1998; Tobin 1995, 1997a,b, 1999, 2000b, 2001a; Tobin & Miyakoda 2001b); and it has been compared to and contrasted with other functional and quantitativeoriented phonological theories (Tobin 1988c, 2000b). 1.1 Historical background PHB may be viewed as part of the historical development of a structural, functional and cognitive tradition in 20 century linguistics. This tradition begins with Ferdinand de Saussure's (1959 [1916]:34-68) concept of system, and the

4

YISHAI TOBIN

dichotomies of langue and parole and phonetics and phonology. The latter, fundamental dichotomy between the abstract code and its realization, based on a classification of sounds according to their articulatory and acoustic features, was further developed by the functional, communication-oriented Prague school phonology developed by Nikolai Trubetzkoy (1969 [1939]) and Roman Jakobson (1968 [1941], 1971). The communication factor adhered to by the Prague school was then supplemented by André Martinet's (1955) introduction of the human factor through the concepts of "assymetry" and "economy of ef­ fort in phonological change." Martinet maintained that phonological systems are arranged asymetrically and change in such a way that the non-random diachronic distribution of sounds reflects the search for equilibrium and harmony within the system as it is affected by the principle of least effort in human be­ havior. The principle of minimal effort postulated by Martinet implies that speakers strive for a minimal number of phonemes which require the least amount of effort to be produced and combined together in what Sampson (1980:112, reviewed in Tobin 1986), referred to as a "therapeutic view of sound change." It is Diver (1974, 1979), however, who has shown that a more complete theory of phonology has to take both the communication factor and the human factor into account together. Diver maintains that there is a constant struggle between our need for maximum communication and our desire for minimum effort. The communication factor (requiring a large number of pho­ nemes demanding a great deal of effort) is in conflict with the human factor (striving for minimal effort), which results in a trade-off between the two. This synergetic compromise between the communication factor and the human factor is reflected in the fact that there is a similar number (20-30) of phonemes of varied proportional degrees of difficulty acquired in a similar order in the languages of the world: less than 20 phonemes would reduce the com­ munication potential and more than 30 would be too difficult to learn, remember and produce. That is, Diver extends and enhances Martinet's more strictly diachronic view of the human factor so that it can become a means of explaining the non-random distribution of phonemes in language in coordination with the needs of communication as originally established by Saussure and the Prague school (discussed in Liberman 1991; Tobin 1988a-c, 1990a, 1996). 1.2 Theoretical background Diver (1995:61) justifies in the following way this synergetic combination of the roles played by the human factor and the communication factor in pho­ nology as orientations which motivate the hypotheses used to discover and analyze the potential observational data of sounds:

PHONOLOGY AS HUMAN BEHAVIOR

5

The observations, the sound waves, are produced by people. The problem is to discover what motivates people to produce particular sound waves on a particular occasion. The purpose of the hypotheses is to reveal motivation by demonstrating the relation of the observations to the orientations, for the orientations represent the con­ trolling circumstances of the entire procedure. The way this works out, in analytic practice, is that the essential motivation is communication, and that the details of how communication is effected are controlled by considerations of the available human physiology, by principles of acoustics, and by normal characteristics of human behavior; that is, by the other orientations.

Although PHB has maintained the basic Saussurian and Praguian notions of phonemes, distinctive features, and morphemes, etc., Diver (1995:61) views these familiar concepts in an innovative way. His particular view of language and linguistic analysis develops from the orientations and the hypotheses out­ lined above: The first task of the hypotheses is thus to establish the identity of the inventory of the units of language (beginning with phonemes and morphemes) and the nature of those units, their nature being an inference from how they are made use of in the process of communication. The other orientations constantly feed in to the process of reaching the primary goal, particularly in regard to why the units are made use of in the way they are. As analysis proceeded (in our work), the hypotheses sorted themselves into three distinct groups, distinct in the way in which they are related to the process of communication: hypotheses about phonological units, about grammatical units, and about lexical units.

Diver (1979, 1995) thus has postulated different kinds of distinctive features in phonology, which are more directly related to human physiology, perception, cognition, and behavior. Based on these new distinctive features, it is possible to classify different sets of phonemes according to the degree of difficulty required to learn, perceive, and produce them. Diver has also introduced the notion of the quantitative analysis of favored and disfavored distributions of phonemes, which can be directly related to their degree of difficulty. These distributional preferences or 'preferred combinations' are further paralleled by similar ones observed in human behavior requiring the control of fine motor movements in extralinguistic contexts (e.g., common daily tasks or sports), that is in other instances of learned behavior, as language is. In practice, Diver's (1995:62) theoretical perspective presents the phonological units of language in a unique way based on the interaction between the communication factor, which enables him to tackle the following questions: why are there certain kinds of sounds in human language and why do these sounds combine in a non-random way to form larger units in a similar way across languages?

6

YISHAI TOBIN The sounds are produced through the control of the musculature of the vocal tract. The first requirement is that audible sound be produced, if the communication is to be transferred from one person to another; clearly audible sound if the communication is to take place over any distance. The second, that a fairly extensive inventory of distinct sounds be built up, if the resources of human communication are to exceed those of animal cries. There are two important considerations here: first, given the variety of the ways in which sounds could be produced by the vocal apparatus, why is there so much agreement in general across languages, in the midst of a great deal of differences in matters of detail? Second, given that an inventory of grammatical and lexical units is to be built up through the combination of phonological units, just how is an expanding inventory of phonemes related to the expanding structure of the mor­ pheme?

Diver (1995:62-65) establishes the criteria for the basic inventory of audi­ ble sounds on the grounds of the physiological and acoustic properties of the human vocal tract, which serve to produce maximum acoustic differentiation with minimum precision of control (e.g., the vowels of the /i, a, u/ triangle can be used independently for the formation of morphemes; cf. the English indefi­ nite article a). He then shows how the phonemic inventory is expanded by ad­ ditional techniques of shaping and exciting the vocal tract to include interme­ diate vowels and consonants (involving the use of the dorsum, lips, apex and other active articulators). These consonants then 'flank' the 'keystone' vowels to form morphemes of the cvc variety. The possible order of phonological development in language is presented and explained by Diver (1995:67) based on the interaction of the communication factor and the human factor as follows: For the 'primary' units we thus get a picture of an imagined sequence of development ordered in terms of the need for precision of control. The units are here symbolized in ways that foreshadow their later phonological status: (1) The single-cavity /a/, furnishing undifferentiated resonance for the excitation of the vocal folds. (2) The development of a two-cavity system, using the dorsum and lips as articu­ lators, introducing /i/ and /u/ and thereby converting /a/ to a third member of a system, maximally differentiated from the other two, rather than a unique unit. (3) The use of the apex in a fairly undemanding way, giving another shape to the cavity, /1/, without recourse to dorsum and lips, still with excitation by the vocal folds. (4) The development of fine motor control over the apex, as it is brought into use as a means of excitation, as well as shaping, of the cavity, in the maximally differ­ entiated positions of/t/ and /s/.

PHONOLOGY AS HUMAN BEHAVIOR

7

Diver (1995:68-69) then shows how the communication and the human factors interact in the relation between phonemes and morphemes in an iso­ morphic way: In general, the principles that underlie the ordering of acquisition, as presented above, are the same as those reflected in the construction of morphemes in current speech; that is, in the utilization of the various members of the inventory of phonemes. Those that can be characterized as being more difficult to control (e.g., in terms of intermediate position), are the same ones that are, in general, made use of less frequently (or perhaps not at all) in the structuring of the morpheme. The most audible members of the inventory - ill, /a/, /u/, together with intermediate units regularly form what may be called the keystone in the arch of the structure of the morpheme. It is rare indeed for a 'word' to be constructed without one of the clearly audible members as keystone.

The flanking members are drawn from the less audible members of the in­ ventory, and the preference among these is very much in terms of how much demand is made on precision of control. Further, the extent to which a number of flanking members are combined in the formation of a single morpheme is very sharply restricted. Diver (1995:69) then attributes the restrictions on the phonotactic distribu­ tion of the consonants flanking the vowels in a syllable, and more generally the phonotactic favorings and disfavorings of sounds in larger units of communication such as morphemes and words, to his particular view of the human factor as a form of precision of control: In terms of precision of control, it is evident that the more phonological units are combined in the formation of a single morpheme, the more control is required, in terms of the number of individual members to be produced and in terms of the coor­ dination of the sequencing. An increase in the number of phonological units in a morpheme type leads to an in­ crease in the potential number of morphemes, but to a decrease in the actual number, in terms of actual numbers as well as in comparison with the potential. To the extent, then, that the selection of a morpheme in the text leads to the characteristics of the sound waves, we see that the "human factor" motivation, preci­ sion of control, is favoring some kinds of combinations of sound waves over others.

Apparent skewings in the form of favorings of certain phonological units on the morpheme level resulting from the interaction between the communica­ tion factor and the human factor play a major role in Diver's (1995:71) theo­ retical approach and even determine the kind of data he uses in his analyses 'words' or canonical morphemes:

8

YISHAI TOBIN A number of hints have been given that there is a quantitative basis for the comments about the favoring of certain uses of the phonological units over others. The following procedures have been used. A collection is made of all 'words' constructed with a single keystone [vowel], within which there is no morphemic boundary. Thus cat would be counted, but not cats. The rationale is that the combination t-s here is determined by something other than purely phonological considerations.

Diver's (1979, 1993, 1995) focus on phonotactics, or combinatory phonology, is reflected in his analysis of the non-random distribution of pho­ nemes in general and in alternative canonical morpheme and word positions in particular, which are directly related to the interaction between the human factor and the communication factor. Despite the fact that consonants are more difficult to produce than vowels, there are more consonant than vowel phonemes in languages because consonants provide clearer communicative distinctions. In morpheme/word-initial position - where the burden of com­ munication is the highest - there is an almost random distribution of conso­ nant phonemes, which, however, reflects their different degrees of difficulty, that is, with a slight favoring of labial, or labio-dental, or apical-dental pho­ nemes, which are perceived both aurally and visually. In syllable/word-final position - where the burden of communication is the lowest - there is a highly significant statistical preference for phonemes requiring the least amount of effort to produce, namely apical (the easiest to produce) and voiceless (which require the excitation of only one set of oral articulators) rather than voiced (which require the excitation of two sets of articulators: the oral articulators and the vocal folds) or nasal (which require the excitation of three sets of articulators: the oral articulators, the vocal folds, and the uvula). 1.3 Methodological background Diver's approach differs from other phonological frameworks not only in its theoretical orientations and hypotheses, but in its methodology as well. Three different procedures for collecting data have been outlined by Diver (1995:71-72) as evidence in support of the hypotheses of his theory: The collection is intended to be complete, but there is a problem as to just what con­ stitutes 'completeness'. Three different kinds of collection will be noted here. In my own research, the collection consisted of all words in my active vocabulary: that is, in the active vocabulary of a literate, well-educated speaker. It is recognized that the resulting list might not be exactly the same as that of another person of a similar education, and that there might well be quite a number of words "in the dictionary" that would not be included.

PHONOLOGY AS HUMAN BEHAVIOR

9

At the other extreme is a study by Shazi Shah Jabeen (Jabeen 1993) of the phonology of Bihar Urdu in the neighborhood of Gaya, India. Here, informants were used, essentially illiterate persons without access to radio or television. The author is herself a native of Gaya, and checked for the occurrence of every possible phonological combination. A third alternative was a study by Joseph Davis (Davis 1987 [1984]). This study based the count on essentially all the entries in a moderately sized dictionary of Standard Italian. In each of these, counts were made of the frequency of occurrence of the phonological units in various positions and in various combinations within the mor­ pheme. In spite of the differences in the inventory of phonological units (between English, Urdu and Italian) and the differences in methods of collection, the general principles that have been sketched above were found to be quite uniform in their manifestation. It was thus possible to check the interrelations among orientations, hypotheses, and observations, and to demonstrate that there is a relationship of motivation flowing from orientation to observation.

In order to uncover all the data necessary for our analyses of initial conso­ nant clusters across languages, Tobin (2000a) examined all such clusters ap­ pearing in monosyllabic words in standard dictionaries. The developmental and clinical studies discussed here were based on recordings of spontaneous speech as well clinical exercises. For similar methodological reasons lexical analyses from standard dictionaries were performed by Tobin (1990b,c) for close to 3,000 triconsonantal (ccc) roots in Hebrew. Specific text or discourse analyses testing the theory on various texts, including a poetic text containing many neologisms, appear in Tobin (1997a: ch. 6). In all cases, regardless of the methodology employed in the collection of the data, the principles established by the theory were supported. 1.4 Theoretical and methodological conclusions supported by the theory The following phonological and phonotactic parameters have been ex­ plicitly derived from the theory (adapted from Diver 1979; Davis 1987 [1984]; Tobin 1990b,c, 1995, 1997a,b): (1) the identification of active articulators (versus the traditional category of place of articulation, which is often a label for passive receptors) and the relative difficulty of learning how to control them; (2) the identification of relative degrees of constriction and turbulent and non-turbulent airflow (versus the traditional category of manner of articulation), which require different articulatory control (mobile and stable) and produce different acoustic patterns for individual sounds and phonation processes (labialization, apicalization, velarization, nasalization, and glottalization);

10

YISHAI TOBIN

(3) the identification of the number of sets of articulators to be controlled (versus the traditional categories of voicing, the fortis-lenis distinc­ tion, and nasality) that require different levels of articulatory control and produce different acoustic patterns; (4) the identification of 'phonemes of constriction' and 'phonemes of aperture' (versus the traditional concepts of consonants and vowels). The following principles have been derived from the quantitative results obtained from the theory (adapted from Diver 1979; Davis 1987 [1984]; Tobin 1990b,c, 1995, 1997a,b, 1999): (1) additional articulators are disfavored; (2) coarticulation by near articulators is disfavored; (3) coarticulation by the same articulators or of the same phonemes is even more highly disfavored, particularly in the roots of Semitic lan­ guages (Tobin 1990b,c); (4) different word or root positions have different communicative force, and thus affect the favoring and disfavoring of different articulatory and acoustic features and phonemes; (5) visual articulators are favored, particularly in word/root initial position; (6) explosive (mobile) phonemes are favored in initial position; (7) turbulent (stable) phonemes are favored in final position; (8) transitions from one distinct constriction to another within a single phoneme are disfavored; (9) consonant clusters concerning different articulatory and acoustic fea­ tures are restricted (e.g., mobility/stability; Diver 1979); (10) among constrictions, maximal constriction is favored, and among apertures, maximal aperture is favored; (11) sequences of phonemes with the same articulators are disfavored unless their juxtaposition is, by virtue of some other factor, mutually beneficial; (12) apical consonants are favored. The following principles have been added to the theory based on the appli­ cation of the theory itself to clinical studies (adapted from the works found and cited in Tobin 1995, 1997a, 1999): (1) the preservation of as many distinctive features as possible (usually 2 out of 3) in substitution processes which require more effort than de­ letion processes;

PHONOLOGY AS HUMAN BEHAVIOR

11

(2) the preservation of as many communicative oppositions as possible in the original word (e.g., the number of phonemes per word) in substitution processes which require more effort than deletion processes; (3) the use of a phoneme already available in the speaker's repertoire in accordance with the immediate phonetic environment in substitution processes which require more effort than deletion processes; (4) the preservation of the original phonetic structure of the word in dele­ tion processes not involving syllable reduction and in reduplication processes; (5) if the original structure of the word is reduced by the deletion of syl­ lables, the stressed syllable bearing the most communicative infor­ mation is maintained; (6) if the original structure of the word is enlarged by epenthesis, the epenthesis makes the transition to or between more difficult sounds easier. PHB supports the idea that there is no pure synchronic phonology, and that instead the dynamic interplay between the communication factor and the hu­ man factor both motivates and serves as a constant control over language change. Therefore the following conclusions may be drawn for synchronic and diachronic phonological analyses: Conclusion 1: Language in general - and phonology in particular - can be seen as a synergetic mini-max struggle: the desire to create maximum communication with minimal effort (Tobin 1990a: ch. 3). As it has been applied to the articulatory errors and processes found in de­ velopmental phonology and in functional and pathological clinical phonology, PHB basically represents a more extreme version of Conclusion 1 where the human factor often overrides the communication factor, and clinical interven­ tion serves as an attempt to balance the two: Conclusion 2: Developmental and clinical speech processes and errors may be viewed as an extreme version of this synergetic mini-max struggle: there is less than maximum communication because of either extreme minimal effort or a lack of control over the articulatory tract or mechanisms. Greater effort will be exerted in order to achieve more efficient or better communication through clinical intervention (Tobin 1995, 1997a,b, 1999).

12

YISHAI TOBIN

2. Phonological processes and phonology as human behavior Natural phonology, one of the leading theories in developmental and clini­ cal phonology, introduced the concept of natural phonological processes (Stampe 1979 [1972], Dressier et al. 1987). The following include most of the major and minor natural phonological processes found in functional language acquisition (adapted from Grunwell 1987; Ingram 1990), accompanied by an explanation derived from the principles obtained from PHB. 2.1 Functional processes influencing syllable structure (1) Final consonant deletion: cvc → cv: out [au], hike [baI] Explanation: word-final position has less communicative force; consonants require more articulatory control (i.e., are harder to make) than vowels. (2) Deletion of unstressed syllable (usually in word-initial position): banana [næena] Explanation: stressed syllables give more communicative, perceptual, and cognitive information than unstressed syllables; the more syllables in the word, the more effort it takes to pronounce it; therefore, the word-initial position, which usually has the greatest communicative force, carries less information in non-initially stressed words. (3) Consonant cluster reduction: CC → C:floor [for], step [tep] Explanation: a consonant cluster requires greater effort than a consonantvowel sequence and may be reduced or replaced at the expense of maximum communication; in addition, coarticulation by near articulators is disfavored; phonemes of constriction give clearer communicative distinctions than pho­ nemes of aperture - that is why there are more consonants than vowels in lan­ guage - but they require more articulatory control (hence the ideal cv syllable). (4) Reduplication: repetition of a syllable or part of a syllable: rabbit [w→wæwæ], noodle [nunu] Explanation: reduplication is often a means to avoid more difficult sound combinations and/or to maintain the number of syllables in the word: se­ quences of phonemes with the same articulators are disfavored unless their juxtaposition is, by virtue of some other factor, mutually beneficial. We also found that newly acquired sounds were often reduplicated as a means of practice or of hypercorrection in the clinical situation.

PHONOLOGY AS HUMAN BEHAVIOR

13

(5) Coalescence: characteristics of 2 consecutive sounds merging into 1 sound: swim [. im], slide [ aid] Explanation: fewer articulatory gestures resulting in fewer distinctive units at the expense of maximum communication. (6) Epenthesis: addition of segments, usually an unstressed vowel: snow [səno , drum [dər m] Explanation: the additional unstressed vowel often eases the transition to more difficult consonants or clusters. The clusters then may be reduced at the ex­ pense of communication. Examples (5) and (6) above, as well as the following assimilation processes (1) and (2) below, may also be explained by the factor: sequences of phonemes with the same or near articulators are disfavored unless their juxta­ position is, by virtue of some other factor, mutually beneficial. 2.2 Assimilation processes (Consonant/Consonant-Vowel harmony) (1) Velar/Nasal/Labial, etc. assimilation: duck [];friend [fire]; top [bap] Explanation: a non-velar/nasal/labial sound changes to a velar/nasal/labial be­ cause of the influence of, or the dominance of, a velar/nasal/labial sound which entails fewer articulatory gestures at the expense of maximum communication. (2) Prevocalic voicing of consonants: pen [ben], tea [di] Explanation: an unvoiced consonant generally becomes voiced before a vowel: the speaker anticipates the control of two sets of articulators in what is usually a longer acoustic phonological segment. (3) Devoicing of final consonants: bed [bet], big [blk] Explanation: additional articulators are disfavored; voiced consonants become unvoiced in word-final position: where the communicative force is least im­ portant or crucial, the speaker opts to activate one set of articulators rather than two. This may also be related to the fact that vowels are shorter before voiceless rather than voiced consonants. 2.3 Substitution processes (1) Processes reflecting the substitution of active articulators: a. Fronting: back (non-apical) consonants are replaced by apical con­ sonants usually preserving the same manner and voicing values:  → t book [but], g → d dug [dd], → s shoe [ u ] , Z → z beige [], r→ n sing [sin]

14

YISHAI TOBIN

Explanation: the apex is the most flexible and easy to control of all the active articulators: the earliest and most frequent examples of the substitution of active articulators are fronting or apicalization, which sharply reduces the number of communicative distinctions made by the speaker. b. Backing: the back pronunciation of front sounds (usually conso­ nants): t/d → k/g: dog [], tiger [kaIgə] Explanation: a later, less frequent (possibly idiosyncratic) process where the dorsum (or another back articulator) replaces the apex (or other front articula­ tors); often found in children who have difficulty controlling the musculature of the apex (for organic or other reasons) and/or try to reduce the number of communicative distinctions made by the apex (or other front articulators), especially as a result of earlier fronting or other processes. (2)

Processes reflecting the substitution of turbulence and airflow: a. Stopping: fricatives/affricates are replaced by stops: s/ts —> t: seat [tit], soup [dup] Explanation: maximum constriction is favored particularly when mobile (stop) phonemes of constriction are easier to control than stable phonemes of less constriction which require greater control of the musculature to create and maintain a small aperture for a stronger turbulent airflow (the most frequent manner of substitution for children). b. Affrication: stops/fricatives are replaced by affricates: t → t : tree [t i] Explanation: maximum constriction precedes and leads into partial con­ striction and turbulent airflow since mobile (stop) phonemes of constriction require less articulatory control than stable phonemes of constriction which produce greater turbulence (a less frequent manner substitution for children, possibly to avoid more difficult consonant clusters or combinations).  Gliding of liquids: l/r → j/w: rock [wak], lap [jæp] Explanation: substitution of a higher for a lower degree of aperture (from con­ sonants to semi-vowels) which may also require less articulatory control. d. Vocalization: nasals and liquids (syllabic consonants) are replaced by vowels: apple [æp], flower [fawo] Explanation: the favoring of maximal aperture particularly when phonemes of aperture (vowels) require less articulatory control than phonemes of constric­ tion (consonants).

PHONOLOGY AS HUMAN BEHAVIOR

15

e. Denasalization: m → b, n → d, → g: no [do], home [hob], sung [s g] Explanation: additional articulators are disfavored; nasal stops become their equivalent oral stops: two sets of articulators are exploited rather than three. f. Deaffrication11 → ', d3 → 31 chips [∫vps], juice [3US] Explanation: transitions from one distinct constriction to another within a sin­ gle phoneme are disfavored; a more complex sound requiring greater effort or control is reduced to a less complex sound after the speaker has acquired the ability to produce the more difficult stable sounds. g. Glottal replacement: butter [Λ?] Explanation: additional articulators are disfavored; a glottal stop replaces an intervocalic consonant or a consonant in syllable-final position: articulatory control of one set of articulators rather than two in an appropriate phonetic environment. h. Gliding of fricatives: /f/ → [w], /s/→ [l,j] Explanation: a more extreme substitution of a higher for a lower degree of aperture (cf. the gliding of liquids in (c) above) requiring less articulatory control: an idiosyncratic process. Most of the processes affecting the turbulence and airflow of phonemes of constriction (manner of articulation) listed above confirm the following factors obtained from PHB: among constrictions, maximal constriction is favored and among apertures, maximal aperture is favored. It should be clear from the above examples that more than one process can appear in the same word and that most, if not all, of these functional processes can be directly related to the principles obtained from PHB. It should also be noted that there is a chronology of natural processes which determines and separates normal processes from deviant functional or organic ones. 2.4 Functional processes in Hebrew-speaking children Shaked (1990) examined the frequency of functional errors of 20 Israeli children (from 1:7 to 2:7 years of age), obtaining the following results: (1) The most frequent functional processes found in the speech of the 20 Israeli children included: a. Fronting: particularly of fricatives: ∫/x → s ∫alom → [salom]L b. Consonant cluster reduction: p a x i m —» [paxim] ("flowers") (fricatives deleted, stops maintained);

16

YISHAI TOBIN

c. Syllable-final consonant deletion / medial consonant cluster reduction: taxtonim → [tatonim] ("underwear"); d. Devoicing of final consonants1 be ez → [be es] ("faucet") (it has also been noted in the literature that children aged 2 to 4 have particular difficulty in the production of voiced stops, which requires activating two sets of articulators for mobile consonants); e. Deletion of unstressed syllables: ambatja → [batja] ("bath"); f. Gliding: 1→j : delet → [jedet] ("door") (+ metathesis); g. Deletion of syllable/word-initial consonant (in unstressed syllable): mita → ita ("bed"); h. Deaffrication: tsav → [tav] ("turtle"). (2)

The least frequent functional errors and processes included: a. Backing (considered to be idiosyncratic by Grunwell 1987 and others): mixtav → [mitax] ("letter") (+ consonant cluster reduction, + metathesis); b. Stopping: sus → [tut] ("horse").

In this and other developmental and clinical studies, the general order of acquisition of phonemes and functional processes conform to the principles of the theory, including: i the systematic order of the exploitation of active articulators (labial apical - posterodorsal); ii the systematic favoring of consonants requiring extreme degrees of constriction and minimal airflow (stops preceding fricatives or mo­ bile preceding stable phonemes of constriction); iii the disfavoring of phonemes of constriction requiring the excitation of more than one set of articulators (when there are phonemic oppositions based on this factor): voiceless (0), most frequent, followed by voiced (+1), and the absence or extremely limited use of nasals (+2). 2.5 Functional processes in a hearing child of deaf parents Samet (1993) investigated the functional processes of a hearing child of deaf parents and compared the frequency of the functional processes found in her study with those found for Hebrew speakers. She analyzed the speech of Elita (2:8), the daughter of deaf, signing parents who are Russian immigrants with a limited knowledge of Hebrew, whose speech is extremely difficult to understand. Elita was brought for therapy because of late language develop­ ment and a medial level of comprehensibility (possibly due to lack of input).

PHONOLOGY AS HUMAN BEHAVIOR

(1) a. b. c. d. e. f. g.

17

Elita showed the following functional processes in descending order of frequency (according to Shaked 1990): consonant cluster reduction: bgadim → [gadim] ("clothes"); deletion of unstressed syllable: tsalaxat → [laxat] ("plate"); syllable-final consonant deletion / medial consonant cluster reduction: taxtonim —» [tatonim] ("underwear"); final consonant deletion: sipuy → [sipu] ("story"); initial consonant deletion 1 yam —» [(?) Jam] ("to-there"); stopping: xatula → [katula] ("kitty"); deaffrication: otsa → [ ota] ("want").

Samet then applied PHB to explain the processes. All of these processes, except for (l)(e) initial consonant deletion, clearly conform to the principles of phonology as human behavior. The possible initial consonant deletion in example (l)(e) appears in an unstressed function word le- 'to- in the phrase [∫am -> [(?) Jam] ("to-there") which, from the point of view of communi­ cation, is less crucial than the stressed lexical item∫am- "there" to which it is attached, which makes this example similar to the previously mentioned dele­ tion of unstressed initial syllables. (2) Elita also has functional processes not found in Shaked (1990): a. assimilation: beged^ [geged] ("garment") /jeled—» [leled] ("boy"); b. epenthesis (very limited and included addition of both consonants and vowels). In the case of Elita, we have a situation where the number and the degree of functional processes require clinical intervention. As we have previously stated, natural functional processes have been as­ signed chronological values and may serve as norms for child language acqui­ sition. If deviations from these chronological values and norms are discovered in a child's speech, that child may be sent for speech therapy. Children who are labeled as having functional disorders in their speech will usually reach the clinic when: i ii iii

early processes continue past their normal period; early processes coexist with later errors and processes; normal functional processes are accompanied by idiosyncratic processes (such as backing, gliding of fricatives, affrication).

18

YISHAI TOBIN

Functional processes which reach the speech clinic clearly reflect an even more extreme case of the struggle for maximum communication with minimal effort, usually under the following circumstances: i

ii

iii

The simultaneous coexistence of several functional and/or idiosyn­ cratic processes of simplification (the human factor) reduce the num­ ber of communicative distinctions (the communication factor) ex­ ploited by the child. If and when the child becomes, or is made, aware that her communi­ cation is impaired, only then will further effort be exerted to produce more communicative distinctions in the quest for maximum commu­ nication with minimal effort. In other words, the child will usually only exert effort in order to enhance communication.

3. Summary and conclusions We can now summarize the major principles of PHB for phonological analysis and its applications for developmental and clinical phonology: (1) We begin with the phonetic observations, both articulatory and acoustic, within phonemes which are abstract and therefore 'unobservable' units (child language inventory and clinical intake). (2) By means of the communication orientation we can establish the number of distinctive units of a language as found in a child or patient. (3) Consideration of the acoustic and physiological characteristics of phonetic units (e.g., the ones elicited from a child or patient) suggests a variety of characterizations. (4) In choosing among these characterizations, it is apparent that the characteristics of the units must be of such a kind that the human user can learn them both in normal developmental and pathological lan­ guage acquisition. (5) We do not know in advance, deductively, in exactly what way the human factor will interact with the communication factor and others. (6) Phonotactic skewings in language (diachronic, synchronic, and in de­ velopmental and pathological data) reflect the learning process of speakers, including children and clinical patients. (7) This skewing, viewed consistently with the human factor against the background of the other factors of communication, acoustics and

PHONOLOGY AS HUMAN BEHAVIOR

(8)

(9) (10)

(11)

(12)

19

physiology, provides us with information concerning the characteris­ tics of the phonological units. Phonology is not random, but motivated; the frequencies of the phonological units and the ways they combine are determined both by their phonetic make-up and by the speaker's (child's or patient's) exploitation of - or coping with - that make-up in the act of commu­ nication. Gestures enhancing communicative distinctiveness are favored, while articulatorily more difficult gestures are disfavored. There is a conflict between the communication and the human factors in language users' search for maximum communication with minimal effort both in the diachronic development and the synchronic state of a language. This conflict is even more keenly felt in language acquistion, where functional errors and processes may be observed, and even more so in the clinic, where developmental and pathological errors and processes are apparent. The theory of phonology as human behavior can explain, in a princi­ pled way, the connection and interrelationship between the phylogeny, the ontogeny, and the pathology of the development of sound systems in human languages. REFERENCES

Azim, Abdul. 1989. "Some problems in the phonology of Modern Standard Urdu". Paper presented at the First International Conference of the Columbia School of Linguistics, Columbia University, 24 August 1989. Azim, Abdul. 1993. "Problems of aspiration in Modern Standard Urdu". Paper pre­ sented at the Third International Columbia School Conference on Linguistics, Rutgers University, 11 October 1993. Azim, Abdul. 1995. "The phonology of the vocalic systems of Modern Standard Urdu". Paper presented at the Fourth International Columbia School Conference on Linguistics, Rutgers University, 20 February 1995. Azim, Abdul. 1997. "Revisiting the phonology of the vocalic systems of Modern Standard Urdu". Paper presented at the Fifth International Columbia School Con­ ference on Linguistics, Rutgers University, 16 February 1997. Davis, Joseph. 1987 [1984]. "A combinatory phonology of Italian". Columbia Uni­ versity Working Papers in Linguistics 8.1-99. Diver, William. 1974. "Substance and value in linguistic analysis". Sémiotext(e) 1:2.11-30.

20

YISHAI TOBIN

Diver, William. 1975. "Introduction". Columbia University Working Papers in Lin­ guistics 2.1 -20. Diver, William. 1979. "Phonology as human behavior". Psycholinguistic Research: Implications and applications, ed. by Dorothy Aaronson & Robert Rieber, 161186. Hillsdale, N.J.: Lawrence Erlbaum Associates. Diver, William. 1993. "The phonology of extremes: The correlation of initials and finals". Paper presented at the Third International Columbia School Conference on Linguistics, Rutgers University, 11 October 1993. Diver, William. 1995. "The theory". Meaning as Explanation: Advances in linguistic sign theory ed. by Ellen Contini-Morava & Barbara Sussman Goldberg, 43-114. Berlin & New York: Mouton de Gruyter. Dressler, Wolfgang U., Willi Mayertaler, Oswald Panagl & Wolfgang Wurzel, eds. 1987. Leitmotifs in Natural Phonology. Amsterdam & Philadelphia: John Benjamins. Fatihi, A. R. 1987. Economy of Articulation in Mewati Phonology. Ph.D. dissertation, Aligarh Muslim University, India. Flores, Nydia. 1997. "The distribution of post-vocalic phonological units in Spanish". Paper presented at the Fifth International Columbia School Conference on Linguistics, Rutgers University, 15 February, 1997. Grunwell, Pamela. 1987. Clinical Phonology. London: Croom-Helm. Hameed, Shahana. 1999. "Interaction of physiology and communication in the makeup and distribution of stops in Lucknow Urdu". Paper presented at the Sixth International Columbia School Conference on Linguistics, Rutgers University, 11 October 1999. Ingram, David. 1990. Phonological Disability in Children. London: Whurr. Jabeen, Shazi Shah. 1993. Economy of Articulation in the Phonology of Bihar Urdu (as spoken in and around Gaya). Ph.D. dissertation, Aligarh Muslim University, India. Jakobson, Roman. 1968 [1941]. Child Language, Aphasia, and Phonological Univer­ sals. The Hague: Mouton. Jakobson, Roman. 1971. Selected Writings I: Phonological studies. The Hague: Mouton. Liberman, Anatoly. 1991. "Postscript". The Legacy of Ghengis Khan by N. S. Trubetzkoy, ed. by Anatoly Liberman, 295-375. Ann Arbor: Michigan Slavic Publications. Martinet, André. 1955. Économie des changements phonétiques: Traité de phonologie diachronique. Berne: A. Franke. Moore, Kate. 1991a. "Speech rate, phonation rate, and pauses in cartoon and sports narrations". Studies in Logopedics and Phonetics ed. by R. Aulanko & M. Leiwo, vol. II, 135-143. Helsinki: Publications of the Department of Phonetics, Univer­ sity of Helsinki. Moore, Kate. 1991b. "A taxonomy of pauses in Finnish". Ibid., 145-150.

PHONOLOGY AS HUMAN BEHAVIOR

21

Moore, Kate. 1993. "Developmental disfluencies in preschool children". Nordic Prosody VI ed. by Björn Granström & Lennart Nord, 173-181. Stockhom: Almquist & Wiksell. Moore, Kate & Anna-Maija Korpijaakko-Huuhka. 1996. "The clinical assessment of fluency in Finnish". Advances in Clinical Phonetics ed. by Martin J. Ball & Martin Duckworth, 171-196. Amsterdam & Philadelphia: John Benjamins. Moore, Kate & Capita Rosenberg-Wolf. 1998. "Perceptions of hesitations in speech". Nordic Prosody VII ed. by Stephen Werner, 195-269. Frankfurt & New York: Peter Lang. Samet, Mirit. 1993. "An analysis of the speech of a hearing child (2:8) of deaf parents according to the theory of phonology as human behavior". Ms. Depart­ ment of Communication Disorders, Speech, Language and Hearing, The Sackler Faculty of Medicine, Tel-Aviv University. [In Hebrew.] Sampson, Geoffrey. 1980. Schools of Linguistics. Stanford: Stanford University Press. Saussure, Ferdinand de. 1959 [1916]. A Course in General Linguistics. Transi, by Wade Baskin. New York: Philosophical Library. Shaked, Galia. 1990. Early Phonological Development: Phonological processes in children 1:7-2:7 years old. M.A. Thesis. Department of Communication Disor­ ders, Speech, Language and Hearing, The Sackler Faculty of Medicine, Tel-Aviv University. [In Hebrew.] Stampe, David. 1979 [1972]. A Dissertation on Natural Phonology. New York: Garland. Tobin, Yishai. 1986. Review of Sampson (1980). Lingua 68.99-108. Tobin, Yishai ed. 1988a. The Prague School and its Legacy. Amsterdam & Phila­ delphia: John Benjamins. Tobin, Yishai. 1988b. "Phonetics versus phonology: The Prague School and be­ yond". Tobin 1988a. 49-70. Tobin, Yishai. 1988c. "Two quantitative approaches to phonology: A contrastive analysis". Beiträge zur quantitativen Linguistik ed. by Hermann Bluhme, 71-112. Tübingen: Gunter Narr. Tobin, Yishai. 1990a. Semiotics and Linguistics. London & New York: Longman. Tobin, Yishai. 1990b. "A combinatory phonology of the Hebrew triconsonantal (CCC) root system". La Linguistique 26:1.99-114. Tobin, Yishai. 1990c. "Principles for a contrastive phonotactics: The Hebrew tricon­ sonantal (CCC) root system a case in point". Papers and Studies in Contrastive Linguistics 26:137-153. Tobin, Yishai, ed. 1995. Phonology as Human Behavior: Theoretical implications and cognitive and clinical applications. Dibur u-shmiya (Speech and Hearing Disorders) 18 (Special Issue on Phonology). Tel-Aviv: The Israel Speech and Hearing Association. [In Hebrew.] Tobin, Yishai. 1996. "Will the real Professor de Saussure sign in, please? The three faces of Ferdinand". Semiotica 112:3/4.391-402.

22

YISHAI TOBIN

Tobin, Yishai. 1997a. Phonology as Human Behavior: Theoretical implications and clinical applications. Durham, N.C. & London: Duke University Press. Tobin, Yishai. 1997b. "Developmental and clinical phonology: Roman Jakobson and beyond". Acta Linguistica Hafniensia 29.127-166. Tobin, Yishai. 1999. "Developmental and clinical phonology: The Prague school and beyond". Travaux du Cercle Linguistique de Prague, Nouvelle série/Prague Linguistic Circle Papers 3.53-68. Tobin, Yishai. 2000a. "Phonology as human behavior: Initial consonant clusters across languages". Signal, Meaning and Message: Perspectives on sign-based linguistics ed. by Wallis Reid & Ricardo Otheguy. Amsterdam & Philadelphia: John Benjamins. Tobin, Yishai. 2000b. "Comparing and contrasting Optimality Theory with the Theory of Phonology as Human Behavior". The Linguistic Review 17:2-4.303-323. Tobin, Yishai. 2001a. "Trying to 'make sense' out of phonological reduplication in Hebrew". Proceedings of LP 2000: Item order and its variety and linguistic and phonetic consequences ed. by Bohumil Palek & Osamu Fujimura. Prague: Char­ les University Press. Tobin, Yishai & Haruko Miyakoda. 2001b. "An analysis of Japanese speech errors based on the theory of phonology as human behavior". Proceedings of the Second Malaysian International Conference on Languages, Literatures and Cultures. Kuala Lampur, 18 April 2001. Trubetzkoy, Nikolai S. 1969 [1939]. Principles of Phonology. Transl, by Christine A. M. Baltaxe. Berkeley & Los Angeles: University of California Press. Trubetzkoy, Nikolai S. 1991. The Legacy of Ghengis Khan. Ed., and with a post­ script, by Anatoly Liberman. Ann Arbor: Michigan Slavic Publications.

SEGMENTAL VS SYLLABLE MARKEDNESS DELETION ERRORS IN THE PARAPHASIAS OF FLUENT AND NON-FLUENT APHASICS

DIRK-BART DEN OUDEN University of Groningen 1. Introduction Where phonological theory has been applied to clinical data, the concept of markedness has often played a significant role (Blumstein 1991). Not often taken into account, however, is the fact that the markedness value of linguistic structures may not be the same at all psycholinguistic levels of processing. What is marked at some linguistic level of representation may well be unmarked at another. In combination with currently maintained assumptions about the generation of aphasic (phonological) errors at different psycholin­ guistic levels (e.g., Kohn 1988), this finding allows for interesting and useful comparisons, through which the influence of different types of markedness on literal paraphasias may be related to specific psycholinguistic levels of processing. For this study, we have compared the literal paraphasias of fluent and nonfluent aphasic speakers on a repetition task, aimed at determining the influence of syllable structure on error patterns. The results of both groups have been shown to be equal to a large extent (Den Ouden & Bastiaanse 1999 and in press), but the observed differences in complex coda cluster reductions lead us to conclude that different types of markedness apply at the different affected levels of processing in fluent and non-fluent aphasies. Our analysis is in terms of conflicting and violable constraints, which leads to the application of phonological Optimality Theory (Prince & Smolensky 1993) to our data. An important section of this paper is devoted to possible and, in our view, necessary, adaptations to mainstream Optimality Theory, for it to correspond more closely to a plausible psycholinguistic model of speech processing and deal with aphasie data such as presented here. The first section discusses phonological markedness, the background to this concept, and its application to the study of aphasia in general. The difference between fluent and non-fluent aphasia is explained and argued for in the second section. After this, we present the data obtained from a

24

DIRK-BART DEN OUDEN

repetition task and our analysis of these data, focusing on the difference between fluent and non-fluent patient groups in the reductions of coda clusters. This analysis is formalized in the fourth section, where we introduce Optimality Theory () and discuss how its tools can be used to give plausible representations of our findings. 2. Phonological markedness An important factor underlying the relative frequency of occurrence of lin­ guistic structures in language is the markedness of such structures.1 In different languages, and also during the course of child language acquisition, more marked structures will be less widely distributed than less marked structures sharing the same domain. According to Jakobson (1971), the sound system of a language starts off with the sharpest contrast between sounds, i.e., a wide vowel (usually /a/) and a stop with occlusion at the front of the mouth (usually /p/). In other words, vowels are as extremely vowel-like as possible, and consonants are as extremely consonant-like as possible. After this first acquired contrast, the next is that between oral consonants and nasal consonants. This order of acquisition is argued to be based on ease of articulation. The generally acknowledged least marked sequence of segment types, namely Consonant + Vowel (cv), however, is argued to develop because the phonemes "need to be correctly identified by the listener, and [...] the best graspable clue in discerning consonants is their transition to the following vowels" (Jakobson 1971:25).2 If markedness effects are the result of factors that are inherent to language, it is interesting to see what the role of markedness is in pathologies that are specific to language, viz. different forms of aphasia. Jakobson himself already stated that "[a]phasic regression has proved to be a mirror of the child's acqui­ sition of speech sounds" (Jakobson 1971:40), when it really had not, but com­ parable arguments did spark off many research projects into the relation be­ tween markedness and aphasia. Blumstein's (1973) starting hypothesis was very Jakobsonian, in that she appears to have expected to find, and indeed did find, similar effects of the relative phonological markedness of segments in patients with different types of aphasic syndromes, as, "regardless of the area of brain damage, the more 1

Here we use the term structure to imply all phonological particles of analysis, i.e., features or feature combinations, segments or segment combinations, syllables or syllable frames, feet or combinations of feet, etc. 2 Evidence of CV as the syllable structure that is first acquired by children is provided by a number of authors, such as Smith (1973) and Fikkert (1994).

SEGMENTAL VS SYLLABLE MARKEDNESS IN APHASIA

25

complex phonological structures are impaired and the less complex phonological structures are relatively preserved" (Blumstein 1973:136). Blumstein studied the error patterns in the speech of patients with Broca's aphasia, conduction aphasia and Wernicke's aphasia. In a nutshell, the phonological errors of Broca's aphasics are claimed to result from the inability to translate a correct phonological speech plan into its correct phonetic counterpart and articulation. Conduction aphasics, in a very broad definition, mainly have (postlexical) problems with the sequencing of correctly retrieved sounds and sound patterns from the lexicon. Wernicke's aphasies' phonological problems are caused by deficient retrieval from the lexicon, or even by distorted lexical representations themselves. These three groups all made more errors on segments that had been classified as more marked beforehand. In segment substitutions, marked segments were generally replaced by less marked segments. Besides segment type, studies by Nespoulous et al. (1984, 1987) took into account syllable structure. They found that the error patterns of Broca's apha­ sies were generally constrained by the relative markedness of different segments and syllable structures, whereas such constraints were not found in the errors of conduction aphasies. Such results of studies into markedness and aphasia lead to hypotheses about the nature of markedness. If aphasie deficits exist at different levels of language processing and markedness effects can be related to specific aphasie syndromes and therefore to specific levels of processing, deepens our knowledge about the origin of markedness effects. One problem, of course, lies in the fact that aphasie syndromes are far from 'specific' (e.g., Poeck 1983; Ellis & Young 1988). Also, as noted by Nespoulous et al. (1984, 1987), if phonological markedness at underlying, ab­ stract levels of processing is heavily influenced, or indeed formed, by phonetic markedness, i.e., motor complexity, it may not be much use looking for the differences between markedness effects at both levels of processing, as they cannot be separated for analysis. 3. Fluent vs non-fluent aphasia and levels of processing A broader classification of patients displaying phonological impairment than that described above (into Broca's, conduction, and Wernicke's aphasies) is through a division between fluent and non-fluent aphasia. In the absence of extralinguistic factors, such as dysarthria, non-fluent patients are generally claimed to suffer from difficulty in the timing and coordination of articulatory movements in speech (Blumstein et al. 1980; Blumstein 1991; Hough et al.

26

DIRK-BART DEN OUDEN

1994). This is related to a deficit at a cognitive phonetic level of processing (Code & Ball 1988). This level is quite peripheral to the language processing system, but it may still be considered linguistic and it is not so peripheral that symptoms of a deficit at this level can be ascribed merely to inadequate bucco-facial muscle strength. Articulatory muscles themselves are intact and all movements necessary for speech production can be correctly executed, but the problem lies in the adequate coordination of and voluntary control over the articulators. An even more cognitive approach says that non-fluent patients are impaired in translating the phonological speech plan into a phonetic plan, which should be fully specified for correct articulation (Code & Ball 1988). Such marginally different interpretations are difficult to disentangle (Code 1998; Croot et al. 1998). Non-fluent aphasics are mostly patients that would be classically diagnosed as suffering from Broca's aphasia with apraxia of speech. The label of fluent aphasia seems to cover a wider range of traditional syn­ dromes. It includes lexical as well as postlexical disorders (Kohn 1988), for example the classical syndromes of Wernicke's aphasia and conduction apha­ sia. What these disorders have in common is that they yield incorrect phonological plans. This may be caused by incorrect lexical access or repre­ sentations, or by incorrect phonemic sequencing, the mapping of speech sounds and features onto metrical frames (i.e., phonological encoding). The difference, then, between fluent and non-fluent aphasics is that fluent aphasies create an erroneous phonological plan that may be correctly executed phoneti­ cally, whereas non-fluent aphasies incorrectly execute, or phonetically imple­ ment, a correct phonological speech plan. Den Ouden & Bastiaanse (1999) argued that this division provides the op­ portunity to investigate whether certain structural markedness effects, such as preferred syllable structure, are the result of phonological or of phonetic level constraints. We studied the effects of positional syllable-internal markedness on the deletion patterns of segments in the paraphasias produced by fluent and non-fluent aphasies on a repetition task. On the basis of a syllable template with relatively strong and weak segment positions (van Zonneveld 1988), we predicted that in a word such as sprints /sprints/, the segments printed in bold would be less susceptible to deletion than the others, if the factors responsible for positional syllable-internal markedness were active. The syllable template model itself was based on language typology (frequency of occurrence of structures) and data from child language acquisition (order of acquisition and tendencies within error patterns).

SEGMENTAL VS SYLLABLE MARKEDNESS IN APHASIA

27

The graphs in Figure 1 show the results of the 9 fluent and the 6 nonfluent Dutch aphasic patients on the monosyllabic repetition task that was presented to them. The items in the task were all Dutch monosyllables, with different syllable structures, which, for analysis, were mapped onto the discussed syllable template. In this particular model, the onset and coda satellites (osat and csat) can only be filled with glides, liquids or nasals (i.e., sonorant consonants), the pre-margin (pre) can only be filled with the segment /s/ and the appendix (app) with coronal voiceless obstruents (/s/, /t/) or, in very rare (highly marked) cases, /k/ or /p/. The pre-margin and appendix positions can be considered extrasyllabic: they violat binary branching and their 'behavior' is exceptional in other ways as well (cf. Harris 1994). Positions dependent on other positions are only filled if the position they are dependent on is filled. For example, the pre-margin and the onset satellite are dependent on the onset core (ons). In this model, everything depends on the peak. The position for this peak (the vowel) is left out in these graphs, as the comparison was only between consonant positions. Vowels were hardly ever deleted in the monosyllabic repetition items. The graphs display the number of deletions within a certain segment posi­ tion, relative to the actual occurrence of that position in the items presented for repetition. Using the scores of individual patients, Wilcoxon tests were em­ ployed to calculate the significance of the difference in the mean proportions of deletions. The illustrative word used here is sprints, which did not actually occur in the item list itself, because it is morphologically complex in Dutch.

s

pre

p

r

ons

osat

n

t

csat

s

cod

s

app

pre

ons

p

osat

r

n

csat

Fig. 1 : Deletion errors in different syllable positions (Den Ouden & Bastiaanse

t

cod

s

app

1999)

Although the tendencies that are visible in these graphs show that the core onset and coda positions are least prone to deletion, significance was only reached for the difference between deletions in onset and onset satellite posi­ tions and between deletions in coda and appendix positions for both the fluent and the non-fluent patients. The fluent patients also deleted significantly more segments in appendix positions than in coda satellite positions. The overall

28

DIRK-BART DEN OUDEN

conclusion of this study was that the literal paraphasias of fluent and nonfluent aphasics show the influence of positional markedness on deletions. Nevertheless, what remains is the fact that the proportions of deletions in coda and coda satellite positions differ less for fluent aphasics than for nonfluent aphasies, whereas there is no particular difference in onsets.3 This coda observation, as we choose to name it, may have been obscured by the particu­ lar template that we used and by the rules for associating particular segments to particular syllable slots. The data analysed in Den Ouden & Bastiaanse (1999) containCVCsyllables, in which a sonorant coda would be associated to the coda core position, as well as CCCVC and CVCCC syllables, which may blur the view on what happens to (pure) CC-onsets and CC-codas. For this reason, we decided to take a closer look at the relevant data, taking into account only those items with CC-onsets and/or CC-codas, in order to abstract away from influences by other structural positions as much as possible. 4. The Coda observation In this section, the relevant data and the methods of collecting them are discussed. We analysed the responses to a subset of target items used in Den Ouden & Bastiaanse (1999). For this renewed visit to the repetition study, four more non-fluent patients were tested. The other patients and their data were taken from the previous, 1999 study. 4.1 The experimental investigation Subjects were 10 non-fluent aphasies, 4 male and 6 female, with a mean age of 61 (range 50-79), and 9 fluent aphasies, 7 male and 2 female, with a mean age of 58 (range 38-84). The non-fluent patients had been diagnosed by their speech therapists as suffering from apraxia of speech, without dysarthria. This diagnosis was confirmed by the examiner. The fluent aphasies did not suffer from apraxia of speech. All patients produced literal paraphasias on language tests and in sponta­ neous speech. All were native speakers of Dutch and more than 3 months aphasie due to a single left-hemispheric stroke. The repetition task consisted of 114 Dutch monosyllabic words, of which 41 were analysed for the present study. These were the items with complex 3

The relatively high proportion of deletions in pre-margin position by non-fluent aphasies is explained by the fact that apraxics generally have problems with initiating movement (Code 1998). Non-fluent patients with apraxia of speech will have difficulty with the beginning of words, independently of syllable position (cf. Dogil & Mayer 1998). All syllable onsets in the monosyllabic items in the test obviously coincided with word onsets.

SEGMENTAL VS SYLLABLE MARKEDNESS IN APHASIA

29

onsets or complex codas, which did not violate the sonority slope, meaning that the sonority value of segments rose from the margins to the peak (Clements 1990). This restriction bars /st/-onsets, for example.4 The restric­ tions on usable items left us with 21 items with complex onsets and 20 items with complex codas. Deletions were scored per segment position within the analysed onsets and codas. In onsets the opposition was between the onset core and the onset satellite, or, in this case, the first and the second position, respectively. In co­ das, the opposition was between the coda satellite (the first position) and the coda core (the second position). As an example, the word print is (orthographically) given below, with the appropriate position labels: p r i onset core onset sat.

n

t coda sat.

coda core

We wanted to know whether the two positions within a complex onset or coda were equally affected by syllable simplification errors. For this reason, we only took into account deletions of these positions, as they are the only true quantitative simplifications of phonological structure, as opposed to segment substitutions. Note that not all patients produced 41 valid responses to the 41 items under scrutiny, as target items yielding neologisms or no-responses were not included in the analysis. All in all, the literal paraphasias of fluent aphasic patients and non-fluent aphasic patients were analysed for the proportions of deletions per syllable position, in onsets and codas. Results are summarized in Table 1 and in Figure 2. Table 1 shows the ab­ solute number of deletions in the different positions. N is the total number of occurrences of relevant onset or coda clusters for this group of patients. The pvalues printed in bold, based on χ2-tests, show significant differences (α= .05). The graphs in Figure 2 show the mean number of deletions, proportionate to the number of occurrences of the relevant position in the target list, for each group of patients. 4

In the more detailed sonority hierarchies that have been proposed, voiceless fricatives, such as /s/, are considered to be more sonorant than stops, such as lil. Detailed sonority hierarcy (Jespersen 1904, cited in Clements 1990:285): low vowels > mid vowels > high vowels > r-sounds > laterals = nasals > voiced fricatives > voiced stops > voiceless fricatives > = voiceless stops In less detailed hierarchies, with a stronger claim on universality, fricatives and stops are joined in the category of obstruents, with equal sonority values, the adjacency of which also does not make for a well-formed sonority slope.

30

DIRK-BART DEN OUDEN

Onsets

deletions in onset core

deletions in onset sat.

N

x2

Fluent

4

16

186

X =6.394 p = 0.0058

Non-fluent

16

30

206

X=4.064

deletions in coda sat.

deletions in coda core

Fluent

13

11

178

X = 0.045 p = 0.6725

Non-fluent

30

13

198

X = 6.679 p =0.0097

Codas

p = 0.0285

Table 1 : Deletions in onset and coda CC-clusters

■ C[-son] C[+son]

fluent nonfluent Fig. 2a: Deletions in onset clusters for fluent and non-fluent aphasic patients

C[+son] ■ C[-son]

fluent nonfluent Fig. 2b: Deletions in coda clusters for fluent and nonfluent aphasic patients

J

SEGMENTAL VS SYLLABLE MARKEDNESS IN APHASIA

31

The graphs visualize the effects shown in Table 1, namely that the patterns of deletions are equal for both groups in onsets, but not,in codas, where only the non-fluent patients delete the sonorant coda position (the coda satellite) significantly more often than the non-sonorant coda position (the coda core). For example, the non-fluent patients' rendition of the target word print (/print/) will characteristically be [pIt], while the fluent patients will turn it into either [pIn] or [pIt], in a seemingly random fashion. The 'coda observation' of Den Ouden & Bastiaanse (1999) thus holds true upon closer inspection. 4.2 Discussion It is clearly not the case that only non-fluent aphasics show effects of markedness relations, while fluent aphasics reveal a random distribution of errors, as the effects are the same for both groups within onset clusters. For our account of these data, we return to the notion of markedness. Jakobson's markedness hierarchy was very much based on a contrast be­ tween segments and segment categories. In order to achieve the biggest con­ trast between vowels and consonants, consonants should be as consonantal as possible, and vowels should be as vowel-like as possible. According to this hierarchy of segmental markedness, consonants are less marked if they are less sonorant. Segmental markedness applies non-contextually; it does not take into account the position of a segment within a syllable. This markedness hierarchy can account for the error pattern of non-fluent patients, but not for the coda observation in fluent patients' errors. If we do look at segments in the context of prosodie structure, a different picture emerges. Clements (1990) argued that the preferred sonority slope of syllables has a steep rise in sonority pre-vocalically and a slow decline in sonority postvocalically. Syllables with sonorant codas are more frequent than syllables with non-sonorant codas. This Sonority Cycle (Clements 1990) allows us to formulate a second markedness hierarchy, which we will call syllable markedness, according to which onsets want to be non-sonorant and codas want to be sonorant. The Sonority Cycle was indeed the object of investigation of Christman (1992), who showed that neologisms (produced by fluent patients) conform to this principle of a steep rise and a minimal decline of sonority in syllables, as formulated by Clements (1990). In a case study, Romani & Calabrese (1998) also investigated the influence of syllable complexity and segmental markedness on literal paraphasias. The Italian mother tongue of their patient, however, has so many restrictions on possible codas, that it does not allow for complex coda analysis in the way that Dutch or English might.

32

DIRK-BART DEN OUDEN

Syllable markedness alone cannot account for the pattern of deletions ob­ served in fluent patients' paraphasias. Note, however, that segmental and syl­ lable markedness reinforce each other in onsets, whereas they are in opposition in codas. It is this combination of the two types of markedness, or rather the crucial conflict between them, that may account for the error pattern of fluent aphasics. Relating the two types of markedness to the results obtained in this study, our claim is that for non-fluent aphasic patients, who have a deficit at a (cognitive) phonetic level of speech production, segmental markedness is dominant. Irrespective of syllable position, these patients 'prefer' to delete sonorant consonants and to end up with non-sonorant consonants. Fluent aphasie patients, who have a deficit at or before the level of phonological encoding, show the influence of both segmental markedness and syllable markedness. For codas, this means that there is a conflict which results in a draw, hence the almost equal distribution of deletions of sonorant and nonsonorant consonants in this syllable constituent. The full analysis of the presented data is therefore as follows: non-fluent aphasies have a deficit at a phonetic level of processing. At this level, where articulatory planning takes place, the markedness of individual segments, or feature combinations, is still an influential factor. The impairment allows this type of markedness to become dominant and this means that when clusters of consonants are reduced, the non-sonorant, segmentally least marked consonant will come out as the winner, irrespective of its position within a syllable. Before this phonetic level of processing, constraints on sonority sequencing, i.e., on preferred syllable structure, are active beside constraints on segmental markedness. At the affected level(s) of processing in fluent aphasies, the conflict between segmental markedness constraints and syllable markedness constraints emerges, as structure-preserving constraints lose control over the output of the speech production process. This yields a pattern of errors in which onsets are relatively systematically reduced to non-sonorant segments, as both types of markedness reinforce each other in onsets, while codas are reduced on a seemingly random basis to either sonorant or nonsonorant segments, as the constraints are in direct conflict over what is a preferred coda. The notion of a conflict between different phonological constraints on well-formedness makes it quite appealing to investigate the representation of the interaction of these constraints within the framework of , as this framework typically allows for constraints with opposing ends and as it has become the most influential phonological framework of the previous and of the current

SEGMENTAL VS SYLLABLE MARKEDNESS IN APHASIA

33

decade. The remainder of this paper will be dedicated to the discussion of how the foregoing analysis may or should be represented in , the starting point being to stay as much as possible within the representational and conceptual limits set out by the still developing formal theory itself. 5. The coda observation in OT Since its introduction in the early 1990s,  has quickly gained ground in phonology, as it is presently doing in the domains of syntax and semantics. Its main appeal lies in two characteristics: the focus on well-formedness of the output, as opposed to a focus on rules that seem to exist for their own sake, and the softness of constraints, where a constraint can be violated in order to satisfy something more important. Constraints on the well-formedness of the output, so-called markedness constraints, compete with each other and with structure-preserving faithful­ ness constraints. A grammar is formed by the language-specific ranking of these violable, universal constraints. An unrestricted number of possible can­ didates for the eventual output form of the utterance are generated by a component named GEN (Prince & Smolensky 1993). These output candidates are compared, on the basis of the input form and the ranked constraints. The output candidate that has the least important constraint violations wins and, consequently, is the optimal output. This optimal output may violate con­ straints, in order to satisfy higher ranked, i.e., more important, constraints. There are two types of faithfulness constraints. Parse (or Max) constraints say that material in the input should also be present in the output. Such con­ straints block deletion. Fill (or Dep) constraints say that material in the output should also be present in the input, thus blocking insertion. Another conspicuous characteristic of classic  is that all constraints on the output should compete with each other at all times and that their hierarchi­ cal ranking is stable for adult speakers. This basically means that the 'con­ struction' of the output occurs in one step,  is therefore minimally deriva­ tional, the only derivation being that from the input to the output. If we consider, then, the focus on derivations and cyclicity of processes in previous decades (e.g., Kiparsky 1979), it is only logical that the main criticism of  has been based on evidence that some phonological processes simply cannot be adequately described without making reference to some notion of cyclicity or multiple levels of processing. Such criticism has led to a number of adaptations to the original theory, all aimed at giving satisfactory descriptions of morphophonological processes in which the output form seems to be opaque, and certain constraints appear to

34

DIRK-BART DEN OUDEN

have been applied only to specific substrings of the eventual output form (a phonological word, mostly) (cf. the contributions to Hermans and van Oostendorp 1999). Examples of such tools that aim to maintain the one-step evaluation are Output-Output Correspondence (McCarthy & Prince 1995), in which the optimal output form wants to be as similar as possible to other out­ put forms it is related to, and Sympathy Theory (McCarthy 1998), in which the optimal output form wants to resemble a fairly arbitrarily chosen other output candidate. Other optimality theorists have chosen to abandon the onestep derivation and to incorporate some type of rule ordering in , allowing multiple levels of evaluation, with constraints that apply only to specific levels of processing, or stages in the derivation (Booij 1997; Rubach 2000). Crucially, the 'founding fathers' of , Prince & Smolensky (1993:79), did not put an absolute restriction on the theory as having only one level of evaluation, although the current practice in 'standard'  is such that multiple levels of evaluation are considered a weakness. Beside the discussion over single or multiple levels of evaluation in the formal theory, stands a large body of evidence for multiple levels of processing, from the fields of psycholinguistics and neurolinguistics, in which the issue is not under debate (see Levelt 1989, among many others). Lesion studies going back to the 19th century have shown that different parts of the brain perform different functions (for an overview, see Whitaker 1998). The only questions left to the field, with respect to this subject, are on the defini­ tion of 'function', whether there are direct, one-on-one relations between brain areas and specific functions and, if so, what areas perform which functions (Poeppel 1996a,b; Démonet et al. 1996). The domain of phonological processing itself has also been dissected (cf. Kohn 1988) and results of studies into temporally successive brain activity point towards a 'phonological loop' (Baddeley 1986) in which abstract and articulatory levels are distinct, though possibly mutually influential,  should aim at ways to incorporate these multiple levels of (phonological) processing, instead of focusing on retaining the one-step hypothesis.  has not yet been applied in a systematic way to the study of language breakdown. However, for our purposes, it is worthwhile to consider the progress it is making in the field of child language acquisition, and consider the theoretical and representational questions raised in that domain of investi­ gation (Sternberger & Bernhardt 1997; Barlow & Gierut 1999; Hayes 1999; Tesar & Smolensky 2000). The relation between aphasia and language acqui­ sition has been sketched above and elsewhere (cf. Den Ouden & Bastiaanse, in press) and it is only natural to take it into account here.

SEGMENTAL VS SYLLABLE MARKEDNESS IN APHASIA

35

5.1  and child language acquisition Whether the constraints of language are considered to be innate and uni­ versal (Prince & Smolensky 1993) or not (Boersma 1998; Hayes 1999), or whether or not there are different rankings for production and perception, the general analysis of child language acquisition in  terms is that children start off with a high ranking of markedness constraints and a low ranking of faith­ fulness constraints. This simply means that they will not be able to produce (or parse) the adult output forms of their mother tongue and that they will mainly show effects of markedness during the first stages of acquisition. Gradually, then, the markedness constraints will lower, relative to the faithfulness constraints, as the child tunes its grammar to that of its mother tongue (Sternberger & Bernhardt 1997). Communication becomes more accu­ rate as the faithfulness constraints become more prominent. It is not the case that children will reset their constraint rankings on the basis of one single piece of positive evidence. Language acquisition, though fast, does not occur in one day, and children hear a lot of adult output forms on the basis of which they eventually set their grammar and optimize their lexicon. Also, during this process, there is much variation in their own output. Boersma (1998) therefore argues that this should be represented by variable distances between different constraints, which may at times even overlap. Thus, every piece of positive evidence (an adult output form) will cause a constraint to change position within the ranking, but this does not necessarily mean that it will immediately outrank another constraint; it may simply move closer. Constraints have a 'moving space' within the hierarchy and as soon as constraint A comes within the scope of the space of constraint B, this means that both ranking A >  and  > A may occur, which accounts for variation in the output, if constraints A and  are in conflict. It is a common strategy in  to capture optionality, also in adult grammars, by representing adjacent constraints as ranked freely with respect to each other (Clements 1997; Demuth 1997).  applications to child language acquisition thus provide us with the tools to represent the domination of the unmarked and to represent variation or optionality. 5.2  and aphasia: constraints at different levels of processing As noted in the above sections, aphasia is generally characterized by a prominence of unmarked structures. Compared to 'normal' speakers, the aphasic speaker is less faithful to the input, the input here being the lexicon or, for example, utterances to be repeated, whether real words or non-words. The most straightforward way of representing this in  is by a lowering of faith­ fulness constraints, relative to markedness constraints. Note that it is theoreti­ cally also possible that the input itself is disturbed, so that the correct (i.e.,

36

DIRK-BART DEN OUDEN

adult) constraint ranking works on an incorrect input, or that the number and/or type of output candidates that are generated is in some way restricted. These options, however, do not directly account for the prominence of the un­ marked, as observed in aphasic speech. Any systematic way of constricting the input or the output candidates would somehow have to be by incorporating extra markedness constrictions on these domains. This would come down to an extratheoretical add-on for which there is no evidence or argument in nonpathological natural language. We start from the hypothesis that language im­ pairment is focal breakdown of the normal language system, crucially within its own terms. The impairment is assumed not to add new features to the nor­ mal system (cf. Caramazza 1991). The aphasic patients in this study show markedness effects in their impaired output, and our  representation of this fact consists in the lowering of faithfulness constraints, which allows markedness constraints to have greater influence on the choice of the optimal output candidate. Aphasie data are never homogeneous. There is much noise and variabity, which is precisely why statistics are used to determine whether some structures are significantly used more often than others. Variation, as we have seen above, can be represented by 'switching' of adjacent constraints. In an algorithm such as Boersma's (1998), with unequal distances between dif­ ferent constraints in the hierarchy, there may be 100% overlap of constraint space, yielding a 50-50 distribution of two forms in a variable pattern, but it is also possible to conceive of a partial overlap, yielding a different distribution, with one of the two forms occurring more often than the other. In this section, we present the  tableau that represent our analysis of the data discussed here, viz. the coda observation. In reduction of consonant clusters, non-fluent patients will render the target word print (/print/) as [pIt], while fluent patients will turn it into either [pIn] or [pIt], in a seemingly ran­ dom fashion. After the presentation and the explanation of the tableaux, we will go deeper into some considerations underlying our particular representa­ tion of the data. The constraints needed for this analysis are in (1): (1)

Markedness *C[+SON] H O N S (Onset Harmony) HCOD (Coda Harmony) *COMPLEX

Do not allow sonorant consonants Onsets do not want to be sonorant Codas want to be sonorant Do not allow complex onsets or codas

Faithfulness PARSE

Preserve input material

SEGMENTAL VS SYLLABLE MARKEDNESS IN APHASIA

37

GEN generates an infinite number of possible output candidates, most of which are rendered irrelevant because they are too deviant from the input form and therefore incur too many violations of faithfulness constraints. The candidates that we will consider in Table 2 are the most relevant to our data and to our example input form print, and we should be able to distinguish between them and choose the correct optimal output form with the constraints given above. Other possible candidates are dealt with by other constraints, but this is outside the scope of this paper. The data discussed all concern cluster reduction. This is represented by the ranking of *COMPLEX over PARSE. This ranking applies to both groups of patients, fluent and non-fluent. Ta­ ble 2 shows how these two constraints and their ranking distinguish between the candidate outputs. The top left cell shows the input. The output candidates are given below that, in the first column. Constraint names are given in the top row, ordered by prominence in the hierarchy from left to right. A candidate's violations of a constraint are marked with '*' in the relevant cell. A crucial violation of a constraint, meaning that the candidate in question is no longer relevant for lower-ranked constraints, as others will always be more optimal, is marked with an exclamation mark !, to the right of which the irrelevant cells are grey. The winning, optimal candidate is marked with a pointed finger

/print/ print pInt

*COMPLEX

PARSE

** !

*!

*'

pt

**

pm

**

nt

**

rm

**

rmt

*!

*

prit

*!

*

prin

*!

* Table 2: Ranking of *COMPLEX over PARSE

The ranking in Table 2 yields four candidates, boldly printed in the tableau, between which the other relevant constraints should differentiate. Non-fluent patients will delete the non-sonorant consonants, regardless of their position within the syllable and we have argued that the constraint

38

DIRK-BART DEN OUDEN

responsible for this is one on segmental markedness, namely *C[+S0N]. The  representation of the impairment of non-fluent patients, on the basis of our data, is therefore fairly straightforward (Table 3): /print/

*COMPLEX

print

** !

pint

* !

PARSE

*C[+SON]

** *

*

**

pit pin

**

* !

[rIt

**

nn

**

* ! ** !

rint

* !

*

**

prIt

* !

*

*

prIn

* !

*

**

Table 3: Non-fluent patients: /print/ → [pit]

The fluent patients may turn example word print into either [pIn] or [pIt]. We have argued that this is because of a competition between a constraint on the preferred sonority value of the syllable constituent coda and a segmental markedness constraint that disallows sonorant consonants. With respect to these two constraints, HCOD and *C[+SON], two rankings are possible, yielding different results, as shown in Table 4a and Table 4b. /print/

*COMPLEX

print

** I

pInt

*!

PARSE

*C[+SON]

HCOD

*

*

**

pit

HONS

** *

pIn

**

*!

rIt

**

*!

rIn

**

Irmt

*!

*

*,*

prit

*!

*

*

prin

*!

*

* *

*

•'*

'f

** i

Table 4a: Fluent patients: /print/ → [prt] ; [pIn]

* /

*

!

SEGMENTAL VS SYLLABLE MARKEDNESS IN APHASIA

39

Note that the specific ranking of H O N S is irrelevant here, as long as it is ranked below *COMPLEX. H O N S , as we have argued, merely 'strengthens' the effect of *C[+SON] on onsets. Following  conventions, this non-crucial ranking with respect to other constraints is marked with a dotted line. /print/ print

* COMPLEX

HCOD

*C[+SON]

HONS

**

* !

pInt

PARSE

** i

pIt

* **

* *!

**

pIn

*

rIt

**

rln

**

** | **

rInt

* !

*

prIt

*!

*

prIn

* !

*

*!

*

*

■ . .

*

"

-

'

.

"

■

.

■

■

.

.

■

'

*

* **

Table 4b: Fluent patients: /print/ → [pIt] ; [pIn]

We have now given an  representation of our interpretation and analysis of the data discussed in this paper, viz. the coda observation. For fluent apha­ sics, the tableaux show competition between and switching of the constraint on segmental markedness and the constraint on preferred sonority value of co­ das. The tableau for non-fluent aphasics shows dominance of segmental markedness and either the absence, or the non-competitively low ranking of the constraint(s) representing the Sonority Cycle. Thus, the data of two groups of patients, with deficits at different levels of speech processing, are repre­ sented with two different tableaux. It will not come as a surprise to the reader that we argue that there is a relation between the different levels of processing and the different tableaux. This relation will be specified further below. 5.3 Considerations In the analysis above, aphasia was represented as a lowering of faithfulness constraints and an increased instability of (markedness) constraints, causing switching of adjacent constraints. A major argument against an analysis in which different types of aphasia are represented only through structural reranking of markedness constraints is the fact that aphasie speech errors hardly ever violate the phonotactics of the mother tongue of the speaker, or, indeed, universal restrictions on well-formedness (see Buckin­ gham 1992). This would be unexplained if markedness constraints changed

40

DIRK-BART DEN OUDEN

position in the hierarchy on a large scale. The adherence to (mother tongue) phonotactics points towards a lowering of faithfulness constraints only. However, the variation found in the patterns of paraphasias belonging to different types of aphasia, such as observed in this study and contra Blumstein's (1973) hypothesis, acts as an argument against the mere lowering of faithfulness constraints in the representation of aphasia. To represent different aphasic symptoms only through different degrees of faithfulness lowering comes down to saying that aphasic 'syndromes', or rather, clusters of symptoms, only differ with respect to the degree of seriousness of impairment. For these reasons, rather than claiming that the constraints H O N S and HCOD are ranked non-competitively low at the level of impairment of nonfluent aphasies, we argue that they are non-existent at this level. In this way, we minimize the structural reranking of markedness constraints, while still being able to represent impairments at different processing levels, with dif­ ferent characteristics. This means, then, that our analysis allows for different levels of evaluation of constraints, where not all constraints are active (i.e., exist) at all levels. From a psycholinguistic and neurolinguistic perspective, this is the only natural way to conceive of linguistic processing, and we also point to the more formally optimality theoretic attempts we have cited to allow for multiple-step derivation by incorporating multiple levels of evaluation in , with the approach taken here is by no means incompatible. In psycholinguistic modeling, it is common practice to minimize the number of levels, modules or stages of processing to those necessary for an accurate representation of empirical findings. A similar principle, Level Minimalism, is formulated by Rubach (2000:313), for his modification of , Derivational Optimality Theory, which allows multiple levels of evaluation. Another principle he formulates to restrict the power of his framework is that of Reranking Minimalism: "[the] number of rerankings is minimal [...] reranking of constraints comes at a cost and needs to be argued for" (Rubach 2000:313). This principle is in line with our approach of unstable rankings to account for variation, but with level-specific constraints instead of structural reranking of markedness constraints to account for the influence of different factors at dif­ ferent levels of speech production processing. The constraint * COMPLEX, which in our analysis is crucially present at the levels of impairment of both fluent and non-fluent patient groups, is formu­ lated here in terms of syllable markedness. It therefore prevents us from claiming that it is syllable markedness that is active besides segmental markedness at the pre-phonetic level of processing and that at the phonetic

SEGMENTAL VS SYLLABLE MARKEDNESS IN APHASIA

41

level of processing, only segmental markedness constraints are active. This is because we have chosen to follow the intuitive notion that the underlying cause of cluster reduction in the first place is the same for fluent and nonfluent aphasics, namely that syllable constituents prefer to be simple (cf. Gilbers & Den Ouden 1994). Although the assumption that all constraints on syllable markedness act as one group and are active at exactly the same levels is not necessarily true, our analysis would gain by a more straightforward division between constraints that work on different domains, at different levels of evaluation/processing. This 'problem' would be solved if *COMPLEX were to be conceived of as a constraint on, for example, adjacent segments, without making reference to syllable constituents. If, however, the formulation of the constraint is indeed similar to 'consonants do not want to be adjacent to consonants', the prediction is that there should be no difference between proportions of reductions within syllable constituents and across syllable constituents. This hypothesis needs to be tested. 6.

Conclusion On the basis of fluent aphasics' and non-fluent aphasies' responses to a monosyllabic real word repetition test, we have argued that there is a differ­ ence between the pre-phonetic and the phonetic level of processing. In our  approach, the constraints responsible for the Sonority Cycle, as formulated by Clements (1990), are active only at pre-phonetic levels of evaluation, whereas a constraint on segmental markedness, saying that consonants should be as consonantal as possible (and therefore non-sonorant), is active at the prephonetic, as well as at the phonetic level. Aphasia, in an  approach, comprises the lowering of faithfulness con­ straints at the affected level of processing. This accounts for the different types of aphasia that are distinguished clinically. Also, aphasia is characterized by unstable ranking of close (adjacent) markedness constraints, which accounts for the high degree of variation found in aphasie error patterns. Along the lines of Boersma (1998), this could be conceived of as a widening of the moving space of constraints from their relatively fixed place in the hierarchy, which increases the opportunity for overlap of constraints. In our study, this overlap and thus switching of constraints was particularly visible in the paraphasias of fluent aphasies, at whose affected level of processing (pre-phonetic) a competition between segmental markedness (*C[+SON]) and syllable markedness ( H C O D ) leads to variation in the output. Where the process of cluster reduction applies, fluent patients will render the example item print /print/ as either [pIt] or [pIn], with an equal distribution of

42

DIRK-BART DEN OUDEN

deletion of the sonorant or the non-sonorant segment. It is our position that language breakdown in the form of aphasia provides a window on the workings of the language system. Linguistic theories should be able to deal with the view thus offered. REFERENCES Baddeley, Alan D. 1986. Working Memory. Oxford: Clarendon Press. Barlow, Jessica A. & Judith A. Gierut. 1999. "Optimality theory in phonological acquisition". Journal of Speech, Language and Hearing Research 42:6.14821498. Blumstein, Sheila E. 1973. "Some phonological implications of aphasic speech". Psycholinguistics and Aphasia ed. by Harold Goodglass & Sheila Blumstein, 123-137. Baltimore & London: Johns Hopkins University Press. Blumstein, Sheila E. 1991. "Phonological aspects of aphasia". Acquired Aphasia ed. by Martha T. Sarno, 157-186. San Diego: Academic Press. Blumstein, Sheila E., William E. Cooper, Harold Goodglass, Sheila Statlender & Jonathan Gottlieb. 1980. "Production deficits in aphasia: A voice-onset time analysis". Brain and Language 9.153-70. Boersma, Paul. 1998. Functional Phonology: Formalizing the interactions between articulatory and perceptual drives. (= LOT International Series, 11.) Ph.D. dissertation, University of Amsterdam. The Hague: Holland Academic Graphics. Booij, Geert E. 1997. "Non-derivational phonology meets Lexical Phonology". Deri­ vations and Constraints in Phonology ed. by Iggy Roca, 261-268. Oxford: Oxford University Press. Buckingham, Hugh W. 1992. "Phonological production deficits in conduction apha­ sia". Conduction Aphasia ed. by Susan E. Kohn, 77-116. Hillsdale, N.J.: Lawrence Erlbaum Associates. Caramazza, Alfonso. 1991. "Data, statistics and theory: A comment on Bates, McDonald, MacWhinney, and Applebaum's 'A maximum likelihood procedure for the analysis of group and individual data in aphasia research'". Brain and Language 41.43-51. Christman, Sarah S. 1992. "Uncovering phonological regularity in neologisms: Con­ tributions of sonority theory". Clinical Linguistics & Phonetics 6:3.219-247. Clements, George N. 1990. "The role of sonority in core syllabification". Papers in Laboratory Phonology I: Between the grammar and physics of speech ed. by John Kingston & Mary E. Beckman, 283-333. Cambridge: Cambridge University Press. Clements, George N. 1997. "Berber syllabification: derivations or constraints?". Derivations and Constraints in Phonology ed. by Iggy Roca, 289-330. New York: Oxford University Press. Code, Chris. 1998. "Models, theories and heuristics in apraxia of speech". Clinical Linguistics and Phonetics 12:1.47-65.

SEGMENTAL VS SYLLABLE MARKEDNESS IN APHASIA

43

Code, Chris & Martin J. Ball. 1988. "Apraxia of speech: The case for a cognitive phonetics". Theoretical Linguistics and Disordered Language ed. by Martin J. Ball, 152-167. London & Sydney: Croom Helm. Croot, Karen, Karalyn Patterson & John R. Hodges. 1998. "Single word production in non-fluent progressive aphasia". Brain and Language 61.226-273. Démonet, Jean-François, Julie A. Fiez, Eraldo Paulesu, Steve E. Petersen & Robert J. Zatorre. 1996. "PET studies of phonological processing: A critical reply to Poeppel". Brain and Language 55.352-379. Demuth, Katherine. 1997. "Multiple optimal outputs in acquisition". University of Maryland Working Papers in Linguistics 5.53-71. Dogil, Grzegorz & Jörg Mayer. 1998 "Selective phonological impairment: A case of apraxia of speech". Phonology 15:2.143-188 Ellis, Andrew W. & Andrew W. Young. 1988. Human Cognitive Neuropsychology. Hillsdale, N.J.: Lawrence Erlbaum Associates. Fikkert, Paula. 1994. On the Acquisition of Prosodic Structure. Ph.D. dissertation, University of Leiden. Gilbers, Dicky G. & Dirk-Bart den Ouden. 1994. "Compensatory lengthening and cluster reduction in first language acquisition: A comparison of different analy­ ses". Language and Cognition 4: Yearbook 1994 of the research group for Theo­ retical and Experimental Linguistics of the University of Groningen ed. by Ale de Boer, Helen de Hoop & Henriëtte de Swart, 69-82. Groningen: University of Groningen. Harris, John. 1994. English Sound Structure. Oxford: Blackwell. Hayes, Bruce. 1999. "Phonetically Driven Phonology: The role of Optimality Theory and Inductive Grounding". Functionalism and Formalism in Linguistics, vol. I: General Papers ed. by Michael Darnell, Edith Moravcsik, Frederick J. Newmeyer, Michael Noonan & Kathleen Wheatley, 243-285. Amsterdam & Philadelphia: John Benjamins. Hermans, Ben & Marc van Oostendorp, eds. 1999. The Derivational Residue in Phonological Optimality Theory. Amsterdam & Philadelphia: John Benjamins. Hough, Monica Strauss, Salvatore Demarco & Donna Farler. 1994. "Phonemic re­ trieval in conduction aphasia and Broca's aphasia with apraxia of speech: Un­ derlying processes". Journal of Neurolinguistics 8: 4.235-246. Jakobson, Roman. 1971 [1941]. Studies on Child Language and Aphasia. The Hague: Mouton. Jespersen, Otto. 1904. Lehrbuch der Phonetik. Leipzig & Berlin: . G. Teubner. Kiparsky, Paul. 1979. "Metrical Structure Assignment Is Cyclic". Linguistic Inquiry 10:3.421-441. Kohn, Susan E. 1988. "Phonological production deficits in aphasia". Phonological Processes and Brain Mechanisms ed. by Harry A. Whitaker, 93-117. New York: Springer. Levelt, Willem J. M. 1989. Speaking: From intention to articulation. Cambridge, Mass.: MIT Press.

44

DIRK-BART DEN OUDEN

McCarthy, John. 1998. "Sympathy and Phonological Opacity". Ms., University of Massachusetts, Amherst. McCarthy, John & Alan Prince. 1995. "Faithfulness and reduplicative identity". Pa­ pers in Optimality Theory. (= University of Massachusetts Occasional Papers in Linguistics, 18), ed. by Jill Beckman, Laura Walsh Dickey & Suzanne Urbanczyk, 249-384. Amherst, Mass.: Graduate Linguistics Student Association. Nespoulous, Jean-Luc, Yves Joanette, Renée Béland, David Caplan & André Roch Lecours. 1984. "Phonological disturbances in aphasia: Is there a 'markedness ef­ fect' in aphasic phonemic errors?". Advances in Aphasiology, vol. 42: Progress in Aphasiology ed. by F. Clifford Rose, 203-214. London: Raven Press. Nespoulous, Jean-Luc, Yves Joanette, Bernadette Ska, David Caplan & André Roch Lecours. 1987. "Production deficits in Broca's and conduction aphasia: Repeti­ tion vs. reading". Motor and Sensory Processes in Language ed. by Eric Keller & Myrna Gopnik, 53-81. Hillsdale, N.J.: Lawrence Erlbaum Associates. Ouden, Dirk-Bart den & Roelien Bastiaanse. 1999. "Syllable structure at different levels in the speech production process: Evidence from aphasia". Paper presented at HILP 4, University of Leiden, January 1999. Ouden, Dirk-Bart den & Roelien Bastiaanse. In press. "Syllable structure at different levels in the speech production process: Evidence from aphasia". Proceedings of HILP 4 ed. by Jeroen van der Weijer, Harry van der Hulst & Vincent van Heuven. Amsterdam & Philadelphia: John Benjamins. Poeck, Klaus. 1983. "What do we mean by 'aphasie syndromes'? A neurologist's view". Brain and Language 20.79-89. Poeppel, David. 1996a. "A critial review of PET studies of phonological processing". Brain and Language 55.317-351. Poeppel, David. 1996b. "Some remaining questions about studying phonological processing with PET: Response to Demonet, Fiez, Paulesu, Petersen & Zatorre (1996)". Brain and Language 55.380-385. Prince, Alan & Paul Smolensky. 1993. Optimality Theory: Constraint interaction in generative grammar. New Brunswick, N.J.: Rutgers University Center for Cognitive Science Technical Report 2. Romani, Cristina & Andrea Calabrese. 1998. "Syllabic constraints in the phonological errors of an aphasie patient". Brain and Language 64.83-121. Rubach, Jerzy. 2000. "Glide and glottal stop insertion in Slavic languages: A DOT analysis". Linguistic Inquiry 31:221\-\1. Smith, Neilson V. 1973. The Acquisition of Phonology: A case study. Cambridge: Cambridge University Press. Sternberger, Joseph P. & Barbara H. Bernhardt. 1997. "Optimality Theory". The New Phonologies: Developments in clinical linguistics ed. by Martin J. Ball & Raymond D. Kent, 211-245. San Diego & London: Singular Publishing Group. Tesar, Bruce & Paul Smolensky. 2000. Learnability in Optimality Theory. Cambridge, Mass.: MIT Press.

