AGING AND COGNITION Knowledge Organization and Utilization
AGING AND COGNITION Knowledge Organization and Utilization
ADVANCES IN PSYCHOLOGY 71 Editors:
G. E. STELMACH
P. A. VROON
NORTH-HOLLAND AMSTERDAM NEW YORK OXFORD TOKYO 9
AGING AND COGNITION Knowledge Organization and Utilization
Edited by
Thomas M. HESS Department of Psychology North Carolina State University No& Carolina, U.S.A.
1990
NORTH-HOLLAND AMSTERDAM NEW YORK OXFORD TOKYO
NORTH-HOLLAND ELSEVIER SCIENCE PUBLISHERS B.V. Sara Burgerhartstraat 25 P.O. Box 2 I I , 1000 AE Amsterdam, The Netherlands
Distributors for the United States and Canada: ELSEVIER SCIENCE PUBLISHING COMPANY, INC 655 Avenue of the Americas New York. N.Y. 10010, U.S.A.
ISBN: 0 444 88369 X ELSEVIER SCIENCE PUBLISHERS B.V., 1090 All rights reserved. No part of this publication may be reproduced, stored in a retrieval system, or transmitted, in any form or by any means, electronic, mechanical, photocopying. recording or otherwise, without the prior written permission of the publisher, Elsevier Science Publishers B.V./ Physical Sciences and Engineering Division, P.O. Box 1991, 1000 BZ Amsterdam, The Netherlands. Special regulations for readers in the U.S.A. - This publication has been registered with the Copyright Clearance Center Inc. (CCC), Salem, Massachusetts. Information can be obtained from the CCC about conditions under which photocopies of parts of this publication may be made in the U.S.A. All other copyright questions, including photocopying outside of the U.S.A., should be referred to the copyright owner, Elsevier Science Publishers B.V., unless otherwise specified. No responsibility is assumed by the Publisher for any injury and/or damage to persons or property as a matter of products liability, negligence or otherwise, or from any use or operation of any methods, products, instructions or ideas contained in the material herein. Printed in The Netherlands
TABLE OF CONTENTS Preface and Acknowledgments List of Contributors Adult Age Differences in Activity Memory: Cue and Strategy Utilization Margaret P. Nor& and Robin L. West
vii
xiii
1
The Assessment of Qualitative Age Differences in Discourse Processing Elizabeth A.L. Stine and Arthur Wingtkld
33
Aging and Schematic Influences on Memory
93
Thomas M. Hess Metamemory in Adulthood: Differentiating Knowledge, Belief, and Behavior Christopher Hertzog. Roger A. Dtwon, and David F. Hultsch
161
Cognition and Aging: A Theory of New Learning and the Use of Old Connections Donald G. MacKag and Deborah M. Burke
213
Semantic-Memory Function and Dysfunction in Alzheimer's Disease Robert D. Nebes
265
A Developmental Approach to the Study of
Visual Cognition in the Elderly Matthew J. Sharps
297
Expertise and Aging: Life in the Lab Neil Chamess and Elizabeth A. Bosman
343
Table of contents
vi
9 Levels of Knowledge Utilization and Visual
Information Processing Willlam J. Hoyer
387
1 0 Aging and Everyday Cognitive Abilities Steven w.comelius
411
11 Affect, Control, and Real World Problem
461
Solving across the Adult Lifespan fiedda Blanchard-Flelds and Cameron J. Camp Author Index
499
Subject Index
511
During the past twenty-five years, there has been a steadily growing level of interest in the study of aging effects on cognition. Much of the work in this area has proceeded within a traditional information processing framework in which the primary focus has been on examining the efficiency of basic cognitive operations. In conjunction with relatively recent developments in cognitive and developmental psychology, however, there has been an increased emphasis on examining cognitive abilities in relation to both the background knowledge possessed by the individual (and their ability to access and use this information) and the speciflc contexts of performance. Specifically. investigators have begun to pay more attention to the role that stable or developing knowledge systems play in determining performance and age differences therein. There is accumulating evidence that the nature of aging effects on cognitive performance is dependent upon the type of task-relevant knowledge that the individual brings into the testing situation. For example, task-relevant expertise in older adults has been found to be associated with high levels of performance in spite of declining levels of efficiency in more domain-general operations. Apparently, individuals can use their experience and knowledge in some cases to maintain performance and/or develop compensatory mechanisms for aging-related changes in more general ability systerns. In a related vein, other investigators have examined the speciflc contexts in which different-aged adults function. Of primary interest here is the notion that development during adulthood is associated with changes both in experiences and in the demands placed upon a n individual's cognitive system, and that age differences in performance need to be interpreted within the context of these changing demands. The efficiency of particular cognitive skills at any point in the lifespan should be directly
vlii
Preface
related to their relevance to the individual's specific life circumstances. Thus, the form of any age function relating to performance will be dependent upon the relation of the specific skill being tested to the general life contexts of the age groups included in the study. The present volume consists of eleven chapters that dfscuss theory and research relevant to the just-discussed issues, with an emphasis on the examination of adult age differences in the representation, organization, and utilization of knowledge in relation to cognitive performance. The authors are all active researchers in the field of aging, and represent a variety of theoretical perspectives relating to the study of aglng and cognitive abilities. Consistent with this diversity, the chapters deal with a wide range of cognitive skills, methodologies, and subject populations. The first four chapters explore various aspects of memory performance. Norris and West discuss research examining adult age differences in memory for activities. They argue that, in contrast to studies using more traditional laboratory tasks, examinations of activity memory have a greater degree of ecological validity due to the practical significance of this skill in everyday life. In accordance with the current focus, their review of the literature indicates that age differences in 'activity memory are moderated by the type of tasks and stimulus items used. Stine and Wingfield examine the evidence for qualitative age differences in discourse processing through the use of relative memorability analyses, in which the recall of specific text units is compared across age groups. Based upon their work, they conclude that there is little evidence for qualitative change with age in the nature of discourse representation, but that the emphasis on speciflc processes across text types may change in conjunction with age-related variations in working memory efficiency. Hess then examines age differences in the extent to which the knowledge possessed by the individual influences memory performance. Using a schema-based framework for interpreting the relevant research, he argues that aging is associated with an increase in the influence of knowledge structures on retention. This increased influence appears to be related both to a greater dependence upon
Preface
ix
the support aspects of knowledge in response to declining general skills and to older adults' greater reliance on knowledge systems in editing memory output. In the fourth chapter, Hertzog. Dixon. and Hultsch explore the concept of metamemory a s it relates to aging and memory. They challenge the utility of the traditional view of metamemory as knowledge of memory functioning. Instead, they propose that metamemory may be better conceptualized using the notion of self efficacy. in which the individual's beliefs about his or her memory abilities is a n important predictor of performance. The next three chapters deal with issues of information access and representation, broadly defined. Mackay and Burke attempt to use Node Structure Theory to explain aging effects on memory performance. They examine age differences in the ability to create and access connections within a memory representation, and propose that a single mechanism--an impairment in the priming of memory nodes--can account for a variety of the aging effects on memory observed in the literature (e.g.. the relatively greater problem that older adults have in accessing new versus old information), In the next chapter, Nebes explores the hypothesis that Alzheimer's disease is associated with a loss of information from semantic memory. He concludes that the basic contents of semantic memory are preserved in Alzheimer's patients, but that the ability to access information in this system is impaired, with successful retrieval being dependent upon the demands of the task. In Chapter Seven, Sharps examines visual cognitive abilities and aging. He advocates the use of methodologies that allow more complete understanding of the conditions under which aging effects in performance will or will not be observed. His review of the research on visual cognition suggests that age differences in the represenation and processing of visual information vary as a function of a variety of task and stimulus factors, such as degree of practice and familiarity. The next two chapters explore issues of cognitive aging in relation to expertise. Charness and Bosman examine the effects of practice in both cognitive and motor domains. They conclude that extensive practice can result in maintenance of skill with age, primarily through the development of compensatory mechanisms.
X
Preface
They also argue, however, that aging-related changes in "hardware" set limits on the extent to which older individuals can benefit from experience. Hoyer examines the interaction between aging and expertise in the realm of visual information processing. He discusses this interaction at several levels of analysis, and attempts to specify the mechanisms through which relevant prior knowledge operates to counteract aging-related impairments that occur in visual cognition. The final two chapters deal with cognitive abilities in everyday contexts. Cornelius examines people's implicit theories of academic and everyday intelligence in relation to traditional ability measures and to development in adulthood. He argues that traditional ability tests that are used in many studies of intellectual development are more reflective of academic intelligence, which is valued most in early adulthood. In contrast, everyday intelligence is seen as increasing in importance with age during adulthood. Unfortunately, this factor is not reflected in traditional ability measures, which leads to a biased picture of intellectual skill in adulthood. Blanchard-Fields and Camp further discuss issues of everyday cognitive skills in their examination of real-world problem solving. They argue that there are qualitative differences in the manner in which different-aged individuals solve problems, and that changes in problem-solving performance can be best understood by examining variations in how individuals interpret and approach problem situations. ACKNOWLEDGMENTS The completion of a project such as this one does not come about solely through the efforts of one individual. First, I would like to thank the contributors to the volume for their hard work and excellent chapters, and for their indulgence of my novice (approaching expert) editor behavior. I would also like to thank George Stelmach (Series Editor) for his assistance in the initial development of this project, and K. Michielsen at North-Holland for his patience and encouragement throughout this endeavor. In a more general frame, I gratefully acknowledge the financial support for my work provided over the years by the National Insitute on
Preface
xi
Aging. Finally, on the homefront, my sincere thanks go to Elisa Fortenberry at North Carolina State University for her expert editorial assistance In putting the fhal product together. And last, but definitely not least, I would like to thank my family-Erica, Stephanie, and Peter--for their support and understanding throughout this project.
T.M.Hess
This Page Intentionally Left Blank
LIST OF CONTRIBUTORS Fredda Blanchard-Fields, Department of Psychology, Louisiana State University, Baton Rouge, Louisiana 70803-5501, U.S.A. Elizabeth A. Bosman, Psychology Department. University of Waterloo, Waterloo, Ontario N2L 3G 1, Canada Deborah M. Burke. Department of Psychology. Pomona College, Claremont. California 9171 1. U.S.A. Cameron J. Camp, Department of Psychology. University of New Orleans, New Orleans, Louisiana 70148. U.S.A. Neil Charness, Psychology Department, University of Waterloo, Waterloo, Ontario N2L 3G 1, Canada Steven W. Cornelius, Department of Human Development and Family Studies, Cornell University, Ithaca. New York 14853, U.S.A. Roger A. Dixon, Department of Psychology, University of Victoria, Victoria. British Columbia V8W 2Y2. Canada Christopher Hertzog, School of Psychology. Georgia Institute of Technology, Atlanta, Georgia 30332-0 170. U.S.A.
Thomas M. Hess. Department of Psychology, North Carolina State University. Raleigh, North Carolina 27695-7801. U.S.A. David F. Hultsch, Department of Psychology, University of Victoria, Victoria, British Columbia V8W 2Y2, Canada William J. Hoyer, Department of Psychology, Syracuse University, Syracuse, New York 13244-2340, U.S.A.
Donald G. MacKay, Psychology Department, University of California, Los Angeles, Los Angeles, California 90024.U.S.A. Robert D. Nebes, Western Psychiatric Institute and Clinic, University of Pittsburgh, Pittsburgh, Pennsylvania 15213. U.S.A. Margaret P. Norris, Department of Psychology, University of Florida, Gainesville, Florida 3261 1, U.S.A. Matthew J. Sharps, Department of Psychology, School of Natural Sciences, California State University - Fresno, Fresno, California 93740-0011. U.S.A. Elizabeth A. L. Stine, Department of Psychology, University of Kansas, Lawrence, Kansas 66045.U.S.A. Robin L. West, Department of Psychology, University of Florida, Gainemille, Florida 3261 1, U.S.A. Arthur Wingfield, Psychology Department, Brandeis University, Waltham, Massachusetts 02254-9110,U.S.A.
Aging and Cognition. Knowledge Organization and Utilization Thomas M. Hess (Editor) Q Elsevier Science Publishers B.V. ~orth-HolfandJ.1990
CHAPTER ONE
ADULT AGE DIFFERENCES IN ACTMTY MEMORY: CUE AND STRATEGY UTILIZATION Margaret P. Nor& and Robin L. West University of Florida
SUMMARY
Activity memory represents a n important new paradigm in cognitive gerontology because it includes tasks that have ecological validity for older adults and tasks that are not a s strongly influenced by strategic processing as traditional verbal memory tasks. This review of the literature, including new data, reveals that the magnitude of age differences in activity memory is strongly influenced by the types of tasks and items that have been employed in the literature. In addition, the impact of motor and object cues, and organization are discussed, with a n emphasis on the notion that age differences may be controlled by the interaction of these task characteristics. Activity memory provides a generally familiar context for learning and it is d e c t e d by the ease of remembering item-specific information. Consistent with the traditional verbal memory literature, age Merences can be explained by the degree to which an activity memory task supports or guides strategic processing. Until recently, the typical mode of representation in memory experiments was based on verbal encoding of words. Because of this tradition, our knowledge of age differences in memory abilities is largely limited to verbal memory tasks such a s word list recall, paired associate learning, and prose recall. These tasks have laid a
2
Norrls and West
strong foundation for understanding memory processing among older adults. Recently, cognitive gerontologists have felt the need to expand task domains, particularly to include nonverbal memory tasks that might be part of the everyday activities of older adults. One fruitful new task domain, to be reviewed here, is activity memory. Activity memory research represents an important new paradigm in cognitive psychology, with considerable heuristic value. R. L. Cohen (1981)has pointed out that investigation of alternative memory tasks such as activity memory may enhance our understanding of memory processes. In order to test the generality of established laws about memory, R. L. Cohen investigated memory for actions and did not find evidence for prirnacy, active attempts to memorize, or effects due to different orienting tasks. He concluded that commonly accepted memory laws about serial position, strategies, and levels-of-processing may be unique to verbal recall (R L. Cohen, 1981). It is clear that traditional verbal memory tasks may differ in many regards from other types of memory. Biiclunan and Nflsson (1984).for instance, have argued that verbal recall and memory for activities are substantially different in terms of stimulus properties. Unlike verbal memory tasks, recall of activities involves multimodal cues because several sensory systems can be involved, including auditory, visual, tactual, gustatory, and olfactory. The investigator can manipulate many modalities of information. Traditional verbal memory tests, on the other hand, are typically unimodal, involving auditory or visual systems. Activity memory research is also valuable because traditional verbal methods of assessing memory may be an inadequate analogue of normal memory functioning and may have poor ecological validity (Neisser, 1978; Poon, 1980). Limiting investigations to verbal memory may tell us little about memory performance outside of the laboratory. Memory for activities, on the other hand, has important practical significance. Recall for what one has done, or what one needs to do is a memory demand that is frequently encountered in everyday life. For example, common memory tasks may include remembering to take medication, or recalling whether or not the bills have been paid.
Actlulty Memory
3
The ecological validity of activity memory appears to be particularly relevant to research on aging. Studies show that memory for names and faces, and memory for activities are among the most common problems identified by elderly adults (Cavanaugh, Grady, & Perlmutter, 1983; Chaffin & Herrmann. 1983). Furthermore, it has been argued that traditional laboratory recall tasks may bias results in favor of younger adults because these tasks are less familiar to older adults (Erickson, Poon, & Walsh-Sweeney. 1980: West, 1986). As a result, patterns of age differences may vary with activity memory studies (e.g., Backman & Nilsson, 1984;Padgett & Ratner, 1987).Thus,investigations with the activity memory paradigm are valuable for verifying whether older adults' secondary memory deficits are limited to verbal recall. It is evident that the investigation of activity memory has important implications both in terms of expanding our knowledge of memory processes in general, and in illuminating aging effects. This chapter will review the research on adult age differences in activity memory, to determine whether or not there is consistent evidence for secondary memory deficits in this alternative domain. The following factors will be discussed the trpes of tasks and items employed, the influence of motor and object cues, and the influence of strategic processes such as organization. The discussion focuses on the impact of these factors on adult age differences in memory. TASKS AND ITEMS A common to-be-remembered item in the activity memory
literature is a discrete one-step action. Series of such actions are often presented in lists. Many authors have investigated Subject Performed Tasks (SPTs) such a s "cross your fingers" or "put the stamp on the book" (Backman & Nilsson, 1984. 1985;R L. Cohen, 1981, 1988;Dick & Kean, 1989;Glover, Timme. Deyloff & Rogers, 1987;Helstrup, 1986, 1987;Nilsson & Cohen, 1988).and similar lists of simple actions (G. Cohen & Faulkner, 1989;Guttentag and Hunt, 1988). Another measure is memory for simulated everyday events such as preparing to go on a trip or taking a tour of a building (Norris & West, 1988;Padgett & Ratner. 1987:Ratner & Bushey. 1984). Finally, recall of cognitive activities such as serial-learning
4
Nor& and West
tasks. WAIS items, and word-picture naming tasks have been investigated by Kausler and his colleagues (Kausler & Hakami, 1983;Kausler, Lichty & Davis, 1985;Kausler, Lichty & Freund, 1985; Kausler. Lichty. Hakami 8t Freund, 1986;Kausler & Phillips, 1989; Lichty. Kausler & Martinez, 1986). Effects of Aging Adult age differences do appear to vary a s a function of type of to-be-remembered item, BBckman's research has compared recall of SPTs to comparable verbal tasks such as sentence recall (recall of the same SPT items without motor action) or word recall (recall of the same nouns as appeared in SPT items). A consistent age by item type interaction has been found such that age differences were eliminated during SPT recall, but remained in the comparable verbal tasks (BBckman, 1984, 1985;Backman 8t Nflsson, 1984, 1985). G. Cohen and Faulkner (1989)investigated age differences in the recognition of actions similar to SPTs. Their activities, however, were limited to the placement of objects (examples: "put the stamp on the book," "put the spoon next to the toothbrush) whereas most SPT studies have included a variety of actions such as "smell", "fold, "look", and so on. Consistent with the earlier SPT data, no age differences occurred in the recognition of performed activities (G.Cohen 8t Faulkner, 1989). In contrast, Guttentag and Hunt (1988)found that young adults recalled substantially more performed actions than older adults. They employed a list of SFTs in which no objects were involved (e.g.. "shake your head yes", "touch your knee"). One possible explanation for these discrepant results is that no object cues were provided. Another distinction is that this investigation involved incidental recall whereas the studies finding no age differences employed intentional recall (Bilckman, 1985;B~ckman&Nflsson. 1984,1985:G. Cohen & Faulkner. 1989). Although intentionality has not been systematically examined in SFT studies, there is consistent evidence that intentional and incidental recall of cognitive activities do not differ (e.g. Kausler & Hakami, 1983). Another possible reason for the discrepant results in this study is list length. Guttentag and Hunt (1988)used lists of 24 items,
Acttufty Memory
5
whereas Backman's studies employed lists of 12 items. Age difllerences in SPT recall have been found with longer lists (R L. Cohen, Sandler. & Schroeder. 1987). A dissimilar dependent measure in the activity memory literature is recall of simulated everyday activities. This has been referred to as "went memory" (Padgett and Ratner, 1987). Rather than discrete one-step actions like SPTs. event memory typically requires the recall of a series of related actions. The complex events that have been studied include clay making (Padgett & Ratner, 1987: Ratner & Bushey. 1984).taking a tour of a building (Padgett & Ratner, 1987).preparation for vacation travel (Norris & West, 1988). and wrapping a parcel (Hashtroudi. Johnson, & Chrosniak. in press). With event memory, the overall event includes component actions which need to be recalled. The component actions are often organizable into categories such as "get the ingredients" for the clay making (Ratner & Bushey, 1984)or "pack a toiletry bag" for the travel preparation task (Norris & West, 1988). Age differences in event memory have varied as a function of organization and motor cues, as noted in later sections of this paper. Kausler and his colleagues have investigated age differences in cognitive activities. They have used a variety of cognitive tasks including perceptual-motor, verbal learning, semantic memory, and problem-solving tasks, and also, several tests from the Reference Test for Cognitive Factors, such as, Card Rotation or Maze Tracing (Ekstrom, French, Harman, & Derrnen, 1976). Their results have consistently shown that recall of cognitive activities is age-sensitive whether tested with incidental or intentional recall (Kausler & Hakami, 1983:Kausler, Lichty. & Davis, 1985:Kausler. Lichty. & Freund, 1985;Kausler et al., 1986;Kausler & Phillips, 1989:Lichty et al.. 1986). The previous evidence, then, was inconsistent, with age difllerences varying a s a function of item type. A recent study in our laboratory investigated these item type differences by comparing recall of cognitive activities and SPTs. Slxteen items (half SPTs and half cognitive activity items) were presented to younger and older adults. Presentation time was also varied. The to-beremembered activities were all performed and then tested with verbal free recall. SpTs were significantly easier for older adults to
6
No&
and West
recall than cognitive activities (see Table 1). A significant age by item type interaction also resulted from substantially greater age differences in recall of cognitive activities than recall of SPTs (Norris & West, 1989). This item type effect could explain why age differences have been prominent in studies of cognitive activity recall (e.g., Kausler et al., 1986)but not in studies of SPT recall (e.g., BBchan, 1985). In summary. various types of measures have been employed in the activity memory literature. Age differences in recall appear to vary as a function of item type and/or task type. While there is evidence that age differences are eliminated when testing for the recall of discrete one-step actions (BAclanan, 1985;BBckman & Nilsson. 1984,1985;G. Cohen & Faulkner, 1989).recall of discrete actions may be age-sensitive under certain conditions such as when long lists are used or when object cues are not provided. Event memory shows a complex interaction between age dmerences and presentation conditions (Norris & West, 1988;Padgett & Ratner, 1987;Ratner & Bushey, 1984).and recall of cognitive activities appears to be age-sensitive. Potential Explanatory Variables What variables account for the item type differences in aging studies? A number of factors could be controlling the outcome across studies because investigators using differing item types have often varied other characteristics as well, such as list length, familiarity, presentation time, and types of cuing. One candidate is list length. Age differences appear to occur more consistently with long lists (R. L. Cohen. Sandler, & Schroeder, 1987;Guttentag & Hunt, 1988;Kausler. Lichty, & Freund. 1985;Lichty, Bressie, & Krell, 1988).The use of strategic processes such as organization, verbal association. and visual imagery may be particularly effective when memorizing long lists. Age differences in recall may be more pronounced with longer lists because young adults are able to spontaneously use these strategies. R L. Cohen and his colleagues found significant age dmerences on an SFT list with 37 items but not on one with 14 items (RL. Cohen, Sandler, & Schroeder, 1987).At the same time, their study did not
Actioity Memory
7
replicate the usual significant age by item type interaction (significant age effects for words and not SPTs) for the 14-item list (R. L. Cohen. Sandler. & Schroeder. 1987). There are no other studies including a direct manipulation of list length so it is dimcult to determine its exact influence on age-related differences. However, across studies, list length does not consistently predict outcome a s older adults have sometimes performed as well as younger adults on lists of 32 SPTs (Dick & Kean, 1989). and have performed worse than younger adults on lists including only 12 cognitive activities (Kausler & Hakami, 1983; Kausler & Phillips, 1989). Nevertheless, list length could still affect performance in interaction with presentation time or other stimulus properties, such as item familiarity. When list length is varied, total presentation time changes as well. It is possible that this time variable explains the discrepant findings in the literature. In Kausler's investigations, total presentation time ranged from 30 to 90 minutes due to activity durations of 45 seconds per activity or longer. In contrast. the SPT studies conducted in BBckman's laboratory have had subjects perform activities for 5 seconds, with a total time of only 60 seconds in most studies (e.g., Bsckman. 1985). This short presentation time would result in at least five items remaining in primary memory at the time of test (see Glanzer. 1972). The longer presentation time, on the other hand, would probably permit subjects to retain no more than one activity name in primary memory. Such presentation differences could account for the fact that investigators using short intra-item intervals have not found age dmerences in activity memory (e.g., B~ckman& Nilsson. 1984). whereas those using longer intervals have (e.g., Kausler & Hakami, 1983). Presentation time may also be a n influential factor in the nonsignificant age by item type interaction found by R L. Cohen and colleagues (R. L. Cohen, Sandler, & Schroeder, 1987). The nonsignificant interaction was due to the elimination of age differences in the word recall list, as well as the SPT list. The authors suggested that their presentation time of 3 seconds per word could account for their failure to find the age by item type interaction. Their total presentation time was therefore shorter
a
No&
and West
than the total presentation time in Backman's (e.g., B&3unan. 1985). Although the time difference does not appear to be a sizable one, Cohen's older subjects may have been able to recall more items from primary memory than BBckman's subjects, and thus performed as well on the word list a s the SFT list (R.L. Cohen. Sandler, & Schroeder, 1987). Clearly, older adults can perform better when recalling from primary memory than from secondary memory (Fozard, 1980). A study conducted recently in this laboratory addressed this question by varying presentation time (see Table 1). Eight SFTs and eight cognitive activities were presented with intra-item intervals of 8 seconds (approximately 4 minutes total time) or 45 seconds (appraximately 14 minutes total presentation time). With respect to presentation time, a significant rate by item interaction was found. The source of this interaction was a significant time effect for the SFTs and not the cognitive activities. This is consistent with an earlier finding that presentation time did not affect recall of cognitive activities (Kausler, et al., 1986). although presentation rate differences between 1.5, and 10 second intervals have not been found with SPTs (RL. Cohen, 1985). In the present study, recall of SFTs was significantly higher than recall of cognitive activities with slow presentation, whereas performance on the two item types was comparable with the more rapid presentation rate. This interaction did not vary by age group (see Table 1). This evidence indicated that the effects of presentation can vary in accordance with other experimental variables such as item type. The results of this study, as well a s the inconsistent effects of list length in the literature, suggest that these two stimulus characteristics -- list length and presentation time -- can influence recall in interaction with other factors (e.g.. item type), but may not always directly predict age Werences (see Norris 81West, 1989). Another possible explanation for discrepant results for SFTs and cognitive activities concerns the familiarity of the tasks employed. Kausler and his colleagues have suggested that age deficits may be pronounced in his research because of the artificial nature of the laboratory tests that subjects perform (Kausler. Lichty, & Davis, 1985). Similarly, Backman argues that SFTs are ecologically valid because they require encoding of meaningful
Activity Memory
9
Table 1
Mean Number of Items Recalled as a Function of Presentation Rate and Item ZlJjIe
SPT Items
Group
Cognitive Activities
Slow
Fast
Slow
Fast
Younger
6.4
5.1
5.2
4.7
Older
4.9
3.7
2.9
2.8
Both
5.5
4.4
4.0
3.7
Overall
Overall
Younger
5.7
5.0
Older
4.3
2.8
Both
5.0
3.9
objects and actions, which would facilitate processing by older adults (Bilckman & Nilsson, 1984). For example, clapping ones' hands or looking in the mirror (Biickman & Nilsson. 1984) are arguably more familiar tasks than performing the Object Assembly test or producing exemplars of a semantic category (Kausler & Hakami. 1983). In fact, in one investigation of cognitive activities, age differences appeared to be less prominent for a set of cognitive activities that included simple mathematics, a more familiar activity than letter cancellation or verbal learning (Kausler & Hakami. 1983). On the other hand, attempts to manipulate familiarity directly have not consistently shown that recall of
10
Norrls and West
SPTs varies as a function of familiarity (R. L. Cohen. Peterson & Mantini-Atkinson. 1987;Helstrup, 1986, 1987;Lichty et al., 1988; Nilsson & R. L. Cohen, 1988). Also, if familiarity were controlling the outcome vis a vis age daerences. event memory studies ought
not to show age differences because the to-be-remembered events are generally very common everyday activities (e.g., Norris & West, 1988).In fact, event memory often does demonstrate age Werences (e.g., Padgett & Ratner, 1987). Overall, familiarity per se does not stand out as a consistently strong explanation for variations across studies, but it could still influence performance in interaction with other factors such as presentation time. Although more study is still warranted, it appears that neither list length, presentation time, nor familiarity can alone account for age differences that vary as a function of item type. At this point, there is no single, clear explanation in the literature. One possibility that has not yet been examined is that recall of cognitive activities places a greater demand on working memory. Subjects are required to recall a task label or description for the cognitive activities, (e.g., "incomplete words". '"word-picture naming"). In contrast, when a person recalls an SFT, they recall the specific action phrase that defined the act they performed, (e.g.. "fold your arms". "stretch your legs"). Thus, recall of cognitive activities generally requires the subject to translate the actions into an unfamiliar label or description, placing a greater demand on working memory. No such transformation is required in SPT recall because the task labels are highly farniliar phrases. When tobe-remembered items require transformation or translation into some kind of coding system, older adults are at a disadvantage (Craik. 1986). This factor needs further investigation. Finally, discrepant findings for different types of activity memory may be due to the kinds of cues that are available. Age differences were eliminated in those experiments in which both motor and visual cues were provided to the subjects, whereas the motor involvement in Kausler's cognitive tasks is mostly limited to paper and pencil activity. Motor activity appears to facilitate the processing of older adults in most of the activity memory studies (e.g, BSckman. 1985;Norris &West, 1988).Visual cues, provided by objects, may also play a role because simple objects have generally
Actlufty M e m o y
11
not been present in the studies which investigated age differences in verbal recall. It has been argued that age differences are eliminated in activity recall precisely because multimodal cues are provided (Bi-ickman, 1985:Bi-iclunan & Nilsson. 1984,1985).The importance of these two particular types of cues will now be discussed. MOTORCUES Motor Encodhg Cues The presence of motor cues is a particularly important distinction between activity memory and verbal recall. Motor cues are always present in action lists but are never presented in verbal recall tasks. The significance of motor cues for recalling SpTs has been supported by findings that older adults were not as adept a s younger adults in recalling sentence descriptions of SPTs (Bi-ickman 8t Nilsson, 1984,1985).although they did recall enacted SPTs as well as the young. Additional evidence for the importance of motor cues comes from Kausler's studies using cognitive activities in which motor action was mostly limited to using a pencil and paper. These studies found age daerences in activity recall (Kausler & Hakami, 1983;Kausler, Lichty, & Davis, 1985: Kausler, Lichty. & keund. 1985;Kausler et al., 1986: Kausler & Phillips, 1989).except in one Lichty study using a list that included 12 motor tasks only (Kausler 81Lichty. 1988). Saltz' work has also suggested that motor cues are beneficial for older adults (Saltz. 1988).His study examined age differences in the motor processing of sentences, that is, the recall of sentences in which the verbs were acted out by the subjects. For example, half of the subjects would enact "eating" when presented the sentence 'The squirrel was eating green acorns". Recall was cued with the sentence verbs. They found that age dmerences were reduced, but not eliminated, when the subjects enacted the sentence verb. This recall task falls along a continuum between activity and verbal recall tasks, because the to-be-remembered item is a sentence that includes an action, not just an action phrase. This may explain why they found only a reduction, and not an elimination of age differences in recall.
12
N o r i s and West
The role of motor cues has also been investigated in the event memory literature. Age differences in recall were eliminated when subjects performed actions in preparation for a vacation (Norris & West, 1988).and when subjects went on a tour of a building (Padgett and Ratner, 1987).However, age differences in recall remained on a clay-making event (Padgett & Ratner, 1987;Ratner & Bushey, 1984). In these studies, list length varied from 24 to 42 items. There are no studies that have examined recall for events using shorter lists. It is reasonable to speculate that age differences in event recall have been magnifled due to list length. Another study presents contradictory evidence for the motor hypothesis (Lichty et al., 1986). Subjects were tested for recall of motor activities such as cutting shapes, connecting rings, and clay modeling. They were also given cognitive activities to recall such as estimation of length, incomplete words, and word search. Younger adults recalled both types of activities better than older adults, and the authors concluded that motor involvement was an unimportant predictor of the magnitude of age differences in activity memory (Lichty et al., 1986). However, later analysis revealed that the age differences in recall of both the cognitive tasks and the motor tasks could be attributed to only four of the twenty-four items (Kausler & Lichty. 1988). The importance of motor cues can also be addressed by a large body of literature which compared enacted activities to imagined. planned, or watched activities. This is an important comparison because watched or imagined actions provide many of the rich visual and kinesthetic cues present in performed actions, but they do not provide motor cues. Younger adults recall actions which they imagine performing as well as they recall actions that are actually performed (BBckman & Nilsson, 1985;Helstrup. 1986, 1987:Saltz & Donnenwerth-Nolan, 1981). In contrast, older adults may benefit from motor cues, and perform more poorly when such cues are not available. Age differences in recall have been found when subjects were given action sentences to recall with instructions to imagine themselves performing each action (BBckman & Nflsson, 1985: Guttentag & Hunt, 1988). These age differences in recall were eliminated when motor cues were provided (BBclunan & Nflsson. 1985).but not when a long list of
Activity Memory
13
items without objects was used (see page 6) (Guttentag& Hunt, 1988). With event memory, there was no age by event type interaction for recall of performed as opposed to imagined events. Younger adults generally remembered more details about events of both types (Hashtroudi et al., in press). Similarly, Kausler and colleagues found that planned and performed activities were less often recognized by their older adults than their younger adults, with no interaction of age and performance condition (Kausler. Lichty, & Freund, 1985). Other investigators have compared subject-performed tasks to experimenter-performed, or watched, tasks. R L. Cohen and his colleagues have typically found no differences in recall between watched or performed activities for younger adults (R. L. Cohen, 1981, 1983;R. L. Cohen & Bean, 1983;R. L. Cohen, Peterson & Mantini-Atkinson. 1987). However, watched activities, even when embedded in a complex sequential event like making clay, are not recalled a s well by older adults a s activities that are actually performed (Norris & West, 1988;Padgett & Ratner, 1987). On the other hand, recognition testing can eliminate age differences in remembering watched actions (G. Cohen & Faulkner. 1989). Overall, these data suggest that degree of motor activity predicts age differences in free recall. Age differences in recall typically remain when actions are only watched or imagined, suggesting that the motor modality is most important for reducing age deficits. There are exceptions to this statement (see, for instance, Guttentag & Hunt, 1988;Kausler & Lichty, 1988)but the exceptions may be explained by other variables, such a s list length and item familiarity, which may interact with motor cues. When studies are placed on a continuum reflecting degree of motor involvement, those that involve complex to-be-remembered items with minimal motor activity show age differences, with the age gap decreasing a s conditions approach the other end of the continuum--simple motor actions. Motor Encoding and Retrieval One valuable aspect of research on activity memory is that it may be able to shed light on the dmerential impact of encoding and
14
No&
and West
retrieval deficits in older adults because motor cues can be manipulated at both stages of processing. Geropsychologists have attributed age deficits on standard memory lists to encoding and retrieval deficits in the elderly (Kausler, 1982). and their differential influence has been debated (see Poon, Walsh-Sweeney, & Fozard. 1980). R. L. Cohen (1981.1983)maintains that encoding mechanisms do not appear to be important in SFT recall. His evidence for this includes the absence of rehearsal and the failure to find orienting task Werences (R L. Cohen, 1981).Kausler and his colleagues have also shown that activity recall is not affected differentially by incidental and intentional recall instructions (e.g.. Kausler & Phillips, 1989). Kausler has concluded that activity memory is independent of rehearsal which occurs during encoding. Hence, the age decline in recall of cognitive activities has been attributed to a retrieval deficit in the elderly (Kausler, 1985). In contrast, Backman and Nilsson (1984, 1985) have emphasized the importance of encoding. An encoding deficit in the elderly is implied in their proposal that older adults are able to compensate for their memory deficits in SPT recall tasks because SPTs provide rich multimodal cues for individuals, thus removing the requirement for the subjects to generate such cues for themselves. The improvement from non-guided memory tasks (e.g., word recall) to guided memory tasks (e.g., SPT recall) may be less pronounced for younger adults because they are able to spontaneously recode sentences, words or imagined actions to other modalities by carrying out organization, imagery, and verbal mediation. Older adults are not able to spontaneously perform these strategic operations, and require the multimodal information in SPTs to support their recall (Backman & Nilsson, 1985). By virtue of the fact that verbal free recall is used in all conditions, most SPT studies do not reveal the role that retrieval could play as the source of older adults' difficulties. Thus, explanations have focused on the fact that SFTs permit older adults to compensate for their strategic encoding deficits. One method which is used to address the encoding versus retrieval issue is to employ recognition tests. In this case, the recognition findings are not definitive because age difllerences occur
Actfufty Memory
15
in recognition of performed activities sometimes (Kausler, Lichty & Freund. 1985).but not always (G. Cohen & Faulkner. 1989). In
general, however, ceiling effects confound studies of activity recognition (Kausler & Lichty. 1988). In most of the extant literature, motor cues were provided only during encoding (BBckman, 1985:BBckman & Nilsson. 1984,1985; R. L. Cohen. 1981). That is, the items were presented with motor cues by having the subjects perform the tasks as the experimenter read the items. Recall, however, was a verbal description of the tasks. This procedure does not allow for an investigation of the influence of motor cues during retrieval. The designs employed by Norris and West (1988)and Salk (1988)manipulated motor cues during both encoding and retrieval to study the impact of motoric processing in both phases of memory. These task conditions were compared: a) verbal encoding/verbal retrieval, b) verbal encoding/motor retrieval, c) motor encoding/verbal retrieval, and d) motor encoding/motor retrieval. In both studies, the presence of motor cues during retrieval did not eliminate age differences. In contrast, age dnerences were reduced in the motor encoding/verbal retrieval condition (Norris and West, 1988). These data are consistent with the findings in SPT studies because motor encoding/verbal retrieval is the typical SPT condition (e.g., BBckman & Nilsson. 1985). This pattern was also seen in a study comparing children and young adults in sentence recall. Motor enactment at input improved recall for both groups. In contrast. motor enactment during retrieval did not significantly benefit either age group (Salk & Dixon, 19821. General conclusions about the age-related impact of motoric cues await further investigation. In particular, there is a need for more studies varying motor cues at both encoding and retrieval with different item types. Research with cognitive activities has shown clearly that retrieval operations are important in understanding age differences in activity memory (Kausler. 1985). Others have argued that motor encoding facilitates older adults' recall of SPTs (e.g.. B~ckman.1984).When motor cues at encoding and retrieval are manipulated, it appears that motor encoding/verbal retrieval is the condition least likely to show age differences. At this point, then, there is insufficient evidence to
16
Nomh and West
argue that motor cues will always improve older adults' performance. Older adults may perform well on SpTs because of the combined effect of using motor encoding with verbal retrieval on highly familiar, simple actions. Together, these conditions may provide the supportive guidance required for older adults to perform well (Perlmutter & Mitchell, 1982). OBJECTCUES The use of external objects is another important distinction between traditional verbal recall and activity recall. Activity memory studies have typically used lists in which some items required the use of objects (e.g. smell the soap) and some items did not (e.g. yawn). In contrast, traditional verbal lists do not include objects. It is possible that older adults' memory performance is improved on activity memory tasks because the objects seen during encoding provide visual cues. Research with younger adults has addressed this issue. Some studies have compared recall of SpTs with objects to recall of SPTs which were mimed, that is, objects were imagined but were not present. The findings are mixed, with symbolic performance leading to recall that is sometimes worse and sometimes equivalent to SPTs with objects (Helstrup. 1986. 1987). Several studies have investigated whether motor cues alone or the combination of motor and visual cues are important in activity recall. The recall of SPTs with objects present was compared to the recall of SPTs without objects. No differences in recall were found in some studies (BWcman & Nilsson, 1984;R. L. Cohen, 1988;R L. Cohen, Peterson. & Mantini-AtMnson. 1987, Experiment 1). although in one case, items without objects yielded better recall than those with objects (R L. Cohen, Peterson, & Mantini-Atkfnson, 1987,Experiment 2). Furthermore, in one study, recall probabilities were calculated separately for the action component and the object component of the SpTs; the results showed that a high recall probability for the action component was more predictive of item recall than a high recall probability for the object component (R. L. Cohen. Peterson, & Mantini-Atkinson. 1987). These results suggest that the motor cues are the critical variable in SPT recall. However, these data
Actiulty Memory
17
were collapsed over age groups, and therefore do not address the question of whether object cues are important in eliminating age differences in SFT recall. In order to test whether object cues alone will eliminate age differences in recall, young and old adults were tested for recall of SFTs and sentences with objects present (BAckman, 1985). The young adults recalled both types of stimuli equally well, whereas the older adults recalled SPTs better than sentences with objects present. B2kkman (1985) concluded that the presence of objects during encoding plays a minor role with respect to the lack of age differences in SFT recall and that the essential task properties for eliminating age differences were the motor cues provided by the performance of the task items. However, it is just as likely that the combined effect of visual and motor cues was important for accurate SFT recall, as no condition provided motor cues without any objects. A recent study in our laboratory investigated the influence of object cues in SFT recall, Older and younger adults were given action lists to recall with the presence or absence of visual and organizational cues. The following four lists were presented under two conditions (verbal encoding/verbal retrieval or motor encoding/verbal retrieval): a) organized list with objects present (categories included actions with office, kitchen, toiletry, and sewing items). b) unorganized list with objects present, c) organized list without objects (categories included hand, feet, torso, and mouth actions), and d) unorganized list without objects. Both age groups performed significantly better with objects than without objects, but the age by object cue interaction did not approach significance, suggesting that the visual cues provided by objects are not important in eliminating age differences in SPT recall. However, object cues did interact significantly with condition (verbal versus motor encoding) and with list organization (see Table 2). These interactions were due to the greater impact of objects when other cues--motor encoding cues and list organization--were not present. In summary, those studies which have tested young and middleaged adults have produced contradictory results. Some have found no dif€erence in recall between those items which use objects and
Norrls and West
18
Table 2 Mean Number of Items Recalled as a Function of Test Condition. Obfect Cues, and Organization Verbal Condition Organized Group Younger Older Both
Unorganized
Object
No Object
Object
No Object
10.0 8.2 9.1
9.9 7.4 8.6
10.8 7.8 9.3
9.0 5.9 7.4
Overall
Overall
9.9 7.7 8.8
9.9 6.8 8.3
Younger Older Both
SPT Condition
Organized Group Younger Older Both
Younger Older Both
Object No Object 11.5 9.0 10.2
11.6 10.0 10.8
Unorganized Object 12.1
9.5 10.8
No Object
11.4 8.4 9.9
Overall
Overall
11.5 9.5 10.5
11.7 8.9 10.3
Activity Memory
19
those that do not. Others have found that object cues sometimes facilitate and sometimes impair recall. Although the elimination of age differences in SFT recall can not be attributed to the provision of object cues, there is evidence to suggest that the presence of objects has other important interactive effects. Object cues improve recall in the absence of other cues, but not when other cues are present. There are two important implications from this cue interaction. First, it suggests that the visual cues provided by objects are less powerful relative to motor and organizational cues. It is possible that object cues do not consistently eliminate age Merences in recall because they are less robust. Second, previous findings on the impact of object cues may be inconsistent because the effects of cuing interact with other test variables. In terns of the aging literature, there is evidence that object cues alone are not sufficient to eliminate age differences in activity recall. Our recent data also show that object cues are not influential in eliminating age differences in SIT recall even when combined with motor and organizational cues. These findings indicate that, overall, motoric processing is the more important variable for eliminating age deficits in recall. STRATEGIC PROCESSING IN ACTIVITY MEMORY
In addition to the influence of visual and motor cues, there has been considerable investigation of the impact of strategic processing on activity recall. It is possible that activity memory, unlike verbal memory, is recalled without the benefit of strategic processing. Verbal recall is thought to be strongly influenced by strategies. Thus,verbal recall favors younger adults who are able to spontaneously use organization, visual imagery, verbal association, etc. It has been argued that age differences in SFT recall are eliminated because SPT recall is nonstrategic (R. L. Cohen, 1983,1984). Several pieces of evidence have been cited to support the notion that SPT recall is nonstrategic or automatic: a) subjects report no attempt to memorize during SPT presentation (RL. Cohen. 1981).b) young children, older adults, and educable mentally retarded adults, who presumably have strategy deficits, are not penalized in
20
Norrls and West
tests of SPT recall (Backman, 1985;Backman & Nilsson, 1984, 1985;R. L. Cohen & Bean, 1983;R. L. Cohen & Stewart, 1982).c) a levels-of-processing effect has not been found in SPT recall (R. L. Cohen, 1981). d) SPT recall was less impaired than word recall under divided attention conditions (R. L. Cohen. 1988;Bilckman, Nilsson, & Chalom, 1986).and e) practice effects are minimal (Kausler & Lichty, 1988). The strategic processes that are most studied in the activity literature are rehearsal and organization. There is agreement that primacy effects are not seen in SPT recall (Backman & Nilsson, 1984,1985;R L.Cohen. 1981. 1983).This isviewed as evidence that SPT recall does not rely on rehearsal strategies (Rundus & Atkinson, 1970;Atkinson & Shiffrin, 1971). An exception to this, however, comes from evidence that primacy effects in SPT recall are seen when serial recall is tested rather than free recall (Helstrup. 1987). This suggests that subjects can be induced to use rehearsal for SPT memory, although they may not spontaneously do so. There is also evidence that cognitive activities, like SPTs. are encoded without rehearsal. Kausler and his colleagues have consistently found that incidental recall was as proficient as intentional memory for cognitive activities. This also provides definitive evidence that activity recall is independent of rehearsal (Kausler & Lichty, 1988). The use of organization in activity memory has also been investigated. Evidence for the use of organization has cast doubt on the notion that SPT recall is nonstrategic in all respects. Among young adult subjects, it has been found that nonorganizable lists cause greater impairment in SPT recall than in sentence recall (Backman et al.. 1986).Backman and Nilsson (1984,1985)suggest that the multidimensional task properties of SPTs enhance the detection of superordinate categories for older adults and this may account for the lack of age differences in SPT recall. However, if this hypothesis were correct, the age by task interaction found on overall performance should also be found on measures of organization, and that was not the case in two studies (Backman & Nilsson. 1984, 1985). Although SPTs were organized to a greater extent than verbal tasks, providing evidence that SPTs are not
Actfufty Memory
21
entirely nonstrategic, the elimination of age differences could not be attributed to the use of organization by the older adult group. A study that examined memory for cognitive and motor activities (Lichty et al., 1986) also did not find a relationship between organization and recall. Although they found no age differences in the use of organization. the younger adults recalled more of the activities than the older adults (Lichty et al.. 1986). However, in this study and the ones mentioned above, only two superordinate categories were included. With only two categories, the potential for organization was necessarily limited, which may account for no association between organization and aging in these studies. The role of organization has also been studied in our laboratory. Older and younger adults were given four SFT lists to recall, two lists that could be organized into four semantic categories and two that could not be organized systematically. (Organizability was also crossed with the presence of visual cues-see Table 2). There were main effects for age and for organization. Also, a significant age by organization interaction resulted from the older adults benefitting more from the organizational cues than younger adults. Organization has also been examined with event memory. When motor cues were varied during encoding and retrieval, younger adults made greater use of organization than older adults in all conditions except the one condition--motor encoding/verbal retrieval--in which no age differences in recall occurred (Norris & West, 1988). In contrast to previously mentioned studies (Bi4ckman & Nilsson, 1984. 1985; Lichty et al.. 19861, this list included seven superordinate categories. It is possible that the relationship between organization and recall is more evident when the lists employed are highly organizable. Another event memory study had mixed results. Padgett and Ratner (1987) found no age differences in recall of the component actions in a tour-taking event, but the young recalled more than the old on a clay-making event. Both events were presented in a n organized structure, however, the items in the clay-making event also were presented in a logically sequential order. The authors interpreted this a s evidence that the younger adults benefitted more
22
No&
and West
from the logical relationships inherent in the clay-making event (Padgett & Ratner, 1987). At the same time, there were no age differences in the recall of the superordinate category actions for both events, suggesting that the organized structure of the to-beremembered items was retained by the older group as well a s the young (Padgett & Ratner, 1987:Ratner & Bushey, 1984). In summary, there is evidence to suggest that activity recall is nonstrategic with respect to rehearsal. This suggests that age differences in activity recall may be eliminated, in part, because the types of strategic processing employed by younger adults in wordlearning tasks are less meaningful when learning activities. On the other hand, memory for actions is apparently not all together different from memory for verbal material. This is demonstrated by the evidence that categorical organization can enhance recall of activities. Organization appears to have a greater influence on performance when the organizational structure is more prominent. It appears that subjects can be induced to use organization in activity recall when the material is highly organizable and that age differences in recall may be eliminated under these conditions. CONCLUSIONS Although it has been commonly accepted that older adults perform more poorly on memory tasks than younger adults, activity memory studies reinforce the notion that age-related memory declines may not be pervasive. A review of the literature revealed that the magnitude of age differences in activity recall is influenced by motor and organizational cues, but apparently not by object cues. The motor cues provided during encoding typically eliminated the age deficit in SFT and event recall, although this finding did not extend to cognitive activities in which the motor cues were limited largely to manipulating paper and pencil. Further evidence for the importance of motor cues came from studies that demonstrated that younger adults consistently recalled watched and imagined activities as well as performed activities, whereas older adults recalled more performed activities than watched or imagined ones. In addition, motor cues at encoding may be more important in
Actiulty Memory
23
eliminating age differences than motor cues at retrieval (this conclusion is made with caution, however, due to limited data). It is important to recognize that the influence of motor cues on age differences does not stand in isolation. The effectiveness of motor cues does not depend on visual cues. At the same time, motor cues may be especially effective when an organizational structure is provided by the to-be-remembered items. In spite of consistent evidence favoring the notion that motor cues facilitate recall by older adults, motor cues do interact with other variables. Therefore, the benefits of motor cuing for older adults may be lost when other test conditions, such as lengthy presentation times, long lists, or unfamiliar task labels. favor young adults. Item type effects appear to have a strong impact on age differences in activity recall. The earlier discussion of item type focused on three domains -- SPTs. event memory, and cognitive activities. The importance of item type in predicting the magnitude of age differences in activity recall is further demonstrated by individual item analyses conducted by Kausler and his colleagues (Kausler & Lichty, 1988). They found that the magnitude of the age differences varied substantially from item to item, regardless of whether the tasks were cognitive or motor in nature. In fact, observed age differences in the recall of twelve motor and twelve cognitive tasks were due to sizable age dmerences on only four of these twenty-four items (Kausler & Lichty. 1988). The age differences were not significant on the remaining items, and on several items, the older adults recalled more than the younger adults. The authors further point out that some activities were substantially more memorable than others. regardless of age. These individual item differences have rarely been examined although one study found that item recall probabilities were not correlated with rated familiarity or vividness (R. L. Cohen, Peterson, & Mantini-Atkinson, 1987). Individual action items vary along numerous dimensions, in spite of many similarities between the different task types (see Kausler & Lichty, 1988). The importance of these dimensions of difference is not yet fully understood. Even within the tasks categorized as SFTs, there is considerable variation among items on many potentially influential factors: number and kind of body
24
Norris and West
movements required, presence of objects, use of isolated a s opposed to repetitive motions, number and kind of mental operations required. familiarity of the to-be-performed activities, and so forth. More systematic analyses of such item characteristics are needed. The existing literature, reviewed here, has explored only a few factors that may influence patterns of age differences. Models for understanding the activity memory data have focused both on encoding and retrieval processes. Our conceptualization combines the models of others with an overall emphasis on contextual issues. The contextual model proposes that retention depends not only on current cognitive processes, but also on the context in which information was learned. According to that model, recall will improve when the to-be-remembered information is presented within a meaningful general context (Hultsch & Deutsch. 1981). There is evidence to suggest that age differences in recall are eliminated when familiar materials are used within a familiar context (Botwinick & Storandt, 1980: Hultsch & Dixon, 1983: Poon & Fozard, 1978). Hence. age differences in SPT recall may be eliminated, in part, because the actions to be recalled can be associated with meaningful everyday settings. In addition. simple motor actions may elicit related contextual information that is distinctive to specific items. making those more memorable. This item-specific information is likely to be processed by older adults. and hence, to improve their level of recall. There is evidence to suggest that successful memorizing by older adults is particularly dependent on the provision (by the experimenter) of item-specific cues that are well integrated with the to-be-remembered items (Burke & Light, 1981). Older adults are thought to be deficient in their spontaneous use of strategies such as rehearsal, organization, and association (Poon, 1985)that would result in memory traces including such cues. Thus, older adults' recall memory is enhanced when the task properties provide integrated item-specific information (e.g., Park, Smith, Morrell, Puglisi. & Dudley, 1987). Consistent with this interpretation. age differences do not occur in activity recall when the task conditions provide subjects with multimodal cues, but remain when the task requires subject-generated strategies. Activity memory tasks that
ActluUy Memory
25
provide motor, visual and/or organizational cues, can be viewed a s guided memory tasks. Individuals are induced to attend to and register cues on such tasks [B~ckman& Nilsson, 1985).and such tasks are less likely to result in age differences in memory performance (Craik. 1986). A similar model has been proposed by Kausler and his colleagues (Kausler & Lichty. 1988). They suggest that information processing requirements are reduced for the encoding of activities. They argue that the activation or initiation of an action sequence automatically creates a memory trace. Their model suggests that there is information automatically processed during the performance of an activity. Hence, when older adults perform the activity they process information that is otherwise not processed during unguided memory tasks. The automatically-encoded information improves the level of activity recall for older adults. At the same time, the model of Kausler and his colleagues suggests that retrieval of activities remains effortful [Kausler & Lichty, 1988). One aspect of their model is particularly intriguing. Retrieval remains effortful because the context-specific information processed at encoding is not necessarily available at retrieval. Hence, age differences in recall may exist. despite encoding of item-specific information, because older adults may not be as efncient as younger adults in the effortful retrieval of that item-specific information (Kausler & Lichty, 1988). DiMerences in retrievability of such detail may vary as a function of item type. and hence, account for the age differences in recall. These notions about retrieval also fit the evidence concerning object cues. Performance may be better without objects when kinesthetic cues (e.g.. moving an arm) processed at encoding are associated with environmental cues (one's arm) that are also available at retrieval. As a n addendum to their model, action familiarity could enter in as an influential factor. When an activity is highly familiar, the regeneration of cues at retrieval may proceed more easily for older adults. When items are less familiar (that is, the general context is one of performing an unfamiliar action), item-specific information that could enhance recall may be less likely to be processed by older adults and/or may be less readily available to the older adult at retrieval. This hypothesis has not had sufficient study.