SEGMENTAL VS SYLLABLE MARKEDNESS IN APHASIA

45

Whitaker, Harry A. 1998. "Neurolinguistics from the Middle Ages to the Pre-Modern Era: Historical vignettes". Handbook of Neurolinguistics ed. by Brigitte Stemmer & Harry A. Whitaker, 27-54. San Diego: Academic Press. Zonneveld, Ron van. 1988. "Two level phonology: structural stability and segmental variation in Dutch child language". First Language Acquisition ed. by Fred van Besien (= Association Belge de Linguistique Appliquée Papers, 12), 129-162. Antwerpen: University of Antwerpen.

. WORDS IN DEAFNESS AND STUTTERING

MORPHOSYNTACTIC FRAGILITY IN THE SPOKEN AND WRITTEN ITALIAN OF THE DEAF ROBERTO AJELLO, GIOVANNA MAROTTA, LAURA MAZZONI & FLORIDA NICOLAI University of Pisa 1. Introduction Morphology has often been characterized as particularly 'fragile', since it is subject to reduction processes of various kinds. Grammatical morphemes are, in fact, highly arbitrary, as well as not particularly salient from the articu­ latory, perceptive or semantic point of view. Such fragility is emphasized in those contexts naturally tending towards linguistic simplification, especially in the emergence of pidgins and the development of interlanguages origi­ nating in the spontaneous acquisition of L2, in formally guided learning of L2, and in the acquisition of L1 (see, e.g., Simone 1988, Berretta 1987; 1990 and Banfi 1993). Both the speech and the writing of subjects with acoustic deficits, like the deaf, show great alterations in morphological structure, although with marked interindividual variability. The present work deals with the linguistic production of some profoundly deaf Italian people who never wore prothesis. We analyze first their speech and then a written, 'not elicited' production, consisting of fax messages sent by a deaf person to other deaf people. The choice of this kind of production is due to the fact that it has many features in common with speech, such as 'ellipsis', brachylogy, topicalization processes, nominal sentences, and in general micro-planning involving very short sentences or even single words. The results of such analysis are then matched with those of a test aimed at checking the morphological ability of some of the subjects who wrote the fax messages and produced the analyzed speech. The coding of the data in the CHAT system is still in progress. All the deaf subjects examined (aged 40-65 years) belong to a generation who received an oralist instruction in specialized Institutes, where neither communication in Italian sign language (LIS = Lingua Italiana dei Segni) nor the use of mimetic gestures was encouraged, since this kind of interactional behavior was considered an obstacle to the correct learning of vocal Italian.

50

AJELLO, MAROTTA, MAZZONI & NICOLAI

We present the state of the art in the field of research dealing with the lin­ guistic production by the deaf, we provide a morphological analysis of the subjects' speech, examine their written production, focus on orthography; we discuss a morphosyntactic test and, finally, offer a global interpretation of the data. 2. State of the art The writing ability of deaf people has not been thoroughly researched. Still, the results of the scanty literature in this field uniformly emphasize the point that most difficulties reside in morphology, and, more specifically, in so-called 'free' morphology. These data, resulting initially from studies on deaf subjects and their mastering of the English language (Kelly 1987, 1993; Maxwell & Falick 1992; Paul & Quigley 1994), have been confirmed by re­ search made on deaf Italian people. The latter results have higher relevance, in consequence of the greater complexity of bound morphology in Italian in comparison to English. Research in this area, carried out with different ex­ perimental modalities, and concerning subjects of different ages, is due mainly to the team of scholars from the Istituto di Psicología, C.N.R. (Rome), who sometimes work in collaboration with outside scholars (see, in particular, Taeschner et al. 1988; Volterra & Bates 1989; Fabbretti 1997, 2000; Fabbretti et al. 1998). In order to evaluate the linguistic competence of the deaf through the me­ dium of writing, researchers have mostly utilised a form of production which they consider as the most spontaneous, namely the epistle or the diary, without excluding other methodologies, such as tales or structured tests (see Quigley, Power & Steinkamp 1977). These written productions are chosen for analysis, on the basis of the commonly shared opinion that spontaneous writing may provide a reliable index of the level of linguistic competence of the deaf, in consideration also of the relevant role played by this modality in various educational programs. Historically, literacy has always been the basis for the education of the deaf, since it was believed that also vocal language learning should start from competence in writing.1 All the research works pertaining to this subject, carried out both on Italian and on English, agree on the following points: -

deaf subjects produce shorter sentences and simpler syntactic structures in comparison to hearing subjects;

1 For a recent development within this methodology based on the central role of writing, see Radelli (1998).

MORPHOSYNCTATIC FRAGILITY IN ITALIAN DEAF

-

51

they also reveal a poor vocabulary and lexical rigidity; they have difficulties with relative, subordinate and pronominal clauses; although they appear to have difficulties in many aspects of the written language (on lexical, morphological, syntactic and pragmatic levels), the major problems concern morphology (omission, substitution, and, to a lesser extent, also addition of morphemes).

Let's consider the most relevant research works carried out on Italian deaf subjects. In Taeschner et al. (1988), 25 deaf students (aged 11-15 years), com­ pared with a control group of 125 hearing students (aged 6-16), were tested in the written modality by way of structured tasks aiming at eliciting answers relative to some morphosyntactic aspects, such as plural, clitics, and definite articles. As for the first two morphological categories, a delayed response is evident in the experimental group in comparison to the control group, but there is no deviance. Errors pertaining to the formation of plural consist mainly in the generalization of the most frequent morpheme -i and in treating feminine singular nouns in -e as plurals; moreover, the majority of errors take place in non-words (as happens with hearing subjects), and this suggests that plural nouns might be learnt by rote. The experimental group's responses to clitic items present a remarkable delay (more noticeable than in the case of plurals) and a prevailing pre-verbal collocation, even in cases when their position should be post-verbal. The general framework concerning the articles appears to be more complex in the experimental group and qualitatively different from the one relative to the control group: deaf subjects make more errors, generally due to an idiosyncratic strategy, since they make the article agree with the final vowel of the noun; thus le (f.pl.) fucile (m.sg.), instead of il (m.sg.) fucile (m.sg.) "the rifle"; i (m.pl.) notti (f.pl.), instead of le (f.pl.) notti (f.pl.) "the nights"; Ie (f.pl.) notte (f.sg.), instead of la (f.sg.) notte (f.sg.) "the night". The authors of this study take into consideration several factors that might influence the linguistic performance of the deaf, such as their delayed exposi­ tion to Italian, the reduced linguistic input they receive, due to their acoustic deficit, and the formal situation of teaching they experience. However, they realize that these general factors are not sufficient to explain why only a few specific morphological aspects are problematic. So, for instance, deaf subjects' difficulty in using articles might be due to the fact that these are homonymic with clitic pronouns, but this does not explain why difficulties concern mainly m.sg. lo and m.pl. gli. One might ascribe this latter type of misuse to the deaf subjects' lack of sensitivity to the phonological conditions requiring the selection of these articles. But also this explanation does not

52

AJELLO, MAROTTA, MAZZONI & NICOLAI

account for the hypergeneralization that makes the article agree with the final vowel of the noun. The errors are ascribed to the fact that free morphology is more difficult and more problematic than bound morphology, since free morphemes are typically short, not stressed, and cany relatively scarce semantic information.2 Volterra & Bates (1989) analyzed the written production (10 letters to a friend) of a congenitally deaf woman who has reached a good competence level in Italian. Her written production is compared with the one of 2 control subjects: one who has learnt Italian as L2 at the age of 16 and a native speaker. The written production of the deaf woman shows complex syntactic constructions and a lexical usage that are comparable with those of an Italian adult with a high literacy level; still, she makes frequent, systematic morphological errors, mostly in the domain of free morphology. The few errors in bound morphology concern long distance agreement. When the same deaf subject takes grammatical tests, she shows good metalinguistic knowledge of many rules governing free and bound morphemes of Italian, and makes very few errors. These results reveal that probably her problem does not lie at the level of morphological knowledge, but at the level of mor­ phological access and use. Also, her errors are different from the bilingual subject's. The deaf subject makes errors equally in substitutions, omissions, and insertions, while the bilingual subject makes error equally in bound and free morphology, but makes many errors of substitutions; additionally, the latter subject fails in making gender agreement, while this error is rare in the deaf subject; and above all, the bilingual subject makes lexical errors. All these data hint that the source of errors is not the same for the two subjects and is not referrable to their shared situation as L2 learners. In the hearing subject's responses, some errors may be plausibly interpreted as interference phenomena from L1, as, e.g., when he invents Italian words as a result of transliteration from English. As for the deaf subject, interference from LIS seems to be excluded, because her errors are similar to those of deaf adolescents who do not know LIs, and because they do not include either errors in the ordering of words, which is an element of great difference between the two languages, or errors in gender agreement, which is a category totally absent in LIS. They think that the morphological errors are not referrable to the fact that L2 was learnt after the critical period, first because, as a matter of fact, the subject in question was exposed to Italian very early, and secondly because the specific scheme of morphological enfeeblement is

This is the reason why deaf Italians very often are not able to recognize them in lip-reading.

MORPHOSYNCTATIC FRAGILITY IN ITALIAN DEAF

53

clearly selective. The explanation offered relates the morphological errors to a degraded input, which to the deaf often appears to consist of 'islands' of content words inserted in syntactic sequences (free morphemes in fluent production are short, uttered rapidly and without stress). Such degraded input may mean that the deaf subject has fewer opportunities to observe the application of morphological processes: he/she knows the rules at a conscious level, but never reaches that level of automaticity which is necessary for a fluent and correct performance. It thus seems that it is not possible to imagine a unique model of linguistic production which can explain the morphological weakness in learners of L2, in deaf subjects, and in agrammatic aphasics, who never show morphological enfeeblement in the presence of integral lexical and syntactic capacity. The results of the latest work by Fabbretti et al. (1998) confirm on the whole those of previous studies (Fabbretti 1997), although with slight differences, namely the probable influence of Lis in the origin of a few forms. In this work, 10 native signing deaf subjects, born of deaf parents, perform 4 writing tasks: a •summary of a short comic movie sequence, a summary of a video shoot in which deaf people play cards and talk about the game in LIs, a summary of another video shoot with a tale in LIS, and the writing down of a letter containing a refusal to an invitation made by a friend. One control group consists of 10 hearing subjects born of deaf parents who are native signers, and a second one consists of 10 hearing subjects who have never had contacts with deaf people. The results are in line with those from previous research, in particular Fabbretti et al. (1998): deaf subjects present a pattern of selective difficulties with morphology, especially with functional words. Interestingly enough, although Italian offers many occasions for errors in the bound morphology of nouns and verbs, such errors are infrequent. Deaf people seem to have access to some aspects of bound morphology, such as the plural, by rote. Although the data in Fabbretti (1997) reveal more frequent non-conventional orthographic forms in the writing of deaf people than in the writing of hearing subjects, the results of Fabbretti et al. (1998), according to which deaf subjects make fewer orthographic errors than the hearing control group, seem to confirm the observation made by Mayer & Welles (1996), who state that conventional spelling is an area of written language in which the deaf show relatively few difficulties. Interference from LIs is excluded because there are no errors in the ordering of words and because the hearing signers do not make the same kind of errors as the deaf subjects. The explanation for the errors is to be sought in deafness: the results are compatible with the statement that acoustic perception plays a special role in the acquisition and in the use of grammatical morphology.

54

AJELLO, MAROTTA, MAZZONI & NICOLAI

Similar observations have been made, among others, by Hanson & Wilkenfeld (1985) in relation to deaf subjects acquiring English: the lack of acoustic/articulatory mediation would have heavy consequences for the acqui­ sition of grammatical morphology, including the production and (perhaps) the understanding of prepositions, articles, pronouns and other free morphemes. Moreover, the specific difficulties the deaf have with the written language might derive from the way in which they learn the vocal language as children, which most of the time involves an intense one-year-long period of formal training; their written language is based on a vocal language they learn late, when the critical period is over. Many recent research works emphasize that the learning of morphology, even the morphology of a sign language, is par­ ticularly difficult after the critical period. It is possible that the explanation based on the critical period and the one based on phonetic/phonological mechanisms are not mutually exclusive: both in vocal and in sign languages, the aspects of morphology more closely connected to the transmission code (the acoustic one in the case of vocal languages, and the visual one in the case of sign languages) must be learnt early. The recent work presented by Chesi (2000) is particularly stimulating. Following the theoretical framework of generative grammar, he analyzed a corpus of verbal and written production of a group of 13 deaf Italian subjects, ranging in age from 6 to 17 years. All the subjects were prothesized, although with different auditory gain. Both verbal and written productions show comparable results. The number of errors is relatively high, especially for the subjects with less auditory gain from the prothesis. For the sake of clarity, Chesi (2000) speaks of non-standard forms instead of errors, since he believes that the grammatical competence of the deaf is completely compatible with the Universal Grammar (UG) hypothesis. Omission of functional elements, like articles, clitics, prepositions, auxiliaries, is well attested. Interestingly, some peculiar asymmetries emerge from the data. For instance, the article appears to be omitted in the Object Noun Phrase more often than in the Subject Noun Phrase. Enclitic forms are omitted more than proclitic ones, which confirms a trend already emerging in the work by Taeschner et al. (1998) and Fabbretti (2000). In verbal inflection, the 3rd sg.ps. is the most frequently used by the deaf children. Finite forms of the verb are often substituted by the infinitive ones. As for the auxiliaries, avere "to have" is often employed instead of ess ere "to be". Therefore, in Chesi's data too, functional morphology does not appear to be perfectly acquired by the deaf children, because of the poor stimulus they get from the input. However, in the interpretation of the data he collected, Chesi (2000) tries to demonstrate that the structure of the spoken and written language

MORPHOSYNCTATIC FRAGILITY IN ITALIAN DEAF

55

of the deaf does not conflict with the basic principles of Universal Grammar. In particular, Chesi (2000:193) believes that structural positions and the features they check are not absent, since they are activated and licensed by special lexical heads, even if some functional elements are omitted in the surface representation of the sentence. 3. Speech production We examined the speech production of 7 subjects, 4 males and 3 females. All of them but one (Al) normally use LIs as a communication language inside the deaf community, although they received an oral education during their childhood in the specialized Institutes they attended. The speech production corpus consists of spontaneous dialogues between two deaf people in the presence of a hearing participant or between a single deaf and a hearing person. The speech was recorded in a sound-proof room with professional instruments and then phonetically transcribed and acoustically analyzed. The total amount of speech recorded and analyzed is about 30 minutes. Let's first consider the bound morphemes in the recorded speech: the un­ stressed final vowels performing morphological functions are very often cen­ tralized or omitted. As a consequence, the marks of nominal as well as of ver­ bal inflection are often no more recoverable, and the meaning of the whole sentence may be affected by that. Table 1 shows the distribution of the possible outputs for the final vowels (correctly produced, omitted or centralized) in our speech corpus, with the in­ dication of the subjects analyzed. As we can see, the rate of final vowel re­ duction (26.0% omission plus 46.2% centralization) reaches the 72.2% of the total. + Total @ Al A2 G R1 R2 R3 S All

24.7 29.0 40.9 32.7 18.5 31.1 21.9 27.9

32.6 22.6 18.2 16.4 30.8 19.7 40.6 26.0

42.7 48.4 40.9 50.9 50.8 49.2 37.5 46.2

100.0 100.0 100.0 100.0 100.0 100.0 100.0 100.0

Table1: Percentage values of final vowels correctly produced {+), omitted (-) or centralized (@)

According to the limited literature on this topic (cf. Savy 2000), the same trend occurs also in the informal speech of hearing subjects, especially in the final position of intonational phrases or utterances, where vocalic reduction

56

AJELLO, MAROTTA, MAZZONI & NICOLAI

appears to be physiologically conditioned. As is well known, at the end of the breath group, glottal pressure goes down, global energy gets reduced, and ar­ ticulatory inertia increases. These same reasons may account for the word-final vowel reduction which can be observed in the speech produced by the deaf, with a crucial difference in the amplitude of the motor program window, which is no doubt wider in the hearing subjects than in the deaf. As a matter of fact, the deaf program their vocal production mainly in terms of phonological words; therefore the final position of a word normally coincides with the final position of an intonational phrase or utterance (see Ajello et al. 1998/1999). On the other hand, the occurrence of phonetic reduction in the speech of hearing people makes the visual input reaching the deaf blurred and ambiguous, which reinforces the natural trend towards segmental weakening in unstressed syllable. The processes of reduction and centralization involving final segments of words cannot be considered a pure performance effect on behalf of the deaf, but are supposedly related to incomplete morphological competence. That is why, in order to check the morphological competence of the deaf, we decided to eliminate the variable of the performance difficulties, and to analyze the written production where mechanical drawbacks are not present. 3.1 Free morphology in speech A second result of the analysis of speech was an imperfect use of free mor­ phemes. The trend towards the omission of determinants, prepositions, auxilia­ ries, and clitics is clear evidence of the difficulty of the deaf in mastering mor­ phological aspects. By limiting the exposition of the data only to the occurrence of determinants (in particular, definite articles) and prepositions, the definite articles are often omitted, although the relative percentage is not the same for all the morphemes. In Table 2, in comparison with a generalized omission of the definite m.sg. article in all its allomorphs (il, lo, l), a rather stable occurrence of the corresponding feminine form (i.e. la) is observed. il lo l la i All

+ 10.0 0.0 33.3 65.2 0.0 35.7

90.0 0.0 66.7 34.8 100.0 64.3

Total 100.0 0.0 100.0 100.0 100.0 100.0

Table 2: Percentage values of production (+) and omission (-)for the definite articles

MORPHOSYNCTATIC FRAGILITY IN ITALIAN DEAF

57

Our data agree with those collected by Chesi (2000:181), who found the highest percentage of omission for i and il (m.plur. and m.sing., respectively). The difference between the gender (masculine = weak versus feminine = strong) in the treatment of the articles is confirmed by the data relative to the acquisition of Italian as L1 (cf. Antelmi 1997), as well as L2. The feminine form appears to be more salient both for hearing children and for the deaf. In particular, la is more salient, since it is an open syllable, produced with the most open, and thus most visible, vowel. Moreover, in the feminine paradigm, the allomorphy is more restricted than in the masculine one. As far as the use of prepositions is concerned, we observe a general trend towards omission. This finding too is not unexpected, since it is often reported in the literature. However, there are some interesting differences among the prepositions occurring in our corpus. As is shown in Table 3, omission is par­ ticularly strong in the case of a ("to"), di ("of") and da ("from"), but rather weak in the case of per ("for "). a di da in con per All

+

-

Total

6.5 46.2 16.7 54.5 58.3 75.0 37.8

93.5 53.8 83.3 45.5 41.7 25.0 62.2

100.0 100.0 100.0 100.0 100.0 100.0 100.0

Table 3: Percentage values of production (+) and omission (-) relative to prepositions

As a matter of fact, the first three of those prepositions not only are very frequently used in Italian, but also play more than one thematic role. These data seem to indicate that the deaf are rather able to use 'unambigous' prepo­ sitions, while they find it more difficult to select a preposition with different syntactic functions. As far as the production and the right use of the preposi­ tion per are concerned, we will see that the same priority emerges as from the analysis of the written production we examined. On the other hand, we would like to recall that in LIS there is a specific sign corresponding to Italian per, while all other Italian prepositions have no corresponding sign. 4. Written production The written production, consisting of 200 fax messages, involved 14 male and 26 female subjects, all of them expert signers, but, as in the previous case, lacking a mother tongue acquired naturally through exposition from birth. All the subjects we have examined learnt the vocal language through a hard ten-

58

AJELLO, MAROTTA, MAZZONI & NICOLAI

year training period in the Institutes, which started at the age of 6 to 9, while they spontaneously acquired a rudimentary sign language, also in the Institutes, simply by communicating with older pupils in limited situational contexts. The interpretation and normalization of fax texts has turned out to be a very complex, not always satisfactory, operation, though one of the authors of the present work, Laura Mazzoni, not only has had life-long experience of the deaf world, but also knows personally and very well the authors of the fax messages and meets them almost daily. The morphological analysis of the written corpus reveals heavy reduction of the bound morphemes, omission or misuse of functional elements, lexicalization of grammatical categories and weak consciousness of some derivational mechanisms. 4.1 Heavy reduction of the bound morphemes The reduction of the bound morphemes and the consequent overextension of a few forms concern mainly the verb forms, and much less the forms of the other grammatical categories. The 3rd ps.sg. of the present indicative and the infinitive tend to function as the basic verb forms in independent clauses: (1) a. io scrive I writes-IND "I writes" b. io rimprovera I scolds-lND "I scolds"  scrive writes-IND "writes" but see also:

target: scrivi write-2nd.PERS-SG-IMP "do write!"

(2) a. io mai pulire I never clean-lNF "I never to clean" b. tu dire tuo marito you-SG tell-INF your-M-SG husband "you do tell your husband" After a coindexed modal verb, an inflected form of the verb is sometimes pre­ sent where an infinitive is expected: (3) a. voglio want-1St.PERS-SG-IND-PRES

"I want comes"

viene COmes-IND

MORPHOSYNCTATIC FRAGILITY IN ITALIAN DEAF

59

b. voleva andato wanted-3rd.PERS-SG-IND-IMPF gone-M-SG. "she/he wanted gone"  puoi tu viene can-2nd.PERS-SG-IND-PRES you-SG comes-lND "you can comes" The present tense is often overextended to past and future contexts: (4) a ieri sera scrivo subito yesterday evening write-1 st.PERS-SG-IND-PRES immediately "last night I immediately write" b. stamattina arriviamo La Spezia this morning arrive-lst.PERS-PL-IND-PRES La Spezia "this morning we arrive La Spezia" The past participle without auxiliary is used to indicate an action achieved in the past, and the aspectual value is sometimes reinforced by the use of the ad­ verb già "already": (5) a. io andato Lucca circolo I gone-M-SG Lucca club "I gone Lucca club" b. già morto in incidente already dead-M-SG in accident "he died in accident" The 2nd ps.sg. of the indicative tends to be extended to cover the corresponding forms of the imperative and the former is also used in the negative imperative: (6)

pensi target: pensa think-2nd.PERS-SG-PRES-lND think-2nd.PERS-SG-lMP "(you) think" "do think! " (7) non fa target: non fare not do-2nd.PERS-sING-IMP do-INF "not do ! " "do not do ! " As for nouns and adjectives, a slight tendency to overextension of the mascu­ line singular form is detectable: (8) a. la gonna azzurro the-F-SG skirt-F-SG blue-M-SG "the blue skirt" b. domenico scorso target: Sunday-M-SG last-M-SG "last Sunday"

domenica scorsa Sunday-M-SG last-M-SG

60

AJELLO, MAROTTA, MAZZONI & NICOLAI

c. nuovo luce target: nuova luce new-M-SG light-F-SG new-F-SG light-F-SG "new light" d. più bello festa target: bella festa more nice-M-SG party-F-SG nice-F-SG party-F-SG "nicer party" at the expenses of the feminine singular, and masculine and feminine plurals: (9)

siamo

molto

contentis simo

are-1 st.PERS-SG-IND-PRES

very

happy-SUPERL-M-SG

"we are very happiest" A slight tendency to isomorphism is detactable in the case of the -e ending of nouns-adjectives-past participles, which tends to be used as a mark of plural: (10) a. baci forte kiss-M-PL strong-M-SG "strong kisses" b. altre amici other-F-PL friend-M-PL "other friends" (11) tutti sono stati gentile all-M-PL are-3rd-PERS-PL-PRES-IND be-PAST-PART-M-PL kind-SG "all of them have been kind" Morphological reduction is related to a phenomenon of substitution of bound forms with free forms, as in the case of verbal morphology, where the use of personal pronouns compensates for the lack of personal endings (see the al­ ready quoted examples 2a, 2b, etc). In the case of nominal morphology, the free morphemes indicating definiteness, such as the articles, are very often replaced by demonstratives: (12) a. quello Beautuful that-M-SG (soap-opera) Beautiful "that Beautiful" b. quello registrator e that-M-SG tape.recorder-M-SG "the tape recorder" c. quella Giovanna I. that-F-SG Joan I. "that Giovanna I."

MORPHOSYNCTATIC FRAGILITY IN ITALIAN DEAF

61

4.2 Omission or misuse of functional elements Functional elements (determinants, prepositions, auxiliaries, copula, clitics) tend to be omitted or misused. The correct usage of prepositions covers approximately 50% of the total amount of contexts, their incorrect usage summing up cases of omission and substitution of the correct preposition needed, and cases of incorrect insertion of a preposition. A more detailed examination of their occurrences reveals that the average percentage just given is illusive, since, if linguistic routines involving prepositions are neglected, the percentage of correct usage of all prepositions greatly decreases. Only one preposition turns out to be extensively used in a correct manner: the preposition per ("for"), which also undergoes a process of overextension at the expenses of a ("to") and di ("of "). This result agrees with what we already found in the analysis of speech production. The reason for this high percentage of correct usage cannot lie in the frequency of that preposition in the input language, because, for instance, the preposition per does not rank as the most frequent one, being preceded by di, a, and in. More probably, its correct usage, which also emerges from the analysis of the speech data, might be due partly to its visual salience, since it begins with a bilabial articulation, and partly to the interference from LIS, where there is a specific sign corresponding to Italian per, while all other Italian prepositions have no corresponding sign. The analysis of the definite article gives approximately the same results as the analysis of the speech data: more than 50% of omissions, but a higher percentage for the masculine forms. As for auxiliaries, they are frequently omitted, as we have seen, but, when used, they often show substitution of essere with avere. (13) a. hanno sconvolto délia morte have-3rd.PERS-PL-PRES-IND upset-M-SG-PAST.PART of.the-F-SG death-F-SG "(they) have upset of death" b. ci ha [target: è] piaciuto to.us-OBL. has-IND [ is] like-M-SG-PAST.PART "we liked it" There are no cases of auxiliary avere "to have" replaced by essere "to be". The only case of a substitution of this kind concerns essere as a main verb, not as an auxiliary: (14)

io non sono [target: ho] problema I not am [ have] problem "I am no problem"

62

AJELLO, MAROTTA, MAZZONI & NICOLAI

There are cases of existential generalized 3rd ps.sg. c'è with the function of possessive avere "to have": (15)

lui c'è un he/him there is a-M-SG "he there is a friend"

amico friend-M-SG

This probably represents interference from LIS. The form c'è is overextended at the expense of other forms such as c'era ("there was"), ci sono ("there are") and so on. Clitics are also often omitted; the use of the form si "oneself, him/herself, one/impersonal") is overextended at the expense of other forms of clitics: (16) a. si ricordo himself-3rd-PERS-SG-CL. remind/remember-1 st-PERS-PRES-IND "I remind himself' b. non candidarsi not candidate-lNF-REFL/3rd-PERS-SG-CL. "not to candidate one-/him-self ' Our data match the results by Taeschner et al. (1988), Chesi (2000) and Fabbretti (2000): enclitic forms are omitted more frequently than proclitic ones. The existential particle ci and the relative pronoun che are also fre­ quently omitted, while che as a complementizer is overextended: (17) a. è moho contento che venire is very happy-M-SG that-COMPL come-INF "(she/he) is very happy that to come" b. voglio sapere che want-1 st-PERS-SG-PRES-IND know-INF that-COMPL. tuo marito è contento your-M-SG husband-M-SG is happy-M-SG "want to know that your husband is happy" Sometimes it is also inserted without justification at the very beginning of a sentence: (18)

Che ieri sono andata a Comune that-COMPL yesterday am-IND gone-F-SG to Town hall "That yesterday I went to Town hall"

Two different determinants are sometimes piled up: (19) a. questo il fax this-M-SG the-M-SG fax "this the fax"

MORPHOSYNCTATIC FRAGILITY IN ITALIAN DEAF b. fra qualche settimana between some week "in a few weeks next"

63

prossima next

The verb fare "to do, to make" used to form causatives, is often omitted: saparti [target: farti sapere "to let you know"]; visitare [farmi visitare "to have myself examined"]; vedere [far vedere "to show"]. The most striking omission of morphemes is that of the main verb, which must be recovered from the co-text or context: (20) a. mattino casa tua, venerdi mercato morning house-F-SG your-F-SG Friday marke-M-SG "in the morning at your home, on Friday market" b. amici Mara dopo pranzo friend-M-SG Mara after lunch "Mara('s) friends after lunch" as happens in the case of the omission of the subject. (21)

è tutto bene il nostro viaggio; is-IND all-M-SG fme-ADV the-M-SG our-M-SG trip-M-SG anche è arrivato esatto also is-IND arrive-PAST.PART-M-SG exact-M-SG "our trip was all fine, even (the train) was on time"

What prevails in these cases is a pragmatic communicative strategy. As a consequence of the defective morphological competence, agreement be­ tween article and noun, possessive/demonstrative adjective and noun, adjective and noun, subject and verb is sometimes incorrect. (22) a. la tua notizie the-F-SG your-F-SG. news-F-PL "your news" b. parole bianco word-F-PL white-M-SG "white words" Sometimes the agreement seems to be oriented to the analogy of endings. (23)

i mani the-M-PL hands-F-PL "the hands"

64

AJELLO, MAROTTA, MAZZONI & NICOLAI

4.3 Lexicalization of grammatical categories This is apparent in at least two cases: the temporal adverb is sometimes the only mark of temporality: (24) a. dopo sto cucendo [future tense] after/afterwards stay-1st-PERS-SG-PRES-IND sew-GER "afterwards I am sewing" b. ieri sera scrivo subito [past tense] yesterday evening write- lst-PERS-SG-PRES-IND immediately "last night (I) immediately write" and the numeral is sometimes the only indication of plurality: (25)

due idraulico two plumber-M-SG "two plumber"

4.4 Weak consciousness of some derivational mechanisms A partial sense of derivational morphology is present in the corpus, as is shown by neoformations such as rabbio [mi arrabbio "I get angry"], accompagnia [compagnia "company"], non abbanzano [non sono abbastanza/non bastano "they are not enough"], intermometro [termometro "thermometer"], which all reveal some sort of consciousness of some derivational mechanisms. 5. Lexicon and interference from LIS Besides the four phenomena illustrated above, other characteristics are worth noticing, especially a fairly good level of lexical competence. Strikingly enough, the lexical competence of the deaf appears to be superior to their mor­ phological competence. The lexicon is not severely limited and is also properly used, with very few exceptions: le tue confidenza [i tuoi pensieri "your thoughts"]; vi auguro come bravi docenti [mi congratulo con voi perché siete dei bravi docenti "I congratulate you, since you are good teachers"]; ti spiego il successo [ti spiego l'accaduto "I explain to you what happened"]; rimango a mio marito [penso sempre a mio marito "I always think of my husband"]. The lexicon is not severely limited even if we include here the phenomenon of grammatical category change: [ho fotocopie = ho fotocopiato "I have photo­ copied"]; devo coraggio la mia vita [devo farmi coraggio "I must take courage"]; io sempre pensiero a te [io pens sempre a te "I always think of you"]; affari [sbrigare faccende "to attend to business"]; avanti [andare avanti "to go ahead"]; accordo [accordarsi "to reach an agreement"]. The lexicon is sometimes affected by interference from LIS, and, very interestingly, this hap­ pens in two types of circumstances. It occurs in cases in which LIS has just one

MORPHOSYNCTATIC FRAGILITY IN ITALIAN DEAF

65

sign corresponding to several Italian words; thus, for example, the expression il treno è esatto [in orario "the train is on time"] derives from the partial cor­ respondence of a LIS sign to several Italian adjectives (puntúale, preciso, etc.). Alternatively, it occurs in cases where Italian is affected by the lexical choice of LIS; for instance, since LIS can choose between QUESTA MATTINA and ADESSO MATTINA, the expression adesso mattina in the written corpus derives from LIS (By convention, expressions written with capital letters represent LIS signs). In­ correct lexical elements or omissions of expected lexical items are sometimes due to interference in cases in which a certain specification in LIS is expressed not by a manual sign, but by a qualitative modification of the movement of the sign, a modification of the facial expression visto tempo [visto da tanto tempo "seen since a long time"] or a modification of the direction of the movement, which may explain the frequent confusion between ricevere/spedire (fax) "to receive/to send (fax)"; arrivare/venire "to arrive/to come". The inevitable loss or the change of information is then due to a linear rendering, on behalf of the deaf, of the multi-layered structure of signs, which does not take into account the holistic multi-componential nature of the sign itself. The interference from LIS at the morphological level seems to be very limited: apart from the case of the preposition per, our data show: 1.

2.

only one case of plural expressed through reduplication of the noun, as is common in LIS (cuori cuori "hearts"), although both occurrences of the noun are marked with the plural ending. The strategy of redu­ plication in the case of plurals may not be due to interference from LIS, but to its iconic expressiveness, which makes it frequently used in all cases of linguistic simplification (acquisition, pidgins, and interlanguages); a few cases of confusion between possessives (tuo/suo "your/his" or "her"); tua/vostra "your", sg. and pl.), which in LIS are expressed with deictic gestures pointing at different but close sectors of the space in front of the signer. The incorrect renderings in Italian may be due to the difficulty in identifying the portion of space relevant to the distinction or to the difficulty in rendering the directionality.

Interference from LIS seems to be on the whole very heavy in the domain of syntax, in particular in the order of linguistic elements, which very frequently follows the LIS ordering, as in interrogative and negative sentences; for instance: (26) a. porta scuola chi takes-IND school who "who takes (him /herto) school?"

66

AJELLO, MAROTTA, MAZZONI & NICOLAI

b. rispos ta non answer-F-SG not "(there is) not answer" The same happens in the case of focalization processes concerning the specifying items: (27) a. panchine dove vicino casa tua bench-F-PL where near house-F-SG your-F-SG "benches where near your place" b. ci vediamo che giorno us/ourserlves-REC see-1 st-PERS-PL-PRES-IND which/what day "What day are we meeting?" 9 gennaio 9 January "(On) January 9th" To speak of LIS interference does not imply that the status of L1 is con­ ferred on LIS: the visuo-gestural communication system is simply the more natural and the more frequently used.3 6.

Orthography As evidenced in the literature, orthography does not represent remarkable difficulties for the deaf, probably because literacy has been traditionally a do­ main of very intensive training (see Ajello et al 2000). The most frequent er­ rors made by our subjects concern: 1.

2.

3.

3

confusion between letters corresponding to voiced and voiceless con­ sonants: rimanco [rimango "I stay"]; statio [stadio "stadium"]; pangia [pancia "belly"]. frequent degemination or misplaced gemination: abracci [abbracci "hugs"]; Ricardo [Riccardo "Richard"]; Coseta [Cosetta]; maremana [maremmana "native or inhabitant of Maremma-F-SG"]; avvisso [avviso "warning, notice"]; occassione [occasione "occasion"]; costanno [costano "they cost"]. substitution of graphemes: golfio [gonfio "swallen"]; pasiensa (pazienza "patience; (it is a) pity"].

We intend to check in detail, in a future work, the hypothesis of syntactic interference from LIS upon the written production of the Italian deaf, by analyzing similar spontaneous texts, very close to speech, produced by a non-signer deaf and addressed to non-signers, either deaf or hearing.

MORPHOSYNCTATIC FRAGILITY IN ITALIAN DEAF

4. 5.

67

iteration of contiguous vowel grapheme: volontieri [volentieri "gladly"]; Monaca [Monica]; sapare [sapere "to know"]. lack of graphic indication of stress: gia [già "already"]; e [è "is"]; lunedi [lunedi "Monday"];perche [perché "why, because"].

On the other hand graphemic clusters are, as a rule, correctly written with only one exception: accompagniare [accompagnare "to accompany"]. Another point of interest is a non-standard use of punctuation. In particular, the interrogation mark is generally overextended (stiamo bene, spero pure di voi? "We are fine, I hope so are you?"; volevo sapere come è andata a Roma? "I wanted to know how did it go in Rome?"), but is sometimes missing when it should be appropriate {porta scuola chi "takes (to) school who?"; panchine dove vicino casa "benches where near house?"). Full stop is often omitted and semicolon is totally absent, while very few commas are used. 7. A test of morphosyntax In order to check the grammatical competence of the deaf, a test was given to a group of the subjects who wrote the faxes examined so far and who pro­ duced the speech analyzed in §§ 3 and 4. We wrote 79 sentences, 12 correct, and 67 incorrect, i.e. containing a grammatical error.4 The typology of the er­ rors includes: agreement within the NP, agreement within the VP, omission of a preposition, selection of a wrong preposition, omission of the article, selection of a wrong allomorph for the article, omission of the auxiliary. Eleven subjects participated in the test. Ten of them are profoundly deaf and one is severely deaf. None of them has ever been prothesized. Their age ranges from 50 to 65 years. Six are females and five are males. All of them have good LIS competence. Their education in special institutes covers the elementary cycle of the school. The test was applied in an informal atmosphere, and it was presented by one of authors, who is a signer (Laura Mazzoni). Each sentence was written on a single piece of paper. Each subject received an envelope containing all the sentences given in a random order. The signer gave the following instructions to the subject: "This envelope contains some pieces of paper, where an Italian sentence is written. The sentences may have errors. Read every sentence and decide whether it is right or wrong. If the sentence is wrong, i.e., it has an error, write the right form on the same paper. Read all the sentences and when you finish, put all the pieces of paper in the envelope. There are no time limits to do the task." An every-day lexicon and a very simple 4

Here and henceforth, the term 'error' is used in a rather general meaning, basically as synonymous of non-standard form.

68

AJELLO, MAROTTA, MAZZONI & NICOLAI

syntax were used in the construction of the sentences, in order to guarantee a relatively easy task. The subjects' comprehension was checked by asking them to reproduce each sentence in LIS after reading it. The relevant parameters referrable to the subjects are presented in Table 4. The LIS competence and the speech intelligibility have been evaluated by the signer with reference to the linguistic abilities shown by the subjects in communicative interactions. Age of education refers to the age when the subject began the education in a special institute for deaf. Time refers to the number of minutes needed to do the task. Subject GI GL GM PB RS SP AS LM BN RM

1 VP

Sex F F F F F F M M M M M

Age of education 6 6 6 7 9 7 7 11 7 6 6

LIS competence 7 6 7 7 7 7 7 7 6 8 8

Intelligibility 4 5 4 4 4 4 2 5 3 3 4

Time 43' 20' 36' 35' 38' 31' 36' 30' 28' 22' 15'

Table 4: List of the subjects who participated in the test

Many subjects observed that the test was too long. Some of them did not understand the meaning of 'error' immediately, and asked for some examples. Almost all the subjects had great difficulties in maintaining their concentration for all the time needed. The attention decreased after roughly ten minutes. A pause was then introduced during the test. No subject knew the goal of the ex­ periment. Three subjects (GL-F, AS-M, RM-M) said that there were no errors in the sentences. These subjects were excluded from the analysis, since they were not reliable. The results are summarized in Tables 5, 6 and 7.5

The abbreviations used in the following tables are:

M: male, -Prep: Sub st Prep Prep: AgrNP:

F: female omission of a preposition substitution of a preposition total of errors referring to prepositions agreement within the Noun Phrase

AgrVP: -Aux: -Det: AlloDet: Det:

agreement within the Verb Phrase omission of an auxiliary form omission of a determinant wrong selection of the allomorph of the determinant total of errors referring to determinants

MORPHOSYNCTATIC FRAGILITY IN ITALIAN DEAF

69

Tables 5 show the percentage values concerning the sentences containing an error, with reference to the total number of errors for each grammatical category we considered. It is divided into three main sections, the first one giving the number of uncorrected errors, the second one giving the wrong corrections, and the third one showing the right corrections made by the subjects. UNCORRECTED ERRORS

Subj

Prep

-Prep

Subst Prep Agr NP Agr VP

-Aux

-Det

Alio Dei

Det 19.6

BN-M

32.6

13.0

19.6

37.0

8.7

2.2

10.9

8.7

LM-M

33.9

16.1

17.9

32.1

10.7

3.6

14.3

5.4

19.6

VP-M

36.0

4.0

32.0

28.0

4.0

0.0

24.0

8.0

32.0

GI-F

100.0

100.0

0.0

0.0

0.0

0.0

0.0

0.0

0.0

GM-F

29.4

5.9

23.5

29.4

5.9

0.0

17.6

17.6

35.3

PB-F

35.6

18.6

16.9

35.6

8.5

1.7

11.9

6.8

18.6

RS-F

48.1

18.5

29.6

29.6

3.7

0.0

14.8

3.7

18.5

SP-F

41.7

25.0

16.7

29.2

4.2

0.0

16.7

8.3

25.0

Subj

Prep

-Prep

Allo Det

Det

WRONG CORRECTIONS

Subst Prep Agr NP

Agr VP

-Aux

-Det

BN-M

57.1

42.9

14.3

42.9

0.0

0.0

0.0

0.0

0.0

LM-M

66.7

50.0

16.7

16.7

0.0

0.0

0.0

16.7

16.7

VP-M

25.0

25.0

0.0

75.0

0.0

0.0

0.0

0.0

0.0

GI-F

33.3

0.0

33.3

66.7

0.0

0.0

0.0

0.0

0.0

GM-F

30.0

20.0

10.0

60.0

0.0

10.0

0.0

0.0

0.0

PB-F

0.0

0.0

0.0

100.0

0.0

0.0

0.0

0.0

0.0

RS-F

50.0

33.3

16.7

50.0

0.0

0.0

0.0

0.0

0.0

SP-F

27.8

11.1

16.7

44.4

11.1

0.0

11.1

5.6

16.7

Subj

Prep

-Prep

Subst Prep

BN-M

30.8

23.1

1.1

23.1

RIGHT CORRECTIONS

Agr NP Agr VP

-Aux

-Det

Alio Det

Det

15.4

1.1

23.1

0.0

23.1

LM-M

0.0

0.0

0.0

100.0

0.0

0.0

0.0

0.0

0.0

VP-M

35.1

27.0

8.1

35.1

13.5

5.4

5.4

5.4

10.8 19.4

GI-F

33.9

17.7

16.1

33.9

9.7

3.2

12.9

6.5

GM-F

38.5

23.1

15.4

30.8

12.8

2.6

12.8

2.6

15.4

PB-F

40.0

20.0

20.0

0.0

20.0

20.0

20.0

0.0

20.0

RS-F

21.2

15.2

6.1

36.4

15.2

6.1

12.1

9.1

21.2

SP-F

33.3

16.7

16.7

33.3

12.5

8.3

8.3

4.2

12.5

Table 5: Percentage values of evaluation of sentences containing an input error

70

AJELLO, MAROTTA, MAZZONI & NICOLAI

In Table 6 we show the results concerning the correct sentences in absolute terms, since here percentage values would be meaningless. The number of corrections wrongly made by the subjects are in the first section, the number of correct sentences judged as correct are in the second section. WRONG CORRECTIONS - Prep Subst Prep

Prep

NO CORRECTIONS

Agr SN Agr SV - D e t Allo Det

- Aux

BN-M

0

0

0

0

0

0

0

0

BN-M

LM-M

0

1

1

2

1

0

0

0

LM-M

12

VP-M

0

1

1

0

0

0

0

0

|VP-M

GI-F

0

0

0

0

0

0

0

0

|GI-F

GM-F

0

0

0

0

0

1

0

0

GM-F

10

PB-F

0

1

1

1

0

0

0

0

PB-F

10

RS-F

0

1

1

0

0

1

0

0

RS-F

10

SP-F

0

0

0

2

0

2

0



|SP-F

8

9 12 8

Table 6: Numerical results for correct sentences (no errors in input)

The subjects showed variable results. With reference to their grammatical competence, as judged on the grounds of the results obtained, our subjects can be grouped into two classes: a) b)

deaf with good grammatical competence (VP, GI, GM, RS); deaf with poor grammatical competence (BN, LM, PB, SP).

The subjects belonging to the first class show the highest percentages of right corrections (on average > 50%), and the lowest percentages of errors (wrong corrections and no correction). The results we collected seem to idicate that there is no strong correlation between LIS competence and grammatical competence: although all our subjects have good knowledge of LIS, with a small difference among them (cf. Table 4), their performance in doing the test was different. If we consider the wrong sentences in input (cf. Table 5), the number of wrong corrections is higher than the number of missed corrections. This is true for both the identified groups of deaf subjects, although the difference is obviously much more evident in the case of group b), that is, in deaf subjects with poor grammatical competence, since the amount of errors they make is greater. Apart from interindividual differences, in all our subjects there is a general trend towards the recognition that a sentence is wrong, but, at the same time, without the ability to select the right form. On the other hand, the number of wrong corrections on right sentences given in input is very low for both groups of deaf subjects (cf. Table 6), which suggests that they are able to judge a right sentence as grammatical. In Table

MORPHOSYNCTATIC FRAGILITY IN ITALIAN DEAF

71

7, the percentage values of the distribution of errors in relation to the three different possible behaviours of the subjects are given. Our aim is to check whether there is a concentration of errors (wrong corrections as well as no corrections) on some categories more than on others. UNCORRECTED ERRORS

Prep

-Prep Subst Prep Agr SN Agr SV

- Aux

-Det

BN-M 32.6

13.0

19.6

37.0

8.7

2.2

10.9

8.7

LM-M 33.9

16.1

17.9

32.1

10.7

3.6

14.3

5.4

19.6

VP-M 36.0

4.0

32.0

28.0

4.0

0.0

24.0

8.0

32.0

100.0

0.0

0.0

0.0

0.0

0.0

0.0

0.0

GI-F

100.0

Allo Det Det 19.6

GM-F 29.4

5.9

23.5

29.4

5.9

0.0

17.6

17.6

35.3

PB-F

35.6

18.6

16.9

35.6

8.5

1.7

11.9

6.8

18.6

RS-F

48.1

18.5

29.6

29.6

3.7

0.0

14.8

3.7

18.5

SP-F

41.7

25.0

16.7

29.2

4.2

0.0

16.7

8.3

25.0

Prep

-Prep Subst Prep Agr SN Agr SV

WRONG CORRECTIONS

- Aux

-Det

Allo Det - Aux

BN-M 57.1

42.9

14.3

42.9

0.0

0.0

0.0

0.0

0.0

LM-M 66.7

50.0

16.7

16.7

0.0

0.0

0.0

16.7

16.7

VP-M 25.0

25.0

0.0

75.0

0.0

0.0

0.0

0.0

0.0

0.0

33.3

66.7

0.0

0.0

0.0

0.0

0.0

20.0

10.0

60.0

0.0

10.0

0.0

0.0

0.0

0.0

0.0

0.0

100.0

0.0

0.0

0.0

0.0

0.0

RS-F

50.0

33.3

16.7

50.0

0.0

0.0

0.0

0.0

0.0

SP-F

27.8

11.1

16.7

44.4

11.1

0.0

11.1

5.6

16.7

Prep

-Prep Subst Prep Agr SN Agr SV

GI-F

33.3

GM-F 30.0 PB-F

RIGHT CORRECTIONS

- Aux

-Det

Allo Det - Aux |

BN-M 30.8

23.1

7.7

23.1

15.4

7.7

23.1

0.0

23.1

LM-M 0.0

0.0

0.0

100.0

0.0

0.0

0.0

0.0

0.0

VP-M 35.1

27.0

8.1

35.1

13.5

5.4

5.4

5.4

10.8

GI-F

33.9

17.7

16.1

33.9

9.7

3.2

12.9

6.5

19.4

GM-F 38.5

23.1

15.4

30.8

12.8

2.6

12.8

2.6

15.4

PB-F

40.0

20.0

20.0

0.0

20.0

20.0

20.0

0.0

20.0

RS-F

21.2

15.2

6.1

36.4

15.2

6.1

12.1

9.1

21.2

SP-F

33.3

16.7

16.7

33.3

12.5

8.3

8.3

4.2

12.5

Table 7: Percentage values of distribution of errors in relation to the three different possible types of behavior of the subjects

72

AJELLO, MAROTTA, MAZZONI & NICOLAI

As a matter of fact, our subjects correct the errors pertaining agreement of VP, as well as omission of Auxiliary or Determinant more than the ones con­ cerning prepositions and agreement within the Noun Phrase. In particular, the subjects with better morphological competence show very clearly the dif­ ference between the two classes of errors. At the same time, the wrong corrections again concentrated on the selection of the prepositions and of the agreement marker within the NP. What is surprising is that even in the case of right corrections, the percentage was still higher in the case of prepositions and agreement of NP. The datum could be constrained, at least in part, by the higher number of sentences containing an error relative to prepositions and agreement within NP. However, we still believe that the selection of the preposition represents the most difficult task for a deaf person. 8.

Conclusions As is apparent from the above-mentioned phenomena, the deaf we exam­ ined made exactly the same errors in Italian as all other learners of Italian as L2 at the initial stages of their interlanguages. Unlike the latter, the deaf sub­ jects showed a discrepancy between a poor competence at the morphological level and a fairly good competence at the lexical level, thus revealing that morphology represents their weak point. Additionally, the deaf subjects made more errors in free rather than in bound morphology, which nonetheless is not sufficiently mastered. The condition of this homogeneous group of deaf people is a very peculiar one, very different from the one that characterizes hearing, spontaneous learners of Italian as L2 for these two reasons: (a) they have no mother tongue acquired from birth, through which they may have access to the principles of UG, since LIS has been acquired naturally but in a late period; (b) the input is always problematic for the deaf, because, even after hard formal training, speech-reading cannot give all the necessary infor­ mation about the target phonological elements. Moreover, the lin­ guistic input for the deaf depends on an interpretation hearers give about their presumed ability to decode the message. The linguistic in­ put is then 'tailored' and in fact reduced, simplified (along lines that have never been investigated so far), so as to fit' their competence, as it is conjectured by hearers: it is an input basically reduced to its lexical content and structured so as to follow semantic and pragmatic strategies.

MORPHOSYNCTATIC FRAGILITY IN ITALIAN DEAF

73

These data could match an interpretation according to which, in the case of deaf people with neurolinguistic parameters of the kind illustrated above, the vocal language does not have a complete access to UG, not even an indirect one, mediated by the characteristics already implied in LI, but is learnt mainly through an explicit and formal approach, which gives as a result a system of rules which is never completely internalized. Our data seem to fit the maturational hypothesis rather than the continuity one, since the grammar of the deaf would be based on semantic and thematic relationships, rather than on merely syntactic ones. The deaf s very peculiar way of learning the language explains also the interindividual variability, which includes cases of very good linguistic performance. The process of learning is based mainly on general, not specifically lin­ guistic, cognitive mechanisms, as is apparent from the discrepancy between a fairly good lexical competence and a poor morphological competence heavily dependent on the input, and a similarly poor syntactic competence which relies fundamentally on pragmatic communication principles. The data we reported suggest the necessity of early didactic approaches with specific targets, so that the new deaf generations may better cope with the difficulties of integration.

REFERENCES Ajello, Roberto, Giovanna Marotta & Florida Nicolai. 1998-1999. "Uno studio sperimentale della produzione vocale di sordi italiani". Quaderni della Sezione di Glottologia e Linguistica,Universitàdegli Studi di Chieti 10/11.17-40. Ajello, Roberto, Antonella Convalle & Florida Nicolai. 2000. "Strategie di apprendimento della scrittura in bambini sordi". Atti del II Convegno Nazionale sulla Lingua Italiana dei Segni, Genova 25-27 setiembre 1998: Viaggio nella città invisibile ed. by Caterina Bagnara, Giampaolo Chiappini, Maria Pia Conte & Michela Ott, 158-180. Tirrenia: Edizioni Del Cerro. Antelmi, Donatella. 1997. La prima grammatica dell'italiano: Indagine longitudi­ nale sull'acquisizione della morfosintassi italiana. Bologna:I1Mulino. Banfi, Emanuele. 1993. L 'altra Europa linguistica.Firenze: La Nuova Italia. Berretta, Monica. 1987. "Per uno studio suH'apprendimento dell'italiano in contesto naturale: il caso dei pronomi personali atoni". L 'apprendimento spontaneo di una seconda lingua ed. by Anna Giacalone Ramat, 329-352. Bologna:I1Mulino. Berretta, Monica. 1990. "Morfologia in italiano lingua seconda". Storie dell'italiano e forme dell'italianizzazione ed. by Emanuele Banfi & Patrizia Cordin, 181-201. Roma: Bulzoni. Caselli, Maria Cristina, Simonetta Maragna, Laura Pagliari Rampelli & Virginia Volterra. 1994. Linguaggio e sordità. Firenze: La Nuova Italia.

74

AJELLO, MAROTTA, MAZZONI & NICOLAI

Chesi, Cristiano. 2000. Inferenze strutturali. Analisi sull'uso degli elementi funzionali neue produzioni verbali dei bambini sordi. Dissertation, University of Siena. De Mauro, Tullio, Federico Mancini, Massimo Vedovelli & Miriam Voghera. 1993. Lessico difrequenza dell'italianoparlato. Milano: Etas. Fabbretti, Daniela. 1997. Scrivere e segnare: la costruzione del discorso nell 'italiano scritto e nella Lingua Italiana dei Segni delle persone sorde. Ph.D. dissertation, University of Rome, Faculty of Psychology. Fabbretti, Daniela. 2000. "L'italiano scritto dai sordi: un'indagine sulle abilità di scrittura dei sordi adulti segnanti nativi". Rassegna di Psicología 17.1-21. Fabbretti, Daniela, Virginia Volterra & Clotilde Pontecorvo. 1998. "Written Language Abilities in Deaf Italians". Journal of Deaf Studies and Deaf Education 3.231- 244. Hanson, Vicki & David Wilkenfeld. 1985. "Morphophonology and lexical organization in deaf readers". Language and speech 28.269-280. Kelly, Leonard P. 1993. "Recall of English function words and inflections by skilled and average deaf readers". American Annals of the Deaf 138.288-296. Maxwell, Madeline M. & Tracey Gordon Falick. 1992. "Cohesion & quality in deaf & hearing children's written English". Sign Language Studies 77.345-372. Paul, Peter V. & Stephen P. Quigley. 1994. Language and Deafness. San Diego: Sin­ gular Publishing Group. Quigley, Stephen P., D. Power & M. Steinkamp. 1977. "The language structure of deaf children". Volta Review 79.73-83. Radelli, Bruna. 1998. Nicola vuole le virgole. Dialoghi con i sordi. Padova: DecibelZanichelli. Savy, Renata. 2000. "Riduzioni foniche nella morfologia del sintagma nominale nel parlato spontaneo. Indagine quantitativa e aspetti strutturali". Fonología e Morfo­ logia dell'italiano e dei dialetti d'ltalia ed. by Paola Benincà, Alberto Mioni & Laura Vanelli, 201-221. Roma: Bulzoni. Simone, Raffaele. 1988. "Fragilità delia morfologia e contesti turbati". L'italiano tra le altre lingue: strategie di acquisizione ed. by Anna Giacalone Ramat. Bologna: I1 Mulino. Spears, Arthur K. & Donald Winford. 1997. The Structure and Status of Pidgins and Creoles. Amsterdam & Philadelphia: John Benjamins. Taeschner, Traute, Antonella Devescovi & Virginia Volterra. 1988. "Affixes and function words in the written language of deaf children". Applied Psycholinguistics 9.385-401. Volterra, Virginia & Elisabeth Bates. 1989. "Selective impairment of Italian gram­ matical morphology in the congenitally deaf: A case study". Cognitive Neuropsy­ chology 6. 273-308.

THE EXPLAN THEORY OF FLUENCY CONTROL APPLIED TO THE DIAGNOSIS OF STUTTERING* PETER HOWELL & JAMES AU-YEUNG University College London 1. Introduction In the following two chapters, we present an overview of the EXPLAN theory. The main focus of EXPLAN is on fluent speech control, but it is also relevant to the diagnosis and treatment of stuttering.1 Fluency failures can af­ fect normally fluent as well as stuttered speech: it occurs where speech control falters even though the speaker does not produce an overt error. An extract in­ cluding several types of fluency failures (detailed later) is "I got on, on the seven ... fffifty ... three train t.to Mac.clesfield". According to EXPLAN, all fluency failures arise from the same generic problem. This problem leads to specific types of fluency failures that are shown by all children. The difference between fluency failures in child and adult stuttering is due to a change in the way the underlying problem is dealt with by people who stutter. The EXPLAN theory offers an account of how differences between fluent and struttered speech arise. This change leads to fluency failures with different characteristic surface (speech output) forms. As all speakers experience the same generic problem, they all could, in principle, tackle the problem in the same way as adults who stutter. According to EXPLAN, the distinguishing characteristics of stuttering emerge in late childhood; a contrasting view is that people who stutter have incipiently different problems in controlling speech from the onset of language. Viewed from the EXPLAN perspective, then, diagnosis of the disorder involves specifying and measuring the types of fluency failures that occur in late adolescence to see if they have changed in a way commensurate with adult stuttering, EXPLAN supports the view that young This work was supported by the Wellcome Trust. Thanks to Scott Yaruss for reading a preliminary version of this chapter. 1 On fluent speech control, see Howell, Rosen, Hannigan & Rustin 2000; Howell & Sackin, submitted. On diagnosis of stuttering, see Au-Yeung, Howell & Pilgrim 1998; Howell, AuYeung, Davis, Charles, Sackin, Williams, Cook, Rustin & Reed, in press; Howell, Au-Yeung & Sackin 1999. On treatment of stuttering, see Howell, Au-Yeung & Vallejo-Gomez, in press; Howell & Sackin, in press.

76

PETER HOWELL & JAMES AU-YEUNG

speakers whom clinicians suspect might develop the disorder need to be monitored. However, the theory suggests that early intervention may not be advisable in such suspected cases (not as long as the child is exhibiting the types of fluency failure common to all children). We will begin by examining how fluency control has been explained in a representative psycholinguistic model (Levelt 1989) that contrasts in many respects with EXPLAN. Levelt's model was selected for this comparison because it attempts to explain all events between intention and action in language. The points made about Levelt's (1989) model also apply to more recent versions by the same author (e.g., Levelt, Roelofs & Meyer 1999). A final reason for choosing Levelt's (1989) model is that it has also been the basis of a model of adult stuttering (the Covert Repair Hypothesis or CRH [Kolk & Postma 1997]). Several implications consistent with EXPLAN are discussed. The problem of stuttering is not expected to be associated with problems of lexical retrieval but, rather, to be dependent on the complexity of the phonological form of words. 2. A psycholinguistic model of language production A summarizing diagram of Levelt's model is given in Figure 1. The lefthand side is the route for language production that goes from the conceptualizer through the formulator to the articulators. The speech compre­ hension system is on the right.

overt speech

Fig. 1: A sketch of Levelt's (1989) model

EXPLAN APPLIED TO THE DIAGNOSIS OF STUTTERING

77

The conceptualizer is responsible for generating a message and monitoring that it is delivered appropriately. Monitoring devices like the one included in the conceptualizer take the output of a process and compare it with what the process intended to produce (the initial input to the conceptualizer). If there is a discrepancy (an error), an adjustment is made to the output to reduce the er­ ror. Feedback is a route by which output is returned to the monitor to be used for comparison with the original intention. Thus the presence of a feedback route implies that a monitoring process is involved. After the message is output from the conceptualizer, it goes through the formulation stage. At the output of the formulation stage, the message is repre­ sented as a phonemic string. In Levelt's model as in many others (Dell 1986; Dell & O'Seaghdha 1992), two sub-stages are identified where the message is represented as a lemma and in phonological forms, respectively. Two different patterns of speech errors occur that have been associated with these two for­ mulation levels: word exchange errors that could arise at the lemma level, and sub-word level speech errors (Dell 1986) that involve phonemic or syllabic transmutations and may arise at the phonemic level. Recently, Caramazza & Miozzo (1997) and Miozzo & Caramazza (1997) have provided evidence on the tip-of-the tongue phenomenon (Brown & McNeill 1966) that they interpret as indicating there is only one stage in formulation. For EXPLAN, it does not matter whether one or two sub-levels are involved in formulation. What is important is that the output of formulation is a phonemic plan generated from left to right. This allows the plan to be available for the first part of the word before the rest of it. Word onset is the point at which most sub-word speech errors occur and is also the point at which speakers sometimes find it difficult to span from word onset to the subsequent part of the word during fluency failures. Articulation is the final step in translating the abstract phonemic represen­ tations to overt speech. In Levelt's model, speech production and perception systems are coupled together on the assumption that audition can recover overt speech from articulatory output and process it through the speech comprehension system (the external loop that will be considered in the following chapter). Information from the speech comprehension system is then transmitted back to the monitor in the conceptualizer. In this chapter, we will only consider internal monitoring as we concentrate exclusively on the production route (the left-hand side of Figure 1, above).

78

PETER HOWELL & JAMES AU-YEUNG

3. Speech repairs The monitor in the conceptualizer is an essential component in Levelt's model. (A similar monitor also appears in Levelt et al.'s more recent 1999 model). Monitors, in general, are rejected by EXPLAN because of the interactions between high and low levels of processing they engender. Such interactions place unrealistic processing demands on language output in terms of the representations that need to be available and the amount of time needed for processing these representations during on-going speech control. We will evaluate Levelt's (1983) work on speech repairs to assess one line of support for the internal monitoring process. Levelt regards structures like "Turn left at the, no, turn right at the crossroads" as instances of self cor­ rection of errors after they have been detected by internal monitoring (termed repairs). When errors are detected by the monitor, speech can be interrupted at any point between conceptualization and articulation and the message can be corrected and restarted. To illustrate with the example used earlier, the speaker gives the wrong direction, realizes this, and replaces the reparandum ("left") with the alteration ("right"). The substitution is not the only thing that happens: The speaker overshoots the reparandum (goes on with the words "at the") and retraces to "turn" when the message recommences (the word said before "left" that is not incorrect). The repair contains an interruption (which can range from a pause to a short phrase; here the word "no"). Most of these parsed components are optional. Levelt considers that speakers can detect and repair errors before they are overtly spoken. These are called covert repairs and an example is "Turn right at the, at the crossroads". Such repairs are problematic to interpret as there is no outward sign of what the error was (i.e., why the speaker hesitated after "turn"), nor even whether an error occurred at all. Though Levelt does not rely on covert repairs in his account of speech control, Kolk & Postma (1997) have described stuttering as a result of covert repairs. Other types of repairs are Different (D) repairs, in which the topic is changed (like non sequirurs), and repairs in which the speech is not pitched at an appropriate level for the addressee (appropriateness repairs). 4. Modelling errors versus fluency failure Basing a model on fluency failures (episodes where speech control falters, but there is no overt error) stands in marked contrast with basing one on error patterns using monitoring processes. It is not denied that errors provide im­ portant information about points of difficulty in a message, but their explana­ tion by a monitoring process is questionable. First of all, errors are infrequent events in language (Shallice & Butterworth 1977). Fluency failures, on the

EXPLAN APPLIED TO THE DIAGNOSIS OF STUTTERING

79

other hand, are common. For instance, estimates from the data of Howell, AuYeung & Sackin (1999) show that fluent speakers produce around 2.57% fluency failures in function words and 0.97% in content words. While a case can be made that it is important to model error patterns, it also seems advisable, at the very least, to consider how the most frequent breakdowns in fluency control could arise. Next, we consider whether errors need explanation by an internal moni­ toring process. Monitoring all speech for errors would be inefficient, given that errors are infrequent. However, this does not necessarily mean that monitoring should be dismissed. We next examine more critical points about the operation of a monitor once an error has been detected. The monitor takes corrective action when a difference is detected between intended and actual versions (the actual version assumed to be in error). To establish whether a difference has occurred or not, like needs to be compared with like, so it must be assumed that the monitor has multiple intended representations equivalent to the actual forms produced at all stages during the output of a message. One problem is that it is not apparent how an errorless version of the intended form at lower levels can be available in the conceptualizer for comparison when an erroneous output form is generated as actual output at these lower levels. If a correct version is available for monitoring, why not use it for controlling the respective level? Another look at the repairs, described earlier from Levelt's perspective, reveals that these actually provide little support for an internal errormonitoring process. Evidence that an error has occurred is only obtained when speech output reveals this. However, this selfsame evidence for an error would suggest either that no monitoring takes place or that the monitor has not worked on this occasion as the speech goes right through to output. The only cases Levelt describes where there may be evidence for the operation of an internal monitor is where errors are intercepted and the result is a covert repair. However, in these repairs another interpretation is that no error occurred in the first place (see the later discussion of stalling fluency failures for an alternative explanation of some surface form features considered to represent components of repairs). In the case of D repairs, the message is abandoned rather than repaired, and appropriateness repairs are a matter of style rather than anything else. Abandonment and restart of a message (as in D repairs) may be a general process applying to all repairs rather than arising from internal monitoring for errors and on-line repair to remove any detected discrepancy. In the abandon­ ment conception, the new message would not be an on-going alteration to the

80

PETER HOWELL & JAMES AU-YEUNG

current message, but would be regarded as a new attempt altogether. This view would not preclude there being a large measure of overlap (overshoot and retrace) between the initial and subsequent messages, as is observed to occur. The abandonment view would also explain why no error is evident in several types of repair and no definitive evidence of errors is available at all until they are translated into concrete articulatory output form. 5. Stutterings and repairs Stuttered speech is often described as fluent stretches that are interspersed with dysfluent events that we term fluency failures. Johnsonet al. (1959) listed eight types of events he considered to be often associated with stuttering, and these are given in Table 1. Stalling

Advancing

1. Word repetitions (e.g., "Katy, Katy"). 2. Phrase repetitions (e.g., "if he, if he").

4. Part-word repetitions (e.g., "k..k..Katy"). 5. Prolongations (e.g., "mmother").

3. Interjections (e.g., editing terms like "uh"). This can include filled and abnormally long silent pauses too.

6. "Other" fluency failures such as broken words (e.g., "di.nosaur"). 7. Idea abandonment (like non sequiturs) 8. Phrase revisions as in "my uunc, my mother's brother".

Table 1 : Stuttering events used by Johnston et al. (1959)

It is apparent that the form of these events is closely related to that of repairs and, as with repairs, none of the stuttering events involve errors at the phonemic level or in articulation (for example, none involve using the wrong phone to start a word). We term events 1-3 stalling fluency failures. These involve whole words (the ones said immediately prior to the point of interruption) being reiterated (types 1 and 2), or interruption by silence or filler word(s) (type 3). We call events 4-6 advancing fluency failures which are, respectively, where the first part of a word is repeated (type 4), is elongated (type 5) or has a gap before the remainder of the word is produced (type 6). The overt speech appears to show that only the first part of the utterance was available (types 4-6). These events could be a consequence of the formulated plan not being complete but, nevertheless, the speaker attempts what is available (the first part), EXPLAN, as described later, develops this idea. The remaining two categories in Table 1 are similar to some classes of re­ pairs Levelt describes. Category 7 events are like D repairs, but, as said earlier, the speaker has not repaired the message, but abandoned it and started another one. The phrase revision, category 8, could be described as an error

EXPLAN APPLIED TO THE DIAGNOSIS OF STUTTERING

81

repair involving lexical selection, as in the "turn left, no turn right" example discussed above. In cases like that and their stuttering equivalents (type 8), we would agree that a high level error has occurred, but, even so, we would not interpret them as showing that the error has been repaired by an internal monitoring process, as the error has not been intercepted. Type 7 and 8 stuttering events can be removed by the procedure, described in Howell, AuYeung, Sackin, Glenn & Rustin (1997), for selecting only fluency failures that do not involve a phonemic or lower-level error. To summarize, EXPLAN regards the surface form of stuttering events (rather than putative errors that may lie behind them, according to Kolk and Postma's CRH) as reflecting two alternative forms of fluency failure (stalling versus advancing) in the vicinity of a word. The EXPLAN view that all fluency failures arise at points where planning cannot keep up with output speed is now considered. 6. The EXPLAN theory of stuttering According to EXPLAN, the fluency failures just described arise because segments take different amounts of time to plan and execute. The linguistic formulator process generates a plan (PLAN), and motor processes execute it (EX). The model assumes that PLAN and EX take place in parallel and that PLAN is independent of EX. This independence precludes feedback loops from EX output to a central monitor that could be used to correct and reinitiate a PLAN. The independence also allows the current word to be executed, while the plan of one to be produced subsequently is being generated. Planning can be put under time pressure when a segment that is difficult and, therefore, time-consuming to generate has to be prepared, and this plan is required quickly as when the planned segment follows a word that is executed rapidly. Though ordinarily the planning process is rapid so that most of the time a plan is delivered ahead of execution (Sternberg, Monsell, Knoll & Wright 1978), in the circumstances just described, the process can fail to deliver the complete plan on time. If this happens, only the left-most part of the plan is ready at the point where execution of the segment should start. Speech can be delayed until the remainder of the plan arrives (stalling) or the speaker can continue and attempt the speech with the partial plan (advancing). Dividing words into function and content types provides a rough basis for characterizing planning time differences of English words. Function words are simpler than content words and require less planning time. Howell, Au-Yeung & Sackin's (1999) transcriptions of samples of spontaneous English speech were analysed to show this (these data include speech from several different age

82

PETER HOWELL & JAMES AU-YEUNG

groups of speakers who stutter). 94% of the function words used were monosyl­ labic and 84% of them carried no primary stress. Among the content words, 62% were monosyllabic and 23% of them carried no primary stress. 88% of the content words started with consonants, whereas only 54% of the function words did. 14%) of the content words started with consonant strings, while only 1% of the function words did. 35% of the content words had a developmentally lateemerging consonant at the start, as defined by Sander (1972), whereas this ap­ plied to only 21% of the function words. The consonant string and later emerging consonant properties co-occurred on the onset of 1% of the function words, but on 12% of the content words. These properties remain roughly con­ stant across normally fluent people and those who stutter, and across ages 2 to 40 years. In sum, then, function words are phonologically simpler than content words, with the onset of content words posing particular difficulties. Parts of wrords that are phonologically difficult do not necessarily present problems in themselves. The difficult word parts are a particular problem when they occur in a message context that itself adds extra pressure on the time at which plans for words must be formulated ready for execution. To demonstrate this, the contexts that cause difficulty when leading up to a word with a phonologically complex onset are described. The segments that can be executed rapidly are those that are simple to plan. There may not be sufficient time during the execution of such simple segments to plan a subsequent word particularly when the later word has a complex focus on it. To be concrete, such demanding contexts are likely to occur at the juncture between a simple function word (executed rapidly) and a content word that starts with a complex phonological structure (planned slowly). In effect, the short execution time for the prior function words puts an extra strain on the planning time of the content word that cannot be met in the short time allowed during the execution of the simple function word. The preceding discussion shows that the function/content word distinction and the context in which these words are combined offer an approximate esti­ mate of regions where fluency may fail in English. Two words of caution about using function and content words are necessary. First, though lexical word type has been discussed, the important property leading to fluency failure is that a segment requiring a long planning time occurs after a segment or segments requiring a short execution time. This could equally well be de­ scribed at the phonological or motorlevels (although the superordinate relation between length of execution and planning times would still need to apply). Second, the situation described for English would not operate in other lan­ guages. In Romance languages, for instance, function words can be long and

EXPLAN APPLIED TO THE DIAGNOSIS OF STUTTERING

83

nearly as complex as content words. Comparison across languages like English and, for example, Spanish potentially allows linguistic and motor involvement in this process to be distinguished. Distinguishing planning and execution has implications for control both where speech is fluent and where fluency fails. Figure 2 shows speech that is controlled fluently. Time runs horizontally with planning shown at the top and execution underneath. The message starts when planning the first word com­ mences (word n). At this time there is nothing to execute. Once PLAN(n) is ready, execution of word n can begin. EX(n) and PLAN(n+l) commences. When EX(n) has been completed, PLAN(n+l) is ready and its execution can start, and the whole process repeats cyclically. PLAN(n) has been completed earlier and is just retrieved for execution at this point (planning of subsequent segments is proceeding independently). Note that speech control does not involve an ex­ ternal timekeeper (it involves what Fowler 1980 would call an intrinsic timing process). The situation that can arise in the contexts described at the beginning of this section is shown at the point in time that EX(n+l) has been completed in Figure 2, of timing patterns (horizontal axis) for PLAN (top line) and EX (bottom line) of serially arrayed words (n, n+1 ...) during fluent speech. If comparison is made vertically, the plan for all words (e.g., word n+1) is always complete prior to when the execution of a preceding word is finished (e.g., word n) so that speech can proceed fluently. The PLAN(n+2) is not complete. The speaker cannot embark on the execution of word (n+2) with a complete plan. PLAN(n)

PLAN (n+1)

PLAN (n+2)

EX(n) 1

\\

EX(n+l)

i

Fig. 2: Timing patterns (horizontal axis) for PLAN (top line) and EX (bottom line)

When the plan is not complete, the message cannot progress fluently. In this situation, the speaker can retrieve the plan of a word recently used and execute it again (Blackmer & Mitton 1991). The word usually selected is the previous (function) word, which leads to the often observed function word repetition prior to a content word (the first type of stalling fluency failure given in Table 1). As planning takes place in parallel with execution, the planning of the content word can continue unaffected during this repetition. After the function word (word n) has been repeated, the plan for the content word n+1 may be complete and, if so, execution can proceed.

84

PETER HOWELL & JAMES AU-YEUNG

In Figure 3 the two lines indicate what words are planned and executed (respectively). At the start, planning and execution are operating appropriately on the first word (word n). After word n is first completed, the plan for n+1 is not ready (its planning continues). In this case, the speaker retrieves plan n and executes that again (leading to a word repetition), at which point the plan for word n+1 is ready and can be completed fluently. A pause introduced at the function-content word juncture or a phrase repe­ tition (which are the other two types of stalling fluency failure in Table 1) would serve the same role of gaining time so the plan can be completed. The consequences of repeating a function word that precedes a content word is to slow message rate down (where 'message' corresponds to Levelt et al.'s (1999) usage as the technical term for the conceptual, rather than the actual, output structure that is ultimately going to be formulated). More specifically, function word repetition leads to a message, but not word, rate adjustment at this point. Speech timing has been controlled within the hand-shaking process and so speech continues to be intrinsically timed (Fowler 1980). Stalling types of fluency failures are not problematic, as the speaker successfully avoided having to commence a word on the basis of a plan that was only partly pre­ pared. The next type of fluency failure is problematic insofar as the speaker does not avoid this situation. PLAN(n)

PLAN(n+l)

EX(n)

EX(n)

EX(n+l)

fig. 3: Scnemanc representation oj a word repetition fluency failure

In this second type of fluency failure, the speaker commences the execu­ tion of a problem (usually content) word when only the first part of the plan is available. Since the generation of the remainder of the plan continues while the first part is being executed, the plan may be completed by the time this part is required for execution, so speech will appear fluent. However, the plan may not be completed and will run out of input as it is being executed. The speaker then only has the beginning of the content word available to work with, and fluency characterised by elongation, repetition or interruption of the first part of the word occurs. EXPLAN regards the advancing fluency failure types 4-6 as cases where this happens. A schematic representation of this situation is given in Figure 4, where, at the start, planning and execution are operating appropriately on the first word in the sequence shown (word n).

EXPLAN APPLIED TO THE DIAGNOSIS OF STUTTERING

85

After word n is completed on the first occasion after planning, the plan for word n+1 is not ready. Even so, the speaker executes the part plan of word n+1. Fluency failure involving parts of words (due to advancing) is problematic, as it is a characteristic of persistent stuttering (Conture 1982). On a priori grounds, it also seems risky to attempt to produce words in this manner. The next chapter describes a simple, biologically plausible check that suffices to identify points where failure has occurred so that this problem can be avoided. This has the result that it changes timing control from an intrinsic to an extrinsic mode. PLAN(n)

PLAN(n+l)

EX(n)

EX(part(n+l))

Fig. 4: A stretch of fluent speech

The EXPLAN model stresses that motor levels are as important as the lin­ guistic planning levels in leading to fluency failure. Fluency failures are not errors and do not involve a monitoring process. Stalling fluency failures and advancing fluency failures both arise as a consequence of the plan not being available in time. No fluency group is precluded from producing any type of fluency failure. However, adult stuttering is characterised by frequent use of advancing fluency failures rather than stalling ones, so this type becomes ha­ bitual and, once acquired, is very hard to lose. Note that though adult stuttering is not easy to treat, it should be reversible, as people who stutter do not have defective brain or articulatory structures. 7. Some lines of evidence consistent with EXPLAN The contexts in which fluency is likely to fail are when EX(n) is short and PLAN(n+l) is long, particularly when PLAN(n+l) starts with a phonologically complex structure. (Problems in these contexts happen to all speakers, not just people who stutter.) Several implications of this view are addressed in this section. First, the problem of stuttering is not expected to be associated with problems of lexical retrieval but, rather, to be dependent on the complexity of the phonological form of words. There is evidence that children (whether or not they stutter) do not have differential problems in lexical retrieval. The re­ maining lines of support are derived from an examination of the contexts

86

PETER HOWELL & JAMES AU-YEUNG

mentioned above. To locate such contexts, extensive use is made of phonological words (Selkirk 1984; see below). The important property of phonological words is that they group a single content word with its adjacent function words. For English, this serves the role of differentiating (approximately) short simple (function) words from long complex (content) words with difficult initial phonemes. In sub-section two, phonological words allow predictions about fluency failure on function words in different positions relative to their content word. Sub-section three uses phonological words to test the prediction that people who stutter change from stalling to advancing fluency failures in late adolescence. The fourth sub-section looks at properties of content words to establish whether those with complex onsets are more likely to lead to fluency failure. Planning problems are going to be exaggerated when speaking at a rapid rate. Analyses of spontaneous speech are reported in sub-section five that establish that those stretches spoken rapidly are more likely to lead to fluency failure. According to EXPLAN, repetition of function word segments is only one way of stalling. In the sixth and final sub-section, evidence that suggests function word repetition and pausing have equivalent roles is reviewed. 7.1 The role of lexical retrieval in fluency failure In the tip-of-the-tongue () phenomenon (Brown & McNeill 1966), speakers often report being in a feeling of knowing state, but cannot retrieve the actual word for articulation. In laboratory studies, it has been shown that speakers in these TOT states, can recall the first sound of the word and the number of syllables in the word reliably and at greater than chance levels. The consensus is that, in TOT, the concept for the word(s) is available, but the com­ plete phonological form of the word is not (see Figure 1). TOT states experienced during spontaneous speech may reflect the same problem as we have described that lies behind fluency failures (i.e., all the speech plan is not available for execution, only its onset is ready). Thus, the onset information shown to be available in TOT states corresponds with the information people appear to have available in advancing fluency failures. Recent work in our laboratory has shown that TOT states occur with the same frequency in age-matched groups of fluent speakers and speakers who stutter. This suggests that, if the TOT state reflects lexical retrieval problems allied to fluency failures, then lexical retrieval failure is no more frequent in people who stutter than in fluent speakers. The fluency failure problem in peo­ ple who stutter would appear to arise, as EXPLAN predicts, because speakers who stutter go on to translate this partial information into the problematic out­ put surface forms (advancing fluency failures).

EXPLAN APPLIED TO THE DIAGNOSIS OF STUTTERING

87

7.2 Phonological word contexts, and stalling and advancing fluency failures Function word position relative to a content word needs to be established objectively to assess the stalling role of function word repetitions. The reason this is necessary is apparent in an utterance like "I look after my mother". Here there are two function words between the content words "look" and "mother". So does one or both of the function words occur before "mother" (and, consequently could it be used to stall) or do they appear after "looK" (when they cannot be used to stall production of "look")? Our research group has developed an analysis procedure based on Selkirk's (1984) phonological words to establish function word position (Au-Yeung & Howell 1998; AuYeung et al. 1998 and Howell, Au-Yeung & Sackin 1999). Phonological words consist of a single content word as its nucleus and any number of function words that serve as prefixes or suffixes to the content word (F n CF m , where n and m can be zero or a positive integer in value). Function words are then associated with content words by sense rules and their position can be determined. The application of this procedure to the preceding example reveals that "after" is part of the phonological word that includes "look" and cannot be used to delay production of the content word, as it occurs in final position in its phonological word, while "my" is part of the phonological word that includes "mother". In this case, as "my" appears prior to "mother", it can be used to delay production of the content word. Au-Yeung et al. (1998) used speech segmented into phonological words to show that the majority of stuttering on function words occurs when the function word appears prior to a content word (e.g., "my" in "my mother") with little stuttering on function words that appear after a content word ("after" in "I look after"). The fact that the likelihood of fluency failure on a function word depends to a marked extent on whether these words appear in initial position in a phonological word (Au-Yeung et al. 1998) is consistent with the view that these words are used to delay production of the content word. 7.3 Function word and content wordfluency failures It was argued above that people who stutter change from producing stalling fluency failures to advancing ones in their teenage years, whereas fluent speakers do not. Phonological words can be used to trace the change from stalling to advancing fluency failures across age groups. Fluency failures on initial function words within phonological words are mainly stalling, whereas advancing fluency failures occur predominantly on content words. Consequently, the relationship between fluency failure rate on function and content words over age groups can be used to establish changes from stalling

88

PETER HOWELL & JAMES AU-YEUNG

(young) to advancing (older) fluency failures in people who stutter (control data were also obtained for fluent speakers in the same age groups). To examine the change from fluency failure on function to content words, phonological words of the form F n CF m , (where the integers n > 0 and m >= 0) were selected so that the relationship between stalling on phonological wordinitial function words and advancing on content words could be investigated. For both fluent speakers and speakers who stutter, very few fluency failures occurred on function words that occupied a position after a content word. For both fluency groups, fluency failure within each phonological word occurred predominantly on either the function word preceding the content word or the content word itself, but not both. Fluent speakers showed, statistically speaking, constant rates of fluency failure on function and content words across age groups, with higher rates on function words. The data for speakers who stutter are shown in Figure 5. 25 T

>17 Age Graup (in year) - Content Word

 Function Word

Fig. 5: Fluency failure rates in groups of stuttering speakers who differ in age

Fluency failure rates are shown separately for function and content words. Function word fluency failures were significantly higher in the two youngest age groups (2-6 and 7-9 years), but were not significantly different in older age groups. This shows the change from function to content word fluency failures (stalling to advancing) as age increases. These data support the view that as speakers who stutter get older, they stop stalling when the plan of a

EXPLAN APPLIED TO THE DIAGNOSIS OF STUTTERING

89

word is not complete, and instead advance and attempt a content word on the basis of an incomplete plan. Similar exchange patterns have been found in German (Dworzynski 1999; Rommel 2000). 7.4 The role of phonological difficulty in fluency failure The issue of word onset difficulty was addressed by Howell, Au-Yeung & Sackin (2000) in a study based on earlier work by Throneburg, Yairi & Padden (1994). The latter authors examined how stuttering on words was affected by phonological properties of their constituent consonants. The consonantal factors examined were: a) whether the word contained a developmentally Late Emerging Consonant (LEC, as defined by Sander 1972); b) whether the word contained a Consonant String (cs) and c) whether the word contained more than one syllable (MS). CS and LEC can occur anywhere in a word. So, for instance, a word could have a simple CS in initial position and an LEC in word final position (e.g., quiz). Throneburg et al. (1994) analysed CS and LEC factors irrespective of the word-position they occupied. EXPLAN suggests that CS and LEC should be more of a problem when they cooccur at word onset than when either occurs alone or when they occur at different word positions because when they co-occur they then require a long planning time. If the word containing CS and LEC is preceded by a simple function word, there may not be sufficient time to complete the plan before it is needed for execution. Howell, Au-Yeung & Sackin (2000) found that when both CS and LEC occurred at the beginning of a content word, stuttering rate on the content word was particularly high compared with that on words that did not contain these factors or where the factors were spread throughout the word. In addition, if phonological words are divided into classes based on whether their content word contains a CS and an LEC in initial position, the difficult stretches of speech can be identified to some extent, as these phonological words are the ones that cause most fluency failure. 7.5 Rate change as a general result of stalling and advancing fluency failures Howell, Au-Yeung & Pilgrim (1999) reported a study designed to examine whether a rapid speech rate in short stretches of speech increases the chance of fluency failure. (It is assumed that the onsets of difficult content words, discussed in the fourth sub-section are randomly distributed across all the stretches so that rate alone can be studied.) Howell et al. (1999) located all tone units in speech and measured speech rate on the fluently produced and stuttered words in each tone unit. They then divided the tone units in which the initial words were fluent, as to whether they were spoken at slow,

90

PETER HOWELL & JAMES AU-YEUNG

medium, and fast rates. They found that speech rate of the fluent words correlated with fluency failure rate, with the most rapid rate of speech being associated with the highest rate of fluency failure. This suggested a local effect of speech rate in spontaneous production that was linked to fluency failure. This finding also carries an implication for treatment: although a global change can lead to improvement in fluency, a local change in these regions should be sufficient to bring about such improvements (see the next chapter). A second prediction about the relationship between speech rate and fluency is that fluent speakers will produce stuttering-like fluency failures if speech rate goes up and speech has to continue, as this will increase the chance of execution getting ahead of planning. An interpretation of a study by Blackmer & Mitton (1991) offers some support for this hypothesis. These authors analysed speech errors and repairs of fluent speakers recorded from late night radio chat shows. They found that fluency breakdown occurred when the speakers were talking rapidly. They reported that this often involved word repetition, which they explained as re-issuing a plan already available (this would be similar to the role we attribute to stalling fluency failures, assuming that the repeated words were function words). Given that Blackmer and Mitton were not in control of the speaking situa­ tion, obviously they had no way of getting the speakers to advance the message at the points where fluency failures occurred. Howell & Sackin (2000) examined this question using a task in which fluent subjects had to provide a running commentary on the actions on a cartoon video they were watching (without the soundtrack). As the speakers were under pressure to continue to keep up with the video action, they tended not to repeat function words, it was found that these speakers produced advancing fluency failures on content words. 7.6 Other ways of stalling The findings of Beattie and Bradbury's study of the temporal structure of speech, support the view that word repetition and pausing act equivalently to slow down speech rate. Beattie and Bradbury reduced pauses in fluent speakers by using an operant procedure while requiring the speakers to maintain a similar articulation rate to that without the procedure. The speakers then produced more word repetitions. This suggests that word repetition has an equivalent role to pausing. Unfortunately, the authors did not report what word type (function or content) was repeated, nor whether the repetitions were of parts of words, words or groups of words. Howell and Sackin (in press) partially replicated Beattie and Bradbury's study and extended the analysis to

EXPLAN APPLIED TO THE DIAGNOSIS OF STUTTERING

91

show that when pauses were conditioned out, it was in fact function words that were repeated. Slowing speech by introducing pauses or encouraging function word repe­ tition (see the next chapter) would be expected to enhance the fluency of peo­ ple who stutter, It is noteworthy in this connection that the strategy of inserting filled pauses has been proposed as a way of reducing stuttering (Christenfield 1996). 8.