26
NorrLs and West
Can it be said that older adults' secondary memory deficits are limited to verbal recall tasks? Even though the pattern of age changes in activity memory has not always conformed to that observed with verbal recall, the evidence thus far does not support such a strong statement. It is important to note that studies of activity memory have often validated earlier evidence obtained with more traditional paradigms. For instance. age differences in memory are less apparent when older subjects are induced to use organizational strategies while learning verbal materials (Hultsch & Deutsch, 1981). This finding is essentially replicated in the activity memory literature. List length effects can occur with SPTs (RL. Cohen, Sandler, & Schroeder. 1987) a s well as with word recall. Changes in stimulus characteristics -- bimodal presentation of sentences (Biickman, 1986) or use of a meaningful encoding context with spatial infomation (Waddell & Rogoff, 1981) have provided supportive guidance for the information processing of older adults and permitted them to compensate for encoding deficits, just as the multimodal properties of SpTs provide supportive guidance for compensation in a similar manner (Backman, 1984). Age differences in activity memory performance appear to vary a s a function of primary and secondary memory demands. just a s they do with word recall (Fozard, 1980). Traditional verbal recall studies consistently demonstrate that older adults perform worse than younger adults when the test conditions do not provide adequate cues (Perlmutter & Mitchell, 1982). and the activity memory literature also demonstrates the value of cuing for older adults. At the same time, activity memory research provides u s with a paradigm to expand our understanding of memory and aging. The relevant stimulus properties are not the same for word recall and activity recall. Activity memory studies have provided new data on the influence of multimodal cues: the investigation of modality cues is an important area of research which the traditional verbal recall literature cannot readily address. Activity memory also represents a task which is not influenced in the same way as verbal recall is influenced by strategic processing. In addition, activity memory research has valuable practical significance since memory for activities has been identified as a common problem for elderly
--
Activity Memory
27
adults. Thus, it is especially useful for the study of memory and aging. Investigations of activity memory may continue to shed new light on age-related changes in cognitive processing. ACKNOWLEDGMENTS
This work was supported by a grant from the National Institute of
Mental Health (MH09542)and a grant from the National Institute on Aging (AGO6014). REFERENCES
Atkinson. R. C.. & Shiffi-in. R. M. (1971).The control of short-term memory. S c i e n t f l American. 224,82-90. Backman. L. (1984).Age differences In memory performance: Rules and exceptions. Unpublished doctoral dissertation. University of Umea, Sweden. B~ckman,L. (1985).Further evidence for the lack of adult age differences on free recall of subject-performed tasks: The importance of motor action. Human Learning,4,79-88. B ~ c h a nL., (1986). Adult age differences in cross-model recoding and mental tempo, and older adults' utilization of compensatory task conditions. Experimental Aging Research 12, 135-140. Backman, L., & Nilsson. L-G. (1984).Aging effects in free recall: An exception to the rule. Human Learning. 3,53-70. BBckman, L., & Nilsson. L-G. (1985).Prerequisites for lack of age differences in memory performance. Experimental Aging Research 11.67-73. B~ckmanL., Nilsson, L-G.. & Chalom, D. (1986).New evidence on the nature of encoding action events. Memory & Cognitioa 14. 339-34.6. Botwinick, J., & Storandt. M. (1980).Recall and recognition of old information in relation to age and sex. Journal of Gerontology. 35.70-76. Burke, D.M., & Light, L. L. (1981).Memory and aging: The role of retrieval processes. Psychological Bulletin, 90. 5 13-546. Cavanaugh, J. C., Grady, J. G., & Perlmutter, M. (1983).Forgetting and use of memory aids in 20- to 70-year olds' everyday life. International Journal of Aging and Human Development. 17. 113-122. Chaffin. R., & Herrmann. D. J. (1983). Self reports of memory abilities by old and young adults. Humn Learning, 2,17-28.
28
No&
and West
Cohen, G.,& Faulkner. D. (1989).Age differences in source forgetting: Effects on reality monitoring and on eyewitness testimony. Psychology and Aging. 4, 1-8. Cohen, R. L. (1981). On the generality of some memory laws. Scandanavian Journal of Psychology, 22, 267-281. Cohen. R. L. (1983).The effect of encoding variables on the free recall of words and action events. Memay & Cqgnition, 6. 575582. Cohen. R L. (1984).Individual differences in event memoqy: A case for nonstrategic factors. Memory & Cognition, 12. 633-641. Cohen, R L. (1985).On the generality ofthe laws of memory. In L.G. Nilsson & T. Archer (Eds.). Perspectives on learning and memory (pp. 247-277).Hfflsdale, NJ: Erlbaum. Cohen, R. L. (1988). Metamemory for words and enacted instructions: Predicting which items will be recalled. Memory & Cognition. 16.452-460. Cohen, R. L., & Bean, G. (1983).Memory in educable mentally retarded adults: Deficit in subject or experimenter? Intelligence. 7.287-298. Cohen, R. L.. Petersen. M.. & Mantini-Atkinson, T. (1987). Interevent differences in event memory: Why are some events more recallable than others? Memory & Cognition, 15, 109118. Cohen, R. L., Sandler. S. P.. & Schroeder, K. (1987).Aging and memory for words and action events: Effects of item repetition and list length. Psychology and Aging, 2,280-285. Cohen, R L.. & Stewart, M. (1982). How to avoid developmental effects in free recall. Scandanavfan Journal of Psychology. 23. 9-16. Craik, F. I. M. (1986).A functional account of age differences in memory. In F. Klix & H. Hagendorf (Eds.),Human memory and cognitive capabilities: Mechanisms and performances (pp. 409422). North-Holland: Elsevier Science Publ. Dick, M. B., & Kean, M. C. (1989).Memory for action events in Alzheimer-type dementia: Further evidence of an encoding failure. Brain and Cognition 9,71-87. Ekstrom, B. R., French, J. W.. Harman, H. H., & Deman. H. (1976). Manual for kit offactor-referenced cognitive tests. Princeton, NJ: Educational Testing Service. Erickson. R. C., Poon, L. W.. & Walsh-Sweeney, L. (1980).Clinical memory testing of the elderly. In L. W.Poon, J. L. Fozard, L. S. Cermak, D. Arenberg, & L. W. Thompson (Eds.), New dfrections in memory and aging: Proceedings of the George A. Talland Memorial Conference (pp.379-402).Hillsdale, NJ: Erlbaum .
Actfulty Memory
29
Fozard, J. L. (1980). The time for remembering. In L. W. Poon (Ed.), Aging in the 1980s (pp. 273-290). Washington, DC: Amerlcan Psychological Association. Glanzer, M. (1972). Storage mechanisms in recall. In G.H. Bower (Ed.). The psychology of learning and motivation Wol. 5. pp. 67193). New York Academic Press. Glover, J. A., TLmme. V.. Deyloff. D., & Rogers, M. (1987). Memory Journal of Educational for student-performed tasks. P ~ y ~ h o l o g79.445-452. y, Guttentag. R E.. & Hunt, R. R (1988). Adult age differences in memory for imagined and performed actions. Journal of Gerontology: Psychological Sciences. 43. 107-108. Hashtroudi. S..Johnson, M. K., & Chrosniak. L. D. (in press). Aging and qualitative characteristics of memories for perceived and imagined complex events. Psychology and Aging. Helstrup, T. (1986). Separate memory laws for recall of performed acts? Scandinavian Journal of Psychology. 27. 1-29. Helstrup. T. (1987). One, two. or three memories? A problemsolving approach to memory for performed acts. A c t a PSyChOlOgica. 66. 37-68. Hultsch. D. F. (1969). Adult age differences in the organization of free recall. Developmental Psychology. 1, 673-678. Hultsch, D. F.,& Deutsch, F. (1981). Adult development and aging: A lge span perspective. New York: McGraw-Hill. Hultsch, D. F., & Dixon. R. A. (1983). The role of pre-experimental knowledge in text processing in adulthood. Experimental Aging Research 9. 17-22. Kausler, D. H. (1982). Experimental psychology and human aging. New York: John Wiley. Kausler. D, H. (1985). Episodic memory: Memorizing performance. In N . Charness (Ed.), Aging and human performance (pp. 101141). Chichester, England: Wiley. Kausler. D. H.. & Hakami, M. K. (1983). Memory for activities: Adult age differences and intentionality. Developmental P~y~hOlogy, 19.889-894. Kausler, D. H.. & Lichty. W. (1988). Memory for activities: Rehearsal-independence and aging. In M. L. Howe & C. J. Brainerd (Eds.), Cognitive development in adulthood (pp. 93131). New York Springer-Verlag. Kausler. D. H.. Lichty, W., & Davis, R. T. (1985). Temporal memory for performed activities: Intentionality and adult age differences. Developmental Psychology, 21. 1132-1138.
30
Norrls and West
Kausler. D. H., Lichty, W.. & Freund, J. S. (1985). Adult age Werences in recognition memory and frequency judgments for planned versus performed activities. Developmental P ~ y c h o w21,647-654. , Kausler, D. H., Lichty, W., Hakarni. M. K.. & Freund. J. S. (1986). Activity duration and adult age differences in memory for activity performance, Journal of Gerontology, 1 , 80-8 1. Kausler, D. H., & Phillips, P. L. (1989). Instructional variation and adult age differences in activity memory, Experimental Aging Research, 14, 195-199. Lichty, W., Bressie, S., & Krell, R (1988). When a fork is not a fork Recall of performed activities a s a function of age, generation and bizarreness. In M. M. Gruneberg, P. E. Morris, & R N. Sykes (Eds.). Practical aspects of memory: Current research and bsues, [vol. 2, pp. 506-511). Toronto: John Wiley & Sons. Lichty. W.,Kausler, D. H., & Martinez, D. R. (1986). Adult age differences in memory for motor versus cognitive activities. ExperimentalAging Research. 12.227-230. Neisser. U. (1978). Memory: What are the important questions? In M. M. Gruneberg, P. E. Morris. & R. N. Sykes (Eds.), Practical aspects ofmemoy (pp. 3-24). London: Academic Press. Nilsson, L-G., & Cohen, R L. (19881. Enrichment and generation in the recall of enacted and non-enacted instructions. In M. M. Gruneberg, P. E. Morris,& R N. Sykes (Eds.),Pmcticd aspects of memory: Current research and issues. (Vol. 1, pp. 427-432). Toronto: John Wiley & Sons. Norris, M. P., & West, R L. (1988, August). Age dinerences in activity memory. Paper presented at the meeting of the American Psychological Association, Atlanta. Norris. M. P., & West, R. L. (1989. July). When do older adults recall perjiormed actions as well as younger adults? Paper presented at the Conference on Action Memory, Toronto. Padgett. R J.. & Ratner, H.H. (1987). Older and younger adults' memory for structured and unstructured events. Eqerimental Aging Research. 13,133-139. Park D. C.. Smith, A. D., Morrell. R W..Puglisi, J. T., &Dudley, W. N. (1987, May). Eflects ofperceptual and semantic contextual integrations ofpicture memory in the elderly. Paper presented at the Cognitive Aging Conference, Atlanta. Perlmutter. M., & Mitchell, D. B. (1982). The appearance and disappearance of age differences in adult memory. In F. I. M. Craik & S . Trehub (Eds.),Aging and cognitiueprocesses (pp. 127144). New York Plenum Press.
Actfufty Memory
31
Poon, L. W. (1980).A systems approach for the assessment and treatment of memory problems. In J. M. Ferguson & C. B. Taylor (Eds.). The comprehensive handbook of behavioral medlclne (Vol. 1, pp. 191-212).New York Spectrum. Poon. L. W. (1985). Differences in human memory with aging: Nature, causes, and clinical implications. In J. E. Birren 81K. W. Schaie (Eds.). Handbook ofthe psychology ofugfng (2nd ed.. pp. 427-462).New York Van Nostrand Reinhold. Poon, L. W..& Fozard, J. L. (1978). Speed of retrieval from longterm memory in relation to age, familiarity, and datedness of information. Journal of Gerontology, 33. 71 1-717. Poon, L. W.. Walsh-Sweeney, L., & Fozard, J. L. (1980).Memory skill training for the elderly: Salient issues on the use of imagery mnemonics. In L. W. Poon. J. L. Fozard, L. S.Cemak, D. Arenberg. & L. W. Thompson (Eds.), New directions in memory and aging: Proceedings of the George A. Talland Memortal Confeence (pp. 461-484).Hfflsdale. NJ: Erlbaum. Ratner, H. H.. & Bushey. N. (1984,August). Adults’ event recall: Putting memory in context. Paper presented at the meeting of the American Psychological Association. Toronto. Rundus. D.. & Atkinson, R. C. (1970).Rehearsal processes in free recall: A procedure for direct observation. Journal of Verbal Learning and Verbal Behavior, 9, 99-105. Saltz. E. (1988).The role of motoric enactment (m-processing) in memory for words and sentences. In M. M. Gruneberg, P. E. Morris, & R. N. Sykes (Eds.). Practfcal aspects of memory: Current research and tssues (Vol. 1. pp. 408-414).Toronto: John Wiley & Sons. Saltz, E., & Dixon. D. (1982).Let’s pretend: The role of motoric imagery in memory for sentences and words. Journal of Experimental Child Psychology. 3 4 , 77-92. Salk, E., & Donnenwerth-Nolan, S. (1981).Does motoric imagery facilitate memory for sentences: A selective interference test. Journal of Verbal Learning and Verbal Behavior. 20.322-332. Waddell. K. J., & Rogoff, B. (1981).Effect of contextual organization on spatial memory of middle-aged and older women. Developmental Psychology. 17, 878-885. West, R. L. (1986).Everyday memory and aging. Developmental Neuropsychology, 2.323-344.
This Page Intentionally Left Blank
Aging and Cognition: Knowledge Organization and Utilization
Thomas M . Hess (Editor) 0Else~k Sden~eP L c b U ~ h e rB.V. ~ North-HOllandl. 1990
CHAPTERTWO
THE ASSESSMENT OF QUALITATIVE AGE DIFFERENCES IN DISCOURSE PROCESSING Elizabeth A. L. S t h e and Arthur Wingfield Brandeis University
SUMMARY It is generally the case that older adults remember less from what they read and listen to than do their younger counterparts. The question remains open, however, as to whether these quantitative age differences are also accompanied by qualitative differences. We address this problem in two parts. First, the issue of whether younger and older adults remember different kinds of information is considered by reviewing data on age Werences in the levels eft”ect (the tendency for gist information to show a higher probability of recall than detail) as well a s by examining results from relative memorability analyses (in which the probabilities of recall of individual text units by young and old are directly compared). Second, the issue of whether younger and older adults depend on different sorts of contextual or linguistic cues, or rely on different processing strategies in understanding and remembering discourse is considered by examining evidence for four different hypotheses of qualitative age differences in discourse processing. It is argued that while the qualitative structure of language processing mechanisms is preserved across age, working memory limitations that accompany advanced age may in themselves result in a qualitatively different balance of these processes, or they may necessitate that certain processes take precedence in later adulthood.
34
StineandWinQpeld
Quantitative memory declines in later adulthood are welldocumented (cf. Burke & Light, 1981). even when the to-beremembered materials are a s ecologically valid a s spoken and written discourse (e.g., Cohen, 1988; Stine. Wingfield. & Poon. 1989). There remains considerable controversy, however, as to whether these quantitative declines are accompanied by qualitative changes as well. In other words, older adults frequently remember less, but do they remember different19 Our goal here will be to explore this question, by suggesting how one might go about answering it, as well as a by providing a current position on the basis of extant data. QUANTITATIVE VERSUS QUALITATIVE AGE DWFERENCES By way of defining the research problem, we begin by first considering what is meant (perhaps, not so obviously) by the contrast between a quantitative and a qualitative difference. According to Webster's New World Dfctionary, to quantify is 'Yo determine or express the [amount] of; [to] measure," and to qualify is "to describe by giving the [characteristic elements, attributes] or characteristics of." Conceptually then, a quantitative difference is a difTerence in amount, whereas a qualitative difference is a Werence in kind. It would be misleading, however, to assume that a qualitative analysis precludes the use of quantitative measurement (to the contrary, we promise to use numbers extravagantly). Consider the importance of measurement in assessing the quality of an empirlcal object or a theoretical system. If we assume that the quality of a n entity may be expressed as a collection of features (e.g., Dersky, 1977),each one having some scale of measurement, then we may express this quality as a vector of quantities corresponding to the feature set. Minimally, such a vector of featural measures would consist of zeroes (i.e., "feature is absent") and ones (i.e.. "feature is present"). So we could say that two things are qualitatively different if one contained one or more features which the other did not, or in other words, if the patterns of zeroes and ones in their feature vectors were different. Such a pattern might emerge if we were to model the cognitive processing for a given task as a collection of processing mechanisms.
Suppose, for example, that we could represent the task of remembering a list of words a s (ITEM ENCODING, INTENTEM CHUNKING, CHUNK RETRIEVAL, ITEM RETRIEVAL). If we presented a list of apparently unrelated words (e.g.. frog, grass, train, bell. .I to two groups of subjects, giving the Experimental Group an organizational orienting task (e.g.. to categorize the list into things that are green, things that make noise, etc.) that we did not give to the Control Group, we would find that the Experlmental Group would remember more words than the Control Group (cf. Einstein h Hunt, 1980, Experiment 2). We might attribute this recall Merence to qualitative differences in processing. in which the processing of the Experimental Group consists of all four processing components (and hence, represented a s ( 1 . 1 , l . l ) ) and the processing of the Control group consists of only two of the four he., (1.0.0.1)). At a slightly more sophisticated level, the values in the feature vector could express how much of a feature were present. For example, we could introduce a second Experimental Group who received the organizational cues, but who were also subject to some condition (e.g.. sleep deprivation) which adversely affected all processes equally. We might represent the efficiency of their processing operations as (.8,.8,.8,.8), So this group would differ quantitatively but not qualitatively from the first Experimental Group. Thus, we could say that two things are qualitatively different if one contained relatively more or less of a feature than another, or in other words. if the values of their feature vectors were not scalar multiples of one another. So for example, in our organization experiment, the Control Group was probably engaging in some kind of (subjective) organizational processing, so w e might more accurately model their processing operations as something like (1*.2,.2,1). The point here is that questions about quality are inevitably reducible to questions about measurable quantities. This, by no means, trivializes our original goal of exploring qualitative age differences, but rather facilitates it. This will be our approach in untangling the different meanings of "qualitative age differences." It seems to u s that one of two things is typically intended by asking whether there are qualititative age differences in discourse
36
StlneandWiq@ld
memory. First, we could be asking about the extent to which younger and older adults remember the same kinds of information. In this case, the elements of our feature vectors would represent particular pieces of information (e.g.. propositions or story nodes), or particular kinds of information (e.g., main points or minor details), and our task would be to compare the feature vectors of young and old. This is an empirical issue. If we choose to classify these items by type, this classification may be theory-driven, but having done this, whether or not the recall protocols of younger and older adults contain the same kinds of information is simply an empirical fact. The existence of qualitative dnerences in recall may or may not co-occur with quantitative Merences. For example, it may be that even though older adults remember less quantitatively, their qualitative recall could be virtually indistinguishable from that of younger adults. That is. older adults remember less, but the deficit may be distributed equally among different kinds of information contained in the text. On the other hand, it may be that the quantitative memory declines in later adulthood are accompanied by the absence of a particular type of information in the recall protocol. A second intent of the question of qualitative age differences in discourse processing could be whether younger and older adults differ in the way they go about processing discourse. In other words, do older adults rely on different sorts of linguistic information or utilize alternative processing strategies? This is a very different question from the first. In this case, the elements of the feature vectors of young and old are cognitive processes. This contrasts with the first case in which the feature vectors contained empirical text units: here the features being compared are theoretical entities. So for example, in our word memory experiment, assuming we represent the processing of younger adults as (1~1.1.1). we might represent the hypothesis that older adults are only 80% as effective at item retrieval but otherwise as emcient as young by (1,1,1,0.8). Like the issue of quality of recall, however, the issue of quality of processing is also distinct from the issue of quantitative age difference. For example, it may be that even though older adults
Qualitative Age Dlffeences
37
remember less, the processing mechanisms in comprehending and remembering discourse may be qualitatively the same, just uniformly slower or less reliable. On the other hand, it could be that quantitative memory decline goes hand in hand with a change in the way information is processed, as would be the case, for example, if age declines in auditory or visual acuity induce a greater reliance on top-down contextual processes. To further c l a m the issues at hand, we might also point out that these two different approaches to the question of qualitative age differences in discourse memory are distinct. For instance, one might observe qualitatively similar performance between young and old, but qualitatively different processing mechanisms. In fact, Salthouse (1984) has discussed the value of a research strategy, "molar equivalence - molecular decomposition." which takes advantage of this principle. As Salthouse (1984) found among expert typists, it may be found in some cases in discourse processing that age-related qualitative change in cognitive mechanism enables qualitatively similar performance. That is. limitations in one part of the processing architecture may be overcome by the older adult who is able to develop m e r e n t ways of using that processing architecture. One corollary to this principle is that one might observe qualitatively different recall performance between young and old accompanied by qualitatively similar processing mechanisms, which could be the case if the older adult did not develop alternative processing strategies to overcome processing limitations. Thus, the existence of qualitative age differences in the empirical arena does not in itself entail the existence of qualitative age difTerences in process. This is not to say that qualitative age differences in process are inaccessible to study. As we will show, interpretation of age daerences in performance in terms of process depends upon one's model of the cognitive system. Our goal in this chapter will be to tackle these two distinct questions. First, we consider whether younger and older adults tend to remember the same kinds of information from prose. While we begin by treating this as the strictly empirical question that it is, we will be able to interpret these data in terms of modem discourse theory. Having done this, we will, in the subsequent section, consider whether younger and older adults process discourse
38
smandwingfleki
differently by entertaining four alternative hypotheses for qualitative age dnerences in processing mechanism. DO OLDER ADULTS REMEMBER DIFFERENT INFORMATION?
For the most part, the question of whether younger and older adults remember different information from discourse has been addressed by exaznining whether there is differential recall of major and minor points of a text across age. Consider the three conceivable answers to this question. The first is often the last bastion of hope for the optimist facing a literature replete with quantitative age declines: perhaps older adults remember less information, but maybe what they are forgetting is simply the trivial detail, while retaining the more important elements reflecting the gist of the discourse. This would be good news indeed, suggesting a functional adaptation to memory loss. On the other hand, the bad news of quantitative age declines could be worse: in addition to older adults remembering less information, they might be unable or less able to discriminate between the important and the trivial, producing a recall protocol with a disproportionate amount of trivial detail. A third possibility is that younger and older adults would show similar distributions of major and minor points in their recall. It is generally the case that a reader or listener is more likely to remember the more important, "higher level" points from a passage than the "lower level'' detail. This leuets effect has been found through various operationalizations of level of importance, including formal prose analysis systems as well a s experimenter or subject ratings. When applied to questions of adult age differences. the issue becomes whether older adults show a sharper discrimination between higher and lower level information than do younger adults (i.e., an "increased levels effect"), a decreased discrimination between higher and lower level information (i.e,, a "decreased levels effect"), or a pattern similar to that found for the young (i.e.. the "same levels effect"). Table 1 summarizes a number of studies which have examined this issue. Immediately apparent are the striking diversity of findings, which cannot be easily accounted for in terms of variability in prose or subject characteristics or by the way text units were analyzed. While some studies have reported
Qualitatiw Age Di@kences
39
that older adults show the same recall advantage for main ideas over detail as younger adults do, others have found that older adults are relatively better at remembering the main ideas, and still others have reported that older adults are relatively more likely to recall lower-level details. We reiterate that "relatively" is a key word here, since both younger and older adults are more likely to remember the major points of a text over the rninor detail, suggesting that both young and old to some extent organize the text, or show some appreciation of text structure. This is an important point. We know of no research at all suggesting that healthy older adults do not discriminate between major and minor points. Thus, the question is one of degree: to what extent do older adults differentiate between the more and less important elements in their recall protocols relative to their younger counterparts? To this point in the argument in discussing whether younger and older adults remember different kinds of information. we have not yet attempted to distinguish clearly between the discrimination among item types in the recall protocols and the cognitive processes underlying this discrimination. This will be addressed in the next two sections. The Discrimination among Text Elements Because we want to isolate the phenomenon before developing a theoretical interpretation, we will first adopt a purely empirical approach in discussing the discrimination among text elements in discourse. In keeping with the traditional empirical definition of discrimination, that "[aln organism is said to discrfmfnate between two stimuli when it behaves differently in the presence of each" (Reynolds, 1975, p. 45), we could say: A dlscrfmfnatfonis made between two text elements when one is remembered and the other is not. We explicitly do not intend for discrimination to refer to an active. effortful choice among units. We, of course, do not necessarily deny the existence of such processes, but to simplify our present discussion, we will use the term discrimination only in its empirical sense. A sensible working hypothesis is that a discrimination among text units occurs when these units are
Table 1 Summary of Studies Testingfor Age X Units Interaction ~~
Passage Characteristics #/
?Lpe
word Length
Modality
Units Analysis System
Subjects
Age X Units Interaction?
Constraining Condition?
Increased levels e&t for 0
...when Ss were instructed
Decreased levels e&t for 0
...when Ss were instxucted
Byrd (1985): 12aa
A
m
Listened: oral recall
Kintsch
25 Y (1826 yrs); 25 0 (64-70 yrs); matched on Mill Hill
to give verbatim recall
to Summarize
&hen (1 979): 1 N
"300
Constituent Listened facts (A 120 vd; ("S " v oral recall
40Y (1829yrs); 400 (65-95yrs); high and low educ'l
~ m z E z * * ] on lwelgrpsmatched WAIS Vac
Same levels effect for Y and 0
Decreased l m l s effect for 0
...when Y and 0 were lower in educational level
36 Y (20-39yrs); 36 M (40-57yrs):
Increased levels effect for 0
...WhenY and owere high
36&(EEoT$erbal grpsmatchedon Voc KFlu-
Decreased levels effect for 0 in vocabulary level
...when Y and 0 were lower
[y=O for highverbal;
OcY for low verbal) Dimn, Hultsch, Simon, & von Eye
6Ex
W3
M
written
...whenY and 0 were high
in educational level
(1984): Kintsch
recall
in vocabulary level
Table 1 (continued) ~~
~
Passage Characteristics #/
Type
word Length
Modality
Units Analysis System
Subjects
Age X Units Interaction?
Constraining Condition?
Decreased levels effect h r 0 & M
...but due to dflemms in
Same units effects for Y and 0
...regardless of recall
D h n , Simon, Notoak, & Hultsch (1 982): 5 Ex
"180
Reador listened; written
Kintsch
d
Ha~-ker,H&y. 4Ex
"400
& Walsh (I 982): Study
Read; oral vs. written
d
30 Y (18-32yrs): 30 M (34-56yr~); 300 (60-81yrs): matched on Voc KFRT
variability,not means
I:
Kintsch 38 Y (18-24yrs); (kinds of 32 0 (66-79 p) propositions)
modality
Hcuker, Hartley. & Walsh (1 9821: Study 2:
4Ex
"400
Read
Kintsch
24 Y (18-26yrs): students; E%276yrs1; on Shipley-Hartford Y
Same units effects hrYand0
Read:
Kintsch
12 M (41-58p); 12 0 (71-89p); matched on abridged
Same levels effect hrMandO
Kernper (1 9871: 12EX
%o
written recall
WAS voc
Decreased levels effectfor 0
...when information was presented in single, independent. or rightbranching clause
...when information was in leftgresented ranching clause
Table 1 (continued) Passage Characteristics #/
?srpe
word Length
Modality
Units Analysis System
Age x units Interaction?
Subjects
Constraining Condition?
Meyer&Ria? (1981):
1Ex
641
Read:
Meyer
written recall
16 Y (20-33 yrs): 16 M (41-55 p); 16 0 (5879yrs): matched on WAIS
Decreased levels effect for 0
148Y,166 0;
Same levels effect hrYand0
...when o>Y in vocabulary
Decreased levels effect
...when Y and 0 were
Same levels effects forYand0
...regardless of vocabulary
VOC
Meyer&Rioe (1983):
2Ex
388
Read: written recall
Meyer
samplesof50 inQukk
WViCWX used for analysis
level, regardless of absolute vocabulary level
low in vocabulary level
Petros. Nargmd, Olson & Tabor (1 989): 6 E)r vs
A210
N
Listened (120 vs. 160 vs
Ysubjects' ratings
30Y(1830yrs):
290(60-84yrs): high and low verbal
GU&FO"
2ooWprr3
level, prose type,or speech late
Petros. Tabor, Cooney, & Chabot (1 983): Experiment 1: 2 N
A400
Listened
(12Owpm): oral recall
Y subjects' latings
27 Y (18-3 1 yrs); 26 0 (62-80 yrs);
high and low verbal
p u p s matched on
WAIS voc
Unsystematic interaction
...regardless of vocabulary level
Table 1 (continued)
Passage Characteristics #/ Qpe
word Length
Modality
Units Analysis System
Age
Subjects
x units
Interaction?
ConsMning Condition?
Peiros, Tabor. Cooney, & Chabot ( I 983): Experiment LI:
4 N
A250
Listened (12OV5.
16OWpm):
Y subjects' ratings
23 Y (18-33 yrs); 1 7 0 (63-87yrs); matched on WAIS Voc
Same levels effects fOrYand0
1 4 6 Y (18-32 yrs): 117 M (40-54 yrs); 159 0 (62-80 F); Y d = O on Quick Word Test
No correlation between age and levels effect
36 Y (18-35 yrs); 360(65-82yrs)
Same units effect for Y d O
oral recall
Rioe & Meym (I 986): 2Ex
388
Smith, Rebok,
4 N
Read;
Meyer
Written recall
Smith Hall & Alvin 11983):
Listened; oral recall
Mandler
Noderecallwases. less differentiad €-KO
...when storieswere
canonical in form or completely scrambled
...when structurewas interleaved
S p i k h (I 983): 2%
A450
Read: oral recall
Kintsch
1 6 Y (18-25 yrs); 1 6 0 (68-91 yrs); 1 6 0-Im (70-90p)
Increased l m l s effect Tor 0 ...when 0 were healthy Decreased l m l s effect ... when 0 were impaired hr 0
Passage characteristics #/ ?srpe
wold Length
Modality
Units Analysis System
Subjects
Age x units Interaction?
Constraining Condition?
Same levels effects foryand 0
...whenYandOwereavrin
Increased levels effect for 0
...whenYandOwerehighin
Stine & Wingjkld ( I 98701 :
4Ik vs.
w
N
Read;
written recall
Kintsch
3oY(iamyrs): 290(60-83p) high and avr verbal u smatchedon %pLy voc
vocabulary level
vocabulary level
...espe-ciallyfor narratives Stine, WhgjkM, & Fbcm (1986):
36Ex
A17
Listened
Izaovs
Kintsch
mvs
O r a l recall
1Ex
1563
&pena-PaeZ
30 Y (1832yrs): 30 0 (64-79 p); matchedonvocd
Same levels effects forYand0
(1984):
Read;
Johnson
Read, wrote recall
Kintsch
wrote recall
Decreasedlevelseffect h0 ...especially when 0 were lower in ~cabulaxylevel ...regadbs of speech rate
F&voc j
m w P & SLU~IT, Kowalski,
24 Y fl7-32 -1: 2401 (61-&yk): 120 2 (61-82 ): ~302<01 on
7 h (1989):
6 N
vs. Ik
"300
Same levels effects forYand0
...regardleas of prose type
Table 1 (continued) Passage Characteristics #/
word Length
Modality
Units Analysis System
Age X Units Interaction?
Subjects
Constraining Condition?
Zelinski Light, & Gilewski ( I 984): Experhent I 1N
aoS
Readalong listening
Meyer
46 Y (16-18yrs); 58 0 (57-82 1; match on Verbal Meaning
Same levels effects hrYand0
65Y (21-35yrs); 58 0 (54-85yrs); matched on PMA Verbal Meaning
Same levels effects
80 Y [ 19-36yrs); 78 0 (5582yrs); o>Y on Shipley Voc
Same levels effects brYand0
...for one passage
Unsystematic interaction
...for the other
&&
(150Wpm);
wmte recall Zelinski, Light & Gilewski [ I 984): Experiment 11: 2N
A210
"
Meyer
fOrYand0
Zelinskf Light & Chewski (1984): Eqeriment 111:
2Es
w30
"
Meyer
Note: Ex = Expository,N = Narrative, Es = Essay a Each subjectwas resented half of these b Voc KFHT = V a x & h ysubtest of the Kit of Factor-Referenced Tests
46
StineandWingjleld
somehow treated daerently by the cognitive processing system to yield daerent degrees of memorability; that is, presumably, more memorable text units were those that were more effectively processed. The demonstration of discrimination among text units, however, does not rely on the demonstration of processing differences. To be sure, both younger and older adults do, indeed, discriminate among text elements, holding some in memory, perhaps for years (Rubin. 1977). while discarding others almost immediately. An analogous situation arises in employee selection. According to the 1978 Uniform Guidelines on Employee Selection Procedures, "the use of any selection procedure which has adverse impact on the hiring, promotion, or other employment or membership opportunities of members of any race, sex. or ethnic group will be considered to be discriminatory. (Section 3)" That is, adverse impact is a sufficient condition to demonstrate discrimination, regardless of the employer's intent to exclude the members of any particular group. Thus. we could adopt an "adverse impact principle" for discrimination in memory: the use of any cognitive process which has adverse impact on the memorability of any text units or any particular types of text units will be considered
...
discriminatory.1
We could carry this analogy further. Of course, we want a n employee selection procedure to discriminate among individuals; specifically, we would like it to function so as to pick out those who could best perform the job for which they are being selected. Consequently, any employee selection procedure is validated against some criterion measure of job performance. We know the employee selection procedure is working when it selects those who perform well on the criterion measure and rejects those who perform poorly. Similarly, we would want our cognitive system that is processing discourse to discriminate effectively among pieces of Ignoring the importance of the empirical fact of discrlmination, the Supreme Court has ncently ruled (Wards Cove Packing v. Atonia) that disparate impact of a selection procedure requires that the employee "isolate and identify the specific employment practices that are allegedly responsible for any observed statistical disparities." Thus. our analogy breaks down. While mgulators of equal employment opportunity m a y in principle have direct access to the "process"that produces empirical descrimtnation(employment practices), the cognitive researcher does not.
Qualitative Age Di@kemes
47
information, picking out those that are best suited for the job for which they are being selected and disgarding others. It is at this stage of the logic that difficulty arises. What is the job for which information is being selected by the cognitive system? What is the appropriate criterion against which to validate selections made by the cognitive system? The Levels Analysis and Qualitative Age Differences in Memory By far and away, the most common criterion used in discourse memory research for evaluating the effectiveness of the selections made by the cognitive system is level of importance. In order to appreciate the nature of this criterion, we must have a more thorough understanding of current views of discourse processing. How Mscourse Is Processed
While several viable models for discourse processing exist in the modem cognitive literature (Aaronson & Scarborough, 1977; Jackson & McClelland, 1978: Just & Carpenter, 1980; Kintsch & van Dijk. 1978; Kintsch. 1988). there is remarkable agreement with respect to the basic processing operations which are involved in understanding and remembering discourse. In general, it is agreed that the comprehension and memory for discourse depends upon a hierarchical array of component processes. More microlevel processes such as the activation of orthographic or phonological code enable access to word meanings, which in turn enables meaning to be constructed at the clausal or sentence level. Processes may become automatic at any level, though it is probably more likely for the more microlevel components because of extensive use of these processes in many contexts (LaBerge & Samuels, 1974). A core assumption of any discourse processing model is that language must be processed within a flnite working memory (WM).
(or "buffer" or "short-term store"). While the verbatim format of the idormation is to some extent retained (Jarvella, 1971; Glanzer. Dorfman. & Kaplan. 1981). the primary function of this buffer in discourse processing is probably to recode the information into a
48
stine and wtngfleld
semantically-based form of representation (Kintsch & van Dijk, 1978). Because the capacity of this buffer is necessarily limited, one of the essential tasks for the human discourse processor must be the effective allocation of processing resources across text units. In other words, except under very special circumstances, the reader or listener simply cannot remember everything, thus requiring a discrimination to be made. One important goal of any text processing model then is to predict which text units will be effectively processed within WM,and therefore, remembered. The model of Kintsch and van Dijk (1978;Kintsch. 1988) is perhaps among the most explicit in terms of describing the theoretical sequence of online operations occurring in WM. In this model, discourse is processed in cycles on a segment-by-segment basis (Jarvella, 1971;Aaronson & Scarborough, 1977). As each segment is input into WM,the phonological or orthographic stream is recoded into a set of propositions (or idea units). Propositions are typically defined in terms of containing a relational term (or "predicate")and one or more concepts to be related (or "arguments") (Kintsch & van Dijk, 1978; Turner & Greene, 1978; Miller & Kintsch, 1980).although propositions are sometimes conceptualized a s being even more primitive than this (Graesser. 1981;Kintsch, 1988). It is also within this W M buffer that interrelationships among these propositions are established. Represented in terms of a hierarchical structure called a "coherence graph," connections among propositions are made in the following way. Within the first input cycle, the highest level proposition is selected. Theoretically, this proposition is the one that is the most important or the one most strongly related to the theme of the passage. Other propositions within this cycle are connected to the Level 1 proposition on the basis of a shared propositional argument. It has recently been suggested (Kintsch. 1988) that it is not only the direct linguistic input that is represented, but that there is also an online elaboration of this input such that closely associated information may be activated as well. Some subset of this information remains activated in W M as the Discourse is said to be next segment is processed. "comprehensible" when the "new" information from the current
QualUativeAge Di@rences
49
input cycle can be integrated into the "old' information already held there. Propositions are most likely to be held over until the next cycle if they are important (or in terms of the Kintsch and van Dijk model, "higher in the coherence graph') or if they are recent (i.e.. along the lower edge of a coherence graph branching from left to right). Thus, within this model, the principle by which the processing system discriminates among propositions is via this "leading edge strategy." There has been much evidence accumulated for this general processing model. For example, all other things being equal, reading time for a text has been found to increase linearly with the number of propositions (Kintsch & Keenan, 1973; Haberlandt. 1984). Especially relevant to our purposes here is the finding that propositions higher in the coherence graph have a higher probability of recall than those lower in the coherence graph. This levels effect is predicted by the theoretical notion that these items receive more processing cycles in W M (Kintsch & van Dijk, 1978). Online measures also support such a model. Reading times, for instance. are not distributed evenly across words of a text, but rather reflect both lower (e.g., number of letters) and higher (e.g., number of propositions, story structure) levels of analysis (Graesser & Riha. 1984; Haberlandt, 1984;J u s t & Carpenter, 1980). In particular, readers often show relatively longer latencies at the ends of sentences and just before clause boundaries; these sharp increments in reading times, or "peaks," lend support to the idea that the discourse is segmented and integrated in cycles. Also, while reading, subjects are able to more accurately and quickly verify concepts that would have theoretically been held Over in Wh4 than those that would not have been (Fletcher, 1981. Exp't 2; McKoon & Ratcliff, 1984). Thus, there is much support for this kind of processing architecture, and especially the important role played by a limited-capacity W M buffer in selecting out certain elements for extensive processing while ignoring others. It is also the case, however, that the precise nature of this processing (at least, parametrically) is not thoroughly understood. In particular, there is evidence that the simplirjring assumption of a leading edge strategy may be overly simple. For example, propositions tend to be retained if they are thematically related to
50
Stine and WLn5lfZel.d
the passage even when they are not necessarily related to the "old information via an explicit shared argument (Fletcher, 1986). Data such as these highlight the importance of modeling global coherence in addition to local coherence of discourse. Even though macrostrategies used to establish such a global organization of text have been explored in some depth (Kintsch & van Dijk. 1978; van Dijk & Kintsch, 19831, the strategies by which readers apply such schematic knowledge are complex (cf. Kintsch & van Dijk, 1978, p.373). Thus, our understanding of exactly how the cognitive system discriminates among text elements is as yet incomplete -- at least to the extent of providing a well-specified algorithm for discriminating among propositions to be retained for further processing. D@lcultiesin Using the Levels Analysis to Assess Age Oiffeences
Given the importance of the levels effect in lending support for the role of W M in discourse processing, one might expect that a comparison between the levels effects of young and old would be easily interpretable into a comparison of processing mechanisms. While the levels effect in itself has been a reliable result, there are several reasons why it may be difficultto assess how it is moderated by age, In particular, caution is needed in the use of propositional level as a measurement device, especially as it is used to provide a criterion against which to assess the effectiveness of discrimination among text units by younger and older adults.
Even though the processes for constructing a coherence graph have to a large degree been operationalized. there is still considerable subjective judgment involved. First, a procedure for the selection of Level 1 propositions has never been well-articulated, and often depends on intuition (Miller & Kintsch. 1980). Efforts to operationalize level in terms of the extent of a proposition's connectedness to other propositions have suggested that while a proposition's degree of relatedness contributes to its memorability. level is not simply relatedness (Manelis, 1980). The selection of the
QualitativeAge Differaces
51
Level 1 proposition to a large extent determines the whole structure of the coherence graph (cf.Miller & Ktntsch. 1980).so the reliability of this judgment is important for making accurate predictions of the recall probabilities of all text units. Having selected a Level 1 proposition, it is possible that coherence graph construction may be completed in a strictly rulegoverned way by connecting propositions that share a propositional argument. It is often the case, however, that among the set of propositions at an input cycle, there are several candidates sharing an argument with the propositions retained in WM. When this happens, the experimenter must again rely on intuition. It is conceivable -- indeed, probable -- that text representation (and in particular, perceived importance of information) depends on many situational factors, such as the perspective of the reader or listener (Pichert & Anderson, 1977).and how much the reader or listener knows about the topic (voss & Bisanz, 1985). Thus. the "situational representation" of discourse (van Dijk & Ktntsch, 1983) can make it dimcult a priori to exactly determine a single text base structure, and hence, assign propositional level. Tyze Input Buffer Size Problem
Because the coherence graph is constructed in cycles, variability
in the number of propositions processed at once could influence its structure. The size of the input buffer is thought to show some flexibility in response to the processing demands of the text (e.g., Miller & Kintsch. 1980). Because segmentation sites between input cycles are probably largely determined by syntax (e.g.. Aaronson & Scarborough. 1977:Wingfleld & Butterworth. 1984).such flexibility would be necessitated for comprehension of complex embeddings. for example, left-branching sentences. Additionally, individual difTerences in the capacity of W M (cf. Daneman & Carpenter, 1983) probably have some impact on the number of propositions that could be processed at once.
52
Stine and Wtngjkld
The Contextual Problem
The Objectivity Problem and Input Buffer Size Problem are the consequences of the discourse processing system being sensitive to contextual and individual difference factors. Note that these "problems" only present dimculties for the researcher using level as a measurement tool, and not for the model a s originally conceived (e.g., Kintsch & van Dijk, 1978). The coherence graph is not in the text waiting to be discovered, but rather is a n individual construction; in other words, it is not so much "in the text" a s it is "in the head." Deviation of one group (e.g., the elderly) from the predicted recall pattern, therefore, does not necessarily suggest less "sensitivity to prose structure," but instead that the model-specified operations for constructing the coherence graph are not complete, Therefore. the logic of constructing a single coherence graph for the sake of defining levels of importance and then comparing the recall performance of two groups as a function of this variable is questionable. The Bins Problem
Another factor that makes it difficult to compare the levels effect across age arises out of the fact that there are typically unequal numbers of propositions at the difTerent levels in a coherence graph's hierarchical structure. This is particularly a problem with the crucial Level 1 propositions which tend to be very few in number. Because reliability of a measurement instrument increases with the number of component items (Lord & Novick. 1968). our ability to assess memory for important ideas, which may rest on what essentially amounts to a one- or two-item test, is limited. In order to avoid this problem, the investigator often combines bins together. Because this process is inevitably somewhat arbitrary and dependent upon the text, it could itself change the steepness of the levels effect.