Conclusions We consider that our experimental investigations may have implications for the practical purpose of diagnosing stuttering. In our research database, we have long-term recordings of eleven children who stutter, who were first seen when they were in the 7-9 year age group and are now in the 10-12 years group. All these children had been admitted to an intensive speech therapy course and the first recording we have was a sample obtained immediately prior to that course. Besides this sample and several follow-up recordings, the children were assessed using therapists' own criteria and researchers' ratings, and on the basis of parents' and children's attitude to communication as re­ vealed in questionnaires. (These criteria were made independently of the speech samples.) In most cases, all the non-speech assessments agreed about whether a child had recovered or not (in the remaining cases there was always a majority in agreement). One classification obtained was based on these scores (recover/persist). After this categorisation had been made, the speech of each child was examined to see what proportion of fluency failure on function (stalling) and content (advancing) word was shown on initial attendance at clinic. The fluency failure rates, provided separately for function and content words, are shown in Figure 6. An 'R' above a child's scores indicates that the independent non-speech measures classified the child as recovered. The children who recovered showed a much higher proportion of function word than content word fluency failures. The recovered cases predicted by applying a criterion of a ratio of function to content word fluency failures of two or greater are indicated with a ' * '. Even this crude measure predicts two thirds (4/6) of the cases of recovery with only a single false alarm in the five recovered cases. This classification procedure appears to offer considerable promise bearing in mind that the analysis was based on a sample of speech obtained before the children received any treatment, and the analysis is limited, as it does not take account of differences in speech rate across these speakers (that is also operative in stuttering

92

PETER HOWELL & JAMES AU-YEUNG

according to EXPLAN). The question that remains to be addressed (see the next chapter) is whether therapies that lead to speech changes that EXPLAN predicts will improve fluency (rate slowing globally or locally, change in type of stuttering pattern) are an effective treatment for stuttering.

Function Word

Conte nt Wond

Fig. 6: Fluency failure rates on function and content words on first attendance at clinic

REFERENCES Au-Yeung, James & Peter Howell. 1998. "Lexical and syntactic context and stuttering". Clinical Linguistics and Phonetics 12.67-78. Au-Yeung, Peter Howell & Lesley Pilgrim. 1998. "Phonological words and stuttering on function words". Journal of Speech, Language, and Hearing Research 41.1019-1030. Beattie, Geoffrey W. & R.J. Bradbury. 1979. "An experimental investigation of the modifiability of the temporal structure of spontaneous speech". Journal of Psy­ cholinguistic Research 8.225-248. Blackmer, Elizabeth R. & Janet L. Mitton. 1991. "Theories of monitoring and the timing of repairs in spontaneous speech". Cognition 39.173-194. Brown, Roger & David McNeill. 1966. "The 'Tip of the Tongue' phenomenon". Journal of Verbal Learning and Verbal Behavior 5.325-337. Caramazza, Alfonso & Michele Miozzo 1997. "The relation between syntactic and phonological knowledge in lexical access: evidence from the 'Tip-of-the-tongue' phenomenon". Cognition 64.309-343.

EXPLAN APPLIED TO THE DIAGNOSIS OF STUTTERING

93

Christenfeld, Nicholas. 1996. "Effects of a metronome on the filled pauses of fluent speakers". Journal of Speech and Hearing Research 39.1232-1238. Conture, Edward G. 1982. Stuttering. Englewood Cliffs, N.J.: Prentice-Hall. Dell, Gary S. 1986. "A spreading-activation theory of retrieval in sentence pro­ duction". Psychological Review 93.283-321. Dell, Gary S. & Padraig G. O'Seaghdha. 1992. "Stages of lexical access in language production". Cognition 42.287-314. Dworzynski, Katharina. 1999. Locating where Fluency Breaks down using the Phonological Word: A cross-linguistic study using speech samples of English and German stutterers. Unpublished MSc dissertation, University College London. Fowler, Carol A. 1980. "Coarticulation and theories of extrinsic timing". Journal of Phonetics 8.113-133. Howell, Peter, James Au-Yeung, Steve Davis, Nicole Charles, Stevie Sackin, Roberta Williams, Francis Cook, Lena Rustin & Phil Reed. In press. "Factors implicated in the diagnosis and prognosis of children who stutter". Proceedings of the Oxford Dysfluency Conference 1999. Howell, Peter, James Au-Yeung & Isabel Vallejo-Gomez. In press. "Why does the position of stuttering change during language development?". Proceedings of the Child Language Seminar, City University, September 1999. Howell, Peter, James Au-Yeung & Lesley Pilgrim. 1999. "Utterance rate and lin­ guistic properties as determinants of speech dysfluency in children who stutter". Journal of the Acoustical Society of America 105.481-490. Howell, Peter, James Au-Yeung & Stevie Sackin. 1999. "Exchange of stuttering from function words to content words with age". Journal of Speech, Language and Hearing Research 42.345-354. Howell, Peter, James Au-Yeung & Stevie Sackin. 2000. "Internal structure of content words leading to lifespan differences in phonological difficulty in stuttering". Journal of Fluency Disorders 25.1-20. Howell, Peter, James Au-Yeung, Stevie Sackin, Kazan Glenn & Lena Rustin. 1997. "Detection of supralexical dysfluencies in a text read by child stutterers". Journal of Fluency Disorders 22.299-307. Howell, Peter, Stuart Rosen, Geraldine Hannigan & Lena Rustin. 2000. "Deficits in auditory temporal resolution in children who stutter and its relation to dysfluency rate". Perceptual and Motor Skills 90.355-363. Howell, Peter & Stevie Sackin. 2000. "Speech rate manipulation and its effects on fluency reversal in children who stutter". Journal of Developmental and Physical Disabilities 12.291-315. Howell, Peter & Stevie Sackin submitted. "Timing interference to speech in altered listening conditions". Perception & Psychophysics. Howell, Peter & Stevie Sackin. In press. "Function word repetitions emerge when speakers are operantly conditioned to reduce frequency of silent pauses". Journal of Psycholinguis tic Research.

94

PETER HOWELL & JAMES AU-YEUNG

Johnson, Wendell & associates. 1959. The Onset of Stuttering. Minneapolis: Univer­ sity of Minnesota Press. Kolk, Herman & Albert Postma. 1997. "Stuttering as a covert repair phenomenon". The Nature and Treatment of Stuttering: New directions (2nd edition) ed. by Richard F. Curlee & Gerald M. Siegel, 182-203. Boston, Mass.: Allyn & Bacon. Levelt, Willem J. M. 1983. "Monitoring and self-repair in speech". Cognition 14.41-104. Levelt, Willem J. M. 1989. Speaking: From intention to articulation. Cambridge, Mass.: MIT Press. Levelt, Willem J. M., Ardi Roelofs & Antje S. Meyer. 1999. "A theory of lexical ac­ cess in speech production". Behavioral and Brain Sciences 22.1-75. Miozzo, Michele & Alfonso Caramazza. 1997. "Retrieval of lexical-syntactic fea­ tures in Tip-of-the-Tongue states". Journal of Experimental Psychology: Learning, Memory, and Cognition 23.1410-1423. Rommel, Dieter. 2000. "The influence of Psycholinguistic Variables on stuttering in childhood". Fluency disorders: Theory, research, treatment and self-help ed. by H.-G. Bosshardt, J.S. Yaruss & H.F.M. Peters. Nijmegen University Press. Sander, Eric K. 1972. "When are speech sounds learned?". Journal of Speech and Hearing Disorders 37.55-63. Selkirk, Elisabeth. 1984. Phonology and Syntax: The relation between sound and structure. Cambridge, Mass.: MIT Press. Shallice, Tim & Brian Butterworth. 1977. "Short-term memory impairment and spontaneous speech". Neuropsychologia 15.729-735. Sternberg, Saul, Stephen Monsell, Ronald L. Knoll & Charles E. Wright. 1978. "The latency and duration of rapid movement sequences: comparison of speech and typewriting". Information processing in motor control and learning ed. by George E. Stelmach, 118-152. New York: Academic Press. Throneburg, N. Rebecca, Ehud Yairi & Elaine P. Paden. 1994. "The relation between phonological difficulty and the occurrence of disfluencies in the early stage of stuttering". Journal of Speech and Hearing Research 37.504-509.

THE EXPLAN THEORY OF FLUENCY CONTROL APPLIED TO THE TREATMENT OF STUTTERING * PETER HOWELL University College London 1. Introduction In this chapter, the question of how speech can be manipulated (i.e., how fluent speech can be made to contain fluency failures and how stuttered speech can be changed to make it more fluent) is addressed with particular emphasis paid to the treatment of stuttering. Generally speaking, a treatment for stuttering is considered successful if it decreases the incidence of fluency failures.1 EXPLAN theory is more specific as it requires a decrease in the inci­ dence of fluency failures that involve production of parts of words (termed advancing fluency failures in the preceding chapter). The current chapter commences by describing the effects Altered Auditory Feedback (AAF) procedures have on fluent speakers and speakers who stutter, AAF manipulations are widely considered as affecting fluency in these speakers in different ways.2 Original explanations of the effects of AAF are then evaluated with particular emphasis on whether an auditory feedback monitor is required to explain their effects. The EXPLAN account of how AAF changes fluency in fluent and stuttering speakers is presented. This explanation, in contrast to those considered earlier in the chapter, does not involve a feedback monitoring process. A contrasting way of manipulating stuttering is then described (verbal operant procedure) and a way this could be explained within the EXPLAN framework is also considered. Finally, lines of

This work was supported by the Wellcome Trust. Thanks to Scott Yaruss for reading a preliminary version of this chapter. * It is recognised that in practical circumstances, treatments bring about various other important changes in the speaker's outlook. However, such effects cannot be used as an outcome measures until there is agreement as to what these changes are, and objective ways of measuring these attributes are available. 2 Inclusion of the word 'feedback' begs the question of how the findings should be explained. Existence of an auditory loop that can be disrupted by altered listening does not necessarily require a monitoring process.

96

PETER HOWELL

evidence that support the EXPLAN account of how speech output control is affected by various procedures are reviewed. 2. Altered listening conditions and auditory feedback monitoring A monitor is defined as a device that takes the output of a process, com­ pares it with what was intended to be produced and makes a correction to re­ duce any discrepancy detected. The auditory exteroceptive loop that features in auditory monitoring accounts (Fairbanks 1955; Lee 1950; Levelt 1989), suggests the speaker recovers speech output through the auditory system and uses this to determine whether there is a discrepancy between intention and output. The main source of evidence for an exteroceptive auditory loop is that alterations interposed between articulation and audition have dramatic effects on speech. Moreover, their effects are somewhat different for fluent speakers and speakers who stutter. Three types of alteration to the auditory loop can be made: timing, spectral content and intensity. Speech timing is almost always studied by delaying time of arrival of the speaker's voice, when it is called Delayed Auditory Feedback (DAF). In fluent speakers, DAF slows speech mainly by elongating vowels. The speech also has a monotone pitch and high amplitude (both effects being easily discernible on the vowels of DAF speech). The effects of DAF on fluent speakers have been explained by proposing that a feedback monitor is used for speech control. Speakers continue to use the altered sound for voice control when the sound of the voice is experimentally altered, and speech problems occur because the monitor acts on this misleading feedback. Timing alterations to feedback lead to adjustment to output timing where none was needed and this causes odd-sounding speech that can contain errors. In speakers who stutter, DAF has the effect of reducing the incidence of stuttering events dramatically. It was originally considered that the response to DAF of adults who stutter was opposite to the response of fluent speakers. However, this position is not sustainable: DAF disruption occurs mainly on medial vowels, (Howell, Wingfield & Johnson 1988), whereas the stuttering events that are lost occur on initial consonants (Brown 1945). From the dif­ ferent distributions of the two types of fluency failure, it can be established that speakers who stutter, while losing stuttering events under DAF, experience the same disruption to speech as fluent speakers. These differences rule out proposals that people who stutter have a timing problem in the monitoring process that once rectified by an external DAF delay allows these speakers to regain fluent control (Webster & Dorman 1970).

EXPLAN APPLIED TO THE TREATMENT OF STUTTERING

97

Fluent speakers shift vocal pitch down when the speech spectrum is shifted up and vice versa. Frequency-shifted Feedback (FSF) has been the subject of intense study since Howell, El-Yaniv & Powell (1987) found that it has equally dramatic effects as DAF in controlling the fluency of people who stutter without any of the side-effects on vowels that occur with DAF. For instance, the intensity of speech output does not change under frequency shift for fluent speakers or speakers who stutter (Howell 1990). Also, speakers who stutter do not make compensatory shifts in voice fundamental frequency as fluent speakers do (Natke & Kalveram, in press). The final form of alteration, intensity, has been extensively studied in fluent speech control. One variant of this manipulation is to amplify or attenuate the speech directly. When speech is amplified, speakers reduce voice level (Lane & Tranel 1971). Another way of investigating voice intensity change is by changing ambient noise level. A high noise level operates like attenuating the voice (i.e., speakers increase loudness when speaking under noise). No work has been done to investigate the changes in voice intensity in people who stutter. However, masking the voice of people who stutter has been investigated and this is reported to improve fluency control in these speakers (Cherry & Sayers 1956). These authors proposed auditory feedback improved speech control in people who stutter by masking the voice. 3. Problematic relevance of the auditory loop It was not necessary to consider an articulation-audition (external) loop, like that in Levelt's (1989) model, in the previous chapter. This loop needs considering here, though, as the powerful effects of AAF may be due to inter­ ference with this loop. As arguments were only given for dismissing monitor­ ing of internal loops in the earlier chapter, nothing said so far requires that an auditory feedback loop should be ruled out. Borden (1979) pointed out two fundamental problems for auditory feedback monitoring loops. First, auditory processing takes time (about 100 ms even for simple reaction time). So, at least an additional 100 ms would be needed after every segment has finished, to process the auditory output from it, establish that it was as intended and then to carry on with the message. This would lead to slower speech rates than speakers achieve. Second, hearing-impaired adults can control their speech even though they are not able to use audition to any great extent (Borden 1979). Thus, it appears that auditory feedback is not necessary for controlling the voice, as has been assumed in auditory monitoring accounts.

98

PETER HOWELL

A third problem, not noted by Borden, is that from an auditory monitoring perspective, it is necessary to be able to retrieve a veridical record of what was produced; otherwise, establishing if and what error has occurred with the intention of correcting it would not be possible. However, it is not clear how veridical the representation of articulatory output provided by the auditory system is about the intended message. The auditory representation the speaker receives while speaking is affected by internal and external noise sources dur­ ing transmission and the external noises vary with speaking environment. Internal noise arises through bone-conduction and is heard along with di­ rect sound. These sound-sources are at approximately equal levels (von Bekesy 1960). Formant structure is heavily attenuated in bone-conducted sound and this sound is dominated by the voice fundamental (Howell & Powell 1984). Bone-conducted sound limits the information about articulation that can be provided by the auditory loop as it is very different from speech output. External sources of noise are mainly a result of extraneous environmental sounds and echo. Extraneous noise affects the intensity-time profile of speech (Rosen & Howell 1981), and rooms with hard walls are reverberant, which affects the acoustic properties of speech salient for perception (Watkins 1992). Thus speech is output into different environments that have unique effects on the sound of the voice. Again, the effect of these variations limits what infor­ mation the speaker can obtain through hearing about how the intended speech was articulated. In brief, internal and external noise added during output makes the distal source of speech a non-reflexive form of the originally conceptualized message that cannot be used for control by a monitor. The first requirement imposed on the architecture of EXPLAN is that plan­ ning is independent of execution. This independence does not allow an audi­ tory feedback monitoring loop so some other explanation why AAF modifies speech is needed. Dismissing a monitor (specifically for monitoring auditory events) was also suggested by the three problems discussed in this section. A second requirement is that a satisfactory theory needs to avoid the latter problems while, at the same time, incorporating (requirement three) an expla­ nation of control by fluent speakers and (requirement four) control by speakers who stutter under normal listening and AAF conditions. An explanation of why disparate forms of treatment (such as AAF and operant procedures) are effective with people who stutter and why the different forms of treatment work over different time periods is also needed. A development of EXPLAN is proposed in the following sections and assessed against these requirements.

EXPLAN APPLIED TO THE TREATMENT OF STUTTERING

99

4. Speech output control without feedback monitoring The basics of EXPLAN as they apply to spontaneous speech produced in normal listening conditions are presented, and developments of the theory to extend it in order to explain the effects of AAF are then given. In this section, EXPLAN's operation is described for word production as in the preceding chapter. Words were used to describe how EXPLAN operates in the previous chapter as most of the evidence considered used these segments. However, syllables are more useful when considering rhythm-control that lies behind the EXPLAN account of speech modification. Words will continue to be used for describing the extension of EXPLAN to speech modification, though monosyl­ labic words should be understood to establish connection with the rhythm of speech. EXPLAN proposes that planning takes place independently of, and in parallel with, execution. Provided that there is enough time during the execu­ tion of one word to generate the next, the following plan can be picked up when the first word is completed and speech will be fluent. The fluent situa­ tion is depicted schematically in Figure la. The diagrammatic representation of the temporal relationship between plan­ ning and execution for three words (n, n+1, n+2) when speech is proceeding fluently is given. Time is along the abscissa. The epoch during which planning (PLAN) and execution (EX) occur are shown as bars in the top and middle rows respectively. The bottom row indicates the computation made on the plan and its execution (the result is zero in these fluent cases).

Fig. la: Planning and execution for three words

Figure lb gives a stretch of fluent speech where, at the start, planning and execution are hand-shaking appropriately on the first word (word n). After word n is first completed, the plan for n+1 is not ready (its planning continues to the right of the bold arrow). In this case, the speaker retrieves plan n, executes that again (leading to a word repetition), at which point the

100

PETER HOWELL

plan for word n+1 is ready and can be completed fluently (the difference computation between the plan and its executed form is zero).

Fig. lb: Stretch of fluent speech on the first word

Figure 1c shows a stretch of fluent speech where, at the start, planning and execution are hand-shaking appropriately on the first word in the sequence shown (wordn). After word n is completed on the first occasion after planning, the plan for word n+1 is not ready. Even so, the speaker executes the part plan of word n+1. After the execution of this part plan has been completed, an efferent copy of the timing pattern is differenced with the complete plan and results in a non-zero difference. This alerts the speaker who then makes a rate change to prevent this situation continuing.

Fig. lc: Stretch ofluent speech on the first word in this frequence

Two types of fluency failure were described in the previous chapter. In both of these, the plan for a word was not available within the time necessary for executing the previous one. This occurs in Figures lb and lc, at the point where execution of word n finishes before the time when the plan of word n+1 is complete. Stalling fluency failures deal with this situation by producing an immediately previous word (or words) again, or by pausing. In the example in Figure lb, word n should be re-executed. Planning of word n+1 continues and

EXPLAN APPLIED TO THE TREATMENT OF STUTTERING

101

when execution of word n has finished the second time, if the plan of word n+1 is then ready it will be executed. Importantly in stalling types of fluency failure, no word is executed until its plan is complete. The situation where the plan is not complete gives rise to different type of fluency failure {advancing fluency failures). Instead of waiting for the whole plan to be ready, the speaker commences execution of the word with the part of its plan that is available. In advancing fluency failures, represented sche­ matically in Figure lc, fluency can fail in ways that involve production of the first part of words. This has been observed to be a feature of persistent stut­ tering (Conture 1982). 4.1 Alerts, stalling and advancing The speech control system ought to detect when speech is advancing pre­ maturely so that something can be done to avoid the inherent danger of fluency failure. The necessary control may be achieved by checking whether the whole plan is supplied for execution. Howell and Sackin (in press) proposed that the check on speech timing is made by taking the difference between the planned and executed versions where the executed version is an efferent copy of the timing plan (e.g., a marker for each vowel onset) delivered to the motor effectors. This aspect of the theory will be illustrated for fluent speech, and then the situation that occurs in stalling and advancing fluency failures will be considered. In the stretch of fluent speech shown in Figure la, the efferent copy (1) of the timing plan of a speech segment is taken at the point in time that its execu­ tion commences (signified for word n as c(EX(n))). A copy of the timing plan of the word is obtained when execution is completed (c(PLAN(n))). c(PLAN(n)) is then differenced with c(EX(n)). If the timing patterns correspond, the dif­ ference is zero, and the two timing signals cancel. This result indicates that execution rate is not too rapid for planning to keep up. The plans of preceding words remain available for a short time after they have been executed (Blackmer & Mitton 1991). So when the plan for the next word is not complete, prior words whose plan is still available can be reexecuted (word repetition). When the plan for word n+1 is not ready, the plan of word n is retrieved (Figure lb) and executed. The timing pattern of the exe­ cuted form of word n (c(EX(n))) will still correspond with the efferent copy of the timing plan of the word (c(PlAN(n))) so the difference will still be zero (as in speech produced fluently). Consequently, word repetition is essentially a fluent speech control mechanism. The point where the speaker has advanced to word n+1 while planning of this attempted word is still going on is depicted after word n has been

102

PETER HOWELL

executed in Figure 1c. If the speaker advances and produces the first part of the word, more, but not necessarily all, of the plan will be produced during the time that this part plan is being executed because the speech plan continues to be updated for execution in this period. When this updated plan, taken at the time execution is completed (c(PLAN(n+l))), is differenced with the efferent copy of the plan at the point in time that execution commences (c(EX(part P(n+1)))), the result is always non-zero. Non-cancellation of the signals provides an alerting response that the speaker is speaking at too high a rate. The intermittency of the alerting response and the relatively crude information it provides (that something has gone wrong, not whether a component of articulation has failed to be realized properly) is a major difference between the current proposal and auditory feedback monitoring accounts. In particular, in EXPLAN no measure of the error in articulation occurs that could be used to minimize any discrepancy between intended and actual speech. Consequently, a monitor has no error information to work with. A block diagram giving an overview of what happens after the EXPLAN process breaks down and the additional processing components that then come into play is given in Figure 2.

Fig. 2: Block diagram of the processes involved in executing speech

Fluent speech and stalling are ways of adjusting local speech rate that do not involve external timing changes (control occurs within the section top left

EXPLAN APPLIED TO THE TREATMENT OF STUTTERING

103

of Figure 2, the response loop). According to EXPLAN, the alerts generated during advancing fluency failures are a series of pulses (the non-zero pulses after differencing each segment). These are input to a timekeeper (bottom right of Figure 2). Because of the selective involvement of the timekeeper during advancing fluency failures, an equivalent way of describing the operation of sending input to the timekeeper is that timing control changes from an intrinsic to an extrinsic schedule (Fowler 1980). Figure 2 shows the inputs that would be received by the timekeeper when a fluent hearing individual is speaking easy material at a comfortable rate and dealing with any occasional incomplete plans by stalling fluency failures. Speech is controlled within the response loop where PLAN and EX handshake in a chaining process (as EX of one word is complete, the PLAN of the next is supplied). Serial activity associated with the execution of each segment arises during this process and, as with all such activity, it is an input registered by the timekeeper. At the same time, the timekeeper also receives a serial input via the auditory system (the speech output). The load that arises from these two synchronised inputs is within the timekeeper's capacity and no change is made. Effectively, the PLAN-EX cycles in the response loop and operates inde­ pendent of the timekeeper. Consideration will be given in the remainder of this chapter, to how changes in the inputs to the timekeeper can change operation of the response loop. In Figure 2 the response loop (top left) performs the operations detailed in Figure 1. The timekeeper is bottom centre. The bulbs represent inputs to the timekeeper that are connected automatically when activity arises in these input lines. When load to the timekeeper is exceeded, the parameters of the coupled oscillator (top centre) are adjusted and the response loop adjusts automatically. The coupled oscillator is under timekeeper control can undergo global (top input line, response to DAF) or local (bottom input line, response to FSF) slowing. 4.2 Timekeeper and load Situations where the inputs to the timekeeper change can now be con­ sidered, starting with the case of advancing fluency failures. In this case, an alert is added to the inputs to the timekeeper. According to EXPLAN, adding an alert sequence raises the number of inputs to the timekeeper. The relationship between the alert and the other concurrent inputs to the timekeeper determines load and can affect the timekeeper's operation. 4.2.1 Number, rate and precision requirements of inputs. In EXPLAN, the load on the timekeeper is determined by sequential internally-generated neural signals, signals arising from other external sensory events and any similar ac-

104

PETER HOWELL

tivity from ongoing motor activities. As alerts are intermittent, they add a neu­ ral input to the timekeeper when they arise. This raises load as the more signals there are, the higher the load. Changing rate of inputs also affects load. If the timekeeper is monitoring a fixed number of inputs, other things being equal, load will increase when the rate of these inputs increases. Load also in­ creases when concurrent tasks are performed that adds inputs and can make demands for more precision on the timekeeper. 4.2.2 Synchrony. Synchronous signals are easier for a timekeeper to deal with than asynchronous ones. Howell, Powell & Khan (1983) illustrate this with canon, a simple form of song in which a synchronised rhythm is heard while singing. Conversely, load increases when sensory or motor inputs are out of synchrony, as when dancing or clapping offbeat (Howell et al. 1983). 4.3 Response of the timekeeper to load variation The timekeeper mechanism is governed by the principle that it needs to keep load within its capabilities. In the spontaneous speech of a fluent speaker speaking at moderate or slow speech rates with sounds that are simple to plan and spoken in an ordinary listening environment (Figure 2), no alerts are gen­ erated. Change in any of these circumstances will cause load on the time­ keeper to increase and potentially to exceed its capacity. If capacity is ex­ ceeded, the mechanism can offset the increase that occurs, though the options for this are limited. Two such ways of reducing load are now considered. Both involve the timekeeper changing the parameters of an oscillator that is coupled with the response loop. This oscillator has its parameters (centre frequency and bandwidth) set initially so that they are in tune with the modulation rate of speech in the response loop (e.g., centre frequency set at word rate, bandwidth set to allow expressive variation around that frequency). The coupled oscillator does not then affect the response loop providing speech rate remains within these limits, preserving the independence of the response generation cycle from the timekeeper. 4.3.1 Global rate control DAF slows overall (global) speech rate because asynchronous inputs that occur with this type of manipulation, overload the timekeeper. The timekeeper responds to asynchronous inputs by shifting the centre frequency of the coupled oscillator to a lower mean value. The coupled oscillator is represented as a resonator in the frequency domain and the shift in centre frequency is shown in the top branch going to the response loop at top right of Figure 2. This alteration to the parameters of the coupled oscillator then perturbs the responses generated in the response loop. That is,

EXPLAN APPLIED TO THE TREATMENT OF STUTTERING

105

resonance principles suggest that the responses are entrained to a lower rate so that they are, once again, not affected by the coupled oscillator. The slower response rate that is induced, besides having an inherently lower load itself, would also remove or reduce the incidence of alert inputs to the timekeeper, as they occur less often at slower rates because planning has more time to catch up with execution rate. 4.3.2 Local rate control Episodes of speech that contain a word that starts with a difficult string of phonemes (Howell, Au-Yeung & Sackin 2000), and where the speech rate in the immediate context is rapid (Howell, Au-Yeung & Pilgrim 1999) are most prone to fluency failure in people who stutter. A change to speech rate focussed on these regions (local rate change) is all that is necessary to reduce advancing fluency failures and reduce load on the time­ keeper. The rate in more extensive stretches of speech, such as phonological words (Selkirk 1984), can be represented as a distribution of rates of segments within these stretches. The segments that lie at the fast rate end of this distri­ bution require slowing. The incidence of these fast segments can be reduced without affecting mean speech rate if an equivalent reduction in the incidence of slow segments is made. This would be revealed if variance in the rate distribution is reduced. FSF has been reported to lead to such a reduction in the variance of rate responses in repeated readings of a target sentence by speakers who stutter (Howell & Sackin 2000). The bottom branch at the top right of Figure 2 depicts the effects of a local rate change. When load is exceeded under FSF, the coupled oscillator is tuned more sharply. The narrower bandwidth entrains the speech rates made in the response loop into a narrow band around the centre frequency (bottom branch). One effect of this is to reduce the incidence of segments spoken fast that can be problematic in that the plan may not be ready for execution in time in these regions. Removing these sections would make speech fluent and remove associated alerts, reducing the load on the timekeeper. 5. EXPLAN addressed to the problems faced by a monitoring account The problems raised in connection with a monitoring account are now re­ considered. Borden's first concern was the amount of time involved in processing feedback for monitoring speech. It is quick to make an efferent copy and perform a differencing operation. The proposed operations in EXPLAN would not, therefore, slow speech down to the extent that obtaining and checking an auditory representation does.

106

PETER HOWELL

Borden's (1979) second point was that auditory feedback is not essential in adults who sustain hearing loss. Sound input goes to the timekeeper, not to the response loop in EXPLAN. So, loss of an input to the timekeeper would not affect operation of the response loop. The loss of auditory input to the timekeeper consequent on hearing loss would reduce load. The timing mechanism responsible for speech output can still operate efficiently when this input is lost. EXPLAN does not include a monitoring mechanism to explain the effects of AAF treatment. Specifically, execution output is not used to revise or tune the speech plan (Postma 2000). The requirement was also made that the speech control mechanism has to operate without veridical information about speech output. In EXPLAN, the auditory input channel only operates as serial input to the timekeeper, not as a channel that provides information about how speech is articulated. So, as the speaker is not retrieving information about placement of articulators to produce the sounds, it does not matter whether the auditory version of the speaker's own speech is veridical, only that its timing pattern does not interfere with speech activity. 6. Account of AAF and secondary tasks in fluent speakers Listening environment, the fluency of the speaker, and the number and complexity of tasks being performed simultaneously, can all affect the time­ keeper and its control structures. All these factors add extra signal/s (in or out of synchrony) or change the rate of signals so the load on the timekeeper alters. The effects of AAF that have been traditionally explained as due to feedback monitoring, really reflect interference to timing control mechanisms (Ivry 1997 locates the timekeeper in the cerebellum). In EXPLAN, FSF and DAF both add an extra input to the timekeeper, increasing its load. The DAF procedure produces an extra signal that is out of synchrony with that due to direct speech output responses from the response loop and unaltered auditory output of the voice. The extra asynchronous DAF input to the timekeeper increases its load. The timekeeper then needs to reduce load, which it does by reducing mean response rate in the oscillator coupled to the response loop (global rate change). FSF also creates a second signal from manipulations of vocal output that is input to the timekeeper. This sound is in synchrony with sensory and response inputs from speech that occur as with DAF. Speakers decrease the bandwidth of the oscillator coupled with the response loop that leads to a reduction in variance of subjects' responses when they are presented with this type of alteration (a local change in response rate that decreases load). The EXPLAN model also incorporates an explanation for secondary task disruption. All serial events associated with sensory, response or neural

EXPLAN APPLIED TO THE TREATMENT OF STUTTERING

107

activity from a secondary task are merely treated as more serial inputs. A con­ current task that involves monitoring and/or responding to a serial signal adds load to the timekeeper because of the additional sensory (and possibly response) input. The extra input/s add load that can lead to response rate or accuracy changes (Howell & Sackin submitted). 7. Effect of rate changes in people who stutter EXPLAN predicts that improvement in speech is always attendant on change in rate control (locally or globally). Both ways of reducing rate will affect how much time is allowed for EX in the EX-PLAN system. Voluntary rate changes should tax or relieve problems where speech planning cannot keep up with execution rate. There is much evidence that reducing the speech rate of speakers who stutter decreases frequency of fluency failure, as EXPLAN pre­ dicts (Johnson & Rosen 1937; Perkins, Kent & Curlee 1991; Starkweather 1985; Wingate 1976). It is also widely reported that rate increase has the opposite effect, again as EXPLAN predicts. Increasing global speech rate increases frequency of fluency failure (Johnson & Rosen 1937; Bloodstein 1987). Besides the evidence where speech rate is altered voluntarily, many clinical procedures known to alleviate stuttering have equivalent effects insofar as they slow speech execution rate. A speculative model about long-term stuttering, suggested by EXPLAN, is that the disorder persists because no adjustment to rate is made when alerts occur. Speakers who stutter experience many fluency failures that lead to alerts. However, although many alerts occur, the speakers do not respond to them. Speech responses that lead to alerts will continue if no action to avoid them is taken. The speaker then gets saturated with alerts that are ignored and loses sensitivity through adaptation (Howell, Rosen, Hannigan & Rustin 2000). The converse applies to fluent speakers: Most of their fluency failures are of the stalling type that do not involve alerts. Responding appropriately to alerts on the rare occasions that advancing fluency failures occur, obviates their occurrence and maintains sensitivity to the alerts by keeping their occurrence infrequent. According to this perspective, treatments should help speakers who stutter by ensuring the timekeeper keeps its sensitivity to alerts, and by treating the alerts as serial signals to a timekeeper that are controllable if speakers respond appropriately. A period where alerts are not experienced as frequently (as ex­ perienced by fluent speakers) may be necessary to regain sensitivity. While DAF and FSF operate on response output, they do this by adding load to the timekeeper that removes alerts. This transient exposure to added load may be too short to restore sensitivity to alerts. To restore sensitivity, it may be neces-

108

PETER HOWELL

sary to prevent alerts for long periods of time or target AAF specifically on stuttered sections and according to prescribed schedules (Reed & Howell 2001). EXPLAN suggests that the extended period of treatment time associated with operant procedures (such as the Australian Lidcombe program) may be reinstating sensitivity by preventing alerts. These procedures may prevent alerts in a different way to AAF (i.e. preventing too rapid execution, thereby influencing the response loop directly). Whichever way alerts are prevented, the timekeeper can regain its sensitivity and fluent speech can be reinstated. 8. Comparing architectures of monitoring and EXPLAN outputs Table 1 gives a summary comparison of error monitoring and the alert view to highlight the differences discussed so far. EXPLAN during advancing fluency failures works on neural alerts whereas the monitors discussed earlier on in this chapter use auditory information to detect equivalent problems (row 1). Neural alert signals are rapid to obtain and process and the alerting activity would be unaffected in people who sustain a hearing loss (Borden 1979). The monitoring and EXPLAN outlooks differ in terms of the signal information they provide (row 2). The error information provided to a monitor is much more detailed than that provided by an alert, as it needs to specify what alterations are needed to correct or retune a speech plan after an error. In the previous chapter, it was pointed out that speech errors are rare and a different strategy for dealing with them is simply to let them happen and then restart speech. Whereas monitoring needs to be continuous and produces a signal all the time, alerts are intermittent (row 3). A knock-on effect of this is that continuous signals have to be continuously monitored, thus affecting overall speech rate (Borden 1979). The two perspectives also differ with respect to the response that is initiated when fluency fails (row 4): in an error monitor, speech has to be replanned or tuned to remove the error, whereas in EXPLAN, rate slows as an emergent property of the timekeeper, not on the basis of a reformulation of the message.

1. 2. 3. 4.

Signal Information provided Existence Response

Error monitors Auditory Contains information what was wrong Signal continuously generated Message replanned when fluency fails

Alerts CNS-internal Indicate where failure occurred, not what the failures were Signal intermittently generated Slower rate is an emergent response

Table 1: Differences between error-monitoring and alert perspectives on speech control

EXPLAN APPLIED TO THE TREATMENT OF STUTTERING

9. Some experimental lines of support for the

EXPLAN

109

control structures

9.1 Is the delayed signal under DAF speech or noise? According to EXPLAN, the delayed sound does not have its effect because the speech information it contains is deciphered and used for control; rather it is a signal with particular rhythmic properties that make it disruptive. Consequently, any noise with the same spectral and intensity properties that appears at the same points as the delayed signal should have an effect equivalent to a delayed speech sound. Howell & Archer (1984) confirmed this with an experiment, in which a non-speech noise with the same intensity profile as the original speech was substituted for the delayed sound under DAF. There was no difference in the time taken to read a list in this stimulus condition compared with that in a delayed speech condition. This suggests that any sound that stands in this temporal relationship with the direct speech will cause equivalent disruption. The interpretation that the delayed signal disrupts the rhythm, not the speech content, would also explain why DAF sounds produce an increase in volume as with noise (Howell 1990). The act of delaying speech creates a sound that disrupts rhythm, which is responded to as noise. 9.2 Does DAF selectively affect the timekeeper process? EXPLAN maintains that the effects of alteration to auditory feedback are due to activity sent specifically to the timekeeper mechanism. Manipulations that influence the timekeeper can be separated from manipulations that affect motor responses using Wing & Kristofferson's (1973) model. As originally developed, the Wing-Kristofferson model decomposes the total variance in a set of tapping responses that have to be made at a specified rate into components associated with the motor and clock processes. The data for the delayed auditory feedback conditions are shown in Figure 3 a and Figure 3b.

Fig. 3a: Motor variance

110

PETER HOWELL

• ♦ - -66ms -

-

133ms

Fig. 3b: Timekeeper (clock) variance

Howell and Sackin (submitted) had speakers do a speech version of the Wing-Kristofferson task (their subjects repeated the syllable /bae/ at pre­ selected rates), and performed the Wing-Kristofferson analysis after syllable onsets had been marked. Subjects performed the experiment at two rates (600 and 800 ms repetition rates) and at three DAF delays (66, 133 and 200 ms). DAF delay of the points connected together was 66, 133 or 200 and the delay used for connected points can be identified from the symbol in the caption. Axes are variance on the ordinate and repetition rate on the abscissa on both sides of the graph. Clock variance increases as repetition rate increases as re­ ported by Wing (1980) in a tapping task. Importantly, clock, but not motor variance also shows dramatic increases as DAF delays are changed from 66 through to 200 ms delays. This demonstrates that most of the effect of DAF is specifically to timekeeper variance, as EXPLAN predicts. 9.3 Serial inputs and speech EXPLAN predicts that serial input signals, in general, are signals going to the timekeeper that should lead to improvement in speech performance in people who stutter because of the effects they have on speech rate. Howell & El-Yaniv (1987) reported that a click synchronous with syllable onset improved the speech of people who stutter as much as a regularly-timed metronome click. The pacing effect of a regularly-timed metronome could lead to improved fluency by slowing speech rate, if (as is usual) the metronome is set to a slow rate. However, the speech synchronous click does not impose a prescribed rate; it is synchronised to the rate the subject uses. The EXPLAN account would pre­ dict that the timekeeper receives an extra input from the metronome pulse what­ ever mode it is operating in, and this leads to the timekeeper regulating its load, affecting speech rate control. Both types of metronome should reduce rate-

EXPLAN APPLIED TO THE TREATMENT OF STUTTERING

111

variability as the clicks are speech synchronous. In turn, this should lead to local slowing that improves speech by reducing the incidence of fast, problematic stretches of speech. Howell et al. (1983) showed that interrupted vowels have some abstract similarities to DAF. Delaying a sound's onset leads the speaker to hear sound at a low level and for the level to increase when the sound onsets after the delay. Interrupting a vowel sound also leads to such steep increases in intensity during vowels. Based on the similarity in disruption to intensity patterns, it would be predicted that, as DAF improves the speech of speakers who stutter, interrupted vowels should also do so, as confirmed recently by Kalinowski, Dayalu et al. (2000). EXPLAN predicts that this would be associated with global slowing as the interruption creates an intensity increase that is asynchronous with direct speech activity. All serial inputs, whatever their modality, are fed to the timekeeper. All serial input signals to the timekeeper would have equivalent effects on load. In people who stutter, the effect of the timekeeper reducing response rate due to a load increase allows planning to catch up with execution and stuttering rate to drop. EXPLAN predicts that even serial inputs from different modalities would produce improvement in much the same way, Kuniszyk-Jozkowiak, Smolka & Adamczyk (1996) have confirmed this prediction. They demonstrated that speech control by speakers who stutter improves when a concomitant flashing light input occurs. 9 4 Affecting fluency failures by operant procedures Beattie & Bradbury (1979) reported changes in stalling fluency failures in fluent speakers using a verbal operant-conditioning procedure. Silent pauses were detected electronically and a light came on when they occurred. Partici­ pants were instructed to try to prevent the light illuminating while maintaining the same speech articulation rate. Findings revealed an increase in the number of word repetitions. Howell & Sackin (in press) replicated this study and established that this repetition occurred mainly on function words. According to EXPLAN, pauses and function word repetitions both have a stalling role. Stalling is essential for maintaining fluent speech, so when the experimenter prevents speakers from using one way of stalling, another automatically emerges. New­ man (1987) also suggested that rate control is important for fluent speech con­ trol. He found that punishing word repetition led to a reduced word output rate« Operant procedures are effective with young speakers who stutter though less so with older speakers (Onslow, Costa & Rue 1990) whereas the reverse has been reported for FSF (Howell, Sackin & Williams 1999). The ineffective­ ness of the operant work with adults may suggest that if the ability to alert the timekeeper is lost it is difficult to re-establish its usual function. Conversely. operant procedures appear to prevent this happening when applied in childhood.

112

PETER HOWELL

10. Conclusions If the EXPLAN account of why operant procedures work is correct, it is only necessary to stop advancing fluency failures in order to prevent alerts. EXPLAN considers that stalling fluency failures are ways of controlling fluency within the response loop and should be reinforced as they may be beneficial in offering a way of avoiding more serious fluency failures. Advancing fluency failures, on the other hand, lead to alerts and should be targeted for reduction to prevent stuttering (assuming that the speaker will then respond by using stalling fluency failures or speak at a slower rate commensurate with fluent speech). Currently my research group is employing a variety of procedures reported to increase stalling fluency failures while at the same time punishing advancing fluency failures. Three children who were diagnosed as non-recovered stutterers based on both the speech and non-speech criteria described in the previous chapter have been treated. The results for one child are given in Figure 4. Percentage of content (mainly advancing) and function (mainly stalling) word fluency failures at baseline, various points during intervention, and six weeks post-treatment are shown. Results show that relative to baseline speech data during diagnosis, stuttering rate of both advancing and (more surprisingly) stalling fluency failures decreased. The reduction in stalling fluency failures may be a result of the operant procedures reducing overall speech rate, thereby cutting down all types of fluency failure. Within the treatment sessions, content word (advancing) fluency failures reduced significantly over sessions. The follow-up six weeks after treatment had finished showed that most of the improvement obtained during therapy was maintained.

—

f

u

n

c

t

i

o

n

word

Fig. 4: Results for subject ED during intervention and post treatment.

EXPLAN APPLIED TO THE TREATMENT OF STUTTERING

113

REFERENCES Beattie, Geoffrey W. & R. J. Bradbury. 1979. "An experimental investigation of the modifiability of the temporal structure of spontaneous speech". Journal of Psycholinguistic Research 8.225- 248. Blackmer, Elizabeth R. & Janet L. Mitton. 1991. "Theories of monitoring and the timing of repairs in spontaneous speech". Cognition 39.173-194. Bloodstein, Oliver. 1987. A Handbook on Stuttering. 4th ed. Chicago: National Easter Seal Society. Borden, Gloria J. 1979. "An interpretation of research on feedback interruption in speech". Brain & Language 7.307-319. Brown, Spencer F. 1945. "The loci of stuttering in the speech sequence". Journal of Speech Disorders 10.181 -192. Cherry, Colin, & Bruce Sayers. 1956. "Experiments on the total inhibition of stam­ mering by external controls and some clinical results". Journal of Psychosomatic Research 1.233-246. Conture, Edward G. 1982. Stuttering. Englewood Cliffs, N.J.: Prentice-Hall. Fairbanks, Grant. 1955. "Selective vocal effects of delayed auditory feedback". Jour­ nal of Speech and Hearing Disorders 20.335-348. Fowler, Carol A. 1980. "Coarticulation and theories of extrinsic timing". Journal of Phonetics 8.113-133. Howell, Peter. 1990. "Changes in voice level caused by several forms of altered feedback in normal speakers and stutterers". Language and Speech 33.325-338. Howell, Peter & Alexander Archer. 1984. "Susceptibility to the effects of delayed auditory feedback". Perception & Psychophysics 36.296-302. Howell, Peter & James Au-Yeung. 2001. "The EXPLAN theory of fluency failure and the diagnosis of stuttering". This volume, 75-94. Howell, Peter, James Au-Yeung & Lesley Pilgrim. 1999. "Utterance rate and lin­ guistic properties as determinants of speech dysfluency in children who stutter". Journal of the Acoustical Society of America 105.481-490. Howell, Peter, James Au-Yeung & Stevie Sackin. 2000. "Internal structure of content words leading to lifespan differences in phonological difficulty in stuttering". Journal of Fluency Disorders 25.1-20. Howell, Peter & Nirit El-Yaniv. 1987. "The effects of presenting a click in sylla­ ble-initial position on the speech of stutterers: comparison with a metronome click". Journal of Fluency Disorders 12.249-256. Howell, Peter, Nirit El-Yaniv & David J. Powell. 1987. "Factors affecting fluency in stutterers when speaking under altered auditory feedback". Speech Motor Dy­ namics in Stuttering ed. by H. Peters & W. Hulstijn, 361-369. New York: Springer. Howell, Peter & David J. Powell. 1984. "Hearing your voice through bone and air: implications for explanations of stuttering behaviour from studies of normal speakers". Journal of Fluency Disorders 9.247-264.

114

PETER HOWELL

Howell, Peter, David J. Powell & Ian Khan. 1983. "Amplitude contour of the delayed signal and interference in delayed auditory feedback tasks". Journal of Experi­ mental Psychology: Human Perception and Performance 9..112-784. Howell, Peter, Stuart Rosen, Geraldine Hannigan & Lena Rustin. 2000. "Deficits in auditory temporal resolution in children who stutter and its relation to dysfluency rate5'. Perceptual and Motor Skills 90.355-363. Howell, Peter & Stevie Sackin. 2000. "Speech rate manipulation and its effects on fluency reversal in children who stutter". Journal of Developmental and Physical Disabilities 12.291-315. Howell, Peter & Stevie Sackin submitted. "Timing interference to speech in altered listening conditions". Perception & Psychophysics. Howell, Peter & Stevie Sackin. In press. "Function word repetitions emerge when speakers are operantly conditioned to reduce frequency of silent pauses". Journal of Psycholinguistic Research. Howell Peter, Stevie Sackin & Roberta Williams. 1999. "Differential effects of fre­ quency shifted feedback between child and adult stutterers". Journal of Fluency Disorders 24.127-136. Howell, Peter, Trudie Wingfield & Michael Johnson. 1938. "Characteristics of the speech of stutterers during normal and altered auditory feedback". Proceedings Speech 88, 7th Federation of Acoustical Societies of Europe conference, Edin­ burgh 1988 ed. by William A. Ainsworth & John N. Holmes, vol 3, 1069-1076. Edinburgh: Institute of Acoustics. Ivry, Richard. 1997. "Cerebellar timing systems". The Cerebellum and Cognition ed. by Jeremy Schmahmann, 555-573. San Diego: Academic Press. Johnson, Wendell & Leonard Rosen. 1937. "Effect of certain changes in speech pat­ tern upon frequency of stuttering". Journal of Speech Disorders 2.105-109. Kalinowski, Joseph, Vikram N. Dayalu, Andrew Stuart, Michael P. Rastatter & Manish K. Rami. 2000. "Stutter-free and stutter-filled speech signals and their role in stuttering amelioration for English speaking adults". Neuroscience Letters 293.115-118. Kuniszyk-Jozkowiak, Elzbieta Smolka Wieslawa & Bogdan Adamczyk. 1996. ''Effect of acoustical, visual and tactile reverberation on speech fluency of stut­ terers". Folia Phoniatrica & Logopedics 48.193-200. Lane, Harlan L. & Bernard Tranel. 1971. "The lombard sign and the role of hearing in speech". Journal of Speech and Hearing Research 14.677-709. Lee, Bernard S. 1950. "Effects of delayed speech feedback". Journal of the Acousti­ cal Society of America 22.824-826. Levelt, Willem J.M. 1989. Speaking: From intention to articulation. Cambridge, Mass.: MIT Press. Natke, Ulrich & Karl T. Kalveram. In press. "Fundamental frequency and vowel du­ ration under frequency shifted auditory feedback in stuttering and nonstuttering adults". Proceedings of the Third World Congress on Fluency Disorders, 7-11 August 2000, Nyborg, Denmark.

EXPLAN APPLIED TO THE TREATMENT OF STUTTERING

115

Newman, Linda L. 1987. "The Effects of punishment of repetitions and the acquisi­ tion of 'Stutter-Like' behaviors in normal speakers". Journal of Fluency Disor­ ders 12.51-62. Onslow, Mark, Leanne Costa & Stephen Rue. 1990. "Direct early intervention with stuttering". Journal of Speech and Hearing Disorders 55.405-416 Perkins, William H., Ray Kent & Richard Curie. 1991. "A theory of neuropsycholinguistic function in stuttering". Journal of Speech and Hearing Research 34.734-752. Postma, Albert. 2000. "Detection of errors during speech production: A review of speech monitoring models". Cognition 77.97-131. Reed, Phil & Peter Howell. 2001. "Presentation of frequency-shifted feedback to sustain long-term improvements in fluency in people who stutter". European Journal of Analysis of Behaviour 1.89-106. Rosen, Stuart & Peter Howell. 1981. "Plucks and bows are not categorically per­ ceived". Perception and Psychophysics 30.1256-1260. Selkirk, Elisabeth. 1984. Phonology and Syntax: The relation between sound and structure. Cambridge, Mass.: MIT Press. Starkweather,  Woodruff. 1985. "The development of fluency in normal children". Stuttering therapy: Prevention and intervention with children ed. by Hugo Gregory, 9-42. Memphis, TN: Speech Foundation of America. von Bekesy, Georg. 1960. Experiments in hearing. New York: McGraw Hill. Watkins, Anthony J. 1992. "Perceptual compensation for the effects of reverberation on amplitude Envelopes - Cues to the Slay-Splay distinction". Proceedings of the Institute of A coustics 14.125-132. Webster, Ronald L. & Michael F. Dorman. 1970. "Decreases in stuttering frequency as a function of continuous and contingent forms of auditory masking". Journal of Speech and Hearing Research 14.307-311. Wing, Alan M. 1980. "The long and the short of timing in response sequences". Tu­ torials in motor behavior ed. by George E. Stelmach & Jean Requin, 469-486. Amsterdam: North Holland. Wing, Alan M. & Alfred B. Kristofferson. 1973. "Response delays and the timing of discrete motor responses". Perception & Psychophysics 14.5-12. Wingate, Marcel E. 1976. Stuttering: Theory and treatment. New York: IrvingtonWiley.

III. MORPHOLOGY AND SYNTAX IN CHILD LANGUAGE DISORDERS

VERB MOVEMENT AND FINITENESS IN LANGUAGE IMPAIRMENT AND LANGUAGE DEVELOPMENT ROELIEN BASTIAANSE, GERARD BOL, SOFIE VAN MOL & SHALOM ZUCKERMAN University of Groningen 1.

Introduction Both normally developing children and children with Specific Language Impairment (SLl) have problems with the production of finite verbs (e.g., Leonard 1998; Wexler et al. in press). The same phenomenon has been de­ scribed for agrammatic aphasic patients (e.g., Saffran et al. 1989; Thompson et al. 1995; Bastiaanse et al. 1995). In the spontaneous speech of these popula­ tions many non-finite verbs are used, usually in combination with a restricted number of finite verbs. In normally developing children this is called the op­ tional infinitive stage. It has been argued that children with SLI show a delay in the development of language skills and therefore the optional infinitive stage takes longer to disappear, if it disappears at all. Rice and Wexler (1996) have named this period in specific language impairment the extended optional infinitive stage. Several authors in the past and present suggested that language loss due to brain damage is characterized by a pattern that is the reverse of language ac­ quisition (e.g., Jakobson 1941; Grodzinsky 1990; Avrutin 1999). At first sight, the characteristics of verbs produced by agrammatic aphasics in spontaneous speech show similarities with those in children in the optional infinitive stage. Agrammatic aphasies produce a significantly diminished number of finite verbs (Thompson et al. 1995; Bastiaanse & Jonkers 1998) compared to non-brain­ damaged speakers, but still produce a considerable number of finite verbs. The present study focuses on two central questions: what is the origin of the problems with finite verbs in these three populations and is it true that children with SLI deal the same way with the problems of the production of finite verbs as normal children do. We will first give a short description of relevant linguistic aspects of Dutch, followed by a summary of the empirical findings in the three populations. Then a test will be presented, meant to tear apart the aspects that are related to finite verbs. The results of testing the three populations will be given and discussed.

120

BASTIAANSE, BOL, VAN MOL & ZUCKERMAN

1.1 Linguistic background Dutch has been analyzed as an sov-language, meaning that the basegenerated position of the verb is after the object (Koster 1975). In the Dutch declarative matrix sentence, the finite verb has to be moved to second position. This movement is known as Verb Second. If the main verb clusters with a modal verb or auxiliary, the main verb remains in situ and the modal verb or auxiliary is moved to Verb Second position. In embedded clauses, the finite verb remains in its base-generated position. This is illustrated in (1-3), where t designates the canonical verb position, which is co-indexed with the Verb Second position. (1)

(2)

(3)

matrix sentence without modal verb/auxiliary de jongen koopti een fiets t¡ the boy buys a bike matrix sentence with modal verb de jongen wili¿ een fiets kopen

ti

the boy wants a bike (to) buy "The boy wants to buy a bike." embedded clause without modal verb/auxiliary (ik denk) dat de jongen een fiets koopt (I think) that the boy a bike buys "I think that the boy buys a bike."

Here it is assumed that a verb is inserted in the structure in fully inflected form (Chomsky 1995; for Dutch, see Zwart 1993). In the embedded clause, the finite verb remains in its base-generated position; in matrix clauses, only the finite verb moves to the left, the nonfinite verb stays behind. Hence, in the matrix clause the finite verb has to fulfil two requirements: it has to move to the second position and it has to check its features. 1.2 Psycho- and neurolinguistic background The speech of young normally developing Dutch children (around the age of 2) is characterized by the use of uninflected verbs (Bol & Kuiken 1990). In Dutch, these verbs are in clause final position, as in (4)-(5). This implies that these verbs are not only non-finite, but also non-moved. This phenomenon oc­ curs both with objects (4) and with subjects (5): (4) (5)

koekje eten cookie eat-lNF papa bouwen daddy build-lNF

VERB MOVEMENT AND FINITENESS

121

In the optional infinitive stage, the children also use a restricted number of finite verbs, usually used as dummy auxiliaries. In Dutch, the third person sin­ gular form of gaan (to go) is often used by children (see, for example, Jordens 1990), as shown in (6). (6)

auto gaat vallen car goes fall-lNF "the car falls" This sentence type is regular in adult Dutch, but it has an inchoative meaning ("is going to"). Children in the optional infinitive stage, however, use this dummy auxiliary for the present tense as well. Other finite verb forms that are used by children in the optional infinitive stage are the copula is and the empty element doet (does). It has been suggested by De Haan (1987) that these dummy auxiliaries which are produced in this stage of acquisition stage are base-generated in Verb-Second position and, hence, no movement is involved. Rice & Wexler (1996) proposed that the problems with the production of finite verbs encountered by children with SLI are caused by a similar mecha­ nism, as shown by their label extended optional infinitive stage. Bastiaanse & Bol (2001) also showed similarities with respect to the production in finite verbs in the two groups of children. The proportion of finite verbs produced by nor­ mally developing children is indeed the same as the proportion produced by children with SLI with the same 'mean length of utterance'. They also demon­ strated, however, that the variation in the proportion of finite verbs in children with SLI is dependent on the diversity of the lexical verbs they produce. The children with SLI can be divided into two subgroups. One group that produces a relatively normal proportion of finite verbs, combined with a relatively low di­ versity of the produced lexical verbs and one group that produces relatively few finite verbs but has an adult-like variation of lexical verbs. This reverse relation­ ship does not appear in the speech production of normally developing children, but there is a similarity with verb production in agrammatic aphasics. Brain­ damaged people suffering from this aphasia type, also known as Broca's apha­ sia, produce a relatively large number of non-finite verbs (both infinitives - to refer to the present - and participles - to refer to the past), as shown in (7)-(8). (7)

(8)

moeilijke zinnetjes vormen difficult sentences form-lNF "to form difficult sentences" na de oorlog gebouwd after the war built-PART "built after the war"

122

BASTIAANSE, BOL, VAN MOL & ZUCKERMAN

Just like in children with SLI, the proportion of finite verbs is reversibly related to the variety of lexical verbs. Bastiaanse & Bol (2001) interpret these data as an inability of languageimpaired people to integrate lexical-semantic and morpho-syntactic informa­ tion, due to a lack of resources needed for complex linguistic operations. In this respect, language impaired individuals differ from children who are acquiring language in a normal way: although the production of finite verbs is difficult to learn, it does not interfere with the retrieval of lexical elements (verbs) from the lexicon. Hence, the similarity between these three (Dutch) populations is the rela­ tive large number of non-finite clauses in their spontaneous speech. One may wonder what makes finite verbs difficult for them. As mentioned before, in order to produce a finite verb in a Dutch matrix verb, two operations must be performed: a morphological operation, that is verb inflection, and a syntactic operation, that is, verb movement. The central question is at makes the pro­ duction of finite verbs in Dutch difficult for these speakers. Is it the morpho­ logical operation (verb inflection) or the syntactic operation (movement)? Ac­ cording to Bastiaanse & Van Zonneveld (1998), for agrammatic aphasics it is movement that causes the problems. In the present paper it will be tested whether this hypothesis can be supported and whether the syntactic account also holds for the two other populations. In order to find an answer to these questions, a test was developed in which finite verbs were elicited in both base-generated position (in the embedded clause) and Verb Second position (in the matrix clause). If the speakers perform at the same level in the matrix and the embedded clause, a morphological account is most likely: they cannot produce finite verbs properly, regardless of movement. If, however, the production of finite verbs in the base-generated position is easier than in Verb-Second position, then support is found for the hypothesis that it is verb movement that is the issue at stake and hence a syn­ tactic operation is causing the difficulties with finite verbs. 2.

Methods

2.1 Subjects Three different groups were tested: 10 children with a normally developing language system (age 3;0-3;11), 10 children with a specific (grammatical) language impairment (mean age 6;2; range 4;10-6;11), as diagnosed by their speech therapists and 6 agrammatic Broca's aphasics (mean age 46.5, range 27-67), for which the aphasia type was established by the Aachen Aphasia test (Graetz et al. 1992) and confirmed by both the

VERB MOVEMENT AND FINITENESS

123

speech therapist and the examiner.1 All patients were right-handed and aphasic due to a single stroke in the left hemisphere, at least 3 months prior to onset (mean time post-onset 35.2 months; range 3-66 months). The choice for normally developing children of this age was motivated by the fact that this is about the youngest age band that can reliably be tested with a language task. The children with SLI were children who were available for testing. One more child participated, but was later excluded from the group, because he was much older (11 years) than the rest of the group. Fourteen older children (age 4;8-5;0) were tested as a control group. 2.2 Materials The participants were presented with two pictures in which the same per­ son was performing the same action with a different person or object (see Figure 1). They were asked to complete a sentence with a finite verb and an object. There were two conditions: 1) a matrix clause that should be comple­ ted with a finite verb - object, 2) an embedded clause with the intended answer object -finite verb. There were 17 sentences in each condition; two of these were used as training items. Two examples are: Condition 1: matrix clause (+movement) Tester: Dit is de man die de tomaat snijdt en dit is de man die het brood snijdt. Dus deze man snijdt de tomaat en deze man "This is the man that the tomato cuts and this is the man that the bread cuts. So, this man cuts the tomato and this man " Participant: "snijdt het brood" "cuts the bread" Condition 2: embedded clause (-movement) Tester: Deze man snijdt de tomaat en deze man snijdt het brood. Dus dit is de man die de tomaat snijdt en dit is de man die "This man cuts the bread and this man cuts the tomato. So this is the man that the bread cuts and this is the man that " Participant: "het brood snijdt" "the bread cuts"

1

Not all children with a normally developing language system finished the entire test. The minimum number of items required for inclusion in the analysis was 16 (50% of the items, equally divided over matrix and embedded clause).

124

ASTIAANSE, BOL, VAN MOL & ZUCKERMAN

The reason to prompt the participants with the verb and the noun several times before s/he was supposed to produce it, was to avoid interference with word-finding problems. Each set of pictures was used twice, once with a ma­ trix clause and once with an embedded clause to complete. The test started with two examples that were repeated until it was clear that the subject under­ stood the task.

Fig. 1: An example of the pictures of the test

2.3 Scoring A correct/incorrect scoring system was used, in which determiner omission was ignored. Self-corrections were allowed and the final answer was the one that was analysed. If requested, the examiner repeated the cueing sentences one time. Post-hoc, an error analysis system was developed on the basis of the most frequently occurring errors: auxiliary insertion, verb omission, word order errors, and a class 'others', which contained semantic paraphasias, nil reactions etc. 3. Quantitative analysis The older children who were used as control subjects performed perfectly in both conditions. Their results will further be ignored. The quantitative re­ sults are given in Table 1. Percentages are used, since not all normally developing children completed the test (see note 1) matrix clause

embedded clause

57

82

normally developing children 49

89

children with SLI

84

! agrammatic aphasics

58

Table 1: Percentages correct for the three populations.

125

VERB MOVEMENT AND FINITENESS

Each population performs significantly better on the embedded clauses than on the matrix clauses (normally developing children t(9) = 6.76, p<0.01; children with SLI t(9) = 2.63, p<0.05; agrammatic aphasics t(5)= 2.79, p<0.05). These results support Bastiaanse & Van Zonneveld (1998), who hypothesized that a syntactic impairment causes the problems in agrammatic production of finite verbs and demonstrate that their hypothesis can be extended to the two other populations. 4. Qualitative analysis The results of the error analysis are given in Table 2. verb omissions

dummy AUX

word order

others

14

0

21

4

normally developing children 22

33

9

10

children with SLI

32

4

28

10

0

0

10

6

normally developing children 2

4

2

6

children with SLI

0

12

4

matrix clause agrammatic aphasics

embedded clause agrammatic aphasies

12

Table 2: Number of errors per error type for the three populations

Verb omission is a frequent error in all three populations. Notice that this error is mainly made in the matrix clause. This poses a problem, since verb omission is allowed in this environment, both in Dutch and in English, as shown in (9)-(10): (9) (10)

deze jongen snijdt de tomaat en deze jongen het brood this boy cuts the tomato and this boy the bread

In the embedded clause these omissions are not allowed: (11) (12)

*dit is de jongen die de tomaat snijdt en dit is de jongen die het brood *this is the boy that cuts the tomato and this is the boy that the bread

126

BASTIAANSE, BOL, VAN MOL & ZUCKERMAN

It was decided to count these reactions as erroneous for three reasons: 1) 2) 3)

no participant was consequent in this error, showing, again, that they were aware that a full VP was required; the errors were also made in the embedded clause by the two groups of children, though to a lesser extent; the resulting structure in which the verb is omitted in a co-ordinating sentence is linguistically very complex and not expected to be pro­ duced by these population on purpose.

The most frequent error the normally developing children produce are socalled 'auxiliary dummies'. These are finite auxiliaries that are supposed to be base-generated in I (De Haan 1987). Interestingly, the kind of auxiliary is de­ pendent on the dialect the children speak. In the South of the Netherlands the children use doet ("does"), which is allowed in their dialect, but not in standard Dutch, the language in which they were tested. In the North of the Netherlands, the children use gaat ("goes"), which has, in standard Dutch, an inchoative meaning ("is going to"). Zuckerman & Bastiaanse (2000) showed that children between 3 and 4 years old do not yet understand the difference between (13) and (14): (13)

(14)

de jongen eet een appel the boy eats an apple "The boy eats an apple." de jongen gaat een appel the boy goes-FUT an apple "The boy is going to eat an apple."

eten eat-INF

Notice that these errors are not hardly ever made in the embedded clause, although both doet and gaat constructions are allowed in the embedded clause. Children with SLI produce dummy auxiliaries occasionally and only in the matrix clause. These are all instances of gaat "goes" (the children are all from the North). The adult agrammatic speakers never produced a dummy auxiliary. Their most frequent error, which is also produced to a large extent by the children with SLI, concerns word order, mainly object - finite verb combinations in the matrix clause, as in (15) (15)

*deze jongen this boy

de tomaat eet the tomato eats

VERB MOVEMENT AND FINITENESS

5.

127

Conclusions The two questions of this study were: 1) are the problems with the pro­ duction of finite verbs in the three populations morphological or syntactic in nature, and 2) are the problems the same in each of the three populations? The quantitative analysis showed the three populations are very well able to produce finite verbs correctly when these are in base-generated position (the embedded clause), meaning that verb morphology is not impaired. It is when these finite lexical verbs have to be moved, that the problems arise, suggesting syntactic problems. The way in which the three populations solve their syntactic problems are different, however. All three populations omit the finite verb in the matrix clause, normally developing children's most frequent error is auxiliary insertion, an error type that never occurs in the agrammatic responses. Agrammatic aphasics' most frequent error is the production of strings in which the obligatory verb movement has not taken place, an error type that is responsible for only a small proportion of the errors produced by the normally developing children. The children with SLI are somewhere in between, but tend to follow the agrammatic pattern: non-movement of the lexical finite verb. How can it be that agrammatics do not use the 'dummy option', which re­ sults at least in a grammatical sentence? First of al, the doet ("does") variant is not available to them, because this is a form from a dialect that they do not master (none of the agrammatic speakers came from the South). This leaves the gaat ("goes") option open to them. As stated above, gaat + inf in Dutch has an inchoative meaning ("is going to ...") and this is not what is shown in the pictures. Nevertheless, both normally developing children and children with use this form. Zuckerman & Bastiaanse (2000) and Van Mol (2000) showed that neither of these groups of children fully understand the meaning of gaat + inf which in their grammar is identical to the present tense. This means that they can choose between, for example, de jongen snijdt de tomaat ("the boy cuts the tomato") and de jongen gaat de tomaat snijden (lit. "the boy goes the tomato cutting": the boy is going to cut the tomato). In such a case, they opt for the form which is most 'economical' in terms of movement (see Zuckerman 1999). Since agrammatic aphasics do know the difference between the inchoative and present tense, this option is not open to them. They produce object - finite verb strings, which are ungrammatical in matrix clause context. These errors were not anticipated, since these constructions never occur in spontaneous speech: although the verb is often produced in clause final position, this is always an infinitive (see Bastiaanse et al. in press). The produced errors suggest that the agrammatic speakers can retrieve finite

128

BASTIAANSE, BOL, VAN MOL & ZUCKERMAN

verbs from their lexicon, but are not always able to 'move' them to Verb Second position. Still, they produce a word order that exists in their language, but only in the embedded clause. Interestingly, English speaking agrammatic aphasics never produce these word order errors when presented with the same test.2 Preliminary data from normally developing children acquiring Frisian (a language with Verb Second in the matrix clause, just like in Dutch) in which the auxiliary - infinitive option is not allowed, show that these children produce the same errors as the Dutch agrammatic speakers: object -finite verb in the matrix clause. Apparently no grammatical option is open to them. To recap, the normally developing children participating in this study have problems with Verb Second and circumvent these problems by inserting dummy auxiliaries, either from their dialect, or from a construction that in adult grammar has a different meaning. Agrammatic speakers (and, antici­ pating on the final results of a study in progress, children with Frisian as their native language) cannot use this solution, as in their grammar it is not available in the way it is in the normally developing Dutch-speaking children. Still, the agrammatic speakers produce a string {object -finite verb) that is al­ lowed in their grammar, but only in the embedded clause. Interestingly, they never produce an infinitive followed by the object, the most frequent error in the English speaking subjects. Bastiaanse & Thompson (2000) suggest that this is because this string is not a part of Dutch grammar and can therefore not be produced. What we see is that there is a clear dichotomy between the errors produced by the normally developing children and the agrammatic speakers. This means that we cannot confirm the hypothesis that language decay is simply the opposite from language learning as suggested by, for example, Jakobson (1941) and, more recently, by Grodzinsky (1990) and Avrutin (1999). Agrammatic aphasics have knowledge of their language that young children have not yet acquired, such as the inchoative meaning of gaat + inf. Although there are certain similarities (both populations have problems with verb movement), the solutions are different. The second question of the present study concerned the consequences of the problems in the production of finite verbs in the three populations. There is a similarity between the two groups of children in their verb movement problem, but the way the children with SLI deal with their problem, resembles the performance of the other language impaired group, the agrammatic apha­ sies, just as suggested by Bastiaanse & Bol (2001). Their spontaneous speech analysis showed that all three populations have problems with the production 2

As expected, the discrepancy between matrix and embedded clause is not found for English, but the English speaking agrammatic aphasies do make errors.

VERB MOVEMENT AND FINITENESS

129

of finite verbs, but that in the two language-impaired groups these problems were related to lexical retrieval abilities. We would therefore like to conclude that, although there are clear similari­ ties between normally developing children and children with SLI with respect to the production of finite verbs, the children with SLI resemble the agrammatic aphasics when it comes to circumventing their problems. This means that the three populations have the same underlying problem (i.e., verb movement), but different ways to solve it.

REFERENCES Avrutin, Sergey. 1999. Development of the Syntax-Discourse Interface. Dordrecht: Kluwer. Bastiaanse, Roelien & Gerard Bol. 2001. "Verb inflection and verb diversity in three populations: agrammatic speakers, normally developing children and children with specific language impairment (SLI)". Brain and Language 77.274-282. Bastiaanse, Roelien, Judith Hugen, Mirjam Kos & Ron van Zonneveld. In press. "Lexical, morphological and syntactic aspects of verb production in Dutch agrammatic aphasics". Brain and Language. Bastiaanse, Roelien & Roel Jonkers. 1998. "Verb retrieval in action naming and spontaneous speech in agrammatic and anomie aphasia". Aphasiology 12.951-969. Bastiaanse, Roelien, Roel Jonkers & Uke Moltmaker-0singa. 1995. "Aspects of lexi­ cal verbs in the spontaneous speech of agrammatic and anomie patients". Lan­ guage and Cognition ed. by Roel Jonkers, Edith Kaan & Anko Wiegel, vol. V, 13-26. Groningen: University of Groningen. Bastiaanse, Roelien & Cynthia K. Thompson. 2000. "Verb finiteness in agramma­ tism: A cross-linguistic study". Brain and Language 74.503-506. Bastiaanse, Roelien & Ron van Zonneveld. 1998. "On the relation between verb in­ flection and verb position in Dutch agrammatic aphasies". Brain and Language 64.165-181. Bol, Gerard & Folkert Kuiken. 1990. "Grammatical analysis of developmental lan­ guage disorders: A study of the morphosyntax of children with specific language disorders, with hearing impairment and with Down's syndrome". Clinical Lin­ guistics and Phonetics 4.77-86. Chomsky, Noam. 1995. The Minimalist Program. Cambridge, Mass.: MIT Press. Graetz, Patty, Ria De Bleser & Klaus Willmes. 1992. Akense Afasie Test. Lisse: Swets and Zeitlinger. Grodzinsky, Yosef. 1990. Theoretical Perspectives on Language Deficits. Cam­ bridge, Mass.: MIT Press. Haan, Gerben J. de. 1987. "A Theory-bound Approach to the Acquisition of Verb Placement in Dutch". Formal Parameters of Generative Grammar: Yearbook ed. by Gerben de Haan & Wim Zonneveld, 15-30. University of Utrecht.

130

BASTIAANSE, BOL, VAN MOL & ZUCKERMAN

Jakobson, Roman. 1968 [1941]. Child Language, Aphasia and Phonological Univer­ sals. The Hague: Mouton. Jordens, Peter. 1990. "The acquisition of verb placement in Dutch and German". Linguistics 23.89-125. Koster, Jan. 1975. "Dutch as an SOV language". Linguistic Analysis 1.111-136. Leonard, Lawrence B. 1998. Children with Specific Language Impairment. Cam­ bridge, Mass.: MIT Press. Mol, Sofie van. 2000. The Relation between Children with Specific Language Im­ pairment and the Production and Comprehension of Moved Verbs. MA-thesis, University of Groningen. Rice, Mable L. & Kenneth Wexler. 1996. "A phenotype of specific language im­ pairment: extended optional infinitives". Toward a Genetics of Language ed. by Mable L. Rice, 215-237, Hillsdale, N.J.: Lawrence Erlbaum Associates. Saffran, Eleanor M., Rita S. Berndt & Myrna F. Schwartz. 1989. "The quantitative analysis of agrammatic production: Procedure and data". Brain and Language 37.440-479. Thompson, Cynthia K., Lewis P. Shapiro, Ligang Li & Lee Schendel. 1995. "Analysis of verbs and verb argument structure: A method for quantification of agrammatic language production". Clinical Aphasiology 23.121-140. Wexler, Kenneth, Jeanette . Schaeffer & Gerard W. Bol. In press. "Verbal syntax and morphology in Dutch normal and SLI children: How developmental data can play an important role in morphological theory". Syntax. Zuckerman, Shalom. 1999. "Economy-based markedness as a universal for language acquisition". Paper presented at GLOW 1999. Berlin, 2-3 March 1999. Zuckerman, Shalom & Roelien Bastiaanse. 2000. "Auxiliary + infinitive structures in child Dutch: Evidence from production and comprehension". Paper presented at the 25th Annual Boston University Conference on Language Development. Boston, 3-5 November 2000. Zwart, . Jan Wouter. 1993. Dutch Syntax: A minimalist approach. Groningen: Grodil.

A-BAR MOVEMENT CONSTRUCTIONS IN GREEK CHILDREN WITH SLI EVIDENCE FOR DEFICITS IN THE SYNTACTIC COMPONENT OF LANGUAGE*

STAVROULA STAVRAKAKI University of Thessaloniki 1. Introduction In recent psycholinguistic studies, there has been increasing interest in in­ vestigating the linguistic abilities of the Specifically Language Impaired (SLI) children. The language development of SLI children is characterized by severe problems in the acquisition of morphosyntax, while their cognitive, motor and social development falls within the normal range (on criteria setting, cf. Stark & Tallal 1981). Based on cross-linguistic research findings indicating that in­ flectional and derivational morphology is severely impaired in SLI children (see Clahsen 1991; Gopnik & Crago 1991; Rice et al. 1995 among others), di­ verse explanatory models of the linguistic deficit in SLI have been developed. They make different hypotheses on the locus of the deficit in SLI grammar as well as on the nature of developmental pattern followed in SLI, i.e. whether language development in SLI is a normal but delayed procedure or an abnormal process. Consider first the hypotheses concerning the locus of the deficit in SLI. An assumption underlying some of those models is that the deficit concerns the grammatical features intrinsic to lexical items. The models differ in the extent to which the grammatical features are supposed to be impaired. Following a 'broad' interpretation of the deficit in SLI, Gopnik and her colleagues put forth the Missing Features Account, according to which the syntactic-semantic features such as person, number, tense, aspect, and gender are missing from the grammar of SLI individuals. In the same sense, Tsimpli & Stavrakaki (1999) and Tsimpli (2001) suggest that only the

My thanks to the audience of the International Conference "Linguistic Theory, Speech and Language Pathology, and Speech Therapy" in Padova and to the conference committee for organizing an exciting, interdisciplinary conference. Thanks also go to Paola Crisma for her useful comments on this paper and to Elisabetta Fava. The usual disclaimers apply. This work has been supported by the National State Scholarship Foundation of Greece (IKY).

132

STAVROULA STAVRAKAKI

non-interpretable features of grammar (Chomsky 1995) are severely impaired in SLI. Adopting a 'narrow' interpretation of the deficit, Clahsen et al. (1997) and Rice et al. (1995) argue that the deficit lies in the optional noninterpretable (p-features of verbs and the missing or underspecified feature of Tense respectively. Taking into account research findings showing that SLI children have problems with syntactically complex structures, van der Lely (1994, 1997) suggests that the deficit causing SLI can be interpreted within the framework of the Representational Deficit Model for Dependent Relations (henceforth, RDDR; cf., van der Lely & Stollwerck 1997; van der Lely 1997). The RDDR model postulates that the deficit lies mainly in the computational component of the language. A direct prediction of that model is that local syntactic relations are expected to be handled correctly but a breakdown of performance will be exhibited on long-distance dependencies involving syntactic operations such as movement. Recently, within the framework of the RDDR model, van der Lely (1998) suggested that the deficit in SLI is rather caused by the absence of the principle that forces movement of unchecked features, i.e., the Must Move principle; due to that absence, movement is optional in SLI grammar. Let us now examine those models with respect to their predictions for the way in which language development occurs in SLI. Except for the hypothesis formulated by Rice et al. (1995), according to which SLI children differ from normal ones only in quantitative terms, i.e. in the higher-than-expected uses of underspecified tense, since they follow the normal acquisition path with some delay, all other hypotheses predict qualitative differences between the normal and SLI development. In particular, the SLI children's grammar is expected not completely to correspond to anyone phase of normal development, due to the fact that there are selective deficits in particular areas of grammar where progress is difficult to occur. The question, which arises then, is concerned with the particular strategies allowing SLI subjects to cope with their linguistic deficits. According to Paradis & Gopnik (1997), such strategies are related to exploitation of pragmatic cues in comprehension and use of metalinguisitc knowledge. Metalinguistic knowledge refers to the explicit knowledge of rules, which is applied deliberately in the controlled construction of utterances. This kind of knowledge is not available automatically in the unconscious process involved in the microgenesis of a sentence. Since explicit knowledge is not converted into implicit knowledge through practice (Paradis 1994), then it is expected that individuals with SLI will never overcome their linguistic deficit and, thus, that they will exhibit a deviant pattern of linguistic development.

A-BAR MOVEMENT CONSTRUCTIONS IN GREEK CHILDREN

133

Against the theoretical background presented above, the primary aim of this study is to present experimental data on the production of structures in­ volving A-bar movement, that is, relative clauses and wh-questions, by Spe­ cifically Language Impaired and Normally Developing Greek children. In par­ ticular, our study deals with the following questions. First, whether SLI performance differs than that of Language Age (LA) control group and in what ways. Second, what the status of the underlying linguistic mechanism in SLI and normal grammar is. Finally, an attempt is made to discuss the implications of the research findings of this study for the core of the theoretical issues con­ cerning SLI, that is, the locus of the deficit and the way language development takes place in SLI children. The overall organization of this chapter is as follows. Firstly, the structure of the A-bar movement constructions (relative clauses and wh-questions) in Greek is sketched and a brief review of the acquisition literature is provided. Then, the experiments are presented. Finally, the findings are discussed and their implications for the theory of SLI are underlined. 2. A-bar movement constructions: On their structure and acquisition 2.1 Greek relative clauses and wh-questions in normal and SLI acquisition Grrek relatives clauses are introduced either with the complementizer pu ("that") or with the relative pronoun o opios ("who"). Opzos-relatives are not so frequent in Greek (Varlokosta 1998:105). Pu-relatives are taken to contain a null relative operator, which originates internally within the sentence and is then moved to the SpecCP position, while in opios-relatives there is movement of an overt operator into SpecCP. The formation of wh-questions requires overt raising of a wh-operator to clause initial position. Also, adjacency be­ tween wh-operator and the verb is required, as in (1): (1)

Ti ekane  Janis? What-ACC do-3SG-PAST the-John-NOM "What did John do?"

As far as the normal language acquisition is concerned, some asymmetries in the availability of the wh-movement are attested between the formation of the interrogatives on the one hand and the relatives on the other. In particular, wh-movement is widely attested in child interrogatives, as shown by crosslinguistic data (Crisma 1992; Guasti 1996, 2000). Research questions are con­ cern the order of emergence of different types of wh-questions (Stromswold 1995; Wilhelm & Hanna 1992). Avrutin (2000) reports results of a compre­ hension experiment with English speaking children who were presented with

134

STAVROULA STAVRAKAKI

four types of wh-questions: who-Subject/Object and which-Subject/Object questions. The children exhibited significantly different performance on which-Object (which-O) questions than that on the other question types. The drop of the children's performance on which-O questions is interpreted as the impact of discourse related operations to the sentence interpretation. In particular, it is suggested that due to their D-linked status (Pesetsky 1987), the which-O questions require the integration of the syntactic and discourse related knowledge simultaneously. In this sense, they are more 'expensive' than the non D-linked questions, the interpretation of which is purely syntactic (Avrutin 2000:300). On the other hand, the cross-linguistic evidence from the acquisition of the relative clauses is rather contradictory, since young children do not always produce adult-like relatives. The non-adult performance of normally developing children gave rise to a hypothesis assuming that no operator movement is used in the formation of child relatives (Labelle 1990, 1996; Goodluck & Stojanovic 1996; Guasti & Shlonsky 1992). Alternatively, it is argued that syntactic movement is used and the non-adult instances, when attested, are rather due to performance factors (McKee et al. 1998; Crain et al. 1990; Varlokosta 1998). Based on some discrepancies attested between the acquisition of the operator movement in w/z-questions and relative clauses, it has been suggested that the relative delay in the formation of relative clauses is rather due to the properties of relative operators, and in particular, to their linking status (Wexler 1991; Guasti & Shlonsky 1995). That is, an operator in an A'position needs to be co-indexed both with its variable and with its antecedent. Consider now the research findings on the production and comprehension of the A-bar constructions in SLI children. As noted by van der Lely (1997), such complex structures are hardly attested in SLI data. An investigation of Subject and Object w/z-questions carried out using an elicitation task (Battell 1998) indicated that English SLI children have great problems especially with the production of Object w/z-questions. Finally, based on Greek s u data from comprehension of reversible relative clauses, Stavrakaki (2001a) argues that the SLI performance is significantly different than that of LA controls on those relatives where the parsing preferences are violated and, thus, the interpreta­ tion is exclusively dependant on syntactic cues. In sum, structures involving A-bar movement seem to cause great difficulties in SLI children crosslinguistically.

A-BAR MOVEMENT CONSTRUCTIONS IN GREEK CHILDREN

13 5

3. Thefirstexperiment: eliciting relative clauses 3.1 Subjects, materials and procedure Two groups of subjects participated in the experiment. A summary of the subject details is presented in Table 1. Chronological age (CA)

Raw Scores

SUBJECT GROUP

RANGE

MEAN

SD

RANGE

MEAN

SD

SLI group N=8

5.4-9.4

7.38

1.567

60-92

79

12.0475

LA controls N=16

3.4-5.2

4.1

.5825

63-93

78

7.9956

Table 1 : Chronological ages and Raw Scores from the Diagnostic Verbal IQ Test

The first group consisted of 8 SLI children, selected according to a set of criteria proposed by Stark & Tallal (1981). Each child in the SLI group was individually matched with two control children on the basis of individual raw scores from the Diagnostic Verbal IQ (DVIQ) Test for Greek children (Stavrakaki & Tsimpli 1999). Therefore, the LA control group consisted of 16 normally developing children, who were selected on the basis of raw scores in the DVIQ. Analysis revealed no significant difference in raw scores between SLI children and LA controls [t (22)=.229 p=.821]. A toy elicitation task (Crain & Thornton 1998) was used. The experi­ menter was the storyteller, who also manipulated the toy props, while a blind­ folded puppet was trying to understand what was going on in the story. Two toy-figures were used for the head of the relative. Subject and Object gap rela­ tives with Subject and Object head were elicited. In particular, 6 Subject gap relatives and 9 Object gap relative clauses were elicited. Examples of the elicited constructions are presented in Table 2, were there are 5 types of rela­ tive clauses: Subject Subject (SS), Subject Object (SO), Object Object (00), Object Object-clitic (OO-clt), Object Subject (OS). There were 3 stories for each relative type, giving a total of 15 answers per child. Consider now an example protocol for the elicitation of SS relatives: One tiger is hitting the elephant and pushing the zebra. Another identical tiger is playing with a doll. The child is asked to tell the blindfolded puppet which tiger is pushing the zebra. Target response: the one that is hitting the elephant

136

STAVROULA STAVRAKAKI

SUBJECT HEAD SS RELATIVES Subject gap

I tigri pou htipai ton elefanta The-tiger-NOM-that-hit-3sthe-elephant-ACC sprohni ti zevra push-3s-the-zebra-ACC "The tiger that is hitting the elephant is pushing the zebra" SO RELATIVES

Object gap

 skilos pou filai I tigri the-dog-NOM-that-kiss-3s-the-tiger htipa rinokero hit-3 s-the-rhinokeros-ACC "The dog that the tiger is kissing is hitting the rhino"

OBJECT HEAD

1

OS RELATIVES ƒ alepou sprohni ti ghata the-fox-NOM-push-3 s-the-cat-ACC pou kinighai to skilo that-chase-3 s-the-dog-NOM "The fox is pushing the cat that is chasing the dog" OO RELATIVES o elefantas kinigha tin kamilopardali pou the-elephant-chase-3s- the-giraffe-ACC-that htipa  rinokeros hit-3 s-the-rhinokeros-NOM "The elephant is chasing the giraffe that the rhino is hitting" OOclt (with clitics) RELATIVES to alogo htipa to provato pou The-horse-hit-the-sheep-that to kinighai to elafi -clitc-chase-3 s-the-deer "The horse is hitting the sheep that the deer is chasing"

Table 2: Experiment 1: The test constructions

3.2 Results and discussion The correct performance in the production of relative clauses of the two groups is presented in Table 3. Mean

SD

SLI children

4.1650

6.1057

LA controls

70.83

25.4

Table 3: The correct performance (%) of SLI children and LA controls

SLI children exhibited significantly lower performance than that of LA con­ trols [t (16.638)= -9.27 p<.001]. The different relative types were codified as follows: SS, OS, SO, OO, and OOclt. The first letter in the code identifies the grammatical function of the head while the second letter indicates the gram­ matical function of the gap within the relative clause. In OOclt relatives, the arguments are morphologically underspecified and the presence of clitic is somewhat necessary. In this type of relatives, the arguments are in neuter gen­ der and, thus, they are morphologically underspecified, since there is no overt

A-BAR MOVEMENT CONSTRUCTIONS IN GREEK CHILDREN

137

morphological contrast between the nominative and the accusative case for neuter gender. It should be noted that 4 out of the 5 correct relatives in SLI data were SS relatives, that is, relatives with SVO word order. On the other hand, LA controls exhibited high level of performance on both Object and Subject extracted relatives, i.e. 00 and SS. Nevertheless, their performance drops on those relatives, where case conflict phenomena occur, that is, the head of the relative is associated with an empty category specified for a different case than that of the head. Such relatives are the SO and OS ones. Relatively low level of performance was also exhibited on ooclt relatives, in which the arguments are morphologically unmarked. The correct performance of the LA controls on all types of relative clauses is presented in Table 4. Mean

SD

Subject Subject relatives (SS)

83.33

29.81

Subject Object relatives (SO)

58.33

31.032

Object Object relatives (00)

85.42

32.13

Object Object-clt relatives (OO-clt)

60.42

32.7

Object Subject relatives (OS)

60.42

32.7

Table 4: The correct performance (%) of the LA controls on relatives

Error analysis indicated that only in 19 out of the 240 (7.92%) responses of the LA controls the relativization strategy was not used.1 Those responses included the production of simple active sentences (N=9) and coordinated structures (N=10) instead of the target. Consider the examples (2) & (3) re­ spectively: (2) a. aflos pu kingai o elefantas this-NOM that chase-3SG the-elephant-NOM "The one that the elephant is chasing" b. o elefantas kinigai ton rinokero the-elephant-NOM chase-3 SG the-rhino-ACC 'The elephant is chasing the rhino"

[target response]

[child's response]

1 There were incorrect responses in which the relativization strategy was used, such as, the relativization of the wrong head, the production of resumprive NPs, the case or gender errors on the head of the relative. For detailed error analysis, see Stavrakaki 2000a.

138

STAVROULA STAVRAKAKI (3) a. afto pu to htipai to liontari [target response] this-ACC that it-ACC-hit-3SG-the-lion-NOM "This one that the lion ishitting" b. to provato kinigai to alogo the-sheep-NOM chase-3SG the-horse-ACC ke to liontari htipai to provato [child's response] and the-lion-NOM hit-3SG the-sheep-ACC "The sheep is chasing the horse and the lion is hitting the sheep"

With regards to the SLI children, the overwhelming majority of errors be­ long to the category simple active sentences (76/120 responses: 63.33%). SLI children produced significantly more simple active sentences than LA controls [t (7.184)^4.035 p=.005]. Some coordinated structures (9/120) were also pro­ duced. Also, relatives with missing heads (15/120 responses: 12.5%) were at­ tested in SLI data. Consider the example below: (4) a. aftos pou kinigai ton elefanta [target response] The-one-NOM that-chase-3SG the-elephant-ACC "This one that chasing the elephant" b. *pou kinigai ton elefanta [SLI child's response] *that chase-3 SG the-elephant-ACC The above error was also taken to show absence of operator movement. If such movement occured, then coindexation between the variable bound by the operator in SpecCP and the head of the relative would be expected. Since heads are missing, no such coindexation can be estasblished. This error type was attested only in SLI data. In total, in 97 out of 123 responses (80.84%), the non-relativization strategy was used. Another asymmetry between the two groups was the predominance of the SVO word order in SLI data. The use of the svo word order in the groups' re­ sponses is presented in Table 5. Mean

SD

SLI children

67.5

6.606

LA controls

30.4

5.42

Table 5: SVO word order (%) in SLI children and LA controls

Notice, also, that the SLI children use the svo word order significantly more than the LA controls [t(22)=14.699 p<.001]. Besides, the overt lexical Subject is omitted in 37 out of 120 responses and, thus, the vo word order was used in those responses.

A-BAR MOVEMENT CONSTRUCTIONS IN GREEK CHILDREN

139

The properties of the performance of SLI children vs. those of the LA con­ trols' performance are summarized in Table 6. SLI children

LA controls

D Incorrect responses inmost of the cases D Predominance of simple active sentences D Predominance of the SVO word order

D Correct responses in most of the cases D Different error types across relative types D Variation in word order

Table 6: First experiment: SLI children vs. LA controls

In general, normally developing children's ability to represent the syntactic structures of relatives seems to be on target. The few correct relatives (with SVO word order) produced by SLI children cannot be taken to indicate the use of relativization strategy. 4. The second experiment: eliciting wh-questions 4.1 Subjects, materials and procedure Since the second experiment was carried out some months later, a re­ assessment of the linguistic abilities of the SLI children was made on the basis of the DVIQ test for Greek children. Following the same procedure, as in the fisrt experiment a LA group was selected. The Subject details are presented in Table 7 below: SUBJECT GROUP SLI group N=8 LA Controls N=16

Chronological age (CA) RANGE MEAN SD 1.4753 6.1-10 8.1

RANGE 64-98

3.6-5.6

63-100

4.4

0.73

Raw Scores SD MEAN 12.5178 82.875 81.18

11.44

Table 7: Chronological ages and Raw Scores from the Diagnostic Verbal IQ (DVIQ) Test

No significance was found between the performance of the SLI children and the LA controls on the raw scores from the DVIQ test [t(22)=.330 p=.744]. As shown in both Tables 1 and 7, the SLI children of this study are school age children. An assessment of their morphosyntactic abilities based on spon­ taneous speech data was conducted in the interval between the two experi­ ments. The focus of the assessment was those morphosyntactic categories that were found to be severely impaired at an early stage of the linguistic development of the SLI children, as soon as they started receiving speech

140

STAVROULA STAVRAKAKI

therapy services. The correct use of those grammatical categories out of their obligatory contexts is provided in Table 8. CORRECT USE Past tense

198/200(99%)

AgrS (2nd S/P) Definite article

78/80 (97.5%)

Object clitic pronouns (3rd person) Prepositions

48/50 (96%) 54/60 (90%)

Case (the marked form: NOM. Masculine gender) Interrogative words

26/26(100%)

284/290 (97.93%)

75/75 (100%)

Complementizers

14/22 (63.64%)

The mood marker na (=to).

35/40 (87.5%)

Table 8: The use of some grammatical categories by SLI children

As shown in Table 8, the grammatical profile of SLI children is far from that described at an early stage of their linguistic development where most of the above morphosyntactic features appeared to be severely impaired (Dalalakis 1994; Stavrakaki 1996, 1999; Tsimpli & Stavrakaki 1999). However, the SLI children of this study appear to be able to build up only sim­ ple syntactic structures and not complex ones (cf. van der Lely 1997). To elicit w/z-questions, games were designed in which the child asked a puppet a question about a scenario acted out with toys. Subject and Object which-NP and who questions were elicited. There is a clear difference between the which-NP and who-questions with respect to discourse presupposition. The former but not the latter are Discourse-linked (D-linked) (Pesetsky 1987). Six exemplars for each question type were tested giving a total of 24 responses per child. For which-NP questions there were three animals, of which two were identical. In particular, suppose there were two monkeys in the workplace and a dog was chasing one of them. Then the Experimenter (E) proceeded as follows: (5)

E:

S afti tin istoria,  rinokeros kinijise enan apo tus pithikus "In this story, the rhino chased one of the monkeys" Rotise tin kukla pion pithiko "Ask the puppet which one" Child: Pion pihtiko kinijise  rinokeros Who-monkey-ACC chase-3 SG the-rhino-NOM "Which monkey did the rhino chase?"

A-BAR MOVEMENT CONSTRUCTIONS IN GREEK CHILDREN

141

For the elicitation of the who-questions, there were used three different animals in each story. Consider the structure in (6).2 (6)

E:

S afti tin istoria, o rinokeros kinijise tonpithiko "In this story, the rhino chased the monkey" Rotise tin kuklapion Ask the puppet who-ACC "Ask the puppet who" Child: Pion kinijise  rinokeros? Who-ACC-chase-3SG the-rhino^TOM "Who did the rhino chase?"

4.2 Results and discussion The correct performance of the two groups of SLI children and the LA con­ trols is presented in Table 9. Questions

who-Subject Which-Suhject Who-Object Which-Object

(who-S) (which-S) (who-S) (which-S)

SLI children

LA controls

Mean

SD

Mean

SD

52.081 62.5 6.25 16.67

10.68 23.15 12.4 26.73

100 91.67 92.71 81.25

0 12.17 12.13 19.13

Table 9: The correct performance (%) on who- and which-Subject/Object questions

LA controls' performance was on target, while SLI children exhibited low performance on all question types, and particularly oil Object questions. The difference between the two groups' performance is significant for all test structures, i.e., who-S questions [t(7)=-12.691 p<.001], which-S questions [t(22)=-4.089 p<.001], who-0 questions [t(22)=-16.348 jX.001] and which-0 questions [t(22)=-6.831 p<.001]. Let us now consider some asymmetries in the within group performance. LA controls exhibited the worst performance on which-0 questions, while SLI 2

Paola Crisma points out that the wh-question in (6) may be ambiguous between a subject and an object interpretation, due to the fact that SLI children have problems with perceiving case markers, and thus, they may not understand that they should produce an object wh-question. Of course, case marking is among the major grammatical problems that SLI children have (cf. Stavrakaki 1996). However, as shown by the correct performance of the SLI group on morphological case marking (see Table 8), the SLI children of this study can deal with aspects of morphological case. This does not mean that the feature of case is not impaired itself. It may be the case that Greek SLI children, exposed to the PF of a language with rich morphology, can learn the morphological paradigm and the relevant thematic associations: for example, accusative case is associated with patient theta-role (cf. Stavrakaki 2001a,b).

142

STAVROULA STAVRAKAKI

children performed slightly better on which-0 questions than on who-0 questions. However, no significance was found between which-0 questions and who-0 questions for both LA controls [t( 15)=-1.696 p=.l 11] and SLI chil­ dren [t(7)=1.929 p=.095]. Error analysis indicated that the main only error type in which-S questions (N=8) and the main error type in which-0 questions (N=15) in the data from the LA controls is the omission of the NP.3 (7) a. Pjos rinokeros kinijise ton elefanta? [target response] Who-rhino-NOM chase-3SG-PAST the-elephant-ACC "Which rhino chased the elephant?" b. Pjos kinijise ton elefanta? [child's response] Who-NOM chase-3SG-PAST the-elephant-ACC "Who chased the elephant?" In this way, the referential which-NP questions were converted into nonreferential who-0 questions. Note that which and who are phonetically identical in Greek ("pjos"). The drop of LA controls' performance on Object D-linked questions seems to be compatible with studies that were conducted with children (Avrutin 2000), Broca's aphasies (Hickock & Avrutin 1995) and normal adults (De Vincenzi 1991, 1996). It seems that the discourse-related operations involved in the interpretation of D-linked questions make them 'expensive' (Avrutin 2000). By converting the which-questions into whoones, LA controls avoid any extra processing cost and show their preference for purely syntactic operations. Consider now the errors produced by SLI children in the obligatory contexts of which-0 questions. The most frequent error type was the production of which-S questions (13/48) instead of who-0 Questions. (8) a. pjon esproxe I kamila? [target response] Who-ACC push-3SG-PAST the-camel-NOM "Who did the camel push?" b. pion... den boro... pion... de thimame [SLI child's response] who-ACC I-NEG-can... who-ACC- I-NEG-remember "Who... I cannot... who... I do not remember" pja kamila esproxe ton rinokero? Who-camel-NOM-push-3SG-the-rhino-ACC "Which camel pushed the rhino?"

3

There are other error types, not theoretically relevant here. 3 yes/no questions instead of the which-0 ones; also, in who-0 questions obligatory contexts, 3 yes/no questions and 2 who-S questions were produced, while there was no response was for just 2 times.

A-BAR MOVEMENT CONSTRUCTIONS IN GREEK CHILDREN

143

Apart from turning Object-questions into subject ones, SLI children convert non-referential questions into referential ones. They seemed to 'struggle' to build a structure as shown by example 8b, in which the SLI child's first response is "I do not remember". Finally, she produces a Subject question by exploiting referentiality although referentiality is not required in the particular context. In this respect, SLI children exhibited a reverse pattern of performance than that of the LA controls. The distinct pattern of performance exhibited by SLI children compared to that of the LA controls was further confirmed by the following error types, not attested in the LA controls' responses. In particular, case errors (11/48) were also found in the obligatory contexts of who-0 questions: (9)

*pjon xtipise ton rinokero? Who-ACC-hit-3SG-PAST the rhino-ACC

Also, the wh-pronoun (8/48) was produced instead of the whole who-0 construction: (10)

Pjon? Who-ACC "Who?" The same error type, that is, the production of the pronoun instead of the whole question was frequently produced (N=23) in which-S questions, (N=23) in which-0 questions4 and (N=8) in who-S questions.5 The general properties of the SLI performance vs. those of the normal one are presented in Table 10. SLI CHILDREN D Low performance especially on Object questions D Reliance on pragmatics: preference for referential representations D Extreme economical choices (pronoun instead of a question, questions with SVO word order)

LA CONTROLS D High level of performance on all structures D Reliance on syntax not on pragmatics D The syntactic representation of wh-questions is on target.

Table 10: SLI children vs. LA controls

4

Other error types that were attested in which-0 obligatory contexts was the production of which-S questions (4/48: 8.33%) as well as split of the which-NP phrase into which and NP (2/48: 416%) as shown in the example below: * Which did the tiger hit one rhino? 5 Case errors were also attested in Who-S questions: 12/48 (25%).

144

STAVROULA STAVRAKAKI

5. General discussion In sum, the data indicate a quite distinct pattern of performance by both groups. On one hand, LA controls were able to produce correctly the target structures in high percentages. Hence, the presence of the A-bar movement mechanism in their grammar. Notice, nevertheless, that their performance drops only on those relatives where case conflict phenomena occur or the ar­ guments are morphologically underspecified. In this respect, an account postulating that the relative operators are generally subject to maturational delay does not seem to hold for our data. Instead, it seems that associating empty categories with heads specified for different case than that of the empty categories or with heads morphologically underspecified is a process causing some difficulties in LA controls. In Stavrakaki (2001b), it is suggested that those difficulties may not be due to lack of syntactic knowledge, since normally developing children appear to be aware of the syntactic processes involved in operator movement in relative clauses; instead, such difficulties are suggested to be related to parsing preferences, i.e., to PF strategies favoring the association of empty categories with heads morphologically marked for the same case as that of the empty category. The knowledge of A-bar movement by the LA controls is also confirmed by the error types attested in our data. First, there were few instances where the relativization strategy was not used in the obligatory contexts of relatives. Second, the main error type, that is, the omission of the NP, in which-0 questions, where the lowest performance of the LA controls was attested, is rather indicative of the LA controls' preference for purely syntactic operations. On the other hand, SLI children produced a few correct instances of relatives and wh-questions. Since most of them were structures with svo word order, they could not be taken as evidence of operator movement. Apart from the use of the svo word order strategy, SLI children resort to the referentiality strategy, as shown by the conversion of the non-referential Object questions into referential Subject ones. Based on the data from this study (for a summary see Tables 6-10), the SLI performance can be described in formal terms by the following principles: i) apply the operation of Merge as less as possible; ii) avoid A' movement in any case; iii) resort to referentiality in order to build a structure. The above principles, which are descriptive in nature, are taken first to express extreme economy in structural terms and second to denote the preference for contentbased representations. In other words, SLI children seem to cope with their computational limitations by favoring the production of simple and referential structures.

A-BAR MOVEMENT CONSTRUCTIONS IN GREEK CHILDREN

145

Let us now consider the above proposal in more details. Under minimalist hypotheses, Merge is an economical operation itself unlike Move. Suggesting that Merge, i.e., an economical operation, is used as less as possible for reasons of extreme economy may be contradictory in theoretical terms. However, the principle i) above does not necessarily imply impairment in the operation Merge itself; it just denotes that SLI children prefer not to apply it quite often, if possible. Such preference was clearly indicated by extreme economical choices as the production of question pronouns only instead of a simple question. This may be due to the fact that SLI children have also difficulties in accessing words and especially verbs. As argued in Stavrakaki (2000b), SLI children have verb retrieval difficulties which may arise from an impairment in the morphological part of the lexical items and not from the lexical categories themselves. It should be noted that Minimalist syntax holds that verb and nouns are retrieved form the lexicon fully inflected. Therefore, what seems to be costly, to some extent, is the operation of Select by which lexical items are taken from the lexicon. As a result, SLI children prefer to Merge as less as possible, since they avoid the operation of Select in this way. In other words, producing only words, where this is possible, and consequently not using Merge, is rather the result of difficulties in accessing items in the lexicon. The second principle indicates that SLI children cannot use the operation of movement; this is rather due to the lack of knowledge of it. This is clearly indicated by their below chance performance on the structures, whose formation requires the use of the movement mechanism, since they cannot be formed by the use of the default strategy subject-verb-object. That is, the Object questions, and most of the relative clauses. The third principle denotes that due to difficulties with the computational component, SLI children exploit referentially in order to build a structure. This was evident in our data, since which-Subject questions instead of who-0 questions were produced. Consider now the implications of these data for the theories of SLI. Most theories of SLI are based on spontaneous speech data from an early stage of SLI children's linguistic development, when most of the SLI children who partici­ pate in the research projects are at preschool age (Rice et al. 1995; Clahsen 1989, 1990, among others). As mentioned before, what most of these theories share in common is the assumption that some of the features of grammar are severely impaired. As shown by the spontaneous speech data as well as by the correct production of simple structures in the elicitation tasks, the SLI children of this study are able to handle correctly grammatical features, such as, tense,

146

STAVROULA STAVRAKAKI

phi-features of verbs, case etc. while producing simple structures. However, they cannot produce structures formed by movement. Obviously, such performance is far from the predictions of those hypothe­ ses postulating either a global (Gopnik & Crago 1991) or a selective deficit in SLI (Clahsen et al. 1997; Rice et al. 1995) for two reasons. Firstly, the per­ formance of the SLI children of this study is high on those grammatical features (tense, case etc.), where the locus of the deficit is expected to be found. Second, no explicit predictions are spelled out for deficits in the com­ putational component of grammar where the core of the deficit in SLI appears to be located, as shown by our data. Predictions for deficits in the computational component, however, are spelled out by the hypothesis that the SLI children will have problems mainly with non-interpretable features, that is, features irrelevant for semantic repre­ sentation (Tsimpli & Stavrakaki 1999). Within the minimalist framework, non-interpretable features are considered to play a crucial role with respect to the nature of the derivation in the computational system. These features have to be checked at some point of the derivation. Therefore, feature checking is triggered by non-interpretable features and linked with movement under the principle of Last Resort (Chomsky 1995). According to it, only features needing to be checked will move. In this respect, non-interpretable features being responsible for checking operations are explicitly associated with the computational component of language, i.e., narrow syntax, in terms of Chomsky (1998, 1999). However, the non-interpretable features deficit hy­ pothesis cannot account for the correct performance of SLI children on the pro­ duction of the grammatical categories encoding non-interpretable features (e.g., case). Note, however, that Greek SLI children of preschool age have se­ vere problems with the grammatical categories encoding the non-interpretable features of grammar (cf. Stavrakaki, 1996; Tsimpli 2001). Also, it cannot ac­ count for the correct production of simple structures, where feature checking procedures are also applied. Let us now consider whether the data from this study can be accommo­ dated under the RDDR. According to that model, local syntactic relations are likely to be handled correctly, because they can be represented in simple phrase structures but more distant relationships involving operations such as movement, agreement and case assignment will not be adequately specified. It should be noticed that our data fit well within the general predictions of the RDDR model, since most problems of the SLI children are related to distant syntactic relations. The question that arises now is why SLI children are able to use syntactic operations, i.e. the checking operation, in simple but not in

A-BAR MOVEMENT CONSTRUCTIONS IN GREEK CHILDREN

147

more complex structures. It is not so clear how an account postulating optionality in movement operations (van der Lely 1998) could explain the systematicity observed in our data: correct simple structures are produced while structures involving movement are absent. An alternative hypothesis would postulate that the syntactic operations at the level of narrow syntax are altogether impaired and, thus, even the correct production of the constructions with svo word order by SLI children would not reflect the same syntactic procedures with those of unimpaired native speakers. However, such a hypothesis has to account for the reasons that SLI children are able to produce correctly some grammatical categories and simple sentences. Such reasons may be related to development issues, that is, to processes involved in language development in SLI. Let us adopt a developmental per­ spective towards SLI. As mentioned before, two main hypotheses for the de­ velopment of SLI grammars have been formulated. First, Rice et al. (1995) postulate a maturational process of SLI children's language development, that is, the normal developmental path is followed in SLI but with some delay. Second, Paradis & Gopnik (1997) suggest that SLI children do not follow the normal developmental path but must make use of non-grammatical communi­ cative cues and metalinguistic knowledge. Based on the data of this study, the second hypothesis seems to be on the right track. In particular, if the SLI children's linguistic development was sim­ ply delayed, then the same pattern of performance with that of normally developing children would be exhibited, that is, no qualitative differences be­ tween the performance of the two groups would be found. However, as the data of this study indicate (see also Stavrakaki 2001b), SLI children exhibited a qualitatively different pattern of performance than that of the LA controls in many respects (see Tables 6-10 for a summary of the groups' performance). According to the hypothesis put forth by Paradis & Gopnik (1997), it is ex­ pected that individuals with SLI, following a deviant pattern of linguistic de­ velopment, must resemble speakers, who express themselves in a second lan­ guage and construct their utterances laboriously with long pauses between phrases. Further support to the view above is given by findings which show that individuals with SLI speak significantly slower than controls (Toblin et al. 1992). In this respect, individuals with SLI use controlled processes to construct an utterance. In particular, they, do it by the deliberate application of explicit grammatical knowledge. In other words, they use grammatical knowledge that has been stored through the application of explicit learning procedures. Paradis (1994) makes a distinction between automatic and controlled processes used by the speakers. Assuming that SLI children use controlled processes while native

148

STAVROULA STAVRAKAKI

speakers use automatic processes, it could be suggested that the correct production of some grammatical features and simple structures by SLI children might be the result of such processes. In this respect, controlled processes are successfully applied to learning the morphological aspect of the non-interpretable features and constructing simple structures but not to building complex structures. That is, producing structures requiring sophisticated knowledge of complex syntactic operations, for example structures with A-bar movement, seems not to be subject to explicit learning processes. Therefore, while implicit grammatical knowledge is crucial for producing complex syntactic structures, controlled strategies, i.e. the SVO word order strategy, seem to be sufficient for the production of simple utterances. Hence, the difficulties of SLI children in producing complex A-bar movement constructions but not simple ones. The hypothesis sketched above can adequately account for the predomi­ nance of the SVO word order in SLI data. In psycholinguistic terms, the SVO word order is argued to be the least complex structure on the grounds that it permits direct mapping of semantic structure onto syntactic forms (Bever 1970). Based on this assumption, Clahsen (1984) accounts for the development of word order in L2 learners of German by postulating that each word order stage reflects the learner's use of varying combination of three speechprocessing strategies. Learners at the first stage {stage X) use the canonical order strategy (svo), that is, they simply sequence words and phrases not on the basis of any grammatical knowledge but according to the underlying semantic relations within the clause. SLI children seem to follow exactly the same path. When they have to produce syntactic structures that require use of grammatical knowledge at the computational level, they resort to the svo word order strategy, which is the less costly speech processing strategy. Interestingly, the predominance of svo word order was also attested in Greek data from aphasia (Kehayia et al. 1990; Kehayia & Jarema 1991) and second language acquisition (Mangana 2000).6 It, therefore, seems that when speakers do not use automatic processes for whatever reason, they resort to the safe harbor of the svo word order strategy (Bates et al. 1988). To summarize the discussion so far, I have suggested that the overuse of the SVO word order in the data from SLI children has been the result of applying controlled processes in sentence production. SLI children have resorted to such processes due to the fact that they have rather been unable to use those syntactic operations required for structure building, that is, movement and checking operations. In this respect, the deficit in SLI has been considered to lie mainly in

Pointed out to me by E. Kehayia.

A-BAR MOVEMENT CONSTRUCTIONS IN GREEK CHILDREN

149

the computational component of language (cf. van der Lely 1996, 1997, 1999). However, the problems with the computational component have not been attributed to the optionaltiy of movement, as suggested by van der Lely (1998). If movement was optional in SLI grammar, then A-bar movement structures with word order other than that of the svo one would be produced. Following a developmental perspective towards SLI, a more radical proposal has been made here. According to it, all syntactic procedures motivated by the noninterpretable features of grammar, e.g., movement and checking procedures, have been expected to be severely impaired. Nevertheless, it has been suggested that SLI children have been able to learn the morphological aspect of the noninterpretable features due to the application of the explicit learning procedures. For the same reason, they have also become able to produce correctly simple structures, i.e. structures with SVO word order. Such performance has been taken as evidence of the non-automatized use of language. However, the syntactic operations at the computational level of language, which are associated to the non-interpretable features of grammar, have not been considered to be subject to explicit learning. Thus, SLI children have been anticipated to have persisting problems with the computational component of language. In this respect, it has been suggested that syntactic operations constitute the core of the deficit in SLI (cf. van der Lely 1999). 6.

Conclusion In this study, I have presented experimental data on the production of rela­ tive clauses and wh-questions by a group of Greek SLI children and their LA peers. As shown by the data, the SLI performance is qualitatively different than that of the control group. In an attempt to provide a formal description of the SLI performance, the principles underlying that performance have been introduced and analyzed. I have pointed out that a preference for producing simple and referential structures lies in the core of those principles. Finally, I have discussed the implications of this study for the theories of SLI. Assuming that linguistic development in SLI is an explicit rather than an implicit procedure, I have argued that SLI children have problems with purely syntactic operations, for instance with the operation of A-bar movement, due to a severe deficit in the non-interpretable features of grammar. Since syntactic operations do not seem to be subject to the controlled process of explicit learning, I have suggested that SLI children experience persistent difficulties with computational operations. In this respect, I have concluded with van der Lely (1999) that the deficit in SLI seems to lie mainly in the syntactic component of language.

150

STAVROULA STAVRAKAKI REFERENCES

Avrutin, Sergey. 2000. "Comprehension of discourse-linked and non-discourselinked questions by children and Broca's aphasics". Language and the Brain ed. by Yoseph Grodzinsky, Lewis P. Shapiro & David Swinney, 295-313. New York: Academic Press. Battell, Jackie. 1998. Wh-Movement in Children with SLI. Bachelor's thesis, Birkbeck College, University of London. Bates, Elizabeth, Angela D. Friederici, Beverly B. Wulfeck & Larry A. Juarez. 1988. "On the preservation of word order in aphasia: cross-linguistic evidence". Brain and Language 33.323-364. Bever, Thomas. 1970. "The cognitive basis for linguistic structures". Cognition and Development of Language ed. by Jerry Hayes, 97-110. New York: John Wiley. Chomsky, Noam. 1995. The Minimalist Program. Cambridge Mass.: MIT Press. Chomsky, Noam. 1998. "Minimalist Inquiries: The framework". Ms, MIT. Chomsky, Noam. 1999. "Derivation by phase". Ms, MIT. Clahsen, Harald. 1984. "The acquisition of German word order: A test case for cognitive approaches to second language acquisition". Second Languages ed. by Roberts Andersen, 88-110. Rowely, Mass.: Newbury House. Clahsen, Harald. 1989. "The grammatical characterization of developmental dyspha­ sia". Linguistics 27.897-920. Clahsen, Harald. 1991. Child Language and Developmental Dysphasia: Linguistic studies of the acquisition of German. Amsterdam & Philadelphia: John Benjamins. Clahsen, Harald, Susanne Bartke & Sandra Gollner. 1997. "Formal features in impaired grammars: a comparison of English and German SLI children". Journal ofNeurolinguistics 10.151-172. Crain, Stephen & Rosalind Thortnon. 1998. Investigations in Universal Grammar. Cambridge Mass.: MIT Press. Crain, Stephen, Cecile McKee & Maria Emiliani. 1990. "Visiting relatives in Italy". Language Processing and Language Acquisition ed. by Lyn Frazier & Jill de Villiers, 335-356. Dordrecht: Kluwer Academic Press. Crisma, Paola. 1992. "On the acquisition of Wh-questions in French". Geneve Generative Papers 2.115-122. Dalalakis, Jenny E. 1994. "Familial Language Impairment in Greek: Linguistic aspects of familial language impairment". Special Issue of the McGill Working Papers in Linguistics 10 ed. by S. Matthews, 216-227. Montreal: Cahiers Linguistiques de McGill. De Vincenzi, Marica. 1991. Syntactic Parsing Strategies in Italian. Dordrecht: Kluwer Academic Press. De Vincenzi, Marica. 1996. "Syntactic analysis in sentence comprehension: effects of dependency types and grammatical constraints". Journal of Psycholinguistic Research 25.117-133.

A-BAR MOVEMENT CONSTRUCTIONS IN GREEK CHILDREN

151

Goodluck, Helen & Danijela Stojanovic. 1996. "The structure and acquisition of relative clauses in Serbo-Croatian". Language Acquisition 5:4.285-315 Gopnik, Mirna & Martha Crago. 1991. "Familial aggregation of the developmental language disorders". Cognition 39.1-50. Guasti, Maria Teresa. 1996. "Acquisition of Italian interrogatives". Generative Per­ spectives on Language Acquisition ed. by Harald Clahsen, 241-270. Amsterdam & Philadelphia: John Benjamins. Guasti, Maria Teresa. 2000. "An excursion into interrogatives in early English and Italian". The Acquisition of Syntax: Studies in comparative developmental lin­ guistics ed. by Marc-Ariel Friedemann & Luigi Rizzi, 105-128. Harlow, England: Longman. Guasti, Maria Teresa & Ur Shlonsky. 1995. "The acquisition of French relatives re­ considered". Language Acquisition 4.257-276. Hickok, Gregory & Sergey Avrutin. 1995. "Representation, referentiality and processing in agrammatic comprehension: Two case studies". Brain and Language 50.10-26. Kehayia, Eva, Gonia Jarema & Danuta Kadzielava. 1990. "Crosslinguistic study of morphological errors in aphasia: evidence from English, Greek and Polish". Morphology and Phonology in Aphasia ed. by Jean-Luc Nespoulous & Pierre Villiard, 140-155. New York: Springer. Kehayia, Eva & Gonia Jarema. 1991. "On agrammatism in Greek". Neuro-und Patholinguistik 35.47-61. Labelle, Maria. 1990. "Predication, who-movement and the development of relative clauses". Language Acquisition 1.95-119. Labelle, Maria. 1996. "The acquisition of relative clauses: movement or no move­ ment?" Language Acquisition 5:2.65-82. Mangana, Anastasia. 2000. "The acquisition of Modern Greek word order by adult L2 Learners". Paper presented at the International Symposium on Theoretical and Applied Linguistics, Department of English, Aristotle University of Thessaloniki, 20-22 April 2000. McKee, Cecile, Dana McDaniel & Jesse Snedeker 1998. "Relatives children say". Journal of Psycholinguistics Research 27:5.573-596. Paradis, Michel. 1994. "Neurolinguistic aspects of implicit and explicit memory: Im­ plications for Bilingualism and SLA". Implicit and Explicit Learning of Lan­ guages ed. by Nick  Ellis, 371-393. London: Academic Press. Paradis, Michel & Mirna Gopnik 1997. "Compensatory strategies in Genetic Dys­ phasia: declarative memory". Journal of Neurolinguistics 10:2/3.173-186. Perez-Leroux Anna Tereza. 1995. "Resumptives in the acquisition of relative clauses". Language Acquisition 4:1/2.105-138. Pesetsky, David. 1987. "Wh-in-situ: Movement and unselective binding". The repre­ sentation of (in) definiteness ed. by Eric J. Reuland & Alice G.B. ter Meulen, 98-129. Cambridge, Mass.: MIT Press.

152

STAVROULA STAVRAKAKI

Rice, Mabel L., Kenneth Wexler & Paul Cleave. 1995. "Specific Language Impairment as a period of extended optional infinitive". Journal of Speech and Hearing Research 38.850-863. Stark, Rachile & Paula Tallal. 1981. "Selection of children with specific language Deficits". Journal of Speech and Hearing Disorders 46.114-122. Stavrakaki, Stavroula. 1996. SLI in Greek: evaluation of person and number agreement, case assignment to overt subject pronouns and tense marking. MA thesis, University of Essex. Stavrakaki, Stavroula. 1999. "The case of a Greek SLI child: Evaluation of person and number agreement, case assignment to overt subject pronouns and tense marking". Working Papers in Theoretical and Applied Linguistics ed. by Athanasios Kakouriotis & Vasilia Bolla-Mavrides, vol. 6:165-187. Stavrakaki, Stavroula. 2000a. "Eliciting Relatives from Specifically Language Impaired and Normally Developing Greek children". Paper presented at the International Symposium of Theoretical and Applied Linguistics, Aristotle University of Thessaloniki, 20-22 April 2000. Stavrakaki, Stavroula. 2000b. "Verb lexicons in SLI: some experimental data from Modern Greek". Journal of Greek Linguistics 1.93-129. Stavrakaki, Stavroula. 2001a. "Comprehension of reversible relative clauses in Spe­ cifically Language Impaired and Normally Developing Greek children". Brain and Language 11'. 419-431. Stavrakaki, Stavroula. 2001b. SLI in Greek: Aspects of syntactic production and comprehension. Ph.D. Thesis, Aristotle University, Thessaloniki. Stavrakaki, Stavroula & Ianthi Maria Tsimpli. 1999. "Diagnostic Verbal IQ Test for Greek preschool and school age children". Poster presented at the 5th European Conference on Psychological Assessment, University of Patras, 25-29 August 1999. Stromswold, Karin. 1995. "The Acquisition of subject and object Wh-questions". Language Acquisition 4:1/2.5-48. Tomblin, J. Bruce, R. Paula Freese & Nancy L. Records. 1992. "Diagnosing specific language impairment in adults for the purpose of pedigree analysis". Journal of Speech and Hearing Research 35.832-843. Tsimpli, Ianthi Maria. 2001. "Interpretability and Language Development: A study of Verbal and Nominal Features in Normally Developing and SLI Greek Children". Brain and Language 77.432-452. Tsimpli, Ianthi Maria & Stavroula Stavrakaki. 1999. "The effects of a morphosyntactic deficit in the determiner system: The case of a Greek SLI child". Lingua 108.31-86. van der Lely, Heather K. J. 1994. "Canonical linking rules: Forward versus reverse linking in normally developing and specifically language impairment children". Cognition 51.29-72.

A-BAR MOVEMENT CONSTRUCTIONS IN GREEK CHILDREN

153

van der Lely, Heather K. J. 1997. "Language and cognitive development in a grammatical SLI boy: Modularity and innateness". Journal of Neurolinguistics 10:2/3.75-108. van der Lely, Heather K. J. 1998. "SLI in children: Movement, economy and deficits in the computational-syntactic system". Language Acquisition 7:2/4.161-193. van der Lely, Heather K. J. & Linda Stollwerch. 1997. "Binding theory and gram­ matical specific language impairment in children". Cognition 62.254-290. Varlokosta, Spyridoula. 1998. "I kataktisi ton anforikon protaseon" (The acquisition of relative clauses). Studies in Greek Linguistics 18.98-108. Wexler, Kenneth. 1991. "Some issues in the growth of Control". MIT Occasional Papers No 44. Cambridge, MIT: Department of Brain and Cognitive Science. Wilhelm, Andrea & Hanna Ken. 1992. "On the acquisition of wh-questions". Galgary Working Papers in Linguistics 15.89-98.

MORPHOLOGICAL ACCESSIBILITY IN ZULU SUSAN M. SUZMAN University of the Witwatersrand 1. Introduction The vulnerability of morphology in language impairment is well-attested, varying cross-linguistically according to language type (Leonard 1998). Differential access to morphology has been observed by several researchers. In a study of English Specific Language Impaired (SLI) children, Rice & Oetting (1993) found that nominal morphology (plural) was easier to learn than verbal morphology (agreement). Rice, Wexler & Cleave (1995) proposed that tense, but not agreement, is problematic for children with SLI, and results in a prolonged period of optional infinitive use, the Optional Extended Infinitive (OEl) period, when children optionally use uninflected verbs. Beyond specific proposals to explain the differential access to morphology, there have been proposals of a more general nature about how children with language impairment may learn language differently from normal children. Using Pinker's (1984) dual mechanism acquisition model, Oetting & Rice (1993) suggest that children with SLI may show increased reliance on rote-rather than rule-learning. In this paper, I present and discuss two case studies of language impair­ ment in Zulu, one of the 350 Bantu languages spoken in sub-Saharan Africa.1 The South-Eastern Bantu languages are spoken in South Africa. They consist of the Nguni and SeSotho subgroups and Venda and Tsonga, (Figure 1). Lan­ guages within the larger subgroups are mutually intelligible. While other lan­ guages are not as closely related, they have very similar underlying structure (see Suzman 1996, for a developmental comparison). These languages are highly inflected, having both extensive Noun Class (NC) (gender) systems and large numbers of agglutinative morphemes (Weimers 1973). Their abundant morphologies provide a rich source of information for the field of language impairment research and wider issues.

1

In this paper, I refer to these languages as African languages, the term preferred in South Africa due to negative political connotations with the term Bantu.

156

SUSAN M. SUZMAN

Fig. 1 : South-Eastern Bantu Languages

I investigate morphological development, by considering the multiple and diverse rules of morphology: the range of NC, the agreement, and the aggluti­ native morphemes used by normal and language impaired Zulu-speaking chil­ dren. The discussion reflects differential access to morphology depending in part on the obligatory or optional status of morphemes (Herbert 1992) in the language. Aim of this paper is to discuss whether access to core grammar is mediated by language specific organization and representation of basic con­ cepts. 2. Research background Morphological use in language impairment is the focus of this paper. Re­ search into language disorders in Zulu and related languages is just beginning (Demuth & Suzman 1997; Suzman & Tshabalala submitted). There is, how­ ever, a background of normal language acquisition studies against which de­ layed language development can be assessed. Studies by Kunene (1979) on Siswati, Connelly (1984) and Demuth (1984) on SeSotho, Tsonope (1987) on Setswana, and Suzman (1991) on Zulu have all considered, to varying degrees, the acquisition of NC and agreement morphology; they concur in finding early, effortless acquisition of nominal and verbal morphology in these languages. Demuth (1984) and Suzman (1991) found precocious acquisition of the passive and relative in SeSotho and Zulu, a finding that counters later acquisition of these structures in languages like English. An analysis of functional categories in SeSotho within the generative framework is provided by Demuth (1992).2 Zulu is an svo, head-initial, head-marking, pro-drop language; the multiple and diverse rules that morphologies plays in this language are illustrated in (1). 2

Demuth (1992) and Suzman (1991, 1995, 1996) have also considered several other developmental issues.

MORPHOLOGICAL ACCESSIBILITY IN ZULU

157

(la) is a SVO sentence with a head noun followed by modifiers.3 (1) a. Aba-ngane ba-khe aba-de ba-phuz-a u-tshwala NC2-friend POSS2-his ADJ2-tall SM2-drink-FV NC 11-beer "His tall friends drink beer" In (lb) and (lc), there is the absence of an overt subject. The agreement prefix SM functions as the subject of this sentence. b. Um-ngane uthel- -el -a u-mama i-thiye NC 1-friend SMl- pour -BENE -FV NC 1-mother NC5-tea "The friend pours tea for Mother"  Uthel- -el -a u-mama i-thiye SMl- pour -BENE -FV NCla-mother NC5-tea "S/he pours tea for Mother" The range of representative tense, agreement and agglutinative morphemes in the complex verbal is(NEG)-SM-(Tense)-(OM)-VStem-VExtension-FV(NEG). (Id) is a sentence with an inflected verb; this 'complex verbal' is very frequent in child speech. d. Abambhal -is -i NEG- SM2- OMl- write -CAUS -NEG "They do not make him write" The examples in (le) illustrate two patterns of the present tense, whose choice of form depends on transitivity, and noun or pronominal OM form of the object. e. 'short' form with transitive verbs i. Uphuz-a ithiye SMl- drink-FV NC5- tea "He drinks tea" The abbreviations used in the text and in glosses are the following: Adjectival prefix Benefactive verb extension Causative verb extension ContingREM Contingent remote tense COP: Copulative marker DP: Demonstrative pronoun FV: Final vowel INF: Infinitive prefix INSTR: Instrumental preposition

ADJ: BENE: CAUS:

NC: NEG:

Noun class prefix Negative

: PART PAST POSS: PRES PS: REL: REM: SM:

Subjunct: Subjunctl :

Object marker Participial prefix Past tense Possessive prefix Present tenses Possessive stem Relative prefix Remote tense gender marker Subject marker (subject-verb agreement) Subjunctive suffix Subjunctive for NCl

158

SUSAN M. SUZMAN ii.

'long' form with present tense marker yaUyaphuz -a (umu-ntu) SMl-

PRES-

drink

-FV

NC 1-person

"He drinks, the person does" Uyayiphuz-a SMl- PRES- 9- drink-FV "He drinks it" (If) shows that the passive, as in English, fronts the understood object of sen­ tence as well as suffixing a passive verb extension -w to the verb. f. Passive Ithiye liphuz -w -a ng- uMa NC5- tea SM5- drink -PASS FV COP- NCla- Ma "Tea is drunk by Mother" Zulu has 13 noun classes (NCS) that determine extensive agreement across the sentence. Modifiers and verb are brought into agreement with the head noun by means of agreement prefixes. A representative sample of the NC system is given in Table 2 (see the full NC system in Appendix A). Class 1 2 9 10

NC umuabaiNiziN-

Poss wabayaza-

ADJ

REL

omabaeNeziN-

0-

abaeezi-

SM ubaizi-

OM mbayizi-

DP lo laba le lezi

Example umu-ntu aba-ntu in-kukhu izin-khuku

Gloss "person" "people" "chicken" "chickens"

Table 2: Representative noun classes and agreement in Zulu

The NC prefix is a marker of noun class and number, the majority of NCs occurring in singular-plural pairs, as seen in NCs discussed in Table 2. The system is variably alliterative, as NCs 1 and 2 show. (1) above illustrates how the NC prefix determines agreement across the sentence. The verb is inflected for number, tense, aspect and agreement as well as a number of negative morphemes, verb extensions, mood, and sentence subordinators. Verb extensions alter or augment the verb's argument structure and in­ clude the benefactive (BENE) in (lb), the causative (CAUS) in (lc), the passive (PASS) in (If), the neuter (NEUT) -ek and the reciprocal (RECIP) -an. Normal children acquire the causative and benefactive by age 2;6 followed by the pas­ sive and reciprocal (Suzman 1991).

MORPHOLOGICAL ACCESSIBILITY IN ZULU

159

3. Sipho and Nompumelelo The subjects of the research are two Zulu-speaking children, Sipho and Nompumelelo. They had both been recognized by family members as having language problems and were subsequently diagnosed as language-impaired professionally. Sipho was professionally diagnosed as language impaired with no attendant hearing or medical problems; however, he did not receive therapy because of the distance from his home to a speech clinic. He was interviewed at ages 2;7 years and 3;7 years at home, his grandmother's accommodation at the townhouse where her employer lived. He was a gentle, shy, co-operative, and wellbehaved child who led a rather isolated life with his grandmother, and saw his mother on the weekend. They occasionally visited the maternal family in Natal and, following these visits, he appeared to be more confident and talked more, as his grandmother's employer reported. At 3;0 years, he started attending a multicultural playschool where teachers spoke English and assistants spoke Zulu and SeSotho to the children who spoke African languages. The second speech sample was taken 7 months after he started school. At 2;7 years, Sipho's spontaneous speech was limited. His discourse strategy was characterized by repetitions of input, a feature often associated with receptive disorders. He did not initiate conversation, answer questions or maintain conversational topic. His words were oddly truncated, e.g., cvcv > cvc, CV. However, he used selected aspects of the noun class and agreement system productively. At 3;7 Sipho's speech showed considerable develop­ ment. He initiated some conversation and talked well on his own, although he tended to go off on his own tangent. He used several Tense Aspect Markers (TAMS), alternated OMS with full noun objects and used simple sentences occasionally with more complex semantic intention. His speech showed a number of non-standard complex verbals. His social interaction was more normal than it had been at 2;7, a development that, together with his speech, initially led to an interpretation that Sipho was at the lower end of normal language development, and that his social isolation contributed to his delayed speech. However, comparing his speech with that of an older language impaired child, revealed similarities between the two children's speech. The similarity strengthened the interpretation of his speech as language impaired (Suzman & Tshabalala submitted). Nompumelelo was a 5;6-year-old girl who had been adopted into her family (parents and children) at 8 months. She was a lively, interactive, outgoing child who enthusiastically entered into all activities. She had a history of minor medical problems, drooling, ear infections, and delayed perceptual and

160

SUSAN M. SUZMAN

motor development. She was brought to the Department of Speech Pathology and Audiology at the University of the Witwatersrand because of her parents' concern with her language development in light of her imminent school atten­ dance. She was diagnosed as having immature speech with some phonological and morphological errors, word-finding difficulties and poor topic mainte­ nance. No neurologically-triggered language problems were found and her hearing was normal. Nompumelelo's speech was voluble, at times incoherent. In her enthusiastic responses, she became quite excited and talked very fast, sometimes stuttering and repeating partial utterances. This was most noticeable in talking on her own and most controlled in ordinary conversa­ tional interaction. She told a structured story, but adhered less to the story line than normal children. Comparison between the two children was partial, given their personality and age differences. Nompumelelo was older and had greater language com­ petence than Sipho. An earlier comparison of Sipho with an age-matched nor­ mal child (Demuth & Suzman 1997) showed that his speech was restricted compared to normal development. Impressionistically, Sipho's speech was re­ stricted, whereas Nompumelelo's was disfluent. According to Suzman & Tshabalala (2000), however, shared properties of the children's speech may allow to identify a syndrome of language impairment in Zulu. Data collection procedures were similar, naturalistic, unstructured conversational interactions between the children and their caretakers and/or the researcher. In addition, Nompumulelo was given a story-telling task. Conversations were tape re­ corded and transcribed. Representative samples of the two children's speech are given in Appendix B. 4.

Analysis Word and morpheme Mean Length of Utterance (MLU) were calculated for the two children and a younger normal control, Nqoba, a child of 3;5 years of age with a loquacious speech style like Nompumelelo. Comparisons were made of 55 word speech samples. MLU

Word Morpheme

Nompumelelo Language-delayed girl, 5;6 years. Narrative 3.87 (213/55) 7.50 (412/55)

Sipho Language-delayed boy 3;7 years. Conversation 1.73 (95/55) 4.2 (231/55)

Table 3 : Word and morpheme ML U counts

Nqoba Normal girl, 3;5 years. Conversation 3.1 7.08

(176/56) (397/56)

161

MORPHOLOGICAL ACCESSIBILITY IN ZULU

Table 3 shows that Sipho had the lowest MLU of the children, slightly un­ der 2 words. His non-imitative speech was comparable to normal children between 2-2;6 years both in length and content. He asked simple questions, and his complex verbals typically contained 3 morphemes. As will be seen, the absence of certain types of morphology in his speech might be due to length limitations on what he said. This was not the case with Nompumelelo, whose word and morpheme MLUs were comparable to the normal younger child, Nqoba. Low MLU on its own provides a diagnostic of language disability (Leonard 1998), but these case studies suggest that low MLU may be optionally associated with language impairment but is not diagnostic in Zulu. More case studies are needed to test the importance of the MLU findings. 4.1 Lexicon The results of a comparison of the number and frequency of words in the major word categories, noun and verb, are in Table 4. Verbs Nouns

Nompumelelo .387 (24/62) .429 (26/61)

.44 .50

Sipho (20/45) (15/30)

.620 .607

Nqoba (36/58) (31/51)

Table 4: Type-token ratios in major lexical classes

The language-delayed children had less elaborate and more repetitious vo­ cabularies than the normal child Nqoba, as the lower type/token ratios of these children indicate. The type-token ratio for verbs is particularly interesting in this respect, indicating that whatever the topic, the normal child had access to greater resources for describing it. However, more data are also needed to estimate the importance of this finding. Nompumelelo was telling a story in­ volving a single actor engaged in somewhat similar activities, a factor that may account for her type/token ratio. Sipho had restricted speech that may be partly a consequence of being shy and living in an isolated environment. 4.2 Morphology The children varied considerably in the extent to which they used mor­ phology, due partly to their age differences. Sipho's spontaneous speech re­ vealed a restricted but productive NC and agreement system, NCs 1 and 9, and their respective SMs at age 2;7 with some evidence of development at 3;7, as shown below in Table 5. He over-generalized rather than omitted morphologi­ cal markers, a finding consistent with younger normal Zulu children (Suzman 1991) and typical of languages with obligatory morphology (Leonard 1998). Naturalistic data do not preclude a more extensive morphology for Sipho than that represented in his speech samples.

162

SUSAN M. SUZMAN

Marker Totals NCPrefix Poss ADJ REL

SM OM

DP

Nompumelelo, 5;6 144 (54) 38% 6% (9) 2% (3) (1) — (54) 38% 2% (3) 14% (20)

Sipho, 3;7 75 33% (25) .04% (3)

— —

— —

57%

(43)

—

(1) (3)

.04%

Nqoba, 3;5

|

150 (56)

37% 5% 3% 3% 39% 14% 13%

(8) (4) (4) (26) (21) (19)

Table 5: Noun class and agreement morphemes

Profiles of the older language-delayed child and of the younger normal control were similar with the exception of the use of OMs and RELs. Neither language-delayed child used either of these markers with any frequency, but otherwise their speech provided evidence of productive NC and agreement morphology. Table 6 summarizes verb morphology used by the three children. Tense Present yaPast-e, -ile Future yoRemote Conting remote Totals Agglutinative Verb part. SeVerb extensions -is -el -iw

Nompumelelo 12 1 23 13 49

Sipho 11 24 1 5 1 42

Nqoba 11 6 6 11

1

—

1

—

1 1

2 2 2

1

2

7

|

34

-ek | Totals

Table 6: Frequency counts of verbal morphology

The profiles of tenses and optional agglutinative morphology are quite dif­ ferent. All three children used a range of tenses and rather fewer agglutinative morphemes. The actual tenses they used reflected what they were talking about. In Nompumelelo's case, she was telling a story and using the Remote

MORPHOLOGICAL ACCESSIBILITY IN ZULU

163

Past tense appropriate for relating a story that had happened some time ago. Sipho's samples included an account of another child's bad experience that he told in several versions in the past tense. Nqoba used mainly the present tense to talk about ongoing activities in the immediate context. More interesting is the difference in use of verb extensions. Although numbers are low, the normal child produced several verb extensions lacking in the speech of the language-delayed children. A larger speech sample con­ firmed her productive use of these morphological devices for extending argu­ ment structure of sentences. Summarizing morphological differences, all children used nominal and verbal morphology. In some respects, children's morphological profiles were very similar, namely in their use of NC prefixes, SMs and DPS. However, results showed differences between normal and language impaired children in OMs, RELs, ADJs and Verb Extensions, morphemes that were both optional and contribute to conceptual complexity in children's speech. 4.3 Syntax With two exceptions of infinitives, Sipho's utterances at 3;7 were simple sentences. The majority were Verb-Subject constructions, e.g., u-ya-dlala um-fana SMl-PRES-play NC 1-boy "he plays, the boy does" and transitive sentences, e.g., um-fana u-dlala i-bhola NCl-boy SMl-play NC5-ball "the boy plays ball". Sipho formed new sentence types building upon early, learned forms that were reported on in Demuth & Suzman (1997). At 2;7 he used frequently iya- and a verb or predicate. Although iya- in Zulu is a SM9-Present tense construction, he used it as an unanalyzed form to create new sentences as in (2). (2)

Sipho a. iya-phuza ???-drink b. iya-m-sula ???-OMl-wipe  iya-phansi d. iya-baba

target ngi-ya-phuz-a 1 sg.-PRES-drink-FV ngi-ya-yi-sul-a lsg-PRES-OM9-V-FV ngi-phansi ngu-baba

Gloss "I drink" "I wipe it (floor)" "I am (sitting) down" "It is Father"

Sipho presents atypical complex verbals. Forms of iya persist in sentences at 3;7 years, as in (3).

164

SUSAN M. SUZMAN (3)

Sipho Hamb -a a. Ayiya-FV NEG- SM9- PRES- go "He doesn't go" b. Ayishek -a NEG- SM9- shake -FV "She (mother) doesn't shake"  U afun' u-ngen-a SMl- PRES- want INF-enter-FV "She wants to enter" d. Iy a- phuz' ama-nzi SM9- PRES- drink NC6-water "It drinks water" e. Iya- dl -e SM9- PRES eat -PAST "She ate"

Target Ayi-

hamb

NEG- SM- go

A-

y i-

s hek

NEG- S M 9 - shake

-i -NEG

-i -NEG

U- fun ' uku-ngen-a SM 1 - want' INF-enter-FV Iphuz' ama-nzi SM9- drink NC6-water /dl -He SM9- eat -PAST

These verbs share the elements SM-ya-V, e.g., i-ya-hamba "he goes", one of the first simple sentence types seen in young children's speech. Sipho used the present tense in his basic structure in the negative (3a), the infinitive (3c), the transitive (3d), and the past (3e), none of which have the present tense marker in Zulu. This type of error occurs occasionally in young normal children's speech between 18 months and 2;4 years but Sipho's speech provided evidence of prolonged reliance on early forms in formulating new sentences. As stated earlier, normal Zulu- and SeSotho-speaking children produce passives as early as 2;6 years (Demuth 1984, Suzman 1991). However, the passive is rarely used by the two language impaired children in this study. Sipho had the concept, word order differences, but lacked the appropriate passive verb morphology, as it appears in (4). (4)

Sipho talked about someone being bitten by the toy horse and fish Sipho Target a. U- lum -ile i-hhashi U-luny-w-e -i-hhashi SMl- bite -PAST NC5-horse SMl-bite-PASS-PAST COP-NC5-horse "He bit, the horse did" "He was bitten by the horse" b. U- lum -ile i-fish U-luny-w-e y-i-fish SMl- bite -PAST NC5-fish SMl-bite-P ASS-PAST COP-NC5-fish "He bit, the fish did" "He was bitten by the fish"  Iya- hamb -a i-hhashi SM9- PRES- go -FV NC5/9-horse "He goes, the horse does"

MORPHOLOGICAL ACCESSIBILITY IN ZULU

165

Sipho did not use the passive verb extension -w. However, the non-linguistic context and the morphological contrast between (4a) and (4b), and (4c) show that (4a) and (4b) have passive meaning for him. When he talked about the person (object) being bitten by the horse, the SM agreed with the object (SM1). When he talked about "the horse (subject) going" (4c), the SM agreed with the subject (SM5/9). Nompumelelo also used primarily simple sentences, but there were many instances of complex sentences in a larger data sample (5). The most frequent of these were compound sentences linked by "and then" (5a) and serial con­ structions (5b). Like Sipho, she relied on a few learned forms, producing not so much deviant forms, but simple svo sentences strung together with or without a fixed conjunction. Other complex sentence types occurred infrequently and are illustrated in (5c)-(5f). The absence of these sentences distinguished Nompumelelo's speech from younger normal children who use infinitives between 2 - 2;6 years and other complex sentences between 2;6 and 3 years. (5)

Compound sentences with aqeda lapho "and then" a. Aceda lapho wayopheka waceda lapho wayofun 'khaya "And then she went to cook and then she was looking for a home" Serial constructions used productively b. Abantwana baye bahamba no Mama "The children went and walked with Mother" Infinitives c. Waceda lapho aceda wayofun ' u-casha la (waqeda lapho aqeda wayofun' uku-casha la) "And then, and then, she was looking to hide there" Consecutive actions marked by present subjunctive suffix -e d. Bamshaya akhal -e they hit her and Subjunctla- cry -subjunct "They hit her and she cried" Concurrent actions marked by participial prefix (irregular) on Noun classes la, 2, and 6. e. a- buy -e bekhal -a SM2- Return -PAST PART2- cry -FV "They returned crying" Relative clause f. Lo-mama wa-shäy-a lo o-gqcok' im-pahla zakhe DP1-mother REM1-hit-FV DP1 REL1-dress' NC1lO-clothes POSSlO- 3PS "This mother hit this one who dressed in her clothes"

166

SUSAN M. SUZMAN

The absence of the relative marker was further probed in two structured tasks, both having picture recognition of relativized complex sentences, instructions and paraphrases. (6)

Instruction a. Ngi-bon-is-e in-tombazane e-khomb-a phezulu lsgOM-see-CAUS-subjunct NC9-girl REL9-point-FV above "Show me the girl who points upwards" Paraphrase b. Na-yi. Um-ntwana lo ukhomb-a eceleni. Here-SM9. NCl-child DPI SMl- point-FV on the side. "Here she is. This child points on the side" Nompumelelo rephrased the relative clause as two simple sentences. Results on 25 input relative clauses are given in Table 7. While the child did use the relative marker, the majority of the paraphrases are simply simple sentences. Total input sentences

Responses: 2 simple sentences

25

Responses: Relative clauses 8 (36%)

16(64%) Table 7: Paraphrases of relative clauses

The weakness in relativizing was further demonstrated in a repetition task (Table 8) where the child demonstrated less ability in repeating this type of sentence than other sentence types. Utterance type Verbal complexes Relatived S V + Caus/Bene N + Adjective

Accurate 13 100% 3 43% 5 100% 5 100%

Errors

— 4 SS (57%)

— —

Total 13 7 5 5

Table 8: Repetition task

Nompumelelo did not use relative clause markers in the majority of the rela­ tive sentences, although, interestingly, she had no difficulty with adjective modifiers that had a similar or identical form. Normal children tend to learn adjectives as part of rote-learned expressions but relatives are rule-governed. These children used early, learned forms in new sentences, producing non­ standard forms or simple sentences rather than complex ones. They did not appear to recognize the role played by certain morphological markers in altering verb arguments (Sipho's passive) and in sentence embedding (relatives). The use of familiar and stereotyped structures is common in language delay. The

167

MORPHOLOGICAL ACCESSIBILITY IN ZULU

language-specific consequence in Zulu was compromised development of morphologically expressed complex syntax. Table 9 summarizes differential access to Zulu morphology. Grammatical Morphology Selected NCs SMs Adjs Tense OMs Vextensions Relative clause markers [ Complex Sentence morphology

Normal X X X X X X X X

Zulu Impaired X X X X Rare Rare Rare Rare

|

Table 9: Normal and language impaired morphology

In Table 9, tenses and some NC morphology were used by all children. Less robust morphology included OMs, VExtensions and complex sentence markers (RELs, participials and subjunctives). The OM's absence may be explained by its lack of salience (SM-(Tns)-OM-V) preceding the verb, comparable to Italian object clitic pronouns (e.g., e poi lo hacia "and then kisses him"; Leonard 1998:92) that are also omitted. The other missing morphemes are associated with the expression of more complex propositions, being either verb extensions or morphemes that embed a subordinate sentence within a complex sentence. Summarizing morphology and syntax results, language impaired children have productive noun class and agreement morphology, but lack many grammatical morphemes signaling structural or conceptual complexity. Their absence results in prolonged reliance on early, structurally simple constructions and avoidance or lack of perception of morphological markers of syntactic complexity. 5. Discussion Grammatical morphology in English involves a handful of morphemes while in Zulu it consists of substantial numbers of grammatical and agglutina­ tive morphemes that combine in extremely varied ways. Pinker (1994) estimates the possible combinations of morphemes on an 8-morpheme verb in Kivunjo, another Bantu language, as approximately half a million. A similar projection can be made for the African languages of Southern Africa. Abundant mor­ phology is predicted to both focus children's attention on morphology and to present learning problems. Research bears out these predictions. Children have

168

SUSAN M. SUZMAN

access to some NCs and agreement markers and have problems with agglutina­ tive, anaphoric morphology. Normal children develop NCs and agreement before verb extensions and OMs (Suzman 1991). Language impaired children use NCs and agreement but generally omit agglutinative morphemes. Aphasics make more mistakes on agglutinative morphology than on NC prefixes and agreement (Herbert 1992). Some selective access to Zulu morphology cleaves along the lines of Herbert's distinction between obligatory and optional prefixes in Afri­ can languages. Based on traditional discussions of Bantu morphology, he notes that NCs and agreement prefixes are obligatory. They do not contribute meaning to nouns and agreeing constituents and their omission generally produces ungrammatical utterances. Verb anaphors (and verb extensions) are optional; they contribute meaning to the sentence and their omission produces incorrect but grammatical sentences. Herbert observes that the two sets of morphemes are not fully comparable. Their lack of comparability is reflected in different patterns of acquisition and breakdown. Acquisition sequences, persistent omissions in lan­ guage delay and difference in type and rate of error by aphasics lend credence to the obligatory/optional distinction. It predicts early acquisition of NC prefixes and SMs compared to OMs. While OMs belong to obligatory morphology, the NC and agreement system, they behave like optional morphemes. Besides being op­ tional, they contribute meaning to an utterance and their absence does not pro­ duce an ungrammatical utterance. The following examples illustrate. The OM is a pronominal object when used alone (u-ya-ba-bona SMl-PRES-OM2-see "he sees them") and a emphasis or definiteness marker when used with a full noun object (u-ya-ba-bona abantwana SM1-PRES-OM2 -see Nc2-child "he sees the children" vs. u-bona abantwana "he sees children". The OM over-generalizes like the SM and is acquired after the NC prefix OM and Poss by normal children (Suzman 1991). It is used infrequently by language impaired children and omitted in repetition tasks by aphasies. Suzman (1991) proposed that SMs and OMs over-generalize because they are learned first as general referential pro­ nouns. Subsequently, the SM becomes part of the obligatory prefix system whereas the OM remains an optional pronominal object. The OM is acquired later because it is less likely to occur. As mentioned earlier, its word position also makes it less perceptually salient than the SM. The relative marker is another exception to the predictions made by the obligatory/optional distinction. Al­ though this issue needs more investigation, its absence in language impairment appears to be associated with sentence embedding. Relative markers on relative stems that function like adjectives tend to be rote-learned (in frequent modifying expressions) in normal acquisition (Suzman 1991). RELs in relative clauses align with VExtensions that add semantic complexity to the sentence. While

MORPHOLOGICAL ACCESSIBILITY IN ZULU

169

VExtensions are not prefixes, they are meaningful and optional and profile like OMs and RELs in delayed acquisition. While the issues are complex, the functional status of morphemes within Zulu partly accounts for patterns of acquisition. Comparability of morphemes on a functional basis may be of relevance cross-linguistically. Superficially, Zulu and English findings are conflicting. From the study of plurals and agree­ ment in English SLI children, Rice & Oetting (1993) proposed that nominal morphology is easier to learn than verbal morphology for children. According to the Zulu data discussed in this paper, nominal and verbal morphology are equally accessible. There is no evidence that there is any difference in acquisi­ tion difficulty between nominal (NC) and verbal (SM) inflection. The fact that the NC prefix is more likely to be omitted than the SM in young children's speech might suggest that verbal morphology is easier than nominal morphology. How­ ever, the invariable inclusion of the NC prefix on nouns following the verb, e.g., bhek'umu-ntu look at'NCl-person "look at the person", indicates that processing factors account for omission of the NC prefix. Ric, Wexler & Cleaves's (1995) proposal that late acquisition of tense, but not agreement, is diagnostic for SLI children is not supported by the Zulu study. Normal and language-delayed chil­ dren have little difficulty with tense, a morpheme that profiles like obligatory morphemes. There is no period of uninflected verb usage by Zulu children com­ parable to the Optional Extended Infinitive (OEl) period found in English SLI children's speech. The lack of difficulty of number, tense, and agreement displayed by Zulu language-impaired children contrasts with the persistently problematic nature of these morphemes and categories in English. The obliga­ tory/optional distinction would predict late acquisition of English grammatical morphology. Concepts like English agreement profile more like optional mor­ phology than Zulu agreement does. It does not have the same morphological status as Zulu agreement. The comparison of Zulu and English suggests that there are language-specific differences in the organization and representation of functional categories. If they do not have the same function and status across languages, they may not be directly comparable. Other aspects of NCs and agreement to consider are its systematicity and its alliterative nature. Zulu NCs and agreement prefixes are also systemic; they are integrated in an exceptionally regular, pervasive, 'closed' system (Levy 1996). She attributes the surprisingly early acquisition of morphological systems like gender in Hebrew, English, German, Polish, French, and Russian, to their being closed, their rules requiring no input from other systems. Further, the system is variably alliterative. Evi­ dence for the significance of this factor is relatively minor to date. Herbert (1992) found that aphasics could repeat agreement morphemes more accurately

170

SUSAN M. SUZMAN

when utterances contained full noun subjects than when had pronominal subjects, e.g., isilwane si-ba-lum-ile NC7-animal SM7-OM2-bite-Past "The ani­ mal bit them" versus siba-lumile SM7-OM2-bite-Past "It bit them". Normal chil­ dren also provide more accurate SMs in conversations with nominal rather than pronominal discourse-established topics. Languages like Zulu hold research potential in the study of homonyms, morphophonological patterning, contrasts (Carstairs-McCarthy 1994), and other system-internal features in acquisition and breakdown. For example, Nompumelelo employs adjective agreement with adjective stems, but does not use the morphologically similar relative agreement markers with relative clauses (see Appendix A for a comparison of AD J and REL forms). Further re­ search will investigate the relative contribution of factors identified in the discussion above, semantics, frequency, systemic and alliterative properties, in explaining patterns of acquisition and breakdown. Beyond morphological issues lies the general concern with explaining dif­ ferent acquisition patterns in normal and delayed language acquisition. In Pinker's (1984) dual mechanism model of acquisition, rule and associative learning dependent on input frequency account for language acquisition. Oetting & Rice's (1993) observation of increased reliance on rote-learning in SLI is supported in the Zulu data. As discussed above, Sipho's and Nompumelelo's speech samples provide ample evidence of reliance on learned forms and lack of progressive restructuring of their grammars along the lines of the adult language. Predominantly simple sentences, often stereotyped, and/or non­ standard sentences, are characteristic. 6.

Conclusions This study of Zulu language impairment confirms findings about how lan­ guage impaired children learn language. It also supports findings of differential access to morphology. Its findings do not support more specific proposals con­ cerning ease of acquisition of particular functional categories and nominal ver­ sus verbal morphology. It identifies factors that partly predict differential access to morphology in Zulu and suggests that they might be of crosslinguistic rele­ vance in explaining developmental acquisition. The view expressed in this paper is that access to core grammar is mediated by language specific organization and representation of basic concepts. Specifically, the status of functional cate­ gories in particular languages influences their accessibility in acquisition. Or­ ganizational principles are anticipated to be general enough to provide general predictions for acquisition across language types and provide an intermediate level of analysis relating language specific and universal properties of language.

MORPHOLOGICAL ACCESSIBILITY IN ZULU

171

REFERENCES Bishop, Dorothy. 1994. "Grammatical errors in specific language impairment: Com­ petence or performance limitations?". Applied Psycholinguistics 15.507-550. Carstairs-McCarthy, Andrew. 1994. "Inflection classes, gender and the principle of contrast". Language 70.737-788. Connelly, Michael. 1984. Basotho Children's Acquisition of Noun Morphology. Ph.D. dissertation, University of Essex. Demuth, Katherine. 1992. "Acquisition of Sesotho". The Crosslinguistic Study of Language Acquisition ed. by Daniel I. Slobin, vol 3, 557-639. Hillsdale, N.J.: Lawrence Erlbaum Associates. Demuth, Katherine. 1992. "Accessing functional categories in Sesotho: Interactions at the morpho-syntax interface". The Acquisition of Verb Placement. Functional categories and V2 phenomenon in language development ed. by Jürgen Meisel, 83-105. Dordrecht: Kluwer Academic Publishers. Demuth, Katherine & Susan Suzman. 1997. "Language Impairment in Zulu". Boston University Conference on Language Development Conference Proceedings 1, ed. by Elizabeth Hughes, Mary Hughes & Annabel Greenhill, 124-135. Somerville, MA: Cascadilla Press. Doke, Clement L. 1954. Zulu Grammar. Cape Town: Longmans. Herbert, Robert K. 1992. "Patterns in language change, acquisition and dissolution: noun prefixes and concords in Bantu". Anthropological Linguistics 33:2.103-134. Kunene, Euphrasia  1979. The Acquisition of Siswati as a First Language: A mor­ phological study with special reference to noun prefixes, noun classes and some agreement markers Ph.D. dissertation, University of California, Los Angeles. Leonard, Lawrence B. 1992. "Specific language impairment in three languages: some crosslinguistic evidence". Specific Speech and Language Disorders in Children: Correlates, characteristics and outcomes ed. by Paul Fletcher & David Hall, 118-126. London: Whurr Publications. Leonard, Lawrence B. 1998. Specific Language Impairment in Childhood. Cam­ bridge: MIT Press. Levy, Yonata. 1996. "Why are Formal Systems Early to Emerge?". Children's Lan­ guage ed. by Carolyn Johnson & John H. Gilbert, 75-84. Hillsdale, N.J.: Lawrence Erlbaum Associates. Oetting, Janna . & Mabel L. Rice. 1993. "Plural acquisition in children with Specific Language Impairment". Journal of Speech and Hearing Research 36.1236-1248. Pinker, Steven. 1984. Language Learnability and Language Development. Cam­ bridge, Mass: Harvard University Press. Rice, Mabel L. & Janna . Oetting. 1993. "Morphological deficits of children with SLI: Evaluation of number marking and agreement". Journal of Speech and Hearing Research 36.1249-1257. Rice, Mabel L., Kenneth Wexler & Patricia Cleave. 1995. "Specific Language Im­ pairment as a period of Extended Optional Infinitive". Journal of Speech and Hearing Research 38.850-863.

172

SUSAN M. SUZMAN

Suzman, Susan M. 1991. Language Acquisition in Zulu. Ph.D. dissertation, Univer­ sity of the Witwatersrand. Suzman, Susan M. 1995. "The discourse origin of agreement in Zulu". The Complete Linguist. Essays in Honour of Patrick Dickens ed. by Anthony Traill, Rainer Vossen & Megan Biesele, 319-337. Cologne: Rüdiger Koppe. Suzman, Susan M. 1996. "Acquisition of noun class systems in related Bantu lan­ guages". Children's Language ed. by Carolyn Johnson & John H. Gilbert, 87-104. Hillsdale, N.J.: Lawrence Erlbaum Associates. Suzman, Susan M. 1998. "Morphology as a diagnostic of Language Impairment in +/- Pro-drop languages". Paper presented at the Annual Meeting of the Linguistic Society of America, New York, Jan. 1998. Suzman, Susan M. & Busi Tshabalala submitted. "An investigation of language im­ pairment in Zulu". The South African Journal of Communication Disorders. Tsonope, Joseph. 1987. The Acquisition of Setswana Noun Class and Agreement Morphology: With special reference to demonstratives and possessives. Ph.D. dissertation, Buffalo: State University of New York (SUNY). Weimers, William E. 1973. African Languages Structures. Berkeley, Calif: University of California Press. Appendix A: Morphological Structure of Zulu NC 1 la 2 2a 3 4 5 6 7 8 9 10 11 14 15

NCPrefix

Poss

Adj

Rel

SM

OM

DP

umuu-

wawababawayalaasazayazaIwabakwa-

omomabaabaomemieliamaesieziNeNeziNoluobuoku-

0-

uubabauiliasiziizilubuku-

mmbabawuyiliwasiziyizilubuku-

To

aba00-

umimi iamaisiizii(N)izi(N)u:ubuuku-

0-

abaabaoeeliaesiezieezioluobuoku-

lo laba laba lo le leli Iowa lesi lezi le lezi lolu lobu lokhu

MORPHOLOGICAL ACCESSIBILITY IN ZULU

173

Appendix B: Speech Samples of Sipho and Nompumelelo S[ipho], 2;7 years, picks up a small broom and sweeps the floor with G[randmother] and R[ose], 11 years. Target forms are given in parenthesis. S.

R.

R.

S.

I(ngiSM1sgI(ngilsgFaka put Fak' (faka put WSMlNahereNahereNa-y' Here-SM9'

yayaPresfun' fun' want phakathi inside phathi phakathi) inside enza do

myiOM9-

amaama NC6

sula sula) wipe nzi -nzi) water

"I am wiping it (floor)"

"I want water"

"Put (the water) inside" "Put (it) inside"

-n' -what

uNCla-

gogo? grandmother

"Here it (=cow) is"

yi SM9 ' SM9' inNC9-COW

"What is Granny doing?"

inNC9komo

komo cow

"Here is the cow" "Here is the cow"

Sipho, 3;7years Yi-ni COP-what INC9G.

S.

u2sgBaSM2-

S. S. S.

WaREM1WaREMlUSMl-

le? DP9?

mali money zoFutthath take shis ' burn yi-

Bheka lapha. Look here.

"What is this? Look here." "Money"

thenga

-ni

OM9-

buy -what mali -e i-PAST NC9- money imali NC9money shisa burn

shis burn

-il' -PAST

inNC9-

nge-mali? Instr-money le

"What will you buy with the money?" "They took this money"

DP9

le

"He burned this money"

DP9

"He burned it (money)" gubo blanket

"He burned the blanket"

174

SUSAN M. SUZMAN

Nompumelelo, 5;6 years. Excerpt from telling the Vava story. Glosses refer to the child's form, not the correct Zulu form, except when the child's form is incomplete. Ba(Ba-

yaya-

hamba hamba

nna-

abaaba-

ntanaa ntwana

ba ba-

SM2PRESgo "They go with their children"

Prep-

NC2-

child

POSS2- 3PS

Lo ?

Vava

lo

I-

ya-

phuma

DPI

Vava

DPI

SM9-

PRES-

go out

khe)

Iyaphuma la ifasitelo SM9PRESgo out there NC5window "This one? Vava this one. She goes o u t . . . she goes out there, the window" Za-

fika la izi-... (Afika la ama-... REM 10arrive there NC 10... "They arrived there, the mice did"

iiamaNC 10-

gundane gundwane) mouse

Nayi indiu bayahleka. ba... le inje Here- SM9 N C 9 - house S M 2 - PRES- Laugh S M 2 . . . D P 9 N C 9 - dog "Here it is, the house. They are laughing (children) They.. .this one, the dog" I-

hlak'

hlek -el -a (ISM9laugh BENE FV "She laughs with her children"

aba-

ntana

ba

abaNC2-

ntwana child

bakhe?) POSS2- 3PS

LANGUAGE PRODUCTION IN JAPANESE PRESCHOOLERS WITH SPECIFIC LANGUAGE IMPAIRMENT TESTING THEORIES

YUMIKO TANAKA WELTY, JUN WATANABE & LISE MENN* 1. Introduction: Japanese as a test of su theories Over the last two decades of the 20th century, theoretical debates about the innateness of grammar have increasingly focused on certain developmental language disorders in children: those which cannot be ascribed to brain damage or overall cognitive insufficiency. Such disorders are currently referred to as Specific Language Impairment (SLI). Researchers reason that if there is such a thing as innate universal grammar - assumed to consist of specific templates for rules and representations (including grammatical categories) - then Specific Language Impairment might be explained by deficits in such templates (Rice & Wexler 1996). Conversely, if such explanations are indeed adequate to describe the data, this may be taken as support for innate grammar. Some general claims are implicit in these arguments. First is the etiological claim: that SLI is a unitary disorder with a single cause. In order to reason across combined evidence from all cases, this claim is necessary; the alternative is to consider that there may be many subtypes of SLI, with different causes. If there are many subtypes, of course, each would have to be studied separately to see if is compatible with the various proposed explanations. Second, and independent of the etiological claim, is the cross-linguistic claim: that SLI (or the different SLI'S, if there are several) will be underlyingly comparable across languages. If both the etiological claim and the crosslinguistic claim are true, then for an explanation of SLI to be valid, it must be capable of applying to all languages. That is, if both of these claims are true, then a theory which fails to explain SLI in even one language cannot be an adequate theory of SLI. If there are several types of SLI, then of course a particular type might not be apparent in a given language because the structure of the language would not be vulnerable to a particular kind of Yumiko Tanaka Welty, International University of Health and Welfare, Tochigi; Jun Watanabe, Osaka University & Lise Menn, University of Colorado.

176

TANAKA WELTY, WATANABE & MENN

problem. For example, a language which does not have any reduced syllables is not vulnerable to problems with reduced syllables; a language that does not have agreement is not vulnerable to a deficit in agreement. But it is not plausible that a particular type of SLI is restricted to a particular language unless it is linked to a gene with limited population distribution (cf. Tay-Sachs disease), and that population only speaks that language. So it is unlikely to be the case that Japanese children with SLI have a different type of disorder from all other children with SLI. (This could be tested by looking at SLI in children of Japanese descent who speak other languages natively, on the model of medical studies. If there were a Japanese-population variant of a genetic SLI, it should be manifested in a certain percentage of children of Japanese ancestry who grow up speaking other languages, and SLI in those children should look different from SLI in other populations speaking those languages. The only reason to even discuss this unlikely possibility is to make clear what would be implied if someone tried to dismiss results from Japanese SLI as being irrelevant to explanations of SLI in other languages.) In summary, if a general explanation offered for SLI cannot apply to a par­ ticular language, but children speaking that language in fact have SLI, then either they have a different type of SLI, violating the etiological claim, or the explanation is not adequate. This is the theoretical standpoint from which we consider SLI in Japanese. 2. Theories of SLI There is no agreement on an exhaustive listing of the symptoms of SLI, but in Indo-European languages and in Hebrew, many children with specific lan­ guage impairment have difficulties using grammatical morphemes (e.g., in English, forms of the copula 'be', present third singular -s verb inflections, and regular past -ed verb inflections), and they tend to use them less frequently than younger, but linguistically normal children (see Leonard 1998 for a recent review). Hansson, Nettelbladt & Leonard (2000) divide current theories of SLI roughly into two groups. One group assumes that children with SLI have a grammatical deficit: they lack some grammatical knowledge that is essential for normal grammatical functioning. The other assumes that SLI children have a processing deficit: they are capable of grasping all grammatical notions, but have limited processing abilities, which slows their progress in identifying and interpreting morphological information from the input. Most theories of SLI are based on data from English-speaking children with SLI; now, however, other languages, specifically German, Italian, Hebrew, and Swedish, are also providing data to test their validity. In this

LANGUAGE PRODUCTION IN SLI JAPANESE PRESCHOOLERS

177

study, we present data from SLI children who are learning Japanese as their first language. Japanese is morphologically very different from Indo-European languages, for example, it has no infinitives, tense-number agreement, gender marking or articles; it almost has no irregular verbs, and its subject and object case markings, although commonly used, are optional, unless they serve to disambiguate the meaning of a clause. 2.1 Grammar deficit theories One account that assumes a deficit in grammatical knowledge in children with SLI is the Missing Agreement account of Clahsen research, based on evi­ dence from German and English (Clahsen 1989, 1991; Clahsen & Hansen 1993). They claim that SLI children have a selective deficit in establishing structural relationships of agreement (thus, e.g., they may fail to produce cer­ tain subject-verb agreements such as "He runs", but not others such as "They run"). This type of deficit cannot exist in Japanese. The inflectional, agglutinative suffixes in this language mark verbs for tense, aspect, voice, negation, and causation, among other things, but not for person, number or gender. Thus, there is no possibility that Japanese-speaking SLI children could have problems with subject-verb agreement. Another account within the grammatical deficit hypothesis is the Extended Optional Infinitives. This account assumes that children with SLI go through an extended period during which they fail to recognize that verbs must be marked for tense in main clauses (e.g., Rice et al. 1995; Rice & Wexler 1996). In its revised version (Wexler et al. 1998), this account allows for the possibility that these children treat agreement, too, as optional. For example, they may either select an appropriate (correctly marked) agreement and/or tense form or select an infinitive (comparable to the infinitive form in adult sentences such as I saw lohn play baseball). This account is also, apparently, irrelevant to Japanese. In Japanese there are no 'bare' or infinitive forms (although there is an extremely common non-finite form, the conjunctive -te). Every other verb form is neces­ sarily inflected for tense/aspect. Since the te- form is tenseless, it is possible, however, that the Extended Optional Infinitive account could be interpreted as a prediction that it would be optionally used as a main clause verb for an extended period of time by Japanese children with SLI. 2.2 Processing limitations One of the accounts that assumes processing limitations on the part of SLI children is the surface account (e.g., Leonard et al. 1997). According to this theory, SLI children have special difficulties with using the information that is encoded in relatively short grammatical morphemes. While they can perceive

178

TANAKA WELTY, WATANABE & MENN

non-syllabic consonantal inflections, weak syllable inflections, and function words, according to this account, they have a limited processing capacity, so their ability to obtain grammatical information from such short forms at nor­ mal conversational speeds is severely taxed. If grammatical morphemes are optional, these problems are made worse, since children would have fewer op­ portunities to observe their use. Japanese case particles, which mark syntactic relations, are short monosyllabic postpositions, such as  (direct object), ga (subject), and ni (indirect object/locative). However, they are not 'weak' like English or Italian inflections, because Japanese does not have a stress accent system, nor does it reduce the vowels of unaccented syllables (except in spe­ cific phonological conditions that case markers never satisfy). The case parti­ cles are, however, frequently omitted in casual conversations, especially if word order and semantics already permit the hearer to decode the basic gram­ matical relations. (When both subject and object are present but case particles are omitted, Subject-Object-Verb word order is assumed. For more informa­ tion on Japanese grammar, see Appendix). Since particle omission occurs even more in the speech addressed to young children (Rispoli 1996), Japanese children with SLI should lag behind their normally developing counterparts in the use of these particles. Another of the accounts that assumes processing limitations on the part of SLI children is the morphological richness account (e.g., Dromi et al. 1993; Leonard et al. 1987). According to this proposal, children with SLI devote their limited resources to the aspect of grammar that is dominant in a given language. Therefore, SLI children who speak English focus on word order, while those who speak Italian and Hebrew concentrate on inflectional morphology. In Japanese, neither word order nor case particles are 'dominant grammatical features', because they are not used consistently. Word order is relatively free (except that verbs appear in sentence-final position). Within the clause, word order is not contrastive or syntactically controlled, and both subject and object may be omitted when retrievable from the context. The most reliable grammatical feature is verb morphology, since verbs are always inflected for tense/aspect or carry the -te marker. It is also perceptually salient, since finite verbs are clause-final and can be followed only by sen­ tence-final discourse markers such as question particles. Japanese thus offers several advantages as a language for studying SLI. For example, if Japanese children show morphological impairment in SLI, this impairment cannot be due to difficulties with representing person-number agreement, or to using a grammar in which the verb of the main clause may optionally be infinitive. Japanese data could therefore require modifications of these particular

LANGUAGE PRODUCTION IN SLI JAPANESE PRESCHOOLERS

179

hypotheses as general explanations for SLI. The regularity of Japanese also permits the researcher to design test items using very high frequency verbs, without needing to use irregular forms. This makes the design of morphological experiments for very young children simpler than in European languages. However, Japanese also has a design disadvantage: the optionality of subject and object case markings in most contexts complicates error scoring. Designing contexts in which production of these case markings is obligatory continues to be a challenge to the researcher. 3. Typical development of verbs and particles Very young typically developing Japanese children appear to use the nonfinite -te form as their default. However, they soon show control of verbal inflections (Clancy 1985), appearing to do so earlier than children acquiring many other languages. Since either -te or tense/aspect markers are required on every form, this would be predicted by Mac Whinney and Bates' Competition Model (MacWhinney & Bates 1987; MacWhinney 1987), in which children most easily learn the cues to meaning that are the most available and reliable. In addition, because person-number agreement is absent, the finite verb system is less complex than in most other languages whose acquisition has been studied. The basic Japanese noun phrase consists of the unit Noun + Particle. Case particles are syntactic markers, and, as in Indo-European languages, they have varying degrees of correlation with semantic functions. Two of their properties, however, make them unlike familiar case suffixes. The Nominative and Accusative markers can be replaced by the Topic marker wa under appropriate discourse conditions, and, as mentioned above, they can be omitted entirely in colloquial discourse styles. Children attempt to produce the Noun + Particle unit early, even before they have learned the functions of the case particles. Children at this stage apparently realize that nouns are typically followed by particles, but do not know yet what kinds of particles exist, nor which particle to use. This parallels the way that children learning English, French, and many other languages learn that nouns are typically preceded by articles before they learn how to use them (Peters & Menn 1993; Veneziano & Sinclair 1997; Peters 2001). At a more advanced stage, Japanese children make systematic errors in using case particles because of the inconsistency of their semantic function. For example, the subject marking Case Particle ga has a wide range of semantic functions. As shown below, the ga-marked nominal indicates agentive/actor subjects of transitive verbs (la), certain types of objects of stative verbs (lb) or experiencers, others are in construction with intransitive verbs (lb) and (2b), and finally the patient of passive verbs (3).