Qualitative Age Dflmences
53
The Floor and Cellfng Problem If we assume for a moment the simplest case that the probability of recall is determined by level and that level influences recall similarly
for young and old alike (i.e.. the same levels effect for young and old), we could effectively "create" a n Age X Levels interaction of either form by sufficiently changing the difficulty of the task. For example, we could "demonstrate" an increased levels effect for older adults by simply raising the level of performance (e.g., by presenting shorter texts, using slower rates, or testing smarter subjects) to the point that younger adults were at ceiling. Similarly. we could "demonstrate" a decreased levels effect for older adults by decreasing the level of performance to the point that older adults were at floor. Thus, it is possible that any Age X Levels interaction is a n artifact of not having a measurement device with sufficient discriminative power at the high or low end. This is not devastating, since we can evaluate the extent to which this is the case. We would know we had a "ceiling effect" if (a) the means were at relatively high values, (b) the variability among the scores were reduced, and (c) the distribution of scores was negatively skewed, or a "floor effect" if (a) the means were at relatively low values, (b) the variability among the scores were reduced, and (c) the distrlbution of scores was positively skewed (cf. Alliger, Hanges. & Alexander, 1988). Should this occur, this does not necessarily mean that we have conducted a "failed experiment." If we are willing to make certain distributional assumptions about the data, there are techniques available for recovering some of the information lost at floors and ceilings. Additionally, defining conditions under which performance is invariably high or invariably low can be interesting in itself, though the levels analysis may not be well suited for this purpose. The Scale of Measurement Problem A final problem for making group comparisons of the levels effect is
that it may effectively be asking an "interval" question of "ordinal" data. Given that a numerical scale is meaningful only to the extent that it represents empirical events (cf. W. Stine. 19891,level in a
54
Stine and Wingfleld
coherence graph is at best mean.in@ul as an ordinal scale. That is. the levels effect has been robust enough to suggest that level is indeed a meaningful way to assign numbers and has a measurable effect on behavior (yet we should point out that this effect is not always monotonic, e.g., Stine, Wingfield, & Poon, 1986; Zelinski, Light. & Gilewski, 1984). There is no reason to assume, however, that the interval between any two adjacent levels is equal to the i n t e n d between any other two adjacent levels. Nor can we assume that the distances between Level 1 and Level 2 propositions, for example, are in any sense equal from story to story. Finally, and particularly important to the point here, is that we cannot assume that the distances between Level 1 and Level 2 propositions, for example, are equal for young and old. In fact, if level impacts on the probability of an item being carried over from cycle to cycle, and hence, to the number of processing cycles through which a proposition is carried, then there could well be disproportionate differences for young and old if the size or character of the input buffer varied across age. When we ask whether or not the levels effect changes with age, we are asking a question about changes in the slope of a line. If the distance between the points on the x-axis are not meaningful (as is the case for an ordinal scale), then neither is slope. Thus, because the distances among levels are not meaningful (in a measurement theoretic sense), neither is any statement about the steepness of the levels effect. What the Levels EffectDoes Reveal about Qualttattue Age Dfferences
For all of the reasons cited, it is perhaps not surprising that the literature on age differences in the levels effect has produced less than a consistent picture. The levels analysis was derived as a way of testing a theory of discourse processing, and as such, it has been an exceptionally valuable tool. As we have seen, however, the leap from this use for which it was developed to a test for sensitivity to prose structure is fraught with problems of measurement. On the other hand, we do not wish to throw out the proverbial baby with the bathwater. There are, at least, a couple of positive generalizations which can be made from the numerous studies
QualitativeAge D@kences
55
which have examined the text units effects in later adulthood. First. just as younger adults, older adults are generally reported to display a levels effect across the wide range of experimental conditions and individual differences shown in Table 1. This consistent finding provides strong evidence that the basic processing architecture for understanding and remembering discourse described earlier is as applicable to the old as it is to the young. To the extent that we expect to find age differences in discourse processing, then, it is likely that they will reside in parametric variation of a single model, rather than in a very different set of processing mechanisms. While all of the measurement problems outlined above make it impossible to give an absolute answer to the question of whether older adults show a levels effect similar to younger adults, there are some striking empirical regularities that emerge from Table 1. Specifically, older adults often show a relative decrease in the levels effect (or relatively less differentiation among units) a s the conditions of text processing become more demanding. Our original question was implicitly phrased in terms of a twoway interaction: is there an Age X Units/Levels interaction? To the extent that there is regularity in the data. it lies at the level of the three-way interaction: when processing requirements become more demanding (either because of individual differences in processing capability. or because of instructional or text demands), older adults are relatively less likely to differentiate among text units. Thus, older adults may show the same units effects as young under less demanding situations, but a depressed levels effect under greater demands. Similarly, if the older adult shows a n increased levels effect under less demanding situations, their units effect may be more similar to younger adults under more demanding conditions. This trend has been noted when subjects are asked to produce summaries. which require active restructuring of the material. rather than verbatim recall (Byrd. 1985).when subjects are lower in verbal ability (Cohen. 1979, Experiment 3; Dixon et al.. 1984; Meyer & Rice, 1983: Stine & Wingfield. 1987a). when syntactic structure increases processing load (Kemper. 1987). when events of a narrative are interleaved so as to require online organization (Smith et al.. 1983). and when older adults are in poor health (Spilich,
56
Stine and Wingfleld
1983). A similar obsewation has been made by Hultsch and Dixon (1984)who have argued that age differences in the “discovery and utilization of the organization of the text are attenuated when the text is well organized and exacerbated when it is poorly organizedf [P. 92). We note that there are some studies which do not conform to this generalization, showing no change in the relative units effects as a function of verbal ability or text demands (e.g.. Petros et al.. 1983;Petros et al.. 1989;Tun, 1989). To our knowledge, there are no studies showing the opposite trend (i.e., with older adults showing a relative increase in unit discrimination with increased task demands). In the next section we discuss an alternative methodology for making age comparisons of units effects, in the hope that it will facilitate the study of this problem. The approach is fundamentally an empirical one. It is not that we are trying to avoid theoretical explanation; rather without reliable findings, a phenomenon does not yet exist to be explained.
Relative Memorability and Qualititive Age Differences in Memory If the empirical question is whether younger and older adults are likely to remember the same text elements, then the most straightforward approach would be to directly compare how memorable different text elements are for young and old. In terms of the analysis of feature vectors described earlier, our goal will be to compare the feature vectors of memorabilities for younger and older adults. These feature vectors contain one element for each text unit under consideration, representing how memorable that text unit is for that group in terms of probability of recall. Thus, a comparison of these vectors will be strictly data-driven. The Text Element as the Unit of Analysts
Previous research considering the text unit a s the unit of analysis has been fruitful (Rubin, 1978). Properties of individual text units are such powerful determinants of recall that it prompted Rubin (1985)to suggest that much recall data can be explained by a
QualitativeAge D@f?rences
57
simple fixed rank ordering among units. According to this ffxed rank order hypothests,the relative ordering of recall probabilities of (text)units is fixed, and does not substantially vary as a function of contextual and individual difference variables. With the possible exception of cohort-specific materials (e.g., Barrett & Wright, 1981; Hultsch & Dixon, 19831,there is evidence that this fixed rank order hypothesis holds in the comparison between younger and older adults. For example, Rubin (1978,Experiment 7)showed that even though older adults recalled a smaller proportion of text units from a narrative than their younger counterparts, the correlation between the unit recall probabilities for the two groups was .85. Similarly, Mandel and Johnson ( 1984) found significant concordance between unit recall probabilities and unit importance ratings of younger and older adults. The direct comparison of text unit memorabilities is free of many of the problems of comparing levels effects. While the levels analysis has sources of unreliability in both construction of the coherence graph and in the scoring of the recall protocols, the relative memorability analysis admits only the possibility of unreliability of scoring. However, because interrater agreement is typically very high, the relative memorability analysis would be expected to produce more reliable results: it is. therefore, not subject to the Objectivity Problem. Similarly. the Buffer Size Problem and the Contextual Problem disappear by foregoing the theoretical analysis and examining only the empirical data. In contrast to level, probability of recall is a continuous scale, thus the Bins Problem does not occur in the relative memorability analysis. Because propositions are not grouped into bins, it also allows a more fine-grained look at performance at higher and lower levels. Thus, the researcher can isolate propositions which show uniformly high or low recall, either for the purpose of determining the properties that contribute to this effect -- or to exclude them from analysis to make comparisons of performance in an intermediate range. Consequently, the Floor and Ceiling Problems can become interesting empirical phenomena when cast in terms of memorability. Finally, probability of recall also has genuine interval properties, and is even a ratio scale. This allows us to make meaningful statements about slope, thus relieving the researcher of
58
Stint?and Wing@ld
the Scale of Measurement Problem. Thus, there is much to recommend this empirical approach to assessing age differences in relative recall probabilities. The Relative Memorability Function
Assume that for each text unit we have measured the probability of recall by the young [p(& IY)] and by the old @(ReI 011. Simply examining the relationship (the relattue rnernorabtltty function) between these two measures can be very informative. First, if these variables show a strong positive correlation (rl, then it would be clear that in an ordinal sense, what is memorable for the young is also memorable for the old. Furthermore, the coeMcient of determination ($1 even gives an estimate of how much of the variance in unit recall probabilities for the old can be explained by the unit recall probabilities of the young. If the recall probabilities of the young and the old did show a significant correlation, we could infer (again, in a n ordinal sense) that the cognitive processing mechanisms of young and old were functioning similarly in the way they were discriminating among text elements, or in other words, that processing resources were being similarly allocated across text units. In fact, as we have seen, previous studies have suggested that there is such a relationship (Mandel & Johnson, 1984:Rubin, 1978).It is also obvious in such an analysis when there is a lack of agreement for particular units. Determination about some common properties of such units, such as, cohort-specificity (Hultsch & Dixon, 1983). or perhaps emotional salience (Blanchard-Fields. 1986). could provide important insights toward explanations of age differences. The virtue of the scale properties of memorability allow u s to go beyond a statement of ordinal comparison, and enables a stronger statement about relative changes in memorability. The interpretation of the slope of the relative memorability function is straighfforward (see Figure 1). A slope of unity (A, B)would suggest that (in a n interval sense) the degree of discrimination among different text elements was similar for younger and older adults. On the other hand, deviation of the slope from unity would indicate differences in the sharpness of that discrimination between more
QualitativeAge Di@rences
59
Probability of Recall for Young Wure 1. Alternative outcomes of a relative memorability analysis (from Stine & Wingfield, 1988).
and less memorable items, with a slope greater than unity (C) indicating a sharper discrimination among text elements by older adults relative to their younger counterparts, and a slope less than unity (D) indicating an attenuated discrimination among text elements.2 Furthermore. such relative age difTerences in discriminabilty are interpretable in theoretical terms. Assuming that the probability of recall is a consequence of the effective allocation of processing resources, a slope of unity would suggest equally effective allocation In Figure 1, the y-intercept is plotted to be less than or equal to zero, 80 as to be consistent with the typical h d i n g s in the literature. m e r e is, of course, nothing in the analysts that requires this.) If the slope is equal to unity, then the y-intercept represents the mean age difference in text unit recall. If the slope is not unity, the mean age difference depends upon the text unit, and the y-intercept is not meaningful.
60
Stlne and Wingfleld
of processing resources by young and old across the range of memorabilities, whereas a slope dflerent from unity would suggest that older adults were not as effective in allocating resources within some range. More specifically, slope could be said to be positively related to the effectiveness of resource allocation to more memorable items. In other words, a slope less than unity would imply that the older adult was differentially less effective in allocating processing resources to the more memorable items. A slope greater than unity, however, would imply that they were especially ineffective in allocating processing resources to the less memorable items. In the Kintsch model, "effective allocation of processing resources to a text unit" corresponds to "being held over in the input buffer." It might be said that more processing resources are being allocated to a text unit when it is held over for more cycles in working memory. Thus, more memorable items are those that received more processing cycles in working memory (Miller & Kintsch, 1980). and relative dflerences in memorability across age would suggest the existence of relative differences in the number of processing cycles in which units participate. For example, a slope greater than unity would suggest that when younger adults are more likely to hold over a unit in working memory, older adults are even more likely to do so. On the other hand, a slope less than unity would suggest that when younger adults are more likely to hold over a unit for the next processing cycle, older adults are relatively less likely to do so. For young and for old, there can be several reasons for not holding a given unit over in working memory until the next processing cycle. First, it may be that coherence was not established, perhaps because there was no explicit overlapping argument already in working memory, and attempts at bridging inferences were unsuccessful. For early processing cycles or at points of a switch in topic, it could also be due to the central executive not selecting the unit a s "important." or necessary for comprehension. A third alternative is that working memory limitations make it impossible to retain the idea even if coherence was established or it was selected as important.
QualitativeAge D@2rences
61
Now assume that a "young processor" has held over a particular unit for the next processing cycle (thus. making it more likely that it will be remembered), but that the "older processor" did not (thus, making it less likely that it will be remembered). Again, there could be three reasons for this difference: (a) coherence was achieved by the young but not by the old, (b) the unit was selected as important by the young but not by the old, or (c)the working memory capacity of the young was large enough to retain it (even though it was not extremely important). but the working memory capacity of the old was not. These alternative explanations can to some extent be differentiated on the basis of what would be predicted if the language processing requirements on working memory were increased. Suppose, for example, that the density of the information were increased such that more propositions had to be processed at each cycle. If the failure of an older adult to hold over a proposition were due simply to reduced working memory capacity, we would expect them to dmerentially lose propositions from the lower levels of the the coherence graph be.. the less memorable ones) a s more propositions had to be processed (cf.Spilich, 1983). That is, if their problem with increased information load were simply computational space, then the extra propositions that were being input would be effectively distinguished along some dimension of importance, appropriately integrated into the current contents of working memory, and quickly lost at the next input cycle. Thus, a reduced working memory capacity explanation predicts that with increased processing load, the slope of the relative memorability function will increase. On the other hand, if the increased processing load made it especially difficult for the older adult to find coherence among propositions or distinguish between more and less important ideas, then they would show a disproportionate loss of propositions that had been held over for more processing cycles by younger adults (presumably those at a higher level in the coherence graph that
62
Sffneand Wingfield
were more memorable). This would produce a decrease in relative memorability slope with increased processing load. The Egects of Prose Characteristics on the Relative M m m b i l U y Slope
Data relevant to this hypothesis was obtained by measuring immediate recall of single spoken sentences of uniform word length that systematically varied in the number of propositions (Stine & Wingfield, 1988). Assuming that younger and older adults are generally similar in their parsing strategies (Stine, in press: Wingfield & Stine, 1986).this would suggest that the old (like the young) would have to cope with the greater number of propositions at each input cycle. Thus, a reduced working memory capacity hypothesis would predict thit increased propositional density would increase the slope of the relative memorability function. In fact, the opposite was true. v i l e the correlation between unit recall probabilities was fairly constant across propositional density at about .8, the relative memorability slopes were .99, .77, .75. and .58 for 4-, 6-, 8-, and 10-proposition sentences, respectively. This suggests that as processing load was increased, the limiting factor on the performance of the older listener was not simply the size of "work space" in working mem01-y.~Rather, it would appear that the older adult had dfliculty organizing the material as the number of idea units increased, either because it was harder for them to choose the more important text units amidst so many competitors,
The logic of this argument characterizes how the interpretation of qualitative empirical differences as qualitative process differences depends on model assumptions. A reduced W M capacity explanation is a hypothesis about quantftatfue change in process, but the organizational deficit hypothesis is a hypothesis about qualftatfw change in process, and yet both yield test implications that suggest y l t t a t t v e differences in recall, of one sort or another Logically, it could be possible that a disorganized memory representation could result from an extremely small W M buffer that could hold only one or two propositions per cycle. Although this might or might not occur in the case of pathology, this explanation does not seem viable in the case of prose processing within n o d lines.
QuclrttatiwAge D~mences
63
or because it was more dimcult to establish coherence among so many propositions, or b0th.5 Such an explanation may also account for the provocative finding that the overall recall of older adults does not suffer disproportionatelywhen propositional density is increased. In other words, when infornational density is increased, the older adult does not particularly remember dflerentially less on the whole than does the younger adult (Stine, Wingfield. & Poon, 1986). Their recall, however. is relatively less likely to contain the units that both young and old found more memorable. So a s load is increased. the older adult is not necessarily losing a disproportionate number of units from working memory on each cycle, but these text units are less likely to be those that are held over by the young. If this explanation is correct, then factors that affect text organization would be expected to have similar effects on the relative memorability slope. That is, texts with more well-defined structures would be expected to show higher relative memorability slopes than those with less well-defined organization. In other words, if the older adult is having dffficulty in online organization operations, then a text that was inherently easier to organize would be expected to show a higher relative memorability slope than one that was more difficult to organize. If, on the other hand, the limitations in working memory were only limits in size. then there would be no reason to expect that text organization would have this effect on relative memorability slope. To evaluate this issue, we consider a relative memorability analysis on the data taken from an earlier report (Stine & Wingfield, 1987a) on the effects of narrativity and density on prose recall. In the experiment, 31 community-dwelling older adults and 37 university undergraduates read a series of four paragraph-length (77-82 words) passages for immediate recall. Passages were selected such that two were narrative and two were expository. The narratives were short anecdotes consisting of a series of episodes unfolding in time. The expository passages were short informative We should point out that even with this clear change in relative memorability slo e as a function of d e n s 8 the Age x Levels x Density interaction was not significant {f. Stine et al., 1986). s underscores the insensitivity of the levels measurement to such effects.
Stine and WLngfIeld
descriptions. One of the narratives and one of the expository passages were relatively low in propositional density, containing 24 and 25 propositions, respectively. The remaining passages were high in density, containing 37 propositions. The mean percentages of propositional recall by the young for the low-density narrative, low-density expository, high-density narrative, and high-density expository were 88.8,77.8,73.0,and 59.1,respectively. These same percentages for the older group were 45.1,42.2.39.8,and 26.2. Thus, both narrative structure and low propositional density were text variables effective in increasing overall recall performance among younger and older adults. Scatterplots of the p(Re 10) as a function of the plRe I Y) are shown for each of the four passages in Figure 2. In this figure are given the equations fitting the data to both linear and logarithmic functions. (Only the logarithmic function is plotted.) These two functions fit the data about equally well except for the most difficult passage which shows less linearity than the rest.6 These data support earlier findings that a substantial proportion of variance in the recall of individual propositions by elderly adults can be accounted for by the recall probabilities of the young. Additionally, systematic changes in slope are apparent. The slope of the regression line was greater than unity for the low-density narrative, t(22)=3.68.p.01,was not significantly and different from unity for the high-density narrative, t(35)=1.65, the low-density expository. t(23)< 1, and was less than unity for the high-density expository7 , t(35)=3.29, p<.Ol. These results were consistent with our earlier findings that low-density texts generally Obviously, there are a number of functions tht would fit these data, and we will leave it to future research to determine the most characteristic function. The logarithmic function was selected since it is a common curvilinear function that gave a reasonable fit. In spite of the simplicity of linear functions, there are several reasons why it might be more advantageous in the long run to entertain such nonlinear discriptions. First, it relieves us of the theoretically impossible solution of negative recall probabilities. Also, it may be possible to integrate this methodology with more established techniques of item analysis (Bimbaum, 1968),in which case nonlinear !It would become a methodologicalnecessity. Note that the correlatfon between p(Re/YJ and p(Re/O) is considerably lower for this passage than for the rest, thus the lower slope cannot be interpreted as simply a relative decrease in text unit discriminability for the older group. Rather there may have been less similarity between young and old in how the passage was processed.
0.0-
6
E
0.6
-
0.6
d x
OA-
o_
0.4
0.2
0.2
-
-
0.0
0.2
0.6
0.4
0.0
1.0
Y=(.o01)10-, y=12ax-.51.
#,76
F$=n
-
0.0 0.0
0.0
High-Density Narrative
0.2
0.4
1.0
-
0.8-
- aso_
.
Low-Density Expository
0s-
o_
y x l . l 2 x - . U , R2-l56
d
6
a40.2
-
0.0 0.0
-
R*xdB
y=(.W8)10-,
0.6
0.0
1.0
PWM
dWv)
*
HigADensity Expository
.
R2=Y
y=(.op0)101=,
y=.!S4x-.o& R2i.41
0.60.4
-
0.2-
1
0.2
.
I
.
I
l
0.4 0.6 P(RI(Y)
.
r
0.8
-
l
1.0
0.0 0.0
0.2
0.4
0.6 p(R.lY)
0.8
1.0
Figure 2. Probability of unit recall among older adults plotted as a function of the probability of unit recall among younger adults. (Each point represents the recall probability of a single proposition.)
66
Stine and WLngfleld
produced a higher relative memorability slope than the high-density texts. Additionally, however, narratives seemed to produce higher relative memorabilty slopes than expository texts, which supports our previous argument that it is organizational ease which produces the hlgher slope and not just memory load per se. Given the implication from these data that older adults have an organizational deficit that impacts on discourse processing a s information load gets high, the question naturally arises whether older adults are able to determine which elements in text are important, in particular, when text becomes more difficult. To consider this in a preliminary way, we recently presented the text bases of the four passages from Figure 2 to independent groups of younger and older adults and asked them to rate the importance of each idea unit. As closely as possible, the Kintschian text bases were translated into English sentences or phrases and presented to subjects in a list. Participants were asked to rate the importance of each element on a seven-point scale, where importance was defined as how necessary the element was to understanding the passage as a whole: that is. it was made explicit that they were to rate the items on importance for comprehensibility, and not personal importance or interestingness. Participants were told to work at their own pace, and were given ample time to complete the task. Ratings on the two narratives were completed by 10 younger adults (19-31yrs; M = 22.5) and 12 older adults (52-73yrs; M = 67.1);ratings on the two expository passages were completed by independent groups of 12 younger adults (17-26yrs; M = 21.1) and 12 older adults (59-76 yrs; M = 67.2). The younger adults were university summer school students and older adults were community-dwelling senior citizens who were members of a social organization that met periodically to hear a speaker. There were no significant diEerences among these groups in years spent in formal education. Additionally, 2 older adults rating the narratives and 3 older adults rating the expository passages were excluded from analysis for using 4 or fewer of the 7 rating scale points. We did this to avoid a finding of age difl'erences simply because of a failure to use the full scale. There were, therefore, no significant differences among the remaining samples in the number of rating scale points used.
.
7: mO-.67 6-
.. /
549
.
/
L
I
/* /
7
c
3
2 1
.
s
.
:j . . . . ./
6
i
I
c
s
l-----00
1
2
3
4
5
6
7
wflCmplv)
Flgum 3. Median importance ratings assigned to each text unit by older adults plotted as a function of the median importance rating assigned to that unit by younger adults. (Each point represents a single proposition.)
68
S t h and wingfteld
For each text unit, the median rating by each age group was calculated. These median importance ratings are plotted in Figure 3, with the median rating of each unit by the old plotted a s a function of that rating given by the young (as we did earlier for memorabilities). We have plotted regression lines (the Pearson f s were invariably within .03of the Spearman p),though the reader should keep in mind that the ordinal nature of the data preclude interval comparisons of the slopes of these lines across texts. In general, there was fair agreement between the two groups in rated importance of the individual text elements, with no substantial systematic Werences among passages, When asked to make judgments of importance, younger and older adults showed similar discrimination among text units. Although more along the lines of pilot data, these results suggest that to the extent that changes in relative memorability slope are due to organizational deficits, this dimculty with decision processes about relative text unit importance may be a problem that occurs in online processing for comprehension or memory, rather than one that arises in omine evaluations. In order to more systematically examine the effects of prose genre on relative memorability slope, we recently conducted another experiment. In this study, there were more passages of each prose type, making it easier to detect effects of text variability on slope independent of their effects on overall difficulty. We were also concerned about the possibility that the very high slopes observed in Figure 2 were due to a lack of variability in recall probabilities for the young. Because subjects had ample time to read and review each passage and wrote their recall, both the encoding and retrieval requirements of the task were very close to optimal without taxing working memory. Previous research, however, has suggested that such conditions could difllerentially favor the young who may be better able to use review time (Dixon et al.. 1982;Taub & Kline. 1978). It could be that the relatively greater discrimination among text items by the old is restricted to such situations in which the young have this kind of advantage. Presumably in this case younger adults distribute their processing resources more evenly among text units, enabling them to achieve uniformly high unit recall, while older adults use their resources to effectively process
Qualftatiue Age Di#mmces
69
only the more memorable ones. Subjects were. therefore, asked to listen to the passages and recall them orally. The participants in this experiment were 18 university undergraduates (18-20 yrs; M = 18.7) and 24 community-dwelling older adults (58- 78 yrs; A4 = 68.6). These groups were similar in
being well-educated (with an average of one or two years of college education) and above average in terms of verbal ability. All participants listened to 12 paragraph-length passages. The passages ranged in word length from 74 to 95 words, and consisted of from 24 to 42 propositions. Half of the passages were narrative and half were expository, and there was a range of difficulties within each group. Participants listened to recordings made by a female speaker reading the passages in normal intonation at about 160 words per minute. There were four orders of presentation distributed among subjects. The spoken recall was recorded for later transcription, and recall protocols were scored using a gist criterion. Figure 4 shows a plot of the mean proportion of propositions recalled from each passage by the older adults as a functlon of this proportion for the young (note that each point represents one passage). The correlation r = .90 (p<.Ol) between these two variables was very high, demonstrating the strong relationship between overall recall performance of the young and the old (y = -.01+ .794.8 Again,the correlations between between p(Re IY') and p(Re I0)for the individual text units within each passage was also moderately high, with a median of 3 5 . and a range of .73 to .90. More interesting was the finding that the relative memorability slope was higher for the narrative passages than for the expository ones. For four of the six narrative passages. the relative memorability slope exceeded .9; this was true for none of the expository passages. The mean slope was .90 (standard error = .06) for the narrative We note that there is one outlier in Figure 4. The easiest passage for the young was not so for the old. Post hoc we can think of various reasons why this passage might have presented particular dimculty for the old (e.g., more inference required for comprehension),especially in an experimenter-pacedlistening situation. If this passage is removed, the correlation between mean p(Re/Y)and mean p(Fk/O)is .96, with a regression equation of y=.08+.99x We flnd this interesting because it puts the slope almost exactly at unity.
Stine and WIngfield
70
y-.009+.79x
r=.m
0.6
5E 0.3
0.4
0.5
0.6
0.7
mean p(Re1Y)
Flgwe 4. Mean probability of unit recall for each passage, with recall performance of the old plotted as a function of recall performance of the young. (Each point represents a single passage: each plus represents a narrative passage: each dot represents an expository passage.
passages, but .76(standard error = .02)for the expositoIy ones: this dlfference was statistically significant, t(10)= 2.16,~x.03.Thus, the degree of text element discrimination was similar for young and old when the prose was well-structured narrative. When the prose was expository, however, older adults' recall of items that the young found the most memorable was suppressed. We should point out also that under these conditions in which both young and older subjects were prevented from reviewing the passages, there was little evidence that older adults actually showed greater discrimination among prose units. There was a single (narrative) passage for which the relative memorability slope was just greater than unity ( b= 1.08). Thus, under listening conditions without opportunity for review there was scant support for the notion that the typical older adult retains gist at the cost of detail. (Note that these data have been analyzed without regard to individual differences in verbal ability. It is quite possible, given our survey of the literature in Table 1, that higher verbal groups would produce a steeper relative memorability slope than lower verbal groups.) These effects of text structure and processing load on relative memorability slope are difficult to explain simply in terms of age
QualitativeAge Dfl~renceces
71
differences in the computational space available in working memory. Rather these results are more supportive of a deficit among older adults in their ability to organize text elements within working memory. Thus, these data suggest the existence of a n important qualitative difference between younger and older adults in the way discourse is processed, specifically. that older adults have particular difficulty with at least some of the component processes necessary for establishing text structure, such a s the online determination of important propositions or in establishing coherence relations. This qualitative difference, however, is not unrelated to quantitative declines in working memory space since organizational difficulty seemed to be exacerbated by a high processing load, when presumably limitations in WM space or efficiency come into play. Recapitulation First, we discussed several problems with relying on a levels analysis for assessing qualitative age differences in discourse processing, pointing out its strength as a method for testing the Kintsch model did not easily extend to testing differences between groups in text organization ability. In spite of the ambiguities in level of importance a s a measurement, we argued that there was a thread of consistency in the literature seeking the elusive Age X Levels interaction: under difficult processing conditions, older adults were relatively less likely to discriminate between more and less important information. Second, we argued for the value of directly comparing the unit recall probabilities of young and old in a relative memorability analysis. Even though its major strength is a s an empirical tool, which can be used with any prose analysis system, its results are interpretable in terms of current models of discourse processing (e.g.. Kintsch & van Dijk, 1978). Finally. we offered data showing that (a) the rank order of text unit memorability is essentially the same for young and old (Rubin, 1985). and that (b) consistent with our conclusions reached with respect to age-related change in the levels effect, older adults show less extensive text unit discrimination relative to younger adults a s
72
Stine and WLngfleld
text difficulty increases. We argued that these data were supportive of the notions that (a) the discourse processing architecture is largely unchanged in later adulthood, but that (b) under conditions of high cognitive load or less well structured text, older adults may have particular difnculty in determining online which ideas are the most important, in achieving coherence among the ideas (cf.Light & Capps, 1986). or both. In spite of its promise, the relative memorability analysis is not without problems. The major disadvantage to this approach is that while we can compare how memorable units are for young and old, the analysis itself will not tell u s directly why that is the case. So its main advantage (freedom from multiple theoretical assumptions) is also its main disadvantage. As we have seen, however, if a discourse processing theory has some notion of a text unit and a mechanism for predicting differential memorability among them (i.e., discrfmination among text units), then relative memorability can be interpreted in those t e r n . Also implicit in this technique is the assumption that the younger adult is the standard against which the older adult is compared. While there are many sound arguments against this approach (e.g., Labouvie-Vief & Chandler, 19781, defense rests on the point that there is much standing evidence to suggest that young adulthood is the optimal time in the life span for effective and eikient memory performance -- assuming the material to be learned is of a generic sort, requiring little expertise in a particular topic. Under these conditions, young adulthood may provide an acceptable criterion. Of course, it is only through aging that we acquire expertise, and we might well expect that for discourse requiring expertise, we would find very different relative memorability phenomena. Finally, we should point out some practical disadvantages of this approach. Reliable estimates of unit recall probabilities are difficult to obtain with small sample sizes. making it ideal that each passage be recalled by a fairly large number of subjects. This can be fairly troublesome for between-subjects designs, or for designs in which the presence of particular passages in conditions is counterbalanced. Also. because the dependent variable of interest is the slope of a function, the unit of analysis becomes an entire passage. This makes it dif'ficult to design experiments that contain
Qualitative Age D i f l m e s
73
enough passages to yield sufficient statistical power, and yet can be completed by a subject within a reasonable amount of time. In spite of these practical difficulties. however, this method does seem to yield fairly reliable results. Thus, we began this section with an empirical question, simply addressing whether younger and older adults remembered the same information from discourse. Changes in this relationship as a function of processing demands, however, shed some light on the extent to which younger and older adults process discourse differently. In the next section, we take a broader look at this process-oriented question, considering several major hypotheses of qualitative age differences in how language is processed. DO OLDER ADULTS PROCESS DISCOURSE DIFFERENTLY? In recent years, there have been four major hypotheses tested which propose qualitative age differences in how discourse is processed: (a) the Response Style Hypothesis, (b)the Distinctiveness Hypothesis, [c) the Structural Support Hypothesis, and (d) the Coherence Hypothesis. Generally, these hypotheses share the common notion that there are some [quantitative) limitations in the discourse processing capabilities of the older adult [e.g., a reduction in processing resources or in WM space or efficiency), and that this produces some qualitative change either in the way discourse is represented or in the strategy used in processing discourse. We describe each of these hypotheses in turn. considering evidence bearing on each. The Response Style Hypothesis According to this view (Adams. 1987. 1988; Labouvie-Vief & Sche11.1982) the goals of the older reader or listener may be fundamentally dlfferent from those of the younger adult. Because of this, there are qualitative age differences in the style of processing (and hence, recalling discourse. Consistent with their recent educational experience and the typical cognitive demands placed on them, younger adults focus on the propositional content of discourse, in an effort to form a reproductive style of representation.
74
stlneandwfngjleld
Older adults, on the other hand, who have fewer processing resources to devote to this kind of indiscriminate absorption of facts, focus on the meaning of the discourse insofar a s it is personally relevant to them. Thus, this hypothesis would predict that while younger adults would recall more of the information actually contained in the text, older adults would be more likely to produce interpretive or evaluative statements which assess a text's personal or social meaning. While not widely researched, some evidence has been supportive of these ideas. For example, Adams (1987, 1988) has reported that older adults are more likely to produce psychological and metaphorical transformations of the story's content in recall and summary protocols, and Adams, Labouvie-Vief, Hobart, and Dorosz (in press) have argued that for non-fable narratives older adults tend to produce more integrative and interpretive statements in recall, while younger adults tend to produce more text-based reproductions and elaborations. Similarly, Boswell ( 1979) has argued that older adults are better able to provide integrative interpretations of metaphor. The possibility that the older individual approaches the discourse processing situation differently, attending to very different text features, remains largely unexplored. It would seem that one very important question that will have to be addressed as this research proceeds is how such qualitative differences in response style are related to the competence-performance distinction (cf. Light, 1988). For example, at what point in the life span does the competence to make sophisticated interpretative judgments develop? Is the change in response style (performance) noted by Adams and her colleagues due to the development of competence, or is it an adaptive strategy to cope with performance declines in other cognitive domains? In other words, to the extent that there are such qualitative differences, do they represent the development of new competencies, changes in performance. or both. 'The Semantic Distinctiveness Hypothesis The Distinctiveness Hypothesis suggests that the semantic representation produced by older adults as they process language is
QualitativeAge Di@kences
75
less distinctive than that produced by younger adults. Within this view, this "fuzzier" representation forces the older adult to place greater reliance on other sorts of information, such as superordinate contextual information or the already existing knowledge base, in order to understand and remember incoming language. These notions have received empirical support from several sources. Cohen and Faulkner (1981) asked subjects to listen to spoken texts and then to detect changes in a written version of this text at varying delay intervals. Older adults were less able to detect semantic changes at all delays, and this difference was particularly pronounced at the longest retrieval delay (40 sec), implicating retention factors as well a s encoding ones. Reder. Wible. and Martin (1986) presented subjects with short passages about which they later responded to a series of true-false statements. Statements varied in plausibility and participants were asked to make either recognition judgments (whether or not the statement had, in fact, occurred in the passage) or plausibility judgments (whether or not the statement could reasonably have occurred in the passage). Response time and accuracy data suggested that older adults tended to rely in general on a plausibility strategy, with age deficits appearing in the recognition task, but not for plausibility judgments. Furthermore, older adults were more likely to make false alarms in the recognition task when statements were plausible. Reder et al. contend that the plausibility strategy allows the older adult to calculate a partial match between the target and "a general or stereotyped encoding of information'' (p. 79). If older adults do tend to produce a less distinctive representation of language input, one might assume (as did Reder et al.) that there would be a greater reliance on stereotypical knowledge-based representations of the discourse being processed. For example, if an older adult listened to a story about two characters having dinner in a restaurant, producing a "fuzzy." nondistinctive representation about their activities, s/he might rely on a probably extensive knowledge base about the sequence of activities that occur while eating out. In fact. Hess and his colleagues (Hess, 1985; Hess, Donley, & Vandermaas. 1989) have reported larger age
76
Stlne and Wtngjleld
differences in memory for atypical actions from such prototypical activity sequences ("scripts"). There are several other studies, however, showing no qualitative age differences in the knowledge and use of scripts. Light and Anderson (1983)found age dmerences neither in the kind of script information generated nor in the typicality effects in recall or recognition memory: for example, younger and older adults were equally likely to recall atypical events and falsely recognize typical ones. Similarly. Zelinski and Miura (1988) have reported that younger and older adults make similar use of thematic infomation in recall and recognition of scripted stories. Thus, while older adults may produce a less distinctive representation of the discourse they process, the extent to which they rely upon generic information as a compensatory source of information may well be under strategic control, and therefore, subject to various demand characteristics of the encoding and retrieval situations. An understanding of how older adults make use of generic information may require more extensive research in which the accessibility of the generic information and the ease of encoding the target discourse is manipulated. The Structural Support Hypothesis The Structural Support Hypothesis is a close cousin of the Semantic Distinctiveness Hypothesis. It, too, posits that deficiencies in encoding force the older adult to rely on other sources of information. Rather than focussing on the usefulness of declarative knowledge about how the world works, it suggests that the older adult places extensive reliance on procedural knowledge about how language works. Some support has been garnered for this hypothesis at different levels of the language processing system. Older listeners appear to depend heavily on linguistic structure for comprehension and immediate memory. For example. whereas older adults have great dlfnculty in remembering short passages of time-compressed speech when it lacks syntactic or semantic structure, age differences are negligable when normal linguistic constraints are present (Wingfield, Poon, Lombardi, & Lowe. 1985).
Qualitative Age Differences
77
In a similar vein, Fullerton and Smith (1980) have argued that the
older adult particularly relies on syntactic structure to discriminate between "given" information (already instantiated in working memory from a previous input cycle) and "new" information (from the current input cycle). Thus, a dependence on surface syntactic form could facilitate finding semantic coherence. Older adults also appear to pay particular attention to prosodic contour and stress patterns in the processing of spoken language, showing enhanced performance in answering comprehension and memory questions about items that are stressed (Cohen & Faulkner, 1986) and depressed immediate recall performance when text is spoken in list intonation without normal variations in intonation, stress and timing (Stine & Wingfield. 1987b: Wingfield. W a r , & Stine. 1989). Older adults also appear to depend upon sentence context for the recognition of words and phonemic information. For example, Cohen and Faulkner (1983. Experiment 2) have reported that older adults are facilitated by sentence context in a lexical decision task to the same extent as younger adults when the context is strongly predictive of the target word and to even a greater extent when the context is weakly predictive. Their data suggest that in everyday discourse comprehension, semantic constraints may dlfferentially support the older reader or listener in efficiently conducting word level processing. Other data come from a recent experiment in our own lab using the "gating" technique (Wingfield, Aberdeen, & Stine, submitted). In this procedure, the subject is asked to recognize the word after hearing only part of the phonemic information. The amount of the word heard is increased across trials. and of interest is how quickly the subject is able to isolate the correct word. While older adults were much slower to access the initial phoneme of the correct word when the word was preceded by a carrier phrase which gave no context for the target word, this age difference was much smaller when the word was preceded by a context which loosely constrained the target word, and there were no age differences at all when the target word was preceded by a highly constraining context. Such use of context in interpreting lower level orthographic or phonemic information may be an important source of support for the older adult a s sensory decline makes these features less
78
StlneandWLngfield
accessible. This interpretation is bolstered by the finding that older adults make particular use of sentence context in auditory word recognition under conditions of moderate noise (Cohen 81 Faulkner, 1983, Experiment 3). Shnilarly, older adults have been found to be particularly reliant on sentence context in visual word recognition when the target word is degraded (Madden, 1988). Another kind of structural support that may facilitate discourse processing among the elderly is provided by the way ideas are arranged within a passage. The narrative passage, which tells a story unfolding in time. is commonly acknowledged to have a welldefined story structure (e.g., Mandler, 1978) that allows the reader or listener to make predictions about the kinds of inf'omation which will be presented at each point. Procedural knowledge about how to use this structure is an important skill in language processing. It is well-established that for younger adults narratives are easier to process and more memorable than expository passages. Older adults have been reported to show a narrative advantage similar in magnitude to that observed among the young (Stine & Wingfleld, 1987a; Tun, 1989; Petros, Norgaard, Olson, & Tabor, 1989). suggesting that older adults are sensitive to these structural differences. Evidence that the older adult may particularly rely on narrative structure comes from a study by Smith, Rebok. Smith, Hall, and Alvin (1983). In this experiment, two-episode narrative passages were presented to younger and older adults in one of three ways: (a) in canonical structure, in which the episodes were presented sequentially, fb) in interleaved structure, in which the elements of each episode were intermixed, or (c) in scrambled form, in which the elements of both episodes were randomly arranged. While recall of the canonically structured stories was equally high for younger and older adults, and recall of the scrambled stories was equally poor (which has also been shown by Mandel and Johnson (1984)).older adults recalled fewer elements than the young when stories were interleaved. Thus, when story structure could facilitate encoding, stories were easily understood and remembered by young and old alike, and when story structure was completely lacking, neither young nor old could reorganize the story for comprehension. When story structure was available but only accessible through complex
QualUatlueAge Diffmences
79
reorganizational operations which places a strain on working memory, older adults had dlfflculty but the young did not. These data provide support for the notion that older adults rely on narrative structure to represent discourse and when that structure is disrupted, the increased organizational load on W M impairs performance. The Coherence Hypothesis Another hypotheslzed source of qualitative age differences in discourse processing is the extent to which coherence is achieved in online processing. According to the Coherence Hypothesis, depressed memory for discourse is, in part, a result of the older adult having dimculty in organizing concepts within input cycles or in integrating concepts between cycles. Thus, this hypothesis is a version of the Semantic Distinctiveness Hypothesis which cites specific cognitive mechanisms a s a cause of decreased distinctiveness. The earliest studies examining the ability of older adults to integrate information from discourse relied on the Bransford and Franks (1971)paradigm of linguistic abstraction. In the original Bransford and Franks ( 1971) experiment, young adults were presented a series of sentences (e.g., The ants ate the jelly. The sweet jelly was on the table.) each representing a subset of the propositions contained in a more complex passage (e.g., The ants in the kitchen ate the sweet jelly which was on the table.). In a subsequent recognition test. subjects' rated confidence that an item was "old' increased with the number of linguistic elements that the item contained, and subjects were only able to marginally discriminate between "old' and "new" items when they contained only one element. In this seminal study, which was certainly a precursor to the important developments in discourse processing models in the last fifteen years, Bransford and Franks used these kinds of data to argue that what was stored in comprehending text was a n "inter-sententially defined." integrated abstraction of linguistic ideas. Thus, recognition errors were taken as evidence that language was [in more modem parlance) integrated across input cycles. Use of this technique with older adults Walsh &
80
Sine and Wingjleld
Baldwin, 1977; Walsh, Baldwin, & Finkle. 1980) has shown that older adults demonstrate effects identical to those reported by Bransford and Franks, leading early on to the conclusion that integration occurred in the same way for young and old. More recent research, however, has suggested that while older adults may not have difficulty with integration under conditions of low memory load (asrepresented by the Bransford and Ranks task) (cf.Light & Capps, 1986). high load conditions may put a strain on integration operations. For example, Light, Zelinski, and Moore (1982, Experiment 3) have reported that older adults have particular difficulty In syllogistic reasoning tasks when the order of the premises is haphazard. Age differences remained even when scores were conditionalized on correct fact retrieval. suggesting that the dimculty for the older adult lies in constructing a n integrated representation when this representation would require excessive information manipulation. Similarly, Light and Capps ( 1986) have found that older adults have particular dmiculty in comprehending pronouns when there is a lot of infomation intervening between the anaphoric referent and pronoun. Additionally, older adults appear to have particular difficulty answering questions about text when the questions require the integration of information that is not adjacent in the passage (Cohen & Faulkner, 1984; Zacks, Hasher. Doren, Hamm,& Attig, 1987). especially when modality and pacing conditions put a strain on W M operations (Zacks & Hasher, 1988; see also Hasher & Zacks, 1988). Other support for the existence of an organizational deficit in discourse processing is provided by Byrd (1985) who found that older adults have particular difficulty in providing adequate summaries of text passages. Online reading data has provided another source of evidence that there are qualitative differences in integration operations. In the measurement of word-by-word reading time, younger adults are known to allocate extra time to sentence-final words (e.g., Aaronson & Scarborough, 1977; Haberlandt, Graesser, Schneider, & Ktely. 1986) and to words immediately preceding major and minor clause boundaries (Aaronson & Scarborough. 1977; Haberlandt, 1984). Thus,as noted earlier. these reading times peaks reflect syntactic structure. The amount of time allocated to these peaks tends to increase as a function of the number of words between peaks.
QualffutlueAge D@krences
81
presumably reflecting within-chunk organization, and a s the total number of prior words in the sentence, presumably reflecting the processing load of integrating across these chunks (Aaronson 81 Scarborough, 1977). In reading single sentences for immediate recall. older adults appear, in fact, to show many similarities to younger adults in how they allocate reading time (Stine. in press). Of interest to our argument here is how they allocate time to syntactic boundaries. Using multiple regression coefficients to determine how much extra processing time was allocated at these critical points, several interesting age differences were found. The time allotted to major and minor constituent boundaries is shown in Figure 5. Because this was measured as a regression coefficient, many of the other important influences on reading time (e.g.. word length and frequency) were statistically controlled. On the whole, older adults allocate just about the same amount of (orjust a little bit more) extra processing time to major and minor constituent boundaries as younger adults do. These groups show differences in immediate recall, however. Thus, the encoding strategy that was effective for the young was not equally as effective for the old. In order to assess what the reading strategy of effective encoders was, a separate analysis was performed on those reading ~ younger adults who times that resulted in perfect r e ~ a l l .While gave perfect recall allocated extra time to major boundaries, older adults allocated extra time to both major and minor boundaries. suggesting that age dmerences in effective reading strategy required that the older adults process discourse in smaller chunks. Additional analyses of reading time peaks also suggested that the reason that similar allocation of processing time by younger and older adults was ineffective was that the quality of this processing time was different for younger and older readers. For both groups, reading time at these peaks increased linearly at the same rate as a function of the number of words since the last peak (about 185 msec per word), suggesting similarity in within-chunk processing. Of course, retrieval failures could have contributed to these recall differences as well, but by isolating perfect recallers, it is clear that at least a subset of the effective encoders was considered: a difference in reading strategy between subjects as a whole and perfect recallers alone would provide supportive evidence that memory differenceswere in part due to encoding differences.
82
Stine and Wtn&Ad
Younger adults also allocated about an extra 200 msec at peaks for each previous peak, for integration time between chunks. The reading time peaks of older adults, however, were insensitive to the number of previous peaks in the sentence, suggesting that they were not allocating integration time. Older adults who were perfect recallers showed a nonlinear trend in the allocation of betweenchunk processing time, alloting a lot more processing time per chunk (about 500 msec per previous peak) early in the sentence only. These data suggest that one source of the memory deficit for single sentence passages was a failure of the older adult to engage in between-chunk integration. Older adults who were effective in gMng immediate recall, on the other hand, engaged in organization and integration operations more frequently and allotted more time for integration for the &st few chunks before the processing load became unmanageably large. Thus, generally, young and old were not qualitatively different in how they allocated processing time: in order to effectively encode the materials, however, older adults had to allocate time dmerently from the young. Finally, support for the Coherence Hypothesis comes from the data on the levels effect and relative memorability analyses discussed earlier. Because it is generally accepted that semantic representation of discourse occurs during comprehension, the qualitative age differences in what is remembered that appear under high load conditions may well reflect age deficits in constructing a coherent representation of discourse under high load conditions.10 Recapitulation We have outlined four hypothesized loci of qualitative age differences in discourse processing. Some of these Werences (e.g., organization and/or integration deficits) appear to arise out of limitations in the kinds of WM processes that can be adequately performed by the elderly adult under high load conditions. Other lo Formally, of course, it could be that under high load conditions, older adults construct the same semantic representation of discourse as the young, but have a difkrentialiy high forgetting rate for (or diffvential &faculty in accessin4 the more important or more memorable information. This explanation, however, does not seem compelling.
QualitativeAge D
~
All Subjects
H
~
e
Perfect Recallers Only
1800-
18001
Major Boundary
E
8
Y
m
83
s
1200'
Minor Boundary
Major
Y
Y
Young
Old
Young
Old
Flgute 5. Amount of extra processing time allocated to major and minor constituent boundaries by younger and older readers. The left panel shows these data for all eubjects: the right panel includes data from perfect recallers only (from Stine, 1990).
qualitative differences (e.g., reliance on contextual information) appear to arise out of these W M differences, perhaps as a kind of compensatory adaptation. These hypotheses are certainly not mutually exclusive. For example. they could all come into play in a scenario in which a failure to create an integrated coherent representation in discourse reduces the semantic distinctiveness of the representation. This, in turn, could cause the older adult to rely on supportive information and/or to change his or her way of responding to the task under some conditions. While all of these hypotheses have some empirical support, the extant data are not uniformly supportive of these qualitative hypotheses. These counterexamples point out possible boundary conditions under which qualitative age differences may be observed, and suggest research tactics for clarifying these issues. As we have seen, age differences in the abilty to construct a n integrated representation of discourse may be more probable when memory load is high or the task of integration is more complex, suggesting the need for more research which explicitly manipulates these factors.
84
Stine and Wingfield
For example, whether or not older adults rely on context in the interpretation of bottom-up information may well depend on the kind of context used. For instance, reliable age differences in the use of context appear to be more often found when that context consists of more procedural knowledge about how to use language at the level of syntax or semantic constraints. We would also expect that the representation of this knowledge (i.e., linguistic competence) would show very little between-subject variability. Age differences in the use of context tend to be less frequently observed. however, when the context under consideration is knowledge about social procedures: such representations of script and thematic information may be less uniform across individuals, and thus use of this kind of context would be, in principle, harder to assess. CONCLUSIONS
The fact that the ordinal relationship among text unit memorabilities remains essentially unchanged in later adulthood suggests that in a very important sense the discourse processing architecture shows little or no qualitative change in later adulthood. We have found numerous empirical examples, however, to suggest that age differences are more likely to appear in discourse processing performance when the demand on processing resources is high. The available data cannot be simply explained in terms of reduced storage capacity in working memory. Rather, it is our position that as the storage demands on working memory increase, the older reader or listener has particular difficulty in creating a coherent representation of discourse. Because such demands are not uncommon in everyday discourse processing, this is not an unimportant qualitative age difference. Future research might be profitably geared toward examining ways to reduce its impact, for example, by exploring structural support or presentation styles which could reduce working memory load. ACKNOWLEDGMENTS
Preparation of this chapter was supported by grants R29 AGO8382 to E. A L. Stine and R37 AGO4517 to A Wingfield. We would like to thank Sarah Wayland, Sheldon Myers, and Alan Corcos for their
QualltatfueAge Differences
85
assistance in data collection and analysis, and David Rubin and Tom Hess for valuable comments on a previous draft of this chapter. We would also like to thank the American Psychological Association, who holds the copyright on Figures 1 and 5, for allowing them to be reprinted in the present manuscript. REFERENCES Aaronson, D., & Scarborough. H.S.(1977).Performance theories for sentence coding: Some quantitative models. Journal of Verbal Learning and Verbal Behavior. 16.277-303. Adams. C. (1987.August). Styles of narrative processing: A qualitatfve shfjlfrom adolescence to mature adulthood Poster presented at the meeting of the American Psychological Association. New York, NY. Adams. C. (1988.August). Qualitative age differences in narrative processing: A story retold. In E. A. L. Stine & A. Wingfield (chairs). The question of qualitative age duerences f n discourse processing. Symposium conducted at the meeting of the American Psychological Association, Atlanta, GA. Adams. C., Labouvie-Vief. G., Hobart, C.J., & Dorosz. M. (inpress). Adult age group differences in story recall style. Journal of Gerontology: Psychological Scfences. Alliger. G.M.. Hanges, P.J.. &Alexander. RA. (1988).A method for correcting parameter estimates in samples subject to a ceiling. P~y~hol~giCaZ Bulletin, 103.424-430. Barrett. T.R., &Wright, M. (1981). Age-related facilitation in recall following semantic processing. Journal of Gerontology, 36, 194199. Blanchard-Fields. F. (1986).Reasoning on social dilemmas varying in emotional saliency: An adult developmental perspective. P ~ y ~ h o l and o g ~As@, 1,325-333. Boswell, D.A. (1979).Metaphoric processing in the mature years. Human Development, 22. 373-384. Bransford. J.D., & Franks, J.J. (1971).The abstraction of linguistic ideas. Cognitive Psychology. 2. 331-350. Burke, D.M., & Light, L.L. (1981).Memory and aging: The role of retrieval processes. Psychological Bulletin. 90.5 13-546. Byrd, M. (1985). Age differences in the ability to recall and summarize textual information. Experimental Agfng Research 11, 87-91. Cohen. G . (1979).Language comprehension in old age. Cognitive Psychology, 11.412-429.
Stine and Wingfleld
86
Cohen, G. (1988). Age Merences in memory for texts: Production deficiency or processing limitations. In L.L. Light & D.M. Burke (Eds.). Language, memo%, and agfng (pp. 171-190). New York Cambridge University Press. Cohen, G., & Faulkner, D. (1981). Memory for discourse in old age. Mscourse Processes. 4.253-265. Cohen, G . , & Faulkner, D. (1983). Word recognition: Age differences in contextual facilitation effects. British Journal of P ~ y ~ h o l o74,239-25 g~, 1, Cohen, G., & Faulkner, D. (1984). Memory for text: Some age differences in the nature of information that is retained after listening to texts. In H. Bouma & D.G. Bouwhuis (Eds.). Attention and Perfomance X (pp. 50 1-5 14). Hfflsdale: Erlbaum. Cohen, G., & Faulkner, D. (1985). Word recognition: Age differences in contextual facilitation effects. British J o u m l of P~ychology,74,239-251. Cohen, G . , & Faulkner, D. (1986). Does "elderspeak" work? The effect of intonation and stress on comprehension and recall of spoken discourse in old age. Language and Communication, 6, 91-98.
Daneman. M., & Carpenter, P A (1980). Individual differences in working memory and reading. Journal of Verbal Learntng and Verbal Behavior, 19, 450-466. Daneman, M., & Carpenter, PA. (1983). Individual differences in integrating information between and within sentences. Journal of Experimental Psychology: Leaming, Memory, and Cognition, 9. 56 1-584.
Daneman. M., & Green, I. (1986). Individual differences in comprehending and producing words in context. Journal of Memory QndLQnguage,25, 1-18. Daneman. M. & Tardif, T. (1987). Working memory and reading sklll re-examined. In M. Coltheart (Ed.), Attention and Pe@b-munce XI (pp.491-508). Hfflsdale: Erlbaum. van Dijk. T.A.,& Kintsch, W. (1983). Strategies of dbcourse comprehension. New York Academic Press. Dixon, R.A., Hultsch. D.F., Simon, E.W., & von Eye, A. (1984). Verbal ability and text structure effects on adult age differences in text recall. Journal of Verbal Learnfng and Verbal Behauior, 37,358-364.
Dixon, R.A., Simon, E.W.. Nowak. C.A., & Hultsch, D.F. (1982). Text recall in adulthood a s a function of level of information. input modality, and delay interval. Journal of Gerontology, 37, 358-364.
Qualitative Age Diflkrences
87
Einstein. G.O.. & Hunt. R.R. (1980). Levels of processing and organization: Additive effects of individual-item and relational processing. Journal of Experimental Psychology: Human Learning and Memory, 6, 588-598. Fletcher, C.R. (1981). Short-term memory processes in text Journal of Verbal Learning and Verbal comprehension. Behnvior. 20,546-574. Fletcher, C.R (1986). Strategies for the allocation of short-term memory during comprehension Journal of Memory and Language, 25,43-58. Fullerton. A.M.. & Smith, A.D. (1980).Age-related Merences in the use of redundancy. Journal of Gerontology,35,729-735 Glanzer, M., Dorfman,D.. & Kaplan, B. (1981). Short-term storage in the processing of text. Journal of Verbal Leamfng and Verbal Behnvior, 20,656-670. Graesser, A.C. (1981).Rose Comprehension beyond the word. New York Springer-Verlag. Graesser. A.C., & Riha. J.R. (1984). An application of multiple regression techniques to sentence reading times. In D.E. Kieras & M.A. J u s t (Eds.) (1984).New methods in reading comprehension research (pp. 183-218). Hillsdale. NJ: Erlbaum. Haberlandt, K. (1984). Components of sentence and word reading times. In D.E. Kieras & MA. Just (Eds.) (1984).New methods in reading comprehension research (pp. 219-252).Hillsdale, NJ: Erlbaum. Haberlandt, K.F., Graesser, A.C., Schneider, N.J., & Kiely. J. (1986). Effects of task and new arguments on word reading times. Journal of Memory and Language, 25,314-322. Harker, J.O., Hartley, J.T., & Walsh, D.A. (1982).Understanding discourse: A Me-span approach. In BA. Hutson [Ed.), Advances in reading/language research (Vol. 1) (pp. 155-202).Greenwich. CT:JAI Press. Hasher, L.. & Zacks. RT. (1988).Working memory, comprehension. and aging: A review and a new view. The Psychology of Leamfrig and Motivation, 22, 193-225. Hess. T.M. (1985). Aging and context influences in recognition memory for typical and atypical script actions. Developmental P~ychol~gy. 21,1139-1151. Hess. T.M., Donley. J., & Vandermaas. M.O. (1989).Age-related changes in the processing and retention of script information. Experimental Aging Research 15, 89-96. Hultsch, D. F., & Dixon, R. A. (1983).The role of pre-experfmental knowledge in text processing in adulthood. Experfmental Aging Research, 9. 17-22.
Stint?and Whgpeld
88
Hultsch. D. F.,& Dixon, R. A. (1984).Memory for text materials in adulthood. In P.B. Baltes & O.G. Brim (Eds.), Life-span development and behavfor (Vol.6, pp. 77-108). New York: Academic Press. Jackson, M.D.. & McClelland, J.L. (1979). Processing determinants of reading speed. Journal of Expertmentaf Psychology: Generd, 108. 151-181. Jarvella, R.J. (1971). Syntactic processing of connected speech. Journal of Verbal Learning and Verbal Behaviw. 10,409-416. Just, MA., & Carpenter, PA, (1980). A theory of reading: From eye fixations to comprehension. Psychological Reulew, 87, 329-354. Kemper. S. (1987). Syntactic complexity and elderly adults' prose recall. Experimental Aging Research, 13, 47-52. Kintsch, W. (1988). The role of knowledge in discourse comprehension: A construction-integrationmodel. Psychological Review, 95, 163-182. Kintsch. W., & Keenan, J. M. (1973). Reading rate and retention a s a function of the number of propositions in the base structure of sentences. CugnUiw Psychology, 5. 257-274. Kintsch, W., Kozminsky, E., Streby, W.J., McKoon, G., & Keenan, J,M. (1975). Comprehension and recall of text as a function of content variables, Journal of Verbal Learning and Verbal Behauior, 14, 196-214. Kintsch, W.,& van Dijk. T.A. (1978). Toward a model of text comprehension and production. Psychological Review. 85,363394.
LaBerge, D., & Samuels. S.J. (1974). Toward a theory of automatic information processing in reading. CognfttvePsychology, 6, 293323.
Labouvie-Vief. G.,& Chandler, M.J. (1978). Cognitive development and life-span developmental theory: Idealistic versus contextual perspectives. In P.B. Baltes (Ed.). Life-span development and behavior (Vol. 1, pp. 181-210). New York Academic Press. Labouvie-Vief. G., & Schell, D. A. (1982). Learning and memory in later Me. In B. B. Wolman (Ed.), Handbook of developmental psychology (pp. 828-846). Englewood Cliffs. NJ: Prentice-Hall. Light, L.L. (1988). Language and aging: Competence versus performance. In J.E. Birren & V.L. Bengtson (Eds.), Emergent theories ofaging (pp. 177-213).: Springer. Light, L.L., & Anderson, P.A. (1983). Memory for scripts in young and older adults. Memory and Cognition, 1 1 . 435-444. Light, L.L., & Anderson, P.A. (1985). Working memory capacity, age, and memory for discourse. Journal ofcerOntofogy.40,737747.
Qualitative Age DIffeences
89
Light. L.L., & Burke, D.M. (1988). Patterns of language and memory in old age. In L.L. Light & D.M. Burke (Eds.). Lunguage, memory, and aging (pp. 244-271).N . Y.:Cambridge University Press. Light, L.L.. & Capps, J.L. (1986).Comprehension of pronouns in young and older adults. Developmental Psychology, 22, 580585. Light, L.L.. Zelinski. E.M.. & Moore, M. (1982). Adult age differences in reasoning from new information. Journal of Experimental Psychology: Learning, Memory, and Cognition, 8, 435-447. Lord. F.M.. & Novick, M.R. (1968).Statbtfcal theories ofmental test scores. Reading, MA: Addison-Wesley. Madden, D.M. (1988). Adult age differences in the effects of sentence context and stimulus degradation during visual word recognition. Psychology and Aging, 3, 167-172. Mandel. R.G., & Johnson, N.S. (1984).A developmental analysis of story recall and comprehension in adulthood. Journal of Verbal Learning and Verbal Behavior, 23,643-659. Mandler. J.M. (1978).A code in the node: The use of story schema in retrieval. Discourse Processes, 1. 14-35. Manelis, L. (1980).Determinants of processing for a propositional structure. Memory and Cognition, 8.49-57. McKoon, G., & Ratcliff, R. (1984). Priming and on-line text comprehension. 1n.D.E. Kieras & M.A. Just (Eds.) (1984).New methods in reading comprehension research (pp. 119-128). Hillsdale. NJ: Erlbaum. Meyer, B.J.F., & Rice, G.E. (1981).Information recalled from prose by young, middle, and old adult readers. Experimental Aging Research, 7. 253-268. Meyer, B.J.F.. & Rice. G.E. (1983).Learning and memory from text across the adult life span. In J. Fine & R.O. Freedle (Eds.), Developmental issues in discourse processing (pp. 29 1-306). Norwood: Ablex. Miller, J.R., & Kintsch. W. (1980).Readability and recall of short prose passages: A theoretical analysis. Journal of Experimental Psychology: Human Learning and Memory, 6.335-354. Petros, T.V.. Tabor, L.. Cooney. T.. & Chabot. R.J. (1983).Adult age differences in sensitivity to semantic structure of prose. Developmental Psychology. 19. 907-914. Petros. T.V., Norgaard. L.. Olson, K.. & Tabor, L. (1989).Effects of text genre and verbal ability on adult age differences in sensitivity to text structure. Psychology and Aging. 4. 247-250.
90
Stine and Wlngfleld
Pichert, J.W., & Anderson, R.C. (1977). Taking different perspectives on a story. Journal of Educatioonal Psych~lOgy,69. 309-315. Reder, L.M., Wible, C., & Martin, J. (1986).Differential memory changes with age: Exact retrieval versus plausible inference. Journal of Experimental Psychology: Learnfng, Memory, and Cognition, 12,72-81. Reynolds, G.S. (1975).A prfrner of operant condftfonfng(Reutsed edftiod. Glenview, IL: Scott, Foresman. and Company. Rice, G.E., & Meyer. B.J.F. (1986).Prose recall: effects of aging, verbal ability. and reading behavior. Journal ofGerontology. 41, 469-480. Rubin. D.C. (1977).Very long-term memory for prose and verse. Journal of Verbal Learning and Verbal Behavior, 1 6.61 1-621. Rubin, D.C. (1978).A unit analysis of prose memory. Journal of Verbal Learning and Verbal Behauior, 17,599-620. Rubin. D.C. (1985). Memorability as a measure of processing: A unit analysis of prose and list learning. Journal of Experimental Psychology: General, 1 14.213-238. Salthouse, TA. (1984).Effects of age and skill in typing. Journal of Eqmimental Psychology: General, 1 13,345-37 1. Smith, S.W.. Rebok, G.W., Smith, W.R., Hall, S.E., & Alvin, M. (1983).Adult age differences in the use of story structure in delayed free recall. Experimental Aging Research, 9,191-195. Spilich, G.J.(1983). Life-span components of text processing: Structural and procedural differences. Journal of Verbal Learning and Verbal Behavior. 22.231-244. Stine, E.A.L. (1989) Online processing of written text by younger and older adults. Psychology and Agfng, 5,68-78. Stine, E.A.L., & Wingfield, A. (1987a).Levels upon levels: Predicting age differences in text recall. Experimental Aging Research. 13, 179-183. Stine, E.A.L., & Wingfield, A. (1987b). Process and strategy in memory for speech among younger and older adults. Psychology and Aging. 2.272-279. Stine, E.A.L., & Wingfield. A (1988).Memorability functions as a n indicator of qualitative age differences in text recall. Psychology and Aging, 3, 179-183. Stine, E.A.L.. Wingfield. A., & Myers, S.D. (1990).Age differences in processing information from television news: The effects of bisensory augmentation. Journal of Gerontology: Psychological Sciences, 45, 1-8.
QualitativeAge Diflerences
91
Stine, EA.L.. Wingfield, A.. & Poon, L.W. (1986). How much and how fast: Rapid processing of spoken language in later adulthood. Psychology and Aging, 1. 303-31 1. Stine, E.A.L., Wingfield. A., & Poon. L.W. (1989). Speech comprehension and memory through adulthood: The roles of time and strategy. In L.W. Poon, D.C. Rubin, & B.A. Wilson (Eds.). Everyday cognition in adult and late 1fJe (pp. 195-221). New York Cambridge University Press. Stine, W.W. (1989).Meaningful inference: The role of measurement in statistics. Psychological Bulletin, 105,147-155. Surber. J.R. Kowalski. A.H.. & Pefia-Paez. A. (1984). Effects of aging on the recall of extended expository prose. Experimental Aging Research, 10,25-28. Taub, H.A., & Mine. G.E.(1978).Recall of prose a s a function of age and input modality. Journal of Gerontology. 33,725-730. Till, R., Mross, E.F.,& Kintsch, W. (1988).Time course of priming for associate and inference words in discourse context. Memory and Cognition, 16,283-298. Tun. P. A. (1989).Age differences in processing expository and narrative text. Journal of Gerontology: Psychological Sciences, 44,P9-15. Turner. A. & Greene. F. (1978). Construction and use of a propositional text base. JSAS: Catalog of Selected Documents in Psychology, 8.3.(Ms.No. 1713) Tversky, A. (1977).Features of similarity. Psychological Review. 84,327-352. Voss. J.F., & Bisanz. G.L. (1985).Knowledge and the processing of narrative and expository texts. In B.K. Britton & J.B. Black (Eds.). Understanding expository texts: A theoretical and practical handbook for analyzing explanatory text (pp. 173-198). Hillsdale: Erlbaum. Walsh. D.A.. & Baldwin, M. (1977).Age differences in integrated semantic memory. Developmental Psychology, 13,509-514. Walsh, D.A.. Baldwin, M., & Finkle. T.J. (1980).Age dnerences in integrated semantic memory for abstract sentences, Experimental Aging Research, 6. 431-443. Wingfield. A.. Aberdeen. J. S.. & Stine, E. A. L. (submitted). Wordonset gating and linguistic context in spoken word recognition by young and elderly adults. Wingfield. A.. & Butteworth, B. (1984). Running memory for sentences and parts of sentences. In H. Bouma & D.G. Bouwhuis (Eds.). Attention and Performance X (pp. 351-363). Hillsdale: Erlbaum.
92
Stine and Wing&Id
Wingfield. A, Lahar. C.J..& Stine. E.A.L. (1989).Age and decision strategies in running memory for speech: Effects of prosody and linguistic structure. Journal of Gerontology: Psychologfcal Sctences, 44, P106-Pl13. Wingfleld. A,. Poon. L.W.. Lombardi. L.. & Lowe. D. (1985).Speed of processing in normal aging: Effects of speech rate. linguistic structure, and processing time. Journal of Cerontobgy, 40,579585. Wingfield. A., & Stine, EAL. (1986).Organizational strategies in immediate recall of rapid speech by young and elderly adults. Experimental Aging Research, 12, 79-83. Wingfield, A, Stine. E.A.L.. Lahar, C.J., & Aberdeen. J.S. (1988). Does the capacity of working memory change with age? Experimental Agfng Research, 14, 103-107. Zacks, R.T., & Hasher, L. (1988). Capacity theory and the processing of inferences. In L.L. Light & D.M. Burke (Eds.), Language, memory,and ugfng (pp. 154-1701,N. Y.:Cambridge University Press. Zacks, RT., Hasher, L., Doren, B.,Hamm, V.,& Attig, M.S. (1987). Encoding and memory of explicit and implicit information. Journal of Gerontology, 42. 418-422. Zelinski, E.M.. Light, L.L., & Gflewski, M.J. (1984). Adult age differences in memory for prose: The question of sensitivity to passage structure. Developmental Psychobgg, 20. 1 181-1192. Zelinski, E.M., & Miura. S.A. (1988). Effects of thematic information on script memory in young and old adults. P s Y c ~ ~andAgtng. ~ ~ Y 3.292-299.
Aging and Cognition: Knowledge Organization and Utilization Thomas M . Hess (Editor) 0 Elsevier Science Publishers B.V. North-Hollad. 1990
CHAPTERTHREE
AGING AND SCHEMATIC INFLUENCES ON MEMORY Thomas M. Hem North Carolina State University
SUMMARY
Variations in content knowledge have been shown to be related to both individual and developmental differences in memory performance. The availability of relevant knowledge appears to influence the efficiency of encoding and retrieval operations and to provide a meaningful interpretive structure for storing information in memory. The present chapter explores the possibility that adult age differences exist in the extent to which knowledge influences memory performance. One specific approach to the study of knowledge effects on memoryschema theory-is discussed and used as a framework for examining the relevant literature on memory and aging. It is concluded that the relation between the individual's knowledge and the nature of the memory task is more important in determining performance in older than in younger adults. Although uncertain at present, it appears as if the age dserences may be related to both the relative efficiency of knowledge-based memory processes and agerelated variations in memory styles. Jenkins (1979) has argued that a full understanding of memory performance in any particular situation requires a consideration of the interaction between the characteristics of the individual, the nature of the materials to be remembered, the strategies required for performance, and the type of criterion task. Prior to the mid1970s, the individual characteristics considered by most
94
Hess
developmental and experimental psychologists studying memory tended to be those having to do with the organization and efficiency of basic cognitive skills. Evidence of this can be obtained by examining reviews of memory research conducted during that time period (e.g., Brown, 1975; Craik, 1977;Crowder, 1976; Hagen, Jongeward, & Kail, 1975). Much of this work on memory in both children and adults concentrated on examining the efficiency of basic processes or identifying the extent to which general strategies (e.g., rehearsal, clustering) were available and/or could be used in creating and retrieving a memory representation. There was a n implicit assumption that the development of these abilities was relatively uniform across individuals, and perhaps related to some general developmental process similar to that described by Piaget (1970). More recently. researchers have come to recognize the importance of considering the knowledge possessed by the individual about specific domains as a characteristic that might influence memory performance. Although it had often been suggested that what individuals know about the world should d e c t what and how they remember, knowledge influences were for the most part ignored in earlier research. For example, Piaget and Inhelder (1973)talked about the influence of "memory in the wider sense" on remembering. yet their research concentrated primarily on the effects of abstract logical structures. In what ultimately has come to be a very influential book, Bartlett (1932)also advocated for consideration of knowledge influences on memory performance. His ideas about schematic memory processes, however. were largely disregarded (at least in the United States) until relatively recently (see Brewer & Nakamura. 1984). Cun-ently, in conjunction with the growth of cognitive science as a discipline, increasing attention is being paid to the role of organized knowledge structures in memory processing (e.g.. Galambos. Abelson, & Black, 1986;Rybash, Hoyer. & Roodin, 1986). Research in this area has made it clear that knowledge has an and impact on each of the four factors discussed by Jenkins (1979). that age-related changes in knowledge structures and knowledgebased processing play an important role in determining developmental differences in memory functioning (see Chi & Ceci,
Schematic InJuences on Memory
95
1987). The purpose of the present chapter is to describe briefly the role of knowledge in memory processing, and then to examine the evidence regarding possible adult age differences in the degree to which knowledge influences retention. It will be emphasized that the consideration of knowledge factors and their relation to aging is necessary for a more complete understanding of everyday functioning in adulthood. Included will be a discussion of schemabased notions of memory, and their possible utility in understanding aging effects on memory performance.
KNOWLEDGE AND BE?WEEN GROUP VAFUAT'IONS IN MEMORY PERFORMANCE Individual Differences Most of the research examining knowledge effects on memory h a s been concerned with the interaction between subject characteristics and the nature of the to-be-remembered materials. However, this same body of research has indicated that a consideration of knowledge influences must take into account all four factors in Jenkins' (1979)conceptual model. This complex interaction is demonstrated nicely in a set of experiments (Chiesi, Spilich, & Voss, 1979: Spilich, Vesonder, Chiesi. & Voss, 1979: Fincher-Kiefer, Post, Greene. & Voss. 1988)in which the memory performance of young adults possessing high and low levels of knowledge about the game of baseball was compared. For example. illustrating an interaction between individual characteristics, materials, and tasks, Fincher-Kiefer et al. (1988)found that Individuals with high knowledge had greater reading spans than low knowledge subjects when the stimulus sentences were baseballrelated and when memorization of these sentences was also required. When baseball-neutral sentences were used, or when the task required comprehension rather than recall, no knowledge effects were obsemed for reading span. In addition, Chiesi et al. (1979)and Spilich et al. (1979)found that high knowledge individuals were superior to low knowledge subjects in memory for a narrative of a fictitious baseball game. The group dflerences in memory performance appeared to be due to
96
Hess
the high knowledge individuals being able to relate the to-beremembered event to their knowledge of the goal structure of baseball, which in turn facilitated retrieval of the event from memory. Those subjects who possessed little knowledge of baseball’s goal structure were unable to systematically organize the material, resulting not only in lower levels of recall, but also in disproportionately poorer recall of information relevant to the goal structure of the game. Fewer group Werences were observed in memory for irrelevant narrative details. The results of these studies demonstrate that, in certain types of remembering situations, high levels of knowledge can reduce the demand on cognitive resources through provision of appropriate processing structures. These structures appear to provide organizational schemes that can be used to guide encoding and retrieval operations. Indeed, the aforementioned researchers have argued that it is primarily this latter function that distinguishes the performance of individuals varying in knowledge. Speciflcally, the flndings of Fincher-Kiefer et al. (1988) suggest that high and low knowledge individuals differ primarily when they are required to use their knowledge to organize input and develop retrieval structures. When such demands are made, only low knowledge individuals exhibit a decrease in reading span for domain-related information, presumably because they need to expend more of their cognitive resources to interpret the text. Analyses of recall data (Spilich et al., 1979) using Kintsch and van Dijk’s (1978) model of text comprehension also support these conclusions. High and low knowledge individuals did not M e r in the size of their working memory buffer during processing of domain-related text, nor in their ability to construct a micropropositional text-base, suggesting no knowledge effects on comprehension. The subjects did diEer, however. in their ability to construct a macrolevel representation of the narrative, which reflects more of an interpretation of the information in the text as opposed to simple comprehension. Apparently, it is this macrolevel representation, and its inherent organizational function, that facilitates memory performance in high knowledge individuals.
Schematic InJuences on Memory
97
Other research, including studies of expertise (e.g., Chase & Simon, 1973). has also shown that individual variations in knowledge of a domain result in differences in patterns of domainrelevant memory performance. The importance of considering knowledge has similarly been demonstrated in studies that manipulate the availability and/or activation of relevant structures in the same individual (e.g.. Bransford &Johnson. 1972; Kintsch, 1977; Mandler & Johnson, 1976). and that vary the relationship of the to-be-remembered materials to the subject's knowledge (e.g., Brewer & Treyens. 1981:Friedman. 1979;Graesser, Woll, Kowalsld, & Smith, 1980). In all cases, memory performance is observed to vary as a function of the relationship between the test stimuli and the individual's activated knowledge. 'Qpically, the availability of a relevant knowledge structure results in enhanced memory performance. However, this does not appear to be the only result of knowledge-based processing. There is substantial evidence suggesting that the individual's knowledge of the world affects the form of the memory representation as well as the overall level of retention. For example, the availability of relevant knowledge appears to result in greater use of interpretive and selective processes, with the remembered event often being transformed to be consistent with existing knowledge (e.g., Brewer & Nakamura, 1984; for a critical review of this research. see Alba & Hasher. 1983). Developmental Differences Researchers interested in memory development across the lifespan are also paying increased attention to the role of knowledge in determining age differences in memory performance. In part, this is based upon the obvious fact that age is positively correlated with the extensiveness of experience, and presumably the related development of organized systems of knowledge. Given the previously discussed findings concerning individual difl'erences, it makes sense then that the development of such systems would be related to changes in mnemonic skills at all stages of life.
98
Hess
Childhood
There is increasing evidence that the growth of mnemonic skills during childhood is partially dependent upon the development of organized knowledge structures (see Chi & Ceci, 1987). For example, Chi (1978) demonstrated that the form of the age effect on a relatively simple memory task is related to the congruence between the the subject's knowledge base and the nature of the to-be-remembered materials. Specifically, she found that the typically observed advantage of adults over children on a memory span task using digits a s stimuli was reversed when the stimuli (meaningful chess patterns) were more familiar to the children (chess experts) than to the adults (chess novices). The implication is that the availability of a richly elaborated knowledge structure facilitates the encoding and retrieval of domain relevant information. For the most part, children possess a more impoverished knowledge base than adults. and therefore will usually perform at a lower level when the task materials have no specific relation to age other than amount of total experience. Other developmental researchers have suggested that the availability and use of strategies in children is at least partially dependent upon knowledge relating to the to-be-remembered materials (e.g., Bjorklund, 1987; Omstein, Baker-Ward. & Naus, 1988). Specifically, strategies such a s rehearsal and clustering are more likely to appear in young children when the connections between stimulus items are highly familiar (e.g., Bjorklund & Zeman, 1982; Chi Br Koeske. 1983). Bjorklund (1987) has argued that organizational strategies develop as children become aware of linkages between items based upon automatically activated associations. It has also been hypothesized that the strength of these connections helps reduce the cognitive requirements of the memory task, thereby facilitating the implementation of mnemonic strategies (Omstein et al.. 1988).
Researchers studying aging processes have also come to recognize the importance of considering knowledge by materials
Schematic InJuences on Memory
99
interactions in the study of adult age differences in memory performance. Several studies (e.g., Barrett & Watkins, 1986; Hultsch 81 Dixon. 1983; Worden & Sherman-Brown, 1983) have found that the degree and direction of age differences in memory for various types of materials depends upon the meaningfulness of the stimuli to the age groups tested. Less work has been done examMng knowledge effects on the use of specific strategies with older adults. However, the results of studies on expertise have suggested that the availability of relevant knowledge may serve to counteract some of the negative changes in general cognitive skills associated with aging (see Charness. 1985). It is also clear from other empirical and theoretical literature in aging that a focus on knowledge-related processes in memory could prove very beneficial in understanding changes in memory functioning in adulthood. For example, empirical work in the psychometric tradition on intellectual skills in adulthood suggests that knowledge-based memory processes may become more influential with age. This research has highlighted the multidimensional nature of intelligence, and has demonstrated that there are different developmental trajectories for different types of abilities (e.g., Schaie. 1979). Broadly speaking, there appear to be a least two major groups of abilities. The first is related to general cognitive skills and structures considered to be part of the basic information processing system. Cattell ( 197 1) and Horn (1970) have referred to these a s fluid intellectual skills, whereas Baltes, Dittmann-Kohli, and Dixon (1984) have termed them the mechanics of intelligence. The second category of abilities, termed crystallized intelligence (e.g.. Horn. 1970) or the pragmatics of intelligence (Baltes et al., 1984). encompass both the specific knowledge systems that have been created through experience, and the procedural knowledge and skills that have developed within these systems. The evidence indicates that, in concert with changing physiological and socio-environmental structures, the mechanics of intelligence decline with age, whereas the pragmatics maintain or continue to develop throughout adulthood. The reduced efficiency of mechanics is often seen as the basis for older adults being less able than younger individuals to benefit from novel experiences and to perform well on standard laboratory
100
Hess
tasks. One might also use this psychometric data, however, to argue that aging is associated with the emergence of a qualitatively different manner of dealing with cognitive tasks. For example, given that aging is associated with a reduction in the cognitive skills necessary to learn by "brute force" (i.e., without the benefit of experience), there may be a shift in the types of sktlls relied upon in remembering situations. Perhaps recognizing that they are no longer a s effective in retention situations as they used to be, older adults might develop alternative strategies relying more on establishing a meaningful relationship between what they know and what they are being asked to remember. That is, remembering activities may become more knowledge-based with age in adulthood. Framing these ideas somewhat differently, there may be age differences in the degree to which individuals use top-down versus bottom-up (or conceptually driven versus data-driven) processing (e.g., Rumelhart, 1984) based upon changes in cognitive abilities. Bottom-up processing requlres the development of a coherent representation of a stimulus event by the individual through analysis of the event components and the subsequent identiflcation of structure between the parts (i.e., from part to whole). Such processing would appear to be dependent upon the mechanics of intelligence. In contrast, top-down processing is guided by the individual's expectations, and the processing of the event consists of a search for confirming or disconfinning evidence for the expectations (i.e.. from whole to part). This type of operation appears to be more reliant on pragmatic abilities. Normally, most memory activities involve both types of processing. However, given the just-described changes in intellectual skills with aging, it might be hypothesized that the memory perfomance of older adults would be characterized by a disproportionately greater emphasis on top-down processing. It would follow that age differences in memory performance in any g M n situation would depend upon the extent to which each of these types of processing are emphasized. Age dmerences in performance should be greatest in situations which rely heavily on data-driven processing, and least in situations where performance is primarily dependent upon conceptually driven processing.
Schematic Influences on M e m o r y
101
Changes in the influences of knowledge-based memory processing during adulthood might also be predicted based on current theoretical notions concerning age-related variations in processing resources and working memory capacity (e.g., Craik & Rabinowitz, 1982;Zacks & Hasher, 1988). Specifically, it has been hypothesized that aging is associated with a reduction in available processing capacity, and that, relative to younger adults, the memory performance of older adults is disproportionately affected a s task demands on capacity increase. These negative effects of aging-related capacity changes, however, might be attenuated in situations where performance is dependent upon access to richly elaborated knowledge systems. As mentioned previously. dealing with information that is highly relevant to existing knowledge systems appears to result in enhanced cognitive efficiency, a s indicated by measures of capacity and processing speed (e.g., Fincher-Kiefer et al.. 1988;but also see Britton & Tesser, 1982). Part of this increased efficiency may be related to the ease of creating and accessing knowledge-based connections for organizing input and output. Thus,it might be predicted that the reduction in general processing capacity hypothesized to occur with aging would not adversely affect performance in meaningful contexts in the same manner a s it would in more traditional laboratory situations. In general, age dtfferences in memory accuracy would be smallest in those situations in which information can be easily related to. and subsequently retrieved through, an e x i s m knowledge structure. Several theorists have argued that a n increase in knowledgebased memory processing may occur with age a s part of normal development in adulthood rather than a s a result of or compensation for deficits in other skills (e.g., Baltes et al., 1984; Hoyer, 1985, 1987;Labouvie-Vief, 1980;Labouvie-Vief & Schell, 1982;Ftybash et al., 1986). These theorists have not ignored the impact of the declining efficiency of basic cognitive and mnemonic skills. but rather emphasize the importance of also considering ongoing development related to specific knowledge systems. For example, Labouvie-Vief [ 1980: Labouvie-Vief & Schell, 1982)has argued that knowledge systems continue to develop during adulthood, with component parts being increasingly subsumed within the larger whole, and superordinate information becoming
102
Hess
more easily activated. With time and experience, processing within these systems results in more interpretive analyses based on the total structure. and less attention to and memory for details. Hoyer and his colleagues (Hoyer, 1985, 1987; Rybash et al.. 1986) have also stressed the importance of considering knowledge systems in understanding aging. They suggest that development during adulthood is characterized by the growth of organized knowledge structures within speciflc domains, and that cognitive skills become increasingly encapsulated within these domains. It is this specialization of skills that is hypotheslzed to be partly responsible for the decreased efficiency of more general cognitive operations. The implication is that basic cognitive skills (i.e.. the mechanics of intelligence) are more important at earlier ages when adaptive knowledge structures are being constructed, but decrease in importance as most behavioral interactions increasingly occur within highly familiar contexts. An obvious implication of these latter theoretical viewpoints is that studying performance in tasks that make some connection to what the subject knows about the world will provide a more ecologically valid and complete view of aging and memory skills (see also Hartley, Harker, & Walsh, 1980). At present, it is unclear whether performance on standard laboratory tests that bear little resemblance to situations encountered in everyday living is predictive of functioning. In fact, preliminary evidence in the aging literature suggests that performance on such tests is uncorrelated with performance in more meaningful situations (e.g., Sinnott, 1986; also see Cornelius,this volume). The viewpoint adopted here is that a consideration of knowledge-based retention factors, within either laboratory or more "real-world" settings, will allow more straighfomard inferences about everyday functioning. One specific approach that has been used to examine knowledge influences on performance in adulthood is the study of expertise and aging (e.g., Chamess, 1981a, 1981b, 1983; KIiegl& Baltes, 1987; Salthouse, 1984). This research has generally been concerned with identifying how expert knowledge is related to the maintenance and continued development of skill within the specific area of expertise across the adult age-span. These studies have provided valuable information about the plasticity of development in adulthood, and
Schematic Influences on Memory
103
demonstrated how expertise may counteract other negative aspects of cognitive aging. The generalizability of this research to understanding everyday memory functioning may be somewhat limited, however, due to the selective nature of the populations and types of knowledge studied. An alternative approach to the study of aging and knowledge influences is to examine the impact of relatively common structures that are presumed to be operative in organizing experience across a wide variety of situations and individuals. Much of daily life involves participation in social interactions and routine activities in which organized, often socially shared knowledge structures guide our actions. For example, when washing clothes, it is assumed that we activate a laundry script, which provides information about the component actions of this activity along with their sequence and conditions of performance (Schank & Abelson, 1977). Similarly, we understand others by integrating our perceptions of their behavior with our knowledge about occupations, personality traits, and social contexts (Black, Galambos. & Read, 1984). The mnemonic consequence of this type of cognitive behavior is that our memory for events occurring during these activities is embedded within the context of the specific knowledge structure activated at the time, thereby influencing both what is retained about the event and the organization of the memory trace. In addition, recall of events is facilitated through the use of conceptually driven retrieval processes in which the originally activated knowledge structure serves as a guide (e.g.. Reiser. 1986). Given the presumed importance of knowledge-based retention activities in everyday functioning. the study of aging and everyday memory functioning would benefit from the examination of related processes. This is especially important given that the organization of common knowledge structures appear to be relatively stable across the adult life-span (e.g.. Hess, in press: Light & Anderson. 1983). One specific framework that has been used to examine knowledge influences on memory is founded in schema-based notions about memory, It is proposed that this relatively welldeveloped framework might be useful in developing an understanding of age differences in knowledge-based remembering
104
Hess
during adulthood by identifying specific types of relevant processes, and thereby providing a focus of study. A brief discussion of this framework follows in order to provide a context for examining the relevant aging research. SCHEMAS AND MEMORY
Findings such as those detailed above concerning knowledge influences have led to a resurgence of interest in the concept of the schema as a possible explanatory mechanism. Bartlett's ( 19321 original work on schematic processes was a direct reaction to the prevailing British Empiricist approach to the study of memory prevalent at the tfme (see Brewer h Nakamura. 1984). In particular, Bartlett was opposed to the atomistic assumptions concerning representation of knowledge and experience in memory. Instead, he argued for the existence of unconscious, organized mental structures representing past experience (i.e., schemas). In addition, he suggested that all incoming information interacts with these structures, and is interpreted and remembered in the context of this interaction, typically resulting in modified representations of events in memory. These notions have formed the basis for more contemporary views on schemas and memory. At present, there is no comprehensive theory of schematic memory, though several fairly formal theories exist that deal with specific types of schemas, such a s scripts (e.g., Graesser & Nakamura, 1982; Schank & Abelson, 1977). Mandler (1984). however, has suggested that there is a general schema framework for the study of memory, consisting of general assumptions shared by most theorists concerning the nature and operation of schemas. These assumptions are briefly outlined below. More detailed discussions may be found in Alba and Hasher (1983). Brewer and Nakamura (1984). Mandler (1984), and Rumelhart (1984). Schema Structure Schemas are typically depicted as organized mental representations of generic knowledge concerning specific domains or classes of events. The components of a schema specify
Schematic Influences on Memory
105
information about objects or events that normally occur within the knowledge domain. Schemas are thought to be hierarchically organized, though the relationship between the components is not one of class inclusion (Mandler, 1984). Rather, the organization is more closely akin to the notion of a collection (see Markman. 1981). in which the whole is defined by the total sum of the parts rather than by a set of defining attributes. For example, one of the more frequently studied types of schemas is the script, which represents information about commonly enacted activities (e.g., going to a restaurant) (Schank & Abelson, 1977). Placement of individual components (i.e.. constituent actions, such as reading the menu, ordering drinks) within the script hierarchy is determined by their role in defining the activity rather than by their possession of a specific set of defining attributes. It is also assumed that interconnections exlst between elements at any given level of the hierarchy, with activation of one component resulting in activation of both other components at that level (e.g., receiving a menu might activate ordering food) and ultimately the entire structure. These connections within levels also distinguish schemas from class-inclusion structures. The connections between components vary in both their nature and processing characteristics. depending upon the type of schema. For example, the components of the just described script sequence may be connected through temporal (e.g., appetizers typically precede the main course) or enabling relations (e.g., getting your meal allows one the possibility of eating it), whereas scene schema components may be linked primarily through spatial connections (e.g., table lamps typically appear on tables in living rooms). Brewer and Nakamura (1984) have suggested that it is this diversity in schematic structures that is one of the main stumbling blocks to the development of a comprehensive script theory. They argue that acceptance of a modularity assumption, in which this diversity is recognized, may be more fruitful in the long run for theory development than the search for more parsimonious principles.
106
Hess
Processing Characteristics As stated before, one of the main assumptions of schema theories of memory is that all retention activities take place within the context of the schemab) currently activated in memory. Schema activation can take place in at least two ways (Rumelhart, 1984). First, the indlvidual may activate a schema before the event occurs, based on prior information. For example, if a person knew they were about to meet a psychology professor, their knowledge regarding characteristics of individuals in that esteemed profession would be activated, and their perceptions of the individual would be influenced by expectations derived from this knowledge. Second, activation can also occur through encountering a schema component in the stimulus array. For example, identifying a stove and a refrigerator in a picture might lead to the activation of a kitchen schema. In both cases, the schema would guide processing in a top-down fashion, but, in the latter case, this would be preceded by bottom-up processing to aid schema identification. The to-be-remembered went is then processed in relation to the activated schema, which provides an interpretive framework for comprehension. This framework serves to facilitate the establishment of a coherent memory trace by relating the event components to each other, thereby preserving episodic information in memory (e.g.. Bransford & Johnson. 1972; Dooling & Lachman, 1971). It is this process of establishing coherence that also often leads to the modification of the memory trace. Specifically, it is assumed that the to-be-remembered event is integrated with the schema, resulting in a memory trace representing the instantiation of the schema with specific episodic information that distinguishes it from other instantiations (e.g., Schank. 1982). In addition to this nonperceptual information being added to the event representation in memory, it is also assumed that many irrelevant surface details are lost through the operation of meaning-preserving abstraction processes. Although there may be reason to question these strong assumptions concerning, for example, the operation of schemabased integration processes (Alba & Hasher, 1983). there is ample evidence indicating that people do have difficulty discriminating
Schematic ZnJuences on Memory
107
between presented and unpresented information that is obligatory or highly typical in nature (e.g., Friedman, 1979: Graesser et al., 1980). This suggests that some schema information is incorporated into the memory trace. The establishment of coherence also results in selective processing and retention, which is dependent upon the relationship of the to-be-remembered information to the schema. Although there are some contradictory findings, in general, research has shown that schema-irrelevant information is retained less well than relevant information, at least with the passage of time (cf. Brewer & Nakamura, 1984). This effect appears at least partially due to the fact that irrelevant information is less likely to be incorporated into the schematically organized memory trace, thereby decreasing the probability that it will be accessed when the schema is used to guide retrieval (e.g., Anderson & Pichert, 1978; Kardash, Royer, 81 Greene, 1988). In addition, relevant infomation that is novel or unexpected tends to be remembered better than predicted or expected relevant information, an effect that appears attributable to variations in the extensiveness of encoding operations. For example, Hastie (1984) found that behavioral information that was inconsistent with an actor's perceived traits was remembered better than consistent information. The difference in memory performance appeared to be due to subjects' more extensive processing of inconsistent behaviors a s they attempted to explain the deviations from expectations. In sum, there appears to be ample evidence suggesting that schematic knowledge has an &ect on retention that is based on the individual's attempt to understand the to-be-remembered event in terms of the schema. While facilitating retention by providing an organizational framework, such processing also results in a modification of the memory trace through the processes of selection and integration, whereby certain types of information are added. deleted. or modified based on their relation to the schema. Naturally, there have been criticisms of schema-based theories of memory. Most of these have centered around the vagueness of the processes specified about schema construction, modification, and operation (e.g., Alba & Hasher. 1983: Thorndyke & Yekovich. 1980). As mentioned before, Brewer and Nakamura (1984) have argued
108
Hess
that at least part of this vagueness results from theorists attempts to develop parsimonious explanations of schematic processes that apply across the various types of schemas. Theorists that have concentrated on more limited schematic structures (e.g., scripts, scenes) have tended to achieve more success in detailing these processes. In spite of this vagueness, the consideration of between age group differences in general processes relating to the supportive and integrative aspects of schematic processing could provide a reasonable starting point for a more systematic exploration of aging and howledge-based retention. Alba and Hasher (1983) have also criticized some of the basic assumptions concerning the representation of infomation in memory. Specifically, they have suggested that schema theories propose that memory for an event is modified through the schemabased processes of selection, abstraction, integration, and interpretation. Alba and Hasher argue, however, that the actual memory trace appears to be richer (in terms of surface details) than would be expected if these processes were operative. They go on to suggest that many of the presumed schematic effects thought to be related to encoding may actually be based in retrieval processes. For example, Anderson and Pichert (1978) found that previously irrelevant, unrecalled information could be retrieved if a relevant retrieval scheme were used. As Mandler (1984) has pointed out, however, a schema-based approach to the study of memory does not preclude the possibility that surface structure is represented in memory; schematic processes do not necessarily operate in an all-or-none fashion. Indeed, there are several models of text comprehension (e.g., Graesser & Nakamura, 1982:van Dijk & Kintsch, 1983)which do allow for the representation of both interpretive and veridical information. It appears that the extent to which schematic processing will influence memory performance is dependent upon the meaningfulness of the task and other task-related factors. Some of the specific task conditions identified by researchers which appear to affect schema use are a s follows: (a) study instructions, such as interpretation/comprehension versus rote memorization (e.g.. Hamilton, Katz, & Leirer, 1980). (b) type of memory test. such as recall versus recognition (e.g., Graesser h
Schematic InJuences on M e m y
109
Nakamura. 19821, (c) time elapsed between study and test (e.g.. Smith 81Graesser. 1981). (d)type of materials (e.g.. Sulin & Dooling. 1974). (e) the knowledge ofthe individual (e.g.. Spilich et al.. 1979). and (fj motivations of the individual (Levine, Chein, & Murphy, 1942; Zadny & Gerard, 1974). With respect to our examination of knowledge factors and aging, distinctions relating to the probability of schema use may be helpful in identifying age differences in the influence of such processes. For example, if older adults are less efficient at performing general memory operations (e.g.. rehearsal), then little daerence in the pattern of memory performance might be observed for these subjects across conditions emphasizing memorization versus comprehension of the stimulus materials. Relatively automatic schema-based top-down processes might be the primary determinants of performance in both cases. In comparison, younger adults might exhibit a greater contrast across conditions due to their more efficient memory skills, the use of which might serve to counteract the schematic effects likely to be found with comprehension instructions. The remainder of this chapter is devoted to a selective examination of aging studies that provide information regarding adult age difl'erences in schematic processing. The overall goals of this review will be to determine whether age differences exist in knowledge-based retention processes. and if so, to identifjr the conditions under which such differences occur, and to evaluate possible explanations for any age differences. For example, if older adults do show greater reliance on knowledge-based processing, is it attributable to the relative ease of such operations, thereby reflecting a compensatory effect due to losses in other ability domains? Or, alternatively, does it indicate a normative developmental progression that is unrelated to deficiencies in other skills? AGING, MEMORY AND KNOWLEDGE
There is much aging research that has direct relevance to the study of schematic memory processes, even though most of this work was not originally conceptualized within a schema
110
Hess
framework. This section will begin with a discussion of some of this research followed by a more detailed examination of investigations from our lab and elsewhere that have focused on the study of specific types of schematic structures, such a s scripts and scene schemas. Research on Schema-Related Issues
Meaningfulness There is a body of aging research that has dealt with the issue of stimulus familiarity or meaningfulness in the study of memory processes. The results of such studies have direct relevance to the present focus on schematic processes by providing evidence regarding age differences in memory performance as a function of the degree to which a supportive knowledge base is available. Some of this work was originally done in response to the possibility that materials used in cross-sectional studies of remembering were more meaningful to the younger participants than to the older ones, thereby resulting in biased views of aging effects on performance. In these studies, researchers have varied the meaningfulness of the to-be-remembered infomation in relation to the age groups studied in order to examine the effects on the nature of the age differences obtained. In general. this research has found that the size and direction of the age effects on memory are directly related to the meantngfulness of the stimuli to the age groups studied (Barrett & Wright, 1981: Barrett & Watkins, 1986; Hanley-Dunn & McIntosh, 1983: Howell, 1972; Hultsch h Dixon, 1983: List, 1986; Worden & Sherman-Brown, 1983). A study by Worden and Sherman-Brown (1983) is fairly representative of this research. They presented young and older adults with four different lists of words in a standard recall task. The lists varied in terms of the frequency of usage of the words in each one, as determined by two sets of cohortspecific norms (Kucera & Francis, 1967: Thorndike, 1921) that were assumed to be reflective of word usage during the youth of the two age groups studied. One test list consisted of dated words that were high in frequency in 1921, but low in 1967 (e.g., bosom, toil). A
Schematic Influences on Memory
111
second list consisted of contemporary words that were high in frequency in the 1967 norms, but low in the older set (e.g., energy. radical). The third list was comprised of popular words that were high in frequency in both sets of norms (e.g., family, standard), and the fourth list contained rare words that were low in frequency in both sets [e.g., brevity, lunatic). It was found that both age groups exhibited generally better memory for words that were high in frequency relative to their cohort than for low frequency words. In addition, the direction of the age effect varied by list, with older adults exhibiting superior performance on the dated list, whereas the young adults performed better on the contemporary list. Part of the reason for these effects may be due to the ease with which familiar information is encoded into memory. Poon and Fozard (1978) found a similar cross-over effect in object naming latencies, with cohort-relevant objects being named faster than cohortirrelevant items. More importantly for present purposes, however, was Worden and Sherman-Brown's (1983) finding that the effects of frequency were greater for the older adults than for the young subjects. That is, the memory performance of older adults was differentially depressed when words were low in frequency. In addition, other studies have found that the age differences in performance with old cohort-speciflc stimuli are smaller than those obtained with young cohort-speciflc materials (e.g.. Barrett & Watkins, 1986; Hultsch & Dixon, 1983). This suggests that the supportive aspects of knowledge may be relatively more important in determining memory performance in older than in young adults. This may be due to the fact that these latter subjects stffl have the cognitive resources necessary to establish an accurate memory trace when meaningfulness is low. The results of other memory research on meaningfulness factors have been somewhat mixed. For example, several earlier studies have examined the effects of the degree to which stimulus materials approximate normal English (Craik & Masani, 1967, 1969; Heron & Craik, 1964; Kinsbourne, 1973; Taub, 1974). Unfortunately, the results of this research have been equivocal with respect to the existence of age differences in the degree to which such structure supports memory. Other research has suggested that
112
Hess
age differences in memory are greater when subjects must impose their own organization on the to-be-remembered materials than when more direct guidance is available (e.g.. Hultsch, 1969, 1971). More recently, Stine and Wingfield (1987) and Wingfield. Poon. Lombard. and Lowe (1985) have found that the sentence recall of older adults was negatively afTected to a greater extent than that of young adults when semantic and prosodic cues were removed. Thus. as with the research on familiarity, the suggestion from this work is that the match between the memory task and the knowledge base available to the individual to support retention is more important in detennfning performance in older than in younger adults.