180

TANAKA WELTY, WATANABE & MENN Agent of transitive verbs: (1) a. Kazuo-ga jusu-o nom-u Kazuo-NOM juice-ACC drink-Non past "Kazuo will drink juice" Agent of intransitive verbs where the subject controls the action: b. Yoshie-ga aruk-u Yoshie-NOM walk-Non past "Yoshie walks" Object (theme) of stative predicates: (2) a. Ringo-ga suk-i apple-NOM like-Non past "(I or some one) like apples" Object (theme) of intransitive verbs where the subject undergoes the action: b. Konoha-ga ochir-u Leaves-NOM fall-Non past "Leaves fall" Patient of passive verbs: (3) Okasan-ga tatak-are-ru Mother-NOM hit-Passive-Non past "The mother is hit"

These three major functions of ga appear at the same time in 2-year-old children (Clancy 1985). However, since ga does not consistently mark agents and is sometimes omitted or replaced by wa (and because the whole subject noun phrase is sometimes omitted), Japanese children have difficulty in acquiring the case marking for transitive events with agents and affected objects, as in (4a). (4) a. Kodomo-ga okasan-o tataku Child-NOM mother-ACC hit-Non past "The child hits the mother" b. Okasan-ga kodomo-ni tatak-are-ru Mother-NOM child-AGT hit-PASS-Non past "The mother is hit by the child" The oblique particle ni also has many semantic functions. It marks agent in passive sentences (AGT) ((4b) above), recipient in benefactor sentences, intermediate agent (causee) in causative sentences, and it also simply means "in'/'on' in locative constructions. Unlike ga and o, it is not optional in any of these uses.

LANGUAGE PRODUCTION IN SLI JAPANESE PRESCHOOLERS

181

The current study is based on the data from a large-scale ongoing study in Japan. Its major objectives are to identify preschool children with SLI, if possi­ ble, and to establish assessment criteria; to begin to characterize the language deficits in Japanese SLI (JSLI); and to contribute to the development of theoretical accounts of bases of SLI in general. 4. Initial study: does J(apanese)SLI exist? 4.1 Procedures: subject identification The identification of SLI children has not been well established in Japan for several reasons: 1) for preschool children (especially boys), not speaking as well as can be expected with age is not perceived as a developmental problem in Japan, as long as they behave appropriately. 2) There is a lack of comprehensive language assessment batteries. 3) The structural characteristics of Japanese - especially the optionality of both subjects and objects - hinder the identification of grammatical deficits. The JSLI identification process began by surveying teachers in regular kindergartens in Osaka, Japan, asking them to list children with whom they were concerned in terms of their language and communicative development in their classrooms. This is an ongoing project, involving the survey of 4 kindergartens with a total enrollment of about 600 children. So far, 34 children ranging in age from 3;10 to 6;4, have been listed as suspects for SLI. After the initial survey, each child was administered the tests for language and cognitive abilities by the first or second author. The tests included the Picture Vocabulary Test (PVT), two subtests of the ITPA (word association and grammatical completion), a subtest (digit span) of the K-ABC, and two of the WIPPSI subtests (picture completion and coding). These tests are standardized in Japanese. The session lasted for about 45 minutes per child. Typical inclusionary and exclusionary criteria were used for identifying SLI subjects: 1) language deficit, defined as -1.25 or more Standard Deviations (SD) below the means of at least two out of three language test scores (the PVT and two subtests of the ITPA); 2) performance IQ not lower than -1 SD on picture completion and/or coding of the WIPPSI; and 3) a confirmation of there being no hearing problem, no evidence of neurological dysfunction, and no anomalies in oral structure and motor function. Based on the scores of these standardized tests, 15 (10 boys and 5 girls) out of the 34 SLI suspects were identified as having SLI. They ranged in age from 4;4 to 6;9, with the mean of 5;6 (SD = 10 months). Fourteen Chronological Age matched children (8 boys and 6 girls) and nine Language Age-matched children (6 boys and 3 girls) served as the control subjects. As

182

TANAKA WELTY, WATANABE & MENN

shown in Table I, the CA-matched controls scored significantly higher than the JSLI children on expressive and receptive language (t=5.05, p<.000; t=2.62, p<.05, respectively). Both control groups performed significantly higher on digit span than the JSLI children (CA: t=3.54, p<.002; LA: t=3.55, p<.002). Thus, a putative group of JSLI subjects was identified on the basis of standardized tests of vocabulary, word association, and grammar. Subject Groups

N

JSLI

15

children 14

CA-

matched 11

LA-

j matched

Receptive LA

Expressive LA

Average

(PVT)

(ITPA)

NVIQa'b

5.6 (10.0) 5.7 (9.8)

4.2 (13.7) 5.5* (17.4)

3.8 (9.9) 5.3" (9.6)

9.8 (2.2) 11.0 (2.2)

Digit Spanc 8.9 (2.5) 12.5" (2.7)

4.3" (7.8)

3.11 (11.2)

3.11 (7.4)

11.9 (1.8)

12.7" (1.9)

CA

a

: the composite score of picture completion and coding, * p<.05, ** p<.01, "* jx.001, **" p<.000

b

and c : the mean = 10, SD=3

Table 1 : Subject Profiles

4.2 Quantitative study: does the putative JSLI group have difficulty with basic morphosyntax? The Syntax Test of Aphasia for Children (Fujita 1984, not officially standardized), consisting of a series of 15 pairs of pictures, was used to elicit children's production. The experimenter describes the first picture of each pair (e.g., "a mother eats a watermelon") and the child has to describe the second picture of each pair with a different subject and object (e.g., "a child eats an apple") but in the same syntactic structure. The child was asked to produce each sentence immediately after each of the experimenter's examples. The syntactic complexity of the 15 examples is varied in terms of linguistic parameters such as number of case relations, kinds of particles, and verb morphology for passive and causative. Three groups of children - the children with putative JSLI, younger Expressive Language Age (ELA)-matched children, and CA-matched children were compared on the elicited production of different types of sentences: Active Subject-Object-Verb (sov); Active Subject-Recipient-Verb (SRV) and Subject-Recipient-Object-Verb (SROV); Passive Subject-Agent-Verb/SubjectAgent-Object-Verb (SAV/SAOV); Causative Subject-Recipient/causee-ObjectVerb (SROV) sentences. Their scores differ for all types of sentences. The post hoc comparisons revealed that the differences were due to the higher percentage correct for the CA-matched children as compared to the children

LANGUAGE PRODUCTION IN SLI JAPANESE PRESCHOOLERS

183

with SLI and/or the younger ELA-matched children. The children with SLI and the ELA-matched children were not significantly different on this elicited production task (Figure 1). At this point, it appears that these 15 children were correctly identified as having JSLI, as they severely lag behind age-matched controls in producing a variety of sentences. However, their difficulty in sentence production seems to be delayed, not deviant. JSLI ELA-matched CA-matched

Causative SROV Fig.l: Elicited production

4.3 Qualitative study: do JSLI children have difficulties with case particles or verb morphology? For this study, data from 12 JSLI children, 8 MLU-M (morpheme) matched children, and 12 Chronological Age (CA) matched children were used (Table 2). Chronological Age (months) SD range MLU-Morpheme SD range Expressive Language Agea SD a

JSLI 70.8 (7.93) 56-81 6.8 (0.94) 5.3-8.0 47.5 (9.02)

MLU-matched 51.8" (6.94) 39-59 7.0 (1.07) 5-8.2 47.0 (9.42)

: the average of two subtests of ITPA (months) Table 2: Subject Group Profiles

CA-matched 70.6 (7.46) 58-79 7.6 (0.48) 6.3-8.1 69.3" (9.49) !

P< .000

|

184

TANAKA WELTY, WATANABE & MENN

As Table 2 shows, the MLU-M matched children averaged 19 months younger than the JSLI children. The production data from these children were obtained with a modified version of the Syntax Test of Aphasia for Children. The modified version had a greater variety of sentences within each grammatical voice, and included sentences where the difference in referents between modeled sentence and the elicited sentence was not so great as in the original version. 4.3.1 Scoring methods. Whenever single or multiple noun phrase(s) and verb were produced, the possible contexts for a case particle were identified. The use of case particles was analyzed based on the percentage of case particles produced in possible contexts and the percentage of case particles correctly used. For example, in (5), the child's utterance okasan-ga banana- taber-u ("the mother eats banana") had 2 possible contexts for a case particle. Only one was produced, and it was correct, resulting in 50% occurrence. (5)

Okasan-ga banana- taber-u mother-NOM banana-0 eat-nonpast "the mother eats banana" The second aspect scored was verb inflection. The number of verbs in­ flected correctly was computed and the percentage correct was obtained. When the child failed to produce the presumed target verb forms, the errors were categorized; the categories of errors included lexical selection, tense, voice and modal (causative). We used sentences of four types: active, dative, passive, and causative; the number of sentence tested per child ranged from 13 to 14. For each measure, the children with SLI, MLU-M matched children and CA-matched children were compared using analysis of variance (ANOVA). Findings of significant differences were followed up by Tukey comparisons at the 0.05 level. 5. Case particles results The percentage of occurrence of case particles in the possible contexts of active, dative, passive, and causative sentences are shown in Figure 2 for chil­ dren with JSLI, MLU-M matched children and CA-matched children. The three groups differed only in dative sentences (F =5.18 (df between groups=2, df within group =29) at the .05 significance level). Tukey comparison revealed that for the dative sentences, the CA-matched children (=98.5, SD=4.91) pro­ duced a higher percentage of case particles than both the children with SLI (M=79.1, SD=22.7) and the MLU-M matched children (=74.3, SD=24.1). The children with SLI did not differ in producing case particles in the potential

LANGUAGE PRODUCTION IN SLI JAPANESE PRESCHOOLERS

185

contexts when compared to the MLU-M matched children in any type of sen­ tences. Unexpectedly, the children with SLI produced the same number of case particles in the potential context of active, passive, and causative sentences as the CA-matched children. However, they did differ in the accuracy of their use of those particles 100 DJSLI  MLU-matched □ CA-matched

Fig. 2: Case particles in the possible contexts

Figure 3 shows group differences in the correctness of case particles marking noun arguments in active, dative, and causative sentences; the passive sentences showed a non-significant difference in the same direction (active: F(2, 29) = 3.35, p < .05; dative: F(2, 29) = 5.67, p <0.01; causative: F(2,29) = 3.33,p<.05).

DJSII I MLU-matched D CA-matched

Artive Sentence

Sentence

Fig. 3: Correct usage of case particles

186

TANAKA WELTY, WATANABE & MENN

According to Tukey comparisons, the children with SU used case particles less appropriately than the CA-matched children especially in active and dative sentences. However, the SLI children did not differ in accuracy of particle use from the MLU-M-matched children. 6. Verb morphology results Figure 4 shows the percentage of verbs that were appropriately inflected for each type of sentences by the three groups. Significant group differences were found all four types of sentences: active, dative, passive, and causative. Tukey comparisons revealed that the differences were attributed to signifi­ cantly higher percentage of correct verb inflections for the CA-matched chil­ dren than for the children with SLI. The SLI and MLU-M-matched groups did not statistically significantly differ from each other. However, the children with SLI showed a great deal of difficulty in producing the verb inflections for dative, passive and causative sentences. The mean percentage of correctly in­ flected verbs for dative, passive, and causative were 45.8%, 34.8%, and 41.7%, respectively. In contrast, the MLU-M matched children correctly inflected 75% of verbs in dative and causative sentences, although they also had difficulty in dealing with passive inflection. Thus, the group differences between the children with SLI and the MLU-M-matched children were not statistically significant.

DJSLI  MLU-matched □ CA-matched

Passive Sentence

Causative Sentence

Fig. 4: Verb morphology

We further inspected the data for types of errors made by children in in­ flecting verbs. Figure 5 shows the frequencies of errors in the four categories Lexical Selection, Tense, Voice, and Modal. The children with SLI made

LANGUAGE PRODUCTION IN SLI JAPANESE PRESCHOOLERS

187

significantly more errors in marking passive voice than either MLU-M-matched and CA-matched children. They also made more errors in selecting lexically correct verbs than the other two groups. The children with JSLI had a higher percentage of errors in marking Tense than the CA-matched children, and were just about the same as the MLU-M-matched group. Both groups of children tended to add -teru (progressive) at the end of verbs. Interestingly, some younger MLU-M matched children used -teru as if it were a passive marker. In some other instances, children added a redundant -teru even after properly in­ flected forms such as tor-are-teru (take-passive-progressive).

 JSLI  MLU-matched D CA-matched

Modal

Fig. 5: Errors in verb morphology

7. Discussion and conclusion Table 3 provides a summary of the predictions of each theoretical account for each of grammatical aspects examined in this study. The table shows com­ parisons between children with SLI, younger controls who are matched to the SLI children on Mean Length of Utterance in Morphemes (MLU-M), and chronological age (CA) matched controls. Children with SLI are assumed to lag behind their CA-matched controls in grammatical aspects of their language re­ gardless of theoretical accounts. None of the leading theories of SLI is well supported by the JSLI findings, so each of them would require modification to be viable as a general theory of this disorder. The grammar deficit accounts assume that the key difficulties of children with SLI revolve around agreement and tense (specifically the Missing Agreement and the Extended Optional Infinitive account discussed in Section 2, above).

188

TANAKA WELTY, WATANABE & MENN

Japanese verb morphology marks no agreement. Since verb morphology makes distinctions according to tense, and since there is a nonfínite inflected form, the Conjunctive form ending in -te/-de, this account should predict that Japanese children with SLI would produce more verb forms + -te/de than the MLU-matched children and CA-matched children. The particular grammar defi­ cit accounts that we have described make no predictions about case particles in JSLl; these would depend on details of the formal mechanisms of case as­ signment. Processing limitation accounts, especially the morphological richness account, assume that the resources which must be devoted to deploying a wide variety of grammatical devices such as word order, case particles and verb morphology will exceed the capacity of children with SLI. However, the sur­ face account predicts that Japanese children with SLI will direct more of their resources to verb morphology, which is a more salient and reliable cue than case particles. Consequently, children with SLI will lag behind MLU controls in the area of case particles but not verb morphology. Processing limitation, especially reduced processing speed, will also have an effect on syntactic cal­ culation of the associations between case particles and verb morphology. These predictions are contrasted in Table 3. The results from Japanese children with SLI did not agree with any of the predicted patterns. Unlike the prediction based on the processing limitation accounts, the children with SLI did not have more difficulty in producing case particles and using them correctly than the MLU-M matched children. Case particles are monosyllabic postpositions and frequently omitted in colloquial registers, especially in child directed speech. Furthermore, as we have seen, some of the basic grammatical particles have a wide range of se­ mantic functions. These surface features of case particles, however, do not ap­ pear to hinder discovery of the grammatical functions by children with SLI, as expected by the processing limitation accounts. Grammatical Aspects

Grammar Deficit

Processing Limitation

Findings

Case Particles Verb Morphology

NotTested

SLKMLuK CA

SLI = MLU
SLKMLU
SLI=MLU
SLI < MLU
Table 3: A summary of predictions andfindings

In contrast, verb morphology is the most reliable and salient grammatical cue, but the children with JSLI performed in this grammatical area much worse than expected by the morphological richness hypotheses and surface accounts. These accounts predicted that Japanese-speaking children with SLI would de­ vote their resources to verb and its inflection because they are perceptually sa-

LANGUAGE PRODUCTION IN SLI JAPANESE PRESCHOOLERS

189

lient and reliable grammatical cues relative to other features such as case particles and word order. The grammar deficit accounts fared no better. We tested the non-finite form Verb + te for Verb morphology, but found no evidence that six-year-old Japanese children with SLI or MLU-matched children of about 4 years of age produce the non-finite form in the active, dative, passive, or causative sentences. There was no verb form which ended with -te in the speech sample of any children, even though very young typically developing children appear to use the non-finite -te form as their default. Based on the grammar deficit accounts as presently elaborated, problems in case particles were not testable. Thus, the results from Japanese did not fit with any of the theoretical accounts of grammatical deficits in SLI. We are aware of that the findings in this study were based on a small number of children, a small number of sentences, one dialect of Japanese, and only one type of elicitation task. However, we also need alternative and adequate explanations for Japanese-speaking children with SLI. 8. A proposal: cognitive overload account The finding of a lexical selection deficit leads us to question the validity of approaches which focus only on morphological problems in SLI. Word retrieval errors are also seen in the written language of older JSLI children, and in rapid naming tasks as well. We suggest that a wider focus on their language deficits would be appropriate. Recall the important findings of Blackwell & Bates (1995) showing that normal young adult subjects doing experimental tasks under severe cognitive overload showed morphological deficits similar to those found in agrammatic Broca's aphasia. We hypothesize that children with JSLI - and by implication, children with SLI in general - have a general language processing deficit, which means that they are under 'overload' conditions whenever they are required to produce (or comprehend) language at the level expected of CA-matched peers. Morphological errors are the result of this overload, and so are the word-retrieval errors. This idea is fully compatible with the notion that there is also a perceptual deficit, which comes into play in languages where grammatical morphemes are reduced in perceptibility, but as we have seen, a perceptual account is insufficient to account for the Japanese data. Observation of the children's actual behavior during testing supports the 'overload' idea: they are clearly putting forth great effort to comply with the test demands. This proposal is harder to disconfirm than the four which we have examined, but since none of them have been supported, it appears to be the best way to deal with the data.

190

TANAKA WELTY, WATANABE & MENN

REFERENCES Bates, Elizabeth & Brian MacWhinney. 1987. "Competition, variation, and language learning". Mechanisms of Language Acquisition ed. by Brian MacWhinney, 157-193. Hillsdale, N.J.: Lawrence Erlbaum Associates. Blackwell, Arshavir & Elizabeth Bates. 1995. "Inducing agrammatic profiles in nor­ mals: Evidence for the selective vulnerability of morphology under cognitive re­ source limitation". Journal of Cognitive Neuroscience 7:2.228-257. Clahsen, Harald. 1989. "The grammatical characterization of developmental dyspha­ sia". Linguistics 27.897-920. Clahsen, Harald. 1991. Child Language and Developmental Dysphasia: Linguistic studies of the acquisition of German. Amsterdam & Philadelphia: John Benjamins. Clahsen, Harald & D. Hansen. 1993. "The missing agreement account of specific language impairment: Evidence from therapy experiments". Essex Research Re­ ports in Linguistics 2.1-37. Clancy, Patricia. 1985. "The acquisition of Japanese". The Crosslinguistic Study of Language Acquisition: The data ed. by Dan I. Slobin, 373-524. Hillsdale, N.J.: Lawrence Erlbaum Associates. Fujita, Ikuyo. 1984. The Syntax Test of Aphasia. Tokyo: The Association of Japanese Communication Disorders. Hansson, K., Ulrika Nettelbladt & Lawrence B. Leonard. 2000. "Specific language impairment in Swedish: The status of verb morphology and word order". Journal of Speech, Language, and Hearing Research 43.848-864. Dromi, Esther, Lawrence B. Leonard & M. Shteiman. 1993. "The grammatical mor­ phology of Hebrew-speaking children with specific language impairment: Some competing hypotheses". Journal of Speech and Hearing Research 36.760-771. Leonard, Lawrence B. 1998. Children with Specific Language Impairment. Cambridge, Mass.: MIT Press. Leonard, Lawrence ., J. Eyer, L. Bedore & B. Grela. 1997. "Three accounts of the grammatical morpheme difficulties of English-speaking children with Specific Language Impairment". Journal of Speech, Language, and Hearing Research 40.741-753. Leonard, Lawrence ., Letizia Sabbadini, Jeannette S. Leonard & Virginia Volterra. 1987. "Specific Language Impairment: A crosslinguistic study". Brain and Lan­ guage 32.233-252. MacWhinney, Brian. 1987. "The competition model". Mechanisms of Language Acquisition ed. by Brian MacWhinney, 249-308. Hillsdale, N.J.: Lawrence Erlbaum Associates. Peters, Ann M. & Lise Menn. 1993. "False starts and filler syllables: Ways to learn grammatical morphemes". Language 69:4.742-777. Peters, Ann M. 2001. "Filler Syllables: What is their status in emerging grammar?". Journal of Child Language 28:229-242.

LANGUAGE PRODUCTION IN SLI JAPANESE PRESCHOOLERS

191

Rice, Mabel L. & Wexler, Kenneth. 1996. "Toward tense as a clinical marker of Spe­ cific Language Impairment in English-speaking children". Journal of Speech, Language, and Hearing Research 39.1236-1257. Rice, Mabel L., Kenneth Wexler & Paul Cleave. 1995. "Specific Language Impairment as a period of Extended Optional Infinitive". Journal of Speech and Hearing Research 38.850-863. Rispoli, Matthew. 1995. "Missing arguments and the acquisition of predicate meanings". Beyond Names For Things: Young children's acquisition of verbs ed. by Michael Tomasello & W. E. Mwrriman, 332-352. Hillsdale, N.J.: Lawrence Erlbaum Associates. Sasanuma, Sumiko. 1986. "Universal and language specific symptomatology and treatment of aphasia". Folia Phoniatrica & Logopedics 38.121-175. Veneziano, Edith & Hermina Sinclair. 2000. "The changing status of 'filler syllables' on the way to grammatical morphemes". Journal of Child Language 27:461-500. Wexler, Kenneth,  Schutze & Mabel L. Rice. 1998. "Subject case in children with SLI and unaffected controls: Evidence for the Agr/Tns omission model". Language Acquisition 7.317-344.

Appendix: Some features of Japanese grammar Basic Japanese morphosyntax differs from English in several ways: basic word order, flexibility of word order, case marking system, and verb mor­ phology (Sasanuma 1986). The basic word order of transitive sentences is SOV. However, word order is relatively free, except for verb position. Within the clause, noun phrase order is not contrastive or syntactically controlled. Adverbial subordinate clauses precede the main clause; relative clauses precede the noun they modify. Therefore, the finite verb of the main clause is the last lexical element in a sentence. It may, however, be followed by discourse-governed final particle; these show the speaker's attitude towards the utterance, such as strong assertion, suggestion, anticipation of agreement, or information question. The simple main clause in Japanese consists minimally of the sentence-final finite verb. Subject and Object noun phrases are present if they cannot be inferred from context or are needed for other pragmatic reasons, but they are syntactically optional; one or both may be omitted. Basically, wherever English would use a pronoun instead of the full noun phrase, Japanese simply omits the noun phrase. Case Marking. The basic Japanese noun phrase consists of the unit Noun + Particle. Syntactic relations are marked by case particles:  (direct object), ga (subject), and ni (indirect object/agent of passive), as illustrated in sentences (a) and (b) below. Japanese also has a modest set of locative postpositions, which are similar to Western European locative prepositions. In addition to its

192

TANAKA WELTY, WATANABE & MENN

role as a case marker, ni can be a locative postposition, with a meaning similar to English 'in'. Each postposition has only one form, regardless of the noun it is attached to. Japanese does not mark either the gender or number of the noun, and there is no gender or number agreement. (a)

(b)

(c)

Kazuo-ga Yosie-o osi-ta Kazuo-NOM Yosie-ACC push-PAST "Kazuo pushed Yoshie" Yosie-o Kazuo-ga osi-ta Yoshie-ACC Kazuo-NOM push-PAST "Kazuo pushed Yoshie" Okasan-ga kodomo-ni tatak-are-ru Mother-NOM child-AGT hit-PASS-Non PAST "The mother is hit by the child"

Either the nominative or the accusative marker can be replaced by the topic marker wa under appropriate discourse conditions. However, if a phrase marked with ni or a locative postposition is topicalized, the topic marker -wa is added after the postposition. The particles -ga and -o may be omitted in casual conversations, especially if word order and semantics permit the hearer to decode the basic grammatical relations. When both subject and object are present but case particles are omitted, sov word order is assumed. Because the case particles are syntactic markers, they have varying degrees of correlation with semantic functions, as discussed in the text. Go-phrases, like subjects in English, may indicate agentive/actor subjects of transitive verbs, themes of stative verbs, experiencers, or other undergoers of intransitive verbs, or the patient of passive verbs. Phrases marked with  may be themes or expe­ riencers; and ni-phrases, most confusingly, may refer to agents, recipients, undergoers, or locations. Ni also denotes the intermediate agent (causee) in causative sentences. Verb Forms. The main verb of a sentence is finite, carrying an obligatory marker for tense/aspect. The tense/aspect marker may be preceded by one or more markers for mood or voice, giving morpheme strings like Stem + Passive + Negation + Past. There is no person or number marking on the verb (or indeed, anywhere in Japanese). The tense/aspect marker is the last morpheme in a sentence except for the discourse-governed sentence-final particles. Other verbs in the main clause, as well as the verbs of subordinate clauses, are in the non-finite 'conjunctive' form, marked with the suffix -te. Progressive forms, consisting of the conjunctive followed by a finite auxiliary, are extremely common, as they are in English. Japanese morphology is

LANGUAGE PRODUCTION IN SLI JAPANESE PRESCHOOLERS

193

basically agglutinative and the forms of these markers affect one another very little. The only important exception to this is when the verb stem is immediately followed by one of the two t-initial inflections: past tense -ta or conjunctive -te. These two endings trigger and undergo regular phonological changes to avoid complex consonant clusters. Japanese has no 'bare' form, infinitive form, or participle. The conjunctive-te form, although it is nonfinite, is not an infinitive, because it is not a noun, nor does it have any noun­ like syntactic or morphological properties. The -te form is not a gerund or gerundive, either, because it cannot be used as an adjective or a noun. The features are summarized in Table 4. eat

ride

write

read

Non past

tabe-ru

nor-u

kak-u

yom-u

Conjunctive

tabe-te

not-te

ka-i-te

yo-n-de

Progressive

tabe-teru

not-teru

ka-i-teru

yo-n-deru

Past

tabe-ta

not-ta

ka-i-ta

yo-n-da

Negative

tabe-nai

nor-anai

kak-anai

yom-anai

Passive

tabe-rare

nor-are

kak-are

yom-are

Causative

tabe-sase

nor-ase

kak-ase

yom-ase

Passive +Negative-+-Past

tabe-rare -nakat-ta

nor-are -nakat-ta

kak-are -nakat-ta

yom-are -nakat-ta

Table 4: Verb inflections and stem changes

IV. ISSUES ON GRAMMAR AND COGNITION

TESTING LINGUISTIC CONCEPTS ARE WE TESTING SEMANTICS, SYNTAX OR PRAGMATICS?

LEAH R. PALTIEL-GEDALYOVICH Ben-Gurion University of the Negev 1. Introduction There are a variety of standardized tests of language acquisition available to speech-language clinicians which are purported to assess linguistic, i.e. syntactic or semantic, knowledge. In this paper the question of what sort of knowledge is being assessed in these tests is discussed. I will argue that in some cases it is pragmatic, rather than syntactic or semantic knowledge, which is required to successfully complete test items. Specifically, I will consider the linguistic knowledge necessary to comprehend and produce coordinated sentences, and the interaction of this knowledge with pragmatic knowledge. In other words, a variety of standardized tests are used to assess children's syntactic or semantic ability. However, target responses may require pragmatic knowledge instead. If a child fails such a test, what kind of deficit does this reflect? I will consider items which assess knowledge of coordinated structures, i.e., structures including including the coordinators and, but and or, of the type shown in (1). (1) a. The dog barked and bit. b. The cat meowed but didn't scratch. c. The bear ate or slept all day. Test items may reflect the interaction between the semantics of coordinators and the influence of the maxim (following form Grice's (1975) pragmatic principle) of quantity. One example coming from the Token Test for Children (TTFC) (DiSimoni 1978) requires children to pick up only one of two items in response to an instruction using the coordinator 'or'. The common ('inclusive') truth-table meaning of 'or' allows truthful use of 'or' both when only one of the disjuncts is true and when all disjuncts are true. The restriction of the use of 'or' to instances where only one disjunct is true may result from the maxim of quantity (e.g., Chierchia & McDonnel-Ginet 1990). The use of or (as opposed to and) in this test item gives rise to an implicature that an appropriate response

198

LEAH R. PALTIEL-GEDALYOVICH

is 'choose only one of the disjuncts'. Choosing both of the items appears to be a failure in pragmatic - and not semantic interpretation. Test items intended to test syntactic knowledge of coordinators may in­ volve knowledge of the maxim of manner in targeted responses. In the Sen­ tence Combining subtest of the TOLD-I test (Hammill & Newcomer 1982), children are given two sentences and required to combine them into one sen­ tence. The example sentence is: stimulus - "I am big. I am tall." The target is "I am big and tall.", i.e., coordination with ellipsis. A sentence not involving ellipsis is syntactically acceptable however appears to violate the maxim of manner. It appears that the failure of children to successfully complete the task reflects impaired pragmatic and not impaired syntactic ability. An analysis of a failure as a pragmatic failure as opposed to a linguistic failure will affect the direction of a treatment program. If we wish to accurately assess children's communicative abilities, and design appropriate remediation programs for them, it is crucial that we accurately analyze the tools we us to assess them. The first part of this paper deals with linguistic and non-linguistic knowledge involved in comprehending and/or producing coordinated senten­ ces. The linguistic knowledge will be said to constitute semantic and syntactic aspects. The specific non-linguistic knowledge to be described will be pragmatic principles, specifically the Gricean concept of conversational implicature. The interaction of linguistic and pragmatic factors is discussed. In the second part of this paper, the effect of this interaction on performance on some test items included in developmental language tests is discussed. I will argue that items that aim to test linguistic knowledge actually require pragmatic knowledge. The third section deals with the clinical implications of this interaction between linguistic and pragmatic knowledge. Implications both for assessment and treatment will be briefly discussed. Modularity issues are discussed in the fourth section. 2. The interface of grammar and pragmatics 2.1 The semantic-pragmatic interface of coordination The meaning of sentences, or the propositions they represent, may be de­ scribed within the framework of a formal (logical) semantics. Within this framework, the meaning of a proposition results from analyzing the meaning of its constituents, and the relationships between its constituents and/or its composite propositions. The meaning of a coordinated proposition will then derive from the meaning of its constituent propositions and the meaning of the coordinating device. The meaning of the coordinators here will be first

TESTING LINGUISTING CONCEPTS

199

defined as their logical form. By logical form, I refer to how these terms are translated into logic. This is to be distinguished from the 'Logical Form' in Chomskyan syntactic theory. I am classifying all coordinators as logical operators. Moreover, for the purpose of this discussion I will be ignoring the debate on the Boolean or non-Boolean nature of and. The meaning of the coordinators is thus defined as in Table 1. p or q

T

p and q p but q T

T

F

F

T

F

T

F

T

F

F

F

F

p

q

T

T

Table 1: Truth tables for and/but and or.

The same conditions which result in the complex expression p and q being true, result in the expression p but q holding true (Table 1). The truth condi­ tions for or are somewhat different. As the difference between and and but is not relevant for this paper, the remainder of the discussion will be restricted to and and or. Examples (2) and (3) show how natural language sentences can be inter­ preted following on the basis of Table 1. First consider (2). (2) a. Pooh bear found a pot of honey. b. Pooh bear ate the pot of honey.  Pooh bear found a pot of honey and ate it. For a coordinated sentence with and to be true, the conjuncts must be true. The coordinated sentence in (2c), represents a coordination of the propositions expressed in (2a) and (2b). By Table 1, the coordinated sentence (2c) will ex­ press a true proposition only in the case where the both the proposition ex­ pressed in (2a) and that expressed in (2b) are true. Now consider example (3). (3) a. Eeyore lost his house. b. Eeyore can't remember where he put his house.  Eeyore lost his house or he can't remember where he put it.

200

LEAH R. PALTIEL-GEDALYOVICH

By Table 1, the proposition expressed in (3 c) will be true under one of three conditions: 1. The proposition expressed in (3a) is true, but not that in (3b), i.e. it is true that Eeyore lost his house (although he definitely remembers where he put it). 2. The proposition expressed in (3b) is true, but not that in (3a), i.e. it is true that Eeyore can 't remember where he put his house (although he did not lose it). 3. Both the propositions of (3a) and of (3b) are true, i.e. it is both true that Eeyore lost his house and that he can't remember where he put it. You will note that by Table 1, this last situation could also be truthfully described by (3d). (3) d. Eeyore lost his house and can't remember where he put it. According to this account, the semantic knowledge associated with the coordinators can then be summarized as knowledge of their truth conditions. It will be this semantic knowledge which we aim to test as part of our assessment of a child's linguistic knowledge of coordination. Yet, if we reconsider (3c) and (3d) it appears that there is an intuitive preference for (3d) over (3c) in that situation where both of the conjuncts are true. Grice (1975) proposed a theory of implicature to account for the part of the meaning of an utterance which the speaker wishes to convey which is not part of what is ordinarily considered the semantic meaning expressed in the utterance.1 The semantic meaning would be the ordinary truth-condition type of meaning which was described earlier. Grice wished to account for other aspects of meaning. These additional aspects of meaning can be illustrated in the rather common type of exchange in example (4). (4) Speaker A: It's cold in here. Speaker B: a. #Yes. (and does nothing else), or b. Yes (and closes the window). In (4), Speaker A's proposition is an assertion, "the temperature in this place is too low". Yet, a response from Speaker  which is merely an affirma­ tion of the assertion would be infelicitous. In Grice's terms, Speaker A is im­ plicating that he would like Speaker  to do something to change the room 1

There have been a variety of versions of this pragmatic theory, improvements, criticisms and revisions. A discussion of these is beyond the scope of this paper. Any theory, however, that recognizes a non-grammatical aspect of meaning, will require an analysis of language tasks in a spirit similar to that discussed in this paper.

TESTING LINGUISTING CONCEPTS

201

temperature, for example, close the window. In addition to the literal meaning of the proposition, Speaker A is communicating the 'extra-linguistic' meaning, i.e., "Close the window (please)". They are two kinds of implicatures defined by Grice. For our purposes, it is the concept of conversational implicature which is of interest. A definition of conversational implicature, after Grice (1975), is given in (5). (5) a. b.  d.

An implicature derived by the hearer about what the speaker means based on: conventional word meaning and the identity of the relevant referents the cooperative principle and the maxims following from it (to be explained below) the linguistic and non-linguistic context of the utterance knowledge of the world.

Grice (1975) defines the cooperative principle as in (6) (6)

... make your contribution such as is required, at the stage at which it occurs, by accepted purpose or direction of the talk exchange in which you are engaged.

He lists four maxims following from this principle: i. relation (be relevant), ii. quantity (be appropriately informative), iii.manner (be clear, including be brief), and iv. quality (be truthful, don't say that which you cannot support). When a maxim is flouted (Grice's term) or violated, the listener will cal­ culate an implicature, i.e., the listener will seek the speaker's motivation for violating the maxim. If the speaker has violated a maxim s/he must want the listener to derive another piece of information. For some reason the speaker wishes to convey this second message by way of an implicature, rather than explicitly. The maxims which are relevant for our discussion are the maxims of quantity and manner. An example of an exchange in which the maxim of quantity is violated is shown in (7). (7) Parent: Did you wash your hands with soap? Child: I washed my hands. Implicature derived by parent: the child did not use soap.

202

LEAH R. PALTIEL-GEDALYOVICH

If the child had used soap, it would have been more felicitous to say simply "yes". By giving more information, "I washed my hands", the speaker is violating the maxim of quantity. This results in the listener seeking a reason for the apparent violation of this maxim. The listener calculates an implicature, something like: the child is mentioning 'washing hands' but not 'with soap' because no soap was used. The child gave information unnecessarily, violating the maxim of quantity, but preserving the maxim of quality, which requires truthfulness. How does this concept of conversational implicature contribute to our un­ derstanding of the use of coordinators? Let us take this formula for calculating a conversational implicature and see what happens when we apply it to coordinated sentences, for instance the sentences in (8). (8) a. The Mad Hatter served Alice tea and cake, b. The Mad Hatter served Alice tea or cake. Also, consider the situations listed in (9). (9) a. It is the case that the Mad Hatter served Alice tea, but not cake, b. It is the case that the Mad Hatter served Alice cake, but not tea.  It is the case that the Mad Hatter served Alice both tea and cake. The sentence in (8a) may be truthfully asserted given the situation in (9c) only. This follows straightforwardly from the truth conditions of and. Sentence (8b) may be truthfully asserted in any one of the situations (9a), (9b) or (9c). Again this follows straightforwardly from the truth conditions for or. The interesting case is the one in which both (8a) and (8b) may be truth­ fully asserted. This is the case described in (9c), the case where the Mad Hatter served Alice both tea and cake. Given this situation, if and is chosen, as in (8a), the choice appears to be straightforward and felicitous. If or is chosen as in (8b), the choice seems infelicitous. The conventional meaning of the coordination includes the truth table meaning of or, i.e. situation (9a) holds true, or situation (9b) holds true, or situation (9c) holds true. In the case where (9c) is true, the speaker appears to be violating the maxim of quantity and therefore not abiding by the cooperative principle. How? In the situation (9c), there are two options available to the speaker for encoding the coordinated proposition. The choice of the coordinator and would provide the listener with the information that both of the conjuncts are true. The choice of the coordinator or, on the other hand, provides the listener with less information. The choice of or informs the listener that either one of the disjuncts is true (which one is unspecified), or both of the disjuncts are

TESTING LINGUISTING CONCEPTS

203

true. This is less informative, because the speaker is not letting the listener know which of these propositions is being asserted. Taken out of context, there appears to be no justification for violating the maxim and the speaker appears to be simply uncooperative. However, if the context of the utterance provides a reason for the speaker to violate the maxim of quantity, for instance the speaker is uncertain about which of the case (9a), (9b) or (9c) is true, then in order not to violate the maxim of quality (say only that which you can support), the speaker must choose or. How this implicature may be derived in context is demonstrated in (10). (10)

Speaker A: Do you remember what the Mad Hatter served Alice at his tea party? Speaker B: I'm not sure. Perhaps he served her tea or cake. Speaker A: You're right, he served both. Implicature calculated by Speaker A: Speaker  used the less informative or, violating the maxim of quantity because of uncertainty.

This is just one example of how a specific context, such as uncertainty, can result in a usually infelicitous use of or being considered felicitous. Adult knowledge of the meaningful use of coordinators includes more then than knowledge of how these coordinators function as logical operators. In addition, adult knowledge includes the interaction of pragmatic principles, specifically the maxim of quantity in the choice of coordinator. The child learning language must master both of these aspects in order to show adult­ like use of coordination. 2.2 The syntactic-pragmatic interface of coordination Let us now turn to the interaction of the syntax of coordination with pragmatics.The syntactic knowledge which we aim to assess in testing chil­ dren's ability to formulate coordinated sentences, includes knowledge of both overt and elliptical syntactic options for formulating these sentences. Consider (11). (11) a. Paul wrote the songs and Paul played the songs and Paul sang the songs, b. Paul wrote, played and sang the songs. Both of the versions in (11) are grammatical. The relationship between the versions in (1 la) and (1 lb) reflects some concept of identity which allows the deletion of the identical elements 'Paul' and 'the songs'.

204

LEAH R. PALTIEL-GEDALYOVICH

Now let us look at implicatures arising from the maxim of manner. This maxim has been described as 'be clear' and includes in its definition be brief. Example (12) shows an example of a violation of the maxim of manner. (12)

Adult Speaker A in the presence of a young child: Do we have any C-H-O-C-O-L-A-T-E? Implicature calculated by the listener: Speaker A does not want the non-spellers present to understand the message.

In example (12) the speaker chooses an obscure manner to express him/herself. By spelling out the word 'chocolate', instead of saying the word normally, the speaker is implicating that his/her message is intended only for the spellers present. The violation of the maxim of manner results in more than merely the literal meaning of the message being conveyed. Let us return now to example (11). Both sentences, (11a) and (lib), are syntactically acceptable. Both express the same coordinated proposition described in (13). (13)

It is the case that Paul carried out three actions: he wrote songs, he played the songs and he sang the songs.

However, the sentence in (11a) appears to be a bit awkward. If someone uses this version the listener is likely to calculate a conversational implicature to explain why the speaker chose this long-winded version. The cooperative principle, specifically the maxim of manner {be brief, part of be clear) appears to be violated. Taken out of context, there appears to be no justification for this long-winded version. However, a certain context may provide a justification, for instance, if the speaker wants to emphasize that it is always Paul who does everything to the exclusion of all others. In example (14), a context is suggested where the violation of the maxim of manner is used for emphasis. (14)

Speaker A: Do you think that all of the Beatles contributed equally to their success? Speaker B: Well, Paul wrote the songs and Paul played the songs and Paul sang the songs. Implicature calculated by Speaker A: Speaker  thinks that it was only Paul, or at least mainly Paul who was responsible for the Beatles success.

Here we see that given a specific context, a speaker may felicitously choose the long-winded version in (11a), despite the apparent violation the

TESTING LINGUISTING CONCEPTS

205

maxim of manner. The choice of the shorter version of (11) is, therefore, a pragmatic, and not a syntactic, decision, made in order to abide by the maxim of manner. Again, adult knowledge of coordination is seen to be more than knowledge of grammar, in this case syntax. The adult speaker is also a dexterous manipulator of the syntactic options with reference to pragmatic principles, in this case the maxim of manner. The child developing language will need a knowledge of the principle of manner, as well as the syntactic options, in order to use coordination in an adult fashion. 3. Test items - do they test semantics, syntax or pragmatics? How does this bear on language tests? First we will have a look at an example of test items which aim to address linguistic semantic concepts. An example is based on the TTFC (similar examples can be found in other as­ sessments, eg., the 1987 CELF-R test). This item is of the type shown in (15). ( 15) a. Visual stimulus : b. Verbal stimulus: Point to the small circle or the large square. The correct response, according to the test's authors, requires pointing to only one shape. Pointing to two shapes would be considered an unacceptable response. Purely on the basis of semantics, it has been argued above that there are three options for correctly executing this item. These are given in (16). (16) a. Point to the small circle only, b. Point to the large square only.  Point to both the small circle and the large square. Each of these three cases is covered in the truth-conditions for or, given in Table 1. So, on what basis is an answer as described in (16c) wrong? The an­ swer appears to be wrong on a pragmatic basis. The use of or, as opposed to and, leads the listener to calculate a conversational implicature, something like the calculation described in (17). (17) a. The speaker used or not and. b. If the speaker had used and, then s/he would have definitely wanted me to point to two items,  Since the speaker chose or, choosing a coordinator which gives me the option of pointing to only one item, s/he must expect me to choose only one. Otherwise the speaker, in observing the maxim of quantity would have chosen and.

206

LEAH R. PALTIEL-GEDALYOVICH

The likelihood of the child calculating this type of implicative is further increased by the linguistic context. The previous item of the TTFC test is of the type shown in (18). (18)

Point to the small circle and the large square.

The use of an item with or immediately following a similar item using and, emphasizes for the listener that the response to the item with or should be different from the response to the item with and.2 The child is 'commanded' to act in such a way as to cause the command to be fulfilled, i.e. true. The intention of the test's authors, and of clinicians using this test, is to accept responses corresponding to (16a) and (16b) as correct. Option (16c) is rejected, although it is consistent with the truth-conditions of the command. The basis for the rejection is that this interpretation is not felicitous, i.e. is not pragmatically appropriate. This leads me to conclude that this type of item assesses pragmatic knowledge. A child who responds according to option (c) is demonstrating a lack of pragmatic, and not linguistic (in this case semantic) knowledge. Examples for test items which aim to address the syntax of coordination may be found in the TOLD-I test (Hammill & Newcomer 1982). One of the subtests of this test battery requires children to formulate complex sentences with coordination or subordination from two related simple sentences. An ex­ ample of the stimulus sentences for such an item is shown in (19). (19) a. I am big. b. I am tall. Sentences (19a) and (19b) would be given and the child would be expected to formulate a complex sentence. Most, if not all, adults would give the response in (19c).  I am big and tall. Yet, there is another possible response, the response shown in (19d). d. I am big and I am tall. Both (19c) and (19d) are grammatical. To produce these sentences the child must know the syntactic structure of coordination. To product (19c), that knowledge must include knowledge of the constraints on deletion. The 2

The influence of context on interpreting test items, even within a test is important. An apparently 'neutral' context, devoid of the contextual influences of 'natural' communication settings, does not necessarily neutralize pragmatic factors, and may in fact emphasize them. My thanks to Yishai Tobin for emphasizing this point.

TESTING LINGUISTING CONCEPTS

207

authors' intention, and that of clinicians using the instrument, is to accept only (19c) as support for the child's knowledge of the syntax of coordination. But (19d) is also grammatical, just awkward. If we return to Grice's cooperative principle and its maxims, we must now consider the maxim of manner which states: be clear (including be brief). Choosing (19d) appears to be infelicitous because it involves unnecessary wordiness. Producing the grammatical (19d), shows lack of pragmatic knowledge, specifically the maxim of manner, and not lack of syntactic knowledge. 4. Summary and what difference does it make anyway? Having addressed the importance of recognizing the relative contributions of linguistic and pragmatic knowledge to successful performance on certain types of developmental test items, I now turn to the practical clinical issues. From a very practical point of view there are two questions which arise here, presented in (20). (20) a. How does failure to complete items requiring pragmatic knowledge contribute to diagnosis of language delayed and disordered children? b. How does a diagnosis of pragmatic as opposed to linguistic deficit affect our treatment and remediation of language delay and disorders? In the following sessions I discuss the assessment of linguistic knowledge with consideration of pragmatic knowledge, the planning remediation based on a pragmatic analysis and the implications for the diagnosis of specific versus non-specific language disorders. 4.1 Assessment of linguistic knowledge and pragmatic knowledge Let us consider a child being assessed on linguistic concepts, who executes instructions containing or as requiring all the actions described. The test protocol would require that this response be rejected. When we describe the language development of this child (of course in the context of performance on a number of tasks), we will describe the child as having a delay or disorder in semantic development. Yet, if we consider only the semantic appropriateness of the response and ignore the pragmatics, the response should be considered correct. And what of the child who formulates coordinated sentences without de­ leting potentially deletable material? The test protocol will require that we reject the response. We will describe the child as having a syntactic impair­ ment. However, if we ignore pragmatics, and consider only the syntactic

208

LEAH R. PALTIEL-GEDALYOVICH

appropriateness of the response, the answer must be accepted as correct (the left side of Figure 1). We present a test item and await a response. This response is judged solely on the basis of its grammatical (syntactic or semantic) appropriateness. If the child fails to perform in accordance to this purely grammatical criterion, the child is considered to have a grammatical deficit. In the case of coordination, this will mean accepting for items containing or any response where at least one of the disjuncts is true (fulfilled). Syntactically, there will be no distinction between responses with and without deletion. Both will be equally acceptable. If we do consider pragmatics, specifically the knowledge of the cooperative principle and the calculation of conversational implicature, interpretation of the same responses will be different (the right side of Figure 1). Present test item

Look at linguistics only; ignore pragmatics.

Look at linguistics and pragmatics.

Accept all grammatical responses.

Accept only felicitous grammatical responses.

Failure on test items

Linguistic deficit. Pragmatic deficit. Fig. 1 : Assessment of linguistic knowledge in view of pragmatic knowledge

Suppose a test item using or is presented to the child and the child's re­ sponse is semantically correct but pragmatically inappropriate. Since only fe­ licitous grammatical responses are accepted, only a response where at least one, but not all of the disjuncts are true will be considered correct. This re­ sponse will be interpreted as showing a pragmatic deficit (in the area of knowledge of the maxim of quantity of the cooperative principle). Semantic knowledge will not be considered to be impaired. In the case of a test item assessing syntax, the child may give a fully overt grammatical coordinated sentence. Since this response is infelicitous, although grammatical, the response will be considered incorrect. The failure will be interpreted as a pragmatic failure, a demonstration of incomplete or absent knowledge of the cooperative principle and the maxim of manner.

TESTING LINGUISTING CONCEPTS

209

Following the right side of Figure 1 one is led to test scoring in accordance with the standard procedures for the types of test items discussed above. However, interpretation of the results is different. The deficit (or competence for that matter) will no longer be considered in linguistic knowledge but in pragmatic knowledge. (Of course, there is the possibility that an individual will show both a linguistic and a pragmatic impairment.) 4.2 Planning remediation based on a pragmatic analysis What happens if we diagnose a child as having a semantic deficit? We set about remediating missing semantic skills. If we consider a child to lack the semantics of or for instance, we will teach the meaning of or. This should in­ clude teaching all the possible cases where the use of or would be truthful. If we are only relating to semantics, we will not address the issue of how the listener knows which of the possible cases is actually true. This does not mean to imply that the teaching of semantics would take place in isolation, without showing a variety of contexts, but only that the emphasis in therapy would be on the possible meanings, rather than on the calculations made by speaker and listener in encoding and deriving meaning. An alternative approach is based on the assumption that the child's seman­ tic skills are intact, while it is the pragmatic competence which requires reme­ diation. With this approach, we will attempt to teach the child to recognize the speaker's intention in teaching or. We will discuss the idea of the amount of information conveyed in an utterance {quantity). The influence of context in interpreting the speaker's intention will be given primary consideration. If we consider the case of failure on syntactic coordination items the situa­ tion would be similar. If we consider the failure to delete potentially deletable material as evidence of a syntactic deficit, we may first set about teaching the child to recognize potentially deletable material. We may then teach the child to always delete this material. What happens then when the child meets an in­ stance where potentially deletable material is not deleted? Will s/he interpret the utterance as ungrammatical? Alternatively, we can interpret the failure to delete in coordinated sen­ tences as evidence of impaired pragmatic knowledge, while syntactic knowledge is intact. In this case, we will teach the child to recognize contexts, in which it is more appropriate to delete (be brief) and those contexts, where it is more appropriate not to delete. In these latter cases the child would be taught to recognize the intentional violation of the maxim of manner as a pragmatic device used to cause the listener to calculate a conversational implicature consistent with the 'extra-literal' message the speaker wishes to convey.

210

LEAH R. PALTIEL-GEDALYOVICH

4.3 The diagnosis of specific and non-specific language disorders Sometimes when we assess a child with difficulty in language per­ formance, the question arises of whether the deficit is specific to the language faculty or if the language deficit is just one manifestation of a more general cognitive malfunction. This may be seen as a question of which aspects of language performance are the manifestation of language specific skills, and which ones of more general cognitive skills. It seems reasonable to assume that syntactic and semantic skills, i.e., linguistic competence is specific to language. What about pragmatics? Kasher (1991) presented a hypothesis of the modularity of pragmatics. He distinguishes between a modular pragmatic knowledge which is purely linguistic and a more general central pragmatics. Conversational implicatures are considered to be part of central pragmatics. Thus, the pragmatic knowledge which I have argued is required to successfully complete the coordination tasks reported here would be part of a more general cognitive knowledge, not specific to language. Following from Kasher's theory, this pragmatic skill would be a developmental skill with improved performance with increased age. Impairment in this field would be reflected in non-linguistic as well as linguistic behavior. Figure 2 shows the relationship between the interpretation of failure on language tasks and how subsequent diagnosis of the child may proceed. If the deficit is interpreted as being a linguistic deficit, there will be an expectation of a specific language disorder. If the deficit is seen to be of a pragmatic nature, the language difficulty will be seen to be part of a broader cognitive disorder. Failure on language test item

Linguistic deficit

Pragmatic deficit

Specific language General cognitive disorder disorder Fig. 2: Interpreting failure on language tasks

Once we analyze our assessment tools as assessing pragmatic as oppose to linguistic skills, our expectation regarding the specificity of the impairment may change. This has very practical implications regarding referral to other professionals, educational expectations and general case management.

TESTING LINGUISTING CONCEPTS

211

5. Conclusion In this paper, I have attempted to show that we need to look very carefully at what we are assessing when we use developmental language tests. The need for pragmatic knowledge to successfully complete test items, which purport to test purely linguistic knowledge, challenges standard interpretation of these tests. This does not negate the usefulness of these tests. Rather, I wish to em­ phasize that we need to recognize the various skills needed in order to appro­ priately use test results in diagnosis and planning treatment of language de­ layed and disordered children. REFERENCES Chierchia, Gennaro & Sally McDonnel-Ginet 1990. Meaning and Grammar. Cam­ bridge, Mass.: MIT Press. DiSimoni, Frank. 1978. The Token Test for Children. Allen, Texas: DLM Teaching Resources. Grice, H. Paul. 1975. "Logic and Conversation". Syntax and Semantics ed. by Peter Cole & Jerry L. Morgan, vol. III, 41-58. New York: Academic Press. Hammill, Don D. & Phillis L. Newcomer. 1982. Test of Language Development Intermediate. Austin, Texas: PRO-ED. Kasher, Asa. 1991. "On the pragmatics modules: A lecture". Journal of Pragmatics 16.381-397. Semel, Eleanor, Elisabeth H. Wiig & Wayne A. Secord. 1987. Clinical Evaluation of Language Function. Revised ed. San Antonio, Texas: The Psychological Corpo­ ration.

SPECIFIC LANGUAGE IMPAIRMENT AND MODULARITY LINGUISTIC AND NON-LINGUISTIC EXPLANATIONS DUŠANA RYBÁROVÁ Slovak Academy of Sciences 1. Introduction During the last few years there have been vivid discussions among lin­ guists, cognitivist scientists, and speech pathologists about the Specific Lan­ guage Impairment (SLI) in connection to modularity and innateness.1 Propo­ nents of linguistic approaches to SLI endorse the idea that SLI results from a genetically determined impairment of the innate autonomous language module, and therefore claim that SLI can be regarded as empirical evidence of an innate modular structure of language faculty. On the other hand, propo­ nents of the functional, non-linguistic processing approach emphasize nonlinguistic deficits in SLI and dynamic explanations of SLI, and deny the innate impairment in the domain-specific language module as an underlying cause of SLI. A big gap thus arises between these two radical positions: the approach stressing an innate, static linguistic explanation of SLI sharply contrasts with the approach emphasizing a dynamic, non-cognitive explanation of the same impairment. The aim of my paper is to show one possible way how the gap between those two positions could be bridged. First, linguistic and non-linguistic ac­ counts of SLI will be outlined. Both approaches will be outlined from the point of view of one of their proponents, so as to show their premises and weak points in relation to the modularity issue. I will argue that, although claiming to be completely different, both linguistic and non-linguistic accounts of SLI accept the same idea of 'strong modularity' (i.e., a conception characterized as nativist, static and uninteractionist) and such understanding of modularity causes them to appear as mutually exclusive. It will be shown that 'strong modularity' is unproductive and inappropriate view to account for the broad array of SLI 1

E.g., Trends in Cognitive Science recently published a wide spectrum of papers and commentaries from proponents of different theories of the underlying cause of SLI (Gopnik 1997; Karmiloff-Smith 1998; Tomblin & Pandlich 1999; Van der Lely 1999; Joanisse & Seidenberg 1998).

214

DUŠANA RYBÁROVÁ

symptoms, symptoms that probably result from different primary deficits. I therefore propose a finer-grained view on modularity, which I call 'weak modularity'. 'Weak modularity' postulates four kinds of domain-specific modules: innate knowledge modules, innate processing modules, acquired knowledge modules, and acquired processing modules. Finally, an alternative model of syntax acquisition combining these different kinds of modules and a domain-neutral device will be suggested. This model allows both linguistic and non-lingustic accounts to converge on one understanding of SLI. 2. Characterization of SLI children Children with SLI have no obvious neurological, hearing or cognitive deficits. They are characterized by a dissociation between the linguistic and their non-linguistic capacities.2 Though scoring at least normal on non-verbal parts of standardized intelligence tests, their verbal scores lag significantly behind their non-verbal performance. In the language domain, the core of their difficulties lies in problems with syntax, morphology, and phonology. However, despite the fact that the dissociation between verbal and non-verbal abilities is the most significant criterion for the diagnosis of SLI, SLI children often exhibit non-verbal deficits as well, namely in memory (e.g., Gathercole & Baddley 1990), and other skills like visual and spatial perception (e.g., Merzenich et al. 1996) or attention. 3. Linguistic and non-linguistic accounts of SLI: 'strong modularity ' If children with SLI only showed a dissociation between verbal and non­ verbal abilities, the nature of their deficits would undoubtedly be linguistic. But, given that these children also exhibit non-linguistic deficits, both linguistic and non-linguistic explanations of SLI are propounded. Proponents of the linguistic approach consider non-linguistic deficits to be only co-occurring with primary linguistic impairments. In contrast, proponents of non-linguistic explanations of SLI consider non-linguistic deficits to be the key to understanding impairments in language of SLI children. I will claim that the main reason for this contrast is the fact that both linguistic and non-linguistic accounts are based on the Same premise, namely 'strong modularity', although they use it differently: proponents of the linguistic approach try to find evidence in favor of it, and proponents of non-linguistic approach try to refute it. However, if they abandon the 'strong modularity' position, they will see that a dynamic view of explanation of SLI can be compatible with modularity and vice versa. For a detailed review of SLI and its possible explanations, see, for example, Leonard (1998).

SPECIFIC LANGUAGE IMPAIRMENT AND MODULARITY

215

3.1 Linguistic accounts Linguistic accounts of SLI stress the morpho-syntactic deficits typical of SLI. Their proponents claim that the primary cause of SLI has genetic roots, and causes impairment in the innate language module. The notion of an innate language module is based on work by Jerry Fodor and Noam Chomsky, while the main proponents of an explanation of SLI in terms of an innate modular deficit are Myrna Gopnik and Heather van der Lely. The latter (1997) argues that there is a specific subgroup of SLI which she calls Grammatical SLI. Grammatical SLI children show the typical profile of language impairment (e.g., they make errors in their production of aspects of inflectional morphology, and overgeneralize rules) in the absence of non-verbal deficits. Van der Lely has developed a theory explaining deficits of these Grammatical SLI children. She claims that they have what she calls a Representational Deficit for Dependent (structural) Relationships (RDDR). According to van der Lely's RDDR combined with Fodor's theory of modularity, children with SLI have an innate modular deficit in the language module (that means they have impaired modular phonology, morphology, and syntax), but their non-modular linguistic abilities, such as semantics and pragmatics, remain intact. Now I will try to consider the pros and cons of van der Lely's account. Her attempt to include linguistic theory into the diagnostic apparatus of Grammatical SLI children and the interpretation of their deficits may valuably serve to shed light on the possible deeper linguistic structure of this impairment. Linguistic regularities of deficits could tell us a lot about the structure of the impairment affecting SLI children, and if their presence were confirmed, this would move our hypotheses further in the direction of an underlying linguistic cause. However, the side effect of her approach, which stresses the modular linguistic deficit in SLI, is the neglect of the possible influence that other cognitive functions (e.g., perception or memory) can have on the output of the language module and on the role developmental dynamics can play in SLI. The reason for this neglect is the specific version of modularity applied by van der Lely (and other proponents of the linguistic approach), which is an interpreta­ tion of Chomskyan or Fodorian modularity.3 Van der Lely's interpretation of Fodor's and Chomsky's modularity is uninteractionist and static. Uninteractionist because it sets aside the influence that other cognitive abilities can have on language acquisition, and static because it does not make provision for any development of the modules, which only become triggered. This 3

In my view, Fodor and Chomsky themselves are not proponents of 'strong modularity'. I will justify this claim at the end of this section.

216

DUŠANA RYBÁROVÁ

interpretation, which I call 'strong modularity', makes no room for dynamic and non-linguistic interpretations of SLI. So, if van der Lely interprets Fodor or Chomsky in the 'strong sense', the dynamic explanation of SLI is incompatible with her linguistic modular explanation of SLI. 3.2 Non-linguistic accounts Proceeding to the non-linguistic accounts of SLI: what they have in com­ mon is the idea that the underlying cause of SLI does not lie in an impaired language faculty, and claim, instead, that SLI can result from a low level cognitive impairment (e.g., in perception or memory). Dorothy Bishop (1997:236) claims that an uneven pattern of language performance, as proposed by van der Lely, does not indicate that the cause of SLI is primarily a linguistic deficit, because non-linguistic deficit can affect the pattern as well as the timing of language learning. She adds that an unusual and distinctive pattern of grammatical difficulty, as seen in the case of Grammatical SLI, "[...] does not in itself constitute evidence for a biological defect in a grammatical module (ibid.)" To illustrate the non-linguistic approach to the explanation of SLI, we can consider the temporal processing deficit hypothesis developed by Paula Tallal and Malcolm Piercy (1973, 1974, 1975). They claim that the cause of problems in children with SLI is a deficit in their ability to correctly process the temporal characteristics of speech. Tallal and her colleagues (Tallal et al. 1996; Merzenich et al. 1996) argue that the important information in speech distinguishing similar syllables such as ba-da occurs too fast for SLI children to catch it. As a result, speech as a perceptual input is not processed properly and this has a disruptive effects on all levels of speech. According to these authors temporal processing deficit in children with SLI is not limited to speech sounds; fast stimuli recognition problems were also found in other modalities including vision and touch. To sum up, in the temporal processing deficit hypothesis, the deficits of SLI children are explained as a result of their underlying non-linguistic perceptual deficit. The dynamic view on development shows it to be conceivable that at the beginning of the 'SLI story' there could be an impairment, say, in the auditory input, which brings about further developmental problems in, say, syntax. So the deficits which we might find in a five-year-old SLI child need not be inter­ preted static manner - that is, it need not be the case that a child has problems in syntax because his/her syntax is innately impaired (see Bernstein & Stark 1985). Rather, an investigator would try to find out the child's whole developmental story.

SPECIFIC LANGUAGE IMPAIRMENT AND MODULARITY

217

However, the shortcoming of this approach, as in the case of linguistic explanations, is the neglect of the claims made by the proponents of the rival (in this case linguistic) approach; namely, the impact that innate factors can have on language development. The logic of the non-linguistic argumentation is as follows: a) modularity emphasizes the static and uninteractionist aspects of impairment; b) there is evidence that developmental dynamics and nonlinguistic factors significantly influence the form of impairment of language in SLl; c) therefore, a modular account can not adequately explain SLI. The premise a) exactly mirrors the 'strong modularity' position characterized in consideration of linguistic accounts. So, if Bishop reads Fodor and Chomsky as suggesting 'strong modularity', it allows her to infer that there is no place for a dynamic approach in the modular explanations of SLI.4 This logic further leads to rejection of the whole modular approach with his further important premise about innate linguistic information. As a result, the proponents of the non-linguistic explanation of SLI don't pay much attention to the role that in­ nate linguistic information can play in the picture of SLI. In the above discussion, I have argued that the source of the clash between linguistic and non-linguistic explanations is the notion of 'strong modularity' - for which the core claims are about a static and uninteractionist view of lan­ guage acquisition. Proponents of the linguistic approach try to support such a view, while proponents of non-linguistic explanations try to reject it. But, in my view, neither Fodor nor Chomsky make such strong claims about modu­ larity. In The Modularity of Mind, Fodor (1983) did not say much about the developmental aspects of modules, since this was outside his interests. In con­ nection to the child's learning of grammar, Chomsky emphasizes the poverty of the stimulus argument.5 He postulates that the necessary innate linguistic information - needed to infer the correct syntactic rules - can be characterized as Universal Grammar (UG). But even if the plausibility of the poverty of the stimulus argument and UG are accepted, the dynamic approach to development of language need not be excluded. The poverty of the stimulus argument shows that there has to be something innate concerning language, but it does not commit us to any radical claims about static or dynamic interpretations in connection to modularity. 4

Coltheart (1999) criticizes Bishop (1997) for reading Fodor as proposing necessary conditions for modularity and trying to define modules. 5 The poverty of the stimulus argument shows that the linguistic input that a child is exposed to in the development is too impoverished to allow the child to infer the correct syntactic rules. Therefore, we have to admit that there is something innate about language that enables us to learn language so easy and fast.

218

DUSANA RYBAROVA

I claim that the gap between linguistic and non-linguistic explanations of SLI results from the 'strong' interpretation of modularity and is artificial in nature. Moreover, I argue that it would be productive to view the two approaches as non-exclusive. It is a truism that the result of syntax processing depends also on perceptual input. This has to be admitted by both linguistic and non-linguistic approaches. The proponents of the linguistic approach are only in principle not interested in the non-linguistic factors and the role they could play in the picture of SLI. However, it is impossible to omit the fact that the outcome of each module depends on the input given, and that, if the input is damaged, this will lead to a damaged output of the module. This thesis does not violate the modularity hypothesis and vindicates the connection between modularity and a non-linguistic dynamic approach to the explanation of SLI. Therefore, in my view, there is no reason to neglect the dynamic aspects of development as they relate to the modularity issue. Moreover, using a dynamic approach in modularity could disentangle the debate between linguistic and non-linguistic accounts of SLI. A dynamic view of development calls for the use of longitudinal methodology in the study of language impairments, and, in my opinion, such a methodology is profitable for both linguistic and non-linguistic approaches. If some studies showed that syntax really developed (to a large extent) separately from other non-linguistic cognitive abilities, this would provide support for linguistic explanations of SLI. If, instead, they showed that this was not that case, the space would be open for non-linguistic theories. I do not see a dynamic approach to the study of language impairment as favoring either a linguistic or a non-linguistic approach. It is also a truism that there is something innate about language. The pro­ ponents of non-linguistic accounts of SLI, however, consider innateness to be very marginal because they connect the notion of nativism with the notion of the static and uninteractionist view of language development. In other words, the reason for this attitude is the fact that 'strong modularity' contradicts their notion of the importance of developmental dynamics and non-linguistic factors. But, as I showed above, if we abandon the 'strong' version of modu­ larity conjoining the notion of innateness, static and uninteractionist explana­ tions, there will be the possibility to accept the innate aspects of language, while dispensing with the static and uninteractionist view. To sum up, it was shown that the 'strong' version of modularity forces us to view the linguistic and non-linguistic accounts as mutually exclusive. Further, it was argued that it is unproductive to exclude in principle either the possible impact of innate linguistic information (as demonstrated in case of nonlinguistic explanations) or the possible impact of developmental dynamics and

SPECIFIC LANGUAGE IMPAIRMENT AND MODULARITY

219

non-linguistic factors (as in case of linguistic approach) on the picture of SLI. It is clear, that after this crucial disagreement is overcome (through abandoning the cause of this disagreement - the 'strong modularity') it will be a matter of further debates what aspects of language are innate, what aspects are learnt and to what extent. But, in my view, such move to finer-grained questions would mean bridging the gap between linguistic and non-linguistic explanations and I consider it to be a significant step forward in the explanation of language development and SLI. But, to be able to accept fully this alternative view of the problem, we need to formulate a new, 'weaker' form of modularity. 4. ' Weak Modularity ' In current cognitive science the notion of modules is defined differently for different purposes. In neuropsychology, according to the principle of dissocia­ tions, it refers to neurologically autonomous units (Ellis & Young 1988). David Marr (1982:325) introduced arguments in favour of the modular archi­ tecture of computational systems based on the notion of efficiency of infor­ mation processing. But the most complex notions of modularity come from Jerry Fodor (1983, 1985) and Noam Chomsky (1984, 1995). The notion of 'weak modularity' that I propose here combines various aspects of Fodorian and Chomskyan modularity with a current critique of modularity, as developed by Karmiloff-Smith (1992). I will try to make clear what I mean by 'weak modularity' by focusing on three points: a) modules are domain-specific units, b) modules can be construed as devices or databases (processing and knowledge modules), and c) modules can be either innate or acquired. According to Fodor (1985:3): "[...] a module is an informationally en­ capsulated computational system - an inference-making mechanism whose access to background information is constrained by general features of cogni­ tive architecture." Informational encapsulation is the most significant feature of modules, but they also exhibit other properties; thus, a module emerges as a 'cluster concept'. Some researchers interpret the properties that Fodor ascribes to modules to be the necessary conditions for any system to be considered modular, and criticize Fodor accordingly. In response, Coltheart (1999) has stressed that Fodor does not propose a definition of modularity nor does he stipulate any necessary conditions for the application of the term module. Coltheart also demonstrates that Fodor's properties of modules are not necessary conditions of modularity. Further, Coltheart offers his own defini­ tion of module as "[...] a cognitive system whose application is domain spe­ cific; here domain-specificity is a necessary condition for the applicability of the term 'module'."

220

DUSANA RYBAROVA

I use this definition of module in the 'weak modularity', but with one ex­ ception: I define module as a cognitive unit, not a cognitive system. The reason for this choice is that, in current usage, both the concept of module and the concept of system are associated with the notion of a processing module. In what follows, I make a crucial distinction between knowledge modules and processing modules. The first characteristic of module, then, is that it is a domain-specific cognitive unit. The second characteristic pertains to the distinction between knowledge and device, as discussed by Fodor and Chomsky. Chomskyan modularity differs from the Fodorian view of modularity in two main respects: first, Chomsky makes no distinction between central systems and input systems; and second, Chomskyan nativism is about innate knowledge, not about psychological devices. It is the second - often neglected - point that is crucial to the weak modularity. In connection with the theory of modularity, an important distinction has to be made: there is a difference between innate information and innate devices or mechanisms. The distinction between innate devices and innate knowledge is also emphasized by Fodor at the beginning of his Modularity of Mind. He claims that what Chomsky thinks is innate is primarily a certain body of information: the child is, so to speak, 'born knowing' certain facts about universal constraints on possible human languages (Fodor 1983:4). In his more recent critique of Fiona Cowie's book (1999) on nativism, Fodor (2001) writes that what Chomsky proposes is a nativism of domain specific propositional attitudes, not one of domain-specific devices. So, Chomskyan nativism is not about innate psychological mechanisms which mediate linguistic behavior, since such matters fall within the domain of performance theories. In contrast, Fodor stresses that we need to postulate devices which operate with the data described by Chomsky. Fodorian modularity is thus about psychological devices processing information. Coltheart (1999:118) has expressed this difference between innate knowledge and innate devices in neo-Fodorian terms as a distinction between a knowledge module vs. a processing module. He considers a knowledge module as a body of knowledge that is autonomous; for example, a linguist might say that syntax is a module. On the other hand, a processing module is a mental information processing system; for example, a psycholinguist might say that during language comprehension, sentences are parsed by a syntactic processing module. I use this distinction between processing and knowledge modules in my model.

SPECIFIC LANGUAGE IMPAIRMENT AND MODULARITY

221

The last feature of modules that has to be delineated in the proposed 'weak modularity' addresses the issue of whether modules can result from learning. Proponents of the non-linguistic approach stress that the acquisition of some skills often results in performance having some of the characteristics that Fodor ascribes exclusively to modules. Karmiloff-Smith (1992, 1997) suggests that modules could be the results of a process of 'modularization' which occurs re­ peatedly as the product of development (1992:4). In line with Karmiloff-Smith's thesis that modules as domain specific units can be produced by learning, two further kinds of modules are postulated in 'weak modularity': a) innate modules which are biologically determined; and b) acquired modules which result from the process of modularization. In connection to the previous distinction between knowledge and processing modules, innate as well as acquired modules can either bear some information (in which case, they are innate or acquired knowledge modules) or can serve as psychological devices (in which case, they are innate or acquired processing modules). To sum up, the necessary condition for a unit to be modular in the sense of the 'weak modularity' hypothesis is domain-specificity. There are four types of modules distinguished in connection to this version of modularity: 1) innate knowledge modules; 2) innate processing modules; 3) acquired knowledge modules; and 4) acquired processing modules.6 5. Alternative model of syntax acquisition If we accept the notion of 'weak modularity', as described above, and combine it with a dynamic approach to the development of language, the de­ bate between linguistic vs. non-linguistic approaches becomes more a matter of degree than a question of a stark choice between two competing alterna­ tives. What is still needed, though, is a model of syntax acquisition that would allow us to grasp the wide range of possible causes of SLI. I will now propose such a model and then reinterpret the mentioned theories of linguistic and non-linguistic approaches in terms of it. The proposed model tries to delineate a finer-grained picture of the organi­ zation of the language faculty, one that might allow us to identify more pre­ cisely the specific causes of the deficits in SLI children. The model consists of transducers, selection unit, a universal grammar database, a general purpose mechanism, and a syntax of a learned language. In this model three types of

6

Automatized skill in driving a car can be used as an example of the acquired processing module. The examples of the other three kinds of modules appear in the following proposed model.

222

DUŠANA RYBÁROVÁ

modules can be found: an innate knowledge module; an innate processing module; and an acquired knowledge module. All these modules preserve the necessary condition of modularity - domain specificity. The notion of an innate body of information (mentioned above in my discussion of 'weak modularity') invites the postulation of what I call the innate knowledge module. The universal grammar database is an innate knowledge module in this sense. Such module stores only innate knowledge relevant to UG, it is not a psychological processing device that operates with this knowledge. In contrast, the selection unit is an innate processing module, which means an innate psychological de­ vice. The unit called the syntax of learned language is an acquired knowledge module. Jane Grimshaw (1987) argues that: "The implications of a particular theory of grammar for learnability cannot be assessed without regard to those principles that mediate between the theory of grammar and the input to language learning." This function of mediating between the universal grammar database and the selection unit is realized by the general purpose device. The functions of the different parts of the alternative model could be de­ scribed as follows. Transducers only change the format of information coming in from the environment into one that is appropriate for further processing (for example, there could be a discrete transducer for auditory or tactile input). They are not inference-making devices. The function of transducer in this model is consistent with Fodor's (1983:40-42) characterization of them. The selection unit chooses from the representations delivered by the transducers that information which is relevant to its domain. In the case of the selection unit for syntax acquisition, those perceptual inputs that are relevant to syntax are domain-relevant; these representations refer to syntax and can be true or false. The universal grammar database stores relevant linguistic information uttered in UG.7 The general purpose mechanism has two functions: first, it checks the information from the selection unit against the information of the universal grammar database and then selects only those mental representations from the primary linguistic data that match the principles of UG (it selects those representations relevant to syntax which match UG and which are, therefore, true about syntax); second, it produces syntactic representations of a concrete language which are finally stored as a learned syntax of a concrete language. 7

The model I propose depicts the developmental story of language acquisition. But it is also interesting to consider the possible changes in this model concerning the steady-state of language abilities in adulthood. To mention at least one probable change, the database of UG is not available in adulthood. Therefore learning language as an adult is much more complicated than in childhood.

SPECIFIC LANGUAGE IMPAIRMENT AND MODULARITY

223

If we admit the idea of a model consisting of different modules, as has been introduced here, we can then try to sketch the possible field of hypotheses about the selective impairment of these modules by applying the principle of dissociations. A defect in transducers should imply broader impairment of syntax without an internal structure. The possible domains of domain-specific transducers concerning language development can be auditory, tactile, or visual. So we can think about three possible transducers that could be impaired in the case of SLI. If there were an impairment in the auditory transducer underlying SLI, children learning American Signe Language (ASL) should not show deficits related to SLI. Conversely, the deficits in children without hearing loss, and not learning ASL, cannot be explained in terms of the impaired tactile transducer. If there were a non-structured grammatical impairment (not showing regularities restricted to very specific parts of UG) in both the ASL and the auditory input speakers, we would infer that there is a deficit in the selection unit. This is because the domain of the selection unit is specified only by syntactically relevant inputs, and not by whether these inputs are auditory, visual or tactile in nature. If there were a selective impairment in some deeper structures of grammar revealed by linguistic analysis, this would probably result from an innate impairment in the universal grammar database. If the children had a generally lowered profile of intellectual abilities, it might arise from problems in the general-purpose device.8

database of UG

syntax of learned language

selection unit

transducer Fig. 1 : An alternative model of syntax

acquisition

8 Of course, this 'hypothesis production' is narrowly connected with the question of whether there are appropriate diagnostic tools for determining which of the hypothesis is relevant in individual cases. I will not try to address this question in this paper. Here I just want to stress, without concrete argumentation, that in my view this methodological problems are not as hopeless as they might appear.

224

DUŠANA RYBAROVA

I will now try to interpret van der Lely's hypothesis by explaining deficits in Grammatical SLI children in terms of the above described model. In general, the representational deficit for dependent relationships would not be interpreted as the problem of innate language module. The alternative model allows one to formulate a finer-grained hypothesis. At first sight, the RDDR might be interpreted either as a problem of innate information in the universal grammar database or as a problem of the selection unit or as a problem of the transducer. In the first instance, the universal grammar could (for example as a result of a genetic disorder) lack some principles. When the representations from the selection unit were compared to UG information in the first procedure of the general purpose mechanism, an appropriate match would not be found and the representations would, therefore, neither be processed further, nor delivered as a grammatical rule of a concrete language. In case the problem lay in the selection unit, this would not deliver appropriate representations from the primary linguistic data and again, the appropriate rule of syntax for a concrete language would not be produced. The representational deficit for dependent structure could also result from a problem in one or another of the transducers. If that were the case, the information from the environment would not be appropriately processed and the input of the selection unit would be defective. Here, the following important question arises: how can we determine which of the three above-mentioned hypotheses is the correct one? The answer lies in the notion of the deficit in Grammatical SLI proposed by van der Lely as grammatical, with some syntactically relevant structure. This means that the deficit is restricted to some specific syntactic structures. Therefore, the most probable candidate for the explanation of deficits in Grammatical SLI is the impairment in the database of UG. I contend that the proposed alternative model is compatible with the views of the proponents of the non-linguistic approach. The new model allows to formulate hypotheses about the underlying causes of SLI in terms of nonlinguistic impairments - say, as impairments in the transducers or the selection unit. Take, for example, Tallal's theory about the temporal processing deficit. If we connect her idea about a deficit in perception with our model, the proposed deficit would turn out to impair the auditory transducer. The output of transducers would be not appropriate and, as long as we assume that the outputs of transducers become inputs of other units, e.g., the selection unit, it would have disruptive effects on further processes, whose results would be also defective. In other words, we would get a grammar with non-structured mistakes. According to Tallal's hypothesis, SLI should not

SPECIFIC LANGUAGE IMPAIRMENT AND MODULARITY

225

occur in ASL speakers, because learning ASL is not dependent on fast structures of auditory or other kinds of fast perceptual input. This claim is consistent with our idea about different specialized transducers. The proposed alternative model of syntax acquisition combines different types of modules falling into the category of 'weak modularity' with the notion of a domain-general device. If we interpret van der Leiy's RDDR in Grammatical SLl-children in terms of this model, we can conclude that the RDDR results from impairment in the database of UG. If we interpret Tallal's theory of temporal processing deficit in terms of the alternative model, the deficit is caused by impairment in auditory transducer. 6.

Conclusion I began this paper by remarking that the current status of the debate be­ tween the proponents of linguistic and non-linguistic approaches to explana­ tion of SLI is characterized by a severe divide, one that would seem to defy any possibility of theoretical bridge-building. Van der Lely's representational deficit for dependent relationships and Tallal's temporal processing deficit hy­ pothesis have been chosen as examples of the linguistic and non-linguistic ex­ planations of SLI respectively in order to illustrate their premises in relation to the modularity issue. It has been argued that the discrepancy between these two approaches arises from the notion of 'strong modularity' and its purported connection to a static and non-interactionist view on which the arguments of both sides are based. I have suggested that one possible way to bridge this discrepancy is to abandon the 'strong modularity' view and replace it with the notion of 'weak modularity'. The 'weak' understanding of modularity con­ siders domain-specificity to be the necessary condition for a unit to be modular and distinguishes four types of modules: innate knowledge modules; innate processing modules; acquired knowledge modules; and acquired processing modules. It was then applied in the alternative model of syntax acquisition that allows one to formulate, in relation to an individual case, both linguistic and non-linguistic explanations of the causes of SLI. Finally, the deficits suggested by van der Lely and Tallal were re-interpreted in terms of the new model and a field of possible hypothesis about the causes of SLI was delineated.

226

DUŠANA RYBÁROVÁ

REFERENCES Bernstein, Lynne E. & Rachel E. Stark. 1985. "Speech perception development in Language-Impaired children: A 4-year Follow-up Study". Journal of Speech and Hearing Research 50.21-30. Bishop, Dorothy V. M. 1997. Uncommon Understanding: Development and disorders of language comprehension in children. Hove: Psychology Press. Chomsky, Noam. 1984. Modular Approaches to the Study of the Mind. San Diego, Calif.: San Diego State University Press. Chomsky, Noam. 1995. "On the nature, use and acquisition of language". Readings in Philosophy and Cognitive Science ed. by Alvin Goldman, 511-535. Cambridge, Mass.: MIT Press. Coltheart, Max. 1999. "Modularity and cognition". Trends in Cognitive Sciences 3.115-120. Cowie, Fiona. 1999. What's Within? Nativism reconsidered. Oxford: Oxford Univer­ sity Press. Ellis, Andrew W. & Andrew W. Young. 1988. Human Cognitive Neuropsychology. Hove: Lawrence Erlbaum Associates. Fodor, Jerry A. 1983. The Modularity of Mind. Cambridge, Mass.: Bradford Books. Fodor, Jerry A. 1985. "Précis of 'The Modularity of Mind"'. The Behavioral and Brain Sciences 8.1-42. Fodor, Jerry A. 2001. "Doing without what's within: Fiona Cowie's Critique of Nativism". Mind 110.99-148. Gathercole, Susan E. & Alan D. Baddeley. 1990. "Phonological memory deficits in language disordered children: is there a causal connection?". Journal of Memory and Language 29.336-360. Gopnik, Myrna. 1997. "Language deficits and genetic factors". Trends in Cognitive Sciences 1.5-9. Grimshaw, Jane B. 1987. "The Components of learnability theory". Modularity in Knowledge Representation and Natural-language Understanding ed. by Jay L. Garfield, 207-212. Cambridge, Mass.: Bradford Books. Joanisse, Marc F. & Mark S. Seidenberg. 1998. "Specific Language Impairment: A deficit in grammar or processing?". Trends in Cognitive Sciences 2.240-247. Karmiloff-Smith, Annette. 1992. Beyond Modularity. Cambridge, Mass.: MIT Press. Karmiloff-Smith, Annette. 1997. "Crucial differences between developmental cogni­ tive neuroscience and adult neuropsychology". Developmental Neuropsychology 13.513-524. Karmiloff-Smith, Annette. 1998. "Development itself is the key to understanding developmental disorders". Trends in Cognitive Sciences 2.389-398. Leonard, Lawrence B. 1998. Children with Specific Language Impairment. Cambridge, Mass.: MIT Press. Marr, David. 1982. Vision: A computational investigation into the human represen­ tation and processing of visual information. San Francisco: Freeman.