Integration Another aspect of schematic processing involves the integration of perceptual information with structures already residing in memory. As discussed previously, one result of this process is the addition of inferential information necessary for the establishment of coherence to the memory representation. One obvious precursor for the operation of such integrative processes is the activation of the appropriate knowledge system during processing. Given available work on priming effects in semantic memory (for review. see Howard, 1988a), it appears that there are few age dmerences in the qualitative aspects of knowledge activation. Older adults may require somewhat more time for activation to occur (e.g.. Howard, Shaw, & Heisey, 1986). but the conditions under which priming takes place and the strength of priming effects are similar across ages. Given that patterns of activation are similar, what about possible age differences in the construction of an integrated memory representation? One way in which this could be examined is by looking at the extent to which implied information is processed and remembered. For example, does knowledge activated in memory to aid comprehension then become part of the memory trace? In one relevant study. Till and Walsh (1980) examined the degree to which information implied in target sentences facilitated recall of those sentences. Relative to free recall. they found that implicational cues facilitated sentence recall In the young adults,
Schematic InJuences on Memory
113
but actually inhibited recall in older adults. The suggestion from this work was that older adults are less likely to spontaneously infer implied information. The results of subsequent studies, however, have indicated otherwise. For example, Burke and Yee (1984) found no age differences in the level of priming for implied sentence content in a lexical decision task. This suggests that such information is processed on-line in a similar fashion by both young and older adults. In another related study, Hess and Arnould (1986) investigated the extent to which inferred information was part of the memory representation by examining false recognitions for implied instruments. They found that the overall false alarm rate for instruments implied in the study sentences was significantly greater than that for nonimplied instruments. More importantly, there were no age dflerences in these false alarm rates, suggesting similar degrees of trace integration in young and old adults. The discrepancy between the results of these two studies and those of Till and Walsh (1980) may have to do with the methodologies used. Burke and Yee (1983)and Hess and Arnould (1986) examined the processing of implied infomation in a relatively direct manner by measuring on-line processing and the presence of the actual information in memory. Till and Walsh (1980). on the other hand, used a somewhat more indirect procedure in which inferences about encoding were made from perfomance on a task heavily reliant on retrieval processes. It may be that their results were due to age differences in the degree to which implicit cues can be used to retrieve specific sentences. Other research related to the above has examined aging effects in inferential skills (for reviews, Light & Albertson, 1988; Zacks & Hasher, 1988: Zelinski, 1988). This work has suggested that age differences in the ability to make inferences are directly related to the importance of the inference to on-line comprehension processes and the demand put upon storage functions in working memory (Zacks & Hasher. 1988). Age differences tend to be minimal if the inference is a straightforward result of knowledge activation or requires reference to recently processed information in working memory (e.g., Belmore. 1981: Light, Zelinski, & Moore, 1982). However, as the task increasingly requires access to
114
Hess
information no longer active in working memory, either in terms of previously presented text or long-term memory conceptual structures. age Merences also increase (e.g., Cohen. 1979:Light et al., 1982:Zacks, Hasher, Doren. Hamm, & Attig. 1987). Work by Pezdek (1980;Pezdek & Miceli, 1982)has also indicated that older adults are poorer at integrating relevant information across modalities (i.e,, sentence and picture), but that increasing the study time for the test items facilitates their integration. Thus, hypothesized limitations in processing capacity may somewhat hinder older adults' ability to create a coherent representation in memory (see also Stine, 1990). Other research, however, has suggested that older adults are able to form integrated memory representations even though they possess poorer general memory skills than younger adults. In a series of studies by Walsh and his colleagues (Walsh & Baldwin, 1977;Walsh, Baldwin, & Finkle, 1980)using Bransford and Frank's (1971)linguistic abstraction procedure, no age differences were found in the ability to construct semantically integrated idea units. This was in spite of the fact that the older adults exhibited poorer performance in a free recall task. Similarly, research in our lab (Hess, 1982:Hess & Slaughter, 1986a, 1986b) has found that there! are few age differences in the ability to abstract central tendency information from a set of visual stimuli. This was true even though the older adults had poorer memorry (both short- and long-term) for individual stimuli and exhibited lower levels of performance than young adults on standard short-term memory tests. Related to the establishment of meaning through the construction of an integrated memory trace is the schema theory assumption that irrelevant details, surface or otherwise. are lost whereas gist is preserved, an effect that increases over time. There are several pieces of evidence relevant to this assumption. First, studies of word memory (Hess, 1984: Hess & Higgins, 1983; Rabinowitz. Craik. & Ackerman. 1982) have suggested that the general semantic aspects of individual words are remembered equally well by young and older adults, whereas young adults appear to retain more about specific surface features or contexts in which the word appeared. For example, Hess and Higgins (1983) tested recognition for target words in the same semantic context
Schematic Influences on Memory
115
presented at study or in a new context that emphaslzed a difTerent meaning of the word. They found that changing the context resulted in a greater reduction in correct recognitions for the older adults than for the younger adults, the implication being that younger adults had encoded more surface detail about individual words which allowed them to be recognized across contexts. Further support for this conclusion is found in studies in which older adults make more semantic confusions than young adults on recognition tests (Hess, 1984;Hess & Higgins, 1983;Rankin & Kausler, 1979; Smith, 1975). Other studies have obtained related flndings using a vaxlety of stimuli. For example, Cavanaugh (1983)found that age differences in memory were greater for details of a television program than for central theme information. especially in subjects of low verbal ability. Research on memory for pictures has suggested that age differences are greater for comprehension-irrelevant details (e.g., picture orientation) than for more general theme-relevant content (Bartlett. Till. Gernsbacher. and Gorman. 1983; Pezdek. 1987). Ratner and her colleagues (Padgett & Ratner, 1987;Ratner, Padgett. 8r Bushey, 1988)have found that age differences in memory are smaller for superordinate than for subordinate event information, though the age effects are eliminated if exact recall is emphasized. This suggests some between-group variation in processing styles, with older adults preferring to concentrate on thematic information and ignoring details. Finally, several studies of text memory (e.g., Spilich, 1983;Zelinski, GilewsM, & Thompson, 1980) have found that age difTerences are smaller for high level text propositions than for low levels ones, though this effect is inconsistent across studies (e.g.. Cohen. 1979; Dixon, Simon, Nowak, & Hultsch, 1982)and may be mediated by ability factors, such as verbal skills (e.g.. Dixon. Hultsch. Simon, & von Eye, 1984). Hultsch and Dixon (1984)have suggested that the degree to which age differences exist for high level versus low level information may also be related to the structure of the stimulus texts. Well-structured texts are more typically associated with greater age dnerences for details, whereas poorly structured texts are associated with greater age differences in higher-level information. If it is assumed that the availability of high-level
116
Hess
information in memory is related to the processing of thematic information, this pattern of results may indicate that older adults are more dependent upon structure for meaning derivation. The results of a study by Smith, Rebok, Smith. Hall, and Alvin (1982) tend to support this interpretation in showing that. relative to young adults, the memory performance of older adults is more adversely affected by moderate degrees of disruption in normal story structure . Both age groups are negatively affected to a similar extent by more severe disruptions (also see Mandel & Johnson, 1984).
In general, this research suggests that age differences are greatest for irrelevant and surface details of information in many situations. Some of these results have been explained with reference to age differences in encoding-based factors, such as the use of organizational strategies or the processing capacity available for performing relevant encoding operations (e.g., Hess, 1984). It may also be the case, however. that age differences exist in retrieval operations. For example, older adults might be more likely to rely on conceptually driven retrieval, which would generally result in poor recall of information unrelated to the conceptual structure used to access information in memory. A study by Reder, Wible, and Martin (1986) has provided clear support for this possibility. They presented young and older adults with a story, and then examined recognition responses to target and distractor items that varied in their plausibility with respect to the story. It was found that older adults were more likely to use a plausibility strategy in making their recognition responses, whereas young adults used a strategy based on direct retrieval from memory. This same pattern of performance was obtained on a semantic memory task involving category judgments. This suggests that the use of such a strategy is not simply a consequence of the manner in which episodic information is encoded into memory, but rather reflects differences in retrieval mechanisms across age groups. Reder et al. (1986) suggested that the older adults rely on plausibility judgments because they are less effortful to perform than a direct search of memory contents. They do not, however, rule out possible age dmerences in retrieval styles due to other factors.
Schematic Influences on Memory
117
We have also obtained some evidence suggesting that older adults are more reliant on conceptually driven retrieval from a study in which we examined age differences in the acquisition of category information (Hess & Wallsten, 1987). Young and older adults were required to learn to distinguish between two sets of paintings by two impressionist artists. Three days later, they were asked to categorize (by artist) a new set of paintings along with the original acquisition set. It was found that the categorization responses of the older adults for acquisition set stimuli were influenced by category centrality. with low central paintings being categorized correctly less often than more central exemplars of a n artist's style. No such effect was observed in the young. This result suggests that the performance of older adults is governed more by conceptually driven retrieval processes, even when the availability of individual stimulus items in memory is controlled across age groups (i.e.. both young and old subjects had learned acquisition set items to a similar level). Interestingly, the recent work on implicit memory seems consistent with this result in showing that explicit tests of episodic memory may result in age differences due to the retrieval characteristics of the task (see Howard, 1988b). Conclusions
This selective review of relevant aging and memory research has suggested the following. First, the types of effects assumed to result from schema-based processing (e.g., support, integration) are observed in adults of all ages. Second. knowledge supports appear to become more important in determining memory performance with age in adulthood. Whereas some of this increased importance in knowledge supports may be due to age-related breakdowns in basic memory skills,there is some evidence that it may also relate to preferential modes of processing (e.g., Hess & Wallsten, 1987; Ratner et al., 1988). Finally, aging-related changes in basic skills may negatively affect schema-based processes, such as integration, when heavy demands are put on storage functions (e.g.. Zacks & Hasher, 1988).
118
Hess
Aging and Schematic Processing
In recent years, much of the work in my lab has concentrated on examining age differences in memory processing associated with
different types of schematic structures. This research is discussed below along with the results of other relevant studies conducted elsewhere that deal with the Impact of three specific types of schemas: scripts, trait schemas, and scene schemas.
Scripts One of the more widely examined types of schematfc structure is the script (e.g., Graesser 81Nakamura, 1982; Nelson, 1986: Schank & Abelson. 1977). These structures are assumed to represent generic information relating to common activities, such as grocery shopping, visiting the doctor. cleaning house, and so on. Scripts contain information about the types of actions that normally occur during the specific activity, the relations between the actions (e.g., temporal order). and the types of objects and people that can and normally do fill roles within the activity. Research has shown that scripts influence both attentional and retention processes during memory processing. Specifically, information that is relevant to the script is processed more quickly than irrelevant lnformation (e.g., Bellezza. 1981; Bellezza & Bower, 1982). presumably because relevant information is more likely to be predicted or expected when the script is activated in memory. In addition, the infomation that is remembered about a specific scripted event is affected by script structural factors, such as typicality. For example, recognition memory is better for atypical information than for typical information, at least initially (e.g., Graesser et al., 1980). This is thought to be due to the fact that specific episodic information is integrated with the generic script, resulting in unstated typical script actions being copied into the memory trace [Graesser & Nakamura, 1982). This results in relatively poor discrimination between stated and unstated typical actions in memory, especially those of very high typicality. Atypical actions are more easily discriminated because they are not integrated with the memory event directly, but instead are hypothesized to be
ScFwmatic Influences on Memory
119
linked to the memory representation indirectly through "tags". This manner of representation, however, also results in a greater rate of forgetting for these actions because of an increased emphasis on conceptually driven retrieval with the passage of time (see Graesser & Nakamura, 1982). Because of their less direct connection to the script-based memory trace, atypical actions are less likely to be accessed than typical actions when the script is used to guide retrieval. Several studies have been conducted that examine adult age differences in script processing. In this research, subjects in different age groups are presented with descriptions of a n individual engaging in a scripted activity, with the actions included in each activity varying in their relationship to the script in terms of typicality or relevance. For example, a typical action in a grocery shopping sequence might be paying the check-out person, whereas an atypical one might be picking a penny up off the floor. Memory for specific actions is then tested with either a cued recall (using the script title) or recognition test. In all of these studies, selection of stimulus actions was controlled by obtaining typicality and/or relevance ratings from both young and older adults, and using only those items on which there was agreement in the ratings across age groups. In an initial study, Light and Anderson (1983) examined the effects of typicality on both the recall and recognition of script actions in young and older adults. Although they found that young adults had better overall memory for the scripted events, there were no age differences observed in the effects of typicality on performance, suggesting that scripts have similar influences on retention across age groups. Several subsequent studies conducted in our lab, however, have obtained results that are somewhat inconsistent with this view. In all of our studies, we have found that the general effects of script structure on memory performance are similar across age groups (e.g., high typicality is associated with poor recognition performance) thereby supporting one aspect of Light and Anderson's (1983)results. We have also found, however, that age differences in memory performance are not consistent across levels of script structure, with factors such as typicality and relevance
120
Hess
appearing to be more influential in determining the performance of older adults. In our first study (Hess, 1985, Experiment 1). we examined recognition memory for script actions a s a function of their typicality. It was found that memory for atypical actions was significantly better than that for typical actions in both young and older adults (see Table 1). In addition. the younger adults exhibited a n overall superiority in memory performance, but the age difference was greater for atypical actions than for typical actions. This effect did not appear to be due to the older subjects exhibiting a greater reliance on the generic script in making recognition responses. If this were true, they should have produced a disproportionate number of typical false recognitions, since such a pattern of performance would have reflected a readout from the generic script. Instead, the pattern of false alarms was similar across age groups, with the age X typicality interaction in recognition performance being primarily due to a reduction in correct recognitions of atypical actions in the older adults. It was argued that this dect could be accounted for by the fact that older adults were less likely to encode distinctive contextual information about the event. It was assumed that information in memory concerning the target event could be accessed through script-based associations (due to activation of the script during presentation) or associations relating to the specific context in which the event occurred. Contextual associations are necessary for distinguishing between dlfferent instantiations of the same script, and for retrieving atypical information, which is less likely than typical information to be directly connected to the scriptbased memory representation because of its low relevance (e.g.. Graesser & Nakamura, 1982). Thus, a reduction in the availability of contextual information on the part of older adults can be used to account for both their overall poorer memory, and for their specific problems with atypical information. Memory for typical actions is less afTected because they can still be accessed through script-based links.
Some support for this explanation can be obtained by examining the change over time in memory for script actions, and comparing the performance of the young adults after three days
Schemattc Influences on Memory
121
Table 1 Memory for Actions as a Function of Script Qpicality and Relevance Time of test and Script relation ~~
~
~
~
3 &Ys
20 minutes
Group
Typ. Atyp. Atyp. rel. rel. irrel.
Qp. Atyp. Atyp.
rel.
rel.
irrel.
1.10
--
Hess (1985)
Exp. 1: recognition
Young Old Hess et al. (1989) Exp. 1:recd Young Old Exp. 2: recognition Young Middle-aged Old
-.a --
1.08
2.39
.48
1.30
.25
---
---
__ __
.49
.61
.28
.a
.50
.12
1.31
2.87
3.09
1.25
3.06 3.05
.60
1.01
2.40
2.26
.30
.67
__
.67 1.55 1.68 1.33 1.43 .85
.76
Note: Recognition memory is indicated by d'. Recall is indicated by proportion Gcalled.
with the older adults after 20 minutes. It is assumed that retrieval becomes more conceptually driven (i.e., dependent upon scriptbased associations) with the passage of time (see Graesser & Nakamura. 1982). This suggests that atypical actions should be forgotten more quickly than typical actions. and indeed this was observed for both age groups. In addition, the similarity between the young adults at three days and the older adults at 20 minutes
122
Hess
suggests that they may simply represent two different points on the same retention curve indicating performance a s a function of the availability of script-based and contextual associations. A second study (Hess, 1985, Experiment 2) provided support for our interpretation of the results by showing that the performance of young adults mimicked that of older adults when contextual cues were made less distinctive. This was accomplished by presenting the subjects with three passages, each containing different versions of the same scripted activities. A second set of experiments (Hess, Donley,& Vandermaas, 1989) was conducted in order to replicate and extend these findings. We were specifically concerned with examining the possibility that age differences in memory for atypical information might be reduced by increasing its relevance to the script. In our previous study, the atypical actions (e.g., Jack wound his watch) included in the event descriptions were also largely irrelevant with respect to the scripted activity (e.g.. Eating out at a restaurant). thereby making integration with the script difficult. We reasoned that if the relevance of these actions were increased, integration would be facilitated along with the ease of retrieval of these actions through script-based connections. In addition, if older adults are more reliant upon such connections in memory due to poorer encoding of contextual information, increasing the association with the script should be especially beneficial to these subjects. To test these hypotheses, we ran two separate experiments in which subjects once again read a series of descriptions of scripted activities. In addition to typical and irrelevant atypical actions, however, these descriptions also contained relevant atypical actions in the form of script interruptions, which represented actions that interfered with the script goal path. For example, in the restaurant script, one interruption was "Jack forgot his wallet." In the first experiment, it was found that age differences in recall were directly related to the degree to which the test information was associated with the script. As both relevance and typicality increased, age Merences in memory decreased (see Table 1). In addition, relevance was observed to account for greater recall variance in the old than in the young. As hypothesized, the facilitating effects of relevance on recall of atypical actions
Schematic XnJuences on Memory
123
appeared to be partly due to the increased probability of integration with the script-based memory trace. This was inferred from the fact that recall of specific typical actions by subjects of both ages had a facilitative effect on recall of relevant atypical actions that immediately followed them in the target passage. No such effect was obsewed for irrelevant actions. This suggests that the relevance of the interruptions to the memory script led to more direct trace integration than found with irrelevant actions. One somewhat unexpected flnding from this study was that the age differences in reading times for individual actions did not match the pattern of performance observed with the recall task. Both young and older adults read relevant actions faster than irrelevant ones, but the degree of facilitation observed was similar across groups. This result suggested two alternative explanations for the age differences in recall performance. It may be that age differences in the encoding of information into memory were minimal, and that the differential impact of relevance across groups was primarily at retrieval. The older adults may use relevance more than young adults to guide retrieval, which would result in a disproportionate reduction in the accessibility to irrelevant information in memory. The fact that no age differences were obtained in the degree to which recall of typical actions facilitated recall of atypical actions, however, would appear to be an argument against such an explanation. Alternatively, it may be that older adults were simply less efficient in general in processing test information. Inefilcient encoding would have the least impact on relevant information, whose relation to the script is easily established, and the greatest effect on irrelevant information, which is more dlfflcult to integrate. To further examine some of these ideas, a second experiment (Hess et al., 1989, Experiment 2) was conducted using the same materials as the first, but recognition memory rather than recall was tested. If the differential impact of relevance across age groups is due primarily to the older adults relying more on this scriptbased factor to guide retrieval, then using a memory test that places less emphasis on conceptually driven retrieval (i.e., recognition) should eliminate the accessibility advantage for relevant atypical actions. This should result in similar levels of retention for all
124
Hess
atypical actions, regardless of relevance. In addition, the age differences observed for relevant and irrelevant atypical actions should be sfrnflar. In contrast, if the previously observed age effects are due primarily to encoding factors, which would have an impact on the availability of information in memory, altering the retrieval requirements of the memory task should not affect the nature of the age differences observed. The results of the study (see Table 1) support this latter interpretation. Once again, age differences in memory performance were observed to decrease as the typicality and relevance of the test actions increased. As in the first study, this effect was obtained in spite of the fact that there were no age Merences in the effects of script structure on reading times. Taken together, these studies suggest that age dserences do exist in the degree to which scripts affect memory performance. Further, it appears that these differences are primarily due to the efficiency with which unpredicted information is integrated with the memory representation. There does not seem to be much evidence for age differences in script-based reconstructive retrieval processes. In addition to the fact that we obtained similar age differences across tasks possessing different retrieval requirements, there is the consistent finding (Hess, 1985; Hess at al., 1989; Light & Anderson, 1983)of no age differences in the pattern of recall intrusions or false alarms. If older adults are dependent upon script-based reconstructive processes at retrieval, we would expect to see them produce a disproportionate number of typical intrusions and false alarms. The lack of any age differences in script-based retrieval processes appears to conflict with the work of Reder et al. (1986). To further explore this issue, we conducted two studies in which we examined age differences in schema-based retrieval. To do this, we utilized a procedure similar to that of Anderson and Pichert (1978) in which subjects read a story about two boys playing hooky at one of their homes from one of two perspectives, that of a burglar or a homebuyer. Subjects recalled the story guided by the same perspective used at reading, and then recalled the story a second time from the alternative perspective. Anderson and Pichert (1978) found that recall of story information was a function of its
Schematic InJuences on Memory
125
relevance to the perspective taken at retrieval, with relevant information being recalled better than irrelevant information. In addition. the shift in perspective at the second test resulted in recall of newly relevant information that had not been produced at the first test, and a decline in recall of previously recalled but now irrelevant information. For example, the piece of information that the side door to the house was always unlocked would have a low probability of recall if the study perspective had been that of the homebuyer. Changing the perspective to that of the burglar, however, would enhance the probability of recalling this information, even if it had not been recalled initially. These results suggest that schemas have a powerful impact on retrieval, and that at least part of the selectivity obsewed in memory is due to schema-based retrieval processes. We used this same basic procedure to examine aging effects on such retrieval processes [Hess & Flannagan, 1990). In the first experiment. 24 young adults (M= 2 1.5 years) and 24 older adults (M = 69.7 years) were tested. and no age differences were found in the degree to which the recall perspective influenced performance (see Table 2, top). Both age groups exhibited similar levels of facilitation at the second test, as indicated by the proportion of newly relevant items recalled for the first time following the change in retrieval perspective. In a second experiment, we attempted to replicate this finding and also included a recognition test as a better measure of the amount of information available in memory. Recent work (Kardash, Royer, & Greene, 1988) has suggested that the effects of schema manipulation on the memory performance of young adults when this procedure is used are primarily based at retrieval. The perspective taken at study does not appear to affect the availability of irrelevant information in memory, but rather influences the accessibility of such information. We wanted to see if this was also true for older adults. Given the results of the previously discussed studies of script memory. we expected that relevance to the study perspective [or schema) would have an impact on availability in older adults. Thirty young adults (M = 20.5 years) and 29 older adults [M = 70.4 years) participated in this experiment.
Hess
126
Table 2 Memory for Znformation vary in^ in Relevance to the Study Schema Relevance to study schema High
Group
Low
Proportion of items reaUed forfist time at second test Exp. 1:
Young
.03
a12
Old
.03
.10
Young
.08
* 12
Old
.04
.10
Exp. 2:
Proportion of items correctly recognized Exp. 2:
Young
.89
.91
Old
.85
.80
The results supported this hypothesis (see Table 2, bottom). Although high levels of both relevant and irrelevant information were available in memory in both age groups, the older adults were significantly poorer than the young adults in correctly recognizing irrelevant information. There were no age differences in memory for relevant information. When recall was examined. and performance was adjusted for the availability of information in memory (i.e.. recall was examined only for those items correctly recognized), once again no age differences were observed in the
Schematic Zn@iences
on Memory
127
degree to which the retrieval schema facilitated production of previously unrecalled information (see Table 2, middle), Thus, these results are consistent with our previous conclusion in showing that age differences in the impact of relevance on memory performance occur primarily at encoding.1 The results of our studies of script memory are obviously different from those of Light and Anderson (1983).At present, the reasons for the discrepancies in results are unclear. W e originally attributed the differences to variations in measurement procedures and to Light and Anderson (1983)using somewhat higher ability subjects. Zelinslci (1988)has suggested that the seemingly greater impact of script structure on the performance of the older adults in our research may be due to the fact that we have used more items on our recognttion tests, thereby increasing interference among items and forcing older adults to rely more on the generic script at retrieval. Several pieces of evidence argue against such a n First, older adults did not produce a interpretation. disproportionate number of typical false alarms, as would be expected if the generic script was guiding retrieval. Second, if increased interference were associated with the number of items on the recognition test, we would expect the older adults performance to become significantly worse over tests (i.e., performance for the first script tested should be better than that for the last script tested), No such effects have emerged in our studies. Finally, the differential impact of script structure on memory performance was observed not only in recognition, but free recall as well. where the type of influence discussed by Zelinski (1988)would be negligible. In sum, with some minor inconsistencies, the evidence suggests that aging is associated with an increased dependence upon memory scripts to support retention of scripted events. Both young and older adults appear to have relatively little difficulty forming an integrated memory representation for event information when it is relevant to the script activated in memory. However. as the relevance of specific actions becomes more tenuous, integration It may be the case that age dflerences do exist in schema-based retrieval, but that
the speciac conditions of obsenmtion in these studies may have masked these effects
by placing a strong emphasis on conceptually driven retrieval processes.
128
Hess
becomes more difkult. especially for older adults. This pattern of results appears consistent with the previously mentioned work on inferences, and suggests that the availability of a meaningful interpretive structure reduces the degree to which age differences are observed in event memory. The results of a recent study by Zelinski and Miura (1988)also tend to support this by showing that young and older adults do not dif€er in the degree to which they are able to integrate thematic information with descriptions of scripted activities when this information is relevant to comprehension.
Person memory Another category of schematic structure that has been widely studied relates to knowledge about traits and stereotypes. It is assumed that people represent trait information in the form of expectations about characteristics and behaviors associated with the trait, and that these trait schemas affect memory in much the same manner as do scripts. Specifically, both attention to and retention of specific aspects of social interactions are dependent upon the relationship of the information to the activated trait schema (e.g., Hastie, 1980). For the most part, the effects of trait schemas on memory processing have been studied by examining retention of behavioral infomation about individuals. Much of this research has shown that information that is highly informative about a n individual is processed more extensively and remembered better than less informative information (e.g., Hastie, 1984; Srull, 1981). For example, behaviors that are inconsistent with perceived traits tend to be remembered better than trait consistent behaviors. This appears to be due to the fact that subjects engage in more extensive processing of inconsistent information in order to establish some rationale for the behavior in the context of schema-based expectations. For instance, if we see a normally friendly person angrily yelling at someone, we might rationalize this behavior by assuming that the person were not feeling well that day, or that the other individual had just insulted him. The generation of such explanations has been shown to be associated with the memory
Schematic Influences on Memory
129
facilitation observed for inconsistent acts. presumably due to the increased interconnections formed between the act and other information in the memory representation. In our research, we were interested in finding out whether young and older adults differed in the degree to which trait schemas affected their perceptions of and memory for information about people. In a first set of studies (Hess, Vandermaas. Donley. & Snyder. 1987). we examined the effects of trait consistency on memory for specific behaviors, and the degree to which schemabased interference occurred in memory processing when traits conflicted with the usual social norms. Young and older adults were presented with a description of a male actor that emphasized either a traditional or nontraditional sex-role. For example, in the traditional condition, the actor had the occupation of banking executive, whereas in the nontraditional condition he was a n elementary school teacher. Subjects were then presented with a list of behaviors that varied in their consistency with respect to the sexrole emphasized, and were told to form an impression of the actor. In each list, the ratio of consistent to inconsistent behaviors was 2: 1. thereby biasing an impression that would be congruent with the initial description of the actor. Impressions were measured by having subjects rate the degree to which 20 masculine and 20 feminine traits (from Bem, 1974) described the actor. An unexpected free recall test for the list of behaviors was then given. This methodology allowed us to examine several aspects of First, the impression schema-based memory processing. formation data would reveal the extent to which the behavioral information had been integrated to form a coherent representation of the actor, If subjects were forming an accurate representation, then the attribution of masculine and feminine traits should vary with the sex-role emphasized. Given our previous work on prototype acquisition (e.g., Hess & Slaughter, 1986b).it was expected that few age differences would exist in the ability to abstract general trait information. Second. the extent to which this integrated representation would affect memory for individual behaviors could be measured by examining the strength of the consistency effect on recall. The work on script memory discussed earlier makes the prediction of age differences somewhat difficult here. Specifically,
130
Hess
given that inconsistent behaviors are highly relevant to the activated trait schema, it might be expected that age differences in the consistency effect would be minimal. Conversely, given the fact that the consistency effect appears to require the production of elaborations for integration (e.g.. Hastie, 19841, it might be predicted that older adults would exhibit less of a recall advantage for inconsistent behaviors due to the presumed effort involved in producing elaborations (e.g., Rabinowitz et al., 1982). Finally, interference from a dominant schema (i.e.* the traditional sex-role) could be observed by measuring the degree to which the consistency effect on recall is altered when schema relevant information is processed in relation to a nondominant schema (isen,the nontraditional sex-role). Specifically, the schema consistent behaviors in the nontraditional condition are inconsistent with respect to the traditional condition, and vice versa. Thus, although a feminine stereotyped behavior would be consistent with a nontraditional sex-role, it might still be processed in relation to a traditional sex-role given the possibility that dominant schemas are easily. and perhaps automatically. activated during processing. The resulting competition between schemas could result in a reduction or reversal of the expected consistency effect. With respect to age differences, there is some evidence that such interference effects would be more likely to occur in older adults (e.g.. Hess, 1982. Experiment 3; Lair, Moon. & Kausler, 1969). The results of this research suggested that under incidental learning conditions, schematic influences did not vary appreciably across age groups. Trait ratings indicated that subjects in both age groups altered their impressions of the actor based upon the sexrole emphasized, with attributions of masculine traits being greater in the traditional condition and feminine trait attributions being greater in the nontraditional condition (see Figure 1). Interestingly, there was also the suggestion that the older adults were less likely than the younger adults to consider sex-role inconsistent information in forming their impressions, a s indicated by the fact that the disparity between their masculine and feminine trait ratings was greater than that observed for the young adults. This result is consistent with our previously mentioned work (Hess &
Schematic In$uences on Memory
131
110
vouNGKKILT8
100
P
ii
g
90
3I FEMINHETRATTS
MASCULINETRMTS
80
70
I
I
I
I
TRAD.
"TRAD.
I TAAD.
1
.)"
SEX ROLE
Flgure 1. Trait attributions as a function of the actor's specifled sex-role. Trait ratings could range from 20 to 140, with higher scores indicating greater trait attributions (from Hess et al., 1987, Ekperiment 2).
Wallsten, 1987)in showing that older adults are less likely than young adults to access specific item infonnation in memory. and more likely to use general information in making judgements. With respect to recall. both young and older adults recalled inconsistent behaviors better than consistent ones in the traditional condition, whereas the trend was reversed in the
Hess
132
TAAMlloNAL SEX-ROLE
”TRAMTK)NAL
SEX-ROLE
mure 2. Proportion of sex-role consistent and inconsistent behaviors recalled as a funcfton of the actor’s specitled sex-role (from Hess et al.. 1987, Experiment 2).
nontraditional condition (see Figure 2). This latter result suggests that both age groups experienced some interference between schernas. In addition, the effects of consistency on recall performance did not vary across age groups, indicating that trait schemas operate in a similar fashion in young and old adults during comprehension. We have recently completed another trait schema study (Hess & Tate. 1990) that examined the types of encoding operations used during the processing of information about individuals.
Schematic Znfluences on M e m o r y
133
Specifically, we were interested in flnding out if young and older adults varied in their use of explanation-based processing (see Hastie, 1984) to support memory. especially when dealing with trait-inconsistent behaviors. To this end, we presented 32 young adults [M= 20.2 years) and 34 older adults (M = 68.4 years) with four lists of behaviors performed by four different actors. Each list contained six behaviors that were consistent with the same personality trait (e.g., honesty) and three that were inconsistent. The behaviors were presented as short phrases followed by a blank line (e.g., Thomas watched professional wrestling on television .) For each list, subjects were instructed to read all the behaviors in order to form an impression of the actor, and then to go back and write a meaningful continuation for each behavior phrase that would form a complete sentence. A recall test of the specific behaviors associated with each actor was then presented after the fourth list. Of primary interest in this study was the degree to which age differences existed in memory for trait-consistent and inconsistent behaviors, and the extent to which these differences could be attributed to the types of continuations produced for each behavior. Following the procedure of Hastie (1984). we classified the continuations into three categories based on their relationship to the behavior: (a) explanations attempted to provide a rationale for the behavior; (b) elaborations expanded upon the circumstances relating to performance of the behavior, and (c) successions listed events that occurred following the behavior. It was found that both age groups recalled inconsistent behaviors better than consistent ones, but the consistency effect was only significant for the young adults (see Table 3,top). When the continuation data are examined (see Table 4). some of the reasons for these age difierences become evident. As observed by Hastie (1984). subjects in both age groups tended to produce more explanations for inconsistent than for consistent behaviors. In addition, adults in both age groups produced more explanations than the other two types of continuations. Relative to each other. however, the young adults produced significantly more explanations, whereas the older adults produced significantly more elaborations. Given that explanations have been found to facilitate
134
Hess
Table 3 Proportion of Behaviors Recalled as a Function of Their Traft Consistency and Explanation Q p e
GmP
Explanation 5Pe
Consistent
Inconsistent
AU behaviors young
.M
.73
Old
.44
.a
Explained behaviors Young Old
Trait-consistent Trait -neutral Trait-consistent Trait-neutral
.58
.80
.61
.61
.50
.54
.4 1
.4!5
retention. part of the age difference in memory performance appears attributable to the older adults being less likely to engage in the type of schema-based integrative processing necessary to support memory. In a subsequent analysis, we examined the probability of recall given that an explanation had been provided for the behavior. It was found that explanations enhanced the recall performance of the young adults to a greater extent than the older adults, especially when dealing with inconsistent behaviors. We hypothesized that this age difference in the degree of explanation-based memory facilitation may have to do with the nature of the explanations produced in each age group. The primary facilitative effect of explanations on memory appears to occur due to the establishment of a connection with the dominant trait-based memory representation. It may be that young and older adults were producing qualitatively different explanations, with the young
Schematic Injluences on Memory
135
Table 4
Proportion of Continuation Types Produced for Trait Consistent and Inconsfstent Behauiors ~
~~
Continuation type
Group
Behavior
Young
Consistent Inconsistent
Old
Consistent Inconsistent
Explanation
Elaboration
Succession
.79
.19
.02
.89
.10
.02
.65
.33
.02
.69
.29
.03
adults being more likely to produce integrative explanations. For example, suppose that an honest person were described a s taking some money that did not belong to him. An integrative explanation might Justify the behavior by assuming that the person was going to return it to the individual it was originally taken from, thereby establishing a degree of consistency with the actor's perceived trait. A nonintegrative response might attribute the behavior to the actor's desire to obtain more money so that he could buy a new stereo. Although explaining the behavior. this latter response does not associate it to the dominant trait, thereby rendering it more difficult to access in memory when this trait schema is used to quide retrieval. To examine the possibility of age differences in types of explanations, we categorized each one in terms of whether it was consistent, inconsistent, or neutral with respect to the actor's attributed trait. It was found that the proportion of integrative explanations (i.e.. trait-consistent) produced by the young adults (.42) was significantly greater than that produced by the older adults (.24). When recall was examined, however, it was found that the production of integrative explanations (relative to trait-neutral explanations) facilitated recall of inconsistent behaviors in both
136
Hess
age groups, though the effect was somewhat greater in the young subjects (see Table 3,bottom). Thus, the results of this analysis suggests that the speciflc problem that older adults have in recall of schema-inconsistent information is related to the fact that they are less likely to carry out effective integrative operations. It should be noted that there is some inconsistency arnong our studies of scripts and trait schemas in the degree to which age interacts with schema structure in determining memory performance. The reasons for this may have to do with variations in study instructions across experiments. In those studies where age X structure interactions did occur (e.g., Hess, 1985; Hess et al.. in press) intentional learning instructions were given and subjects expected the memory test. It may be that younger adults benefit more when memorization instructions are given since they can utilize their more efficient general memory skills in processing the target information (see also Craik 8r Ftabinowitz. 1985). Interestingly, Hess and Tate (1990) found that the memory performance of younger adults was somewhat enhanced when they were instructed to memorize the target behaviors in addition to forming an impression of the actor. In contrast, memorization instructions actually suppressed the recall performance of older adults. This suggests that the extent to which age differences are observed in remembering situations may be related to the degree to which the task emphasizes schema-based comprehension versus memorization. When the former process is emphasized, and memorization is not explicitly called for (e.g., Hess at al.. 1987, Experiment 2). the influence of age-related general memory processes on performance may be reduced, resulting in less between group variation in performance patterns. I
The final area in which we have been conducting schema research is in the area of scene memory. Although there has been a great deal of research on memory for pictorial and scene-like materials in the aging field, little of this work has concentrated on what might be termed schematic influences on scene memory. A notable exception is the work of Waddell and Rogoff (1981. 1987).
Schematic InJuences on Memory
137
who have examined adult age differences in the effects of scene organization on memory. In general, their research has suggested that age differences in memory performance are minimal when the structure imposed on the materials to-be-remembered is meaningful. In their first study (Waddell & Rogoff, 1981). they presented middle-aged and older women with spatial arrays containing 30 common objects (e.g., animals, cars, people). In one presentation condition, the objects were randomly placed in a bank of cubicles, whereas in the other they were placed on a panorama that provided a meaningful contextual organization (i.e.. a small town). After studying each array, subjects were asked to select the objects that were in the study array from a larger pool. and to place these objects in the positions in which they originally appeared. Using the proportion of objects both correctly recognized and placed, it was found that there were no age differences in memory performance in the panorama condition, but that the middle-aged women were superior in the cubicle condition. In a subsequent study (Waddell 8t Rogoff, 1987). the finding of no age differences in the panorama condition was replicated. In addition, it was observed that the performance of the older adults was best when they were able to impose a narrative structure on the materials. This structure may have been particularly beneficial to older adults because it provides a n integrated representation of the stimuli that utilizes a meaningful organizational scheme based upon their experience. The results of this research suggest that the availability of schema-based organizational information is a more important determinant of memory performance in older adults, a conclusion consistent with much of our work with schemas and aging. Our subsequent research on scene memory extends this initial work by examining schematic influences on object and spatial memory separately. The tasks used in the Waddell and Rogoff studies did not discriminate between memory for object versus spatial information. A perusal of the literature on scene processing (e.g., Mandler, 1984). however, indicates that these two types of information are both affected by knowledge structures, and that memory for each is influenced by dmerent factors.
138
Hess
For object memory, the primary influence on performance relates to the likelihood of occurrence of the individual stimulus objects in similar scenes. Research has shown that subjects attend less to expected than to unexpected objects, at least initially (Friedman, 1979;Loftus & Mackworth. 1978).presumably because objects predicted by the activated schema require less analysis for identification purposes (e.g., Palmer, 1975). In addition, unexpected objects also are remembered better than expected objects (e.g., Friedman, 1979).an effect that is related to at least two different factors. First, the memory representations of expected objects appear to be less elaborate, as indicated by poor memory for detail information (e.g., Mandler & Parker, 1976;Mandler & Ritchey, 1977). Second, subjects also have a more difficult time discriminating between expected objects that were presented and those that were not (e.g., Friedman. 1979).This may be due to the integration of normative schematic information with the memory trace or to the operation of schema-based reconstructive processes at retrieval. Interestingly, spatial organization has been found to have little impact on object memory, at least at immediate test (e.g., Mandler & Johnson, 1976). In contrast, researchers have studied knowledge effects on location memory primarily by manipulating the types of spatial organization found in the stimulus scenes. Mandler and her colleagues (Mandler & Johnson, 1976: Mandler & Parker, 1976; Mandler & Ritchey, 1977)have found that violations of normative spatial relations (e.g., having the sun below a house) result in poorer spatial memory than when the relations between stimulus objects conform to expectations. In addition, it was also observed that errors in reconstruction of unorganized scenes tended to approximate normative relations (e.g., the aforementioned s u n would be incorrectly placed above the house). This suggests that schemas provide support for episodic spatial infomation. and that reconstructive processes are operative when accurate location information is unavailable. We have conducted two studies of age differences in scene memory that specifically examine the impact of schematic knowledge on memory for object and spatial information. In the first study (Hess & Slaughter, in press, Experiment 1). we presented
Schematfc Injluences on Memory
139
young and older adults with a series of eight pictures depicting different types of scenes (e.g.. a picnic, a kitchen). Each scene contained ten objects, five of which had been rated as highly likely to appear in the scene, and five of which had been rated a s less likely. In addition, half of the pictures were presented in a n organized fashion, presenring normative interobject relations, and half in a n unorganized fashion. The stimuli in this latter condition were constructed by inverting the organized scenes and then rotating each individual object 180 degrees. This procedure effectively eliminated normative spatial relations while preserving an identical degree of interobject density a s found in the organized scenes. Separate memory tests were given for object and spatial location information for each scene. Consistent with the previous discussion of schematic &&As on scene memory, it was expected that scene organization would have a greater impact on memory for spatial information than for objects, whereas object memory should be afTected primarily by likelihood of occurrence. In addition, consistent with the hypothesis that schematic factors increase in importance in determining memory performance with advanced age, it was expected that the impact of these factors would be greater in the older adults. Object memory was tested using a recognition test containing the ten objects for each scene plus an identical number of distractors that had been matched for size and likelihood of occurrence with the targets. Consistent with previous work (e.g., Friedman, 1979).it was found that discrimination memory was better for low likelihood than for high likelihood objects (see Table 5. top). In addition, we also found that this likelihood effect was greater in the older adults than in the young, primarily due to the disproportionately greater number of high likelihood distractor objects incorrectly recognized by the older subjects. No effects of organization on object memory were obtained. Memory for spatial location was tested using a procedure developed by Mandler (e.g., Mandler & Parker, 1976) in which subjects were given all the target objects and asked to reconstruct the original study scene. Of primary interest was the extent to which interobject spatial relations were maintained rather than exact duplication of the presentation location. Mandler (1984)has
140
Hess
Table 5 Recognition Memory and Scene Reconstruction Performance Scene type and Object likelihood Organized Group
High
LOW
Unorganized High
LOW
0bject recognftion Exp. 1:
Young Old
2.61
2.87
2.70
2.61
1.49
1.88
1.81
1.98
Young
2.57
2.87
2.34
2.46
Old
1.80
2.23
1.73
2.30
Exp. 2:
Scene reconstruction Exp. 1:
Young Old
.85
.64
.42
.27
.65
.36
.05
.24
Young Old
.65
.66
.23
.37
.48
.44
-.01
.24
Exp. 2:
Note: Recognition performance is indicated by d'. Reconstruction performance is indicated by Kendall's tau, a rank-order correlation.
argued that scene schemas are organized in topological terms emphasizing relational information, and therefore the primary impact of schematic knowledge should be on such information. To test memory for relational information, rank-order correlations were obtained between the orderings of the objects at presentation
Schematic Injluences on Memory
141
and reconstruction in both the vertical and horizontal dimensions. Only the results for the vertical dimension will be discussed since it has been shown to be more influenced by schema-based factors (e.g., Mandler & Parker, 1976). As expected, spatial memory was better for the organized than the unorganized scenes in both age groups (Table 5, bottom). However, there were variations in the degree to which organization facilitated memory for specific objects. In the young adults, spatial memory was enhanced at both levels of likelihood when a meaningful organization was provided. In contrast, only high likelihood items were affected in the older adults. This finding is consistent with expectations. Specifically, older adults were less able than young adults to remember spatial information when the stimulus information did not readily fit in with their experience. Their performance was poor for high likelihood objects in the unorganized condition and for low likelihood in both conditions. In each of these cases, older adults could not draw upon normative relational information to support memory. Young adults, on the other hand, appeared to be more flexible in their ability to organize information, a s indicated by the fact that memory for low likelihood objects was enhanced by organization in spite of the fact that relations between these objects and others in the scene were not necessarily predicted by the scene schema. The fact that organization had a greater impact in determining age differences in spatial memory than in object memory suggests that the organizational manipulation in Waddell and Rogoff (1981) may have had its primary effect on memory for relational information. In fact, this does appear to be the case when separate indices of spatial memory (objects correctly placed plus substitutions of similar objects) and object memory (objects correctly placed plus location errors) are obtained from Table 1 in that paper. The between age group differences in total items correct are 20.4 for spatial memory and 11.5 for object memory. In our second study (Hess & Slaughter, in press, Experiment 2). we wanted to replicate the memory results of the first experiment, and therefore essentially duplicated the basic design. We were also interested in examining age diflerences in schematic influences on attention during study to see if the types of age effects found in
142
Hess
memory performance could be related to variations in attention. For example, were older adults paying disproportionately less attention to high likelihood objects, thereby accounting for the fact that age Merences were greatest in memory for these items? The design employed also allowed u s to see if the lack of organization affected the older adults ability to detect the overall relationship between objects in the scene (i.e., identlfy the schema). To examine attention. we measured the amount of time that subjects fixated on individual objects during study. Friedman (1979) had found that low likelihood objects were fixated more initially than high likelihood objects, indicating the influence of the schema in guiding perception. If the age differences in object memory are due to variations in attention during study, the likelihood effect for fixation durations should be greater in the older adults. In addition, if the lack of a meaningful spatial organization hinders schema activation, the typically observed interaction between likelihood and fixation number should be disrupted in these same individuals. The patterns of age-related performance on the memory tests for this study essentially replicate those from Experiment 1, and therefore will not be discussed (see Table 5). With respect to the attention data, it was found that the pattern of object fixations did not vary across age groups in the organized condition (see Figure 3). The initial fixations in both age groups were longer for low Ijlce~lihoodobjects, and the slze of the likelihood effect across age groups did not differ reliably. In addition, there was no effect of organization on the performance of the young adults, indicating that they were able to identify the interstimulus structure even in the face of poor contextual infomation. The older adults, however. exhibited some difficulty in schema integration, as indicated by significantly greater overall fixation durations for unorganized scenes. and by the disruption of the usually observed relationship between likelihood and fixation when studying these scenes. The interesting aspect of this result, however, is that this dflerence in the impact of organization across age groups did not translate directly into age differences in either object or spatial memory. Apparently, the eventual schema integration achieved by the older
Schemattc Injuences on Memory
143
F Y g m 3. Mean fixation dumtlona (inmsec.) for individual objects in scenes.
adults was sdicient to provide memory support for the objects in the unorganized scenes. The results of these two studies suggest that the relation between the to-be-remembered stimuli and the individual's knowledge structure is more important in determining the performance of older adults than of young adults. The lack of organizational cues was associated with difficulty in schema activation in the older adults. In addition, the types of schematic effects on memory performance normally found in young adults were enhanced in the older adults. Finally, the older adults' memory for spatial information was more adversely affected than that of the young adults when the depicted relations between objects were not based in usual experience.