SPECIFIC LANGUAGE IMPAIRMENT AND MODULARITY

227

Merzenich, Michael M., William M. Jenkins, Paul Johnston, Christoph Schreiner, Steven L. Miller & Paula Tallal. 1996. "Cortical plasticity underlying perceptual, motor, and cognitive skill development: implications for neurorehabilitation". Cold Spring Harbor Symposia on Quantitative Biology 90.1 -6. Tallal, Paula & Malcolm Piercy. 1973. "Developmental aphasia: impaired rate of nonverbal processing as function of sensory modality". Neuropsychologia 11.389-398. Tallal, Paula & Malcolm Piercy. 1974. "Developmental aphasia: rate of auditory processing and selective impairment of consonant perception". Neuropsychologia 12.83-93. Tallal, Paula & Malcolm Piercy. 1975. "Developmental aphasia: the perception of brief vowels and extended stop consonants". Neuropsychologia 13.69-74. Tallal, Paula, Steve L. Miller, Gail Bedi, Gary Byma, Xiaoqin Wang, Srikantan S. Nagarajan, Christoph Schreiner, William M. Jenkins & Michael M. Merzenich. 1996. "Language comprehension in Language-Learning Impaired children im­ proved with acoustically modified speech". Science 271.81-84. Tomblin J. Bruce & Jennifer Pandlich. 1999. "Lessons from children with Specific Language Impairment". Trends in Cognitive Sciences 3.283-285. Van der Lely, Heather K. J. 1997. "Narrative discourse in grammatical Specific Lan­ guage Impaired children: A modular language deficit?". Journal of Child Language 24.221-256. Van der Lely, Heather K. J. 1999. "Learning from grammatical SLI". Trends in Cognitive Sciences 3.286-288.

THE LANGUAGE/COGNITION INTERFACE LESSONS FROM SLI AND WILLIAMS SYNDROME

VESNA STOJANOVIK, MICK PERKINS & SARA HOWARD University of Sheffield If we cannot now end our differences, at least we can help make the world safe for diversity. John F. Kennedy (Address given at American University, Washington, D.C., 10 June 1963)

1. Introduction: the modularity debate Scholars from a number of disciplines have long viewed the language system as a gateway to the capacities and functions of the human mind (Caroll 1994). The possibility of a genetic-innate basis of domain specific aspects of human language has been a hotly debated topic in the last decade. Often termed modularity, the domain specificity of language is a fundamental feature that differentiates competing linguistic theories and accounts of language acquisition (Bates et al. 1988; Bates et al. 1995; Fodor 1983, 1985; Levy 1994; Roeper & Seymour 1994). At one end of the scale are the 'cognitivists', largely within the Piagetian camp, who take the view that language is dependent on cognition. Considerable evidence has been adduced in support of this contention by showing, for example, that cognitive structures must develop before their linguistic counterparts can be acquired (e.g., Tomasello 1996). This position is also adopted by those who are mainly concerned with artificial systems that stimulate real-time comprehension. For computational linguists, limitations in attentional capacity, memory, and motor programming are regarded as fundamental constraints that shape the form that languages take. The aim is to explain grammar in terms of natural constraints on language processing, rather than to regard grammar as something entirely independent or autonomous, whose characteristics can only be accounted for by appealing to innateness. At the other end of the scale are the 'nativist' theorists such as Chomsky (1965, 1981, 1986) and Lenneberg (1967), who believe that one is born with linguistic knowledge already in the brain, and therefore that language is inde­ pendent of cognition. Grammatical knowledge is characterised as a static en-

230

STOJANOVIK, PERKINS & HOWARD

tity, and language is assumed to be part of our innate mental ability, i.e., lan­ guage 'competence', embodied in a specialised brain module. This view was further developed in Fodor's (1983) Modularity of Mind, which had a pro­ found impact on the study of language, its claim being that many of the processes involved in comprehension were undertaken by special mental systems termed modules. Fodor argues for a distinction between a central system responsible for rational thought and the fixation of beliefs, and a num­ ber of modular input systems, one for each of the senses, which feed the cen­ tral system. The language faculty is viewed as an input system on a par with the senses such as vision. Various criteria are required for a system to be counted as a module, which, apart from the domain specificity, crucially in­ clude information encapsulation. In addition, the criteria include specific neural architecture, which subserves the module and may be subject to idiosyncratic pathological breakdown. It is the neural specificity of modular architecture that is claimed to be genetically determined, i.e., innate (Pinker 1984, 1991; Fodor 1983; Chomsky 1965, 1986). The possibility of a genetic-innate basis to modular and domain specific aspects of human language has been a highly controversial issue over the past decade or so. Developmental dissociations can be of two kinds. The first occurs within the language system itself (thus, for instance, the lexi­ con is supposed to develop independently of grammar; see Locke 1994; Clahsen & Almazan 1998). At issue here is whether language is a unitary entity where all levels (phonology, morphology, lexicon, syntax, and pragmatics) interact with each other or whether these levels operate as stages in a processing chain and are largely independent of one another. The second do­ main of dissociations is between language and cognition. A considerable body of literature on adult aphasia has been devoted to describing and explaining selective impairment and sparing of particular linguistic and cognitive compe­ tencies (Bates & Thal 1991; Caramazza & Berndt 1978; Grodzinsky 1986). At issue in much of this work is whether language is a specialised and specific aspect of the human mind or whether generic cognitive mechanisms give rise to linguistic competencies. It is this second aspect of modularity, i.e., the language/cognition interface, that is addressed in the present paper. 2. Modularity and developmental disorders A great deal of evidence in support of or against modularity comes from reports of individuals with conditions such as Williams Syndrome (ws) and Specific Language Impairment (SLI), which manifest an atypical relationship

LANGUAGE/COGNITION IN SLI AND WILLIAM SYNDROME

231

between cognitive development and linguistic competence. WS is a rare genetic disorder (1 in 25,000 live births) which causes a variety of cardiovascular difficulties, failure to thrive in infancy, infantile hyper­ calcemia, delayed language and motor development, and abnormal sensitivity to sounds (hyperacusis). The syndrome has attracted a great deal of attention because of the resulting uneven neurolinguistic profile which is characterised by relative strengths in language, facial processing and social cognition, and profound impairment in spatial cognition, planning, and problem solving (Bellugi et al. 1988; Bihrle et al. 1989; Karmiloff-Smith et al. 1995; Mervis et al. 1999; Bellugi et al. 2000). The average IQ in individuals with ws is ~ 50, and yet they are hyperlinguistic (Cromer 1988; Pinker 1991). They can score at ceiling levels when knowledge of grammar is required, such as use of reversible passives, reflexive anaphors and regular past tense inflection (Bellugi et al. 1990; Bellugi et al. 1994; Clahsen & Almazan 1998), thus showing 'intact' linguistic competence in the face of moderate to severe cognitive deficits. The inference that language can therefore be dissociated from cognition has been drawn (Pinker 1991) and has encouraged some (Pinker 1994) to even propose the existence of a 'grammar gene'. Further evidence in support of brain modularity comes from SLI (Clahsen 1989; Gopnik & Crago 1991; Ullman & Gopnik 1999; Van der Lely 1996; Van der Lely & Stollwerck 1997), a disorder where difficulties with morphology and syntax co-exist alongside 'intact' non-linguistic abilities, i.e., cognition, and non-modular abilities, i.e., pragmatics.1 Problems with semantics and pragmatics are not unknown, though. Thus the morphosyntactic impairments revealed in children with SLI are taken to reflect impairment of an innate grammatical module; in consequence, the child lacks some of the specialised knowledge crucial for grammatical acquisition, despite normal cognitive development in all other areas of learning. Nevertheless, the reports on the relative strengths of individuals with WS in the language domain are controversial, and it has been shown that morphosyntactic abilities in WS can actually be impaired and more in line with mental age (Crisco, Dobbs & Mulhern 1988; Karmiloff-Smith 1997; Arnold et al. 1985; Volterra et al. 1996). No less controversial are the claims about the existence of 'intact' cognitive functioning in the face of linguistic deficits in SLI. Thus, for instance, the proponents of connectionist networks claim that learning of grammar is dependent upon general cognitive processes of per­ ception, recognition, attention, classification, memory retrieval, and planning, 1

Please note that Sperber & Wilson 1986 have provided an outline of crucial distinctions between modular, syntactic and non-modular, pragmatic representations and processes.

232

STOJANOVIK, PERKINS & HOWARD

and it is in one or more of these processes that the basis for morpho-syntactic impairments is sought. Hence the language deficit manifested inSLIis claimed to be dependent upon more general cognitive processing deficits (Baddeley 1986; Bishop 1992; Leonard et al. 1987; Tallal 1980, 1984; Tallal et al. 1985). Despite years of research in the area, the issue of language/cognition dissociations as evidenced in cases of ws and SLI is far from being resolved. Furthermore, there have been hardly any studies which consider in detail linguistic and non-linguistic functioning in the same subjects with WS or SLI, respectively, or that provide linguistic and non-linguistic data from both populations with ws and populations with SLI. Most of the studies usually concentrate almost exclusively on either SLI or ws and refer to the work of others only if making comparisons between the two. In addition, very few studies (Udwin & Yule 1990) have looked at conversational ability in ws, most of them considering linguistic abilities in terms of morpho-syntax. To better understand the complex mechanisms involved in ws and SLI, we need to look beyond the issue of double dissociations or lack of such. To this end, the aim of the present paper is to explore the language/cognition interface by attempting to investigate in greater detail the complex patterns of impairment in ws and SLI, and to show that when detailed information about linguistic, cognitive, and communicative functioning is obtained in the same subjects, the two profiles may not be as opposed to each other as has been claimed in the literature. This has important implications both for linguistic theory and for speech and language therapy. 3.

Method A battery of standardised verbal and non-verbal measures was administered, supplemented with a detailed analysis of the conversational functioning of four children with ws and four children with SLI. The reason for focusing on detailed profiles of a small number of children was to avoid the situation often encountered in larger group studies, where average performance may reflect that of none of the individual subjects', or in the best case scenario may reflect the performance of a majority of cases, but hide the existence of divergences from the average in a significant number of cases. Although such a small number of subjects may not be fully representative of ws and SLI, respectively, it allows us to identify patterns of behaviour on which to concentrate in larger scale studies.

LANGUAGE/COGNITION IN SLI AND WILLIAM SYNDROME

233

4. Participants: children with WS and children with SLI The participants with ws, two girls and two boys, were recruited through the UK ws Foundation and local speech and language therapists. They had a mean age of 9;06 (range 7;06-12;01). The participants with SLI, three boys and one girl, were also recruited on the judgements of speech and language therapists. The exclusion criteria for the SLI participants were: a) no hearing loss (such as would require regular visits to an audiologist) and no history of recurrent Otitis Media; b) no autistic spectrum tendencies; c) no history of clinically significant behavioural, emo­ tional or neurological problems; d) performance IQ not lower than 85; e) no oral motor or sensory defects. The mean age of children with SLI was 9;07 (range 9;01-11;03). The subjects were not matched on any other category apart from chrono­ logical age. This reflects the fact that the study is data-driven, i.e. we ap­ proached the research setting with no a priori hypotheses, in order to remain sensitive to the data and record what was actually happening. 5. Materials and procedure The standardised language measures used were: TROG (receptive gram­ mar), BPVS (receptive vocabulary), CELF-E (expressive grammar), and the Bus Story (the child's ability to give a coherent description of a continuous series of events and at the same time, and the child's use of grammar in context). The standardised non-verbal measures included: the Ravens Coloured Matrices Test and four non-verbal performance tests which form part of the wIsc-R (Picture Completion (PC), Picture Arrangement (PA), Block Design (BD) and Object Assembly (OA)). The rest of the data was spontaneous speech. Every child was seen on at least 5 occasions either at home or at school and every session lasted between 45 and 60 minutes depending on the atten­ tion span of each individual child. All sessions were tape recorded and most of them were video recorded as well. 5.1 Conversational sampling procedure Semi-structured conversational situations were used in order to avoid excessive influence of topic variation on the nature of the interaction. A conversation was developed around a set of photographs thought to represent everyday scenes of which the children might have had some personal experience. 150 conversational turns were then selected and analysed. The inadequate utterances were analysed adopting the categories of inappropriacy

234

STOJANOVIK, PERKINS & HOWARD

of Table 1. The framework was devised by Bishop & Adams (1989) and Adams & Bishop (1989). ESS:

Expressive semantics/syntax

FILIM: Failure to comprehend literal/inferred meaning VES:

Violation of exchange structure (Pragmatics I) Nil response, Ignores initiation

FCIC:

Failure to use context in comprehension (Pragmatics II)

TLI:

Too little information (Pragmatics III) Inappropriate presupposition ('pseudo-ellipsis '), Unestablished referent, Minimal verbal response where more elaborate response is expected, Provides less information that than expected Too much information (Pragmatics IV) Unnecessary assertion/denial, Excessive elaboration, Unnecessary reiteration

TMI: CICS:

O:

Unusually or socially inappropriate content or style Topic drift, Unmarked topic shift, Stereotyped/formulaic, Inappropriate questioning, Socially inappropriate (utterance), Socially inappropriate (behaviour), Inconsistency Other Lack of knowledge/experience,

U:

Repetitions, Ambiguity

Unclassified Table 1 : Categories of inappropriacy

6. Results 6.1 Verbal and non-verbal ability Tables 2 and 3 show the Chronological Age (CA) and z-scores for both groups of children on the above-mentioned measures. +/- indicates whether the score is above or below the mean, and the number gives the distance between the subject's score and the mean in terms of Standard Deviations (SD), the average being between -0.85 and +0.85. The figures in bold indicate performance within the normal range. Table 2 shows that three of the four children with ws have very poor un­ derstanding of grammatical structure as evidenced by their low scores on the TROG. Based on their very low scores on the CELF-E (between 2-4 SDs below the mean), the same three children also show very poor expressive language skills. On the BPVS, two out of four children perform below average while the other two are within the normal range. The scores on the non-verbal measures are more consistent. All four children with WS score below average on the Picture Completion, Picture Arrangement, Block Design, and Object Assem-

LANGUAGE/COGNITION IN SLI AND WILLIAM SYNDROME

235

bly tasks. On the Ravens Coloured Matrices, only one of the children scores within the normal range. Child BW CW DW MW

CA 11;04 12;01 8;06 7;06

TROG -2.26 -2.86 - 1.66 1.73

*PC -3 -2.33 -1.66 -1.66

BPVS CELF-E -0.20 -4.06 -1.87 -2.95 -1.07 -5.36 -0.6 0.12

*PA -2.66 -3 -2.66 -3

*BD -3 -3 -2.33 -2.33

*OA Ravens -2 -6.43 -1 -3.89 -2 -2.84 -2 -0.90

Table 2: Standardised verbal and non-verbal performance - children with WS

Table 3 shows the performance of the children with ws on the Bus Story Task. Child CW BW DW MW

CA 12;01 11;06 8;06 7;06

Info (age) 17(4;3) 13(3;11) 8 (3;9) 11 (4;0)

Sent.Length (age) 12(7;4) 9(5;6) 7(3;10) 13 (7;9)

Subclause (age) 1 (4;2) 1 (4;2) 0(3;10) 3 (6;3)

Table 3: Bus Story - children with WS

All four children score much below their CA on the information they provide. On sentence length three out of the four children with ws again score much below their CA; however one child (MW) performs above her CA. On the use of subordinate clauses, three out of the four children score below the average ex­ pected for their CA, whereas the score of one child (the one who scored above her CA on sentence length) approaches the expected average for her CA. As shown in Table 4, all the children with SLI perform below average on the standardised verbal tests, scoring lower on the expressive (CELF-E) than on the receptive measures (TROG and BPVS). Child BS SS JS TS

CA 9;01 10;03 11;03 9;01

TROG -1.46 -1.86 0.8 -1.73

BPVS CELF-E -2.24 - 1.46 -2.16 - 1.66 -1.33 -1.6 -0.87 -2.39

*PC -1.33 -1 0.33 1.66

*PA 1 1 -0.33 0.66

*BD -1.33 -0.66 -0.33 1

*OA Ravens -1.33 -1.58 -2 -4.33 -1.86 0 0 -0.85

Table 4: Standardised verbal and non-verbal performance - children with SLI

On the other hand, they generally perform better on the non-verbal measures, though their scores vary considerably. They all perform within the normal range on the Picture Arrangement Task. On the Picture Completion and Object Assembly tasks two children out of four show average performance

236

STOJANOVIK, PERKINS & HOWARD

while the other two perform below average. On the Block Design Task, three out of the four children with SLI perform above average whereas only one child performs below the average expected for that chronological age. The scores on the Ravens Coloured Matrices are most variable: two of the children performe at age appropriate levels and the other two children score below, with subject sw scoring as low as 4 SDs below the mean. The children with SLI have a similar spread of scores across categories on the Bus Story (Table 5). Thus their scores on sentence length are slightly better than on information and use of subordinate clauses. Yet they score slightly higher than the children with ws on Information they provide. Child BS TS SS JS

CA 9;01 9;01 10;03 11;03

Info (age) 22 (4;11) 14(4) 25 (5;5) 36 (8;2)

Sent.Length (age) 13 (7;9) 12 (7;4) 14 (8;2) 14 (8;2)

Subclause (age) 4 (7;0) 2 (4;8) 1 (4;2) 3 (6;3)

Table 5: Bus Story - children with SLI

6.2 Conversational ability Table 6 shows the number of inappropriate utterances produced by the children with ws in a 150-utterance conversational sample. ESS FILM VES FCIC TLI TMI SICS Other U

BW 12 (8%) 4 (2.6%) 5 (3.3%) 0 6 (4%) 2(1.3%) 2(1.3%) 0 0

DW 24(16%) 2(1.3%) 3 (2%) 0 7 (4.6%) 2(1.3%) 4 (2.6%) 6 (4%) 2(1.3%)

CW 3 (2%) 0 8 (5.3%) 1 (0.6%) 8 (5.3%) 0 1 (0.6%) 5 (3.3%) 6 (4%)

MW 1 (0.6%) 5 (3.3%) 1 (0.6%) 8 (5.3%) 22(14.6%) 0

 6 (4%) 2(1.3%)

Table 6: Inappropriate utterances - children with WS

The rows show the categories of inappropriacy whereas the columns show the number of inappropriate utterances out of 150 in the relevant categories. The figures in brackets are percentages that correspond to the number of inappro­ priate utterances out of 150. Two of the children with ws are characterised by high rates of inappropriacy in the categories of expressive syntax/semantics.

LANGUAGE/COGNITION IN SLI AND WILLIAM SYNDROME

237

All the ws children had a high percentage of inappropriate utterances in the category of TLI (providing too little information) with one of the children having 14.6% of her inappropriate utterances falling in this category. The conversational profiles of children with SLI (Table 7) are rather dif­ ferent in that they show high rates of inappropriacy across several categories, i.e. in ESS, VES, TLI,TMIand sIcs (see Table 1), with exceptionally high levels of inappropriacy in TMI (providing too much information).

ss

ESS FILM VES FCIC TLI TMI SICS Other

u

TS BS 12(8%) 4 (2.6%) 14 (9.3%) 2(1.3%) 3 (2%) 2(1.3%) 3 (2%) 3 (2%) 10(6.6%) 0 3 (2%) 3 (2%) 8 (5.3%) 7 (4.6%) 5 (3.3%) 22(14.6%) 20(13.3%) 10(6.6%) 5 (3.3%) 13 (8.6%) 1 (0.6%) 2(1.3%) 6 (4%) 10(6.6%) 5 (3.3%) 6 (4%) 2(1.3%)

JS

1

3 (2%) 1 (0.6%) 3 (2%) 3 (2%) 2(1.3%) 12(8%) 1 (0.6%) 0 1(0.6%)

Table 7: Inappropriate utterances - children with SLI

7. Discussion The aim of the present study was to explore the language/cognition inter­ face by attempting to investigate complex patterns of impairments in indi­ vidual cases of ws and SLI, and to show that when detailed information about linguistic, cognitive, and communicative performance is obtained in the same subjects, the two profiles (ws and SLI, respectively) may not be as opposed to each other as it has been widely claimed in the literature. The children's functioning was considered in 3 areas: verbal and non-verbal performance on standardised tests, conversational ability, and narrative discourse. The results on the standardised language tests suggest that claims of 'in­ tact' linguistic functioning in ws should be reconsidered, and caution should be observed when making generalisations. The ws children's performance across standardised language tests was highly variable and almost always in some way deficient, with three of the four children scoring between one and five SDs below the mean, which is indicative of serious problems both in re­ ceptive and expressive language. The results from the ws children's per­ formance across standardised language tests appear to be comparable to those of the children with SLI. Thus linguistic functioning in ws could hardly be considered as 'relative strength'.

238

STOJANOVIK, PERKINS & HOWARD

Three of the four children with ws failed on comprehension of reversible passive sentences, which required an understanding of the underlying syntax of the sentence. They also failed on the embedded structure sentences, such as the subject and object relatives. It follows that the ceiling level performance on syntactic structures reported by Bellugi et al. (1994) and Clahsen & Almazan (1998) does not seem to be the case for all individuals with ws. Although the BPVS scores were in general higher than the grammar comprehension scores, two of the children with ws still performed below what was expected for their chronological age. Children with ws did not show a particular advantage with regard to their expressive language skills either, with three of them performing between 2 and 5 SDS below the mean, and producing structures which were both syntactically and semantically deviant (e.g., after the race I won the race', the car was in the driving car, the boy fell down but he is hurt). Therefore it seems that the children with ws could show a constellation of difficulties across formal domains of language functioning which was often in line with their general cognitive functioning. Although this was not always necessarily the case and there were examples of children with ws who per­ formed exceptionally well in their linguistic functioning (cf. MW), the majority of subjects manifested a constellation of difficulties that cannot be ignored and should be explored in greater detail. The linguistic impairment of children with ws was also apparent in the conversational data. Some utterances produced by children with ws were from 0.66% to 14% inappropriate due to problems with expressive syntax/semantics. What was also striking was their apparent tendency to provide little information, given that it has been claimed that ws children are verbose. In the present study even the child who scored above his chronological age on grammar comprehen­ sion and production (MW) appeared to provide too little information in conversational settings. A possible explanation for this tendency in the ws children may be their cognitive deficits or circumscribed knowledge of the world. They simply did not know what to say, unless it was their favourite topic. The conversational data, however, sheds some light on another feature of children with ws, namely, their apparent willingness to communicate. All the ws children in the study hardly ever used socially inappropriate language, and very rarely failed to use context in conversation, which are key factors in being a successful conversationalist (Table 6). This might be the reason why they gave the impression of being linguistically more able than they actually were. Caution should also be exercised when talking about 'intact' cognition, since the children with SLI varied considerably in their non-verbal perfor­ mance, with one of them scoring as low as 4sDs below the mean on some of

LANGUAGE/COGNITION IN SLI AND WILLIAM SYNDROME

239

the non-verbal tests. However, even the children who performed well on the non-verbal tests revealed considerable inadequacy in their conversational in­ teractions. They often provided too much information, failed to use context in conversation, violated the exchange structure, and used socially inappropriate language by either abruptly changing or drifting away from the topic in order to talk about what they felt most comfortable with. Thus they gave the impres­ sion that they were less able than their performance on non-verbal measures might have suggested. This is presumably because their linguistic difficulties prevent them from revealing their true cognitive abilities. The language/cognition interface cannot be profitably considered without focusing some attention on the psycholinguistic and communicative demands of the production of connected discourse. For this purpose, reference will be made to the Bus Story, which tapped into the ability to produce a short narrative and use grammar in context. The Bus Story consists of a sequence of events experienced by a set of characters located in time and space. Consequently, construction of the narrative entails describing a series of events related both temporally and causally. A skilful narrative does not simply consist of a linear chain of successive events located in time and space. Rather, events must be packaged into hierarchical constructions (Berman & Slobin 1994). In order to achieve this, cognitive and linguistic factors must interact in complex ways. Thus, for example, the number and length of subordinate clauses produced will inevitably influence complexity of thought. That is, subordinating conjunctions represent the key concepts or operators that are necessary for the arrangement of subordinate clauses into various logical relationships with their main clauses, and perform an essential cohesive function in narrative production, i.e., they provide an invaluable tool for refining and manipulating thought, and, thereby, abstract analytical ability. This seems to be lacking (to a varying degree) in both the children with WS and those with SLI (see Appendix for transcripts of the ws and SLI children's Bus Stories). Their overreliance on the coordinating conjunction and results in their stories becoming very factual, resembling enumerations of facts often not logically connected to one another (as a result of the unestablished causativity between them), and at certain points difficult to follow as well. It is worth noting that production limitations both in children with ws and those with SLI occurred not so much in sentence length, but rather in the amount of information that the children provided, and in their handling of structure complexity (i.e., the number of syntactic operations used within a sentence). This suggests that language and cognition interact in complex ways. Cognitive abilities at different stages of development make certain meanings available for expression (Cromer 1974). Thus, if a child with ws

240

STOJANOVIK, PERKINS & HOWARD

does not have the cognitive potential for understanding causal relationships, failure to express them linguistically is not surprising (although the actual lin­ guistic structures may be present in the child's expressive language). On the other hand, certain specifically linguistic capabilities must develop which will allow these meanings to be expressed in language. Hence, although the child may sometimes possess the cognitive potential for understanding causation, the underdevelopment of the child's linguistic abilities will prevent that cogni­ tive potential from being fully realised linguistically. 8. Conclusion The results of the present study suggest that the language/cognition inter­ face is still terra incognita and may have been approached too simplistically in the past. Linguistic knowledge alone does not guarantee appropriate language use (cf. the case of MW), neither does unimpaired performance on cognitive measures (cf. the case of TS). A primary either linguistic or cognitive deficit may contribute to inappropriate language use in conversation and continuous discourse, i.e. may affect communicative ability. Certainly, narrative discourse, which depends upon the manipulation of extended language forms, is not possible until complex language is learned. However, being equipped with complex language forms does not guarantee successful narration because this also requires the involvement of cognitive factors. We should not finish without commenting on MW, the ws child who reveals an uneven profile of relative strengths in language structure, but deficits in cognitive abilities. It is this type of profile which has been assumed to be representative of ws in general and which has led many to postulate that language develops independently of cognition. The only comment we can make at the moment is that this is a very interesting case with implications on linguistic theory, and that such rare but theoretically informative cases are to be welcomed when they are found. The question to be addressed here is what aspects of language might develop independently, and whether pragmatics belongs to the linguistic or the cognitive system or whether it is a domain where both linguistic and cognitive modules interact (Perkins 1998). It is extremely important for clinicians to be aware of the degree of individual variation possible in conditions such as ws and SLI. Given the current state of our knowledge about those language deficits, intervention programmes should be sensitive to the unique interplay of linguistic, cognitive, and pragmatic abilities obtaining in each individual, and should not assume that all cases of ws andSLIconform to an even relatively homogeneous profile.

LANGUAGE/COGNITION IN SLI AND WILLIAM SYNDROME

241

REFERENCES Adams, Catherine & Dorothy Bishop. 1989. "Conversational characteristics of chil­ dren with semantic-pragmatic disorder. I: Exchange structure, turn-taking, repairs and cohesion". British Journal of Disorders of Communication 24.211-239. Arnold, Rosemary, William Yule & Neil Martin. 1985. "The psychological characteristics of infantile hypercalcaemia: A preliminary investigation". Developmental Medicine and Child Neurology 27.49-59. Baddeley, Alan. 1986. Working Memory. Oxford: Oxford University Press. Bates, Elisabeth, Inge Bremerton & Lynn Snyder. 1988. From First Words to Gram­ mar: Individual differences and dissociable mechanisms. Cambridge: Cambridge University Press. Bates, Elisabeth, Philip S. Dale, & Donna Thal. 1995. "Individual differences and their implications for theories of language development". Handbook of Child Language ed. by Paul Fletcher & Brian MacWhinney, 96-151. Oxford: Blackwell. Bates, Elisabeth & Donna Thal. 1991. "Associations and dissociations in child lan­ guage development". Research in Child Language Disorders: A decade of progress ed. by Jon F. Miller, 147-168. Austin, Texas: PRO-ED. Bellugi, Ursula, Shelly Marks, Amy Bihrle & Helene Sabo. 1988. "Dissociations between language and cognitive functions in Williams Syndrome". Language Development in Exceptional Circumstances ed. by Dorothy Bishop & Kay Mogford, 177-189. Edinburgh: Churchill Livingstone. Bellugi, Ursula, Amy Bihrlie, Terry L. Jernigan, Doris Trauner & Sally Doherty. 1990. "Neurospychological, neurological and neuroanatomical profile of Williams Syndrome". American Journal of Medical Genetics 6.115-125. Bellugi, Ursula, Paul P. Wang & Terry L. Jernigan. 1994. "Williams Syndrome: An unusual neuropsychological profile". Atypical Cognitive Deficits in Develop­ mental Disorders: Implication for brain function ed. by Sarah H. Broman & Jordan Grafman, 23-56. Hillsdale, N.J.: Lawrence Erlbaum Associates. Bellugi, Ursula, Liz Linchtenberger & Zona Lai. 2000. "The neurocognitive profile of Williams Syndrome: A complex pattern of strengths and weaknesses". Journal of Cognitive Neuroscience 12: supplement. 7-29. Berman, Ruth A. & Dan Issac Slobin. 1994. Relating Events in Narrative: A cross linguistic development study. Hillsdale, N.J.: Lawrence Erlbaum Associates. Bihrle, Amy M., Ursula Bellugi, Dean Delis & Shelly Marks. 1989. "Seeing either the Forest or the Trees: Dissociation in visuospatial processing". Brain and Cognition 11.37-49. Bishop, Dorothy. 1992. "The underlying nature of Specific Language Impairment". Journal of Child Psychology and Psychiatry 33.1-64. Bishop, Dorothy & Catherine Adams. 1989. "Conversational characteristics of chil­ dren with semantic-pragmatic disorder. II: What features lead to a judgement of inappropriacy?". British Journal of Disorders of Communication 24.241-263.

242

STOJANOVIK, PERKINS & HOWARD

Caramazza, Alfonso & Rita S. Berndt. 1978. "Semantic and syntactic processes in aphasia: A review of the literature". Psychological Bulletin 85.898-918. Carroll, David W. 1994. Psychology of Language. Pacific Grove, CA: Brooks & Cole. Chomsky, Noam. 1965. Aspects of the Theory of Syntax. Cambridge, Mass.: MIT Press. Chomsky, Noam. 1981. Lectures on Government and Binding. Dordrecht: Foris. Chomsky, Noam. 1986. Knowledge of Language: Its origin, nature and use. Dordrecht: Foris. Clahsen, Herald. 1989. "The grammatical characterisation of developmental dyspha­ sia". Linguistics 27. 897-920. Clahsen, Herald & May ella Almazan. 1998. "Syntax and morphology in Williams Syndrome". Cognition 68.197-198. Crisco, Jeffrey J., June M. Dobbs & Reymond K. Mulhern. 1988. "Cognitive processing of children with Williams Syndrome". Developmental Medicine and Child Neurology 5.650-656. Cromer, Richard F. 1974. "The development of language and cognition: The cogni­ tion hypothesis". New Perspectives in Child Development ed. by Bryan Malzard Foss, 184-253. Hamondsworth, Middlesex: Penguin. Cromer, Richard F. 1988. "The cognition hypothesis revisited". The Development of Language and Language Researchers: Essays in Honour of Roger Brown ed. by S. Frank Kessel, 223-249. Hillsdale, N.J.: Lawrence Erlbaum Associates. Fodor, Jerry A. 1983. The Modularity of Mind. Cambridge, Mass.: MIT Press. Fodor, Jerry A. 1985. "Precis of the modularity of mind". Behavioural and Brain Sciences 8.1-42. Gopnik, Myrna & Martha B. Crago. 1991. "Familial aggregation of a developmental language disorder". Cognition 39. 1-50. Grodzinsky, Yosef. 1986. "Language deficits and the theory of syntax". Brain and Language 27.135-159. Karmiloff-Smith, Annette. 1997. "How intact is 'intact'?". Child Development 68.246-262. Karmiloff-Smith, Annette, Edward S. Klima, Ursula Bellugi, Julia Grant & Simon Baron-Cohen. 1995. "Is there a social module? Language, face processing, and theory of mind in individuals with Williams Syndrome". Journal of Cognitive Neuroscience 7.196-208. Lenneberg, Eric Heinz. 1967. Biological Foundations of Language. New York: John Wiley. Leonard, Lawrence ., Letizia Sabbadini, Jeanette S. Leonard & Virginia Volterra. 1987. "Specific Language Impairment in children: A cross-linguistic study". Brain and Language 32.233-252. Levy, Yonata. 1994. Other Languages, Other Children: Issues is the theory of lan­ guage acquisition. Hillsdale, N.J.: Lawrence Erlbaum Associates.

LANGUAGE/COGNITION IN SLI AND WILLIAM SYNDROME

243

Locke, John. 1994. "Phases in the child's development of language". American Scientist 82.436-445. Mervis, Carolyn ., Colleen A. Morris, Jacquelyn Bertrand & Byron F. Robinson. 1999. "Williams Syndrome: Findings from an integrated programme of re­ search". Neurodevelopmental Disorders: Contributions to a new framework from the cognitive neurosciences ed. by Helen Tager-Flusberg, 65-111. Cambridge, Mass.: MIT Press. Perkins, Michael R. 1998. "Is pragmatics epiphenomenal? Evidence from communi­ cation disorders". Journal of Pragmatics 29.291-311. Pinker, Steven. 1984. Language Learnability and Language Development. Cam­ bridge, Mass.: Harvard University Press. Pinker, Steven. 1991. "Rules of language". Science 253.530-535. Pinker, Steven 1994. The Language Instinct: The new science of language and mind. London: Penguin Books. Roeper, Thomas & Harry N. Seymour. 1994. "The place of linguistic theory in the theory of language acquisition and Language Impairment". Other Languages, Other Children: Issues is the theory of language acquisition ed. by Yonata Levy, 305-330. Hillsdale, N.J.: Lawrence Erlbaum Associates. Tallal, Paula. 1980. "Auditory temporal perception, phonics, and reading disabilities in children". Brain and Language 9.182-198. Tallal, Paula. 1984. "Temporal or phonetic processing deficit in dyslexia? That is the question". Applied Psycholinguistics 5.167-169. Tallal, Paula, Rachel E. Stark & David E. Mellits. 1985. "Identification of LanguageImpaired children on the basis of rapid perception and production skills". Brain and Language 25.314-322. Tomasello, Michael. 1996. "Piagetian and Vygotskian approaches to language acqui­ sition". Human Development 39.269-275. Udwin, Orlee & William Yule. 1990. "Expressive language of children with Williams Syndrome". American Journal of Medical Genetics, Supplement 6.108-114. Ullman, Michael T. & Myrna Gopnik. 1999. "Inflectional morphology in a family with inherited Specific Language Impairment". Applied Psycholinguistics 20.51-117. Van der Lely, Heather K. J. & Linda Stollwerck. 1996. "A grammatical Specific Lan­ guage Impairment in children. An autosomal dominant inheritance?". Brain and Language 52.484-504. Van der Lely, Heather K. J. 1997. "Language and cognitive development in a grammatical SLI boy: modularity and innateness". Journal of Neurolinguistics 10.75-107. Volterra, Virginia, Olga Capirci, Grazia Pezzini, Letizia Sabbadini & Stefano Vicari. 1996. "Linguistic abilities in italian children with Williams Syndrome". Cortex 32.663-677.

244

STOJANOVIK, PERKINS & HOWARD

Appendix: The Bus Story A. Children with SLI there was a naughty very naughty bus/ and then the bus drew driven on the road without a driver/ but it didn 't know the brakes were broken/ they had a race but he didn't know who got to win/ but instead the train went under the tunnel and it (..)/and the bus driven very very very very faster than the train/ the policeman tried to stop him 'Stop Bus Stop! /you'll knock everyone flying'/ 'I'm tired of the road'/ 'can I get off this road'/a bus thinks 'I'll probably jump over a hill '/'well done good idea jump over the hill'/'eee I can't believe my eyes moo'/the bus just drove very fast/he span and he didn't know there was a water below/Stop Bus Stop /'you left without me'/so the bus had to be driven by a crane out of the the the water/ so he had to (..) so his (.) he went back to street garage/ ( T S ' s story) one day there was a naughty bus/when his driver was trying to mend him he decided to run off/he ran and ran/he was running beside a train/and they was both pulling faces at each other/he had to go alone because the train was going in a tunnel/he hurried into the city/he met a policeman/he blew his whistle/he didn't pay any attention/he was going/he was going/he got tired of staying on the road/so he jumped over a fence and saw a cow who went 'Moo I can't believe my eyes '/and he was going down the hill/and then he saw a pond underneath it/'help I am stuck '/then he went in the pond and stuck underneath/the driver ran where the bus was/and men find a crane/and then the crane pulled him out/and he was back on the road (JS' s story) again/and he wasn 't very happy/ once upon a time there was a lovely little bus/there was a driver driving the bus on the road/the driver was trying to fix him/but he had no chance/then the bus started to run until he got far away from the driver/this is a (.)/ then he saw and met a train and er (.) I don't know what to say/ they raced each other/the bus went on the road/the policeman told him to stop/but he took no notice/he told the bus to go on the road/he said he was tired/they go to the road/so he jumped over the fence onto the grass/the bus saw a cow/the cow said: 'I'd like to ride into that bus '/then the bus went downhill/fell to a pond/then the bus thought he was going to sink/mind you buses do sink/the bus driver went back into his bus/the bus set off and he said to the driver: 'I will never do that again / (BS's story) and there was a naughty bus/and his driver was (.)/his driver was driving him and he broke down/s the bus driver was mending him but the bus thought 'Oh I'll have a run and go somewhere to do something '/so he was along the road/ and he saw this train/and he was pulling faces/and then the train went under the tunnel so the bus had to go on his own/the bus ran into town/ and there was a policeman/and he blowed his whistle and said: 'Stop bus'/and the bus was alone and to the countryside/and he says: 'I had enough going on the road"/and he jumps over the fence and goes down this hill/and he goes on top of the hill/and he sees a cow/and the cow says: 'Moo I can't believe my eyes '/he went down the hill really fast/and he was at the bottom/and he saw some water but he didn't know how to put his brakes on/and he got into the water and it was muddy there/ (SS's story)

LANGUAGE/COGNITION IN SLI AND WILLIAM SYNDROME

245

B. Children with WS Once upon a time there was a stupid horrible bus/he lived in a bus carriage with his driver/and when the driver jumped out to put this thing on his head the bus didn 't stop/he went on/but the driver raced after it/now this is a funny bit/he passed the train but the train went into a tunnel/but a policeman blowed a whistle/and then he was (.)/decided not he's not staying on the road/he's going to the village/so off he went into the field/and then he saw the little cow who couldn 't believe his eyes/but he isn't scared of it/he raced/he raced down and into the pond/and he was taking him out but now he was good/ (MW's story) the bus wouldn 't work/one two three four/he ran away/to train track/the fireman from (.) he whistle/he says 'Stop '/then he ran away/then he (.)/then he had enough and say 'Ooh, he's had enough ' said the cow/that's the end/ [The adult interferes with an open-ended question in order to elicit further narration from the child, by asking: what happened here?] /he ran in the pool/the bus driver gets a number one/a number one/ (DW's story) one day there was a naughty bus/it was (.)/it was driving along the road/it had a funny face on his back/he runned over in front that policeman/they had a argue that train and that the erm the train and that bus/the train couldn't beat the bus/the bus went (.)/the bus wanted to go into town/the policeman blowed the whistle and the bus didn't know where to stopped/it went at a town/he said: 'I am tired of walking along the road'/he jumped over and went into the field/'mmm I can't believe my eyes '/it got stuck in the lake/the crane(.) the bus had been pulled out by a crane and (BW's story) then it behaved itself/ once upon a time there was a naughty bus/the bus driver was driving it/the bus on the road the driver stopping it/and he runned away/the bus met a train pulling funny faces at each other/the train went under the tunnel/and the bus went on the road/the bus went into the city/a policeman was whistling his whistle: 'stop it! stop '/the bus went over the fence into the country and down the road/and the cow said: 'Moo I can't believe my eyes '/the bus went rushing down the grass and it splashed into the water/driver phones to get the bus out of water/ (CW's story)

V. GRAMMATICAL STRUCTURE IN APHASIA

GRAMMAR AND FLUENT APHASIA* SUSAN EDWARDS The University of Reading 1. Introduction Aphasia is an acquired language problem subsequent to cerebral damage. Aphasia can be sub-divided into two major types, non-fluent aphasia, associated with lesions in the frontal part of the left cerebral cortex, and fluent aphasia, as­ sociated with lesions in the posterior areas of the left cerebral cortex (Goodglass & Kaplan 1983). Non-fluent aphasia is recognised by poor sentence construc­ tion, specifically errors in verb inflection, the use of determiners, the use of ne­ gation and the construction of questions. Speakers with non-fluent aphasia are thought to have a syntactic deficit. In contrast, fluent aphasic speakers use well formed sentences, although they may use less embedding than non-aphasic speakers and make many lexical errors. These speakers are thought to have a lexical semantic problem. However, they occasionally make tense and agree­ ment errors on verbs, make agreement errors when using pronouns and use prepositions incorrectly and sentences may lack obligatory verb arguments. These kinds of errors can be regarded as grammatical errors or the result of lexi­ cal retrieval problems. In this paper I want to discuss the possibility that at least some of the errors made by fluent aphasic speakers arise because access to the grammar, or the implementation of grammatical rules and processes is faulty. Observations that grammatical errors occur in fluent aphasia are not new, but there is little supporting evidence to date compared with what is known about grammatical errors in non-fluent aphasia. The view that grammar is preserved in fluent aphasia and that any errors observed result from faulty lexical retrieval has been largely unchallenged, perhaps because the unequivocal lexical errors in fluent aphasia are more obvious, often more flamboyant and more frequent than the subtle, less frequent, grammatical deficits. Fluent aphasia is usually charac­ terised as a disorder of lexical semantic access where problems listed above and problems with sentence structure are considered to arise from difficulties in ac­ cessing lexical items which in turn arise from either semantically based

Work contributing to this paper has been supported by awards from The University of Reading Research Endowment Fund, the Nuffield Foundation, and the British Council.

250

SUSAN EDWARDS

problems or from problems in phonological representation. However, although lexical problems are characteristic of all fluent aphasic speakers, lexical deficits cannot account for all the errors found in fluent aphasia. Errors in verb inflection are part of a phenomenon known as paragrammatism, a classical symptom of fluent aphasia. Although this portmanteau term suggests a grammatical rather than a lexical deficit, this term does, es­ sentially, include lexical substitution errors known as paraphasic errors. Paragrammatism is also recognised by the substitution of grammatical morphemes (Davis 2000:117). Errors in the use of verb inflection can be seen as a problem in either the lexical domain, where the wrong item is selected, or in the syntactic domain, if one takes the view that verb inflection takes place after lexical selection. In addition to inflectional errors, sentence structure is disrupted: sentence structure is said to be 'tangled' (Goodglass & Kaplan 1983). The no­ tion that grammar in fluent aphasia may be disrupted has been around since the term paragrammatism was first coined by Kleist at the beginning of last cen­ tury. Research reports continue to support this view. For example, Gleason, Goodglass et al. (1980) reported on five patients with moderate to severe Broca's aphasia and five with moderate to severe Wernicke's aphasia. Data from a story-telling task were analysed and a number of features of language, including 'syntactic organisation', were quantified. The researchers found that their Wernicke's aphasic (a type of fluent aphasia) subjects did not produce as many complex sentences as the control subjects. There were fewer embeddings, disjunct clauses, causal clauses and infinitive verb complements. The Wernicke aphasie subjects produced many more simple concatenations and, which the authors interpreted as a preference for the syntactic rather than semantic connectives, such as because. So the interpretation in that study was that semantic rather than syntactic deficits were responsible for at least some of the reduced grammatical abilities Moreover, the fact that and has a different syntactic function from because was not discussed. Research has continued over the last twenty years, with attempts to describe and define grammatical disorders or limitations in this population and, although research reports are few and far between, their results indicate a range of grammatical problems in fluent aphasia. We now know that fluent aphasie speakers make inflectional errors, use fewer complex sentence structures, use fewer embedded clauses, fewer non-canonical sentence structures and fewer complex subject NPs than non-aphasic speakers do (Martin & Blossom-Stach 1986; Butterworth & Howard 1987; Bates, Friederici & Wulfeck 1987; Niemi 1990; Hermann & Kolk 1992; Goodglass, Christiansen & Gallagher 1993, 1994; Edwards 1995; Bastiaanse, Edwards & Kiss 1996; Edwards & Bastiaanse

GRAMMAR AND FLUENT APHASIA

251

1998). Although the phenomena reported would seem to be within the grammatical domain, some researchers deny that these speakers manifest a grammatical deficit and various explanations have been proposed for the errors found. Butterworth & Howard (1987), for example, suggested that the presence of complex sentences in the fluent aphasic speech ruled out a grammatical explanation. They proposed a 'control' deficit, where 'control' is conceived as some kind of cognitive module that is independent of language and yet influences output. Performance variation provoked under different test and sampling conditions led Hermann & Kolk (1992) to propose the adapta­ tion/strategic explanation. Under this account, speakers with fluent aphasia do not adapt to their language deficit and produce errors within normal sentences. In contrast, the non-fluent aphasic speakers do control their output and as a result are not able to produce normal sentences, but over-use sentence ellipsis. While Butterworth and Howard correctly observed the presence of complex sentences, they did not compute the frequency of these sentence types. In a critique of Butterworth and Howard's hypothesis, Harley (1990) makes the necessary distinction between the representation of syntactic rules and syntactic processes and proposes a Reduced Syntactic Disturbance to account for Butterworth and Howard's data. Harley points out that it is possible that paragrammatic speakers retain the rules, but misapply them. Edwards & Bastiaanse (1998:114) made a similar claim, that there is a reduction of grammatical abilities in fluent aphasie speaker, not accounted for by either a lexical or a processing account. Niemi (1990) found differences between his Finnish speakers with fluent aphasia and non-aphasic speakers on a number of variables, such as preference for certain sentence types and elaboration of NPs. Although the extent of studies of paragrammatism does not match those of agrammatism, there is mounting evidence that at least some patients with posterior lesions and Wernicke's aphasia have grammatical problems, which cannot be explained by defective lexical accessing alone. Data presented in this paper illustrate some of the problems encountered in English. Examples of inflectional errors, errors in case assignment and in the use of verb arguments, speakers' facility in sentence construction and the use of complex and non-canonical sentences are discussed. 2.

Methodology Much of the data used in this chapter is taken from one subject, MG, a 65 year old man at the time of testing. Eight years previous to testing he had suf­ fered a cerebral haemorrhage in the temporo-parietal region. Additional data are taken from three other subjects who had also suffered posterior cerebral lesions, but detailed neurological reports were not available. All four subjects

252

SUSAN EDWARDS

were diagnosed by their clinicians as fluent aphasics. Two of the subjects MG and JoH had a clear diagnosis of Wernicke's aphasia confirmed by the Boston Diagnostic Aphasia Examination test (Goodglass & Kaplan 1983) and the experimental English version of the Aachen Aphasia Battery test (Miller, Willmes & De Bleser 2000). Wernicke's aphasia was less severe in the third subject, MF than it was in MG and JOH. TR, the fourth subject presented as an anomic aphasie except he had some comprehension problems as did all the other three subjects. In all cases aphasia was the consequence of Cerebral Vascular Accident (CVA). All subjects were at least 212 months time post onset (T.P.O) of CVA. This information is displayed in Table 1 below. Subject JoH MG MF TR

Sex M M F M

Age 76 65 81 72

T.P.O

5 8 2 1

yrs.

Aetiology CVA CVA CVA CVA

Aphasia type Wernicke's severe Wernicke's moderate/severe Wernicke's moderate Anomie mild

Table 1 : Subject information

Two types of data are used, spontaneous speech data, that is story telling, with and without a picture, and data from constrained tasks. Tasks include sentence completion, repetition tasks, producing inflected verbs and anagram tasks. Some of these tasks were taken from the VAST test (see Rispens, Bastiaanse & Edwards, this volume). The tasks and results will be discussed under aspects of language production; verb inflection; sentence construction; lexical selection; complex sentences and thematic role assignment. We will also consider the ability of one subject to judge whether a sentence is well formed. 3. Investigations and results 3.1 Verb inflection in spontaneous speech Speakers with fluent aphasia produce inflectional errors in their spontaneous speech, when we might assume there is little conscious planning. Butterworth & Howard (1987) found inflectional errors in the spontaneous speech of each of the five fluent aphasie speakers they examined. Fifty-six of the total 226 errors, that is a quarter of all errors logged, were classified as 'in­ flectional errors' including errors of verb inflection, such as: (1) (2) (3)

He go and set on. Right, and I wented with it. He's went to picks the [dikis].

GRAMMAR AND FLUENT APHASIA

253

In (1), the speaker fails to use the correct 3rd person singular inflection. This can be seen as an error of omission or, as an example of the speaker using the base form of the verb. Inflectional errors in paragrammatism typically include additions as well as omissions. (2), for instance, is an example of a 'double' past tense form; (3) is an example of an additional auxiliary verb (he's), as well an unnecessary and illegal past tense inflection on a verb infinitive. Butterworth and Howard also included agreement errors in this category, such as: (4)

Want a good towels, yes.

We have found similar errors in the patients seen in our clinic. Moreover, some further examples from conversational data collected from MG include errors in 3rd person singular (come/comes), as well as agreement (pint/pints). (5) (6) (7)

They had Wogan ... he come yesterday ... Wogan ... he came in ... that is very good ... that was ... very good. All the water ...waters... means you comes up the thing ... all the boats. I go to the Spread and they're very ... they can get about... I used to do that... he goes about three or four .... Two chaps him and her ... he has about three or four (therapist: what, whiskies?) no ... pint.

In (5) the speaker is either incorrectly substituting the null inflection in­ stead of past tense, come for came, or is using the verb stem. For MG, this is an incorrect 3rd person singular, past tense form, as can be seen as he corrects himself, he come yesterday ...he came. In (6), he uses the 3rd person singular form although in this position, the 2nd person is required. In (7), there is an agreement error, pint for pints. Admittedly, it is spread across a conversational turn, but it remains an error. These errors are infrequent for this particular speaker: there were only three errors in a 300 word sample, less frequent than in Butterworth and Howard's data or that reported by Hermann & Kolk (1992) for their eight Dutch speaking Wernicke's aphasic subjects. The MG subject, whose data are discussed in this paper, does not make frequent errors, but that is why the in­ crease that occurs in constrained tasks is interesting. Such tasks serve to high­ light problems that will now be described.

254

SUSAN EDWARDS

3.2 Verb inflection in constrained task In this section we will examine subjects' abilities to provide an inflected verb within highly constrained tasks, where, we might assume, there is some level of conscious planning. For example, the task might be to produce an in­ flected verb when attention is focused on the production of the verb. The goals of these tasks are not usually made explicit to the subject, although the procedure was. In all cases reported in this paper, subjects were aware of task requirements, despite their comprehension problems. This is not always the case for fluent aphasic speakers, but the data we consider are selected only if the examiner was confident that the subject understood the explanations that were given. In the first task reported, the subject is given a word, either an inflected verb or noun and is asked to repeat it. The second is a sentence repetition task and the third is a sentence completion task, where the speaker is required to provide the inflected form of the verb in one half of the exercise, and the infinitive form in the second half of the exercise. In this section we will discuss first the ability of the MG subject to repeat inflected verbs and nouns, and then look at some further data from three other subjects with fluent aphasia. MG performed poorly when repeating regularly inflected words, omitting inflectional endings in five of the twenty words given to him: Target: Rocks canned smiled kissed freed Response'. Rock can smile kiss free In these examples, despite his aphasia and associated problem with repetition, MG can repeat the lexical stem. What he fails to do is to repeat the stem with the correct inflectional morphology. He also makes errors in sentence repeti­ tion tasks and, given the nature of these errors and the errors made in repeating single words, it is unlikely that these errors are the result of poor auditory memory. It was established that MG was able to repeat simple phrases and sentences. For example he could repeat eat cake, man running, he went home, who saw the man, but failed with other short sentences. Consider: (8) (9) (10)

Response He he's the running. The cat was angry now. Were this lady was going.

Target He's running. The cat is angry now. Where is the lady going?

In (8), he repeats he's after a false start, but fails to complete the verb phrase correctly. He changes the tense in (9) and in (10). In each case he pro­ duces a meaningless sentence.

GRAMMAR AND FLUENT APHASIA

255

A new repetition task was given using sentences with very simple structures, but this time containing either syntactic or semantic violations. MG found it more difficult to repeat the sentences if they had syntactic errors (3 correct out of 13: 23%), than if the sentences had semantic anomalies (6 out of 13: 46%). This result suggests that syntactic violations were better detected, at some level, than were semantic errors. His, at least partial, awareness of the syntactic errors disrupted his repetition of sentences containing such errors, whereas he was more likely to overlook semantic anomalies and thus gained a higher repetition score. Although his score was better for sentences containing semantic anomalies than for those containing syntactic errors, overall scores on this test were poor. It had been established that there were limitations in his auditory memory (he was only able to recall four digits) and these limitations would affect his ability to repeat. However, if he produces ungrammatical sentences, he doesn't attempt to correct them. It is not simply a matter of not remembering the sentence; in this task he is not always able to structure a well-formed sentence. In a sentence completion exercise taken from the VAST test the speaker is required to supply ten finite verbs in the first part of the exercise and ten in­ finitive verbs in the second part of the exercise. The subject is shown a picture and read a sentence, e.g., the man (digs) his garden or, in the case of the in­ finitive exercise, the man ¡ikes to (eat). The task is to supply the missing verb in either the finite or the infinitive form, according to the sentence structure. Nonaphasic speakers make no errors on these tasks. MG, however, found the task difficult. As we have noted above, although he makes few verb inflection errors in his spontaneous speech, in this exercise, for nine out of ten of the finite verb test items, he either produced the non-finite {washing) or the verb stem (smoke, water). When he was given a sentence to complete, such as the man {waters) his garden, he supplied the lexical verb, but failed to inflect it. However, given that he was asked to produce single words, we cannot confidently classify his responses as verbs. In the example given, he may have been naming a salient feature in the picture. We have found this to be true for other fluent speakers. Results from this test for four fluent aphasic speakers are in Table 2. Subject JoH MG MF TR

Finite 0% 1 5 3

Infinitive 3% 8 9 9

Table 2: Percentage correct of finite and infinitive verbs

256

SUSAN EDWARDS

All four subjects are much better at producing verbs in the infinitive than in a finite form although the severity of their aphasia varies across subjects. TR rarely produced inflectional errors in his spontaneous speech yet he only produced three correctly inflected verbs. 3.3 Sentence construction Speakers with fluent aphasia have problems with sentence construction and examples can be found in spontaneous speech data, picture description and anagram tasks, and so on. In this section some results from two types of tasks are given. In the first, the subject is required to provide a sentence to describe a picture. In the second, an anagram task, the subject has to order written sentence constituents to form a sentence. Other tasks reported in this paper include picture description, sentence completion, repetition and anagrams: procedures differed with the presence or absence of pictures, and /or the presence and absence of written language accompanying the task. Speakers with severe aphasia find such tasks difficult while those with a mild deficit may, like TR, do well. Nevertheless, this subject TR had occasional difficulties with sentence construction. For example, he found difficult to complete a sentence with an initial subordinate clause. TR who had progressed from a diagnosis of Wernicke's aphasia to that of a fluent, anomic aphasia, produced the following in a picture description task: (11)

Response Target

A chap who is throwing the stick She/the girl is throwing the stick.

In (11) the speaker tried to form a complex sentence, although the task did not require this, but he failed. Having started with a NP followed by a wellconstructed subordinate clause, the speaker was unable to complete his own sentence. In this case the subject made it clear that he wasn't satisfied with his effort. For TR, this was the only error on a 20-item sentence construction task. The other three subjects had more difficulty with the picture description task. Examples of errors are in (12)-(16) and scores are in Table 3. (12) (13) (14) (15) (16)

Response for the target The girl has a stick pushing. Walking the pavement. The girl is slapped him. A Gibraltar taking a monkey. The cat is ... dog eats the cat.

Target The girl is throwing a stick. The man is walking. The boy is hitting the girl. The monkey is eating the banana. The dog is biting the cat.

In (12), the selection of pushing, rather than the target throwing, renders the sentence non-grammatical, as it lacks one of the three obligatory verb

GRAMMAR AND FLUENT APHASIA

257

arguments. In contrast, in (13) the subject, MG, appears to be able to retrieve two of the required lexical items, but without a NP in subject place and an auxiliary, he is unable to construct a sentence. In (14)-(16) there is an incorrect thematic role assignment. (14) also has also a tense error. In table 3, the results for 20 items are given. Subject JoH MG MF TR

Correct sentences: 3 4 15 19

Table 3: Scores for the sentence construction task

In the anagram tasks in the VAST, subjects are required to order sentence constituents to make either active or passive sentences, with and without pictures, and to construct questions with pictures. Constituents are laid before the subject in random order. An example is: the boy the girl

is kissing is kissed

the girl by the boy

There is an additional implicit requirement in these tasks and that is that the subject has to create a sentence that matches the thematic roles depicted in the picture. This is achieved by the ordering of the NPs around the VP. As there is no requirement to speak, some researchers regard these tasks as mapping into comprehension rather than production skills. In fact, all the subjects I have tested read at least some of the cards aloud and are often mislead by their own mis-reading of the constructions. Fluent aphasic speakers find these tasks difficult and make errors of constituent order. Although they may indicate that they are unhappy about their performance, they are unable to correct their er­ rors. This does not hold for all fluent aphasic speakers, of course. Performance varies a great deal (Table 4). Subjects JOH MG MF TR

Anagrams +pictures Anagrams -pictures 10 10 12 11 18 20 20 20

Table 4: Number anagrams arranged correctly

258

SUSAN EDWARDS

All these subjects spent some time considering the pictures and reordering the sentence constituents until they were satisfied that the sentence they had created matched the picture. The results, considering 20 items in each section, were not always successful. 3.4 Lexical errors Speakers with paragrammatism make errors with closed class words as well as with open class words and these errors have consequences for wellformed sentences. Examples of pronoun and preposition errors taken from some informal tasks given to MG are in (17)-(21). In (17)-(19) he was required to produce a sentence, while in (20) and (21) to describe a picture. (17) (18) (19) (20) (21)

John misses you friends very much I don't know if his wants to I asked Sue if she liked John she said she likes his He's having a drink on the table The little one is going his dad

Examples (17)-(19) may be classified as errors of lexical substitution or as problems with case assignment; in (20) there is a pragmatically strange prepo­ sition; in (21) a preposition is omitted. Some fluent aphasic speakers occasionally fail to provide obligatory argu­ ments in contexts where they appear to be happy with their production. Such errors seem to be different from failure to provide obligatory arguments because of the intonation used. The following examples are taken from MG's data. (22) (23)

He ' s putting wallpaper I was taking my wife up

The result of errors of case on pronouns or of verb argument, whether con­ ceived as lexical, or grammatical, is that the sentence produced is ungrammatical. Completing argument structures and providing pronouns, correctly marked for case, involve both semantic and grammatical processes and it is the involvement of the two domains that is apparent in these errors. Problems involving both semantic and syntactic processes are also apparent when fluent aphasic speakers are required to process or manipulate sentences that have non-canonical word order, as we will see in the next section. Errors that result from the tasks discussed below suggest that, like some agrammatic speakers, there are problems with thematic role assignment and movement of sentence constituents. It could be argued speakers who produce, such utterances make, such errors because of their incomplete understanding of the

GRAMMAR AND FLUENT APHASIA

259

lexical items in the sentences, but the differential performance on the different sentence types suggests that this an unlikely explanation. 3.5 Complex sentence construction in spontaneous speech We have so far considered how some fluent aphasic speakers have problems with sentence formulation and this is highlighted when given various sentence construction tasks. Looking at transcriptions of spontaneous, connected speech, it becomes obvious that a significant proportion of that speech is in sentence fragments. Sentences are aborted as lexical retrieval fails, but they may be aborted for other reasons. Problems with sentence construction may not just be a consequence of the lexical problem, but diffi­ culty with constructing complex sentences may contribute. Edwards (1995) presented analyses of two fluent aphasic speakers, both of whom scored below the non-aphasic controls on Saffran, Berndt & Schwartz (1989) embedding index. Both aphasie subjects had an embedding index of 0.09 compared with the 0.53 index given by Saffran et al for the non-aphasic controls. This finding was replicated by Edwards & Bastiaanse (1998) re­ porting on groups of Dutch and English speaking fluent aphasie speakers. They found that only one of the English subjects had a proportion of subordi­ nate clauses that fell within the normal range. This factor could not be ex­ plained by problems with either verb or noun retrieval as there was no clear relationship between lexical retrieval and the ability to construct complex sentences. Although four of the ten English subjects used a smaller variety of verbs than the control subjects, all but one subject had the same number of verb tokens as the normal speakers. There was not, then, a simple explanation along the lines that the fluent aphasie speakers produced fewer subordinate clauses because they produced fewer verbs. The ability to produce a number of verb tokens that fell within normal limits was not associated with the ability to produce subordinate clauses. 3.6 Sentence construction in tasks and exercises The difficulty with complex sentence production is also seen in tasks where the subject has only one sentence to think about. We have seen above how fluent aphasie subjects have difficulty in ordering sentence constituents in an anagram task. If we take a closer look at the anagram task we find that there is some evidence that, for at least some fluent aphasie speakers, the type of sentence structure required has an effect on performance with noncanonical sentences being more difficult than canonical sentences except for subjects that perform at ceiling. Two types of non-canonical sentence types are examined, passives and WH questions.

260

SUSAN EDWARDS

Table 5 shows that the performance of three fluent aphasic speakers varied within the anagram tasks containing active and passive sentences (TR has been omitted as he performed at ceiling.). The performances were not so much in­ fluenced by the presence or absence of a picture, but influenced by sentence type. Subjects JoH MG MF

Actives + pictures - pictures 90% 100% 80% 100% 100% 100%

Passives + pictures - pictures 10% 0% 30% 20% 80% 100%

Table 5: Sentence anagrams, with and without pictures : percentage correct

For two subjects MG and JoH, anagrams of active sentences are easier to construct than anagrams of passive sentences. If this difficulty were a conse­ quence of lexical retrieval alone, then we might expect the presence of the picture to enhance lexical retrieval and thus sentence construction. Martin, Wetzel, Blossom-Stach & Feher (1987:177) have suggested that aphasic per­ formance may improve in tasks where pictures are involved as the presence of pictures leaves "more capacity for the development of syntactic structure or the retrieval of function words or inflections". Caplan, Waters & Hilderbrandt (1997) found some evidence that the presence of pictures might enhance per­ formance on certain tasks. However, for MF, the presence of a picture produced a poorer performance, which, at first glance, may seem difficult to explain. However, the explanation lies with the type of sentences used. Only those passives with pictures included reversible passives and it was on these sentences that she made two errors. For the two subjects with more severe Wernicke's aphasia, the passive structure was more difficult than the active. These subjects frequently failed to place the NPs in the correct place, placing the agent in first NP position in the passive sentence, even when the sentence was non-reversible. It might be suggested that if they had intact access to the meaning of the NPs and the VPs then their semantic knowledge would prevent, such errors at least on the non­ reversible sentences. However, both JOH and MG scored poorly on the non­ reversible sentences suggesting that they were not able to assign thematic roles reliably. In the question anagram task, subjects were required to order the cards con­ taining sentence constituents to match the action shown in the picture. All four subjects found this a difficult task and either frequently placed the constituents in orders where the NPs clearly did not match the thematic roles depicted in the pictures (JoH and MG), or worked very slowly (MF and TR). So, for example,

GRAMMAR AND FLUENT APHASIA

261

given the anagram task of constructing the question who is the boy pushing in the water', MF (a subject with severe to moderate fluent aphasia) constructed: (24)

Who is pushing the boy in the water.

Although (24) is a grammatical sentence, it does not match the picture, where the boy is clearly the agent not the beneficiary. For the target question what is the girl photographing in the grass the same subject constructs: (25)

The girl in the grass is photographing what.

It might be argued that this is a grammatical sentence, which it is within certain pragmatic constraints, but it also can be seen as an example of failure of movement. Movement of what is required to produce the target, more common, sentence. MG, who had a more severe aphasia, produced reversed thematic role order even in sentences where the NPs had dubious reversibility as he had in the sentence anagrams. (26) (27)

What is kissing the princess by the pond. When is the child washing the mother.

Subjects with severe aphasia are unable to do this task. JoH only assembled two of the twenty questions correctly yet in attempting the task, he always selected the card with the WH word to start the sentence and this, coupled with his reaction to his performance, suggested that he understood the task. The scores shown in Table 6 demonstrate how this task differentiates between those with moderate/severe aphasia and those with mild aphasia. Non-aphasic controls can do the task with a greater than 90% success rate. Subject JoH MG MF TR

Percentage correct: 20 items 10 63 85 100

Table 6: WH anagrams: percentage correct

In the anagram task, the subject is not required to retrieval lexical items and sentence construction may still be problematic. 3.7 Judgement tasks Further tests with MG that tapped into his knowledge of the inflectional system, showed that his grammar was better preserved than we might have as­ sumed if we just looked at his ability on the production tasks. Given a listening task, MG tended to make more reliable judgements on syntactic well-

262

SUSAN EDWARDS

formedness than he did on semantic appropriacy although judgment was not always 100%. Errors occurred and he indicated that he was not sure of many of his responses in the tasks reported in this paper MG. correctly detected all grammatically correct sentences (100%) and he was 80% correct at detecting grammatical violations. He was less successful detecting sentences that were correct as opposed to those that were incorrect and this held for both syntacti­ cally and semantically violated sentences. Overall, he was less successful de­ tecting semantic violations. In this task he identified only 66% of sentences containing semantic anomalies. This performance is in line with his relatively superior ability to repeat sentences with semantic violations than those with syntactic violations. The results of this task suggest that he could identify syntactic violations although his slow and hesitant performance suggested that he did not have ready access to this knowledge. He was much better at detecting errors than repeating sentences with grammatical violations, but in both tasks we see that although his syntactic skills seem less impaired than his semantic, he still has difficulty with tasks that tap into his grammar. If grammatical knowledge is more or less intact, access to it and application of that knowledge is faulty, es­ pecially in production tasks. In a task where no production was required, he has access to grammatical information: access to grammar is much better than access to semantic information, though, that access to grammatical knowledge is fragile. His scores do not match scores of normal controls who would be expected to score at or near to 100%. MG was given a variety of judgment tasks during therapy. By varying sen­ tence type it was possible to explore whether syntactic complexity was a factor in his ability to detect errors. Given the task of judging whether a preposition was correctly used within a sentence, MG could judge correctly if a sentence was grammatically correct, but his performance was better if the sentence had a simple, active structure. When the sentence had a noncanonical structure, he made more errors. We can see below that he made more errors of judgment with passive than with active sentences. In all three conditions, he finds it easier to identify a correct sentence than an incorrect sentence and this is true irrespective of sentence type as can be seen in Table 7 below (10 items are considered in each section). Correct sentences Incorrect sentences

Active sentences 0 errors 7 errors

Syntactic passives 4 errors 6 errors

Lexical passives 2 errors 9 errors

Table 7: Judgment of the suitability of prepositions within different sentence types

GRAMMAR AND FLUENT APHASIA

263

4. Summary of findings and discussion In this paper we have examined the speech of four speakers with fluent aphasia, concentrating on one speaker, MG who has a classic Wernicke's pro­ file. It has been demonstrated that although speakers with fluent aphasia can use inflection, mark case appropriately and use complex and non-canonical sentences, they make errors in all these domains. They can correctly assign thematic roles within sentences, but they also make errors. They can produce verbs and argument structures correctly, but they also produce verbs without all obligatory arguments. Inflectional errors occur, errors in case marking oc­ cur, the frequency of complex sentences is reduced and the assignment of the­ matic roles may malfunction. Performance on error detection tasks suggest that the fluent aphasic speaker examined in this paper retained some ability to judge syntactic well-formedness that was superior to his ability to detect se­ mantic appropriacy. He and the other subjects showed some awareness of their grammatical errors although were not always able to self-correct. The data reviewed suggest that Blanken et al (1992:100) claim that speakers with paragrammatism 'are able to apply all the rules', is only partially correct. The tasks reported in this paper highlight difficulties in producing inflection and in structuring sentences, but at what level does the difficulty lie? It is unlikely that the grammar (i.e., grammatical representations and processes involved in sentence construction, the formation of questions, the inflection of verbs etc.) of the aphasic speaker is damaged. As we have argued elsewhere for agrammatism, the fact that the aphasie speaker has access to correct grammatical structure and can implement grammatical processes, at least some of the time, suggests that aphasia results not from a damaged grammar, but from intermittent access to an intact grammar (Arabatzi & Edwards 1999; Edwards & Lightfoot 2000). What is proposed in this paper is that although speakers with fluent aphasia can apply grammatical rules much of the time, application, is not reliable. There is a deficit in the application of rules and processes, and this, along with the semantic, lexical deficit, is part of fluent aphasia. Errors in inflection suggest that the implementation of the inflectional process (and this will be conceptualised in different ways, using different terms depending on the theoretical model selected), is not always functioning normally. Taking one model, Chomsky's Minimalist Program, it could be that faulty application of the processes of checking and merging allows inflectional errors to be produced, errors which in the healthy speaker, crash. The grammatical processes that generate embedded and non-canonical sentences are also faulty and so fewer complex or non-canonical sentences are produced.

264

SUSAN EDWARDS

Errors discussed in this paper may be conceptualised as 'maximally' a processing deficit (Niemi 1990:402), but alternatively can be seen as being at the mild end of the same grammatical deficit exhibited by agrammatic apha­ sics. Neuro-imaging, while largely confirming the role of Broca's area in grammatical production, shows more widespread neural activity than the clas­ sic syndromes suggest and, furthermore, the effect of neural damage in one cortical area is likely, through the complex neural networks of the brain, to have consequences for other cortical areas. It is not surprising then that while damage to Broca's area has the more serious consequence for grammar, damage to posterior regions may result in similar although milder grammatical deficits. In both cases, the same language system is vulnerable and similar errors occur. Grodzinsky (2000:14), while making a case for syndrome specificity, does acknowledge the possibility of some syntactic abilities in the posterior region of the brain, but notes that disturbances observed are less tangible and less characterisable than those arising from damage to the frontal cortex. 5.

Conclusions In both types of aphasia it would seem likely that errors arise as the speaker tries to access or implement the grammar. Sometimes it works, sometimes it does not. For the fluent aphasic speaker, there is an additional difficulty in lexical retrieval. All aphasie speakers have problems with lexical retrieval, but these are especially marked in those speakers whose aphasia is a consequence of damage to the posterior region of the dominant hemisphere. Error in sentence construction may arise from problems in lexical access and difficulties in preposition and pronoun selection may be interpreted as lexical rather than grammatical errors, although this distinction is not especially useful here. But lexical accessing deficits will not explain all errors or the difficulties encountered on the anagram tasks. The different task performances, reviewed above, suggest that Butterworth and Howard's 'diminished control' explanation is hard to sustain. We might expect that a control module would be functioning at maximum efficiency where the speaker's attention is focused on a task. We might expect that is a condition where the sole effort is to produce an inflected verb, a 'control module' would be better able to detect errors. This, however, was not the case for the four fluent aphasie speakers we have reviewed. Their performance diminished. Caplan et al (1997:552) claim that results on their sentence comprehension tasks demonstrate that "clinical classification is not a good guide to patient

GRAMMAR AND FLUENT APHASIA

265

performance in syntactic comprehension". This is a claim too strong to adopt for production abilities, given the data discussed in this paper. For example, all the subjects described here were better at sentence construction than typical agrammatic patients. They all produced utterances that were considerably longer than agrammatic speakers can produce. However, the errors made by the fluent aphasic speakers were of the same kind if not the same magnitude as those made by agrammatic speakers. The conclusions drawn at this stage of our knowledge about this group of patients has to be that, in the same way as non-fluent aphasic patients, they have faulty access to their grammar. We can no longer assume that syntactic abilities are intact in this particular aphasie population.

REFERENCES Arabatzi, Marina & Susan Edwards. 1999. "Optionality and inflections in Agram­ matic speech". Brain and Language 69.270-272. Bastiaanse, Roelien, Susan Edwards & Katalin Kiss. 1996. "Fluent aphasia in three language: aspects of spontaneous speech". Aphasiology 10.561-575. Bates, Elizabeth, Angela Friederici & Beverly Wulfeck. 1987. "Grammatical mor­ phology in aphasia: evidence from three languages". Brain and Language 32.19-67. Blanken, Gerhard, Jurgen Dittmann, Hannelore Grimm, John  Marshall & ClausWallesch, eds. 1992. Linguistic Disorders and Pathologies. Berlin: Walter de Gruyter. Butterworth, Brian & David Howard. 1987. "Paragrammatisms". Cognition 26.1-37. Caplan, David, Gloria Waters & Nancy Hildebrandt. 1997. "Determinants of sen­ tence comprehension in aphasie patients in a sentence-picture matching task". Journal of Speech, Language and Hearing Research 40.543-555. Davis, Albyn D. 2000. Aphasiology: Disorders in clinical practice. Needham Heights, Mass.: Allyn & Bacon. Edwards, Susan. 1995. "Profiling fluent aphasie speech: A comparison of two methodologies". European Journal ofDisorders of Communication 30.333-345. Edwards, Susan & Roelien Bastiaanse. 1998. "Diversity in the lexical and syntactic abilities of fluent aphasie speakers". Aphasiology 12.99-117. Edwards, Susan & David Lightfoot. 2000. "Intact grammar, intermittent access". Brain and Behavioral Sciences 23.31-32. Gleason, Jean Berko, Harold Goodglass, Loraine Obler, Eugene Green, Mary Hyde & Sandra Weintraub. 1980. "Narrative strategies of aphasie and normal-speaking subjects". Journal of Speech & Hearing Research 23.370-382. Goodglass, Harold, Julie Ann Christiansen & Roberta Gallagher. 1993. "Comparison of morphology and syntax in free narrative and structured tests: Fluent vs. Nonfluent aphasies". Cortex 29.377-407.

266

SUSAN EDWARDS

Goodglass, Harold, Julie Ann Christiansen & Roberta Gallagher. 1994. "Syntactic constructions used by agrammatic speakers: Comparison with conduction apha­ sics and normals". Neuropsycholog 8.598-613. Goodglass, Harold & Edith Kaplan. 1983. The Boston Diagnostic Examination of Aphasia. Philadelphia: Lea & Febiger. Grodzinsky, Yosef. 2000. "The neurology of syntax: Language use without Broca's area". Behavioral & Brain Sciences 23.1-21. Harley, Trevor. 1990. "Paragrammatism: Syntactic disturbance or breakdown of con­ trol?". Cognition 34.85-91. Harmann, Henk & Herman Kolk. 1992. "The production of grammatical morphology in Broca's and Wernicke's aphasics: Speed and accuracy factors". Cortex 28.97-112. Martin, Rand & Carol Blossom-Stach. 1986. "Evidence of syntactic deficits in a fluent aphasie subject". Brain and Language 28.196-234. Martin, Rand, Fredrick Wetzel, Carol Blossom-Stach & Edward Feher. 1987. "Syntactic loss versus processing deficit: An assessment of two theories of agrammatism and syntactic comprehension deficits". Cognition 32.157-191. Miller, Nick, Klaus Willmes & Ria De Bleser. 2000. "The psychometric properties of the English language version of the Aachen Aphasia Test (EAAT)". Aphasiology 14.683-722. Niemi, Jussi. 1990. "Non-lexical grammatical deviations in 'paragrammatic' apha­ sia". Folia Linguistica 24.299-404. Rispens, Judith, Roelien Bastiaanse & Susan Edwards. 2001. "The verb and sentence test: Assessing verb and sentence comprehension and production in aphasia". This volume, 279-298. Saffran, Eleanor, Rita S. Berndt & Myrna Schwartz. 1989. "The quantitative analysis of agrammatic production: procedure and data". Brain and Language 37.440-79.

FAILURE TO AGREE IN AGRAMMATISM * ANNA GAVARRÓ Universitat Autónoma de Barcelona 1. Introduction The purpose of this paper is to consider, in respect of inflection, the rearing on linguistic theory of aphasia. Several researchers undertake the grammatical analysis of linguistic disruption on the assumptions of contempo­ rary generative grammar. In a similar vein, we will examine the inflectional deficits reported in the literature on aphasia, which includes analyses in terms of phonology, morphology and syntax, and we will see how they can be accommodated in contemporary minimalism, holding minimal impairment as a basic assumption. The association between agreement (to which inflection relates) and movement embedded in the minimalist theories will also be considered and shown to be at least potentially problematic. The proposal will be made to accommodate for the facts of agrammatism in a natural way. 2. The empirical domain The empirical domain which we aim to explore is that of inflection in agrammatism.1 The deficit of inflectional markers (including Case markers) and function words associated with aphasia was recorded long ago in the literature: the example (1) was observed by Deleuze in 1819, reported by Pitres (1898) and quoted in Goodglass (1976). (1)

Souhaiter bonjour, rester, mari venir. wish-INF good-morning remain-INF husband come-INF "Said good-morning, stayed, the husband came."

This paper was presented at the conference Linguistic Theory, Speech and Language Pathology held in Padova in August 2000. I am grateful to the audience there for their comments and suggestions; any remaining errors are my own. I would also like to acknowledge the MEC for its financial support through project BFF2000-0403-C02-02. 1 Inflectional deficits have also been noted in patients suffering from fluent aphasia, both in their spontaneous speech and in experimental tasks (Edwards 2001, this volume). It would be interesting to see to what extend an account of inflection in agrammatism covers some of the facts of fluent aphasia, but we will not pursue this matter here.

268

ANNA GAVARRÓ

In this example, the French finite verbs have been replaced by their nonfinite, infinitive, counterparts. Forms departing from the normal adult form are also illustrated for English in (2) (Goodglass 1976:239), for Catalan in (3) (Gavarró 1993:3), for Spanish in (4) (Gavarró 1993:4), and for Hebrew in (5) (Grodzinsky 1990). The Catalan and Spanish data were recorded by the author from the spontaneous speech of several aphasic patients at the Hospital Clinic de Barcelona (see Gavarró 1993).2 (2)

Uh, oh, I guess six month ...my mother pass away.

(3)

No me'n recordo. No sortir. notREFL of-it remember-IS not come out-INF "I don't remember. It doesn't come out." (4) a. casas "houses" (repetition of 'casa', "house") b. La niño tiene un vestido. DET-F child-M has a dress (repetition of 'La niña tiene un vestido', "The girl has a dress") c. Ser correcto.  e. INF correct "That's correct." d. Mañana iré al cine tomorrow go-FUT-lS to-the cinema (repetition of 'Mañana iremos al cine', "Tomorrow we'll go to the cinema") (5) a. xamesh yamin five-F days-M b. tiylu anaxnu ba'ali ve'ani walked-3PL we my-husband and I "My husband and I walked."

The examples above suffice to show that the inflectional deficit associated with agrammatic aphasia involves omissions of inflectional markers (2), as well as substitutions, both in concatenative (1), (3), (4), and non-concatenative morphology (5). The first approach to inflectional disorders in agrammatism within genera­ tive grammar was that of Kean (1977). She claimed that phonological clitics, as opposed to phonological words, are omitted in agrammatism; her prediction was thus the occurrence of omission errors only. Lapointe (1983) offered a 2

Mismatches in agreement have been found accepted as well (e.g., Zurif & Grodzinsky 1983).

FAILURE TO AGREE IN AGRAMMATISM

269

reanalysis of the same data in morphological terms. Grodzinsky (1990) argued convincingly that the deviant forms of the kind exemplified (empirically more diverse than predicted by Kean & Lapointe) result from syntactic impairment, rather than a morphological or phonological one. More recently, Bastiaanse & van Zonneveld (1998) have dismissed a morphological account of verb in­ flection errors like the one in Bates & Wulfek (1989) showing that no mor­ phological disorder can explain the problems in verb placement (not morpho­ logical ill-formedness) exhibited by their Dutch speaking patients. The categories affected by this syntactic impairment are either tense, aspect and person on the verbs, as illustrated in (1), (2), (3), (5), or gender and number on D/N categories, as in (3)-(5). The claim has been made that there is a more selective inflectional im­ pairment which affects only a subset of the inflectional categories above. Thus Hagiwara (1995) and Friedmann & Grodzinsky (1997, 2000) describe the fol­ lowing cases of impairment of a subset of categories: CP for Japanese in (6), TP (but not AGRP) for Hebrew in (7) and (8). (6) a. Taro-wa kinoo ryokou-ni dekake-*ru/ta Taro-TOP yesterday a trip-to go *PRES/PAST "Taro *goes/went on a trip yesterday." (grammaticality contrast detected) b. Taro-wa sono hon-o doko-de kaimasita-*0/ka. Taro-TOP that book-ACC where buy-PAST 0/Q "Where did Taro buy that book?" (grammaticality contrast undetected) (7) a. Etmol ha-yeled katav. Yesterday the boy write-PAST b. *Etmol ha-yeled yiktov. Yesterday the boy write-FUT (ungrammaticality undetected)  *Etmol ha-yeled katvu. Yesterday the boy write-PAST.PL (ungrammaticality detected) (8) *Etmol ata telex. Yesterday you go-FUT-2MS (ungrammaticality undetected) Other sources can be adduced for selective impairment: De Bleser & Luzzatti (1994) found verb agreement spared in some Italian patients, etc. (see Friedmann & Grodzinsky 1997 for further references).

270

ANNA GAVARRÓ

However, Hagiwara expressly acknowledges the existence of patients whose impairment is general. Regarding Friedmann & Grodzinsky (1997, 2000), it is unclear whether the existence of general (i.e., non-selective) im­ pairment is precluded or not. In any case, their theory does not exclude such a possibility. For these reasons, and while we retain the case of selective im­ pairment in mind, general impairment remains part of the empirical domain to be characterised. This general impairment has been noted for production and grammaticality judgement (see Zurif & Grodzinsky 1983). More recently in the history of aphasiology, another area in which agrammatics display deficits has been pointed out: that relating to displacement of constituents within a structure (e.g., Thompson, Shapiro, Jacobs & Schneider 1994, for disruption of w/h-questions). Grodzinsky (1990) first asserts that the operation Move is altered in agrammatism in view of the disruption in the interpretation of sentences involving movement, notably passives as opposed to actives.3 The principle he formulates is known as the Trace Deletion Hypothesis (TDH). The Trace Deletion Hypothesis claims that traces of syntactic movement of phrasal constituents are deleted from grammatical representations of patients [of agramma­ tism], resulting in a selective syntactic impairment. (Grodzinsky & Finkel 1998:282)

Trace deletion affects comprehension as well as grammaticality judge­ ments (for which see Grodzinsky & Finkel 1998), and, if it affects grammatical representation as claimed, should have repercussions in sentence production too. Notice that trace deletion does not disrupt Xo movement, according to the empirical evidence. Lonzi & Luzzatti (1993) demonstrate that Xo movement remains intact in agrammatism; they examine word order alternations in Italian, in particular the relative position of adverbs and verbs. In Italian, nonfinite verbs may follow or precede adverbs, while finite verbs must always precede adverbs, as a result of verb raising to I. Three agrammatic patients were shown to have knowledge of these alternations by means of a constituent ordering task. Nonetheless, evidence of impaired head-raising is attested: van Zonneweld & Bastiaanse (1999) show cases of V2 being impaired in Dutch agrammatism.

3

Cf. Grodzinsky (1986, 1990, 1995) and Grodzinsky & Finkel (1998:282). What should be the exact formulation of the TDH is not relevant here; see, for discussion, Hickok & Avrutin (1995).