144
Hess
One somewhat p w l i n g aspect of these results was the fact that age dmerences in object memory were smallest for items that were least related to the schema that was presumably activated during processing. This appears to be in direct contradiction both to the results of the spatial memory task and to our research on script memory (Hess et al.. 1985). where the opposite pattern was obtained. At present, the reasons for the discrepancies in findings are unclear, but may have to do with the nature of the stimuli and the memory test used. The scenes used in our studies did depict groups of schematically related objects, but the objects were not interrelated in the same way that script actions are. In the latter case, there are temporal, causal, and enabling conditions which connect actions both to each other and to the goal of the scripted activity. Beyond general relational information, however. no such connections were present between objects in our scenes. The absence of such connections, plus the use of a memory test that deemphasized relational cues, may have resulted in the stimulus objects being treated in terms of categorical (e.g., living room contents) rather than schematic relations, thereby affecting the manner in which the objects were processed. In support of this explanation. Mandler (1984, p. 9 1) has reported that event-related organizational information is more important in determining memory in stimulus scenes depicting people interacting than in those depicting passive people or inanimate objects. Our scenes were non-interactive in nature, perhaps explaining the daerences in results between the scene and script studies. CONCLUSIONS
The goal of this chapter was to examine the evidence for age differences in knowledge effects on memory in adulthood using a schema-based framework a s an organizational structure for interpreting the relevant evidence, It is apparent from this review of both schema-based and nonschema-based memory research that there are many similarities across age groups in the nature of schematic effects on memory performance. Regardless of age, schematic knowledge appears to provide a n organizational
Schematic Influences on Memory
145
framework that supports the comprehension and retention of specific events. Studies in our lab have also suggested that the integrative functions of schemas are similar across age groups. Both young and older adults are able to use relevant stimulus information to form integrated schematic representations of an event (e.g., Hess et al., 1987). The similarities in patterns of recall intrusions and false recognitions (e.g.. Hess et al.. 1989)across age groups also indicate that subjects of all ages incorporate normative, but unpresented information into these representations. It is also clear, however, that there are age differences in the degree to which remembering is affected by schematic structures. In general, the memory performance of older adults appears to be more dependent than that of young adults upon the relationship between the information to be remembered and the available schematic structure. Age differences in retention tend to be smallest for schema-relevant information (with the possible exception of memory for individual scene objects IHess & Slaughter, in press]). This suggests that the supportive functions of schemas in memory are similar across age groups for those aspects of the situation that are consistent with expectations or can be easily predicted. Such information is readily encoded in terms of the activated interpretive structure. In contrast, age differences in memory performance are usually greatest for information that is novel or irrelevant to the schema activated during study. Such information is assumed to require extensive processing on the part of the individual to discover or construct its relationship to the schema, and thereby facilitate its integration within the schematically organized memory trace (e.g., Black, Galambos, & Read, 1984). Interestingly, the observed age effects in memory performance cannot be attributed to simple variations in patterns of schema-influenced attention allocation. Measures of on-line processing be.. reading times and fixation durations) indicate no age differences in the relative amounts of attention allocated to pieces of information across levels of schema-relevance. at least when the stimulus materials are presented within a familiar organizational context (Hess et al., 1987:Hess & Slaughter, in press: Hess et al.. 1989).Both young and
146
Hess
older adults devote more processing time to unexpected than expected infoxmation. This suggests that the observed age differences in memory performance are based either in retrieval or in more qualitative aspects of the encoding process. Our work on schema memory has strongly suggested that at least part of the age effect stems from older adults being less efficient at integrating unexpected information with the memory representation. For example, older adults are less likely to spontaneously engage in the types of encoding operations that facilitate integration of and support memory for novel infomtion Mess & Tate, 1989). In addition, the fact that the pattern of age diEerences in memory for script actions remains the same across tasks emphasizing difTerent levels of schema-based retrieval (e.g., Hess et al., 1989) also provides support for an encoding-based explanation for the observed age-related variations in performance. At present, there is less direct supportive evidence for the possibility that the differential effects of schema structure on memory across age groups are based in retrieval processes. The studies of script memory suggest few age differences in the use of script-based reconstructive retrieval. This does not preclude the possibility of age differences in retrieval strategies, however, with older adults perhaps being more likely to access the memory representation through schema-based associations, thereby resulting in a reduction of schema-irrelevant information retrieved from memory. Note that this type of process would not automatically result in the production of a n inordinate amount of unpresented, typical information at test, a s would be expected if subjects were reconstructing their memory responses using the generic script as a guide. Our research on impression fonnation (Hess at al., 1987) and categorlzation (Hess & Wallsten. 1987) has indicated that older adults are less likely than young adults to access information with little direct association to the schema, even though it is present in memory. Such results, along with those of Reder et al., (1986). suggest that age differences do exist in the use of schema-based retrieval processes. and that more systematic studies are needed to consider the nature of these effects and the conditions under which they occur.
Schematic Influences on Memory
147
Given the existence of these age differences in schematic influences on memory, how might they be best explained? One possible explanation pertains to changes in the constellation of cognitive skills and resources available to the individual. The pattern of results just presented appears to be consistent with expectations derived from descriptions of changes in the mechanics and pragmatics of intelligence (Baltes et al., 1984). Specifically. age differences in performance are small, and older adults can perform relatively well, when existing knowledge can be utilized to support performance. Older adults are at a relatively greater disadvantage, however, when they cannot utilize prior experience. and instead must rely on their less efficient basic cognitive skills in support of memory. Interestingly, our work on trait memory (Hess et al.. 1987; Hess & Tate. 1989) has suggested that age Werences in performance patterns may be minimized when the memory task de-emphasizes the use of such skills and encourages schema-based processing. The apparent disproportionate benefits experienced by older adults when they can utilize existing knowledge may be related to the ease with which schematic structures are activated and operate when the match between the external event and the internal interpretive structure is high. Research has suggested that schemarelevant information is both encoded and retrieved more easily than less relevant information (e.g.. Galambos & Rips, 1982: Sharkey. 1986: Yekovich & Walker, 1986). Given hypothesized limitations on working memory capacity in older adults (e.g., Craik & Rabinowitz, 1982; Zacks & Hasher, 1988). it may be that the availability of a relevant schema during processing might facffltate integration by providing a ready link between schema components. This might overcome potential processing overloads that occur when subjects must generate their own connections, and storage capacity is taxed, as would presumably occur in the older adults when dealing with unexpected information. Likewise, the ease of retrieval associated with schema-related information might also play a part in determining age differences in memory as a function of schema-relevance. If the use of direct retrieval strategies places a high demand on capacity, as suggested by Reder et al. (19861, then older adults might be more likely to use less effortful schema-based retrieval schemes. This would presumably have a negative impact
148
Hess
on the accessibility of information with little apparent relevance to the schema. Although the previously discussed results of work on age differences in inferential abilities (e.g., Light & Albertson, 1988) make the just-described scenario highly plausible. other factors may also help in determining the observed age dmerences in performance. For example, Hess and Tate (1989)had observed that older adults were less likely to perform effective integrative operations at encoding, which might conceivably be attributed to the demand placed by such operations on their reduced working memory capacity. In that same study, however, it was also found that all the older adults did engage in such operations at some level, and that their memory was facilitated to the same extent as that of younger adults when they did. This suggests that the age Werence in production levels for effective integrative operations may have been related to variations in control processes governing the approach taken in performing the task. In a similar vein, our studies of scene memory (Hess & Slaughter, in press) provided both formal and infonnal evidence that age-related variations in schematic memory effects may be related to stylistic factors, and not just to differences between groups in the efficiency of basic memory processes. Formally, we investigated the degree to which schema-based reconstructive processes (e.g., false alarm rates) were evident in memory performance as a function of the amount of information available In memory. It was reasoned that the use of such processes might be more prevalent when little accurate information is available in memory, and subjects supplement their responses using their schematic knowledge. In the young adults, a consistent signiflcant relationship was observed, with reconstruction being more prevalent when veridical information was unavailable. In contrast, this relationship was much weaker in the older adults, suggesting that their use of reconstruction may have reflected a response style rather than a compensatory mechanism. In the same research, and related to the above point, we also informally noted that older adults were more likely t o change a correct, but schema-inconsistent response (e.g., placing a coffee table behind a sofa) to an incorrect, but consistent one (e.g., placing
Schematic InJuences on Memory
149
the coffee table in front of the sofa). This suggests that knowledge might exert a form of influence on the memory performance of older adults that has nothing to do with factors relating to the efficiency of cognitive resources or the nature of the memory representation. Instead, they may be using schematic knowledge to edit their responses, thereby producing information at test that is coherent and consistent with their experience in the world. Results such as these indicate that further research into, what could be termed, styles of remembering might be fruitful in helping us to understand aging-related changes in memory. In conclusion, it is clear that more research needs to be done in order to provide a more coherent picture of adult age differences in knowledge effects on memory. At present, it appears as if there is age-related variation in knowledge influences, and that these effects may be due to both changes in the efficiency of various ability systems as well a s developmental differences in remembering styles. More systematic study of these factors within the context of models of adult development that consider these various components of cognition as well as the social environment should provide us with a greater understanding of the changes in everyday memory functioning associated with aging. ACKNOWLEDGMENTS Preparation of this chapter was supported by grant number RO1 AGO5552 from the National Institute on Aging. I would like to thank Lynne Baker-Ward and Elizabeth Stine for helpful comments on an earlier draft. REFERENCES Alba. J.W., & Hasher, L. (1983). I s memory schematic? Psychological Bulletin. 93, 203-23 1. Anderson, R.C.. & Pichert, J.W. (1978). Recall of previously unrecallable information following a shift in perspective. Journal of Verbal Learning and Verbal Behavior, 1 7 , 1- 12.
150
Hess
Baltes, P.B., Dittman-Kohli, F.. & Dixon, R.A. (1984). New perspectives on the development of intelligence in adulthood: Toward a dual-process conception and a model of selective optimization with compensation. In P.B. Baltes & O.G. Brim, Jr. (Eds.), Lsfe-spandeuebpment and behavior (Vol. 6, 33-76). New York Academic Press. Bartlett, F.C. (1932). Rernernberlng: A study in experlmentul and social psychology. Cambridge, England: Cambridge University Press. Bartlett, J.C., Till, RE., Gernsbacher, M., & Goman. M. (1983).Agerelated differences in memory for lateral orientation of pictures. Journal Of Genmtology,38,439-446. Barrett, T.R., & Watkins, S.K. (1986). Word familiarity and cardiovascular health as determinants of age-related recall differences. Journal of Gemntobgy, 41,222-224. Barrett, T.R. & Wright, M. (1981). Age-related facilitation in recall following semantic processing. Journal of Gerontology, 36, 194199.
Bellezza. F.S. (1983). Recalling script-based text: The role of selective processing and schematic cues. Bulletin of the P~~chonomrc Society,21, 267-270. Bellezza, F.S., & Bower, G.H. (1982). Remembering script-based text. P~etics,11. 1-23. Belmore, S.M.(1981). Age-related changes in processing explicit and implicit language. Journal of Gerontology, 36. 3 16-322. Bem, S.L. (1974). The measurement of psychological androgyny. Journal of Consulting and Clinicaf Psychology. 47, 155-162. Bjorklund, D.F. (1987). How age changes in knowledge base contribute to the development of children's memory: An interpretive review. Deuebpmental Review, 7.86-92. Bjorklund, D.F., & Zeamon. B.R (1982). Children's organization and metamemory awareness in their recall of familiar information. Child Development. 53. 799-810. Black, J.B.. Galambos, J.A.. & Read, S.J. (1984). Comprehending stories and social situations. In R.S. Wyer. Jr., 8t T.K. Srull (Eds.), Handbook of social cognition (Vol. 3. pp. 45-86). Hillsdale, NJ: Lawrence Erlbaum. Bransford, J.D., 8t Franks, J.J. (1971). The abstraction of linguistic ideas. Cognitive PsycbZcyy. 2, 331-350. Bransford, J.D., & Johnson, M.K.(1972). Contextual prerequisites for understanding: Some investigations of comprehension and recall. Journal of Verbal Learning and Verbal Behavior. 1 1 . 717-726.
Schematic Influences on Memory
151
Brewer, W.F., & Nakamura, G.V. (1984). The nature and functions of schemas. In RS. Wyer. Jr., & T.K. Srull (Eds.). Handbook of social cognition (Vol. 1, pp. 119-160). Hfllsdale, NJ: Lawrence Erlbaum. Brewer, W.F., & Treyens, J.C. (1981). Role of schemata in memory for places. CcgnftfvePsychdogy, 13. 207-230. Britton, B.K., & Tesser, A. (1982). Effects of prior knowledge on use of cognitive capacity in three complex cognitive tasks. Journal of Verbal Learning and Verbal Behavior. 21,42 1-436. Brown, A.L. (1975).The development of memory: Knowing. knowing about knowing, and knowing how to know. In H.W. Reese (Ed.), Advances in child development and behavior (Vol. 10, pp. 103152). New York Academic Press. Burke, D.M.. & Yee, P.L.(1984). Semantic priming during sentence processing by young and older adults. Developmental
Psychology, 20,903-910.
Cattell, R.B. (1971). Abilities: Their structure, growth, and action. Boston: Houghton Mif€lin. Cavanaugh, J.C. (1983). Comprehension and retention of television programs by 20- and 60-year olds. Journal of Gerontology, 38, 190-196.
Charness. N. (1981a).Aging and skilled problem solving. Journal of Experimental Psychology: General, 1 10, 2 1-38. Charness, N. (1981b). Search in chess: Age and skill dmerences.
Journal of Experimental Psychology: Human Perception and Performance, 7,467-476. Charness, N. (1983).Age, skill. and bridge bidding: A chronometric analysis. Journal of Verbal Learning and Verbal Behavior. 22. 406-416. Charness, N. (1985).Aging and problem-solving performance. In N. Charness (Ed.), Aging and human performance (pp. 225-259). New York John Wiley & Sons. Chase, W.G., & Simon, HA. (1973). Perception in chess. CognUive
P ~ y ~ h o l o4g,~55-81. , Chi, M.T.H. (1978). Knowledge structure and memory development. In R.S. Siegler (Ed.), Children's thinklng: What develops? (pp. 73-96). Hillsdale. NJ: Lawrence Erlbaum. Chi, M.T.H., & Ceci. S.J. (1987). Content knowledge: Its role, representation, and restructuring in memory development. In H.W. Reese (Ed.), Advances in child development and behavior (Vol. 20. pp. 91-142). New York Academic Press. Chi, M.T.H., & Koeske, R.D. (1983). Network representation of a child's dinosaur knowledge. Developmental Psychology, 19. 2932.
152
Hess
Chiesi, H.L., Spilich, G.J., & Voss, J.F. (1979). Acquisition of domain-related information in relation to high and low domain knowledge. Journal of Verbal Learnlng and Verbal &W, 18, 257-274. Cohen, G. (1979).Language comprehension in old age. Cognitive P~ychology,11,412-429. Craik, F.I.M. (1977).Age differences in human memory. In J.E. Birren & K.W. Schaie (Eds.), Handbook of the psychology of aging (pp. 384-420). New York Van Nostrand Reinhold. Age differences in the temporal Craik, F.I.M., & Masani, P.A. (1967). integration of language. British Journal ofPsychology, 58. 291299. Craik, F.I.M., & Masani, P.A. (1969).Age and intelligence differences in coding and retrieval of word lists. British Journal Of Psycholosy, 60,315-319. Craik, F.I.M.. & Rabinowitz, J.C. (1982).Age differences in the acquisition and use of verbal information: A tutorial review. In H. Bouma & D.G. Bouwhuis (Eds.),Attention and p e ~ o m n c X: e Control of language processes (pp. 471-499). Hillsdale. NJ: Lawrence Erlbaum. Craik. F.I.M., & Rabinowitz, J.C. (1985).The effects of presentation rate and encoding task on age-related memory deficits. Journal of Ger~ntol~gy, 40,309-3 15. Crowder, R.G. (1976).Principles of learning and memory. Hillsdale. NJ: Lawrence Erlbaum. Verbal Dixon, FUL, Hultsch, D.F., Simon, E.W.. & von Eye, A. (1984). ability and text structure effects on adult age differences in text recall. Journal of Verbal Learning and Verbal Behavior, 23,569578. Dixon, R.A., Simon, E.W., Nowak, CA., & Hultsch, D.F. (1982). Text recall in adulthood as a function of level of information, input modality, and delay interval. Journal of Gerontology, 37,358364. Dooling, D.J., & Lachman. R. (1971). Effects of comprehension on the retention of prose. Journal of Experimental Psychology. 88, 216-222. Fincher-Kiefer, R , Post, T.A., Greene, T.R., & Voss, J.F. (1988). On the role of prior knowledge and task demands in the processing of text. Journal of Memory and Language, 27,416-428. Friedman, A. (1979).Framing pictures: The role of knowledge in automatized encoding and memory for gist. Journal of Experimental Psychology: General, 108, 316-355. Galambos, J.A., Abelson, R.P.. & Black, J.B. (1986).Knowledge structures. Hillsdale, NJ: Lawrence Erlbaum.
Schema& Influences on Memory
153
Galambos, J.A., & Rips. L.J. (1982). Memory for routines. Journal of Verbal Learning and Verbal Behavior, 21,260-28 1. Graesser, A.C.. & Nakamura, G.V.(1982).The impact of a schema on comprehension and memory. In G.H. Bower (Ed.), T h e psychology of learning and motivation (Vol. 16, pp. 59-109). New York Academic Press. Graesser, A.C., Woll, S.B..Kowalski. D.J.. & Smith. D.A. (1980). Memory for typical and atypical actions in scripted activities.
Journal of Experimental Psychology: Human Learning and Memory, 6. 503-513. Hagen, J.W., Jongeward, R.H., & Kail. R.V. (1975). Cognitive perspectives on the development of memory. In H.W. Reese (Ed.), Advances in child development and behavior (Vol. 10, pp. 57103). New York:Academic Press. Hamilton, D.L.. Katz, L., & Leirer. V. (1980). Organizational processes in impression formation. In R. Hastie. T. Ostrom. D. Hamilton. E. Ebbesen, R. Wyer. & D. Carlton (Eds.). Person memory: The cognitive bask of social perception (pp. 121- 153). Hillsdale, NJ: Lawrence Erlbaum. Hanley-Dunn. P.. & McIntosh, J.L. (1984). Meaningfulness and recall of names by young and old adults. Journal of
Germtolwy, 39.583-585. Hartley, J.T.. Harker. J.O.. & Walsh. D.A. (1980). Contemporary issues and new directions in adult development of learning and memory. In L.W. Poon (Ed.). Aging in the 1980s (pp. 239-252). Washington, D.C.: American Psychological Association. Hastie, R. (1980). Memory for behavioral information that confirms or contradicts a personality impression. In R Hastie, T. Ostrom. D. Hamilton, E. Ebbesen, R Wyer, & D. Carlton (Eds.), Person memory: The cognitive basis of social perception (pp. 155-177). Hillsdale, NJ: Lawrence Erlbaum. Hastie, R (1984). Causes and effects of causal attribution. Journal of Personality and Social Psychology. 46. 44-56. Heron, A,, & Craik, F.I.M. (1964). Age differences in cumulative learning of meaningful and meaningless material. Scandinavian Journal of Psychology, 5 , 209-2 17. Hess, T.M. (1982). Visual abstraction processes in young and old adults. Developmental Psychology. 18, 473-484. Hess, T.M.(1984). Effects of semantically related and unrelated contexts on recognition memory of different-aged adults.
Journal of Gerontobgy. 39, 444-45 1. Hess, T. M. (1985). Aging and context influences on recognition memory for typical and atypical script actions. Developmental P s ~ c ~ O Z 21. ~ Y .1139-1151.
Hess
154
Hew, T.M. (in press). Adult age differences in script content and structure. In R. L. West & J.D. Sinnott (Eds.). Everyday memory and aging: Current research and methodology. New York: Springer-Verlag. Hess, T.M.. & Amould. D. (1986).Adult age differences in memory for explicit and implicit sentence infonnation. Journal of Gerontdogy, 41, 191-194.
Hess, T.M.. & Flannagan, D.A. (1990). Adult age dflerences in schema-based retrieval processes. Manuscript submitted for publication. Hess, T.M., & Higgins, T.M.(1983).Context utilization in young and old adults. Journal of Cemntology, 38,65-71. Hess, T.M., Donley, J . , & Vandermaas, M.O. (1989). Aging-related changes in the processing and retention of script information. Ex;perlmentaZ Aging Research, 15, 89-96. Hess. T.M.,& Slaughter, S.J. (1986a).Aging effects on prototype abstraction and concept identification. Journal of Gerontology, Ii,214-221. Hess, T.M., & Slaughter, S.J. (1986b). Specific exemplar retention and prototype abstraction in young and old adults. Psychology Md A m , 1,202-207. Hess, T.M.. & Slaughter. S.J. (in press). Schematic knowledge influences on memory for scene information in young and older adults. Developmental Psychotogy. Hess, T.M.,& Tate, C.S. (1989). Adult age differences in
explanations and memory for behavioral information.
Manuscript submitted for publication. Hess, T.M.. Vandennaas, M.O., Donley. J., & Snyder, S.S. (1987). Memory for sex-role consistent and Inconsistent actions in young and old adults. Journal of Gerontology, 42, 505-5 11. Hess, T.M., & Wallsten, S.M. (1987). Adult age differences in the perception and learning of artistic style categories. Psychology and A m , 2.243-253. Horn, J.L. (1970). Organization of data on life-span development of human abilities. In L.R.Goulet & P.B.Baltes (Eds.), Lge-span developmental psychology: Research and theory (pp. 423466).New York Academic Press. Howard. D.V. (1988a).Aging and memory activation: The prlming of semantic and episodic memories. In L.L. Light & D.M. Burke (Eds.), Language, memory, and aging (pp. 77-99). New York: Cambridge University Press.
Schematic Influences on Memory
155
Howard, D.V. (1988b).Implicit and explicit assessment of cognitive aging. In M.L. Howe & C.J. Brainerd (Eds.), Cognitive development In adulthood (pp. 3-37).New York: SpringerVerlag. Howard. D.V., Shaw. RJ.. & Heisey, J.G. (1986). Aging and the time course of semantic activation. Journal of Gerontology. 41. 195203. Howell, S.C. (1972).Familiarity and complexity in perceptual recognition. Journal of Gerontology, 27, 364-371. Hoyer, W.J. (1985).Aging and the development of expert cognition. In M.T. Schlecter & M.P. Toglia (Eds.), New directions In cognitive sctence (pp. 69-87).Norwood, NJ: Albex. Hoyer, W.J. (1987). Acquisition of knowledge and the decentralization of g in adult intellectual development. In C. Schooler & K.W. Schaie (Eds.), Cognitfvefunctionfng and socfal structure over the life course (pp. 120-141).Norwood. NJ: Albex. Hultsch, D.F. (1969).Adult age differences in the organization of free recall. Developmental PsychoZogy, 1 , 673-678. Hultsch. D.F. (1971). Adult age differences in free classification and free recall. Developmental Psychology. 4 , 338-347. Hultsch, D.F.. & Dixon, RA. (1983). The role of pre-experimental knowledge in text processing in adulthood. Exper6nental Agfrtg REZWC~,9, 17-22. Hultsch. D.F., & Dixon, R.A. (1984).Memory for text materials in adulthood. In P.B. Baltes & O.G. Brim, Jr. (Eds.), Life-span development and behavior (Vol. 6, pp. 77-108).New York: Academic Press. Jenkins, J.J. (1979).Four points to remember: A tetrahedral model of memory experiments. In L.S. Cennak & F.I.M. Craik (Eds.), Levels of processhg tn human memory (pp. 429-461). Hillsdale, NJ: Lawrence Erlbaum. Kardash. C.A.M., Royer, J.M., & Greene, B.A. (1988).Effects of schemata on both encoding and retrieval of information from prose. Journal of Educatfonal Psychology. 80, 324-329. Kinsbourne, M. (1973).Age effects on letter span related to rate and sequential dependency. Journal of Gerontology, 28.317-319. Kintsch, W. (1977). On comprehending stories. In M.A. Just & P.A. Carpenter (Eds.), Cognitive processes In comprehensfon(pp. 3362).Hfllsdale, NJ: Lawrence Erlbaum. Kintsch, W., & van Dijk, T.A. (1978).Toward a model of text comprehension and production. Psychologkal Revfew, 85, 363394.
156
Hess
Kliegl. R., & Baltes, P.B. (1987). Theory-guided analysis of mechanisms of development and aging through testing-thelimits and research on expertise. In C. Schooler & K.W. Schaie (Eds.), Cognltiuefunctionlng and social structure ouer the lge course (pp. 95-119). New York Albex. Kucera. H., & Francis, N.W. (1967). Computational analysfs of present-day American English. Providence, RI: Brown University Press. Labouvie-Vief, G. (1980).Adaptive dimensions of adult cognition. In N. Datan & N. Lohmann (Eds.), Zlansitions ofagfng (pp. 3-28). New York Academic Press. Labouvie-Vief, G., & Schell, D.A. (1982) Learning and memory in later life, In B.B. Wolman (Ed.), Handbook of developmental psychology. Englewood Clas, NJ: Prentice-Hall. Lair, C.V.. Moon, W.H., & Kausler. D.H. (1969). Associative interference in the paired-associate learning of middle-aged and old adults. Deuelopmental Psychology, I, 548-552. Levine. R.. Chein. I., & Murphy, B. (1942). The relation of the intensity of a need to the amount of labeling bias. Journal of COnSul@ and Clinical P ~ y ~ h o l ~13, g y 283-293. , Light, L.L., & Albertson, S.A.(1988). Comprehension of pragmatic implications in young and older adults. In L.L. Light & D.M. Burke (Eds.), Lunguage, memory,and aging (pp. 133-153). New York: Cambridge University Press. Light, L.L., & Anderson, P.A. (1983). Memory for scripts in young and older adults. Memory & Cognition, I 1,435-444. Light, L.L., Zelinski. E.M.. & Moore, M. (1982).Adult age differences in reasoning from new information. Journal of Experimental Psychology: Learning, Memory, and Cognition, 8, 435-447. List, J.A. (1986). Age and schematic differences in the reliability of eyewitness testimony. Developmental Psychology. 22. 50-57. Loftus, G.R., & Mackworth, N.H. (1978). Cognitive determinants of fixation location during picture viewing. Journal of Experimental Psycho logy; Human Perception a n d Performance, 4 , 565-572. Mandel, R.G, & Johnson, N.S. (1984).A developmental analysis of story recall and comprehension in adulthood. Journal of Verbal Leming and Verbal Behaulor. 23.643-659. Mandler, J.M. (1984). Stories, scripts, and scenes: Aspects of schema theory. Hillsdale, NJ: Lawrence Erlbaum. Mandler. J.M., & Johnson, N.S. (1976). Some of the thousand words a picture is worth. Journal ofExperlmental Psychology: H u m n Learning and Memory. 2.529-540.
Schematic Znjluences on Memory
157
Mandler, J.M., h Parker, RE. (1976).Memory for descriptive and spatial information in complex pictures. Journal of Experimental Psychology: Human Learning and Memory, 2,38-
48.
Mandler. J.M., 81 Ritchey, G.H. (1977).Long-term memory for pictures. Journal of Experimental Psychology: Human Learning and Memory. 3.386-396. Markrnan, E.M. (1981).Two different principles of conceptual organization. In M.E. Lamb and A.L. Brown (Eds.). Advances @I developmental psychology (Vol. 1, pp. 199-236).Hillsdale, NJ: Lawrence Erlbaum. Nelson, K. (1986).Event knowledge: Structure and function in development. Hillsdale, NJ: Lawrence Erlbaum Associates. Omstein. P.A., Baker-Ward, L.. & Naus. M.J. (1988).The development of mnemonic skill. In F.E. Weinert & M. Perlmutter (Eds.), Memory development: UnfuersaZ changes and individual differences (pp. 31-50).Hillsdale. NJ: Lawrence Erlbaum. Padgett, R.J., & Ratner, H.H. (1987).Older and younger adults' memory for structured and unstructured events. Experimental Aging Research 13,133-139. Palmer, S.E. (1975).The effect of contextual scenes on the identification of objects. Memory & Cognition, 3. 519-526. Pezdek. K. (1980).Life-span dflerences in semantic integration of pictures and sentences in memory. ChW Development. 51. 720729. Pezdek. K., & Miceli. L. (1982).Life-span differences in memory integration as a function of processing time. Developmental P~y~hology. 18,485-490. Piaget, J. (1970).Piaget's theory. In P.H. Mussen (Ed.), Carmichael's handbook of child psychology (Vol. 1. pp.703-732).New York Wiley. Piaget, J., & Inhelder, B. (1973).Memory and intelligence. New York: Basic Books. Speed of retrieval from long-term Poon. L.W..& Fozard. J.L. (1978). memory in relation to age, familiarity, and datedness of information. Journal of Gerontology. 33,71 1-717. Rabinowitz. J.C.. Craik. F.I.M.. & Ackerman. B.P. (1982).A processing resource account of age differences in recall. Canadian Journal of Psychology. 36.325-344. Rankin, J.L., Kausler, D.H. (1979). Adult age differences in false recognitions. Journal of Gerontology, 34.58-65. Ratner, H.H., Padgett. R.J., & Bushey. N. (1988).Old and young adults' recall of events. Developmental Psychology, 24, 664-671.
158
Hess
Reder. L.M., Wible, C.. & Martin, J. (1986).Differential memory changes with age: Exact retrieval versus plausible inference. Jounuxl of Experimental Psychology: Learning, Memory, and Cognftion, 12,72-81. Reiser, B.J. (1986). Knowledge-directed retrieval of autobiographical memories. In J.K. Kolodner & C.K. Riesback (Eds.), Experience, memory, and reasonlng (pp. 75-93). Hillsdale, NJ: Lawrence Erlbaum. Rumelhart, D.E. (1984).Schemata and the cognitive system. In RS. Wyer, Jr.. & T.K. Srull (Eds.). Handbook of social cognftlonWol. 1, pp. 161-188). Hillsdale, NJ: Lawrence Erlbaum. Rybash, J.M., Hoyer, W.J., & Roodin, P.A. (1986).Adult cognition and aging. New York Pergamon Press. Salthouse, T.A. (1984).Effects of age and skill in typing. Journal of Experimental Psychology: General, 1 13, 345-371. Schank, R C . (1982).Dynamic memory:A theory of reminding and learning in computers and people. New York: Cambridge University Press. Schank, RC., & Abelson, RP. (1977).Scrfpts. plans, goals, and understanding. Hfflsdale, NJ: Lawrence Erlbaum. Sharkey. N.E. (1986).A model of knowledge-based expectations in text comprehension. In J. A. Galambos et al. (Eds.), Knowledge structures (pp. 21-47).Hillsdale. NJ: Lawrence Erlbaum. Sinnott , J.D. ( 1986). Prospective/intentional and incidental everyday memory: Effects of age and passage of time. P~ycholOgY& Aging, 1 , 1 10-1 16. Smith, A D . (1975).Partial learning and recognition memory in the aged. International Journal of Aging and Human Development. 6,359-365. Smith, D. A.. & Graesser, A.C. (1981). Memory for actions in scripted activities a s a function of typicality, retention, interval, and retrieval task. Memory & Cognition, 9,550-559. Smith, S.W.. Rebok, G.W., Smith, W.R., Hall, S.E., & Alvin, M. (1983). Adult age differences in the use of story structure in delayed free recall. Exper-lmental Aging Research, 9,191- 198. Spilich. G.J. (1983).Life-span components of text processing: Structural and procedural changes. Journal of Verbal Learning and Verbal Behavior, 22.231-244. Spilich, G.J., Vesonder, G.T., Chiesi, H.L., & Voss,J.F. (1979). Text processing of domain-related information for individuals with high and low domain knowledge. Journal of Verbal Learning and Verbal Behavior, 18.275-290.
Schematic InJuences on Memory
158
Srull, T.K. (1981). Person memory: Some tests of associative storage and retrieval models. Journal of Experimental Psychology: H ~ m a Learning n and M e m ~ r y 7.440-463. .
Stine, E.A.L. (1990). On-line processing of written text by younger and older adults. Psychology and A g w , 5.68-78. Stine, E.L., & Wingfield, A. (1987). Process and strategy in memory for speech among younger and older adults. Psychology and Agfng, 2,272-279. Sulin. R.A.. & Dooling, D.J. (1974). Intrusion of a thematic idea in retention of prose. Journal of Experimental Psychology, 103. 255-262.
Taub, H.A (1974). Coding for short-term memory a s a function of age. Journal of Genetic Psychology, 125,309-314. Thorndike. E.L. (1921). The teacher's word book. New York: Columbia University. Thorndyke. P.W., & Yekovich, F.R. (1980). A critique of schemabased theories of human story memory. Poetics. 9.23-49. Till, RE,, & Walsh, DA. (1980). Encoding and retrieval factors in adult memory for implicational sentences. Journal of Verbal Leamfng and Verbal Behavior, 19, 1- 16. van Dijk, T.A.. & Kintsch, W. (1983). Strategies of discourse comprehension. New York Academic Press. Waddell, KJ.. & Rogoff, B. (1981). Mfect of contextual organization on spatial memory of middle-aged and older women. Developmental PsychoZwy, 17,878-885. Waddell, K.J.. & Rogoff, B. (1987). Contextual organization and intentionality in adults' spatial memory. Developmental P~ychobgy.23.514-520. Walsh, D.A.. & Baldwin. M. (1977). Age differences in integrated semantic memory. DeveZopmental Psychologg, 13, 509-514. Walsh, D.A., Baldwin, M.. & Finkle, T.J. (1980). Age differences in integrated semantic memory for abstract sentences. Experfmental Agfng Research 6.43 1-444. Wingfield, A.. Poon, L.W.. Lombardi. L., & Lowe, L. (1985). Speed of processing in normal aging: Effects of speech rate, linguistic structure, and processing time. Journal of Gerontology, 40, 579585.
Worden, P.E.. & Sherman-Brown. S. (1983). A word-frequency cohort effect in young versus elderly adults' memory for words. Developmental Psychology. 19, 52 1-530. Yekovich, F.R, &Walker, C.H.(1986).Retrieval of scripted concepts. Journal of Memory and Language. 25, 627-644. Zacks, RT., & Hasher, L. (1988).Capacity theory and the processing of inferences. In L.L. Light & D.M. Burke (Eds.), h n g u a g e ,
Hess
160
memory, and wing (pp. 154-170). New York: Cambridge University Press. Zacks, RT., Hasher. L., Doren, B..Hamm, V.. & Attig, M.S.(1987). Encoding and memory of explicit and implicit information. Journal of Gerontology, 42, 418-422. Zadny, J., & Gerard, H.B. (1974). Attributed intentions and information selectivity. Journal of Experimental and Social PSyCholOgy. 10.34-52.
Zelinski. E.M. (1988). Integrating information from discourse: Do older adults show deficits? In L.L.Light & D.M. Burke (Eds.). Language, memory, and aging. (pp. 117-132). New York Cambrldge University Press. Zelinski, E.M., Gilewski, M.J..& Thompson, L.W. (1980). Do laboratory tests relate to self-assessment of memory ability in the young and old? In L.W. Poon. J.L.Fozard. L.S. Cermak, D. Arenberg. & L.W. Thompson (Eds.),New directions in memory and aging (pp. 519-544). Hillsdale, NJ: Lawrence Erlbaum. Zelinski, E.M., & Miura, S.A. (1988). Context effects on script memory in young and old adults. Psychology and Aging. 3,292299.
Aging and Cognition: Knowledge Organization and Utilization Thomas M. Hess (Editor) 0 Elseuier Science Rhlishers B.V. Il\rorth-Holland), 1990
CHAPTER FOUR
METAMEMORY IN ADULTHOOD: DIFFERENTIATING KNOWLEDGE, BELIEF, AND BEHAVIOR Christopher Hertzog Georgia Institute of Technology
Roger A. Dixon and David F. Hultsch University of Victoria
SUMMARY
This chapter reviews the literature on metamemory -defined broadly as cognitions about memory -- in adults. Our own research program has examined two self-report questionnaires measuring multiple dimensions of metamemory. two memory tasks (word and text recall), and memory task predictions in cross-sectional and longitudinal samples of adults. Knowledge about how memory functions is weakly related to beliefs about one's own ability to remember, defined as memory self-efficacy (MSE). We have obtained evidence for the convergent and discriminant validity of the metamemory questionnaire measures of MSE. MSE has a significant, but modest, relationship to recall performance. Predictions of task performance also correlate significantly but modestly with recall, but the relationship increases following task experience. Preliminary longitudinal results show that questionnaire measures of MSE have high stability of individual differences. Results from this series of studies are consistent with the view that MSE is a stable, hierarchically differentiated belief system regarding memory that (a) is not necessarily accurate, and (b) is distinct from both declarative knowledge about memory
162
H e r t z q . D W n & Hdtsch
functioning and strategic behaviors in memorydemanding situations. Psychologists interested in age changes in memory have become increasingly aware of the complex interplay between multiple sets of variables in determining memory performance. and the ways in which aging effects on memory tasks can be exacerbated or attenuated by experimentally manipulated variables and subject characteristics (Hultsch & Dixon, 1984; in press). There are normative age-related changes in memory task performance (Kausler. 1982). It is also clear, however, that individual differences in memory task performance are influenced by (a) traditionally studied memory mechanisms, such a s encoding and retrieval, (b) background subject characteristics, such a s prior knowledge and familiarity with the material to be remembered, and (c) contextual influences on memory task behaviors (Dixon & Hertzog. 1988; Hultsch & Dixon, in press). One domain of variables to be considered regarding acquisition and retention of to-be-remembered information involves the general construct of metamemory (Cavanaugh, Morton, & Tilse, 1989; Dixon, 1989; Dixon & Hertzog, 1988; Lachman & Lachman, 1980: Perlmutter. 1978; Zelinski. Gilewski, & Thompson, 1980). The term metamemory, defined broadly as cognitions about memory (e.g.. Wellman, 19831, is in essence a label for multiple aspects of knowledge, beliefs, and behaviors related to memory (Cavanaugh, Kramer, Sinnott. Camp, & Markley, 1985; Dixon, 1989; Gilweski & Zelinski, 1986; Hultsch, Hertzog, Dixon. & Davidson, 1988). We have been collaborating on an ongoing series of studies of adults designed to (a) assess the structure of metamemory, (b) validate existing questionnaire measures of metamemory, (c) c l a w the relationships among multiple operational definitions of metamemory employed in the literature, especially questionnaires and memory task predictions, and (d) elucidate the relationships between the multiple dimensions of metamemory and performance on memory tasks. All of these objectives have been framed within the superordinate goal of clarifying the ways in which
Metamemoy in Addthouti
163
metamemory is related to observed age differences and age changes in memory functioning. In this chapter we will selectively review the general literature, with a n emphasis on reviewing results from our own studies. We attempt to integrate these results into an updated summary statement regarding metamemory and its relevance to cognltive aging research. We begin with a brief review of the metamemory construct in developmental psychology and t h e early gerontological literature studying aging, metamemory, and memory performance. METAMEMORY IN ADULTHOOD: AN OVERVIEW It is by now well-accepted
that metamemory is a multidimensional domain, and several reviewers of the literature have attempted to categorize the plethora of metamemory-related constructs explored in the cognitive and developmental literature (e.g.. Cavanaugh et al., 1989: Cavanaugh & Perlmutter, 1982: Dlxon. 1989; Hultsch et al., 1988; Schneider, 1985). Certainly, the original usage of metamemory, as coined by Flavell and colleagues (see Flavell, 1977). emphasized the connotations of awareness of or knowledge about memory. However, the construct domain subsumed under the term metamemory is clearly more complex. Recently Hultsch et al. (1988) identified four broad aspects of metamemory: (a) factual knowledge about memory tasks and memory processes -- defined a s knowledge about both how memory functions and the viability of strategic behaviors for tasks requiring memory processes: (b) memory monitoring -- defined as awareness of how one typically uses memory as well a s the current state of one's memory system: (c) memory self-efficacy -- defined a s one's sense of mastery or capability to use memory effectively in memory-demanding situations; and (d) memory-related affect -defined as a variety of emotional states that may be related to or generated by memory demanding situations, including anxiety, depression, and fatigue. For cogent reasons, the first two categories dominated early work on metamemory, particularly in the domain of child development. As we shall see, however, the latter two
164
Hertzog, Dlxon & Hultsch
categories have begun to emerge as important areas for understanding metamemory in adulthood and old age. The Dominant Metaphor: Metamemory a s Knowledge Metamemoy and Strategu Utiflzatfon
Flavell (1977: Flavell & Wellman. 1977) focused on metamemory as a possible explanation for developmental changes in episodic memory task performance during the transition from early t o late childhood. He suggested that children's task performance improved in large part because they became aware of the causal linkage between different behaviors during encoding and retrieval and the subsequent probability of successful recall. Moreover, as children grow older they are better able to monitor the extent to which they have successfully attained a representation of a to-be-remembered informational unit in memory. Some of the improvements in performance were therefore attributed to the volitional application of effective strategies, due in large part to (a) the emerging knowledge that such strategies exist and can be effective, and (b) the awareness that behaviors designed to enhance encoding and/or retrieval are required, given the current state of the memory system. This perspective led to a metacognitive version of the production de..lency tqpthesfs: deficient memory performance in individual children, or specialized subgroups such a s the learning disabled, might be attributed to deficient metamemory, and in particular. suboptimal use of memory strategies, rather than to any inherent deficiencies in basic memory processes themselves (e.g., Belmont, 1978; Borkowski, Carr,& Pressley, 1987; Brown, 1978). Sirnilar concerns were raised regarding development during adulthood. Lachman and Lachman ( 1980) explicitly identified metamemory with knowledge which facilitates executive control through effective monitoring of the status of the memory system. They identified metamemory a s a crucial mediating variable in determining successful retrieval and utilization of world knowledge. Perlmutter (1978) actually studied a wide variety of questions and indices of metamemory, but her theoretical
Metarnemory in Adulthood
165
orientation clearly emphasized metamemory a s a proximal determinant of strategy selection and utilization (see especially Perlmutter, 1980). It is not surprising, then, that gerontologists would seek to test the production deficiency hypothesis in old age, examining whether deficient metamemory knowledge and awareness might determine inefficient memory task strategies in older persons, and in turn, produce obsemed age differences in memory task performance. The hypothesis of a link between metamemory, age-related production deficiencies in memory strategies, and age diEerences in memory performance, is analogous to other hypotheses in the cognitive aging literature that have explained poor performance by older adults in terms of age differences in task-specific performance confounds (e.g., fast pacing of stimulus presentation) or subject behaviors (e.g., conservative response criteria: Botwinick. 1967; Kausler, 1982). Some studies suggested less spontaneous and effective use of appropriate mnemonic strategies by older persons. including organizational strategies at encoding and maintenance strategies prior to recall (e.g.. Hulicka & Grossman, 1967; Hultsch, 1969, 1975;Treat, Poon, Fozard. & Popkin. 1978). Early work with orienting tasks, thought to control processing at encoding, suggested that age dmerences in recall tasks might be in large part a function of suboptimal processing strategies (e.g.. Eysenck, 1974).although subsequent research demonstrated that attempts to control encoding processes (e.g., via orienting tasks) were insufficient to account for all age differences in recall performance (Craik & Rabinowitz, 1984:Smith, 1980). A modicum of support for the strategy deficiency hypothesis can be found in the recent literature. Sanders, Murphy, Schmitt. and Walsh (1980)compared adults on a free recall task using categorizable nouns, both with and without explicit instructions to rehearse. They found age differences in both clustering, a s measured by a ratio of repetitions measure, and recall, but found no age differences in category intrusion errors. Sanders et al.'s (1980) overt rehearsal data suggested that young persons shifted from serial to category rehearsal in the latter part of list presentation, but that old subjects showed no increase in categorically clustered rehearsals across items. A subsequent study by Schmitt. Murphy,
166
Hertzog, Dixon & Hultsch
and Sanders (1981) demonstrated that older persons could be tratned to use categorical rehearsal strategies on the task, and that performance improved when they did so. T w o other studies from the same laboratory suggested deficient strategies by older persons in a serial recall paradigm. Murphy, Sanders, Gabriesheski, and Schmitt (1981) found that, given unlimited time to study sets of line drawings, younger subjects studied longer and achieved better serial recall than older subjects. A separate experiment demonstrated that older persons who were simply required to study longer performed better than either older persons given strategy training or older controls. Murphy et al. (1981) attributed age dmerences in the study time paradigm to a memory monitoring deficiency: older adults were less aware of readiness to recall and hence did not select an optimal study time. Murphy, Schmitt, Caruso, and Sanders (1987) replicated the finding that, under standard recall instructions, young subjects tended to study longer, rehearse more, and recall better than old adults. However. older adults improved their utilization of study time and rehearsal under explicit instructions, and performance improved. Rabinowitz (1989) recently reported results suggesting that providing additional opportunity to utilize strategies is. by itself, insufllcient to guarantee optimal strategy utilization by the elderly. He provided optimal study conditions, in the form of unlimited study time, permission to take notes, and explicit encouragement to use strategies. The optimal conditions yielded improved recall over standard procedures. but the improvements in recall were greatest for the young subjects. Study time was, however, equivalent for both age groups, suggesting to Rabinowitz that any strategy deficiencies were in how the study time was used (poor strategy selection or i n a c i e n t strategy use). The previous studies suggest a possible strategy production deficiency in at least some memory task paradigms, but do not explicitly demonstrate that such deficiencies arise in the presence of either inadequate knowledge about potential strategy efficiency or lack of awareness of the need for strategy utilization in a memory performance context. Brigham and Pressley (1988) presented more persuasive evidence for a link between metacognitive deficiencies and deficient strategy utilization in
Metarnmry in Adulthood
167
older persons. They required subjects to learn new. esoteric vocabulary after exposure to two alternative mnemonic strategies: key word generation and semantic context generation. Subjects' use of the two strategies during practice was controlled by alternating the presentation of processing instructions for one of the strategies along with the new word. Subjects predicted performance in one of two conditions: judgements and strategy choices after, or both before and after, performance on the practice list. After study and recall, they gave postdictions (performance evaluations for each strategy) and indicated which strategy they would choose if asked to learn a new vocabulary list. Brigham and Pressley (1988) found that the keyword method was superior to the semantic context method for both age groups. Young subjects were more likely to adjust their postdictions in the main experiment to reflect differences in strategy effectiveness, and were more likely to nominate the keyword method for subsequent use. Older persons were apparently less aware of the relative superiority of the keyword strategy, even after employing it. However, the preceding literature does not definitively show that metacognitive deficits are linked to a strategy production deficiency. Moreover. the studies cited do not provide support for the hypothesis that qualitative age difTerences in utilization of task strategies can account for all the age-related variance in memory task performance. Although Schmitt et al. (1981) noted that the performance of their strategy-trained older subjects on high strength lists (nouns with high category typicality) approximated performance of young subjects in the Sanders et al. (1980) study, comparisons of the serial position recall curves for all lists suggests superior performance by the young (who had not, of course, received strategy training). Murphy et al. (1987) found that, even under conditions of strategy instructions. older persons recalled supraspan list materials more poorly than young subjects. Moreover. Murphy et al. (1981, 1987) determined the number of elements in the series of line drawings presented to subjects on the basis of individual span estimates, and there were age Werences in the initial memory spans. The Murphy et al. (1987) finding of equivalent percentage of serial recall by older persons under strategy instructions in span and span+2 length series does not
168
H e r t z q , DLxon & Hultsch
imply equivalent absolute performance levels for different age groups. Their paradigm seems well-suited for examining instructional effects within individual subjects, but is problematic as a basis for between-person (i.e., age) comparisons of memory performance. Finally, Brigham and Pressley (1988) found that age differences in recall were greater when subjects were forced to encode the items using the superior keyword method. In general, then, the literature suggests that memory tasks can be constructed which maximize the probability of performance being influenced by strategies during acquisition, and that such tasks will adversely affect performance of older adults, relative to younger adults (Craik & Rabinowitz, 1984). Although age differences on memory tasks are at least partly due to age-related. structural deficiencies in memory mechanisms, such differences can be amplified by age differences in strategy selection and utilization during acquisition and maintenance of information in memory prior to retrieval. Metamemory, in turn, may be a contributing factor to deficient strategy utilization by older persons in two ways: (a) failure to construct and/or identify the strategic behaviors necessary to optimize task performance, perhaps secondary to deficient knowledge of task strategies and their effectiveness, and (b) inadequate ability to monitor the contents of memory (epistemic awareness). Deficient memory monitoring would reduce the likelihood of (a)selective application of encoding, maintenance, and retrieval enhancing processes to the information at greatest risk for subsequent forgetting, and (b) adaptive selection of alternative, appropriate strategies given the state of the memory system. A rather dflerent strand of evidence regarding age differences in strategy utilization comes from the questionnaire literature on memory strategies in everyday life. One issue is whether older adults report that they intentionally use behaviors designed to aid remembering (either internal mnemonics [e.g., rehearsal] or external memory aids [e.g., appointment books]). Questionnaire studies indicate that older persons do report using such memory aids, but there is some disagreement regarding the existence of age dflerences in use of memory strategies in everyday life (Cavanaugh & Poon, 1989: Dixon & Hultsch, 1983a; Hultsch. Hertzog. & Dixon,
M e t a m m r y in Adulthood
169
1987; Loewen. Shaw, & Craik. in press). A second issue is whether self-reported strategy use relates to memory performance. Unfortunately, the literature to date has only employed laboratory memory tasks. There is at best a weak relationship between selfreported strategies and memory task performance. For example, Dixon and Hultsch (1983a) found that self-reported memory strategies significantly predicted text recall in some, but not all, of their three young and middle-aged subsamples. The same relationship was not statistically reliable in their older adult samples. Zelinski et al. (1980) also did not find a significant relationship of self-reported memory strategies in everyday life to memory task performance In an aggregated sample of young and old subjects. In summary, the evidence for age differences in strategy utilization is mixed, but suggests that there may be age dmerences in formulation and use of optimal strategies on laboratory cognitive tasks. Older persons may be less likely to spontaneously develop effective strategies and recognize the differential utility of strategies to which they have been exposed. It does not necessarily appear, however, that any deficiencies in strategy utilization identified via experimental task paradigms are related to differences in ability to monitor status of the memory system (epistemic awareness). Furthermore, age differences in selfreported use of external aids and mnemonics in everday life do not parallel observed age differences in strategy selection and utilization in experimental tasks.
Metamemory and Recall Predictions One method often used for operationally defining metamemory in studies of both children and adults has been some form of prediction of memory task performance. Consistent with the original emphasis on metamemory as self-referent knowledge and awareness of memory processes, memory predictions have often been conceptualized as an index of knowledge about one's own memory (Cavanaugh & Perlmutter, 1982; Cavanaugh et al., 1989; Schneider. 1985). The implicit logic is as follows: the more accurate the prediction, the better one's knowledge about one's own memory
170
Hertzag, Dkon & Hultsch
and how it functions, Cavanaugh et al. (1989) interpreted predictions a s a n aspect of on-line awareness of memory functioning, a construct closely tied to the concept of memory monitoring discussed briefly above. Much of the early literature on adult age differences in metamemory involved some type of memory prediction paradigm (e.g., Bruce, Coyne, & Botwinick. 1982: Murphy et al., 1981; Perlmutter, 1978). A common procedure is to give subjects a description of a task, or limited experience with examples or practice, followed by a prediction of performance. The central question for such research is usually whether there are age differences in the accuracy of performance predictions. Several studies have suggested that old subjects overestimate their performance on cognitive tasks (Brigham & Pressley, 1988: Bruce et al.. 1982: Coyne, 1985; Lachman and Jelialian 1984: Lovelace & Marsh. 1985; Murphy et al., 1981). In contrast, other studies have found relatively accurate memory task predictions by older adults (Camp, Markley, & Kramer, 1983: Lachman. Steinberg, & Trotter, 1987; Perlmutter, 1978). Recently, Lachman et al. (1987) reported results suggesting that older persons can increase the accuracy of their predictions after task experience. In their study, multiple regression analysis indicated a significant relationship between prior recall performance levels and predictions prior to a second recall trial (controlling for predictions on the first trial). Moreover, there seems to be relatively little age difierence in either (a) accuracy of predictions of future performance based on item-by-item ratings of memorability (Lovelace & Marsh, 1985: Rabinowitz. Ackerman, Craik, & Hinchley, 1982) or (b) feeling of knowing judgements collected during or after the study phase of a memory task, but prior to recall (e.g. Lachman, Lachman, & Thronesbexy, 1979: Buttefield, Nelson, & Peck, 1988). The different pattern of age differences across prediction paradigms suggests that general performance ratings after instructions or practice, but prior to the study of memory task materials, may measure a different aspect of metamemory than other prediction paradigms (Lovelace & Marsh, 1985). Be that as it may, these types of prediction paradigms offer little evidence of meaningful age differences in ability to assess the
Metamemory in Adulthood
171
status of one's own memory. It appears, therefore, that deficiencies in utilization of strategies in laboratory tasks by older adults may not be a simple function of problems in memory monitoring. The major exception to this conclusion is the data reported by Brigham and Pressley (1988).in which they found less accurate postdictions on the part of the elderly, as well as a lesser influence of differential strategy utility on postdictions and self-reported strategy preferences after the task. An implicit assumption of most prediction studies is that prediction accuracy is determined by both (a) accuracy of declarative knowledge about one's own memory system, and (b)the degree of awareness of current status of information held in memory. Accurate knowledge about one's own memory system implies accurate knowledge about how memory functions. in general. How accurate is the knowledge adults of different ages have about memory systems? A recent study by Shaw and Craik (1989) showed that neither young nor old subjects vary predicted item recall to accurately reflect the influence of encoding conditions on subsequent cued recall performance. Both age groups were more likely to predict successful recall for items that actually were recalled, and young subjects were slightly more likely to accurately predict subsequent recall. Rabinowitz et al. (1982)similarly showed that item predictions were not influenced by an imagery manipulation known to d e c t recall. Brigham and Pressley (1988) found that both young and old subjects failed to show a preference for the keyword strategy after being introduced to it. but prior to actually using it to study vocabulary items. Such findings suggest that experimental psychologists may make unwarranted assumptions about the extent of knowledge subjects have about memory functioning, as well as the degree to which individuals can evaluate the memory processing demands of tasks with which they have little or no prior experience (e.g.. Glenberg. Sanocki. Epstein, & Morris, 1987; Herrmann. Grubs, Sigmundi. & Grueneich, 1986). Indeed, Shaw and Craik (1989) explicitly questioned the validity of the prediction-recall discrepancy a s an index of accuracy of memory monitoring, in part because of the variable magnitude of discrepancy scores across within-subjects experimental conditions.
172
Hertzog, Dtwon & Hultsch
More generally, one can question the assumption that predictions -- whether at the level of overall task performance or at the level of individual items -- are based upon knowledge of memory functioning and awareness of the contents of memory. Indeed, much of the recent work on adult age and metamemory has been conducted under the assumption that self-knowledge is not necessarily the appropriate metaphor for understanding these constructs. An Alternative Metaphor: Metamemory a s a Belief System
The concept of metamemory as a set of beliefs has been manifest in both the clinical and experimental literatures on memory and self-reported memory functioning. In gerontology, use of the beliefs metaphor stems in large part from work done on presentation of memory complaints in assessment contexts. Although memory complaints can be manifested as reported memory dysfunction to a physician or a mental health professional, in the research literature complaints are usually operationally defined by responses to self-rated memory problems questionnaires (Gilewski & Zelinksi, 1986). A common finding is that older persons complaining of poor memory or memory loss actually perform within normal ranges on standardized memory tests, as judged by age-based norms or non-complaining comparison groups (e.g., Kahn et al., 1975;Zarit. Cole, & Guider, 1981). Several studies have found relatively weak correlations between memory complaints and memory task performance, indicating that individual differences in complaints may not predict individual differences in memory (Sunderland, Watts, Baddeley, & Harris, 1986). Some studies have suggested, in fact, that memory complaints are more likely to be related to depressive affect than memory functioning (Popkin, Gallagher, Thompson, & Moore, 1982;West, Boatwright, & Schleser. 1984;Zarit et al.. 1981). A study by O'Hara, Hinrichs. Kohout, Wallace, and Lemke (1986)is representative of this literature. They compared groups of older subjects with clinically diagnosed depression, self-reported depressive affect (without clinical diagnosis), and low self-reported depressive afl'ect. The two depressive groups reported more memory complaints, as measured
Metamemory in Addthood
173
by an abbreviated version of the Zelinski et al. (1980) frequency of forgetting measure, and greater change in memory from young adulthood, than older persons with low depressive affect. Free recall performance was slightly lower in the depression and high affect groups, but the differences were not statistically reliable. Moreover, complaints did not correlate with recall performance. Other studies have confirmed a depression/memory complaint relationship, but have found some significant, but modest correlations of complaints and memory task performance (e.g., Niederehe & Yoder. 1989; Sunderland et al., 1986). Findings of weak relationships of complaints to memory performance led many individuals to suggest that memory complaints were often invalid beliefs associated with negative affect rather than accurate reflections of the status of one's memory system (e.g.. Zarit. 1982). Treatments of metamemory as a set of beliefs also emerged outside of the clinical memory literature. Sehulster (1981)was one of the earliest cognitive psychologists interested in self-reported memory to advocate a conceptuallzation of metamemory as a set of beliefs (see also Herrmann, 1982. in press). He characterized perceptions of memory ability as components of a self-theory of memory, and emphasized that beliefs about one's memory ought to be viewed as a subset of the person's beliefs about self. Sehulster (1981) emphasized that use of the concept of memory self-theory had multiple advantages. First, it grounded metamemory within the larger framework of self-theory. rendering the burgeoning findings regarding self conceptions and theories relevant to the understanding of how beliefs about memory are formed and maintained. Second, the concept of belief calls into question the veridicality of beliefs about memory (Langer. 1981). Third. the concept of memory beliefs suggests that certain behaviors (e.g., the amount of risk one takes in situations perceived a s demanding memory) may be more highly related to memory beliefs than actual memory ability (Sehulster, 1981). Along with others, we have argued that beliefs in one's ability to remember is an important domain of metamemory that is differentiable. theoretically and empirically, from other aspects of metamemory (Berry, West, & Dennehy. 1989; Cavanaugh et al., 1985; Dixon & Hertzog. 1988; Hertzog. Dixon. Schulenberg. &
174
Hertzcg, D M n & Hultsch
Hultsch, 1987;Hultsch, Hertzog, & Dfxon, 1985;Perlmutter. Adams, Berry, Kaplan, Person, & Verdonik, 1987). In particular, we have emphasized the importance of differentiating knowledge about memory mechanisms and processes from beliefs about one's own memory abilities, strengths, and weaknesses (Dixon, 1989;Hertzog et al., 1987).subsumed under the term memory setf-efncacy. Memory self-efficacy can be defined a s beliefs about one's own capability to use memory effectively in different situations (Bandura, 1989;Berry et al., 1989;Cavanaugh et al., 1989). The differentiation of knowledge about memory functions from memory self-efficacy allows for the possibility that a n older individual may have extensive and accurate knowledge about how memory functions but may also believe that his or her ability to remember in a given context is poor. The concept of memory selfefficacy also makes it possible to entertain questions concerning the accuracy of memory beliefs, a s noted above, and to develop intervention techniques designed to identify and ameliorate negative self-efficacy beliefs s o a s to enhance the effective functioning of the individual (Bandura, 1986, 1989; Rebok & Balcerak, 1989;Weaver & Lachman, 1989;Zarit, 1982). As we noted at the beginning of this chapter, the knowledge/beliefs distinction is inadequate for the purposes of categorizing all metamemory-related constructs. However, it seems likely that the construct of memory self-efficacy helps to explain much of the observed age differences in multidimensional metamemory questionnaire responses. The most robust flnding in the literature is that the greatest age differences are observed on measures of perceived change in memory ability during adulthood (Hultsch et al., 1987; Perlmutter, 1978). It also appears that measures of current memory complaints and perceived memory ability show age differences, with lower perceived memory capacity and a greater degree of memory complaints by older persons (Dixon & Hultsch, 1983b: Hultsch et al., 1987;Zelinski et al., 1980). although some studies have failed to find differences in current memory complaints (e.g., ChaMn & Herrmann, 1983). Dixon and Hultsch's (1984)Memory in Adulthood (MIA) questionnaire contains three subscales that seem closely related to the concept of memory self-efficacy (Capacity -- measuring
Metamemory fn Adulthood
175
perceived ability: Change -- measuring perceived change in memory ability; and Locus -- measuring perceived control over one's memory). Several studies have found that there are reliable age differences on these three scales (Cavanaugh & Poon, 1989: Carroll & Drevenstedt, 1987: Dixon & Hultsch. 1983b. Hultsch et al.. 1987: Loewen et al., in press). Age differences on other scales from the MIA, especially the Task scale which measures general knowledge about memory functions, have been found much less frequently. As already noted, age differences in self-reported strategy use are also less robust than differences on measures that appear to be related to memory-self efficacy. METAMEMORY QUESTIONNAIRES: WHAT DO THEY MEASURE?
As indicated above, one approach for measuring metamemory has been the use of self-report questionnaires. There is currently a plethora of questionnaires available for work with adults (Dixon. 1989: Gilewski & Zelinski. 1986: Herrmann, 1982). The four questionnaires most frequently used in aging studies are (a) the Short Inventory of Memory Experiences (SIME. Herrmann & Neisser, 1978). (b) the Everyday Memory Questionnaire (EMQ, Sunderland, Harris, & Baddeley, 1983). (c) the MIA (Dixon, Hultsch. and Hertzog, 1988). and (d) the Memory Functioning Questionnaire (MFQ, Gilewski & Zelinski, 1988). The Memory Self-Efficacy Questionnaire (MSEQ; Berry et al., 1989) is a recently developed instrument that explicitly focuses on measurement of task-specific memory self-efficacy. and may also eventually achieve widespread usage. Given the two rather different metaphors that have guided metamemory research - - metamemory a s knowledge and metamemory as beliefs -- it is reasonable to ask what constructs have, in fact, been measured in different metamemory instruments. Comparison of the questionnaires suggest major differences in content, format, and breadth (Dixon. 1989; Gilewski & Zelinski, 1986). One difference, for example, is the extent to which the questionnaires obtain assessments of memory problems in very specific domains (such as remembering names and appointments), a s does the SIME or MFQ. as opposed to obtaining more general
176
Hertzcg, D b n & Hdtsch
ratings of memory ability, Moreover, scales such as the MIA and MFQ measure multiple aspects of metamemory, including perceived change in memory and strategy utilization. One motivation for our recent work on metamemory questionnaires was our working hypothesis that respondents' memory self-enicacy beliefs play a central role in their responses to memory questionnaire items. MSE may influence responses to questions that arguably are measures of different metamemory constructs (e.g., achievement motivation about memory, or knowledge about memory functioning). In addition, questions about different kinds of memory functioning may be measures of MSE despite differences in content, domain, and format. MSE may be the chief construct measured by a metamemory scale, independent of whether it was designed to measure specific kinds of memory complaints or beliefs about memory ability. We doubted that frequency of forgetting scales were actually measuring a different construct than scales of self-rated memory ability. Although it might be the case that actual incidents of memory dysfunction form the evidential basis for beliefs about memory ability and age-related changes in memory, we suspected that selfrated frequency of forgetting was not necessarily based directly upon the actual frequency of incidents of forgetting. Although the actual frequency of forgetting is likely to be a cause of memory selfefficacy beliefs, the proximal cause of self-reported frequency of forgetting is not necessarily true frequency of forgetting but, rather, the memory self-efficacy beliefs themselves. Individuals' responses to questionnaires are often based upon generalized beliefs and selfschemata -- the stable representations of self that are products of a self-appraisal process (e.g.. Markus & Wurf, 1987; Srull & Wyer, 1989). There is good evidence from other research on questionnaire responses (e.g., self-reported frequency of doctor visits in health surveys) that individuals usually do not base frequencyjudgements upon a n exhaustive retrieval search of memory, but rather upon more general schematic beliefs about themselves (e.g., Bradburn, Rips, & Shevell, 1987). Accordingly, it is plausible that frequency of forgetting estimates are based upon individuals accessing their memory self-emcacy beliefs and then converting these beliefs into a frequency estimate. If so, then the surface differences between
M e t a m o r y in Adulthood
177
memory complaints and memory capacity scales may actually have relatively little functional impact on individual differences in self-ratings of memory functioning. A major implication of emphasizing the memory self-efficacy construct is that certain assumptions about criteria for evaluating the validity of metamemory questionnaires are called into question. Two chief criteria cited by Gilewski and Zelinski (1986) for validation of metamemory scales are (a) predictive validity for memory performance, and (b) sensitivity to age differences. If one conceptualizes frequency of forgetting estimates as reflections of memory self-efficacy beliefs, then the issue of predictive validity for memory task performance becomes a theoretical question of the accuracy of the memory beliefs rather than a basis for judging the construct validity of the scale. Predictive validity for memory performance, then, is cast rather differently. It becomes solely a question of the empirical utility of the questionnaire measure a s a predictor of performance, and by implication, a method for determining the accuracy of subjective memory assessments. Similar issues arise regarding the criterion of age sensitivity. Gilewski and Zelinski ( 1986) criticized several metamemory questionnaires for their lack of sensitivity to age differences. Their implicit premise seems to be that, given age dflerences in actual memory functioning, a valid measure of metamemory ought to show age dflerences. This assertion presumes that memory selfefficacy beliefs are accurate reflections of memory functioning. However, a measure of memory self-efficacy beliefs may show no age differences if a sample of older individuals does not perceive itself a s having low memory self-efficacy, irrespective of the accuracy of those beliefs. Conversely, to the extent that older persons' memory self-efficacy beliefs are low, or they perceive a high degree of decline in memory from young adulthood, then a valid memory beliefs questionnaire should reflect such perceptions even if true memory ability is not impaired or has not changed. The memory beliefs metaphor, then, underscores the need for a further work investigating the construct validity of metamemory questionnaires (Dixon, 1989; Gilewski 8t Zelinski, 1986). This construct validation process requires greater attention to both the convergent and discriminant validity of metamemory
178
H e r t z q , Dtxon & Hultsch
instruments. Dflerent scales may converge to measure the same construct (e.g., memory self-emcacy or knowledge about memory functioning]. Nevertheless, convergent validity of measures of multiple aspects of metamemory cannot be assumed, but rather, must be empirically investigated. Discriminant validity of metamemory scales must be demonstrated relative to measures of other, related construct dimensions such as self concept and personality, as well as to other variables related to questionnaire responding (e.g., affective states: Dixon & Hertzog. 1988;Hultsch et al., 1985). The concern of broadening the scope of construct validation efforts (Cronbach & Meehl, 1956;Messick, 1981)was a primary motive for the studies we describe below. The Annville and Victoria Validation Studies We designed the Annville and Victoria validation studies to address the construct validity of the MIA and the MFQ, a s well as to address other empirical issues involving metamemory. Both studies involved assessment of cross-sectional samples of adults, using sets of core measures (for more details see Hertzog, Hultsch, & Dixon. 1989:Hertzog, Dixon, & Hultsch, 1990.in press: Hultsch et al., 1988). Both studies included the two metamemory questionnares, measures of self-reported affective states and depression, locus of control, general self-efficacy, a vocabulary test, and two recall tasks: free recall of words and narrative texts. Both studies administered these materials in two sessions, with the metamemory questionnaires and affective status measures in the first session, and the vocabulary test and memory tasks in the second session. There were also some dflerences in the design of the two studies. The Annville sample was a cross-sectional sample of 447 persons drawn from a family medical practice in Annville, Pennsylvania, and included individuals ranging in age from 20 to 78. The Victoria sample included a group of 100 young University of Victoria students and 278 adult volunteers, ranging in age from 55 to 80. The Annville study included a task prediction component, the Victoria study did not. Finally, the Annville study incorporated a two-year longitudinal retest in order to assess the stability of memory performance and metamemory.
Metamemory In Adulthood
179
Construct Validation of Two Metamemory Questionnaires
Convergent Vdidity of the MIA and MFQ Hertzog et al. (1989)reported a series of analyses demonstrating good convergence between the MIA and the M F Q in the measurement of memory self-efficacy. perceived change in MSE, These and use of memory strategies in everyday life. questionnaires each contain one principal scale we hypothesized to measure MSE: the Capacity scale from the MIA and Frequency of Forgetting Scale from the MFQ. MIA Capacity requires Likert scale ratings of questions like, "I am good at remembering names." The MFQ Frequency of Forgetting scale requires frequency judgements on how often one forgets specific types of information (e.g., names, appointments). In addition, the MFQ contained additional measures that had already shown factorial convergence with Frequency of Forgetting (Gilewksi & Zelinski. 1986). including a general rating of memory problems. Our previous factor analysis of the MIA had indicated that three other MIA scales also loaded with Capacity on a memory self-emcacy factor: Locus (perceived control over memory), Change (perceived change in memory capacity). and Anxiety (anxiety about memory). We hypothesized that the MFQ memory problems factor and the MIA self-efficacy factors are, in fact, alternative measures of the memory self-efficacy construct, and that therefore they would have a perfect factor correlation of 1.0 (see also Hultsch et al.. 1988). We also predicted that the measures of perceived change and memory strategies from each questionnaire would load on the same factors, an alternative method for demonstrating convergent validity. The results of a confirmatory factor analysis did demonstrate convergent validity for these three subsets of metamemory scales. In particular, there was a high degree of convergence between the MSE factor measured by MIA scales and the MSE factor measured by MFQ scales. However, although the estimated factor correlation for these two MSE factors was very high (just greater than .9 averaged over samples), the hypothesis of perfect convergence (a 1.O correlation) was rejected. We also conducted a simultaneous factor analysis on multiple age groups, and demonstrated equivalent
180
Hertzog, DLwon & Hultsch
factor loadings on the MSE factors at all age levels. Indeed, all factor loadings appeared to be equivalent across age groups. There was one indication of an age dinerence in the factor structure: the Change factor correlated more highly with MSE in the older groups, relative to the University of Victoria student group. Other memory problems questionnaires besides the MFQ appear to have convergent validity with the MIA, consistent with our arguments that MSE drives responding on both types of questionnaires. Cavanaugh and Poon (1989) reported sizable correlations between the multiple, domain-specific memory problems subscales of the SIME and the Capacity and Change subscales of the MIA. The results of the Hertzog et al. (1989)analysis also underscored the fundamental difference between memory knowledge. selfreported strategy use. and MSE. The MSE and Change factors had very low correlations with the Strategy factor and the MIA Task scale, which measures knowledge about memory processes, replicating results from the original MIA validation samples (Dixon & Hultsch, 1983b; Hertzog et al., 1987). Discriminant Validity of Metamernory Factors
Metmemory scales do show some significant correlations with measures of personality, depression, affective states, and general measures of self-efficacy and locus of control (Broadbent, Cooper. Fitzgerald, & Parkes, 1982; Hultsch et al., 1988). However, these correlations are sufficiently small to support the argument that MSE and other metamemory factors are related to, but distinct from, constructs such as extraversion, neuroticism, depressive affect, and general self'-efficacy beliefs. Given that zero-order correlations between scales are attenuated due to measurement error, it is important to test hypotheses of discriminant validity with latent variable models in which estimated covariances among latent variables are disattenuated for measurement error (Judd, Jessor. & Donovan, 1986). Hertzog et al. (1990) conducted a series of structural regression models which demonstrated discriminant validity of metamemory scales from the MIA and MFQ within a latent
Metamemory In Adulthood
181
variable framework. A factor measuring general self-efficacy and control beliefs (as measured by the Paulhus and Christie (1981) spheres of control and the Levenson (1974)internal locus of control scales) correlated about .5 with the MSE factor, but other factor correlations were considerably lower. The MIA Anxiety scale was shown to load both on MSE and on a trait anxiety factor. We interpreted both these findings within a general/domain specific differentiation of beliefs (e.g.. Lachman. 1986). MSE can be considered self-efflcacy beliefs in the domain of memory, related to but distinct from general self-efficacy beliefs: likewise, anxiety about memory is influenced both by general dispositions toward anxiety and the domain-specific influence of memory self-efficacy beliefs (Bandura, 1988). The significant relationships were therf'ore theoretically plausible, and the magnitude of the factor covariance points to related but differentiable underlying constructs. Another interesting finding of the Hertzog et al. (19901 discriminant validity analysis was that the MFQ and MIA MSE factors had differential correlations with Neuroticism and Psychological Distress, as measured by scales from the Jackson Personality Inventory (Jackson, 1976). the Mental Health Inventory (MHI, Veit & Ware, 1983)and the CES-D depression scale (Radloff. 1977). The MFQ MSE factor correlated significantly more highly with these dimensions than did the MIA MSE factor, in spite of the .9 factor correlation between the two MSE factors. Indeed, we were able to force both MIA and MFQ indicators to load on the same MSE factor without a loss of statistical fit when the MFQ scales also loaded on the Neuroticism and Psychological Distress factors. The latter factor was estimated by scales measuring depression, anxiety, and subjective well-being from the CES-D and the MHI. Hertzog et al. (1990)concluded that the slight diirerences between the two scales could therefore be attributed to higher relationships of memory complaints measures to negative affect. The difference in MSE factor correlations with measures of negative affect may help to reconcile discrepant findings regarding age differences in the Frequency of Forgetting and Capacity scales. Hultsch et al. (1987)found significant age diirerences in Capacity, but not Frequency of Forgetting, in the Annville sample. Older persons in that sample report lower frequency of depression and
182
Hertzog, Dtwron & Hdtsch
anxiety symptoms on these scales than younger adults (Hertzog. Van Alstine, Usala, Hultsch, & Dixon. in press). Given that age differences in MSE factor scores are in the opposite direction (Hertzog et al., 1989). with young reporting higher MSE, the lack of age dmerences in Frequency of Forgetting appears to be caused by two offsetting components of variance in Frequency of Forgetting scores: (a) lower self-reported psychological distress and higher self-reported well-being in older adults, coupled with (b) lower MSE. Formulation of a theory accounting for the differential relationship of memory complaint and memory ability questions to trait and state measures of negative affect is an important avenue for the future. In general, then, the results from the construct validation analyses support the thesis that metamemory questionnaires have acceptable validity as well as reliability, if one construes construct validity in terms of consistent patterns of questionnaire responses pointing to constructs related to beliefs about memory. Predictive Validity of Metamemory for Memory Task Performance Hertzog et al. (1990) also examined the predictive validity of metamemory questionnaires for memory task performance. Participants in the validation studies performed three times on two different memory tasks: free recall of words and free recall of narrative texts. Three word lists of 30 nouns were constructed, using six words from each of five taxonomic categories (e.g., metals, fruits). Three narrative stories were constructed, and analyzed into a hierarchical propositional system using the techniques developed by Kintsch and van Dijk (1978). Recall protocols were scored for presence of the gist of each proposition. Lists and texts were presented in the same alternating order to all subjects (see Hertzog et al., 1990,in press, for more details). Table 1 gives selected zero-order correlations between metamemory scales, education, vocabulary and recall of word lists and texts. There were modest, but significant relationships of multiple MIA and MFQ scales to recall performance, with correlations in the .2 to low .3range. Both the questionnaires showed significant correlations with education (see Perlmutter.
Table 1 Metamemory and M m r y Performance Correlations in the Annville Validation Sample Variables
MIA Scales Task Strategy Capacity Change Anxiety Achievement Locus MQ Scales GEN RETRO
FOF Reading Remote
Age
Sex
Educ.
Vocab.
WR1
WR2
WR3
TR1
TR2m
-.09 -.03
.04 .23 .01
.25 .18 .25 .27 -.23 -.05 .05
.27 .14 .21 .15 -.21
.12 .08 .23 .23 -.15
.15
.10
.24 .17 .24 .21
.22 .18 .21 .17
-.09
-.06 .03
-.02
.21 -.01 .30 .26 .20
.12 -.07 .29 .25 .20
.13
.ll -.04 .27 .20 .15
-.17 -.36 .01 -.01 -.14
-.09
.21 .07
.oo
-.05 -.12 -.04
-.07
-.09
-.11 .05
-.02
.oo
-.07
- .02
-.03
.27 .21 .18
.08
.09
.25 .23
.24 .26 -.13 -.04
-.09
.oo
- .06 .03
.29 .19 .27 .29
-.03 .06
- .09 .02 .06
.oo
-.01
.01
.13 -.01 .28 .20 .16
.10 -.07 .23 .24 .17
-.05 .17 .19 .18
.12
.03
-.03
Nde: AU correlations based upon pairwise deletion for missing data. N ranges from 418 to 447. Given the large sample size. significancelevels are not provided. Abbreviations: W R -- Word Recall; TR -- Text Recall; GEN -- General Rating: REI'RO -- Retrospective: FOF -- Frequency of Forgetting
.24 .24 .21
184
Hertzog, D h n & Hultsch
1978, for similar findings) and vocabulary a s well as recall performance. The largest correlations were for measures of MSE: the MFQ Frequency of Forgetting and the MIA Capacity and Change scales. These two scales had roughly equivalent magnitudes of correlation. The Frequency of Forgetting and Capacity scales did not differ in the magnitude of their correlations with text and word recall. This finding was somewhat surprising to us. in that there were more frequent reports of memory beliefs scales correlating with text recall than word list recall (e.g., Dixon & Hultsch, 1983a; Sunderland et al.. 1986; Zelinski et al.. 1980). Indeed, studies failing to detect significant correlations of memory tasks with metamemory scales in older samples often involve some type of recall or recognition of words (e.g., Cavanaugh & Murphy, 1986; West et al., 1984). The equivalent correlations for text and word list recall in our study are, however, consistent with the recent findings of Cavanaugh and Poon (1989). We analyzed relationships among metamemory scales and memory task performance by means of structural regression models, which disattenuated the predictive validity coefficients for measurement errors in the questionnaires, and which estimated the relationships at the level of the latent variables (e.g., MSE). The disattenuated validity coefficients were higher. Although there was a significant relationship between the MSE factor and the word recall and text recall factors, the analyses pointed to two interesting failures to obtain specijkity in predictive validity. First, self-reported problems with remembering information read in magazines, newspapers, and books (as measured by the MFQ's Frequency of Forgetting, Reading subscale) did not covary more highly with text recall than the more general MSE factor, and there was therefore no evidence of significant residual covariance of these measures with text recall, partialling for the general MSE factor. Second, strategy use in everyday life, as measured by the MIA and MFQ, did not correlate with the adjusted ratio of clustering (ARC; Roenker, Thompson, & Brown, 1971). The ARC is a measure of the clustering of recalled words by semantic category. Although it is a measure of clustering at output, it is commonly thought to be highly influenced by categorical clustering
Metarnemoy in Adulthood
185
strategies during encoding. Moreover, the sample correlation of the clustering measure with the self-reported mnemonics subscale of the overall MIA Strategy scale (Dixon & Hultsch. 1983b) was slightly lower than its correlation with the subscale measuring self-reported use of external memory aids, even though clustering during acquisition is a mnemonic strategy. In general, the Hertzog et al. (1990) results support the conclusion that MSE and other metamemory factors have significant. but modest relationships to actual memory performance. It appears, then, that individual differences in MSE beliefs are not necessarily valid reflections of relative memory ability as measured by experimental tasks (see also Herrmann. in press). Moreover, the lack of specificity in predictive validity to text recall and the experimental measure of strategy use in free recall appears to constrain inferences based upon such questionnaire measures (see below). Relations of Memory Self-Efficacy and Memory Task Predictions Another major goal of the Annville validation study was to examine the relationship of memory self-efficacy, a s measured by metamemory questionnaires, to memory task predictions. As noted above, memory performance predictions have often been conceptualized as reflections of knowledge about memory processes and the current state of one's memory system. However, the memory beliefs metaphor suggests a radically different way of viewing predictions. From the beliefs perspective, a key to appreciating the linkage between memory self-efficacy, a s measured by the MIA and MFQ, and memory task performance predictions may be t o distinguish general or global memory selfefficacy beliefs from local efficacy judgements in a particular context (i.e.. a particular memory task, testing environment. and the concurrent physiological and psychological state of the rememberer; Dixon & Hertzog. 1988). Given that memory selfefficacy is probably a highly schematized system of beliefs regarding one's ability to utilize multiple types of memory in a variety of contexts, it is reasonable to presume that stored propositions about self-as-rememberer are hierarchically
186
H e r t z q , D h n & Hultsch
dmerentiable on a global to local continuum. One may have global beliefs about self-as-rememberer (e.g., "I never forget a face"), but also have local beliefs about ability to remember specific types of information in speciflc contexts (e.g.. "I am good at remembering new names and faces at conventions"). Questionnaires like the MIA and MFQ measure memory self-efficacy either (a) by a general rating question, which is obviously more associated with global memory self-efficacy, or (b) by aggregating ratings of memory capacity and forgetting across a variety of specific memory functions (e.g., remembering names. faces, telephone numbers). The items are also generalized, in that they are also divorced from a speciflc temporal and spatial context. It seems reasonable, then, to expect that questionnaire measures are more related to global memory self-efflcacy beliefs. Performance predictions, on the other hand, may be conceptualized as self-efficacy judgements that are specific to the immediate memory-demanding context (Cavanaugh, et al.. 1989: Beny et al.. 1989:Lachman & Jelalian, 1984). As such, predictions are more closely related to the operationalization of self-efficacy employed by Bandura (1986)in other domains. In that sense, requiring subjects to predict overall task performance (as opposed to the type of item-by-item predictions used in some paradigms, e.g., Lovelace & Marsh, 1985)is likely to be influenced by both global self-efficacy beliefs and local self-efficacy beliefs. which may be activated in part on the basis of similarity of the current memorydemanding context to memory-demanding contexts previously experienced by the subject. Although most prediction studies have simply requested quantitative predictions in terms of the natural stimulus units employed (e.g., number of words). more elaborate scaling of the predictions is possible. The MSEQ (Berry et al.. 1989)requires a yes/no judgment (with confidence ratings) regarding whether one will be able to achieve certain performance levels on laboratory and everyday memory tasks. Multiple performance levels are given for each task, and the responses are then aggregated over levels to yield a self-efficacy judgment scaled in Likert rating scale units rather than units recalled. It seems clear that relatively global measures of memory self-efficacy, as measured by questionnaires
M e t a m m r y in Adulthood
187
like the MIA and MFQ, are at least superficially distinct from taskspecific efficacy judgements a s obtained by the MSEQ (which is, in essence, an elaborate performance prediction paradigm). However, to the extent that responses on the MSEQ are based upon relatively global self-efficacy beliefs, such as in the case where individuals have little relevant experience on a speciflc task, the two types of questionnaires may yield very similar results. We propose that a performance prediction is based upon (a) one's global and local memory self-efficacy. (b) a n appraisal of the memory task, and (c) an as yet unspecified set of processes translating one's memory self-efficacy into a prediction by using a representation of the distribution of task performance derived from the task assessment (see also Cavanaugh et al., 1989). A performance prediction, then, will be based upon a generalized memory self-efficacy belief in the absence of speciflc experience in the memory domain assessed by the task, or upon more specific beliefs about one's memory self-efflcacy In familiar situations. In either case, however, the belief system must be combined with the task appraisal to produce a performance estimate. This conceptualization of beliefs and efficacy judgements has important implications for explaining prediction accuracy. It suggests three classes of possible reasons for inaccurate predictions: (a) inaccurate memory self-efficacy. at either the global or local level; (b) inaccurate appraisal of the memory task and, by implication, inaccurate representation of the distribution of task performance: and (c) faulty mapping of memory self-efficacy onto the subject's representation of the performance distribution. Thus an overestimate of performance by an older person might reflect neither overconfidence in memory ability (inappropriately high memory self-efficacy) nor poor memory monitoring skills. Instead, an overestimate might reflect inaccurate task appraisal -for example, an inaccurate assessment of how well the average individual performs on that memory task. This latter possibility seems especially likely for complex laboratory memory tasks that are unfamiliar to the subjects. and is consistent with the studies reviewed above (e.g., Shaw & Craik, 1989) showing predictions to be insensitive to experimental manipulations known to affect memory task performance.
188
Hertz*,
D h n & Hultsch
Hertzog, Dixon, and Hultsch (in press) assessed memory selfefficacy/performance prediction relationships in a subset of the subjects from the Annville validation sample. Given our expectation of multiple influences on memory predictions, we hypothesized that correlations between questionnaire measures of memory self-efficacy and performance predictions in the standard prediction paradigm would be attenuated by faulty task appraisal. Conversely, the correlation between general memory self-efficacy beliefs and predictions would be maximized when individuals already possess, or are explicitly given, information about normative levels of task performance. Moreover, given task experience, individuals should utilize on-line awareness of performance to form task-specific performance evaluations and self-efficacy beliefs (Cavanaugh et al., 1989; Cavanaugh & Perlmutter, 1982). This process should improve prediction accuracy over recall trials (Herrmann et al., 1986;Lachman et al.. 1987)but produce lower correlations of performance predictions with questionnaires influenced by more global memory selfefficacy beliefs. Roughly half the groups of subjects had been randomly assigned to a performance prediction condition. In order to minimize the influence of individual differences in task assessments on memory task predictions, subjects in the prediction condition were given prior information about average performance levels on the task. They were given a brief description of the task and then told: "Before you actually do the task described above, we would like to ask you to tell us how well you think you will do. It may help you to know that on this task the average person is able to remember about....". They were told that average performance on the categorized word recall task is 15 out of 30 words, and that the average performance on text recall is 25 out of 50 ideas. These values were not based on empirical or theoretical estimates of expected performance levels, but were instead arbitrarily selected to anchor predictions in the middle of the range of possible responses. Figure 1 plots the word recall predictions. as well as the average recall performance (pooled over trials). As can be seen. the older and middle-aged subjects predicted lower performance levels than younger subjects on the first recall trial. Actual mean performance
Metamemory InAdulthood
189
21 20
i
Ig
h 18
s
16
8
15
8 17
t
14
13
'
I
1
I
2 Trials
I
I
3
Average
Performance
mure 1. Word list predictions and performance for 3 age groups (from Hertzog, Dixon, & Hultsch. in press).
was higher than the predictions, as well as the normative performance levels provided. On subsequent trials. all three age groups increased their performance predictions, but the age difllerences in the amount of increase were not significant. in spite of the fact that younger adults' performance levels were considerably higher than initially predicted. Moreover, all three age groups showed comparable upgrading in the accuracy of predictions. Similar initial age differences were found for text recall predictions, with old subjects predicting poorer performance than young subjects. However, age group means in text recall predictions did not change as much as word recall predictions. The apparent difference between text and word recall in the degree of prediction upgrading' over trials was even more obvious when the correlations of predictions and task performance were examined. There was a modest correlation of first word recall task with first word recall prediction (.24), but subsequent correlations between predictions and word recall performance increased
190
H e r t z q , Dtwon & Hultsch
dramatically. The highest correlations were associated with the immediately prior task performance with the subsequent prediction. For example, List 3 prediction and List 2 performance produced the highest correlation of all (.71). The initial correlation of text recall prediction and text recall task performance was .44. higher than the initial word prediction/recall correlations, but the text prediction/recall correlations increased less across trials, reaching a maximum of .58 at the third text recall trial. The Capacity, Change, and Frequency of Forgetting scales did correlate significantly with memory performance predictions (Table 2). These scales all relate to the higher-order M S E factor identified by Hertzog et al. (1989). In contrast, the correlations of other metamemory scales with performance predictions were smaller. This finding suggested that predictions were related primarily to MSE, with significant but weaker relationships to Strategy and Task. The correlations between predictions and MSE measures were largest for the flrst prediction, but then appeared to decline slightly across trials. In order to better address the relationship of MSE to memory performance and recall task predictions, we developed a structural regression model for the data which reflected both (a) the issues raised above regarding self-emcacy beliefs and predictions and (b) a set of causal hypotheses, framed within a latent variable framework, regarding relationships between predictions, MSE, memory ability. and memory task performance. First, we assumed no causal relation between trials of recall performance, instead assuming that recall of each word list reflects an underlying latent ability for free recall of words from a categorized list, and that each trial of text recall reflected a latent text recall ability. Second, consideration of likely causal mechanisms suggested that underlying memory ability, as reflected in these latent text and word recall variables, has a direct causal influence on (global) memory self-efficacy beliefs. Third, these global memory selfefficacy beliefs directly influence memory task performance predictions. The model also reflected a specific hypothesis regarding the upgrading of performance predictions across trials: the proximal cause of the predictions in the later trials of a multitrial design is
Metamemory In Adulthood
191
Table 2
Correlations Between Metamemory Scales and Recall Predictions for Word Lists and Texts. ~
Predictions List1
Text1
List2
Text2
List3
Text3
Strategy
.06
.16*
.17*
.18*
.25**
.13
Task
.14*
.20**
.07
.22**
.14*
.17*
Locus
.14*
.06
-.04
.01
-.01
-.07
Achievement
.03
.08
-.03
.09
.o 1
.02
Anxiety
-.25**
-.14
-.lo
-.11
-.08
-.10
Capacity
.39**
.37**
.20**
.32**
.33**
.27**
Change
.41**
,38**
.25**
.31**
.27**
.26**
kequency of Forgetting
.36** .30**
.24**
.26**
.27**
.2W*
Note: Correlations based on a subsample of the Annville Validation sample asked to provide performance predictions (N=l57). Adapted from Hertzog, Hultsch, & Dixon (in press).
*
p<.05 p<.Ol
not the latent ability to recall text or words per se. but rather, the actual performance on the prior recall trial. One evaluates the perf'omance on the current task, and revises subsequent estimates accordingly. This perspective suggested that (a) latent memory ability determines MSE and, indirectly, initial performance predictions, but that (b) actual performance on a particular recall trial determines changes in predictions between trials.
192
Hertzog, Dixon & Hultsch
Flgure 2. Causal model for latent memory and memory self-emcacy. with dynamic prediction/memory relations.
Figure 2 depicts the implications of this conceptualization, with reference to a single latent memory factor. The latent memory factor determines the observed recall performances ( M l , M2, and M3). The latent memory factor also influences MSE. as measured by the metamemory scales. In turn, MSE influences the first prediction, with subsequent predictions influenced by prior predictions and by the prior recall performance. The broken line between the latent memory variable and the first prediction represents the issue of domain specificity in self-efficacy beliefs. Is it the case that the context-specific self-efficacy judgment, represented in the performance prediction, is influenced by latent ability independent of the mediating (direct) effect of general memory self-efficacy beliefs? The model we actually attempted to fit was more complex than that shown in Figure 2, primarily because it simultaneously modeled prediction and performance variables for both word and text recall. A n initial important outcome of attempting to fit the model to the sample data was that we found it necessary to use a higher-order verbal memory factor, determining both text and word recall, a s the latent variable mediating the memory ability/memory self-efficacy relationship. Figure 3 graphs the
Metamemory in Adulthood
193
standardized factor loadings and structural regression coefficients for the accepted model. The model has several noteworthy features. First, the memory latent variables were well-defined, as shown by the high loadings of the recall measures on the text recall and word recall factors, and, in turn, the high loadings of these factors on the second-order verbal memory factor, Second, the MSE factor also had strong relationships to the MIA Capacity and MFQ Frequency of Forgetting scales. Third, the model successfully represented the modest correlation between the first word recall performance and the first word recall prediction a s being actually mediated by a salient path from verbal memory to MSE and, in turn, a salient path from MSE to the first word recall prediction. Fourth, the relationship of MSE to the first text recall prediction, although statistically significant, was relatively weak. Instead, there was a direct effect of text recall to the first text prediction. Fifth, there was evidence of upgrading of perfonnance predictions for both word recall and text recall on the basis of past perfonnance. This effect was definitely more pronounced for word recall, a s evidenced by (a) the lower autoregressive coefficients for word recall predictions and (b) the larger lagged coefficients from prior word recall performance to the next word recall prediction. Finally, sex (but not age) had a direct effect on MSE independent of verbal memory ability, with females demonstrating lower MSE than males. On the other hand, all covariances of performance predictions with age and sex were successfully represented as being mediated by MSE and text recall ability. What should one conclude from these analyses? First, it is apparent that memory self-efficacy is related to task-specific performance predictions. When adults predict recall performance after being told normative performance levels, older adults predict poorer recall performance than younger adults. This pattern of mean age differences in predictions is similar to the patterns of age differences seen for questionnaire measures of memory selfefficacy (e.g.. Hultsch et al., 1987). Moreover, the zero-order correlations and structural equation models also showed a salient relationship of the MSE factor to the h t word recall performance prediction.
mure 3. Structural regression model for r e d performance and predictions (standardized solution: from Hertzog, Dlxon. & Hultsch. in press).
Metamemory in Adulthood
195
Second, the salient, but moderate, relationship between the two latent variables supports the conclusion that memory self-efficacy beliefs may be based upon actual ability to remember, but are not necessarily veridic a1. A third important outcome of the study was the salient influence of text recall ability to text performance predictions, independent of the relationship of memory self-efficacy to text predictions. This specific relationship is consistent with the hypothesis that individuals have more implicit knowledge about their text recall abilities than their word recall abilities (e.g.. Dixon, 1989). It is curious, then, that text recall does not correlate higher than word recall with a domain-specific measure of selfrated memory for information gathered from printed materials taken from the MFQ (Hertzog et al., 1990). The discrepancy could be explained by arguing either (a) implicit knowledge about text recall skills are better represented in the prediction than the questionnaire scale, or (b) the differential correlation was influenced by the order of recall task administration (word recall, then text recall). We attempted to model effects of word recall performance on text recall predictions, without success, but resolution of the discrepancy will require new data from a different design. The fact that word recall prediction accuracy increased much more than text recall prediction accuracy was quite striking. Predictions of word recall performance showed strong upgrading in degree of accuracy across trials, consistent with other studies (Herrmann et al.. 1986; Lachman et al.. 1987). This relationship was best modeled as a lagged effect of word recall performance on subsequent performance predictions. The most plausible explanation for pattern of results is that it is easier for individuals to monitor their own performance levels on the word recall task. This accuracy in memory monitoring could be due to on-line awareness of performance (Cavanaugh et al.. 1989). or due to some kind of post-recall performance evaluation. The reduced level of upgrading for text recall predictions probably reflects the greater dimculty in gauging how well one is doing on that task.
196
Hertzcg, Dtwon & Hultsch
Longitudinal Data on the Stability of MSE An important issue regarding memory self-efficacy beliefs
is
the degree to which they are malleable and highly influenced by successes and failures in memory functioning that occur in everyday life (Dixon & Hertzog, 1988;Hultsch et al., 1985). Dixon and Hertzog (1988)suggested that memory self-ratings might behave more like measures of affective states than personality traits, and hence be influenced by characteristics of the specific context in which the memory ratings are made. Broadbent et al. (1982)noted that there are actually advantages of instability in individual differences in memory complaints and other cognitive attributes, provided that changes validly reflect transient changes in the effective memory functioning of the individuals. To be effective in clinical memory screening, subjective memory complaints measures should be sensitive to current levels of memory functioning, even if the changes are relatively transient. However, Broadbent et al. (1982)found test-retest correlations of about .8 (over either 21 or 65 week intervals in two different samples) in their Cognitive Failures Questionnaire. including its memory complaints subscale. Recently, Hertzog, Saylor, Mobley, Hultsch, and Dfxon (1989) reported preliminary data from a two-year followup study of the Annville Validation sample. There was no evidence of longitudinal change in either mean levels of MSE, as measured by Frequency of Forgetting and Capacity, or perceived change, a s measured by the MIA Change scale in any age group. We also found a high degree of stability of individual differences in metamemory scale scores, especially Capacity and Change, with test-retest correlations exceeding .8 Over the two-year interval. It therefore appears that individual differences in questionnaire measures of M S E are highly stable, consistent with the idea that the questionnaires are tapping into a stable system of self-beliefs. Although this finding does not rule out fluctuation in perceived memory functioning, it does suggest that, contrary to the speculations of Dfxon and Hertzog (1988).the types of global memory self-efficacy measured by the MIA do not resemble affective states in their intraindividual variability. Indeed, Usala and Hertzog (1989) examined stability of
Metamemoty in Adulthood
197
affect measures in the same sample. The test-retest correlation for a measure of trait anxiety was also quite high. Test-retest correlations were significantly lower (approximately .5) for measures of state anxiety. Taken together, this pattern of correlations indicates that memory self-emcacy measures are more trait-like than state-like in terms of stability of individual differences.