FAILURE TO AGREE IN AGRAMMATISM

271

Then, X movement cannot be assumed to be impaired, but impairment can result from independent sources. There are also indications that comprehen­ sion of head raising structures is spared in other Germanic languages; see Friederici & Frazier (1992). We will come back to the theoretical implications of Xo movement being generally preserved. Before we turn to the theoretical analysis of the phenomena summarised so far, it should be pointed out that the asymmetries that may arise between production and comprehension, and also grammaticality judgement, in agrammatic subjects are not central to this paper. We attempt to characterise the dis­ ruption that affects grammatical derivations, and the reasons why the disruption is not apparent in all cases remain a topic for future research. 3. Background to agreement in minimalism The account of agreement and movement proposed in the minimalist pro­ gram constitutes a departure from previous models of generative grammar and, even within minimalism (Chomsky 1995[1992], 1998, 1999), the formalisation of the phenomena has undergone a considerable shift. All minimalist proposals establish some connection or another between agreement and movement, which is relevant to our concerns. When compared to the previous principles-and-parameters models, Chomsky (1992) introduces the modification of determining that structural Case and agreement only occur in Spec-head configurations. The sentential structure is the one in which these relations hold: cp[Spec [  AGRSP[ Spec [ AGRS Tp[ Spec [T AGROP [Spec

[AGRO VP[SU

[V OB]]

In this structure, the subject SU is assumed to raise to the specifier of AGRS, and the object OB to the specifier of AGRo; in the two resulting Spec-head con­ figurations, both subject and object can acquire morphological case and/or come to agree with the inflected verb - since V raises successively to AGRo, T and AGRs. Thus subject and object enter into two kinds of relations with a ver­ bal predicate: agreement, which consists of feature sharing, and Case, which manifests itself in the NP alone. The AGR projection is the only one involved in agreement, while Case involves the raised T and V projections, depending on their lexical properties. It was suggested in Gavarró (1993) that this new theoretical approach granted us some generalisations based on the work of Grodzinsky (1990) but not foreseen in his original work, which had been carried out in former ver­ sions of principles-and-parameters. Grodzinsky (1990) dealt separately with the inflectionally deviant forms and the impairment related to movement, the

272

ANNA GAVARRÓ

second by a version of the TDH introduced above. It was pointed out in Gavarró (1993) that the TDH suffices to predict disruption of inflection if agreement and Case assignment are a product of movement, as is the case al­ ready in the first version of minimalism. Moreover, this stance allows us to generalise over other facts, such as the distribution of DPS with respect to finite and non-finite verbs. For example, the presence of a null DP as in (9) is expected if feature checking, being mediated by traces, is disrupted and thus the +/- finite feature cannot be properly controlled by the speaker. (9)

... mari venir husband come-INF

However, this proposal suffered from an empirical shortcoming (pointed out to me by Y. Grodzinsky): it predicted that all sorts of movement (or displacement) are impaired, contrary to fact. Thompson, Fix & Gitelman (1999) and Izvoski & Ullman (1999) argued for minimalist accounts of agrammatism not unlike that in Gavarró (1993): the first considered overproduction and substitution of inflection by an English speaking patient and proposed impairment in feature-checking (as formulated in Chomsky 1995); the second equated inflectional deficits to impairment of the operations of concatenation and/or movement (Chomsky 1995) with a cu­ mulative effect (with higher nodes being more affected) - the reasons for this cumulative effects left unclear. 3.1 Competing minimalist approaches Later versions of minimalism retain the inherent relation between move­ ment and agreement, albeit in a different way. Chomsky (1998) redefines the basic operations that intervene in the syntax. These are: Merge, Agree and Move. Agree is defined as establishing "a relation (agreement, Case-checking) between a linguistic expression and a feature F in some restricted search space" (Chomsky 1998:14; as summarised by Carstens 2000). (10)

Agree operates between a probe a and a goal ß iff a has uninterpretable -features ß has identical, interpretable  -features ß has an unchecked feature of structural Case a c-commands ß there is no potential alternative goal  such that a c-commands  and  c-commands ß f. the structural relation between (a, ß) was not created by Merge

a. b. c. d. e.

(α,ß)

FAILURE TO AGREE IN AGRAMMATISM

(11)

273

[α [ [ [ ß ] ] ] ] probe goal Agree

Move is an operation composite of Merge and Agree (plus an extra step). That Move should be a composite of simple operations brings in a theoretical problem with respect to economy: "good design conditions would lead us to expect that simpler operations are preferred over more complex ones, so that Merge or Agree (or their combination) preempt Move" (Chomsky 1998:14). To overcome this contradiction, a new metric of economy is introduced by Chomsky. This goes to show that the composite character of Move is not free of shortcomings. We have seen that, just as agreement is found to be deviant in agramma­ tism, word order patterns (partially determined by movement) are altered, al­ though in a more selective manner. Now, assuming Chomsky 1998 has the following consequence: if the application of Agree is problematic in agram­ matism, we would expect movement to be problematic too, insofar as Move requires the application of Agree. Contrary to this, movement seems to be par­ tially but significantly spared in agrammatism. With respect to the theory as stated in Chomsky 1998, and given the empirical evidence above, agramma­ tism could be taken as an argument for reconsideration of the basic operations Move and Agree. This could involve either the restoration of Move as a primi­ tive operation, or granting Move a status whereby it was not a composite of Agree. Notice as well that there are independent reasons to make Agree and Move entirely independent operations: that Move should be preeempted, due to economy, by Merge and Agree is problematic in itself (see Chomsky 1998:14). Carstens (2000) points out some shortcomings of Chomsky's (1998) account of agreement, and separates this from concord (this last one understood as agreement within DP, between sister constituents). Chomsky (1999:34, n.5) also signals that "there is presumably a similar but distinct agreement relation, concord, involving Merge only"; the details of it remain unspecified. Roeper & Eisenbeiss (2000) assume the distinction between concord and agreement to account for the early vs. late development of inflectional patterns in acquisition. In view of the fact that some authors have argued that there is preserved agreement within DPS in some patients' agrammatism (see De Bleser, Bayer & Luzzatti 1995/96 for German, and Friedmann & Grodzinsky 1997 for Hebrew), the proposal of Carstens may, therefore, be relevant for the study of agrammatism too. This is a matter for future research.

274

ANNA GAVARRO

The issue of concord aside, there is a possible alternative to a redefinition of the basic operations that encompasses the facts associated with agramma­ tism; it consists in analysing the inflectional deficits at the level of lexical in­ sertion. It can be argued that the features characterising a lexical item have been empoverished or misassigned in the lexicon in such a way that appro­ priate lexical insertion is impossible. So, for example, in (12): (12)

Uh, oh, I guess six month ...my mother pass away.

month can have been wrongly specified as [+ plural], or pass been specified as [+ past], or simply unspecified for tense, so that matching of features proceeds as normal. The locus of the inflectional deficit does not lie in the application of any syntactic operation, but in the lexical specification of the item entering the enumeration. However, such an analysis relies on the misassignment or unspecification of a particular set of features: -features and verbal features of tense, aspect and mood. There is no principled way of accounting for the limits of the set of features affected, i.e., -features and verbal inflectional features being affected becomes accidental. This represents a loss of explanatory adequacy with respect to the syntactic account of agrammatism initiated by Grodzinsky (1990). The puzzle that a minimalist account of the kind outlined above poses is then: how can we account for the inflectional deficit associated with agrammatism together with the deficit of XP-movement (but not Xo movement), if Move is stated in terms of Agree? That is, how can we predict general impairment in the application of an operation without automatically predicting impairment in the application of the operation that is the composite of the first? The basic operations Move and Agree remain unchanged in Chomsky (1999): a) a relation Agree, holding between a and ß, where a has interpretable features and ß has iminterpretable ones, which delete under Agree. (p. 3) Matching of probe-goal induces Agree. eliminating uninterpretable features. (p. 4) b) The combination of Agree/Pied-Pipe/'Merge is the composite operation Move, preempted where possible by the simpler operations Merge and Agree. (p. 7) c) Case-assignment is divorced from movement and reflects standard properties of the probes, indicating that it is a reflex of Agree holding of (probe, goal). (p. 13)

However, a neater distinction is drawn between Xo and XP movement: There are good reasons to suspect that a substantial core of head-raising processes, excluding incorporation in the sense of Baker (1988), may fall within the phonological component. [...] The interpretative burden is reduced if, say, verbs are interpreted the same way whether they remain in situ or raise to T or C. [...] Verbs

FAILURE TO AGREE IN AGRAMMATISM

275

are not interpreted differently in English vs. Romance, or Main Scandinavian vs. Icelandic, or embedded vs. root structures. More generally, semantic effects of headraising in the core inflectional system are slight or nonexistent, as contrasted with XP-movement, with effects that are substantial and systematic. That would follow insofar as head-raising is not part of narrow syntax. [...] Overt V-to-T raising, T-to-C raising, and N-to-D raising are phonological properties, conditioned by the phonetically affixal character of the inflectional categories. (Chomsky 1999:30-31)

This new approach undoubtedly raises many questions, especially with re­ spect to PF (e.g., what kinds of operations are allowed on the way to PF, if Xo movement is among them?). However, on the other hand, it offers advantages over previous versions of minimalism in allowing a natural formulation of part of the agrammatic deficit. The dissociation of Xo movement, spared in agrammatism according to the literature, from XP-movement resolves the puzzle presented above. In actual fact, the data of agrammatism outlined not only give support to this theoretical distinction, but render it necessary. In the present framework, if the operation Agree is applied in a deficient manner by agrammatic patients, as a consequence Move will be deficient, but not Xo movement, because this last operation belongs to the PF component and it is hot subsumed by Move. This prediction meets the empirical evidence. Evidence that would run contrary to the analysis here would involve an in­ flectional deficit without XP-movement deficits, or an intact inflection with XPmovement deficits. I have not been able to find any such case in the literature. 4.

Conclusions Inflectional and word order deviations constitute a natural class in virtue of the hypothesis put forward in this paper, which can be stated as follows:

Agrammatism results in part from unability of the application of the operation Agree. This results in impairment of (i) overt inflectional patterns related to agreement (and structural Case), (ii) grammatical phenomena resulting from the application of Move, since Move is a composite operation of Agree; these include XP-movement: wh-movement and raising (in passives, etc.). Note that, with respect to Case, only structural Case may be considered in this paper, as well as in Grodzinky's 1990 work.4 Other phenomena that are not covered by (15) are e.g., absence of determiners, and governed preposi­ tions (Grodzinsky 1990), which by hypothesis do not conform a natural class 4

Our predictions run contrary to what seem to be the facts of Lukatela, Crain & Shankweiler (1988) for Case in Serbo-Croatian, which seems to be spared in contexts of subcategorisation.

276

ANNA GAVARRO

with the phenomena considered here. There is another area in which agrammatic subjects fail in a systematic way, namely the resolution of pronominal reference (Grodzinsky et al. 1993), and the introduction of discourse referents in general (Avrutin & Manzoni 2000); as argued by Avrutin & Manzoni, this may result from an impairment in discourse devices, rather than a grammatical breakdown. However, an analysis in terms of discourse disruption is unlikely to extend to the facts of e.g., gender inflection treated here, for which a gram­ matical account is called for. In principle, (15) should grant disruption in the production of inflection and grammatical markers, but not necessarily in their comprehension (unless comprehension depends exclusively on inflectional markers in a particular construction); production and comprehension of sentences with displaced con­ stituents should be affected, assuming that phrasal movement has effects on interpretation. Regarding the TDH of Grodzinsky (1986, 1990, 1995), the minimalist for­ mulation of 1999 renders its reformulation necessary, since traces are dispensed with, and so on; this need not be more than a technical matter. However, the analysis presented here would seem to subsume the TDH (or at least partially subsume it, to an extent that I leave for future research). This follows from the fact that the TDH singles out grammatical representations in which relations are mediated by an XP-trace; these are equivalent to represen­ tations which result from the application of Move. Further, the application of Move fails if Agree fails, because the first is a composite of the second. In this paper I have tried to show how linguistic theory informs the study of agrammatism. More importantly, the linguistic evidence provided by agrammatism has a bearing on the evaluation of competing linguistic theories. In the case at hand, our analysis of some phenomena associated with agrammatism lends support to the last version of minimalism (Chomsky 1999) over previous ones, because this helps tease apart impaired agreement and XPmovement from spared Xo movement.

REFERENCES Avrutin, Sergey & Dario Manzoni. 2000. "Grammatical constraints on agrammatic speech: Evidence from Italian". Paper presented at the conference Linguistic Theory, Speech and Language Pathology, University of Padova, Padova, August 2000. Bastiaanse, Roelien & Ron van Zonneveld. 1998. "On the relation between verb in­ flection and verb position in Dutch agrammatic aphasics". Brain and Language 64.165-181.

FAILURE TO AGREE IN AGRAMMATISM

277

Bates, Elisabeth & Beverly Wulfeck. 1989. "Comparative aphasiology: A cross-lin­ guistic approach to language breakdown". Aphasiology 3.111-142. Carstens, Vicki. 2000. "Concord in minimalist theory". Linguistic Inquiry 31:2.319-355. Chomsky, Noam. 1992. "A Minimalist program for linguistic theory". MIT Occa­ sional Papers in Linguistics, 1. MIT. (Also published in The Minimalist Pro­ gram, 167-217 Cambridge, Mass.: MIT Press, 1995.) Chomsky, Noam. 1998 "Minimalist inquiries: The framework". Ms., MIT. Chomsky, Noam. 1999. "Deriving by phase". Ms., MIT. De Bleser, Ria, Josef Bayer & Claudio Luzzatti. 1995/96. "Linguistic theory and morphosyntactic impairments in German and Italian aphasics". Journal of Neurolinguistics 9:3.175-185. De Bleser, Ria & Claudio Luzzatti. 1994. "Morphological processing in Italian agrammatic speakers: Syntactic implementation of inflectional morphology". Brain and Language 46.21-40. Friedend, Angela & Lyn Frazier. 1992. "Thematic analysis in agrammatic compre­ hension: Syntactic structures and tasks demands". Brain and Language 42.1-29. Friedmann, Na'ama & Yosef Grodzinsky. 1997. "Tense and agreement in agram­ matic production: Pruning the syntactic tree". Brain and Language 56.397-425. Friedmann, Na'ama & Yosef Grodzinsky. 2000. "Split inflection in neurolinguistics". The Acquisition of Syntax: Studies in comparative developmental studies, ed. by Marc-Ariel Friedemann & Luigi Rizzi, 84-104. London: Longman. Gavarró, Anna. 1993. "A Note on Agrammatism and The Minimalist Program". Ms., Universitat Autónoma de Barcelona. Goodglass, Harold. 1976. "Agrammatism". Studies in Neurolinguistics, vol. I, ed. by Harry A. Whitaker, 237-260. New York: Academic Press. Grodzinsky, Yosef. 1990. Theoretical Perspectives on Language Deficits. Cam­ bridge, Mass.: MIT Press. Grodzinsky, Yosef & Lisa Finkel. 1998. "The neurology of empty categories: Apha­ sics' failure to detect ungrammaticality". Journal of Cognitive Neuroscience 10:2.281-292. Grodzinsky, Yosef, Kenneth Wexler, Yu-Chin Chien, Susan Marakovitz & Julie Solomon. 1993. "The breakdown of binding relations". Brain and Language 45.396-422. Hagiwara, Hiroko. 1995. "The breakdown of functional categories and the economy of derivation". Brain and Language 50.92-116. Izvorski, R. & Michael Ullman. 1999. "Verb inflection and the hierarchy of func­ tional categories in agrammatic anterior aphasia". Brain and Language 69.288-291. Kean, Mary-Louise. 1977. "The linguistic interpretation of aphasie syndromes". Cognition 5.9-46. Lapointe, Steven. 1983. "Some issues in the linguistic description of agrammatism". Cognition 14.1-41.

278

ANNA GAVARRO

Lonzi, Linda & Claudio Luzzatti. 1993. "Relevance of adverb distribution for the analysis of sentence representation in agrammatic patients". Brain and Language 45.306-317. Lukatela, K., Stephen Crain & D. Shankweiler. 1988. "Sensitivity to inflectional morphology in agrammatism: Investigation of a highly inflected language". Brain and Language 33.1-15. Pitres, Albert. 1898. L'aphasie amnésique et ses variétés cliniques. Paris: Alean. Roeper, Thomas & Sonja Eisenbeiss. 2000. "Acquisition insights into UG: Making an agreement/concord distinction". Ms., University of Massachusetts, Amherst. Thompson, Cynthia, Stephen Fix & Darren Gitelman. 1999. "Selective impairment of morphosyntactic production in a neurological patient: evidence for impaired feature processing". Brain and Language 69.285-288. Thompson, Cynthia, Lewis Shapiro, Beverly Jacobs & Sandra Schneider. 1994. "Training wh-questions in agrammatic aphasia: An analysis of lexical and syntactic properties". Brain and Language 47.405-408. van Zonneweld, Ron & Roelien Bastiaanse. 1999. "Finite verbs in agrammatism". Brain and Language 69.253-256. Zurif, Edgar & Yosef Grodzinsky. 1983. "Sensitivity to grammatical structure in agrammatism: A reply to Linebarger et al.". Cognition 15.207-213.

THE VERB AND SENTENCE TEST: ASSESSING VERB AND SENTENCE COMPREHENSION AND PRODUCTION IN APHASIA JUDITH RISPENS, ROELIEN BASTIAANSE & SUSAN EDWARDS University of Groningen University of Reading 1. Introduction Aphasia is a language disorder caused by neurological damage. The symptoms of the disorder vary according to the localisation and the size of the lesion. Several aphasia tests have been developed with which the type and se­ verity of the aphasia can be established.1 These kinds of diagnostic assess­ ments provide valuable information. However, they do not give much insight into the underlying linguistic disorder. An example of a test battery that inves­ tigates certain aspects of linguistic processing in aphasic patients is the Psycholinguistic Assessment of Linguistic Processing in Aphasia: PALPA (Kay, Lesser & Coltheart 1992). The numerous tasks of this battery give means of examining the linguistic abilities of a patient at the word and sound level; information which plays a very important role in the planning of aphasia treatment. From clinical experience and theoretical investigations, it has become clear that disorders at the level of verbs and sentences are frequent in aphasic patients (e.g., Edwards 2000; Grodzinsky 2000 and Jonkers 2000). Verbs play an important role in sentence comprehension and production; a disorder in processing verbs therefore has a great effect on linguistic processing and, very importantly, on communicative ability. There are currently few standardised assessment materials which systematically investigate disorders at the level of verbs and sentences. It seems that a gap exists between neurolinguistic findings and clinical application of this knowledge. Therefore, we developed a test battery for verb and sentence processing which 1) uses linguistic insights and is theoretically motivated, 2) is suitable for different types of aphasie pa­ tients and 3) is clinically relevant. This means that only tasks have been in­ cluded in the test that reveal impairments which have been described ade­ quately in the literature and for which therapy programs or methods exist. 1 E.g., the Boston Diagnostic Aphasia Examination (Goodglass & Kaplan 1983) and the Aachen Aphasie Test (Huber, Poeck, Weniger & Willmes 1983).

280

RISPENS, BASTIAANSE & EDWARDS

These three principles have led to the inclusion of ten sub-tests which assess the processes involved in comprehending and producing verbs and sentences: verb retrieval, processing information regarding the grammatical and thematic roles, 'mapping' of thematic roles onto grammatical roles and, for verb and sentence production, retrieving morpho-syntactic elements such as subjectverb agreement inflection. The VAST was originally developed in the Netherlands (Bastiaanse, Maas & Rispens 2000) and is now being adapted and translated into English, French and German. The Dutch version has been standardised and has proven to be valid and reliable. Below, some neurolinguistic investigations into verb and sentence processing in aphasia will be discussed. Then, the test contents will be de­ scribed, after which two case studies will be discussed to demonstrate the clinical use of the VAST. 2. Neurolinguistic framework The sub-tests of the VAST investigate linguistic aspects involved in com­ prehension and production of verbs and sentences that are known to be problematic for aphasic persons. We have chosen to select only tasks for the test battery that give insight into deficits described in the literature and for which treatment programmes/methods are available. For example, even though it is known from clinical practice that pronouns are often difficult for aphasic patients, a pronoun specific deficit has not been explained theoretically. Therefore, we have not developed a task assessing pronouns. Below, a range of problems that aphasie people can experience with verbs and sentences will be discussed. 2.1 Verb deficits Verbs play an important role in language: they do not only carry lexical meaning, but they also determine the linguistic environment: verbs specify the number of arguments (argument structure) and the kind of phrasal categories that may follow the verb (subcategorization). For example, the verb repair specifies that two participants are involved (one who repairs and something that is being repaired) and that the constituents following the verb need to be an NP and - optionally - a PP (see la-1b): (1) a. John repairs the car in the garage b. *John repairs in the garage

THE VERB AND SENTENCE TEST

281

Relatively little is known about the range and the nature of verb deficits in aphasia. Some researchers found that Broca's aphasics were better in re­ trieving nouns than verbs, in contrast to Wernicke's patients who showed the reverse (cf. Miceli et al. 1984; Zingeser & Berndt 1988; Berndt et al. 1997a,b). Other studies report that aphasic patients perform similarly on naming objects and actions (Basso, Razzano, Faglioni & Zanobio 1990) or that they are better in retrieving nouns compared to verbs (Williams & Canter 1987; Kohn, Lorch & Pearson 1989; Jonkers & Bastiaanse 1996). The last two research groups have pointed out that the results of studies on verb processing may be different due to the fact that a number of linguistic factors need to be controlled for when selecting test items. Jonkers & Bastiaanse (1996) show, for example, that transitivity and name relatedness with a noun influence naming actions much more than word frequency does, which is generally the only variable that researchers have taken into account when designing experimental materials. In this test, all these factors have been controlled for. The ability to name actions is not necessarily related to verb retrieval within sentential context, which is a better approximation of verb use in daily life (cf. Jonkers 2000). Different aspects come into play when verbs need to be used within a sentence; inflecting them for tense and agreement is one of them. Bastiaanse & van Zonneveld (1998) have investigated verb inflection in Dutch Broca's aphasies and found that the ability to retrieve verbs in a sentence depends on the position in the sentence. Dutch is an sov language (Koster 1975), meaning that the finite verb is in the last position (the canonical position) in an embedded sentence (see 2a) and that it moves to the second position in a matrix clause (this is called Verb-Second, see 2b). (2) a. Ik zie dat de man een boek leest I see that the man a book reads "I see that the man reads a book" b. De man leest een boek The man readsj a book t¡

Infinitives remain in the canonical position (see 2c):  De man gaat een boek lezen The man goes a book read "The man is going to read a book"

282

RISPENS, BASTIAANSE & EDWARDS

Dutch agrammatic Broca's aphasics were asked to complete a sentence (matching a picture) in which a finite verb was missing. The results show that producing inflected verbs in second position (for example 2b) was signifi­ cantly more difficult than retrieving finite verbs in the canonical position (2a). The difference between finite verb production in matrix and embedded sen­ tences has not been found in English (Bastiaanse & Thompson 2000), although the number of errors was comparable between the Dutch and English Broca's aphasics, demonstrating that verb inflection is problematic. Fluent aphasie patients may also experience problems with verb inflection (see Edwards in this volume for a discussion). Furthermore, spontaneous speech studies show that almost all fluent aphasies have problems using verbs; they may use fewer verbs than healthy speakers and/or the diversity of the produced verbs is lower (Bastiaanse, Edwards & Kiss 1996; Edwards & Bastiaanse 1998). The tasks of this test battery assess aspects of verb processing as described above; verb retrieval at the word level, but also at the sentence level. The pro­ duction tasks differentiate between eliciting finite and nonfinite verb forms. All test items have been matched on transitivity, thematic structure, name relatedness with a noun and word frequency as these factors can influence the ability to retrieve/comprehend verbs. 2.2 Sentence comprehension deficits Failure to understand a sentence can take different forms; it may be that aphasie patients do not understand the meaning of (some of) the words in a sentence. Other patients may have difficulty with interpreting word order. This is noticeable in semantically reversible sentences, as the plausibility of this kind of sentence is not affected by changing the order of the subject and the direct object. This is in contrast with irreversible sentences: if a patient has difficulty with word order information s/he can deduct the word order of irre­ versible sentences on the basis of general world knowledge. For example, the thematic roles of the sentence the girl reads a book can be interpreted based on the knowledge that books cannot read girls. However, a sentence such as the horse kicks the cow also has a plausible interpretation when one changes the order of the thematic roles: the cow kicks the horse. In the latter kind of sentence, world knowledge cannot be used to infer 'who is doing what to whom'; only processing the right word order leads to correct understanding of the sentence. If patients suffer from a severe grammatical comprehension deficit, an ac­ tive reversible sentence (such as the horse kicks the cow) may be misunder­ stood. In case of a less severe disorder, grammatically complex sentences

THE VERB AND SENTENCE TEST

283

(such as passives and object-cleft sentences) in which the word order differs from the more frequent one (agent-verb-theme) are prone to misinterpretation. There is an ongoing debate in the literature about the theoretical analysis of this disorder observed in agrammatic Broca's aphasics. Some researchers claim that the comprehension deficit of semantically reversible sentences, in which the NPs have been moved out of their original positions (such as in pas­ sives and in object-cleft sentences), results from deleted traces (cf. Grodzinsky 1995, 2000). Other researchers suggest that the deficit in interpreting word order stems from a problem in 'mapping' the thematic roles onto the grammatical roles (cf. Marshall 1995; Schwartz, Saffran et al. 1994). Even though comprehension of grammatically complex sentences has been studied extensively in Broca's aphasia, fluent patients may also have difficulty interpreting this type of structure. Zurif (1995) has shown that Wernicke's aphasics have problems with understanding passive sentences on a pointing-to picture task. The performance pattern is, however, different from that of Broca's aphasies. Whereas the latter pointed most frequently to the distracter picture that displayed the same action but with reversed thematic roles, the former group of patients tended to choose the pictures that depicted a different action. This suggests that the comprehension problems of the fluent patients originate from a lexical-semantic deficit rather than from a grammatical deficit. With these aphasie performance patterns in mind, we have developed a sentence comprehension task (a pointing-to-picture design), testing semanti­ cally reversible sentences with different word orders (for English: actives, passives, object and subject-cleft sentences) including lexical-semantic distracters and thematic role reversal distracter pictures. Both fluent and non-fluent patients may also have trouble with judging the grammaticality of certain syntactic structures (cf. Linebarger, Schwartz & Saffran 1983; Grodzinsky & Finkel 1998; Balogh & Grodzinsky 2000), espe­ cially with respect to judging word order. Detailed investigations have shown that some patients are able to judge a sentence correctly on its grammaticality, even though they do not understand the sentence. The dissociation between sentence comprehension and the ability of judging a sentence on its grammati­ cality has been taken into account by incorporating a grammaticality judge­ ment task in the test battery. 2.3 Sentence production deficits Producing a sentence not only entails retrieving the lexical items, but also linking the thematic roles to the grammatical roles (for instance, in the sen­ tence the girl eats a biscuit, the thematic role of agent needs to be mapped

284

RISPENS, BASTIAANSE & EDWARDS

onto the grammatical role subject (the girl) and the thematic role of theme has to be mapped onto the grammatical role direct object (the biscuit). A final step in the process of sentence production is placing the words in the right order and inserting grammatical morphemes (for example to express subject-verb agreement and tense). Difficulties can arise in either stage involved in sentence production; the VAST therefore, contains several tasks to assess these steps. It has been demonstrated that grammatical deficits lead to disturbances within the production process. The majority of studies focuses on Broca's aphasic patients, who tend to omit thematic roles (Schwartz et al. 1994), or use verbs that carry few grammatical roles (Byng 1988; Thompson, Lange et al. 1997; Bastiaanse & Jonkers 1998). These patients furthermore seem to opt for 'simple' syntactic structures, such as subject-verb clauses. Experimental studies have demonstrated that Broca's aphasics have difficulties with producing more complex grammatical structures, such as relative clauses, passives and questions. The nature of the deficit is controversial; several strong theoretical explanations have been formulated (see for example Hagiwara 1995; Schwartz, Fink & Saffran 1995; Friedmann & Grodzinsky 1997; Bastiaanse & van Zonneveld 1998). Fluent aphasies may also have problems with constructing sentences. Their severe word retrieval deficit seems to play a role in this. According to some researchers, paragrammatical utterances (typical for the speech of fluent patients) are the result of word finding problems (Butterworth & Howard 1987; Butterworth, Panzeri, Semenza & Ferreri 1990; Bird & Franklin 1996). Others suggest that there are also grammatical problems not related to the lexical deficit, impacting on the ability to produce sentences (cf. Edwards & Bastiaanse 1998). 3. Contents of the VAST The test battery can roughly be divided into two parts; one testing verb/sentence comprehension and one testing verb/sentence production. In to­ tal, it contains ten sub-tests which will be described below. 3.1 Verb comprehension Background: Comprehending a sentence requires grasping the meaning of the lexical items and interpreting the way they are structured grammatically. This task investigates the ability to comprehend verbs; part of the first step in the process. There are other tests available to assess comprehension of nouns (e.g., PALPA), therefore no such test has been incorporated in the VAST.

THE VERB AND SENTENCE TEST

285

Outline: The patient is presented with four pictures and is asked to look at them all. Then the examiner reads aloud a verb and the patient is asked to point to the picture matching the spoken word. The three distracter pictures are representing 1) an object that has a close semantic relation to the target verb (e.g., target: biting, closely related noun distracter: teeth) 2) an action that is closely related to the target action (e.g., target: biting, closely related action distracter: scratching) and 3) an object that is related to the action distracter (e.g., target: biting, noun distracter related to action distracter: nails). The forty items are controlled for transitivity, word frequency and name relatedness with a noun (e.g., cycling-cycle). An example of a test item is presented in Figure 1.

Fig. 1 : The verb comprehension task. Rowing

3.2 Grammaticality judgement Background: Misinterpretation of reversible sentences may originate from a problem with parsing the grammatical structure, or from a difficulty with mapping the thematic roles onto the arguments. Studies have shown that aphasic patients who were disturbed in understanding reversible sentences were able to judge a sentence on syntactic violations (Linebarger, Schwartz & Saffran 1983), indicating that the comprehension deficit originated from a difficulty with thematic roles, rather than from a grammatical problem. This task investigates whether a patient is able to parse a sentence, which is important to know for therapy, especially when the comprehension task shows s/he has difficulties understanding reversible sentences. Outline: The examiner reads aloud the sentence and asks the patient to say whether this sentence is 'good' or 'bad'. The test starts with a detailed in­ struction and with four examples so that the examiner can be confident

286

RISPENS, BASTIAANSE & EDWARDS

that the patient understands the task. The test contains forty irreversible sentences of which twenty are grammatical and twenty are ungrammatical. Each group contains four different types of sentences: - Actives (word order: agent-verb-theme): the woman drills the hole - Passives (word order: theme-verb-agent): the hole is drilled by the woman - Object-clefts (word order: theme-agent-verb): it is the hole that the woman drills - Subject-clefts (word order: agent-verb-theme): it is the woman who drills the hole In this way, the influence of the order of thematic roles on the patient's ability to judge a sentence can be determined. In the ungrammatical sen­ tences, the order of agent and theme has been changed (e.g., the hole drills the woman/the woman is drilled by the hole etc.). Sentences only appear once, either grammatically or ungrammatically. Thus, if there is a sentence the pipe smokes the man no item the man smokes the pipe exists. 3.3 Sentence comprehension Background: Many aphasic patients have problems understanding sentences. In this task, sentences containing moved constituents (theme-verb-agent; in passives (e.g., the woman is saved by the man) and theme-agent-verb in object-clefts (e.g., it is the woman who the man saves)) have been in­ cluded, together with sentences in which the order of the thematic roles is canonical (agent-verb-theme; in active (e.g., the man saves the woman) and subject-cleft sentences (e.g., it is the man who saves the woman)), so that it can be investigated whether the order of the thematic roles influences the ability to comprehend a sentence. Outline: The patient is shown four pictures and is asked to look at them all. The examiner reads aloud a sentence and the patient is asked to point to the picture matching the sentence. There are three types of distracters: 1) a picture in which the action and objects are the same but the thematic roles are reversed (e.g., target: the cow kicks the horse, reversed roles distracter: the horse kicks the cow), 2) a picture in which a different action is depicted (e.g., target: the cow kicks the horse, lexical distracter: the cow bites the horse) and 3) a picture in which both the thematic roles are reversed and in which the action is different (e.g., target: the cow kicks the horse, reversed roles/lexical distracter: the horse bites the cow). This format allows the examiner to determine what kind of errors the patient

THE VERB AND SENTENCE TEST

287

makes (lexical errors/ thematic errors or a combination of both). An example of a test item is presented in Figure 2.

Fig. 2: The sentence comprehension task. The woman is saved by the man

Action naming as single words ckground: In order to produce a sentence, one has to retrieve the different content words from the lexicon and one has to form a grammatical structure. This task investigates the ability to retrieve verbs. The items in this sub-test have been controlled for transitivity, name relatedness with a noun and word frequency. tline: The patient is presented with a picture and is asked to describe in one word what is happening in that picture. The test contains forty items that are all agentive verbs, meaning that they select the thematic role of agent in case of an intransitive verb and agent and theme in case of a transitive verb. An example of a test item is presented in Figure 3.

Fig. 3: Action naming as single words. Target: skating

288

RISPENS, BASTIAANSE & EDWARDS

3.5 Filling in verbs in sentences Background: For some aphasic patients, a sentence framework facilitates verb retrieval, whereas for others sentential context has a detrimental effect. Some patients will have more trouble using inflected verbs than infinitives. This task gives the opportunity to investigate these factors and can be viewed as a step in between retrieval of verbs as single words and constructing a sentence without help. Outline: This task comprises two sets of sentences: sentences 1) in which the verb is finite (e.g., the boy eats an apple) and 2) in which the verb is an in­ finitive (e.g., grandmother likes to knit a jumper). The patient is presented with a picture that has a sentence printed under­ neath of which the main verb is missing. The experimenter reads aloud that sentence and 'hums' at the place of the verb. The patient is asked to say the missing word. The action words are controlled for frequency and transitivity. Two test items are shown in Figure 4.

The girl likes to

The farmer .... the cow Fig. 4: Filling in infinitives and finite verbs. Left picture: filling in infinitives Right picture: filling in finite verbs Target: cycle Target: milks

3.6 Sentence construction Background: An oral picture description task is one of the best ways to esti­ mate the ability of a patient to make sentences in daily life. The responses on such a task can be analysed with respect to the ability to retrieve lexical items and to construct grammatical sentences because one has a good idea of the target sentence, which is not always the case in spontaneous speech. The twenty test items are controlled for three factors: frequency (all verbs are high frequent), transitivity and reversibility of thematic roles. Outline: A picture is presented to the patient who is asked to describe that picture in one sentence.

289

THE VERB AND SENTENCE TEST

3.7 Anagram task with pictures Background: A sentence construction task, such as the one described above, is a good tool to investigate sentence production in aphasics. It also has some serious limitations. First of all, as mentioned before, it can be difficult to score the performance of a patient systematically. Furthermore, it is hard to elicit other grammatical structures than simple actives with a picture de­ scription task. Also, a patient's ability to express him/herself can be com­ promised, for instance by dysarthria or an oral dyspraxia. Therefore, two anagram tasks have been included in the VAST. In the first one, twenty items are divided between ten semantically reversible and ten irreversible sentences of which ten are passive and ten are active sentences. Outline: The patient is presented with a picture and three cards with con­ stituents printed on them. The patient is asked to use these cards to form a sentence that matches the picture (e.g., an active sentence: the boy/ kisses/ the girl, a passive: the girl / is kissed by / the boy). An example of a test item is presented in Figure 5.

grandmother

is drawn by

grandfather

Fig. 5: Sentence anagrams with pictures

3.8 Sentence anagram without pictures Background: This task has been included for the same reasons as described above on special request of speech and language therapists, as it was found that some patients perform differently when the sentences do not have to match pictures. In this task twenty semantically irreversible sentences have been included of which ten are active and ten are passive. Outline: The patient is presented with three cards with constituents printed on them and is asked to make a sentence with them. Examples of sentences are: the bike/ is fixed by/the man and the child/ throws/ the ball.

290

RISPENS, BASTIAANSE & EDWARDS

3.9 Wh-anagrams Background: Aphasic patients may have difficulty forming questions. Thompson, Shapiro, Tait, Jacobs & Schneider (1996) have studied this in agrammatic aphasia and describe a treatment program based on linguistic principles. They describe two types of questions: constructions in which the wh-word represents an NP (who and what questions) and constructions in which the wh-question word represents a PP (where and when). The former structure is derived by argument movement, whereas the latter is an adjunct transformation. Thompson et al. (1996) have found that this linguistic difference is important with regards to therapy and therefore the two types of questions are distinguished in this task. It is difficult to elicit questions orally in a structured way, so anagrams are used. Twenty items are included in this task; five items for each wh-question word (who, what, where and when). Outline: The examiner shows the patient a picture and presents the patient with five cards containing the sentence constituents (e.g., what/ is/ the farmer/ pushing /in the stable/). The patient is asked to form a question that matches the picture. Figure 6 shows an example of this task. the princess what kissing at the pond

Fig. 6: The wh-anagrams task

4. Two case studies During the standardisation procedure, aphasic patients (as well as healthy control subjects and patients with right-hemisphere lesions) have been tested with the VAST. In this section, the performances of two of these aphasie par­ ticipants will be discussed. The first is an English patient. His results on the comprehension tasks and the wh-question anagram test will be discussed in the light of how these might be used in drawing up a suitable treatment plan. The second patient is Dutch; his performances on the production tests will be described and suggestions for therapy will be discussed.

THE VERB AND SENTENCE TEST

291

4.1 The English patient Mr. L is an 84 year-old retired butcher who suffered from a CVA in his left hemisphere 5 months prior to testing. Initially he was, according to his speech and language therapist, a non-fluent aphasic but when tested, he was quite fluent and had mild to moderate word finding difficulties, sometimes resulting in phonemic paraphasias. He does not have any physical problems. He needs reading glasses and has some hearing loss, probably age-related. An example of his spontaneous speech is given below: Can you tell me what happened to you? Yes .... that ... tenth of July ... tenth of January ... second of January I had a full stroke and I couldn't speak and I couldn't speak at all. I member sitting in the kitchen in the fire place and the ... and he just said that ehm and I didn't know what to do and I was just standing there and all over sudden I fell down and I s I was thinking of things and I could thinking but I couldn't do anything and then my son in law came in and he came up and came up and set me and myi wife found me then they saw the ... the ... anyhow the doctor came but... and then I was took into Reading.

A lexical analysis of 300 words reveals that the number of noun tokens falls just within the range of the healthy speakers (22, mean healthy speakers 46.1, range 22-69) and that the diversity of noun use is normal (20, mean healthy speakers 35.3, range 18-64). The Type Token Ratio for nouns is high: 0.9, meaning that he uses a wide variety of nouns even though he does not use that many compared with the healthy speakers.2 The number of tokens and types of verbs is normal (tokens: 41, mean healthy speakers 40.2, range 28-51 and types: 21, mean controls 25.6, 17-32). 4.1.1 Test results Verb comprehension: 36/40 correct. Three times he chose the related verb and once the closely related noun. Sentence comprehension: 25/40 correct. The majority of the errors were due to pointing to the distracter with the reversed roles (12/15 errors), but the errors were almost equally divided between the actives, object and subject-cleft sentences. Remarkably, the passive sentences were done best: 8/10 correct. Two of the three items that triggered the choice of the lexical distracter were of the agent-verb-theme type. 2

The Type Token Ratio measures the diversity of use. It is calculated as follows: the total amount of words of the word class one wants to measure is counted. This is the number of tokens. Subsequently, the different words of that word class are counted: the types. The number of types is divided by the number of tokens: the resulting number will be between 0-1. The higher the type-token ratio; the higher the variety of the used word class.

292

RISPENS, BASTIAANSE & EDWARDS

Grammaticality judgement: Mr L showed no problems with this task: 39/40 correct. Wh-anagram task: 12/20 correct. The majority of the errors were due to re­ versing the thematic roles (5/8 errors) and incorrect placing of the whword, or a combination of the two (e.g., 1. Target: who is drawing Santa on the blackboard? Response: who is Santa drawing on the blackboard, 2. Target: where is the tourist filming the girl? Response: the tourist is filming where the girl? and 3. Target: what is the girl hugging in the sta­ ble? Response: is what hugging the girl in the stable?). The majority of errors occurred in questions with the question word where (1/5 correct, versus who and what 4/5 correct and when 3/5 correct). The test results are presented in Figure 7. 80|

70J patient EL controls

verb comp

sentence comp

gramm judgment

wh-anagr

Fig. 7: Percentages correct of the English patient

4.1.2 Conclusions and directions for therapy. This patient has mild problems in comprehending actions as single words, but he is more impaired in under­ standing reversible sentences. The pattern of errors indicates that he has a problem with understanding 'who is doing the action to whom', even in sen­ tences in which the thematic roles have not been moved out of their original positions. The results on the grammaticality judgement task indicate that the patient is able to parse structures in which NPs have moved out of their canonical positions. The discrepancy between his performance on the comprehension task and the grammaticality judgement task points to a problem with mapping thematic roles onto grammatical roles, but suggests that he has an intact sensitivity with respect to word order since the irreversible sentences can only be judged correctly if the word order is parsed correctly. For instance, the sentence * the woman is baked by the cake will be judged as being incorrect because it is not allowed the order: animate NP verb - by inanimate NP. A mapping deficit will not interfere with the ability to

THE VERB AND SENTENCE TEST

293

judge sentences on their grammaticality. Understanding irreversible sentences is different from reversible sentences as extralinguistic knowledge (cakes cannot bake women and can therefore not be the agent of the sentence) will guide the patient in assigning the thematic roles correctly to the NP's in an irreversible sentence. Thus, the difficulties that this patient experiences with comprehending reversible sentences do not seem to stem from a parsing deficit as he is able to judge irreversible sentences on their grammaticality, but from a mapping deficit that only surfaces with reversible sentences where extralinguistic information cannot be used to interpret the sentence. The difficulty with understanding reversible sentences, even actives, is quite severe and a therapy programme may improve Mr L's insight into thematic role assignment, using for instance a design described in the literature (e.g., Marshall 1995). The relation between sentence structure and sentence meaning is the central point of interest in mapping therapy. Tasks that are often used are dividing the sentence into the constituents (with a special focus on the verb) and deciding on the roles of the constituents (often by asking 'who does what'?). During the course of treatment, different structures are trained, including sentences with moved constituents. Therapy studies, reporting on the efficacy, reveal that treatment can be very successful, but individual differences must be taken into account in the planning of the therapy. The exact nature and the severity of the mapping deficit, cognitive problems and other issues influence outcome (Marshall 1995). Forming questions is problematic for this patient; there the difficulty often also comes from thematic role assignment. A block of therapy can be used to facilitate the production of questions, for instance based on the design of Thompson et al. (1996). The order and function of thematic roles is very im­ portant with respect to the production of this kind of structure. In this therapy program, patients are trained to produce wh-questions {who, what, where and when questions). The treatment consists of training several steps leading to question production. Patients are presented a sentence and are asked to recognise the verb, its arguments and the thematic roles. Then, they are asked to move the appropriate argument or adjunct and replace it by the correct whword. Finally, they are asked to produce the question. Thompson et al. (1996) discuss that the patients who participated in the study improved in question production and that generalisation occurred from trained to untrained questions of the same type. They furthermore found that the number of arguments and adjuncts in the spontaneous speech of the patients had significantly increased after therapy.

294

RISPENS, BASTIAANSE & EDWARDS

4.2 The Dutch patient Mr.  is a 47 year old left hemisphere CVA patient, who is married with two children. Before the CVA, he worked as a bank director. He does not suffer from a hemiparesis/plegia, cognitive problems, nor from hearing and/or visual problems. His spontaneous speech is fluent, but he experiences word-finding diffi­ culties which interferes with his fluency. Below, a sample of his spontaneous speech is shown: Wat zijn uw problemen met spreken? Alleen zinnen .. alleen zinnen ... ja begin dat dat dat was niets eigenlijk .. ja nie eens nie eens nie eens mijn naam nog nie eens geen aantallen niets en toen dacht ik wel ja toen dacht ik ... een dag dacht ik, toen dacht ik twee dagen nou wel en nie eens kranten. Toen dacht ik prima je kunt met benen en armen dat dat is prima maar alleen die zinnen en een krant ... dat dat gaat niet. Can you tell me about your problems with speaking? Only sentences ... only sen­ tences .. yes beginning that that that was nothing really .. yes not even not even not even my name not even no numbers nothing and then I thought well yes then I thought.. a day I thought, then I thought two days now well and not even papers. Then I thought fine you can with legs and arms that that is fine but only those sen­ tences and a paper ... that that does not work.

An analysis of 300 words shows the following. The mean length of utterance is lower than that of healthy control speakers (5.0, mean healthy speakers 8.22, range 5.7-10.7) and he does not use any embedded clauses (mean healthy controls 8.00, range 3-15). The number of modal and copular verbs is normal (11, mean healthy speakers 12.75, range 4-18). The typetoken ratio of nouns is normal (0.61, mean healthy controls 0.75, range 0.540.88). The type-token ratio of verbs is extremely low (0.23, mean healthy speakers 0.74, range 0.63-0.86). 4.2.1 Test results Action naming: 31/40 correct. Word frequency, transitivity and name relatedness with a noun do not play a role. The majority of the errors are semantic paraphasias (for example roeien > kanoën, "rowing > canoeing") or he substitutes a verb for a noun (for example schilderen > schilder, "paint > painter"). He also sometimes changes a noun (related to the picture) into a verb (e.g., persen > sappen, "squeezing > juicing"). Filling in infinitives and finite verbs in a sentence: Filling in infinitives does not pose too much problems: 8/10 correctly. When he has to produce a fi­ nite verb, the score reduces to only 50% correctly (5/10 correct).

295

THE VERB AND SENTENCE TEST

Inflecting the verb is not problematic, but phonemic and a few semantic paraphasias occur. Sentence anagrams without pictures: The first four passive sentences are con­ structed in the following manner: agent-verb-theme. After four sentences he detects that this is wrong and does not make any more mistakes after this. Score: 16/20 correct. Sentence anagrams with pictures. Mr.  does not have any problems with this task: score 20/20 correct. Sentence construction. Mr.  tries to make sentences but often he does not succeed due to phonemic paraphasias in combination with great word finding difficulties. He experiences difficulties with verb production, which is surprising considering his score on the action naming task which shows that verb retrieval itself is relatively spared. The verb is sometimes omitted (target: the boy hits the girl > the boy and the girl), sometimes the infinitival form is used (target: de jongen duwt het meisje "the boy pushes the girl > de man ging omdraaien "the man went to turn"). Inflection of the verb is sometimes problematic (target: de baby kruipt "the baby crawls"> de baby kruip, kruip, kruipt "the baby crawl, crawl, crawls") and sometimes a noun is changed into a verb (target: de clown lacht "the clown smiles" > de man clownt "the man clowns"). Score: 7/20 correct. An overview of his results is displayed in Figure 8.

action naming

infinitives finite verbs

anagram -pict

anagram +pict

sentence const

Figure 8: Percentages correct of the Dutch patient

296

RISPENS, BASTIAANSE & EDWARDS

4.3 Conclusions and directions for therapy Figure 8 shows that retrieving verbs is only mildly impaired, but when the patient has to use them in a sentence, this becomes very difficult. The results on the task 'filling in verbs in sentences' show that producing a finite verb is particularly difficult in contrast with retrieving an infinitive. Therefore, it is assumed that verb movement, which underlies the production of finite verbs (as he has to in the sentence construction task), is impaired. The problems with fmiteness have a spin-off effect on lexical properties of the verb: semantic and phonemic paraphasias occur when he has to produce a finite verb. This can also be observed in his spontaneous speech; the diversity of the verbs is extremely low (in the sample printed above, only two different lexical verbs are used: think and work). In therapy, one can focus on expanding a linguistic strength of this patient, namely the ability to use verbs which are not finite. Treatment may consist of training sentences in which verbs do not need to be moved (e.g., hij is een boek aan het lezen (lit. he is a book to read; "he is reading a book") rather than hij leest een boek "he reads a book"). The Dutch therapy program of Bastiaanse, Jonkers, Quack & Varela Put (1996) can be used for this. Teaching this strategy is aimed at not only improving his ability to produce sentences, but also at increasing the diversity of verbs. 5. Conclusion Only few standardised assessment materials focusing on verbs and sen­ tences have been published. Considering the problems that aphasic patients have with verbs and sentences, it was decided to develop a test battery: the VAST. This test is based on theoretical insights and neurolinguistic findings. Deficits can be pinpointed for which treatment programs/methods exist or have been described in the literature. In the text above, two case studies have been described illustrating the clinical relevance of the VAST. We hope that the VAST contributes to the diagnosis and treatment of verb and sentence deficits in aphasia.

REFERENCES Balogh, Jennifer E. & Yosef Grodzinsky. 2000. "Levels of linguistic representation in Broca's aphasia: Implicitness and referentiality of arguments". Grammatical Disorders in Aphasia: A neurolinguistic perspective ed. by Roelien Bastiaanse & Yosef Grodzinsky, 88-104. London: Whurr.

THE VERB AND SENTENCE TEST

297

Basso, Anna, Carmelina Razzano, Pietro Faglioni & M. Ester Zanobio. 1990. "Confrontation naming, picture description and action naming in aphasic pa­ tients". Aphasiology 4.185-195. Bastiaanse, Roelien, Susan Edwards & Katalin Kiss. 1996. "Fluent aphasia in three languages: Aspects of spontaneous speech". Aphasiology 10.561-575. Bastiaanse, Roelien, Roel Jonkers, Christina Quack & Maria Varela Put. 1996. Werkwoordproductie op woord- en zinsniveau. Lisse: S wets & Zeitlinger. Bastiaanse, Roelien & Roel Jonkers. 1998. "Verb retrieval in action naming and spontaneous speech in agrammatic and anomic aphasia". Aphasiology 12.951-969. Bastiaanse, Roelien, Edwin Maas & Judith Rispens. 2000. De Werkwoorden- en Zinnentest (WEzT). Lisse: Swets & Zeitlinger. Bastiaanse, Roelien & Cynthia Thompson. 2000. "Verb finiteness in agrammatism: A cross-linguistic study". Brain and Language 74.503-514. Bastiaanse, Roelien & Ron van Zonneveld. 1998. "On the relation between verb in­ flection and verb position in Dutch agrammatic aphasies". Brain and Language 64.165-181. Berndt, Rita S., Anne N. Haendiges, Charlotte  Mitchum & Jennifer Sandson. 1997a. "Verb retrieval in aphasia: 1. Characterising single word impairments". Brain and Language 56.68-106. Berndt, Rita S., Charlotte  Mitchum, Anne N. Haendiges & Jennifer Sandson. 1997b. "Verb retrieval in aphasia: 2. Relationship to sentence processing". Brain and Language 56.107-137. Bird, Helen & Sue Franklin. 1996. "Cinderella revisited: A comparison of fluent and nonfluent aphasie speech". Journal of Neurolinguistics 9.187-206. Butterworth, Brian & David Howard. 1987. "Paragrammatisms". Cognition 26.1 -37. Butterworth, Brian, Marta Panzeri, Carlo Semenza & T. Ferreri. 1990. "Paragramma­ tisms: A longitudinal study of an Italian patient". Language and Cognitive Processes 5.115-140. Byng, Sally. 1988. "Sentence processing deficits: Theory and therapy". Cognitive Neuropsychology 5.629-676. Edwards, Susan. 2000. "A clinical assessment of verbs in an agrammatic patient". Grammatical Disorders in Aphasia: A Neurolinguistic Perspective ed. by Roelien Bastiaanse & Yosef Grodzinsky, 191-213. London: Whurr. Edwards, Susan & Roelien Bastiaanse. 1998. "Diversity in the lexical and syntactic abilities of fluent aphasie speakers". Aphasiology 12.99-117. Friedmann, Na'ama & Yosef Grodzinsky. 1997. "Tense and agreement in agram­ matic production: Pruning the syntactic tree". Brain and Language 56.397-425. Goodglass, Harold & Edith Kaplan. 1983. The Assessment of Aphasia and Related Disorders. Philadelphia: Lea & Febiger. Grodzinsky, Yosef. 1995. "A restrictive theory of agrammatic comprehension". Brain and Language 50.27-51. Grodzinsky, Yosef. 2000. "The neurology of syntax: Language use without Broca's area". Behavioral and Brain Sciences 23.1-73.

298

RISPENS, BASTIAANSE & EDWARDS

Grodzinsky, Yosef & Lisa Finkei. 1998. "The neurology of empty categories: Apha­ sics' failure to detect ungrammaticality". Journal of Cognitive Neuroscience 10.281-292. Hagiwara, Hiroko. 1995. "The breakdown of functional categories and the economy of derivation". Brain and Language 50.92-116. Huber, Walther, Klaus Poeck, Dorothea Weniger & Klaus Willmes. 1983. Der Aachener Aphasie Test. Göttingen: Hogrefe. Jonkers, Roel. 2000. "Verb finding problems in Broca's aphasics: The influence of transitivity". Grammatical Disorders in Aphasia: A Neurolinguistic Perspective ed. by Roelien Bastiaanse & Yosef Grodzinsky, 105-122. London: Whurr. Jonkers, Roel & Roelien Bastiaanse. 1996. "The influence of instrumentality and transitivity on action naming in Broca's and anomie aphasia". Brain and Lan­ guage 55.37-39. Kay, Janice, Ruth Lesser & Max Coltheart. 1992. Psycholinguistic Assessments of Language Processing in Aphasia (PALPA). London: LEA. Kohn, Susan E., Marjorie P. Lorch & Dawn M. Pearson. 1989. "Verb finding in aphasia". Cortex 25.57-69. Koster, Jan. 1975. "Dutch as an SOV language". Linguistic Analysis 1.111-136. Linebarger, Marcia, Myrna Schwartz & Eleanor Saffran. 1983. "Sensitivity to gram­ matical structure in so-called agrammatic aphasies". Cognition 3.361-392. Marshall, Jane. 1995. "The mapping hypothesis and aphasia therapy". Aphasiology 9.517-539. Miceli, Gabriela, M. Silveri, Giampiero Villa & Alfonso Caramazza. 1984. "On the basis of the 'agrammatics' difficulty in producing main verbs". Cortex 20.207-220. Schwartz, Myrna, Ruth Fink & Eleanor Saffran. 1995. "The modular treatment of agrammatism". Neuropsychological Rehabilitation 5.91 All. Schwartz, Myrna, Eleanor Saffran, Ruth Fink, J. Myers & Nadine Martin. 1994. "Mapping therapy: A treatment program for agrammatism". Aphasiology 8.19-54. Thompson, Cynthia, K. Lange, Sandra Schneider & Lewis Shapiro. 1997. '"Agram­ matic and non-brain damaged subjects' verb and verb argument structure pro­ duction". Aphasiology 11.473-490. Thompson, Cynthia, Lewis Shapiro, Mary Tait, Beverly Jacobs & Sandra Schneider. 1996. "Training wh-question production in agrammatic aphasia: Analysis of ar­ gument and adjunct movement". Brain and Language 52.175-228. Williams, Sarah E. & Gerald J. Canter. 1987. "Action naming performance in four syndromes of aphasia". Brain and Language 32:124-136. Zingeser, Louise B. & Rita S. Berndt. 1988. "Grammatical class and context effects in a case of pure anomia: Implications for models of language production". Cognitive Neuropsychology 5:473-516. Zurif, Edgar. 1995. "Brain regions of relevance to syntactic processing". Language ed. by Lila Gleitman & Mark Liberman, 381-398, Cambridge, Mass.: MIT Press.

CASE ASSIGNMENT AS AN EXPLANATION FOR DETERMINER OMISSION IN GERMAN AGRAMMATIC SPEECH* ESTHER RUIGENDIJK University of Utrecht 1. Introduction Agrammatic patients are known to have problems with the production of determiners and pronouns. Determiners are often omitted and a relatively low number of pronouns is used in (spontaneous) speech production (Saffran et al. 1989). Determiners and pronouns are function words, and these are vulnerable in agrammatic aphasia. Problems with function words, or grammatical mor­ phemes, is one of the main features of agrammatic aphasia (Saffran et al. 1989; Caramazza & Berndt 1985).1 Recently, it has been suggested that considering all grammatical morphemes as a homogeneous class is an overgeneralization (e.g., Friedmann & Grodzinsky 1997; Ruigendijk et al. 1999). According to Friedmann & Grodzinsky (1997) the production deficit in agrammatism is restricted to realisation of functional projections high in the tree (i.e., CP and TP). Ruigendijk et al. (1999) suggested that the production of determiners and pronouns is not impaired because they are grammatical morphemes, but because their realisation is dependent on the production of (finite) verbs. According to Ruigendijk et al., the basic problem is the production of (finite) verbs; poor determiner and pronoun production can then be considered a side effect (see also: Bastiaanse et al. in press). In contrast to function words (closed-class words), content words are said to be relatively intact in agrammatism. It has frequently been shown, however, that verbs are relatively difficult, and that finite verbs are even more difficult in agrammatism (Bastiaanse & Van Zonneveld 1998; Bastiaanse & Jonkers 1998; Kim & Thompson 2000). One can wonder whether these phenomena, that is, The authors thank Fedor Jalvingh, Martina Penke and Luise Springer for referring patients for this study. We are grateful to Dirk den Ouden for is comments on an earlier version of this paper. 1 'Function words' and 'closed-class words' refer to the same word classes: determiners, prepositions, pronouns, etc. As it is also mentioned that 'grammatical endings' are often missing, we prefer the term 'grammatical morphemes' as this refers to both fuinction/closedclass words and grammatical endings.

300

ESTHER RUIGENDIJK

reduced number of determiners and poor verb production, are related. Based on linguistic theory, which will be discussed in the next section, the expectation is that there is a causative effect, as suggested by a former study (Ruigendijk et al. 1999) in which the spontaneous speech of Dutch and German agrammatic aphasics was analysed. In the study presented here, we will test the hypothesis that there is a relationship between the production of de­ terminers and the production of (finite) verbs in agrammatic speech. This hypothesis is neurolinguistically interesting, because it theoretically relates two well-known characteristics of agrammatic speech, and it is also clinically relevant, since it may have implications for therapy. The first section of this paper will give the linguistic background to our topic. The next section describes the psycholinguistic background. In the third section, relevant aphasiological studies are presented. This section is followed by a description of the methods of the study. Subsequently, a quantitative and qualitative analysis of the results will be given and these will be discussed in the final section of this paper. 2. Linguistic background According to Chomsky, all languages have an abstract case system. Case may or may not be morphologically realised. In a language like English or Dutch, it can only be seen on the pronouns (7 vs. me; he vs. him). In the lan­ guage under investigation in this study, German, not only pronouns, but also determiners are marked for case (see (1) for the German case paradigm). (1) nominative genitive dative accusative

masculine der Mann (the man) des Mannes dem Mann den Mann

feminine die Frau . (the woman) der Frau der Frau die Frau

neutral das Kind (the child) des Kindes dem Kind das Kind

plural die Kinder (the children) der Kinder den Kindern die Kinder

Case, that is, abstract case, expresses the syntactic relation between a verb and its arguments. It specifies, for example, which noun phrase is the subject of the sentence and which one is the object. Noun phrases therefore require case. Chomsky (1986, 1995:111) formulates this in the Case Filter, "every phonetically realised NP must receive (abstract) Case". This means that a sen­ tence with a noun phrase that has not been assigned case is ungrammatical. Following Chomsky (1986), the head of a noun phrase is the noun, therefore it is analysed as an NP (noun phrase). Abney (1987) suggests another analysis, in which the determiner of the noun phrase is the head, therefore a noun phrase

CASE ASSIGNMENT IN GERMAN AGRAMMATIC SPEECH

301

should be analysed as a DP (determiner phrase).2 Ouhalla (1993) refines Chomsky's Case Filter, using the DP-analysis of noun phrases. He suggests that every DP, i.e., every complete noun phrase, must have abstract case. Bare noun phrases (NPs in his terminology) are not subject to the Case Filter. In other words: only the occurrence of DPS is restricted by the Case Filter, not the occurrence of NPs. The question is: how do DPS get case? According to Chomsky (1986, 1995), abstract case is assigned by a verb or a preposition in the following way: the finite verb, which has moved to AgrS, assigns nominative case to the subject of a sentence (in [Spec, AgrS], see figure 1). The subject DP is assumed to be base-generated vp-internally, following Koopman & Sportiche (1991), which means that the subject DP has to move to the specifier position of AgrS to receive its nominative case from the finite verb. AgrSP Spec

AgrS'

Fig. 1 : Case assignment to the subject and object noun phrase

Accusative and dative are both assigned by the verb to the direct and indi­ rect object of a sentence, respectively. Without a finite verb, nominative case cannot be assigned and thus no subject DP should be present. This means that the production of properly casemarked subject determiners depends on the presence of a case assigning finite verb that has moved to AgrS. In the same way, without a verb, no case can be assigned to the object of a sentence and therefore no object determiners can be present when there is no verb. Case-marked object determiners are thus de­ pendent on the presence of a verb.

See Abney (1987) for technical details.

302

ESTHER RUIGENDIJK

The German examples (2) and (3) show this. The finite verb wird (will) in (2) and (3) assigns nominative case to the subject of the sentence der Mann (the mannom). The object of sentence (2) den Jungen (the boyacc) is assigned accusative case by the verb tragen (to carry). German allows for dative case assignment to the object of a sentence as well (also depending on the verb), as can be seen in (3). This dative assignment is associated with the theta role 'benificiary'. In (3) the verb helfen (to help) assigns dative case to the object of the sentence dem Jungen (the boydat). (2) der Mann

wird

NOM

den Jungen ACC

the man-NOM will the boy- ACC "the man will carry the boy" (3) der Mann

tragen

wird

dem Jungen

carry helfen

DAT

the man-NOM will "the man will help the boy"

the boy-DAT

help

Whether a verb assigns dative case - or even genitive - to an object noun phrase has to be learnt for each individual verb. This information can be seen as part of the lexical entry of the verb (Haider 1984; Haegeman 1994). This kind of case assignment is called inherent case assignment, as opposed to structural case assignment, which is dependent on the syntactic structure, as described above, i.e., nominative case assignment to the subject of a sentence in [Spec, AgrS] and accusative case assignment to the object of a sentence, by V.3 3. Psycholinguistic background The information in the lexical entry of a verb and the retrieval of this in­ formation are important for case assignment and thus for the production of DPS. According to Levelt (1989) a lexical entry contains information about the meaning of a specific item, and about the syntactic properties of this item, for example, to which category it belongs (V for hit) and the syntactic arguments it can take {hit takes an external subject and an internal object). The semantic and syntactic information of an item constitute the lemma. Together with mor3

For a detailed discussion on the difference between 'structural' and 'inherent' or 'lexical' case in German see, for example, Haider (1984), Czepluch (1998).

CASE ASSIGNMENT IN GERMAN AGRAMMATIC SPEECH

303

phological and phonological information (i.e., the word form or lexeme) the lemma forms the lexical entry of a word. Systematic relations exist between these different kinds of information in the lexical entry. In his blueprint for the speaker, Levelt (1989) distinguishes several stages from intention to articulation. First a conceptualizer generates a preverbal message. This message is translated into a linguistic structure by a so-called formulator that can be divided into two components, a grammatical encoder and a phonological encoder. For grammatical encoding, the semantic and syntactic information, i.e., the lemma information, of a lexical entry is needed. Phonological encoding is triggered by word form information (the lexeme) and results in a phonetic plan. This phonetic plan, finally, is needed for articulation. For this study, specifically the retrieval of the syntactic information on verbs is important, in other words, the retrieval of verb lemmas. This is used for grammatical encoding. A lemma can only be activated when the conceptual conditions of the preverbal message can be matched with the semantic information of this lemma. When the lemma is retrieved, the syntactic information also becomes available. According to Haider (1984), the information on the verb not only contains how many and which arguments a verb can take, but also how these arguments are assigned case: Lexically (L) or Structurally (s). In (4) two examples are given of the case assignment information in a verb's lexical entry: note that in these examples A1 is normally realised as 'external argument' and A2 as 'internal argument' (examples from Haider 1984). In (4a) both arguments of the verb schlagen must be assigned structural case, which will result in nominative for the external argument (Al s ), the subject, and accusative for the internal argument (A2S), the object. In (4b) only the external argument (AI s ) of the verb helfen, the subject, is assigned structural nominative case, the internal argument (A2L), the object, is assigned lexical case, namely dative. (4) a. schlagen (A1s A2S) to hit b. helfen (A1s A 2 L ), L=Dative to help It can be concluded from psycholinguistic and linguistic theory that the verb is important for case assignment. First, the syntactic information of the lexical entry of the verb specifies how case is to be assigned to the arguments that a verb takes: lexically or structurally. Second, structurally, the subject DP of a sentence depends on the finite verb of a sentence for its case and the ob-

304

ESTHER RUIGENDIJK

ject DP is dependent on the verb for its case. In other words, the presence of a DP depends on the presence of a case assigning verb, and the correct case of this DP depends on the complete retrieval of the syntactic information of that verb. Verbs are known to be vulnerable in Broca's aphasia, the effects of this vulnerability on the ability of Broca's aphasics to produce determiners and as­ sign case will be clarified in the next section. 4. Aphasiological background Agrammatic aphasics have problems with the production of verbs on naming tasks. They also produce fewer verbs in their spontaneous speech than non-brain damaged speakers do, and when they produce verbs, these verbs are often non-finite (Bastiaanse & Jonkers 1998; Bastiaanse & Van Zonneveld 1998). Lexical information must be encoded grammatically for the production of sentences (Levelt 1989). Bastiaanse (1995) as well as Kim and Thompson (2000) ascribe part of the problems Broca aphasies have with verbs to an im­ pairment in grammatical encoding; the lemma information of the verb cannot be processed or accessed correctly. Structural case assignment, and consequently determiner production, de­ pends on the presence of a (finite) verb. Ruigendijk et al. (1999) analysed the spontaneous speech of German and Dutch agrammatic aphasies with regard to determiner production in relation to verb production and inflection. They showed that there is such a relationship. No object determiners (i.e., no accu­ sative or dative marked noun phrases) were produced in the absence of a case assigning verb. The patients did produce some nominative case marked noun phrases, but these were explained with a so-called default strategy: nominative case is assigned to a subject NP, when this cannot be done structurally by a finite verb in [Spec, AgrS]. This default strategy was originally proposed for normal speech (Van Zonneveld 1994), but can be used to describe agrammatic data as well. The most important finding of this study was that when no proper case assigner was present, the determiner was often omitted. German determiners are, as mentioned above, marked for case, but also for number and gender. De Bleser et al. (1996) demonstrated that a group of German agrammatic aphasies was able to inflect isolated noun phrases for gender. When these same patients had to produce the same noun phrases in the context of a sentence, where not only gender, but also case plays a role, their performance was far less accurate. This was interpreted as a problem with case marking, rather than with gender marking. Information about how case has to be assigned is assumed to be part of the lexical entry of a verb (Haider 1984).

CASE ASSIGNMENT IN GERMAN AGRAMMATIC SPEECH

305

Incompletely or incorrectly retrieved lemma information will negatively influence grammatical encoding. The syntactic information on the verb speci­ fies, among other things, how case is to be assigned to the arguments of a verb. If this information is not available for a patient, case errors are expected. Con­ sidering the issues described above, the expectation of this study is that there is a relationship between verb production and determiner production. Impaired verb retrieval will have a negative influence on determiner production. This leads to the following hypothesis for agrammatic speech production: (5) The production of complete noun phrases (DPS), that is, the pro­ duction of determiners, is related to the realisation of a case assigning verb. 5.

Methods

5.1 Subjects and materials Ten agrammatic Broca's aphasics participated in this study (8 male, 2 fe­ male; mean age 55,4). All patients were examined more than a year post onset. Nine patients were right-handed and one left-handed. All were aphasic due to a single stroke in the left hemisphere. The patients were diagnosed as Broca's aphasies on the basis of the Aachen Aphasia Battery (AAT, Huber et al. 1983) and this diagnosis was confirmed by the speech therapist and the examiner. The speech production of all aphasie patients was characterised as telegraphic. The experimental data of this group were compared to the data of a group of 16 German non-brain damaged speakers (7 male, 9 female; mean age 55,6). First, a spontaneous speech sample of each patient was elicited. This was done by a semi-structured interview, according to AAT standards. In this inter­ view, patients were asked to describe their history of illness, job, family and hobbies. These speech samples were then analysed to examine whether and how the patients produced determiners, and whether the production of deter­ miners depended on the production and inflection of verbs. Subsequently, the data of the spontaneous speech analysis were compared to the data of the two experimental tasks. The first experimental task was a sentence completion test that contained 30 items. For each item, subjects were presented with a picture and an incom­ plete svo-sentence (see figure 2) that was to be completed with a noun phrase. Either the subject or the object noun phrase was missing. The patients were asked to read the sentence (or repeat it in case of reading difficulties) and to complete it with the missing noun phrase. For figure 2 this would result in, e.g., der Junge ähnelt... [dem Mann] (the boynom looks like .... [the mandat]).

306

ESTHER RUIGENDIJK

This task meant to examine whether the patients were able to produce a deter­ miner when the case marking verb was given. After this task, the patients were presented with a sentence production task in which the same pictures were used, but now only the non-finite verb was given (see figure 2). The patients were asked to tell in one simple sentence, using the given verb, what was happening in the picture. The sentence for the picture in figure 2 aims to elicit something like die Frau überwacht den Mann (the womannom guards the manacc). The patients were told that they should try to change the infinite verb into a finite verb. This was done to avoid that the patients would think they had to produce a sentence with the infinite verb. Through this task it was possible to test the patients' ability to inflect a verb for tense and agreement in combination with the production of the correct de­ terminers for the noun phrases in the sentence.

Fig. 2: Two examples of the test for the production of determiners (first: sentence completion; second: sentence construction)

These two tasks aimed to evoke nominative subjects, accusative objects and dative objects. In each task, there were 10 items for each constituent, which makes 60 items in total. Note that in German only masculine nouns are unambiguously marked for case (see the German paradigm in 1), therefore the critical constituent was always represented by a masculine noun. 5.2 Scoring Since the question of this study was whether there is a relationship be­ tween the production of determiners and the production and inflection of verbs, the spontaneous speech data were analysed with regard to determiner production in relation to verb production. The noun phrases were counted, and then each noun phrase was examined to check whether a correctly case-marked determiner was produced, and whether a case assigning verb was present. The

CASE ASSIGNMENT IN GERMAN AGRAMMATIC SPEECH

307

difference between (1) noun phrases plus determiner, plus case assigning verb, and noun phrases without determiner without case assigning verb and (2) noun phrases without determiner plus case assigning verb and noun phrases plus determiner without a case assigning verb was statistically tested. Occurrences such as (1) are expected according to the linguistic theory described above, whereas occurrences such as (2) were not expected. In sentence (6a) and (6b), examples are shown of respectively a noun phrase plus determiner, plus case assigning verb, and a noun phrase without a determiner, without a case assigning verb. Examples (7a) and (7b) show the unexpected combinations, namely a noun phrase plus determiner, but without a case assigning verb (7a) and a noun phrase without a determiner, but plus a case assigning verb (7b). (6) a. dann habe ich ein Kind than have I one child "I have one child" b. Kohlrabi langsam kohlrabi slow (7) a. das Essen fertig the food-NOM ready b. Krankenwagen bestellt ambulance called-PART For the tests, the noun phrase was considered to be correct and complete when it was produced with a correctly case-marked determiner. For the sen­ tence production task it was also examined whether the given verb was in­ flected correctly by the patient. After this, the errors on the noun phrases were analysed for a qualitative comparison of the data. Four error types were identi­ fied: case substitution, gender substitution, determiner omission and other er­ rors (nil reactions or irrelevant reaction). See (8) for examples. (8) a. case substitution der Junge begegnet the boy meets Goal: der Junge begegnet the boy meets b. gender substitution Katze kratzen cat Goal: die Katze kratzt the cat scratches

den Vater the father-ACC dem Vater the father-DAT die Mann scratch-INF the man-FEM den Mann the man-ACC

308

ESTHER RUIGENDIJK

c. determiner omission das Mädchen beißt Mann man the girl bites Goal: das Mädchen beißt den Mann the girl bites the man-ACC d. other error (irrelevant completion) das Mädchen erwürgt mit immer die Klassenarbeit the girl strangles with always the examination Goal: das Mädchen erwürgt den Jungen the girl strangles the boy-ACC 6.

Results First, a quantitative analysis of spontaneous speech and the test results will be given. Table 1 shows the results of the spontaneous speech analysis of ten German agrammatic speakers (the expected combinations are bold). All pa­ tients omitted determiners in their spontaneous speech. As a group they pro­ duced 378 noun phrases. In 227 (60%) of these noun phrases the determiners were omitted, in 151 (40%) cases the determiners were produced. For 198 noun phrases no case assigning verb was present. There was a significant dif­ ference between (1) the expected combinations, namely the noun phrases plus determiner, plus case assigning verb, and the noun phrases without determiner, without a case assigning verb and (2) the unexpected combinations: the noun phrases without a determiner, but plus case assigning verb and the noun phrases plus determiner, but without a case assigning verb (t=5,586, df=9, p<0.001). Only one accusative and no dative marked noun phrases without a case assigning verb were produced. In other words, all noun phrases plus determiner but without a case assigning verb (n=21) had nominative case. No gender errors were made in the spontaneous speech.

total number

+ Case assigning verb + determiner - determiner 50 130

- Case assigner verb + determiner - determiner 21 177

Table 1 : Determiners in the spontaneous speech in relation to the use of a case assigner

The Quantitative Analysis of Experiments shows that there was no dif­ ference between patients' results on the two experimental tasks, (t=-0,536, df=9, p=0,605), therefore these results will be taken together per patient as can be seen in figure 3 and table 2. The agrammatic aphasics produced 64,2% of all noun phrases correctly, which is significantly worse than the control group of non-brain damaged speakers, who produced 94,5% of the noun

CASE ASSIGNMENT IN GERMAN AGRAMMATIC SPEECH

309

phrases correctly (t=-5,943, df= 24, p<0,001). The patients produced 88% of the subjects correctly, which is worse than the control group (99,7%, t=2,373, df=24, p<0,05). 65,5%) accusative objects were produced correctly, and 39% dative objects, both significantly worse than the control group (accusative 96,9% correctly, t=-4,232, df=24, p<0,001; dative 86,9% correctly t=-5,272, df=24, p<0,001). The patients were usually able to inflect the nonfinteverb that was given in the sentence production task, in only 21 cases (out of 300) the verb was not inflected. Thirteen of these infinite verbs were produced by patient 8. patient 1 2 3 4 5 ! 6 7 8 9 10 total non-brain-damaged speakers

nominative 85 90 100 95 100 100 100 35 95 80 88 99,7

accusative 80 45 90 80 30 90 90 5 70 75 65,5 96,9

dative \ 10 85 10 35 95 60 40 0 20 35 39 86,9

Table 2 : The percentages of correct noun phrases produced on the sentence completion and sentence production task

D patients □ non-brain-damaged speakers

nominative

accusative

dative

Fig. 3: Determiner production in sentence completion and production: percentage correct

310

ESTHER RUIGENDIJK

The Qualitative analysis shows the results of an error analysis over the data of the experimental tasks (Table 3). The patients made a few gender errors (14, which is 6,4% of all errors), some determiner omissions (51, which is 23,3% of all errors). These determiner omissions were most frequently made by one patient (8) who omitted the determiner 44 times. The majority of the errors (130, which is 59,4% of all errors) were case substitutions. #case substitution # errors

130

# gender substitution

# determiner omission

# other errors

total

14

51

24

219

Table 3: Error types: total number of errors for each category

As Table 2 shows, all patients, except patient 8, do relatively well on the production of nominative determiners for subject noun phrases, although significantly worse than the normal controls. More problems arise for the ac­ cusative and dative objects, on which many errors are made, most often case substitutions. The scores for the accusative and dative objects are both significantly lower than the scores for the nominative subjects (t=3,391, df=24, p<0,05 and t=5,389, df=24, p<0,05 respectively). The performance on the dative and accusative objects does not differ significantly (t= 1,793, df=24, p=l,07). When the individual scores are analysed, there is one particularly no­ table difference. Patient 8 scores low on all constituents. The other nine pa­ tients show two reversed patterns of object case marking. The first pattern can be seen in seven patients who are relatively good in producing accusative objects, but poor in the production of dative objects. The most frequent error for these dative objects is case substitution: the patients often produce accusa­ tive case instead of dative (73,5% of all incorrect dative objects). The other two patients show the opposite pattern. They are good in the production of da­ tive objects, but make many errors when they have to produce an accusative object (45%) and 30% correctly respectively): they produce the dative form in­ stead of the accusative (72%> of all incorrect accusative objects). Hardly any objects were realised with nominative case (4,5%). 7.

Discussion The hypothesis of this study is that the production of complete noun phrases (DPS), that is, the production of determiners, is related to the realisation of a case assigning verb. The results presented above confirm this hypothesis. There is such a relationship in agrammatic speech production. Although all patients omit determiners in their spontaneous speech, they are, with one exception (patient 8), very well able to produce complete noun

CASE ASSIGNMENT IN GERMAN AGRAMMATIC SPEECH

311

phrases with a case marked determiner when presented with the verb, as the results of the two experimental tasks show. There is only one patient who omits determiners on both the sentence completion and sentence production task. This patient frequently produces an infinite verb on the sentence production task (n=13). For all other patients, it holds that when the (finite) verb is given, it is produced and inflected, and the determiner of the noun phrase is hardly ever omitted (7 times out of 540 noun phrases). Moreover, the results of the spontaneous speech analysis show that if a case assigning verb is realised, the determiner is often produced as well, confirming the results of Ruigendijk et al. (1999). This means that the production of determiners as such is not impaired in agrammatism. If the patients are not able to produce the proper case assigning verb, as often happens in spontaneous speech, they are usually also unable to produce a complete noun phrase. The problems with determiners seem to be caused by an underlying problem with case assignment. An underlying problem with gender can be ruled out, since only 14 gender error and many case errors (n=130) are made, thus confirming the results of De Bleser et al. (1996), who found that German agrammatic patients are able to inflect an isolated noun phrase for gender, but had problems in inflecting the same noun phrases in a sentence where case as­ signment plays a role. In other words, there is a relationship between the presence of a (finite) verb that assigns case and the realisation of a determiner in the production of these German agrammatic aphasics. A linguistic notion, namely case assignment, can thus be used to describe a phenomenon in agrammatic speech: the relationship between verb production and determiner production. However, this is not the whole story. Even if they have retrieved the verb, the patients still have some problems assigning the correct grammatical role to the object noun phrases. Subject noun phrases are case marked correctly most of the time, but object noun phrases are more problematic. The patients make case substitution errors on the object noun phrases. With regard to these case substitution errors, there are two different patterns. Seven of the patients overgeneralize accusative case, thus scoring well on the accusative objects, but poor on the dative objects, and two patients show the opposite pattern, namely overgeneralization of dative case, which results in a high score on the dative objects, but a low score on the accusative objects. It looks as if the patients know the difference between a subject and an object, which is determined structurally, that is, it depends on the position in the sentence structure and the presence of a case assigning verb. When they have to case mark an object, they prefer objective case, that is, accusative or dative, even if not always

312

ESTHER RUIGENDIJK

correctly. They almost never produce a nominative case marked object noun phrase, which would result in a sentence with two nominatives. It has been mentioned above that agrammatic aphasics have problems in accessing the lemma of a verb, which includes the syntactic information. This implies that they do not retrieve the information correctly or completely. Note that according to Haider (1984) the lemma of a verb also contains information about which kind of case assignment applies to each argument that a verb selects: lexical or structural case assignment. If this information is not retrieved correctly or if it is retrieved incompletely, the patient does not know whether an object must receive accusative or dative case, and will thus make case substitution errors. As Kim & Thompson (2000:15) assert: "disrupted access at the lemma level will inevitably affect operations involved in the subsequent stages during sentence production". One of the affected operations is case assignment. Most importantly, these results demonstrate that once the verb is realised, the production of determiners is less impaired in agrammatic aphasics who omit determiners in their spontaneous speech production. This has interesting implications for treatment of agrammatic patients. Training the production of isolated DPS will not be successful, since these are related to verbs. It is better to train verb production, as the results of the present study suggest that this will increase the production of complete noun phrases. Results of a study per­ formed by Springer et al. (2000) show that this is indeed correct. They found in some of their severely agrammatic patients an increase in the production of complete noun phrases after treating these patients with a program that aimed at the production of (infinite) verbs combined with noun phrases. 8. Conclusion The results of this study on German determiner production and case as­ signment demonstrate that once the case assigning verb is realized, the pro­ duction of determiners is possible in agrammatic aphasies. This means that determiner production as such is not impaired, but rather that the problems with determiners are closely related to the problems with the production of verbs. When no case assigning verb is realized, no determiners can be pro­ duced. Incorrect or incomplete retrieval of the lemma information of the verb can account for case substitution errors that are made with object noun phrases.

CASE ASSIGNMENT IN GERMAN AGRAMMATIC SPEECH

313

REFERENCES Abney, Steven. 1987. The English Noun Phrase in its Sentential Aspect. Ph.D. dis­ sertation. Cambridge, Mass.: MIT Press. Bastiaanse, Roelien. 1995. "Broca's aphasia: A syntactic and/or a morphological disorder? A case study". Brain and Language 48.1-32. Bastiaanse, Roelien & Roel Jonkers. 1998. "Verb retrieval in action naming and spontaneous speech in agrammatic and anomic aphasia". Aphasiology 12.951-969. Bastiaanse, Roelien, Judith Rispens, Esther Ruigendijk, Onésimo Juncos-Rabadán & Cynthia . Thompson. In press. "Verbs: Some properties and their consequences for agrammatic Broca's aphasia". Journal of Neuro Unguis tics. Bastiaanse, Roelien & Ron Van Zonneveld. 1998. "On the relation between verb in­ flection and verb position in Dutch agrammatic aphasies". Brain and Language 64.165-181. Caramazza, Alfonso & Rita S. Berndt. 1985. "A multicomponent deficit view of agrammatic Broca's aphasia". Agrammatism ed. by Mary-Louise Kean, Orlando: Academic Press. Chomsky, Noam. 1986. Knowledge of Language: Its nature, origin and use. New York: Praeger. Chomsky, Noam. 1995. The Minimalist Program. Cambridge, Mass.: MIT Press. Czepluch, Hartmut. 1988. "Kasusmorphologie und Kasusrelationen: Überlengungen zur Kasustheorie am Beispiel des Deutschen". Linguistische Berichte 116.275-310. De Bleser, Ria, Josef Bayer & Claudio Luzatti. 1996. "Linguistic theory and morphosyntactic impairments in German and Italian aphasies". Journal of Neurolinguistics 9.175-185. Friedmann, Naama & Yosef Grodzinsky. 1997. "Tense and agreement in agrammatic production: Pruning the syntactic tree". Brain and Language 56.397-425. Haegeman, Liliane M. V. 1994. Introduction to Government and Binding Theory Oxford: Blackwell, Haider, Hubert. 1984. "The Case of German". Studies in German Grammar ed. by Jindrich Toman 65-102. Dordrecht: Foris. Huber, Walter, Klaus Poeck, Dorothea Weniger & Klaus Willmes. 1983. Der Aachener Aphasie Test (AAT). Göttingen: Hogrefe. Kim, Mikyong & Cynthia K. Thompson. 2000. "Patterns of comprehension and pro­ duction of nouns and verbs in agrammatism: Implications for lexical organiza­ tion". Brain and Language 74.1-25. Koopman, Hilda & Dominique Sportiche. 1991. "The positions of subjects". Lingua 85.211-258. Levelt, Willem J. M. 1989. Speaking: From intention to articulation. Cambridge, Mass.: MIT Press. Ouhalla, Jamal. 1993. "Functional categories, agrammatism and language acquisi­ tion". Linguistische Berichte 143.3-36.

314

ESTHER RUIGENDIJK

Ruigendijk, Esther, Ron Van Zonneveld & Roelien Bastiaanse. 1999. "Case as­ signment in agrammatism". Journal of Speech, Language and Hearing Research 42.962-971. Saffran, Eleanor M., Rita S. Berndt & Myrna F. Schwartz. 1989. 'The quantitative analysis of agrammatic production: Procedure and data". Brain and Language 37.440-479. Springer, Luise, Walter Huber, Klaus J. Schlenk & Claudia Schlenk. 2000. "Agrammatism: Deficit or compensation? Consequences for aphasia therapy". Neuropsychological Rehabilitation 10.279-309. Zonneveld, Ron van. 1994. Kleine Syntaxis van het Nederlands. Dordrecht: ICG Publications.

THE ROLE OF VERBAL MORPHOLOGY IN APHASIA DURING LEXICAL ACCESS EVIDENCE FROM GREEK*

KYRANA TSAPKINI, GONIA JAREMA & EVA KEHAYIA Université de Montréal 1. Introduction The dissociation between regular and irregular morphology in aphasia has been documented in different languages (e.g., English, German and Italian). In studies of English and German, dissociations between regular and irregular verbal inflection have been found in both normal processing and language im­ pairment (for a review, see Pinker 1999 and Clahsen 1999), suggesting two distinct mechanisms: the rule-based and the full-form-based. Aphasic patients with frontal lesions, diagnosed as Broca's aphasics, were found to be impaired in the application of the past-tense inflection but while having no difficulty with irregular stored forms (Ullman, Corkin et al. 1997). Moreover, the authors claimed that the regularity-irregularity distinction is the manifestation of two different brain mechanisms, the 'procedural' and the 'declarative' systems, which are subserved by anterior and posterior brain areas, respectively. However, connectionist accounts of language tried to challenge this position by showing that artificial lesions to language-simulation models elicited novel/irregular dissociations without assuming different mechanisms (Joanisse & Seidenberg 1999). Thus, lesions at the level of phonological representations of words produced an impairment of rule-governed verbal inflection (e.g., wug-wugged), whereas lesions in the semantic representations The research reported here was supported by an MCRI grant from the Social Sciences and Humanities Research Council of Canada (412-95-0006) awarded to Gonia Jarema (Université de Montréal), director and co-principal investigator; Eva Kehayia (McGill University), coprincipal investigator, and Gary Libben (University of Alberta), co-principal investigator. We are grateful to patients IS and MH for their participation and their patience throughout the testing period. We would also like to thank Dr Artemis for referring patient MH to us and allowing us access to the files of the AHEPA hospital of Thessaloniki. We are indebted to Cristina Manouilidou for her unfailing efforts in collecting the control participants' data. Finally, we would like to thank Elisabetta Fava, Alberto Mioni, Carlo Semenza and Brenda Rapp for their insightful and constructive discussions on previous presentations of this work.

316

TSAPKINI, JAREMA & KEHAYIA

caused erroneous performance on irregular (or stored) inflection (e.g., taketook). A possible artifact in English is that regular past-tense forms are usually more phonologically similar to their base forms (e.g., mark-marked) than irregular forms (e.g., teach-taught) so that regular forms are more sensitive to phonological similarity effects. Therefore, morphological effects may be confounded by form effects. In order to establish whether morphologically regular and irregular forms are represented differently they would have to be matched with respect to degree of form similarity between the present and past-tense forms. Since this is impossible in English, it is important to address the issue of morphological regularity in languages that allow to dissociate the effects of morphological regularity from those of form and semantic similarity. This would allow to maintain the claim of universality of the distinction between rule-based and storage mechanisms in language processing and breakdown (Pinker 1999). Greek is precisely a language that offers the opportunity to investigate morphological regularity while keeping form and meaning overlap between base and past-tense forms constant across regular and irregular forms. Aim of this paper is to establish whether Greek morphologically regular and irregular forms are represented differently. 2. The regular/irregular distinction Conflicting results have been obtained in different languages on the question of whether regular and irregular inflection can be described as two qualitatively different operations, or a single mechanism based on statistical probability or associative memory. For example, from the field of language acquisition both approaches have gathered considerable evidence for English and Italian (for English, see, e.g., Marcus, Pinker et al. 1992 and Plunkett & Marchman 1993; for Italian, see, e.g., Say & Clahsen 1999 and Orsolini, Fanari & Bowles 1998). However, another possibility that has not been entertained thus far is that regularity or irregularity could, in some languages, represent the two ends of a continuum, with intervening degrees of idiosyncracy. The study of the regular/irregular distinction in Greek presents an interesting new dimension because regularity is not an 'all-or-nothing' con­ cept in the language, since past-tense formation yields a more complex picture in Greek as compared to other languages such as English. More specifically, verbal inflection for perfective past-tense forms in Greek encompasses phe­ nomena ranging from phonological change during the application of the regu­ lar past-tense rule to the most unpredictable stem-allomorphic changes in past-tense formation, with the in-between case where an allomorphic stem combines with the regular past-tense rule. Thus, the purpose of this study was

THE ROLE OF MORPHOLOGICAL STRUCTURE

317

to explore the role of regularity in language breakdown by investigating how these phenomena are manifested in the performance of Greek-speaking apha­ sics. As mentioned above, the study of morphologically regular and irregular verbs in Greek allows to address the issue of morphological regularity per se, independently of form and semantic considerations. In addition to aiming at shedding light on universal patterns of lexical processing in aphasia, this study aimed at clarifying whether the difficulties Greek-speaking aphasics have with the past tense are due to the different morphological operations postulated, or whether they reflect task-specific deficits related to the lexical access proce­ dures involved in each task. Several recent models of lexical access claim that inflected words are ac­ cessed by both whole-word and morpheme-based routes (Baayen, Dijkstra & Schreuder 1997; Frauenfelder & Schreuder 1992, and Caramazza, Laudanna & Romani 1988). However, there is still a substantial debate surrounding the is­ sues of when and how these routes interact and what factors influence their implementation (see Bertram, Schreuder & Baayen 2000, for a discussion). It has also been claimed that the differences between regular and irregular pasttense forms in English mirror the different routes that are involved in each category, i.e., morpheme-based for regulars and whole-word for irregulars. In Greek, however, this distinction is not valid because, since all words are in­ flected, the decomposition route should apply for the irregulars as well. It has previously been claimed that the mere presence of a suffix triggers decompo­ sition (Taft 1994, 1988; Taft & Forster 1975). Since regular and irregular forms in Greek are affixed with inflectional suffixes, decomposition can be assumed for all forms, at least in the initial stages of lexical access and for the inflectional suffixes that are common to both (i.e., not the rule-based aspectual marker -s). This does not exclude the possible application of the whole-word route. If whole-word access takes place, it does so for both regular and irregular forms that are rendered equal in that respect. As already claimed, both routes may apply in the initial stages of lexical access (Tsapkini, Jarema & Kehayia in press,b). What differentiates regular and irregular (rule based and non-rule based) morphological processing is what happens in later stages, after the stripping of the common inflectional affixes, i.e., what happens at the stem level. This is particularly important in a stem-based language such as Greek. In past-tense formation in Greek the stem morphemes accessed are also subject to allomorphic variation. Furthermore, it is the very presence of a stem-allomorph that marks the most 'irregular' (in the sense of idiosyncratic and unpredictable) past-tense formation. The rule-based aspectual marker may or may not be combined with the allomophic stem to form the past tense. It has been claimed

318

TSAPKINI, JAREMA & KEHAYIA

for another highly inflected language, Finnish, that whatever morphophonological adjustment occurs at the level of the stem, it does not affect mor­ phological processing (Niemi, Laine & Tuominen 1994). "Fusionality has not been shown to affect the speed and accuracy of recognition of inflected forms, leading to the hypothesis that the stem variants have independent representa­ tions at the visual input lexicon" (Laine, Vainio & Hyönä 1999). However, in Greek, stem changes are not uniform and they can vary from allomorphy to stem-final phonological adjustment as described below. 2.1 Greek inflectional system: real-time and off-line morphological processing In our study of the effects of morphological and morphophonological op­ erations in the processing of past tense in active non-derived verbs in Greek, we adopted Ralli's (1988) linguistic analysis of verbs as a theoretical framework. In general, all stems are bound in the language, and active verbs are formed by the addition of the inflectional suffix -o after the present tense stem, e.g., pez-o. There are three different types of active past tense formations in the language: 1) verbs with stem-internal change, e.g., plen-o, e-plyn-a, "I wash", where the perfective stem plyn- is an allomorph of the imperfective stem plen-, 2) verbs with phonological change, e.g., graf-o, e-grap-s-a, "I write", and lin-o, e-li-s-a "I untie", where the presence of the perfective aspectual marker -s- triggers a phonological alternation of lfl to /p/ and a stem-final consonant deletion, respectively, and 3) verbs with both a perfective allomorph and the addition of the aspectual marker -s-, e.g., mil-o, mili-s-a, "I speak", where mili- is the perfective allomorph of mil-. The second category can be seen as the equivalent of a rule-based paradigm, and the third category as the combination of rulebased and stored-allomorph paradigms. Previous neurolinguistic investigations of Greek addressed the issues of access and representation of inflected and derived lexical items as revealed by the performance of aphasic subjects in off-line and on-line experiments (Kehayia 1988; Kehayia, Jarema & Kadzielawa 1990; Kehayia 1997). The main finding of these studies was that rules of word-formation and underlying representations of morphemes may not lost in non-fluent aphasia; rather, what is lost is access to morphemes and to the mechanisms that combine them. This, in turn, impairs the processing of morphologically complex words. In order to investigate whether morphological deficits in aphasia reflect an impairment at the level of lexical representations or at that of lexical access mechanisms, Kehayia (1997) conducted an on-line word recognition experi­ ment which probed, among other issues, the role of stem transparency in the processing of past tense forms. Twenty-four non-impaired adults and two stroke patients with non-fluent aphasia participated in a simple and a primed

THE ROLE OF MORPHOLOGICAL STRUCTURE

319

visual lexical decision experiment, where they performed word-nonword judgements for regular and irregular verbs. It was found that the non-impaired and aphasic subjects exhibited the same patterns in response latencies, even though the aphasic subjects manifested an overt morphological deficit in off­ line comprehension, production and repetition tasks. Kehayia (1997) concluded that an acquired morphological deficit in aphasia does not necessarily imply an impairment at the level of underlying mental representa­ tions, but rather attenuates lexical access and impairs production. Given the theoretical framework outlined above and previous findings on Greek, the question is how the full range of the regular-irregular distinction in Greek is manifested in aphasia performance. One possible assumption is that there is a dissociation between the forms that involve the aspectual rule and those that do not. This is what we found in a previous study (Tsapkini, Jarema & Kehayia 2001) in which aphasie patients experienced more difficulties with the morphological categories that involve the aspectual rule (e.g., mil-o, mili-s-a; and lin-o, e-li-s-a) than with the elicitation of the stem-allomorphic form (e.g., plen-o, e-plyn-d). The opposite pattern, i.e. fewer difficulties with the aspectual-rule categories and more difficulties with the stem-allomorph-only category has also been documented as well (Kehayia 1988; Tsapkini, Jarema & Kehayia in press, a). Aside from the rule vs. no-rule dissociation, another possible distinction is conditioned by the presence of allomorphy. Therefore, in theory, verb categories that entail a stem-allomorph (e.g mil-o, mili-s-a; and plen-o, e-plyn-a) could dissociate from the category that does not involve stem-allomorphy (e.g., lin-o, e-li-s-a). The question specific to Greek is whether in the case of the category that requires both the perfective stemallomorph and the rule-based aspectual marker, it is the allomorphic change or the addition of the aspectual marker that bears the processing weight. In summary, the goal of the present experiment was to investigate the issue of morphological regularity in past-tense formation in the real-time performance of Greek-speaking aphasies. Our rationale was that Greek would allow us to isolate morphological effects from the confounding effects of form and semantic similarity. Moreover, we wanted to compare previous off-line results with present real-time performance of the aphasie patients tested for the same verb categories in order to distinguish between access and representational deficits in aphasia. The particular experimental parameters used are described in the next section.

320

3. Neurolinguistic

TSAPKINI, JAREMA & KEHAYIA

investigation

3.1 Patient profiles In the present experiment, we tested aphasic patients and age- and educa­ tion-matched control participants on a primed lexical decision task that ad­ dressed the issue of inflectional regularity during on-line processing of verbs. In addition to the on-line testing, the patients were tested on the Greek adapta­ tion of the Bilingual Aphasia Test (BAT) (Paradis 1987) for an assessment of their off-line language deficits. Below we provide a description of each case. Patient MH was a right-handed 56-year-old woman who suffered a Cerebral Vascular Accident (CVA) resulting from a blockage of blood flow of the left internal carotid. The CT-scan showed small hypodensity areas in the left parietal lobe. The MRI-scan confirmed the presence of an ischemic infarct with haemorrhagic signs at the left parietal lobe and the basal ganglia. The neurological diagnosis was CVA with motor aphasia and right hemiparesis. Right-hand motor difficulties were still present at the time of testing, 17 months post-morbidly. Her speech production was characterized by perseverations, repetitions, automatizations, and short sentences. She did not show any signs of visual or auditory deficits, or of anterograde or retrograde amnesia. She exhibited no agnosia or apraxia and she was well-oriented in time and space. Her short-term memory span was four digits. Her formal education ended with elementary school, but she was literate. The BAT examination revealed no deficits in word-picture matching, but did reveal impaired syntactic comprehension in sentence-picture matching. More specifically, she could not decode passive structures or pronouns. By contrast, she performed accurately most of the time on semantic category, synonymy, antithesis and morphological relatedness judgements. Patient IS was a right-handed 59-year-old man who had suffered a left CVA 24 months prior to testing. His CT-scan revealed a hypodensity area at the basal ganglia; the neurological diagnosis was right hemiplegia with motor aphasia. His speech was relatively fluent but dysarthric, and he still showed right upper-limb hemiparesis. Visual and auditory acuity, along with orientation in time and space were normal. He showed no signs of retrograde or antergorade amnesia, nor any signs of agnosia or apraxia. He had finished elementary school. His short-term memory span was three digits. The BAT examination revealed no deficits in syntactic comprehension of passive constructions or pronouns in sentence-picture matching. However, his semantic category, synonymity, antithesis and morphological relatedness judgements were impaired.

THE ROLE OF MORPHOLOGICAL STRUCTURE

321

3.2 Method 3.2.1 Participants and materials. We tested the two non-fluent aphasic patients MH and IS, along with 11 matched control subjects (60-year-old, ±5). All participants were monolingual native speakers of Greek and tested in Thessaloniki, Greece on a portable Macintosh Powerbook 1400 with active matrix screen. All of them had normal or corrected-to-normal vision. None of the control participants reported any past or present neurological illness. In this study we compared the following categories of verbs reflecting the different types of past-tense formation in the language: 16 verbs with stem-internal change, e.g., plen-o, e-plyn-a, "I wash, I washed", where the perfective stem plyn- is an allomorph of the imperfective stemplen-; (2) 32 verbs with phonological change, 16 of which feature a phonological alternation of /f/ to /p/, or /h/ to /k/ in the presence of the aspectual marker -s- e.g., graf-o, e-grap-s-a; "I write, I wrote", and 16 of which undergo a stem-final consonant deletion in the presence of the aspecutal marker -s-, e.g., lin-o, e-li-s-a, "I untie, I untied"; (3) 13 verbs which form their past tense by the addition of the aspectual marker -s- to their perfective allomorph, e.g., mil-o, mili-s-a, "I speak, I spoke".

(1)

related prime COMBINED mili-s-a KS REGULARS epex-a eli-s-a S REGULARS STEM-ALLOMORPHS eplyn-a

VERBS target mil-o pez-o lin-o plen-o

translation I speak I run I untie I wash

Table 1 : Examples of stimuli materials

The second category can be considered the equivalent of a rule-based para­ digm and the third category as the combination of rule-based and stored-allomorph paradigms. All verb categories were matched for frequency and length. All verbs were simple, i.e. not derived, and the frequency of the past tense was always lower than that of the present. Given that there are no frequency ratings for Greek, we obtained familiarity ratings. Forty adults aged 40-60, rated 250 Greek nouns and verbs on a scale of 1 to 5 on how often native speakers from their social background would use these words. Sixty-one disyllabic verbs of medium to high familiarity were selected. The familiarity of their past forms was lower than (but not extremely divergent from) that of

322

TSAPKINI, JAREMA & KEHAYIA

their present forms. The same number of nonwords were constructed by changing the first letter of the stem of the experimental stimuli. Nonwords were created from existing verbs without violating the phonotactic rules of the language. Thus, they were of the same length, inflectional type and complexity as the experimental words. 3.2.2 Design and procedures. Two priming conditions were used: either the past tense or an unrelated item primed the present first person singular form of each verb. The unrelated prime was the first person singular past tense form of another semantically, phonologically and morphologically unrelated verb. This form was matched for frequency and length with the experimental prime. Moreover, we purposely chose unrelated items that belonged to the same inflectional class as the matched prime so that all inflectional operations would be the same in the related and unrelated conditions, and would thus not influence response latencies. Each testing session was preceded by 20 practice trials. There were four testing sessions separated by one week each, so that every participant would see all the test items in all the priming conditions. No single item or its inflectional variant appeared more than once in the same testing session as either a prime or a target. Each testing session lasted 10 to 15 minutes for non-impaired participants, and 15 to 20 minutes for patients. There were 210 items in each session (50% words and 50% nonwords) preceded by 15 practice trials. We used a visual masked priming paradigm where the prime was presented for 150 ms following a 200ms mask with immediate presentation of the target (ISI=0 ms). The inter-trial interval was 1000 ms. Testing was carried out on a Macintosh Powerbook 1400 using the Psyscope presentation program. All participants performed lexical decision judgements using the index and middle fingers of their left hand. They were all instructed to respond to the second item that appeared on the screen as fast as possible and to decide whether they recognized it as a legitimate Greek word. All participants saw the full set of stimuli in all priming conditions and the presentation order was randomized individually by the computer program used. 3.3 Results Errors and extreme Response Times (RTS) over 4000ms were eliminated from the analysis. Each patient was analyzed separately and outliers (more than 2 Standard Deviations (SDs) of each verb category) were removed from the analysis. Errors constituted 2% of responses for the controls. Patient MH had a 13% error rate for words and 17% for nonwords, and patient IS had an error rate for words of 15% and 30% for nonwords.

THE ROLE OF MORPHOLOGICAL STRUCTURE

323

We performed two analyses of the data for both controls and patients: one on raw RTs for each priming condition of each morphological category to verify whether there was an effect of priming or not, and another on the magnitude of facilitation per morphological category, i.e., the difference scores between the two priming conditions, to verify whether there were significant differences in the facilitation elicited by each morphological category. We also report on error rate analyses with respect to the morphological categories and the priming conditions. For controls, we report both the subject and item analyses, while for patients we report only the item analyses. Given the small number of items, particularly in the allomorphic categories, and the fact that extreme RTs are not an unusual phenomenon for patients, we decided to perform non-parametric tests of significance for both the patients' and the controls' by-item analyses (including extreme RTs would skew the distribution of items per category while excluding them would result in loss of critical data points). We opted for non-parametric tests for the bysubject analyses as well, because the small number of age-matched control subjects (N=11) may have resulted in a non-canonical distribution of subjects' performance scores. Mean reaction times, error rates and number of observations are shown in Table 1 for both patients and controls for each verb category and priming condition.

COMBINED REGULARS KS REGULARS S REGULARS STEM-ALLOMORPHY

CONTROLS unrelated related

unrelated

related

Unrelated

787(79) 762(51) 743(93) 778(96) 768(34)

1581(687) 1555(494) 1803(589) 1526(523) 1528(308)

1096(311) 1649(428) 1432(394) 1805(371) 1220(236)

1607(413) 1468(414) 1421(361) 1523(430) 1542(323)

701(59) 730(57) 759(124) 707(73) 667(34)

MH

IS related 1460(359) 1437(350) 1383(265) 1436(422) 1230(108)

Table 2: Mean RTs and SDs in parenthesis for controls and patients

3.3.1 Response times and error rates. For controls, the analyses of RTs (Wilcoxon matched-pairs signed-ranks test) revealed significant priming effects for the allomorph-plus-aspectual marker category (e.g., mili-s-a, mil-o) for both the by-subject and by-item analyses (Zl=-2.845; p<.005 for by-subject and Z2=-2.844; p<.005 for the by-item analyses). However, in the rule-based category, with the addition of the aspectual-marker-only (e.g., e-li-s-a, lin-o), the difference between the two priming conditions was significant only in the by-item analyses (Zl=-1.467; p=.142 and Z2=-2.113; p<.04). Finally, there was a significant priming effect for the stem-allomorph-only category (e.g., e-plyn-a, plen-o) for both analyses (Zl=-2.578; p<.01 and Z2=-3.18; p<.002).

324

TSAPKINI, JAREMA & KEHAYIA

For patient MH, in the by-items analysis we performed Wilcoxon matchedpairs signed-ranks tests comparing the control and past-tense priming condi­ tions for each verb category. For the purposes of clarity, we will follow the same presentation order for both patients' and control subjects' results. Thus, for the allomorph-plus-rule category there was a significant priming effect (z=-2.4, p<.02). No priming effect was found for the most regular, rule-based category (z=-0.669, p=.5). Finally, there was a significant priming effect for the stem-allomorph-only category (z=-2.667, p<.008). The pattern was similar to that of normal subjects, thus strengthening the assumption that for some aphasic patients the residual capacities of lexical processing are intact, whereas their impaired performance is manifested only in production and is thus task-specific. The analysis of patient IS's data showed no priming effects for either the allomorph-plus-rule or regular-rule-only conditions (z=-0.866, p=.386; and z=-0.17, p=.98, respectively). However, the stem-allomorph-only category elicited a significant priming effect (z=-1.84, p=.06). An analysis of the patients' error rates was also performed in order to de­ tect any differential patterns in the categories involved. Neither patient demonstrated a significant effect of verb category on error percentages (Pearson x2=1.565 p=.45, for patient MH; and Pearson x2<1, for patient IS). 3.3.2 Difference scores. In an additional analysis we measured the magnitude of priming between the three verb categories in order to capture relative differences in facilitation or inhibition. We thus computed the difference scores between the control and the past-tense priming conditions and compared them for both the by-subject and by-item analyses. As for the control subjects, we performed a Friedman two-way ANOVA in the by-subjects analysis and a Kruskal-Wallis one-way ANOVA in the by-item analysis. Both analyses were significant (Fr(2)6.73; p<.04; K W ( 2 ) = 9 . 7 8 , p<.008; for the bysubject and by-item analyses respectively). Since the obtained values were both significant, we performed multiple comparisons between the verb categories to determine to which difference the effect can be attributed. The critical value for the planned comparisons in the by-subject analysis was z=1.02 and the differences between the regular rule-based category and the allomorph-plus-rule and stem-allomorph-only categories were close to significant, i.e., they just failed to exceed the critical difference (z=l and z=.09, respectively). In the by-items analysis, the difference between the regular rule-based category and the stem-allomorph-only category was significant (16.44>13.98, where the latter is the critical value for the comparison), and the difference between the regular rule-based category and

THE ROLE OF MORPHOLOGICAL STRUCTURE

325

the allomorph-plus-rule one just failed to reach significance (11.78<12.93, where the latter is the critical value for the comparison). Next we performed the Kruskall-Wallis test to verify whether there was any difference in the magnitude of priming between the three verb categories for patient MH and found it to be significant ( K W ( 2 ) = 6 . 3 , p<.05). We pro­ ceeded to identify the critical differences that yielded significance between (a) the regular-rule-based and the stem-allomorph-plus-rule categories, (b) the rule-based and the stem-allomorph-only categories, and (c) the two allomorphic conditions. The difference between the rule-based (e.g., e-li-s-a, lin-o) and the stem-allomorph-plus-rule-based (e.g., mili-s-a, mil-o) was marginally significant (10.97<11.73, where the first is the critical value for the compari­ son). With respect to the difference between the rule-based (e.g., e-li-s-a, lin-o) and the stem-allomorph-only categories (e.g., e-plyn-a, plen-o), it just missed significance (9.07< 11.92, where the first is the critical value for the compari­ son). Finally the difference between the two allomorphic conditions was far from being significant (1.9<13.43, the first being the critical value for the comparison). When we compared the priming difference scores between the morphological categories for patient IS, we did not find any significant differ­ ence in the magnitude of facilitation ( K W ( 2 ) = 2 . 7 6 , p=.25). It becomes evident that the regular rule-based category elicited significantly less priming than the other two categories. It should also be noted that the priming effect for this category did not reach significance in the by-subjects analysis. Moreover, this category is the most regular one in Greek and the analogous categories in other languages elicit more or at least the same amount of priming when compared to the irregular categories. What could be the source of this discrepancy in our data? Although all items fulfilled the same morphological criteria to be included in the same category, i.e., a single stem with a phonological change in the stem-final consonant triggered by the aspectual marker -s-, this marker was not visually salient for half of the items. In these items, the phoneme Is/ was not represented by the grapheme -s-, but instead was 'fused' with the phonologically changed consonant of the stem. Thus, the phonological strings /ps/ or Iksl in Greek are each graphemically represented by one letter only. Given that the task administered was visual recognition, we investigated the possibility that such a mismatch may affect processing. Our working hypothesis was that if the lack of correspondence between the graphemic and phonological codes does not play a role, the same results should be obtained when these items are excluded as when they are included in the data analysis.

326

TSAPKINI, JAREMA & KEHAYIA

3.3.3 Corrected response times and corrected error rates. We first extracted the items presenting a visual mismatch, i.e., items in which the aspectual marker -s- was not visually represented. We thus excluded 16 out of 32 items and proceeded with the same statistical analysis for the regular-rule-based category for both controls and patients. For controls, there was significant priming effects for both the by-subjects and the by-items analyses (Zl=-2.845, p<.005; Z2=-2.844, p<.005). For the other two categories there was no change, the results remaining as described above, i.e., both categories show significant priming effects in both analyses. Turning to patient MH, when only the items with the salient aspectual marker are included in the analysis, the difference between the control and the past-tense primes is significant (Z=-2.411, p<.02). However, the direction is opposite, i.e., the past-tense primes inhibit processing as compared to the control primes, as shown in RTs (Table 1). For patient IS the exclusion of the 16 mismatched items did not change the results, i.e., there was no significant difference between the control and the past-tense primes for the regular rulebased category (Z=-.459, p=.65). In the error rates analysis for past-tense primes, neither patient exhibited any significant effect of verb category (Pearson x 2 = l-19, p=.55, for patient MH; and Pearson x2 =1.81, =.4 for patient IS). 3.3.4 Corrected difference scores. We calculated the magnitude of facilitation for the three morphological categories after the removal of the orthographically opaque items for both controls and patients. For controls, there was no significant difference in either the by-subjects or the by-items analysis. Thus, the Friedman two-way ANOVA did not reach significance in the by-subjects analysis (Fr(2)=1.27, p=.53), and Kruskal-Wallis one-way ANOVA did not reach significance in the by-items analysis ( K W ( 2 ) = . 9 8 , p=.61). Therefore, the weak difference in the magnitude of facilitation was an artifact produced by the inclusion of the orthographically opaque items because, in the present analysis, the difference scores were almost identical for the three categories. For the patients we used the same statistical test that was employed for the item analysis of controls (Kruskal-Wallis one-way ANOVA). For patient MH there was a significant difference in the magnitude of facilitation between the three verb categories ( K W ( 2 ) = 1 3 . 1 6 , p<.002). We thus proceeded by com­ puting the critical values for each comparison between the two groups. The difference between the stem-allomorph-plus-rule category (e.g., mili-s-a, mil-o) and the rule-only category (e.g., e-li-s-a, lin-o) was significant (14.39>10.41, where the latter is the critical value for the comparison). Moreover, the

THE ROLE OF MORPHOLOGICAL STRUCTURE

327

difference between the stem-allomorph-only category (e.g., e-plyn-a, plen-o) and the rule-only category was also significant (13.04>10.65, where the latter is the critical value for the comparison). The difference between the two allomorphic categories did not reach significance (1.35<10.65, where the latter is again the critical value for this comparison). Thus, the inhibition found in previous comparisons for the rule-based category produced a significant difference in the amounts of priming between the three categories. For patient IS, there was no significant difference in the magnitude of priming between the three verb categories ( K W ( 2 ) = 2 . 1 7 4 , p=.33). 4. General discussion In this study we investigated the performance of two aphasic patients during the processing of different types of past-tense formation in Greek involving either aspectual marker affixation, stem-allomorphy, or a combina­ tion of these two processes. The language provided the opportunity to take some distance from the dichotomy of rule-based vs. stored-allomorph-based mechanisms postulated by dual-route models, because of the presence of a category that combines both mechanisms. Two aphasic patients and eleven control subjects performed a primed visual lexical decision task, and their reaction times and magnitude of priming were analyzed in order to gain insight into lexical processing of past tense. For controls, priming effects were obtained for all morphological categories. Moreover, we did not find more priming for the most regular rule-based category that involves the affixation of a single stem (e.g., lin-o, e-li-s-a). Patients MH and IS elicited different patterns of priming. In particular, patient MH showed priming effects for both allomorphic conditions (with and without the implementation of the rulebased aspectual marker -s-), but failed to show any priming effects for the most regular category, which was also found to be significantly different from the other two categories in the magnitude of facilitation comparisons. Patient IS, on the other hand, did not show any priming effects for either of the categories that require the computation of the rule-based aspectual marker, but he did show priming effect for the most irregular category, where past-tense formation involves the processing of only the stem allomorph without any rule-computation (e.g., plen-o, e-plyn-a). Furthermore, in the magnitude of facilitation comparisons, the same category elicited significantly more priming than the others. The dissociatation reported for patient MH indicates that on-line past-tense processing in aphasia may yield a pattern similar to that of non-brain-damaged subjects. This finding is in line with previous investigations of aphasia in Greek (Kehayia 1988, 1997), as well as with the

328

TSAPKINI, JAREMA & KEHAYIA

claim that the problem in the aphasic lexicon is impaired access or slowed activation (Prather, Zurif et al. 1992; Prather, Zurif et al. 1997). The other dissociation observed was between the rule-based and stored-allomorph categories (patient IS). Whenever the rule needed to be computed, performance was slowed down, suggesting that the rule-based mechanism was impaired as was also the case in the same patient's performance in off-line production (Tsapkini et al. 2001). This is a common finding for many aphasic subjects who exhibit difficulties in inflectional morphology. Thus, the commonly observed problem that aphasies have in the processing of rules and suffixes is manifested in this Greek-speaking patient in the production difficulties he demonstrated for both categories that involve rule-based affixation, irrespective of the presence of a stem-allomorph. 4.1 Morphological processing impairment: a phonological deficit? Since we found that the different morphological categories elicit different priming effects for these patients, the next challenge was to locate each patient's impairment within the lexical system. Morphological impairments have been documented repeatedly in the aphasie literature, but they are not always attributable to the same processes. For example, connectionist approaches have claimed that a deficit in irregular morphology can be attributed to an impairment at the semantic level, whereas a deficit in regular morphology can be attributed to an impairment at the phonological level (Joanisse & Sternberg 1999). However, the patients did not exhibit any comprehension deficits in the lexical semantics of past tense (see Tsapkini et al. 2001). It has also been shown that the morphological errors agrammatic patients make in production are independent from the errors they make in comprehension (Goodglass, Christiansen & Gallagher 1993). Furthermore, if their deficit was in accessing the semantic features of verbs, it should have affected all categories equally, which was not found to be the case. It has also been documented in a speeded lexical decision task that Broca's and Wernicke's aphasies can automatically access the semantic lexicon (Hagoort 1993). In that study, the author claims that difficulties experienced by Broca's aphasies to integrate lexical-semantic information in context can be attributed to the failure in either the morphological parsing of complex forms into stem and affix, or the on-line retrieval of the syntactic features of the inflectional suffix. This differentiation corresponds to Badecker's (1997) distinction between morphophonological and morphosyntactic deficits, respectively. However, for our patients, the locus of the deficits cannot be at (or at least only at) the morphosyntactic level, since they were tested on isolated words. This does not exclude the possiblitiy that their morphophonological deficit

THE ROLE OF MORPHOLOGICAL STRUCTURE

329

may parallel their morphosyntactic deficit. It has to be noted, however, that a deficit in inflectional morphology in aphasia is more often attributed to the morphosyntactic level (e.g., Badecker 1997; Luzzatti & De Bleser 1996; De Bleser & Luzzatti 1994). There are only a few reports of 'pure' morphological deficits, i.e., those that do involve morphological operations only at the level of single words (Nadeau & Rothi 1992; Luzzatti, Mondini & Semenza 1999). More specifically, Luzzatti et al. (1999) report a patient (MB) who could easily access the syntactic structure of complex words, but not their phonological forms/Thus, they concluded that the patient's deficit was not at the 'lemma' level, where morphosyntactic features are represented (see Levelt, Roelofs & Meyer 1999) but rather at the lexeme level, where morphophonological information is represented. As argued earlier, both our patients have a deficit at the morphophonological level manifested during word recognition, i.e., patient IS has difficulties in decomposing the aspectual marker suffix, and patient MH has difficulties retrieving the phonological form of regular stems that have undergone suffix-triggered phonological change. To fully interpret the normal and aphasic recognition of complex forms in Greek we have to resort to models of lexical access that minimally assume affix-stripping. 4.2 Morphological regularity in Greek and implications for models of lexical access Two basic processes take place during word recognition in dualmechanism models (Baayen et al. 1997; Frauenfelder & Schreuder 1992; Niemi et al. 1994; Caramazza et al. 1988), decomposition and morpheme retrieval. These models specify how the two processes operate and the factors that influence their involvement. Thus, it has been documented that factors such as word formation type (inflection vs. derivation), productivity, affixal homonymy, affix frequency and semantic transparency favor one route over the other during lexical processing (for further discussions, see Baayen, Burani & Schreuder 1997 for Italian; Bertram, Laine & Carvinen 1999 for Finnish; Bertram et al. 2000 for Dutch). In the present study the only factor at play was the presence or absence of the rule-based aspectual marker, since, as mentioned in the method section, we matched the familiarity ratings of the past and present-tense forms and chose only the verbs with present-tense forms rated higher than past-tense forms. Thus, whatever the role of the whole-word route before any affix-stripping may be in Greek, it influenced lexical access to the same extent for all past-tense morphological categories. The question, therefore, is how decomposition proceeds and whether there are other types of processes analogous to whole-word access but occuring at the stem-level in Greek. The models discussed above do not explicitly specify

330

TSAPKINI, JAREMA & KEHAYIA

how inflected forms are processed in a highly inflected language such as Greek, which features a category with both stem-allomorphy and stem-final phonological change in the presence of the aspectual suffix. If lexical organization and access is morpheme-based, as has been claimed for Finnish (Niemi et al. 1994; Laine et al. 1999), then stem-allomorphs would be instrumental during access. But what then is the role of phonological change at the stem-level, or of the lack of grapheme-to-phoneme correspondence? We will first focus on the non-impaired subjects' results, where the most regular, rule-based category with a phonological change triggered by the aspectual marker elicited less priming than both the stem-allomorph and the stem-allomorph-plus-rule category. However, when we excluded the cases where the aspectual suffix was not visually salient, there was no difference in priming amongst the three categories. Moreover, the RTS for the control and the past-tense priming conditions were almost identical (see Table 1). This finding shows that the stem-final phonological change, conditioned by the aspectual marker, does not affect processing. More cautiously stated, whatever the effect of phonological mismatch between the input and the morpheme entry in earlier stages, it is resolved at the SOA used (150 ms). On the other hand, what does influence processing is the lack of grapheme-tophoneme correspondence, or rather the lack of suffix saliency for the input modality involved (visual), a finding replicated in young participants as well (Tsapkini, Jarema & Kehayia in press). In the present study, the processing importance of nesting at the level of the stem was manifested when there was no priming effect for the verbs where the phoneme representing the aspectual marker (-s-) was 'fused' with the stem-boundary consonant into one grapheme, representing a double sound in the language {/ks/ or /ps/). Since the task was visual, the lack of aspectual marker specification at the visual input level resulted in slowed processing. This claim about the processing system also finds justification within a linguistic framework where, the difference between the two regular categories where only one stem is postulated (Ralli 1988) could be attributed to the different phonological rules involved. Thus, in the case of e-li-s-a, lin-o, there is a consonant deletion in the coda of the imperfective stem {lin-) where -n- is deleted in the presence of the aspectual marker -s-. On the other hand, in the case of e-pex-a, pez-o, there is a con­ sonant substitution of /z/ to ík/ in the presence of the aspectual marker -s-. Since the present experiment was conducted in the visual modality, the possible effects of processing the different phonological rules implicated in past-tense formation could not explicitly explored. We are currently investi-

THE ROLE OF MORPHOLOGICAL STRUCTURE

331

gating these effects in a cross-modal experiment that implies not only the visual but also the auditory modality. The question that arises is why, for patient MH, the pattern is diametrically opposed to that of controls when the two categories are separated: marginally significant priming for the orthography-phonology mismatched category and inhibition for the phonological change category with a salient aspectual marker. It must also be noted that when all verbs of the most 'regular' category were combined, there was no significant priming effect for that category. It could therefore simple be assumed that the phonological adjustment rule is impaired and patient MH cannot perform the phonological adjustment on-line. However, that is not easily justifiable since she can use the rule in the visually opaque cases. The only possible explanation would thus be to postulate a deficit in the orthographic input lexicon where the system must make an adjustment and retrieve the single stem, e.g., lin-, from the form it has in the past-tense li-. It is thus hypothesized that the inhibition results from a deficit in retrieving morphemes from orthographic variants at the level of the orthographic input lexicon. Additionally, patient MH is assumed to rely on the phonological lexicon which is not impaired, since she seems able to do the adjustment for the orthographically opaque aspectual marker. To recapitulate, we propose that when the patient is confronted with a form such as e-li-s-a, the system has no problems parsing -a or even -sa, but cannot easily exclude li- as a correct candidate (since it is a nested morpheme) or adjust it to the morpheme -lyn- in the orthographic input lexicon, and that results in some extra processing cost. On the other hand, when the patient is confronted with a form such as e-pex-a, the system after parsing the suffix -a and unable to find the morphem pex- in the orthographic lexicon, has to resort to the phonological lexicon to compute the post-phonological rule. By doing so, the system is informed of the presence of the aspectual marker -s- and has no difficulties making the morphological adjustment of the non-existing pekto the morpheme pez-, since there is no impairment at that level. The lack of impairment at the level of the phonological output lexicon is also evidenced by the off-line results for the same patient, who showed no deficit in producing the past tense of any morphological category of verbs. As for patient IS, as discussed above, his deficit is in the affix-stripping and morphological decomposition process. He had no problems with irregular verbs, which shows that his stem-allomorph retrieval mechanisms were not impaired. Given his parallel performance in off-line tasks (Tsapkini et al. in press), one could also claim that decomposition was impaired not only at the orthographic input level, but also at the phonological output level.

332

TSAPKINI, JAREMA & KEHAYIA

5.

Conclusions This study allowed us to investigate morphological processing in the performance of Greek-speaking aphasic patients in a primed visual lexical decision task. In relation to the deficits revealed in patients MH and IS and the theoretical assumptions involved, the fundamental issue addressed was whether aphasic deficits are generalized or task-specific, i.e., whether they involve impaired access as manifested in real-time tasks or impaired representations as manifested in off-line tasks. Our results show that for patient IS, access is conditioned by the efficiency of decomposition which translates into processing of the aspectual rule. For patient MH, it is conditioned by the orthographic transparency of the available form relative to the morphemic stem. For patient IS, who demonstrated the same deficit in his on-line and off-line performance, we can conclude that his impairment is not task-specific. He thus showed difficulties not only in morphological decom­ position during word recognition, but a more generalized problem in processing rules, manifested both as an affix-stripping deficit in lexical decision and as an impairment of regular morphology in oral production. In the case of patient MH, the deficit is task-specific, since she showed no difficulties in off-line tasks, and is a manifestation of an impairment in the orthographic input lexicon. REFERENCES Allen, Mark & William Badecker. 1999. "Stem homograph inhibition and stem allomorphy: Representing and processing inflected froms in a multilevel lexical system". Brain and Language 41.105-123 Baayen, Harald. 1994. "Productivity in language production". Language and Cognitive Processes 9.447-469. Baayen, Harald, Cristina Burani & Robert Schreuder. 1997. "Effects of semantic markedness in the processing of regular nominal singulars and plurals in Italian". Yearbook of Morphology 1996 ed. by Gert Booij & Jaap van der Marie, 13-33. Dordrecht: Kluwer. Baayen, Harald, Ton Dijkstra & Robert Schreuder. 1997. "Singulars and plurals in Dutch: Evidence for a parallel dual-route model". Journal of Memory and Language 37.94-117. Baayen, Harald & Robert Schreuder. 1999. "War and peace: morphemes and full forms in a noninteractive activation parallel dual-route model". Brain and Language 68:1/2.27-32. Badecker, William, Argye Hillis & Alfonso Caramazza. 1990. "Lexical morphology and its role in lexical process: Evidence from a case of acquired dysgraphia". Cognition 35:3.205-243.

THE ROLE OF MORPHOLOGICAL STRUCTURE

333

Bertram, Raymond, Matti Laine & Katja Karvinen. 1999. "The interplay of word formation type, affixal homonymy, and productivity in lexical processing: Evidence from a morphologically rich language". Journal of Psycholinguistic Research 28:3.213-226. Bertram, Raymond, Robert Schreuder & Harald Baayen. 2000. "The balance of storage and computation in morphological processing: The role of word formation type, affixal homonymy, and productivity". Journal of Experimental Psychology: Learning, Memory and Cognition 26:2.489-511. Bastiaanse, Roelien & Gerald Bol. 2001. "Verb inflection and verb diversity in three populations: agrammatic speakers, normally developing children, and children with specific language impairment (SLI)". Brain and Language 77:3.274-82. Caramazza, Alfonso, Alessandro Laudanna & Cristina Romani. 1988. "Lexical access and inflectional morphology". Cognition 28.297-332. Clahsen, Harold. 1999. "Lexical entries and rules of language: A multidisciplinary study of German inflection". Behavioral and Brain Sciences 22.991-1060. De Bleser, Ria & Claudio Luzzatti. 1994. "Morphological processing in Italian agrammatic speakers' syntactic implementation of inflectional morphology". Brain and Language 46:1.21 -40. Frauenfelder, Uli H. & Robert Schreuder. 1992. "Constraining psycholinguistic models of morphological processing andrepresentation: The role of productivity". Yearbook of Morphology 1991 ed. by Gert Booij & Jaap van Marie, 165-183. Dordrecht: Kluwer. Hagoort, Peter. 1993. "Impairments of lexical-semantic processing in aphasia: Evidence from processing of lexical ambiguities". Brain and Language 45:2.189-232. Goodglass, Harold, Julie Ann Christiansen & Roberta Gallagher. 1993. "Comparison of morphology and syntax in free narrative and structured tests: Fluent vs. nonfluent aphasics". Cortex 29:3.377-407. Grodzinsky, Yosef. 1986. "Language deficits and the theory of syntax". Brain and Language 37.480-499. Jarema, Gonia. 1998. "The breakdown of morphology in aphasia". Handbook of Neurolinguistics ed. by Brigitte Stemmer & Harry A. Whitaker, 221-234. San Diego, Calif: Academic Press. Joanisse, Mark & Mark Seidenberg. 1999. "Impairments in verb morphology after brain injury: A connectionist model". Proceedings of the National Academy of Science of the United States 96.7592-7597. Kehayia, Eva, Gonia Jarema & Danuta Kadzielawa. 1990. "Cross-linguistic study of morphological errors in aphasia: Evidence from English, Greek, and Polish". Morphology, Phonology and Aphasia ed. by Jean-Luc Nespoulous & Pierre Villiard. 140-155. New York: Springer. Kehayia, Eva. 1997. "Morphological representation of inflected and derived words in Greek speaking individuals". Proceedings from the Second International Con­ ference on Greek linguistics, October 1997.

334

Kehayia, Eva. 1988. Morphological Impairments in Agrammatic Aphasia: A comparative study. Ph.D. dissertation, McGill University, Montreal. Laine, Matti, Seppo Vainio & Jukka Hyönä. 1999. "Lexical access routes to nouns in a morphologically rich language". Journal of Memory and Language 40.109-135. Levelt, Willem, Ardi Roelofs & Antje S. Meyer. S. 1999. "A theory of lexical access in speech production". Behavioral and Brain Sciences 22.1-75. Luzzatti, Claudio & Ria De Bleser. 1996. "Morphological processing in Italian agrammatic speakers: Eight experiments in lexical morphology". Brain and Language 54.26-74. Luzzatti, Claudio, Sara Mondini & Carlo Semenza 1999. "Storage or parsing of morphologically complex words: A case study in agrammatism". Brain and Cognition 40.183-195 Marcus, Gary F., Steven Pinker, Michael Ullman, Michelle Hollander, T. John Rosen & Fei Xu. 1992. Overregularization in Language Acquisition. (Monographs of the Society for Research in Child Development 57.4). Chicago: University of Chicago Press. Nadeau, Stephen E. & Leslie J. Gonzalez Rothi. 1992. "Morphologic agrammatism following a right hemisphere stroke in a dextral patient". Brain and Language 43:4.642-667. Niemi, Jussi, Matti Laine & Juhani Tuominen. 1994. "Cognitive morphology in Finnish: Foundations of a new model". Language and Cognitive Processes 9.423-446. Orsolini, Margherita, Rachele Fanari & Hugo Bowles. 1998. "Acquiring regular and irregular inflection in a language with verb clases". Language and Cognitive Processes 13:4.425-464. Paradis, Michel. 1987. The Assessment of Bilingual Aphasia. Hillsdale, N.J.: Lawrence Erlbaum Associates. Pinker, Steven. 1999. Words and Rules: The ingredients of language. New York: Basic Books. Plunkett Kim & Virginia Marchman. 1993. "From rote learning to system building: Acquiring verb morphology in children and connectionist nets". Cognition 48:1.21-69. Prather, Penny, Edgar Zurif, Catherine Stern & T. John Rosen. 1992. "Slowed lexical access in nonfluent aphasia: A case study". Brain and Language 43.336-348. Prather, Penny, Edgar Zurif, Tracy Love & Hiram Brownell. 1997. "Speed of lexical activation in nonfluent Broca's aphasia and fluent Wernicke's aphasia". Brain and Language 59:3.391 -411. Ralli, Ageliki. 1988. Eléments de la morphologie du grec moderne: La structure du verbe. Ph.D. dissertation, Montreal: Université de Montréal. Say, Tessa & Harald Clahsen. 1999. "Words, rules and stems in the Italian mental lexicon". Essex Research Reports in Linguistics, December 1999, 1-47. Taft, Marcus. 1988. "A morphological-decomposition model of lexical representa­ tion". Linguistics 26.657-667.

THE ROLE OF MORPHOLOGICAL STRUCTURE

335

Taft, Marcus. 1994. "Interactive-activation as a framework for understanding mor­ phological processing". Language and Cognitive Processes 9.271-294. Taft, Markus & Kenneth Forster. 1975. "Lexical storage and retrieval of prefixed words". Journal of Verbal Learning and Verbal Behavior 14.638-647. Tsapkini, Kyrana, Gonia Jarema & Eva Kehayia. 2001. "Manifestations of morpho­ logical impairments in Greek aphasia: A case study". Journal of Neuro Unguis tics 14.281-296. Also in Manifestations of Aphasic Symptoms in Different Languages ed. by Michel Paradis, 197-212. Oxford & New York: Pergamon Press. Tsapkini, Kyrana, Gonia Jarema & Eva Kehayia. In press, a. "A morphological processing deficit in verbs but not in nouns: A case study in a highly inflected language". Journal of Neurolinguistics (special issue on noun and verb processing ed. by Kenneth Shapiro & Alfonso Caramazza). Tsapkini, Kyrana, Gonia Jarema & Eva Kehayia. In press, b. "Regularity revisited: Evidence from lexical access of verbs and nouns in Greek". Brain and Language. (special issue on the Mental Lexicon ed. by Gonia Jarema, G. Libben & Eva Kehayia). Ullman, Michael T., Suzanne Corkin, Marie Coppola, Gregory Hickok, John H. Growdon, Walter J. Koroshetz & Steven Pinker. 1997. "A neural dissociation within language: Evidence that the mental dictionary is part of declarative memory, and that grammatical rules are processed by the procedural system". Journal of Cognitive Neuroscience 9:2.266-276.

INDEX OF SUBJECTS A. A-bar movement; 133; 134; 144; 148; 149 accusative; 141; 192; 301; 302; 308; 309; 310; 311; 312 action; 281; 283; 286; 294 active; 182; 184; 185; 186; 260; 262; 270; 283; 286; 289; 291; 293;318 advancing; 80; 86; 87; 88; 90; 95 affix; 327; 328; 329; 331; 332 African languages; 168 agent, agentive; 180; 192; 283; 291 agglutinative; 155; 157; 162; 167; 168; 177 agnosia; 320 agrammatism, agrammatic; 119; 126; 127; 128; 189; 191; 263; 264; 265; 267; 268; 270; 271; 272; 273; 274; 275; 276; 282; 283; 299; 300; 304; 308; 311; 312 Agree; 272; 273; 274; 275; 276 agreement; 52; 63; 72; 146; 155; 156; 157; 158; 161; 168; 169; 177; 187; 192; 249; 253; 267; 268; 271; 272; 281; 306 agreement phrase (AGRP); 269 allomorph, allomorphy; 56; 57; 316; 317; 318; 319; 321; 323; 324; 325; 327; 328; 330; 331 American Signe Language (ASL); 223; 225 anomic aphasia; 252 anterior; 315 aphasia, aphasic, aphasiology; 23; 24; 25; 26; 28; 29; 31; 32; 34; 35; 36; 39; 40; 41; 148; 168; 230;

249; 251; 252; 254; 259; 261; 263; 264; 265; 267; 270; 279; 281; 283; 286; 290; 291; 315; 317; 318; 319; 320; 324; 327; 328;329; 332 appropriate, inappropriate; 79; 208; 233; 236; 238; 240; 262; 263; 293 apraxia; 28; 320 argument; 163; 185; 249; 257; 258; 263; 280; 293; 300; 302 aspect, aspectual; 131; 158; 159; 177; 269; 274; 318; 319; 321; 323; 325; 326; 327; 329; 330 auditory; 96; 216 auxiliary; 54; 56; 61; 126; 127; 253 B. Bantu; 168 basal ganglia; 320 behavior; 5; 27; 71; 210 benefactive; 158; 180 bound; 50; 52; 53; 55; 58; 60; 72; 318 Broca; 25; 121; 122; 189; 250; 264; 281; 282; 283; 284; 304; 305; 315; 328 . case; 146; 177; 178; 185; 188; 191; 258; 263; 267; 271; 272; 275; 300; 301; 302; 303; 304; 305; 306; 307; 308; 310; 311; 312 Catalan; 268 causative; 63; 158; 180; 182; 184; 185; 186; 192; 250 cerebral haemorrhage; 251

338

INDEX OF SUBJECTS

Cerebral Vascular Accident (CVA); 252 checking; 146; 148; 149; 263 chronological age; 181; 183; 187; 234;238 clitic, enclitic, proclitic; 51; 54; 56; 61; 62; 268 cognition, cognitive; 5; 32; 73; 210; 215; 229; 230; 231; 232; 238; 239;240 competence; 56; 63; 64; 68; 70; 72; 73; 86; 209; 210; 230; 231 computational; 146; 149; 219 conduction aphasia; 25 consonant; 17; 89; 96; 178; 193; 325;330 conversation, conversational; 233; 237; 238; 239; 253 conversational implicature; 197; 198;201;202;204;210 cooperative principle; 201 ; 204; 208 coordination; 197; 198; 199; 200; 202; 203; 208; 209; 239 cortical area; 264 cross-linguistic; 134; 155; 169; 170; 175 D. dative; 184; 185; 186; 301; 302; 308;310;311;312 deafness; 16; 49; 50; 51; 52; 53; 54; 56; 57; 66; 67; 73 deficit; 25; 32; 49; 51; 149; 176; 189; 209; 210; 214; 215; 216; 221; 224; 240; 249; 251; 263; 264; 274; 275; 293; 296; 317; 318; 319; 328; 329; 331; 332 representational deficit; 132; 215; 225;319

Delayed Auditory Feedback (DAF); 96 determiner; 249; 275; 299; 300; 301; 304; 305; 306; 307; 308; 310; 312 discourse; 192; 237; 276 disjunct; 198; 202; 208; 250 domain specificity; 213; 219; 221; 222; 229 dominant hemisphere; 264 Dutch; 27; 28; 31; 45; 119; 120; 121; 122; 125; 126; 127; 128; 129; 130; 253; 259; 269; 270; 276; 280; 281; 282; 290; 294; 295; 296; 297; 298; 300; 304; 313;329;332 dysarthria; 28; 289; 320 E. embedded; 124; 128; 238; 249; 250; 263 empty; 121; 137; 144 English; 3; 6; 9; 31; 43; 50; 52; 54; 74; 81; 82; 86; 93; 114; 125; 128; 133; 134; 150; 151; 155; 156; 158; 159; 167; 169; 176; 177; 178; 179; 190; 191; 192; 251; 252; 259; 266; 268; 272; 275; 280; 282; 283; 290; 291; 292; 300; 313; 315; 316; 317; 333 error; 11; 23; 24; 31; 39; 51; 52; 53; 54; 66; 67; 68; 69; 71; 77; 78; 79; 81; 90; 138; 160; 184; 186; 187; 189; 205; 249; 250; 251; 252; 253; 254; 255; 257; 258; 260; 262; 263; 264; 265; 268; 269; 287; 291; 292; 322 Esperanto; 3

INDEX OF SUBJECTS

F. failure; 75; 198; 209 faithfulness constraints; 33 feature; 5; 55; 131; 146; 188; 193; 219; 272; 274; 328 -features; 272; 274 non-interpretable; 132; 146; 149 feedback monitor; 96 finite, non-finite; 54; 119; 121; 122; 127; 177; 188; 191; 192; 193; 255; 268; 272; 281; 282; 288; 294; 296; 299; 300; 301; 302; 304; 306; 309; 311 Finnish; 3; 20; 21; 251; 318; 329; 334 fluency failure; 75; 78; 84; 86; 88; 91; 92; 96 fluent, non-fluent; 23; 25; 26; 27; 28; 29; 31; 32; 41; 75; 90; 95; 96; 249; 250; 251; 252; 254; 256; 257; 258; 259; 260; 263; 264; 265; 267; 283; 284; 291; 294; 318;320 French; 150; 151; 169; 179; 268; 280 Frisian; 128 frontal cortex, lesion; 264; 315 function; 34; 54; 55; 58; 61; 81; 169; 170; 180; 299 G. gender; 131; 177; 269; 276; 304; 307; 308; 311 genetic; 176; 229; 230 German; 89; 93; 130; 148; 150; 169; 176; 177; 190; 273; 277; 280; 300; 302; 304; 305; 306; 308; 311; 312; 313; 315; 333 Germanic languages; 271

339

grammar; 33; 35; 127; 131; 132; 149; 217; 230; 238; 239; 249; 262; 263; 265 generative grammar; 267; 268 grammatical, grammaticality; 6; 49; 52; 53; 54; 64; 67; 70; 146; 147; 148; 167; 176; 177; 178; 187; 188; 189; 203; 208; 216; 223; 224; 229; 234; 249; 250; 251; 258; 261; 262; 263; 264; 270; 271; 276; 280; 282; 283; 284; 285; 288; 292; 293; 299 grammatical relations; 178 grammatical role; 283; 292 Greek; 131; 133; 134; 135; 139; 141; 142; 146; 148; 149; 150; 151; 152; 153; 316; 317; 318; 319; 320; 321; 322; 325; 327; 329; 332; 333; 335 Gricean; 198 H. head; 55; 156; 157; 158; 270; 271; 300 hearing; 50; 51; 53; 57; 223; 233 Hebrew; 3; 9; 15; 16; 21; 22; 169; 176; 178; 190; 268; 269; 273 I. impaired, impairment; 25; 36; 40; 122; 129; 140; 147; 155; 159; 160; 167; 170; 198; 207; 208; 209; 210; 213; 214; 215; 216; 217; 218; 224; 230; 240; 262; 267; 269; 270; 271; 272; 274; 275; 276; 296; 299; 304; 305; 315; 320; 328; 331; 332 Indo-European languages; 176 infelicitous; 208

340

INDEX OF SUBJECTS

L. L1; 49; 52; 57 L2; 49; 52; 57; 148 Late Emerging Consonant (LEC); 89 left hemisphere; 28; 291; 294; 305 left parietal lobe; 320 lexical, lexicon, lexicalization; 25; 35; 51; 58; 64; 72; 73; 81; 85; 122; 145; 184; 186; 187; 189; 230; 249; 250; 251; 252; 254; 255; 257; 258; 259; 260; 263; 264; 271; 274; 283; 287; 302; 303; 317; 318; 319; 324; 327; 328; 331; 332 linguistic; 198; 200; 205; 207; 209; 210; 211; 213; 214; 215; 216; 217; 218; 229; 238; 240; 267 non-linguistic; 198; 210; 213; 214; 216; 217; 218; 221;224 lip-reading; 52 Lingua Italiana dei Segni (LIS); 49; 52; 53; 55; 61; 64; 67; 68; 70; 72 literal paraphasias; 23; 28; 29 J. locative; 180; 191; 192 Japanese; 3; 22; 175; 176; 177; 178; infinite, infinitive; 54; 58; 127; 177; 193; 250; 253; 268; 281; 288; 294; 296; 306; 311; 312 inflect, inflection, inflectional; 131; 158; 169; 176; 177; 178; 184; 186; 187; 188; 215; 249; 250; 252; 253; 254; 255; 256; 260; 261; 263; 264; 267; 268; 269; 271; 272; 274; 275; 281; 282; 304; 305; 306; 315; 317; 318; 322;328; 329 innate, innateness; 35; 213; 214; 215; 217; 218; 220; 221; 222; 224; 225 interference; 52; 61; 64; 65 intransitive; 192; 287 Italian; 3; 9; 19; 31; 49; 50; 51; 52; 53; 54; 57; 61; 65; 66; 67; 72; 74; 150; 151; 167; 176; 178; 269; 270; 276; 277; 297; 313; 315; 316; 329; 332; 333; 334

179; 180; 181; 188; 189; 190; 191; 192; 269 Japanese SLI; 181; 182; 187; 189

K. Kivunjo; 167 Klingon; 3 knowledge; 197; 198; 207; 208; 209; 210; 211; 220; 221; 231; 260 pragmatic knowledge; 197; 198; 207;209;210; 211 semantic knowledge; 197; 208; 270 syntactic knowledge; 144; 197; 198; 203; 207; 209

M. marked, markedness; 23; 24; 26; 28; 31; 32; 33; 35; 39; 40; 41; 137; 144; 161; 166; 176; 187; 300 marker; 158; 159; 164; 168; 169; 187; 192; 319; 327; 329 matrix; 123; 124; 127; 128; 281 maxim of manner; 198; 204; 205; 208 maxim of quality; 202 maxim of quantity; 197; 202; 203 mean length of utterance (MLU); 121; 160; 161; 188; 294

INDEX OF SUBJECTS

morphemes (MLU-); 184; 185; 187;188 Merge; 144; 145; 263; 272; 273 Mewati; 3 minimalism; 146 263; 267; 271; 272; 276 modal, modality; 58; 184; 186; 216; 294 module, modularity; 198; 210; 213; 214; 215; 216; 217; 218; 219; 220; 221; 222; 223; 224; 225; 229; 230; 240 monitor, monitoring; 95; 96 mood; 192; 274 morpheme; 6; 7; 51; 63; 169; 183; 192; 317; 318; 329; 330; 331 morphology, morphological; 27; 49; 50; 51; 52; 53; 54; 55; 56; 60; 63; 64; 65; 72; 73; 122; 127; 131; 136; 141; 148; 155; 156; 160; 161; 163; 164; 165; 166; 167; 168; 169; 170; 177; 178; 188; 189; 191; 214; 230; 231; 267; 268; 269; 303; 316; 317; 318; 319; 320; 322; 323; 326; 328; 329; 331 morphophonological; 33; 318; 328; 329 morphosyntax, morphosyntactic; 51 ; 122; 131; 191; 232; 139; 140; 328; 329 motor aphasia; 320 Move, movement; 120; 121; 127; 132; 133; 145; 146; 148; 149; 258; 261; 267; 270; 271; 272; 273; 274; 275; 276; 292; 293 Must Move principle; 132

341

N. negation, negative; 65; 158; 177; 192; 249 Nguni; 155 nominative; 192; 301; 302; 308; 310;312 noun; 51; 53; 63; 65; 155; 156; 157; 158; 163; 167; 168; 169; 185; 191; 193; 254; 259; 281; 284; 294; 295; 321 noun phrase (NP); 54; 67; 72; 142; 144; 184; 250; 256; 257; 260; 261; 271; 280; 283; 290; 292; 293; 300; 301; 304; 306; 307; 308; 310; 311; 312 number; 131; 158; 163; 169; 177; 269; 304 O. object; 54; 120; 128; 165; 283; 286; 310 omission; 52; 54; 56; 57; 58; 61; 65; 72; 268 Optimality Theory (); 23; 33; 34; 35; 36; 41 optional; 119; 121; 155; 156; 162; 168; 169; 177; 178; 180 Optional Extended Infinitive (OEl); 155;177; 187 oral dyspraxia; 289 orthography, orthographic; 66; 331; 332 P. paragrammatic, paragrammatism; 250; 251; 258; 263; 284 paraphasias, paraphasic; 23; 26; 32; 40; 250; 291; 294; 295; 296

342

INDEX OF SUBJECTS

passive; 156; 158; 164; 165; 180; 182; 184; 185; 186; 187; 192; 231; 238; 259; 260; 262; 270; 283; 284; 286; 289; 291; 295; 320 past; 176; 192; 253; 315; 316; 317; 318; 319; 321; 322; 324; 326; 327 perfective; 316; 318; 319; 321 performance; 51; 56; 137; 141; 147; 181; 198; 207; 210; 214; 216; 220; 233; 237; 238; 239; 240; 251; 257; 259; 260; 261; 262 263; 264; 283; 289; 290; 292; 304; 310; 317; 318; 319; 323; 324; 327; 328; 332 peripheral; 26 person; 60; 131; 165; 177; 253; 269 phoneme, phonemic; 5; 7; 86; 295; 296;325 phonetic, pre-phonetic; 4; 26; 32; 40; 41; 54; 56; 300; 303 phonology, phonological; 3; 5; 6; 7; 11; 12; 23; 24; 25 26; 29; 32; 33; 34; 51; 54; 56; 82; 85; 86; 87; 88; 89; 160; 178; 193; 214; 215; 230; 250; 267; 268; 269; 303; 315; 316; 318; 321; 322; 325; 328; 329; 330; 331 phonotactic; 7; 39; 40; 322 planning; 81; 86; 231; 254 plural; 51; 53; 60; 65; 155; 158 Polish; 151; 169; 333 posterior; 251; 264; 315 pragmatic; 51; 63; 73; 132; 197; 198; 203; 205; 207; 208; 209; 210; 211; 215; 230; 231; 240; 258;261 preposition, postposition; 54; 56; 57; 61; 65; 72; 178; 192; 249; 258; 264; 275; 299

principle of manner; 205 of quality; 201 of quantity; 197; 201; 209 process, processing; 11; 14; 17; 26; 35; 73; 78; 147; 148; 176; 177; 188; 220; 221; 249; 251; 264 progressive; 187; 192 pronoun, pronominal; 51; 54; 157; 170; 249; 258; 264; 280; 299; 300; 320 Q. questions; 249; 257; 261; 263; 284; 290; 293 R. regular, irregular; 231; 315; 316; 317; 319; 325; 326; 327; 329; 332 relative; 51; 62; 133; 134; 135; 136; 137; 149; 156; 166; 167; 168; 191; 238; 284 repair; 78; 80; 81; 90 retrieval, retrieve; 122; 189; 231; 249; 259; 260; 261; 280; 283; 288; 281; 305; 311; 329 reversible; 231; 238; 260; 293 right-handed; 123 right-hemisphere; 290 Romance; 82; 275 rule, rule-; 31; 155; 217; 251; 315; 318; 319; 321; 322; 324; 325; 327;328; 329; 330; 331 Russian; 16; 169 S. segmental; 32; 40; 41 semantic; 33; 49; 73; 131; 170; 178; 180; 192; 197; 198; 200; 208;

INDEX OF SUBJECTS

209; 210; 215; 231; 236; 238 250; 255; 258; 260; 262; 263 294; 295; 296; 302; 315; 320: 322; 328; 329 SeSotho; 155; 156; 159; 164 Setswana; 156; 172 sign language; 49; 54; 58; 67 singular; 158; 176; 253 Siswati; 156; 171 Spanish; 3; 20; 83; 268 Specific Language Impairment (SLI); 119; 121 123; 126 127; 131 132; 133 134; 138 139; 140: 144; 145 146; 147 148; 149 155; 169 175; 177 178; 181 183; 184 186; 187 188; 189: 213; 214 215; 216 217; 221: 223; 224 225; 230 231; 232: 233; 237; 238; 239; 240 speech; 28; 78; 90; 140; 213; 216 232; 252; 255; 259; 267; 289:

291; 293; 294; 308 spontaneous; 252; 255; 259; 267 293; 294; 308 stalling; 80; 87; 88; 90 Standard Deviations (SD); 181 234; 237 stem; 192; 253; 254; 255; 316; 317; 318; 319; 321; 322; 323; 324; 325; 327; 328; 330; 331 stutter, stuttering; 75; 80; 81; 82; 86; 88; 90; 92; 95; 96; 160 subject; 54; 63; 120; 157; 164; 168; 169; 182; 165; 265; 283; 286 subordinate; 51; 191; 236; 239; 259 substitution; 52; 60; 61; 250; 268 Swedish; 176; 190 syllable; 23; 31; 32; 32; 40; 41; 176; 178; 188

syntax, 68; 140; 178; 204; 214; 230; 251; 264; 284;

343

syntactic; 33; 51; 65; 66 73; 122; 127; 131; 134 144; 146; 148; 149; 167 182; 191; 197; 198; 203 205; 207; 208; 209; 210 217; 218; 221; 222; 224 231; 236; 238; 249; 250: 255; 258; 260; 262; 263 265; 267; 269; 274; 283 300; 302; 303; 305; 312

temporo-parietal region; 251 tense; 131; 146; 155; 157; 158; 159: 162; 167; 169; 177; 178; 184 186; 187; 192; 249; 257; 269: 274;281; 299; 306 test; 181; 197; 206; 208; 211; 279; 283; 296 Aachen Aphasia Battery Test (AAT); 122; 252; 279; 305 Bilingual Aphasia Test (BAT); 320 Boston Diagnostic Aphasia Examination Test; 252; 279 CELF-R Test; 205 Diagnostic Verbal IQ Test (DVIQ); 135; 139 Kruskall-Wallis Test; 325; 326 ITPATest; 181 K-ABCTest; 181 Picture Vocabulary Test (PVT); 181 Psycholinguistic Assessment of Linguistic Processing in Aphasia (PALPA); 279 Ravens Coloured Matrices Test; 233

344

INDEX OF SUBJECTS

Reduced Syntactic Disturbance Test; 251 Syntax Test of Aphasia for Children; 182; 184 Token Test for Children (TTFC); 197;205; 206 TOLD-l Test; 198; 206 VAST Test; 252; 255; 284; 289; 296 Wilcoxon Test; 27; 323; 324 wIPPSl Test; 181 theta role, thematic; 57; 73; 141; 252; 257; 258; 260; 261; 263; 280; 282; 283; 284; 285; 286; 287; 288; 292; 293; 302 tip-of-the tongue (); 77; 86 trace; 270; 272 Trace Deletion Hypothesis (TDH); 270; 272; 276 transitivity, transitive; 157; 180; 192; 281; 282; 287; 288; 294 Tsonga; 155 Type Token Ratio; 291 U. universal; 29; 33; 35; 39; 170; 220; 316;317 Universal Grammar (UG); 54; 55; 72; 175; 217; 221; 222; 223; 224 Urdu; 3 utterance; 147; 170; 191; 200; 203; 209; 238; 284; 294 V. Venda; 155 verb; 53; 54; 63; 120; 122; 158; 166; 170; 178; 184; 186; 188; 191; 192; 249; 250; 251; 252; 253;

254; 255; 257; 259; 263; 274; 279; 280; 281; 282; 287; 293; 294; 295; 296; 300; 301; 302; 304; 305; 324; 331 verb phrase (vp); 67; 72; 257;260; 301 voice; 177; 184; 186; 192 vowel; 51; 55; 56; 96; 178

269; 284; 299; 321; 126;

W. Wernicke; 25; 250; 252; 253; 256; 260; 263; 283; 328 wh - (movement, question, word); 133; 134; 140; 141; 149; 259; 270; 275; 290; 292; 293 Williams Syndrome (ws); 230; 231; 232; 233; 234; 235; 236; 237; 238; 239; 240 word; 7; 8; 51; 52; 53; 81 89; 148; 189; 315; 317; 318; 329 class word (closed, open); 258; 299 content word; 81; 82; 83; 84; 86; 87; 88; 89; 90; 91; 287; 299 function word; 53; 81; 82; 83; 84; 87; 88; 91; 178; 260; 267; 299 phonological word; 56; 86; 87; 88;89;268;315 word order; 164; 178; 188; 191; 192; 258; 275; 282; 283; 286; 292 sov;120; 182; 191; 192; 281 svo; 144; 147; 148; 149; 156; 157;305

Zulu; 155; 156; 158; 159; 160; 161; 163; 164; 167; 168; 169; 170; 171; 172; 174

LIST OF CONTRIBUTORS Roberto Ajello is a Professor of Applied Linguistics and Chair of the Department of Linguistics of the University of Pisa and also teaches Linguistics at the Faculty of Medicine at the same University. His main research interests include: Armenian, Iranian, Somali and Chadic linguistics, Italian Sign Language and other communicative systems of the deaf. He has published over 60 articles on these topics and collaborated on the edition of the first Somali-Italian dictionary. Besides his interest in languages that have no written form, he is investigating the communicative systems of the deaf other than sign languages or labialization, namely, the so-called 'mouth gestures' of the deaf, which are based on proprioceptive mechanisms. James Au-Yeung is Wellcome Research Fellow at the Department of Psychology, University College London, where he has been the senior full-time staff member of the Speech Research Group for ten years. He joined the group after completing a Ph.D. in Computational Linguistics at Manchester University. He set up the UCL bilingualism survey which is one of four developmental psychology tests listed in the American Psychological Society of Psychological Research (http://psych.hanover.edu/APS/exponnet.html) main­ tained by John Krantz, whose work is described in the recent book on internet research by M. H. Birnbaum (New York, 2000). He is currently developing multilingual tests of syntax development for use with children from the age of two upwards. Roelien Bastiaanse has been Full Professor at the Department of Linguistics of the University of Groningen since 1997. She worked as a clinical linguist in the Aphasia team of Rehabilitation center het Roessingh in Enschede, The Netherlands, from 1984-1990. Since 1990 she has been a full-time staff member of the Department of Linguistics of the University of Groningen, where she has developed several courses in Neurolinguistics and Clinical Linguistics. Prof. Bastiaanse is co-ordinator of the European Masters in Clinical Linguistics (EMCL), an international MA program that runs at several European universities. Her main topic is aphasia, both from a theoretical and a clinical perspective. She has published many international research papers, two co-edited books, several Dutch diagnostic tests and treatment programs,

346

CONTRIBUTORS

some of which have been translated into other languages (German and English). She is an elected member of the Academy of Aphasia and member of the editorial board of several international (Brain and Language, Lingua, Aphasiology) and national journals. She collaborates with researchers from many European and American universities in projects that study the phenomena of aphasia from a cross-linguistic perspective. Susan Edwards is a Senior Lecturer at the School of Linguistics and Applied Language Studies at the University of Reading. She is also a speech and language therapist with wide clinical experience. She is currently director of a BSc and a MSC programme in language pathology, both of which lead to a professional qualification in speech and language therapy. Her field of research is language pathology and has included work with children with development problems as well as adults with acquired problems, such as aphasia. She has published articles in international journals and chapters of books. She is co-author of a clinical assessment of language in young children and she is about to publish a test of verbs and sentences in aphasia with Roelien Bastiaanse and Judith Rispens. She is currently writing a book on fluent aphasia for Cambridge University Press and a text book on aphasia with Roelien Bastiaanse for the same publisher. Elisabetta Fava is Full Professor of General Linguistics at the Humanities Faculty at the University of Ferrara and head of the degree board for the Faculty. Concurrently, she also holds teaching appointment at the Faculty of Medicine of the University of Padova. She has had research appointments at the University of Padova and at the University of California, Berkeley. From 1987 to 1999, she was Associate Professor of General Linguistics, first at the Humanities Faculty of the University of Trieste, then at the University of Padova, where she was the co-ordinator of the linguistic and clinical linguistic program both at the Faculty of Psychology and at the Faculty of Medicine. She has authored or edited five books, published widely in peer referred papers and contributed chapters to books in the field of General Linguistics. A large part of her research deals with the action-character of many linguistic phenomena, insights into which date back to Greek grammatical speculations, by investigating the grammatical features relevant to the representation of illocutionary force devices in different Indo-European languages. Anna Gavarro is Lecturer in the Department of Catalan Philology at the Universitat Autónoma de Barcelona. She did her Ph.D. in the Department of Linguistics at the University of Edinburgh in 1990 and was a visiting scholar

CONTRIBUTORS

347

at the Department of General Linguistics at the Universiteit van Amsterdam in 1990-1991. She has published articles, in Yearbook of Morphology, Probus, Gramática del Catalá Contemporani and various collections by international publishers. Her current interests are the grammar of language disruption and language acquisition. Sara Howard is a Senior Lecturer in Clinical Phonetics and Linguistics at the Department of Human Communication Sciences, University of Sheffield. She is Secretary of the British Association of Clinical Linguistics (BACL). Among others, she has published two books, co-edited with Mick Perkins: Case Studies in Clinical Linguistics (London, 1995) and New Directions in Language Development and Disorders (The Hague, 2000) and has produced papers in the areas of clinical phonetics and phonology. She is currently working on a book on Clinical Phonetics and Phonology for Academic Press with Barry Heselwood as co-author. Peter Howell has worked at the Department of Psychology University College of London since 1972. He was made the department's first Professor of Experimental Psychology in 1993. He has authored or edited three books, published 63 peer refereed papers, contributed 26 chapters to books. He was responsible for introducing frequency shifted feedback as a treatment for stuttering, which is now available in many states in the US. He has been a member of the European Union's Expert Advisory Group on Language Engineering Standards. His group offered the first automatic speech analysis over the world wide web. Gonia Jarema is Professor of neuro- and psycholinguistics of the Department of Linguistics of the Université de Montréal, where she has been a faculty member since 1980. She is also Head of the Language and Cognition Unit at the Research Center of the Institut Universitaire de Geriatrie de Montréal. She has been part of the international NIH- and NSF-funded Cross-Language Aphasia Study, in which she contributed the work dealing with Polish (principal investigator) and French (co-investigator). As Director of an international collaborative research project on the mental lexicon, funded by the Social Sciences and Humanities Research Council of Canada, she organized the First and the Second International Conferences on the Mental Lexicon in 1998 and 2000, thereby creating a forum for world-wide exchange of findings in this area. In addition to numerous publications in the fields of cross-language neuro- and psycholinguistics, she has recently guest-edited two special issues of Brain and Language on the mental lexicon.

348

CONTRIBUTORS

Eva Kehayia is Associate Professor in the School of Physical and Occupational Therapy in the Faculty of Medicine at McGill University, Montréal, Canada, and an Adjunct Professor in the School of Communication Sciences and Disorders in the same university. She is also director of the Centre for Interdisciplinary Research in Rehabilitation of Greater Montreal, responsible for psychosocial research. She holds a Ph.D. in Linguistics and Neurolinguistics and received postdoctoral training in Psycholinguistics. Her research focuses on two areas, the study of the representation, organization and access of language in non-impaired individuals across different languages and the study of language breakdown in acquired and developmental language disorders. She heads the McGill University branch of the Social Sciences and Humanities Research project investigating ihe architecture and nature of the mental lexicon and the Language and Communication Research Laboratories at the Jewish Rehabilitation Hospital in Laval, Canada. Giovanna Marotta is Professor of Phonetics and Phonology at the University of Pisa. She also teaches Phonetics at the Faculty of Medicine of the same University. Her main interests of research include phonological theory and acoustic phonetics. In particular, recent research focuses on prosody and its relation with syntax and pragmatics. The languages investigated are basically contemporary Italian and their dialects. She has published two books and over forty articles on these topics and collaborated in the compilation of Dictionary of Linguistics edited by GianLuigi Beccaria (Torino, 1994). Her study of spontaneous speech, both normal and pathological, has been carried out with reference to experimental techniques and special software for acoustic analysis. Lucia Mazzoni, native signer, is a professional conference sign language interpreter and a certified Italian sign language (LIS) teacher. She is Ph.D. candidate of the department of Linguistics of the University of Pisa where she got her MA degree in Applied linguistics with a thesis that concerned the error analysis in interpretation processes where sign language is involved. She currently holds a temporary lectureship in Italian sign language at the department of Linguistics in Pisa. Her area of research covers issues of language and deafness, whit a particular reference to the interpretation processes, interpreters training, the linguistic education of the deaf, the structure of LIS at lexical, morphological, syntactic and pragmatic level and the sign language comprehension process of Italian deaf signer. She has published some articles on sign language issue including the labial pattern of the sign language, the interpretation in LIS of technical languages, the production and comprehension of Fingerspelling and the project named "deafness, information, health".

CONTRIBUTORS

349

Lise Menn has been Professor in the Linguistics Department and a member of the Institute for Cognitive Science at the University of Colorado, Boulder, since 1986. She received a Ph.D. in linguistics in 1976 from the University of Illinois, with a dissertation on child phonology. From 1977 until 1986 she held research appointments at the Aphasia Research Center, Boston University School of Medicine. Her co-edited books include Exceptional Language and Linguistics (1982) and the three-volume Agrammatic Aphasia: A Cross-Language Narrative Sourcebook (1990), both with Loraine K. Obler; Phonological Development'. Models, research,implications, co-edited with Charles A. Ferguson & C. Stoel-Gammon (1992); and Methods of Studying Language Production, with Nan Bernstein Ratner. She is also the co-author of Non-fluent Aphasia in a Multi-lingual World (1995). Sofie van Mol is both a speech pathologist and a neurolinguist. She received her BA-degree from the School for Speech Therapy in Antwerp. After graduation she came over the the University of Groingen in The Netherlands for an MA-degree in Neurolinguistics. Her final work focused on the production of finite verb forms by children with Specific Language Impairments. This work she presented at several international conferences on language impairments. Currently she is working as a speech and langauge therapist in Belgium and as a clinical linguist on consultancy basis in The Netherlands. Florida Nicolai works as a researcher in the Department of Linguistics at the University of Pisa, where she has completed her graduate and postgraduate studies. Her main fields of interest are psycholinguistics and neurolinguistics. She is currently working on issues of language and deafness, with particular reference to the linguistic education of deaf, the development of reading and writing skills in deaf students, the structure (lexical, morphological, syntactic and pragmatic) of LIS (Italian Sign Language) used by Italian deaf signers. Dirk-Bart den Ouden is a Ph.D. candidate in linguistics at the University of Groningen, The Netherlands. His current project is on the role of phonological encoding and self-monitoring in conduction aphasia. He has previously conducted research into phonological structure in child language acquisition and creolisation. Leah Paltiel-Gedalyovich is Speech-Language Pathologist and a Ph.D. Candidate in Linguistics at the Ben-Gurion University of the Negev, in the Department of Foreign Literatures and Linguistics. She worked for over fifteen years with communication impaired children and adults in a variety of

350

CONTRIBUTORS

health and educational settings, concerning the communication impairments resulting from a variety of aetiologies including mental retardation, deafness, autism, learning disability, specific language disorders. As a result of her clinical work she became interested in developmental linguistics, specifically first language acquisition of semantics. Mick Perkins is a Reader in Clinical Linguistics in the Department of Human Communication Sciences, University of Sheffield, where he has worked for ten years. The main focus of his research is the relationship between semantics, pragmatics and cognition in communication impairment. He has published numerous articles and several books including Modal Expressions in English (London, 1983), Case Studies in Clinical Linguistics (London, 1995) and New Directions in Language Development and Disorders (The Hague, 2000), the latter two co-edited with Sara Howard, and is currently writing a book entitled Pragmatics and Language Pathology for Cambridge University Press. He is President of the British Association of Clinical Linguistics (BACL) and Vice President of the International Clinical Phonetics and Linguistics Association (ICPLA). Judith Rispens is a qualified speech and language therapist. She works as a Ph.D. student in the Department of Linguistics, the University of Groningen. The topic of her research is processing syntactic information in developmental dyslexia, using behavioural and electrophysiological measures. After her Masters in Linguistics, she is co-author of a test to investigate verbs and sentences in aphasia. She has published cross-linguistic studies on negation in agrammatic aphasia. Esther Ruigendijk is currently working as a postdoctoral fellow at Utrecht University in the project on comparative psycholinguistics, focusing on the syntax-discourse interface in aphasia and language acquisition. Before this, she worked as a Ph.D. student in the department of Dutch Linguistics at the Univer­ sity of Groningen. Before starting her Ph.D., she worked as a clinical linguist in a rehabilitation centre in Germany, where she diagnosed and treated aphasic pa­ tients. Her Ph.D. study is a cross-linguistic study about Case assignment in Agrammatism. For this study, she examined the speech production of Dutch, German, Russian and Hungarian patients. The results of this study have been presented at several international conferences. Part of her research results has been published in peer-reviewed papers in international journals and in confe­ rence proceedings. Her main areas of interest are aphasiology, especially agrammatism, syntax, child language, and aphasia therapy.

CONTRIBUTORS

351

Dušana Rybárová is currently Fulbright visiting researcher at the University of Arizona. She received her MA in clinical psychology from Comenius University in Bratislava, Slovakia and she is Ph.D. candidate at the Institute of Philosophy of Slovak Academy of Sciences in Bratislava. She has been working on various topics concerning the modular organization of the language faculty. Stavroula Stavrakaki, Ph.D., currently holds a temporary lectureship in language disorders at the Department of Educational and Social Policy in Thessaloniki. She studied Psychology and Linguistics at the University of Thessaloniki and got her MA degree in Language Acquisition at Essex University. She completed her Ph.D. studies at the University of Thessaloniki in 2001. Her Ph.D. thesis is concerned with the processes of syntactic production and comprehension in Greek children with SLI. Her research interests are centered on Psycholinguistics and Neurolinguistics. These include topics such as SLI, Broca's Aphasia and language development in genetic syndromes of mental retardation. She has published articles and chapters in journals and books such as Brain and Language, Lingua, Journal of Greek Linguistics, Advances in Clinical Linguistics and Phonetics. Vesna Stojanovik is a Ph.D. candidate at the Department of Human Communication Sciences at the University of Sheffield. She completed an MA in Linguistics and English Language Teaching at the University of Leeds in 1998. Her major interest is the modularity debate and the language/cognition interface in children with Specific Language Impairment and Williams Syndrome. Susan M. Suzman is Research Fellow in Speech Pathology and Audiology at the University of the Witwatersrand, Johannesburg. Prior to this, she was Senior Lecturer in the Department of Linguistics. Dr. Suzman has worked extensively in the field of language acquisition in Bantu languages. Her publications include a forthcoming book on Norms of Language Development in Zulu. She is pioneering the development of Speech Therapy in African languages and has recently received a South Africa Netherlands Programme for Alternative Development grant to research childhood language disorders in Zulu. Yumiko Tanaka Welty, Ph.D., is Associate Professor at the Department of Speech, Hearing and Science at the International University of Health and Welfare in Tochigi, Japan. She teaches language disorders in children. Her area of specialization is cross-linguistic comparisons to uncover the nature of language and reading disorders.

352

CONTRIBUTORS

Yishai Tobin is a Professor of Linguistics and Chair of the Department of Foreign Literatures and Linguistics at the Ben-Gurion University of the Negev, where he also teaches in the Department of Behavioral Sciences. Concurrently, he is a Professor of Phonetics and Linguistics in the Department of Communication Disorders, Speech, Language and Hearing in the Sackler Faculty of Medicine at Tel-Aviv University. His research interests include: articulatory and acoustic phonetics and phonology, discourse and text analysis, semiotics. He has published over 15 books and 150 articles representing a cognitive sign-oriented approach to language, based on Saussurean tenets, and functional guidelines to language as a system of signs used by human beings to communicate. He views language as an instance of human behavior and looks for connections between linguistic and nonlinguistic behavior patterns. His research in phonology as human behavior includes case studies in developmental and clinical phonology as well as bilingualism. Kyrana Tsapkini is a Post-Doctoral Fellow at the University of Montreal on the Valorisation Recherche Québec Fellowship. After receiving her B.A. in Psychology from the Aristotle University of Thessaloniki, she continued her studies at Dartmouth College, U.S.A., where she earned her MA. She went on to study neurolinguistic and psycholinguistic aspects of morphological processing in Greek at the University of Montréal and the Research Center of the Institut universitaire de gériatrie de Montréal under the supervision of Gonia Jarema and Eva Kehayia. She obtained her Ph.D. in neuropsychology in 2001 and has published several papers in Brain and Language and Journal of Neurolinguistics on off-line and real-time processing of morphological structure in nouns and verbs in aphasic patients and non brain-damaged populations. Jun Watanabe, M.D., is Professor at the Department of Nursery, University of Arts Junior college, Osaka Japan. He has been clinically active in psychologically supporting children and adults with mental retardation or with learning disabilities, and their families. His main research areas are the assessment and treatment strategies for children with language-learning disabilities, the programs for training preschool Japonese teachers and the effect of computer use on preschoolers.

CONTRIBUTORS

353

Shalom Zuckerman, Ph.D., is a linguist and psychologist by training. He received his MA from Tel Aviv University in Israel. After his graduation he came to the University of Groningen, where he did his Ph.D. His main topic of research is language acquisition. His Ph.D-thesis is on the so-called optionality of syntactic constructions. For this, he works in a minimalist framework. He defends the thesis that true optionality does not exist, not even in the child's grammar. Since 2001, he works at the University of Utrecht on a project of syntactic relations in discourse in language acquisition. He published several papers in international journals.