THEORETICAL SYNTHESIS AND CONCLUSIONS In order to place our research in context, it is necessary to
return to the two metaphors that have guided research on metamemory in adulthood -- knowledge and beliefs. As we have seen, our results are consistent with the view of Sehulster (1981) and others that responses to metamemory questionnaires often access a system of self-referent beliefs individuals have about their memory functioning. These beliefs are associated with -- and probably influence -- performance predictions in ways not fully appreciated by the researchers who first used the prediction paradigm as a method of studying individuals' awareness of their memory system and its functions. Moreover, it seems apparent that general beliefs about memory self-efficacy are related to local memory self-efficacy beliefs, as manifested in task-specific performance predictions and evaluations. It is likely however. that the degree of relationship between more schematic, generalized propositions about self-as-rememberer and local self-efficacy beliefs is reduced by the degree of relevant experience one has in using memory in a particular context. Thus, as with the free recall task, it seems plausible that subjects who are unfamiliar with a memory task will initially give more weight to global self-efficacy beliefs, but wffl elaborate local self-efficacy beliefs on the basis of a performance appraisal which is necessarily colored by but not isomorphic with global memory self-efficacy. Certainly it is the case that both the knowledge and beliefs metaphors are appropriate for different aspects of the domain of constructs subsumed under the label of metamemory. In the context of the perfonnance prediction paradigm, we have argued that the initial task appraisal is a crucial part of the prediction
198
Hertzqq. DLmn & Hultsch
generation process, and it seems likely that knowledge about memory tasks and their demands play a n important part in determining the prediction and its accuracy. The meaning of inaccurate predictions, however -- a finding emphasized in the literature -- is called into question when one appreciates that the prediction requires both an assessment of the task, in terms of performance distributions. a s well a s a n evaluation of oneself (based in all likelihood upon MSE and other attributes). Can predictions then be taken as a measure of self-knowledge or general knowledge about memory? Only if we make some rather strong assumptions about how knowledge about memory relates to task appraisal, or even if any kind of rational task appraisal takes place at the time the prediction judgment is computed. The notion that task predictions are a measure of self-knowledge, a s in the old metaphor of metamemory, seems in retrospect rather simplistic and inaccurate. Of course, one could argue that beliefs are a better metaphor for task appraisal, and that one should differentiate subjects' implicit theory about task environments from their implicit self-theory about memory. A meaningful treatise on knowledge and belief would take us quickly into some epistemologically deep water, and is certainly beyond the scope of this chapter. Nevertheless, two things seem clear. First, general knowledge about how memory operates probably has relatively little relationship to global or local memory self-efficacy beliefs. This argument is supported by the low correlation of the MIA Task scale with the M S E factor. Second, metamemory researchers often make unwarranted assumptions about the knowledge individual subjects have about memory processing requirements of memory tasks, and then assume validity of operational definitions in the absence of compelling theoretical and empirical support. Indeed, it seems in general that we actually know very little about the structure and nature of memory self-efficacy, and how it influences behavior in memory-demanding situations. The longitudinal data on questionnaire measures of M S E suggests that these beliefs are relatively stable over a two-year period, at the level of both means and individual differences. At this point we do not know very much about the relative differentiation of memory self-
Metamemory in Adulthood
199
efficacy beliefs across difllerent contexts and processes, nor do we understand the structure of the belief system and the way aging affects mechanisms relating to access and evaluation of propositions about self stored in memory. Like Sehulster ( 1981). we believe that framing memory beliefs in terms of a self-theory inevitably leads to application of methods and principles derived from the growing social cognition literature. Recent work on self-theories (e.g.. Markus & Wurf. 1987)may help us understand the dynamics of memory beliefs and their influences on memory-related behavior (see also Cavanaugh, Feldman, & Hertzog. 1990). As reviewed in detail by Markus and Wurf (19871, current conceptualizations of self-concepts emphasize both the multidimensional nature of the self-concept and its dynamic nature. Conceptions of self are schematic and complex. and the particular propositions about self that are retrieved in a particular context are subject to a host of influences (Srull & Wyer, 1989). Thus, important topics for further research on memory selfefficacy include domain-specificity of beliefs (and differential accuracy of beliefs across domains), the question of whether global beliefs, a s identified by standard questionnaires, accurately reflect local beliefs that may be more proximal influences of actual behaviors, and the relationship of global and local self-efficacy beliefs to memory-related behaviors such as strategy utilization. A crucial issue for understanding global memory self-efflcacy beliefs is the extent to which they are derived from inferences about actual memory-related experiences or internalization of stereotyped aspects of aging. As Perlmutter (1978)and others anticipated, a key to understanding the self-theory about memory may be the implicit theories people have about the effects of aging on memory (Langer, 1981; Perlmutter et al., 1987; Person & Wellman, 1988;Ryan, 1989).To the extent that individuals believe that decline in functioning is impending, they may internalize the stereotype of decline and subsequently attribute memory failures to age-related loss of effective functioning. To be sure, current theories of self-concept would not consider these two types of influences to be incompatible. Evaluations of experiences are inherently grounded in the self-theory of the individual, and experiences of memory failures will be interpreted in light of the
200
Hertzog, Dimn & Hultsch
implicit theory the individual has adopted regarding the possible interpretations of forgetting and other lapses of memory functioning (Person & Wellman, 1988). Nevertheless, it is reasonable to ask about the extent to which beliefs identified by metamemory questionnaires are influenced by biased interpretations of everyday events determined by schematic beliefs about memory and its age-related changes. For example, is the robust phenomenon of perceived change in memory by older persons a function of actual memory loss as opposed to acceptance of the self-relevance of a widely shared view on the effects of aging on memory (e.g., Langer, 198113 One advantage of the self-efficacy beliefs perspective on memory self-theories is that it helps to explain why there may be age-related increases in self'-reported anxiety about memory, as well a s performance anxiety in laboratory testing situations (Bandura. 1989). Individuals with low MSE may experience anxiety and apprehension conceming test performance, especially if they have volunteered for the study because they are concerned that their memory is indeed declining, and are seeking reassurance that it is not. However, given wide individual differences, even among older persons, regarding self-efficacy beliefs, it becomes plausible that failures in the earlier gerontological literature to consistently detect "overarousal" by older subjects (see the review by Kausler, 1982) were because of the unmeasured moderating effects of memory self-efficacy on the age/ test anxiety relationship. Moreover, MSE may prove to be an important mediating construct that explains low task-specific motivation, negative attributions about poor performance at early stages of skill acquisition, and the like (e.g.. Bandura, 1989; Person & Wellman, 1988). Unlike Bandura (1989). however, we do not necessarily endorse the idea that self-efficacy theory is the principal means by which one can account for age differences in strategic behaviors in memory demanding situations. and ultimately, age differences in memory task performance. To be sure, this is a largely untested empirical question, and one worth investigating. It cannot be tested by merely computing correlations of measures of MSE and strategic behaviors (e.g., study time). Our research found that MSE actually correlates weakly with both self-reported strategy use and
Metamemory in Adulthood
201
strategic behavior, as measured by ARC scores, in our free recall task. The latter finding might be discounted by arguing against ARC scores as measures of organizational strategies. Nevertheless, even when observed correlations between strategic behavior and MSE are significant (e.g.. Berry, 1987).one cannot in general assume that those who do not engage in optimal strategies are manifesting a strategic production deficiency induced by low memory self-efficacy. A n account of age differences in strategic behavior during memory task performance requires identification of the complex interrelationships between memory beliefs, epistemic awareness of the state of the memory system, and formulation and use of strategies in memory-demanding situations. Knowledge about past experiences using strategies, as well as self-efficacy beliefs, will in principle influence strategy identification, evaluation, and utilization in cognitive tasks -what has often been defined a s self-regulation in performance settings (Kanfer, 1987). Nevertheless, the strategy utilization literature suggests that strategic behaviors are variable and complex, and that perhaps they should be best viewed as analogous to executive processes in intelligence tests and complex problem solving tasks (e.g.. Borkowski et al.. 1987;Lachman & Lachman, 1980;Sternberg. 1985). Our perspective on self-regulation theory suggests that epistemic awareness of the memory system may be an important influence on monitoring the effectiveness of current strategies and the need to modify strategic approach (Brigham & Pressley. 1988;compare Kanfer. 1987). However, adequacy of the mental model for the task, as constructed from knowledge about the task and its requirements, will necessarily interact with epistemic awareness in influencing memory task behaviors. Thus, although age differences in MSE may influence the probability of optimal strategy selection and utilization in memory tasks, it is only one of many influences. Hypotheses that MSE may explain age differences in acquisition have merit (Bandura, 1989;Berry. 1987) but are undoubtedly incomplete, given that there will be multiple influences on task behaviors by older and younger adults. It is interesting to note that Hertzog, Dixon. and Hultsch (in press) found no direct effect of predictions on recall, independent of
202
H e r t z q , Dba)n & Hultsch
the lagged relationship between past recall and predictions, as might have been expected if self-efficacyjudgments -- as manifested in predictions (Bandura, 1986; Berry et al., 1989) -- negatively influence contiguous recall performance via increased test anxiety. reduced motivation, and the like. Lachman's (1983) and Lachman and Lefs (19891 studies also found a lagged longitudinal influence of fluid intelligence on perceived control over intellectual functioning by elderly adults, but not vice versa, a flnding recently replicated by Grover and Hertzog (1989) in a larger cross-sectional sample. This pattern of results may indicate that cognitive abilities and perfonnance experiences affect beliefs at least as much as self-efficacy beliefs affect performance. Our models therefore treated global MSE beliefs as an outcome, in part, of latent memory ability, and found no additional influence of local MSE, as operationalized by performance predictions, on performance. Instead, performance influenced subsequent predictions. As Bandura (1986) notes, there can be cyclic and reciprocal effects of self-efficacy and performance, and we certainly are not inclined to dismiss the type of MSE-mediated performance suppression effects hypothesized by Bandura (1989) and Berry (1987) on the basis of our limited empirical results. It might be the case that such effects are more readily observed for individuals with very low levels of MSE (or very high levels of negative affect), resulting in a weak degree of association when calculated across the entire distribution of these constructs in a general population. It may also be the case that individuals at highest risk for poor performance, and with lowest prior selfefficacy beliefs, are much less likely to volunteer for memory experiments. Nevertheless, it should be clear that adoption of the self-theory approach to memory beliefs does not necessarily imply endorsement of the hypothesis that MSE can account for strategic deficiencies in adults' memory task performance. It also does not imply enthusiasm for the hypothesis that low MSE is a direct cause of lower levels of memory performance by older adults. Intervention studies designed to improve self-efficacy and enhance self-esteem regarding memory may prove to be beneficial for the subjective well-being of older adults. Whether they wffl materially
Metammory tn Adulthood
203
influence effective memory functioning is another matter (e.g., Rebok & Balcerak, 1989;Weaver & Lachman. 1989). In summary, we anticipate two major new directions in metamemory research: (a) examination of the contents, structure. and processes associated with memory-related aspects of selfconcepts, and (b) determination of the ways in which MSE beliefs of adults are activated in memory task situations and influence taskrelated behaviors. To what extent are strategic production deficiencies in older adults, when observed, influenced by the selfefficacy belief system? How does the self-theory interact with knowledge and beliefs about memory functioning to influence task behaviors? To what extent can production deficiencies be attributed to deficient mental models for task requirements in isolation from self-efficacy beliefs and epistemic awareness of the memory system? Attention to this latter class of issues will require a n integration of the two different traditions in metamemory research (analysis of strategic behaviors and subjective memory assessments), and may well require methods of tapping into the complex influences of task-specific self-efficacy judgements that are fundamentally different from the type of metamemory questionnaires that we and others have developed and validated. ACKNOWLEDGMENTS The research reviewed here was supported by a grant to Hertzog from the National Institute on Aging (RO1-AGO61621 and by a grant to Hultsch from the Social Sciences and Humanities Research Council of Canada. Hertzog was also supported by a Research Career Development Award from NIA (K04-AG00335). Dixon was also supported by a grant from the Natural Sciences and Engineering Research Council of Canada. The cooperation of Robert K. Nielsen, MD, the other physicians, and the members of the Annville Family Practice, Annville, PA, is deeply appreciated. Thanks also to the research assistants who made this work possible, both at Georgia Tech (Paul Usala, Judith Van Alstine. Lori Blank, Laurie Saylor. and Michele Mobley) and at the University of Victoria (Mark Hammer and Barbara Reeves). Correspondence concerning this chapter can be directed to the first author at the School of Psychology, Georgia Institute of Technology, Atlanta, GA, 30332-0 170.
204
Hertzog. D b n & Hdtsch
REFERENCES Bandura, A (1986).Socialfoundation of thought & action: A soclal cognitive theory. Prentice Hall: Englewood Cliffs, NJ. Bandura. A. (1988).Self-efficacy conception of anxiety. Anxiety & s ? w c ~1.77-98. , Bandura, A. (1989).Regulation of cognitive processes through perceived self-efficacy. Developmental Psychology, 25, 729735. Belmont, J. M. (1978).Individual differences in memory: The cases of normal and retarded development. In M. M. Gruneberg & P. Morris (Eds.), Aspects of memory (pp. 153-185).London: Methuen. Berry, J. M. (1987,August), A sew-efficacy model of memory performance. Paper presented at the 95th Annual Meeting of the American Psychological Association, New York, NY. Berry, J. M.. West, R. L., & Dennehy, D. M. (1989).Reliability and validity of the Memory Self-Efficacy Questionnaire (MSEQ). Dewlopmental Psychology. 25. 701-713. BorkowsM, J. G., Cam, M., & Pressley, M. (1987)."Spontaneous" strategy use: Perspectives from metacognitive theory. Intelligence, 11. 61-75. Botwinick, J. (1967).Cognftfve processes in maturity and old age. New York, Springer. Bradburn, N. M., Rips, L. J , & Shevell. S. K. (1987).Answering autobiographical questions: The impact of memory and inference on surveys. Scfence, 236, 157-161. Brigham, M.C. & Pressley, M. (1988).Cognitive monitoring and strategy choice in younger and older adults. Psychology and Aging, 3,249-257. Broadbent, D. E., Cooper, P. F., Fitzgerald, P.. & Parkes, K. R. (1982). The Cognitive Failures Questionnaire (CFQ) and its correlates. British Journal of Cltnnical Psychology, 21, 1-16. Brown, A. L. (1978).Knowing when, where, and how to remember: A problem in metacognition. In R. Glaser (Ed.), Advances in fnstructional psychology (pp. 77-165).Hillsdale, NJ: Erlbaum. Butterfleld, E,C.. Nelson, T. 0.. & Peck, V. (1988).Developmental aspects of the feeling of knowing. Developmental Psychology, 24,654-663. Bruce, P. R., Coyne, A. C., & Botwinick, J. (1982).Adult age differences in metamemory. Journal of Gerontology, 37, 354357.
Metamemay in Adulthood
205
Camp, C. J., Markley, R P. & Kramer, J. J. (1983).Spontaneous use of mnemonics by elderly individuals. E d u c a t f onal Gerontology. 9. 57-71. Carroll, J. D., & Drevenstedt, J. (1987). Metamemoy in adulthood. Age differences and predictive validity. Unpublished Manuscript. Cavanaugh, J. C.. Feldman, J.. & Hertzog, C. (1990). Know thy questionnaire: How people answer rnetamemory questions. Unpublished Manuscript. Cavanaugh, J. C.. Kramer, D. A.. Sinnott, J. D.. Camp, C. J., & Markley, R. P. (1985). On missing links and such: Interfaces between cognitive research and everyday problem solving. Human Deuelopment, 27, 146-168. Cavanaugh, J. C., Morton, K. R., & Tilse, C. S. (1989). A selfevaluation framework for understanding everyday memory aging. In J. D . Sinnott (Ed.), Everyday problem solving: Zkory and application (pp. 266-283). New York Praeger. Cavanaugh, J. C., & Murphy, N. 2. (1986). Personality and metamemory correlates of memory performance in younger and older adults. Educational GerontotogY, 12. 385-394. Cavanaugh, J. C., & Perlmutter. M. (1982). Metamemoxy: A critical examination. Child Development 53. 11-28. Cavanaugh, J. C. & Poon, L. W.(1989). Metamemorial predictors of memory performance in young and old adults. Psychology and A g m , 4,365-368. Chamn R.. & Herrmann, D. J. (1983). Self-reports of memory ability by old and young adults. Human Leamfng, 2, 17-28. Coyne. A. C. (1985). Adult age, presentation time, and memory performance. Expr0nentu.l Aging Research, I I , 147- 149. Craik, F. I. M.. & Rabinowitz, J. M. (1984). Age differences in the acquisition and use of verbal information: A tutorial review. In H. Bouma & D. G. Bouwhuis (Eds.). Attention and Performance X(pp. 471-499). Hillsdale, NJ: Erlbaum. Cronbach, L. J., & Meehl, P. E. (1955). Construct validity in psychological tests. PsychologicalBulletin. 52, 28 1-302. Dixon, R. A. (1989). Questionnaire research on metamemory and aging: Issues of structure & function. In L. W. Poon. D. C. Rubin. & B. A. Wilson (Eds.), Everyday cognitton in adulthood and old age (pp. 394-415). New York Cambridge University Press. Dixon. R. A. & Hertzog, C. (1988). A functional approach to memory and metarnemory development in adulthood. In F. E. Weinert & M. Perlmutter (Eds.), Memoy development: Universal changes and individual differences (pp. 293-330). Hillsdale, NJ: Lawrence Erlbaum Associates.
206
Hertzog, Dtwon & Hultsch
Dixon, R A., & Hultsch, D. F. (1983a).Metamemory and memory for text relationships in adulthood: A cross-validation study. Journal of Gerontology. 38, 689-694. Dixon. R A , & Hultsch, D. F. (1983b).Structure and development of metamemory in adulthood. Journal of Gerontology, 38. 682688. Dixon, R A, & Hultsch. D. F. (1984).The metamemory in adulthood (MIA) instrument. Psychologtcd Documents, 14.3. Dixon, R A , Hultsch, D. F., & Hertzog, C. (1988).The Metamemory In Adulthood (MIA) questionnaire. Psychopharmacology Bulletin, 24,671-688. Eysenck, M. W. (1974). Age differences in incidental learning. Developmental Psychology, 10, 936-941. Flavell (1977). Cognitive development. Englewood Cliffs, NJ: Prentice-Hall. Flavell, J. H., & Wellman, H. M. (1977).Metamemory. In R V. Kail & J. W.Hagen (Eds.), Perspecttoes on the development of memory and cognitfon. Hfflsdale, N J : Erlbaum. Gilewski, M. J., & Zelinski. E. M. (1988). Memory Functioning Questionnaire (MFQ). Psychopharmacology Bullettn, 24, 665670. Gilewski, M. J.,& Zelinski, E. M. (1986). Questionnaire assessment of memory complaints. In L. W. Poon (Ed.), Handbook for cltntcal memory assessment of older adults (pp. 93-107). Washington, DC: American Psychological Association. Gilewski, M. J., Zelinski. E. M., Schaie, K. W., &Thompson. L. W . (1983, August). Abbrevtating the metamemory questtonnatre: Factor structure and norms for adults. Paper presented at the 9 1st annual meeting of the American Psychological Association, Anaheim, CA. Glenberg, A. M., Sanocki, T., Epstein, W., & Morris, C. (1987). Enhancing calibration of comprehension. Journal of Experimental Psychology: General, 1 16, 119-136. Grover, D., & Hertzog, C. (1989). Relationships between tntellectual
control beliefs and psychometric intelligence In adulthood. Unpublished manuscript. Herrmann, D. J. (1982). Know thy memory: The use of questionnaires to assess and study memory. Psychological Bulletin, 92, 434-452. Herrmann, D. J. (in press), Self perceptions of memory performance. In K. W. Schaie (Ed.). Self dlrectedness and
e m m y : Causes and effects throughout the life course.
Metarnemoy in Adulthood
207
Herrmann. D. J., Grubs, L.. Sigmundi. R.. & Grueneich. R (1986). Awareness of memory ability before and after relevant memory experience. Human Leamfng. 5.91-107. HerrmaM. D. J., & Neisser, U. (1978).An investory of everyday memory experiences. In M. M. Gruneberg, P. E., Morris, & R. N. Sykes (Eds.), Practical aspects ofmemay (pp. 35-51). London: Academic Press. Hertzog, C.. Dixon, R A.. & Hultsch, D. F. (in press). Relationships between metamemory. memory predictions, and memory task performance in adults. Psychology and Aging. Hertzog, C., Dixon. R. A.. & Hultsch. D. F. (1990). Construct validation of metamemory questionnaires through structural equation models. Unpublished Manuscript. On Hertzog, C.. Dixon, R. A.. Schulenberg. J. & Hultsch. D. F. (1987). the differentiation of memory beliefs from memory knowledge: The factor structure of the Metamemory in Adulthood scale. ExperfmentdAging Research, 13. 101-107. Hertzog, C.. Hultsch, D. F., & Dixon, R A. (1989).Evidence for the convergent validity of two self-report metamemory questionnaires. Developmental Psychology, 25, 687-700. Hertzog, C., Saylor, L., Mobley. M., Hultsch. D. F.. & Dixon, R. A. (1989,August), Stability of metamemory in adults: A longitudinal study. Paper presented at the 97th Annual Convention of the American Psychological Association. New Orleans, LA. Hertzog. C.. Van Alstine. J.. Usala, P..Hultsch, D. F.. & Dixon, R A. (in press). Measurement properties of the Community Epidemiological Studies - Depression (CES-D) scale in older populations. PsychoZogical Assessment. Hulicka, 1. M.. & Grossman, J. L. (1967).Age group comparisons for the use of mediators in paired associate learning. Journal of Gerontology, 22. 46-5 1. Hultsch, D. F. (1969).Adult age differences in the organization of free recall. Developmental Psychology, 1 , 673-678. Hultsch, D. F. (1975).Adult age differences in retrieval: Tracedependent and cue-dependent forgetting. Developmental P S Y C ~ O ~ O1~1.Y 197-201. . Hultsch, D. F. & Dixon. R. A. (1984).Memory for text materials in adulthood. In P. B. Baltes & 0. G. Brim, Jr. (Eds.). Life-span development and behavior (Vol. 6).New York Academic Press, Hultsch, D. F.. &Dixon. R A. (in press). Learning andmemory and aging. In J. E. Birren & K. W . Schaie (Eds.), Handbook ofthe psychology ofaging (3rd Ed.). San Diego: Academic Press.
208
Hertzag, D m n & Hultsch
Hultsch, D. F.. Hertzog. C., & Dixon, R. (1987). Age dmerences in metamemory: resolving the inconsistencies. Canadian Journal ~ f P ~ y c h o l ~41, g y .193-208. Hultsch, D. F., & Hertzog, C., & Dixon, R. A. (1985).Memory perceptions and memory performance in adulthood and aging. Canadian Journal onAging. 4. 179-187. Hultsch, D. F., Hertzog, C., Dixon, R. A., & Davidson. €3. (1988). Memory self-knowledge and self-efficacy in the aged. In M. L. Howe & C. J. Brainerd (Eds.), cognitive developmental in adulthood: Progress in cognitive development research (pp. 6592).New York Springer. Jackson, D. N. (1976). Manual for the Jackson PersonaZfty Inventory. Goshen, Ny: Research Psychologists Press. Judd, C. M.. Jessor, R., & Donovan, J. E. (1986). Structural equation models and personality research. Journal of Personality, 5 4 , 149-199. Kahn, R. L., Zarit, S. H., Hflbert, N. M., & Niederhe, G. (1975). Memory complaint and impament in the aged: The effects of depression and altered brain function. Archives of General P~ycNatry,32,1569-1573. Kanfer. R. (1987).Task-specific motivation: An integrative approach to issues of measurement, mechanisms, processes, and determinants. Journal of Social and Clhical Psychology. 5,237-264. Kausler, D. H. (1982).Experimental psychology and human aging. New York John Wiley. Kintsch, W. & van Dijk. T. A. (1978).Toward a model of text Comprehension and production. Psychological Review, 85. 363-394. Lachman. J. L.. & Lachman, R. (1980).Age and actualization of world knowledge. In L. W. Poon. J. L. Fozard, L. S. Cermak, D. Arenberg, & L. W.Thompson (Eds.), New directions in memory and aging I pp. 285-311). Hillsdale, NJ: Lawrence Erlbaum Associates. Lachman, J. L., Lachman, R., & Thronesbery, C. (19791. Metamemory through the adult life span. Deuelopmental P ~ y ~ h O l ~15, g y 543-551. , Lachman. M. E. (1983).Perceptions of intellectual aging. Antecedent or consequence of intellectual functioning? Developmental Psychology, 19, 482-498. Lachman, M. E. (1986). Locus of control in aging research: A case for multidimensional and domain-specific assessment. PSychOZ~gyand Aging. 1.34-40.
M e t a m y in Adulthood
209
Lachman. M. E. & Jelalian, E. (1984).Self-eMcacy and attributions for intellectual performance in young and elderly adults. Journal of Gerontology. 39. 577-582. Lachman, M. E.. & Leff, R. (1989).Perceived control and intellectual functioning in the elderly: A 5-year longitudinal study. Developmental Psychology, 25,722-728. Lachman, M. E., Steinberg, E. S., & Trotter, S.D. (1987).Effects of control beliefs and attributions on memory self-assessments and performance. Psychology and Aging, 2.266-271. Langer, E. (1981).Old age: An artlfact? In J.McCaugh & S. Kiesler (Eds.). Agtng: Biology and behavior (pp. 255-282).New York Academic Press. Levenson, H. (1974).Activism and powerful others: Distinctions within the concept of internal-external locus of control. Journal of Personality Assessment, 38,377-383. Loewen. E. R., Shaw, R. J., & Craik. F. I. M. (in press). Age differences in components of metamemory. Experimental Aging Research Lovelace, E. A,,& Marsh, G. R. (1985). Prediction and evaluation of memory performance by young and old adults. Journal of Gerontology, 40. 192-197. Markus. H., & Wurf, E. (1987). The dynamic self-concept: A social psychological perspective. Annual Reulew of Psychology, 38, 299-338. Messick. S . (1981).Constructs and their vicissitudes in educational and psychological measurement. Psychologfcal Bulletin, 89, 575-588. Murphy, M. D., Sanders, R. E.. Gabriesheski, A. S., & Schmitt. F. A (1981).Metamemory in the aged. Journal of Gerontology. 36, 185-193. Murphy, M. D., Schmitt, F. A..Caruso. M. J.. & Sanders, R E. (1987). Metamemory in older adults: The role of monitoring in serial recall. Psychology and Aging. 2,331-339. Niederehe, G. & Yoder. C. (1989). Metamemory perceptions in depressions of young and older adults. The Journal of N e w u s and Mental Disease, 177,414. O'Hara. M. W.. Hinrichs. J.V.. Kohout. F. J., Wallace, R. B.. and Lemke, J. H. (1986). Memory complaint and memory performance in the depressed elderly. Psychology and Aging. 1. 208-214. Paulhus, D.. & Christie R. (1981).Spheres of control: An interactfonist approach to assessment of perceived control. In H. Lefcourt (Ed.), Research with the locus of control construct Wol. 1. pp. 161-188).New York Academic Press.
210
Hertzog, Dtrcon & Hultsch
Perlmutter, M. (1978). What is memory aging the aging of? Developmental Psychology, 14,330-345. Perlmutter. M. (1980).An apparent paradox about memory aging. In L. W. Poon, J. L. Fozard, L. S. Cermak, D. Arenberg. & L. W. Thompson (Eds.),New dfrections in memory and aging (pp. 345354). Hillsdale, NJ: Erlbaum. Perlmutter, M., Adams. C., Berry, J., Kaplan. M., Person, D.. & Verdonik, F. (1987).Aging and memory. In K. W. Schaie (Ed.), Annual review ofgerontology and gerfatrfcs (Vol. 7,pp. 57-92). New York Springer. Person, D. C., & Wellman, H. M. (1988).Older adults' theories of memory dtJjlculties. Unpubltshed Manuscript. Popkin, S. J., Gallagher, D., Thompson, L. W., & Moore, M. (1982). Memory complaint and performance in normal and depressed older adults. Experimental Aging Research, 8,141- 145. Rabinowitz, J. C. (1989).Age deficits in recall under optimal study conditions. Psychobgy and Aging. 4,378-380. Rabinowitz, J. C., Ackerman, B. P.. Craik, F. I. M.. & Hinchley, J. L. (1982). Aging and metamemory: The roles of relatedness and imagery. Journal of Gerontology. 37, 688-695. Radloff, L. S.(1977).The CES-D scale: A self-report depression scale for research in the general population. Applied Psychological Measurement. 1, 385-401. Rebok, G. W.. & Balcerak, L. J. (1989).Memory self-efficacy and performance differences in young and old adults: Effects of mnemonic training. Developmental Psychology. 25, 714-721. Roenker, D. L.. Thompson, C. P., &Brown, S.C. (1971). Comparison of measures for the estimation of clustering in free recall Psychological Bulletin, 76, 45-48. Ryan, E. B. (1989).Bellefs about memory changes across the lifespan. Unpublished manuscript. Sanders, R E., Murphy, M. D., Schmitt, F. A, & Walsh, K. K. (1980). Age differences in free recall rehearsal strategies. Journal of Gerontology, 35. 550-558. Schmitt, F. A., Murphy, M. D., & Sanders, R. E. (1981).Training older adult free recall rehearsal strategies. Journal of Gerontology. 36, 329-337. Schneider, W. (1985).Developmental trends in the metamemorymemory behavior relationship: An integrative review. In D. L. Forrest-Pressley, G. E. Mackinnon, & T. G. Waller (Eds.), Cognition, metacognitfon. and human performance (Vol. 1, pp. 57-109).New York Academic Press. Sehulster. J. R. (1981).Structure and pragmatics of a self-theory of memory. Memory and Cognition. 9, 263-276.
Metamemory in Adulthood
211
Shaw, R. J.. & Craik, F. I. M. (1989).Age differences In predictions and performance on a cued recall task. Psychology and Aging. 4,131-135. Smith, A. D. (1980).Age differences in encoding, storage, and retrieval. In L. W. Poon. J. L. Fozard. L. S. Cermak, D. Arenberg, & L. W. Thompson (Eds.), New directions in memory and aging (pp. 23-45).Hillsdale, NJ: Erlbaum. Srull. T. K., & Wyer, R S.,Jr. (1989).Person memory and judgment. Psychological Review, 96.58-83. Sternberg, R J. (1985). Beyond IQ:A Marchic theory of human intelligence. New York Cambridge University Press. Sunderland, A.. Harris, J. E., & Baddeley. A. D. (1983).Do laboratory t e s t s predict everyday memory? A neuropsychological study. Journal of Verbal Learning and Verbal Behavior, 22,341-357. Sunderland. A , Watts, K., Baddeley, A. D., & Harris, J. E. (1986). Subjective memory assessment and test performance in elderly adults. Journal of Gerontology, 4 I , 376-384. Treat, N. J.. Poon,L.W.. Fozard. J. L.. & Popkin. S.J. (1978). Toward applying cognitive skill training to memory problems. Experimental Aging Research 4 , 305-319. Usala. P. D.. & Hertzog. C. (1989).Evidencefor diflerential stability ofstate and trait anxfety in adults. Unpublished Manuscript. The structure of psychological Veit, C., & Ware. J. E.. Jr. (1983). distress and well-being in general populations. Journal of Consulting and Clinical P~y~hology. 51, 730-742. Weaver, S. L., & Lachman, M. E. (1989.August). Enhancing memory self-conceptions and strategies in young and old adults. Paper presented at the 97th Annual Meeting of the American Psychological Association, New Orleans, LA. Wellman. H. M. (1983). Metamemory revisited. In M. T. H. Chi (Ed.). Trends in memory development research (pp. 31-51).Basel: Karger. West, R. L.. Boatwright. L. K., & Schleser. R (1984). The link between memory performance, self-assessment. and affective status. Experimental Aging Research 10. 197-200. Zarit, S. H. (1982). Affective correlates of self-report about memory in older adults. International Journal of Behavioral Geriatrics, 1. 25-34. Zarit, S. H., Cole, K. D.. & Guider, R. L. (1981).Memory training strategies and subjective complaints of memory in the aged. The Gerontologfst.21, 158-164.
212
Hertzqg, Dtwon & Hultsch
Zelinski, E. M.,Gilewski, M. J., & Thompson, L. W. (1980). Do laboratory test relate to self-assessment of memory ability in the young and old? In L. W. Poon, J. L. Fozard, L. S. Cermak, D. Arenberg, & L. W. Thompson (Eds.),New dtrections in memory and aging: Proceedings of the George A. Talland Memorial Conference (pp. 5 19-544). Hillsdale, NJ: Erlbaum.
Aging and Cognition: Knowledge Organhtion and Utilization 'Ihomas M. Hess Witor) 0 Elsevim Science RCbZf~shersB.V. (North-HOlladJ.1990
CHAPTER FIVE
COGNITION AND AGING: A THEORY OF NEW LEARNING AND THE USE OF OLD CONNECTIONS Donald G. MacKay University of California, Los Angeles Deborah M.Burke Pomona College
SUMMARY
This chapter describes a detailed theory of perception, production and memory for language and applies it to the problem of cognitive decline in old age. Altering a single parameter in the theory (rate of priming) was shown to account for a wide range of established age Werences in cognitive ability, and to suggest an alternative framework for understanding some findings which in the past have seemed contradictory. Examples of these findings are effects of age on learning, rate of processing (general slowing), and the tip of the tongue (TOT) phenomenon. The theory postulates different mechanisms for retrieving existing representations in memory vs. learning new or unique representations and predicts that new learning will be especially vulnerable to aging. Specifically, the theory predicts that age differences will increase with the number of new connections required in a memory task, but will diminish if already established connections are sufficient to accomplish the task. This prediction cuts across specific paradigms and theoretical distinctions and applies to a broad range of memory phenomena. By way of illustration, we review findings from experimental studies of encoding specificity, implicit versus explicit memory, and semantic versus episodic priming. and show how the observed pattern of age differences is consistent with disruption of new
214
MacKay and Burke
learning and preservation of memory involving existing connections. The theory also makes some interesting and genuinely new predictions for future research that are spelled out here, for example, an age-linked decline in the detection of speech errors. Psychology is currently witnessing an upsurge of interest in theory. especially within advanced areas of psychological theorizing such as cognitive psychology. However, there has been little development of theory in cognition and aging. Salthouse (1988)recently lamented this absence of theory and noted that explanations of aging effects often invoke vague concepts such a s processing resources or mental effort which have not been integrated into a well defined cognitive model. Salthouse urged investigators in cognition and aging to begin development of testable and well specifled cognitive models. This recognition of the importance of theoretical development represents a recent shift in aging research and experimental psychology at large: over the past sixty years the field has been dominated by an empirical epistemology whose primary goal is to develop a body of reliable facts. Under this empirical epistemology, theories emerge spontaneously when a large enough body of data has been amassed (MacKay, 1988). However, although the accumulation of data may reveal empirical laws, it does not produce theory. Theories originate as products of cognition rather than observation. In the history of science, theorists have often developed highly successful theoretical constructs, such as atoms and sound waves, quite independently of the accumulation of experimental data (see MacKay. 1988). Following in this spirit, the present paper analyzes selected findings in cognitive aging within the context of a detailed and explicit theory of language and memory. The theory postulates specific memory representations and processing mechanisms and views age-related memory deficits as the result of an impairment in a particular mechanism that is basic to all perception, action, and learning, namely, the priming of memory representations. Our aim is to make a theoretical contribution, rather than an empirical one, and our theoretical analyses provide an explanation for well established aging patterns such as general slowing (e.g., Birren.
Cognition and Aging
215
1965) and differential decline in tasks involving new versus old learning or fluid versus crystallized abilities (e.g.. Horn. 1982;Horn & Cattell. 1966; Light & Burke, 1988; Salthouse, 1982, 1988). Further, the theory suggests a n alternative framework for understanding some existing findings which previously seemed contradictory, and it generates some interesting new predictions for future research. The theory presented is the Node Structure Theory (NST) developed originally by MacKay (1982;1987)to provide a general and explicit account of the perception and production of language and other cognitive skills. The theory has been extended here to include processes that are rarely seen in models of language perception and production, but are necessary for explaining memory phenomena such a s the recall of new information (see also MacKay, 1990).The theory likewise incorporates processes that are rarely seen in memory models, but are necessary for other reasons, for example, producing speech sounds in correct order in words, and words in correct order in sentences. This chapter is organized into seven sections: Section I presents a brief qualitative description of the basics of the NST, focusing on the representation and use of already established knowledge. Section I1 outlines the Transmission Deficit hypothesis, the claim that priming transmitted across connections between nodes in the NST declines in rate and amount as a function of age. Section I11 shows how this Transmission Deficit explains general slowing, the reduction in processing rate that accompanies aging. Section N examines the differences between two types of learning processes in the NST: commitment learning, the process whereby new connections are formed or new combinations of familiar information are represented versus engrainment learning, the process whereby existing connections among familiar information are strengthened when the information is retrieved or utilized. We show that under the Transmission Deficit hypothesis, commitment learning should be much more vulnerable to the effects of aging than engrainment learning. Using this princple, Section V provides an account of older adults' pattern of impaired and spared memory functions seen in studies of encoding speciflcity. semantic versus episodic priming, and direct versus indirect memory tasks.
216
MacKay and Burke
Section VI examines a striking exception to this general pattern, namely. the age-related increase in tip of the tongue (nrr)states in which retrieval of existing word knowledge is impaired. We show how the Transmission Deficit hypothesis can explain this specific deficit. Finally, Section VII outlines some new and untested predictions concerning the relation between aging and phenomena such a s error detection that follow from the Transmission Deficit hypothesis. BASICS OF THE NODE STRUCTURE THEORY
Mental processes in the NST consist of inhibitory and excitatory interactions occurring in parallel between hypothetical processing units or nodes in a highly interconnected network. Like the parallel distributed processing (PDP) models of McClelland and Rumelhart (e.g., Rumelhart & McClelland, 1986)the model allows simultaneous integration of many sources of information. However, nodes within the NST involve local representation (where a unique concept corresponds to a single node) rather than the mass action representation of some PDP models (where a unique concept corresponds to a pattern of activity among a large number of nodes. and no node codes any concept uniquely). Some NST nodes represent higher level cognitive components such a s phonological segments, syllables, and words, and play a role in both perception and production. These "mental nodes" are organized hierarchically into a sentential or semantic system that represents semantic and grammatical information, and into a phonological system for speech perception/production, and an orthographic system for reading/writing (see Figure 1). Other systems do not share perception/production functions. For example, one system contains sensory analysis nodes which represent patterns of auditory input for perceiving speech. Another system contains muscle movement nodes which represent patterns of muscle movements for producing speech. For purposes of illustration, Figure 2 shows a sample of topdown connections for producing the sentence "The dragon ate the fudge." The highest level node representing the entire thought underlying the sentence has the content "the dragon ate the fudge."
Cognition and Aging
217
MENTAL NODES (SENTENTIAL SEMANTIC SYSTEM)
t MENTAL NODES (e.g. PHONOLOGICAL
AND ORTHOGRAPHIC
I
SYSTEMS)
MUSCLE MOVEMENT
SENSORY ANALYSIS
NODES (PRODUCIION)
(PERCEPTION)
Rgure 1. Organization of the basic systems within the Node Structure Theory.
This particular node is connected to two other nodes: One represents the thought "the dragon" and the other represents the thought "ate the fudge". The node representing "the dragon" is connected to two lexical nodes: one represents "the" and the other represents "dragon". Nodes such a s these are part of the sententialsemantic system, and will be the primary focus of the present paper because most of the memory phenomena discussed here occur
MacKay and Burke
218
Propositional Node SENTENTIAL/ SEMANTIC
I
(active, declarative) the dragon ate the
L
fudge
-
I
Syllable Nodes PHONOLOGICAL SYSTEM Phonological Compound Nodes Phonological Nodes
(initial consonant group) dr
(vowel group) a
(initial consonant) d
FYgure 2. A sample of top-down connections for producing the sentence "the dragon ate the fudge".
Cognition and Aging
219
within this system. Note, however, that lexical nodes are connected with specific phonological nodes organized hierarchically into syllables, phonological compounds, segments, and features. Activation of these nodes allows retrieval of the phonology of the word corresponding to the concept activated at the lexical level. The phonological nodes map onto a system of nodes assumed to underlie muscle movements so that the person could actually speak the word or sentence, not just think it. (See MacKay. 1987, for details of node structures within the phonological and muscle movement systems). Nodes exhibit four processing characteristics that are relevant here: Activation, priming, linkage strength, and self-inhibition. Node activation in the NST differs from the concept of spreading activation in other network theories (e.g.. Anderson, 1983; McClelland & Rumelhart, 1985). Node activation never spreads, is all-or-none in nature. and has potential behavioral consequences: activation of the appropriate nodes results in perceptual recognition and/or production of a word. An activated node simultaneously primes all nodes connected to it and a primed node sends priming to its connected nodes with degree of priming decreasing as a function of distance. However, priming dflers from the concept of spreading activation in other network theories. Priming prepares a node for possible activation, and a node must be primed first before it can be activated. Figure 3 illustrates the priming and activation of a node. Although priming summates spatially and temporally, it only summates up to an asymptote, and cannot by itself activate a node: a special activating mechanism is required for activation. Linkage strength determines how much and how rapidly priming crosses the connections. Linkage strength represents a relatively long-term characteristic of a connection and has been used to explain a wide range of practice effects: Highly practiced connections have greater linkage strength and therefore transmit priming more rapidly and up to a higher asymptotic level than do relatively unpracticed connections (MacKay. 1982). Another factor influencing linkage strength is recency of activation: all other factors being equal, linkage strength will be greater for connections between nodes that have been recently activated. Conversely, connections between nodes that have not
220
MacKay and Burke
been activated over long periods of time (years) will exhibit diminished linkage strength. After a node becomes activated, it undergoes a brief period of self-inhibition during which its level of priming falls below normal or resting level (see Figure 3). The mechanism underlying self-inhibition is a n inhibitory 'satellite'. After receiving sufficient priming from its activated parent node. the satellite becomes activated and inhibits its parent node. The nodes discussed so far are known as content nodes, because they represent the form or content of a memory representation. We turn now to sequence nodes, the mechanisms for activating content nodes. Sequence nodes connect with a set of content nodes called a sequential domain, and all content nodes belong to at least one sequential domain. Nodes in a sequential domain all share the same sequential privileges of occurence. For example, nodes in the semantic system that are in the same sequential domains, e.,g.. color adjectives, exhibit identical patterns of sequential Occurence in a sentence. By way of illustration, sequential domains for the content nodes in Figure 2 are indicated in parentheses; the node representing the thought "the dragon" is in the noun phrase domain, "ate the fudge" is in the verb phrase domain, "the" is in the determiner domain, and "dragon" is in the noun domain. Once a sequence node becomes activated, it repeatedly multiplies the priming of every node in its domain by some large factor within a relatively brief period of time. This multiplicative process has no effect on an unprimed node, but soon serves to activate (that is, bring to threshold) the content node with the greatest degree of priming in the domain. Normally, the one node with more initial priming than all other nodes in the domain will reach threshold first, and this "most-primed" node will become activated. Thus, the all-or-none nature of activation and the control of activation by sequence nodes allow production of responses in appropriate order, for example, correct sequencing of words in sentences. Once a content node becomes activated, it inhibits its corresponding sequence node, thereby ensuring that one and only one content node in a domain becomes activated at any one time.
Cognitlon and Agirg
22 1
RECOVERY FUNCTION
b
‘2
‘1
TlME
Ffgure 3. The priming, activation and recovery functions for a single node. Activation is achieved through a special activating mechanism which multiplies the priming of all nodes within a sequential domain so that the first node to reach threshold is activated. THE TRANSMISSION DEFICIT HYPOTHESIS
The Transmission Deficit hypothesis provides the basis for explaining effects of aging within the NST. Under the Transmission Deficit hypothesis. age weakens the linkage strength of connections between nodes in the network so that the rate and asymptotic amount of priming transmitted across any given connection declines as a function of age. From a historical point of view, this Transmission Deficit account of aging is not entirely new: To explain the word finding problems of older adults, Bowles, Obler and Poon (1989)and G. Cohen and Faulkner (1986)proposed a n age-related reduction in transmission of priming from semantic to phonological representations. And Salthouse (1988) can be interpreted a s suggesting the possibility of a more general reduction in the rate of priming a s one of three possible explanations for
222
MacKay and Burke
observed declines in new learning and fluid intelligence. The other explanations involved either a decrease in how many nodes can be simultaneously active, or a decrease in the maximum available activation (see also Salthouse, 1985). However, the NST postulates no principled limit on how many nodes can be simultaneously primed and/or activated, and views activation as a local or localized activity (that is specific to particular nodes) rather than a global or network resource that can become depleted. Moreover, the NST views aging as just one of several factors contributing to transmission deficits in both young and older adults, others being recency of activation (time since last activation) and history of prior practice (number of previous activations over the course of a lifetime). Finally, the NST allows greater specificity about the nature of transmission deficits in particular connections and the implications of these deficits within a detailed account of perception, production, and recall in language and other cognitive skills. Figure 4 illustrates the specific form of the age-linked transmission deficit by comparing the transmission of priming across a single connection to a YOUNG versus an OLD node. The YOUNG and OLD nodes are part of an otherwise comparable domain of nodes with spontaneous or resting level Lo. and both are receiving priming transmitted from a single other node beginning at time to and ending at time t3. All other characteristics of the nodes are assumed to be equal in the theory, especially their history of prior practice and their recency of activation. The summation characteristics of the nodes in Figure 4 illustrate how priming summates over time until asymptote is reached, and as can be seen there, priming builds up more slowly and to a lower asymptote for OLD versus YOUNG nodes. Figure 4 also shows how the priming on a node decays over time when a connection no longer transmits priming. Note that decay rate is unrelated to age under the Transmission Deficit hypothesis: priming decays at the same rate for YOUNG versus OLD nodes. Because frequent and recent use (activation) also influence linkage strength and enable more efficient transmission of priming across connections, these factors can in principle mask an age-linked transmission deficit in older adults. In general,
Cognition and Aging
223
however, effects of frequency and recency of activation seem likely to trade off or cancel one another with advancing age. Frequencyof-use (repeated activation of already coded information) can only increase with age, whereas recency-of-use will tend to decrease with age. As Burke, Worthley. and MacKay (1989) point out, age provides greater opportunity for long intervals following, say, last use of the name of an acquaintance such a s a grade school teacher. Such decreases in recency-of-use may explain why older adults experience disproportionately more TOTS for the names of acquaintances (see Burke et al., 1989). The Transmission Deficit hypothesis is consistent with neurobiological characteristics of aging. An example is the observation of Landfield (1988) and Landfield and Lynch (1977) that neurons in the aged hippocampus (of rats) exhibit reduced potentiation and increased time to peak. The Transmission Deficit hypothesis is also consistent with the finding that the time required for conducting neural activity between brain structures, including different areas within the cortex. increases with age (see e.g., Aston-Jones. Rogers, Shaver, Dinan & Moss, 1985; Dorfman & Bosley, 1979; Streng & Hedderich. 1982). Under the Transmission Deficit hypothesis. this age-linked increase in conduction times may reflect increased synaptic delays, a possibility that remains to be tested in the case of cortical neurons. The exact nature of such synaptic delays also remains to be determined, with possibilities including reduced postsynaptic responsivity and decreased uptake of neurotransmitter substances (see Decker, 1987). The remainder of the present paper elaborates on various consequences of the Transmission Deficit hypothesis illustrated in Figure 4. One consequence explored in Section I11 is reduced reaction times and rates of responding in a wide variety of tasks. Another consequence is that OLD nodes will sometimes transmit insufficient priming to enable their connected nodes to become activated. This consequence may underlie the age-linked deficits in the encoding of new information: Section IV and V argue that transmission deficits can make the learning of new information imposssible by causing activation failures in the uncommitted nodes that are essential for coding new information. Section VI argues that transmission deficits also cause activation failures in
MacKay and Burke
224
c3
DECAY CHARACTERISTICS
SUMMATION CHARACTERISTICS
z
U
c
U
a! k
Young Node
OldNode
-----
Ftc
0
;k
W
W a!
c3 W
\ \
CI
\ \
\
MEAN NOISE LEVEL
t0
‘1
t2
T
\
\
‘\
‘3
I
M
E
Flgure 4. The Transmission Deficit hypothesis: The hypothetical functions
for the summation and decay of priming over time for a YOUNG and an OLD node within an otherwise comparable domain of nodes with spontaneous or mean noise level Lo. Both nodes are receiving priming transmitted from a single other node beginning at time to and ending at time t3 (see text for explanation). Summation characteristics illustrate how priming builds up on these nodes over time up to asymptote, while decay characteristics illustrate how priming decays over time to the spontansous or mean noise level.
Cognition and Aging
225
existing connections and are responsible for retrieval failures such as TOTS.
GENERAL SLOWING UNDER THE TRANSMISSION DEFIClT HYPOTHESIS The Transmission Deficit hypothesis readily explains the general slowing that occurs with aging in a wide range oftasks (for a summary. see e.g., Salthouse, 1985). However, the relation between slowing and transmission deficits is neither simple nor direct under the NST and requires a clear understanding of the age-related decrement in priming transmission illustrated in Figure 4. Implications of this decrement are discussed below. How m-iming Transmission Influences Speed and Errors in the NST Under the NST, both reaction time and general rate of output depend on how soon following onset of priming a triggering mechanism (sequence node) is applied to a domain for the full set of nodes in an output hierarchy. That is. a triggering mechanism can be applied to its domain at any voluntarily determined point in time after onset of priming (to in Figure 4) for every node in an output hierarchy. If the triggering mechanism is applied soon after onset of priming, say at ti in Figure 4, the overall rate of output will be fast because activation of each of the many hundreds of nodes in the output hierarchy wffl occur soon after priming onset. (Recall that the nodes illustrated in, for example, Figure 2 represent only a small subset of the total number of nodes in such a n output hierarchy). However, if the triggering mechanism is applied long after onset of priming, say at t2 in Figure 4. the overall rate will be slow, because activation of each of the nodes in the output hierarchy will OCCUT longer after priming onset. Consider now the relation between rate and errors in the theory. For a given priming function, the faster the rate (that is, the sooner the triggering mechanism ts applied after the content node begins to receive prfming), the greater will be the likelihood of error (that is, the probability of activating the wrong content node). This is
226
MacKay and Burke
because the target node is less likely to have greater priming than all other extraneous nodes in the domain (the current noise level) soon after priming onset a s compared to later (see Figure 4). Or equivalently in the theory, the more rapid the priming summation characteristics of nodes, the faster will be the potential rate of output for any given probability of error. In particular, with probability of error held constant, the overall rate of output can be faster for YOUNG nodes than for OLD nodes because of the more rapid priming summation characteristics of YOUNG nodes (see Figure 4). However, the theory does not necessarily assume that young and older adults are equally intolerant of errors, that is, that they adopt equivalent error criterions. For both young and older subjects it is possible to vary the error criterion within wide limits: all subjects can reduce the probability of errors by decreasing speed or they can increase the rate of output at the expense of allowing more errors. To illustrate these points more concretely, let the YOUNG and OLD nodes in Figure 4 represent the concept blue (color adjective) for young and old adults participating in a standard Stroop task. The stimulus in this hypothetical experiment is the word "redt printed in blue ink, and the subject must name the color of the ink, that is, produce the word "blue" as quickly a s possible in this example. Now in order for the correct response to be produced, blue (color adjective) must acquire more priming than any other node in the color adjective domain so a s to become activated under the most-primed-wins principle. In particular, blue (color adjective) must acquire more priming than red (color adjective), its main competitor in the color adjective domain. Now, red (color adjective) automatically acquires priming bottom-up from the orthographic characteristics of the word, and because of the frequency with which we read words such as red (see MacKay, 1987).this priming rises very rapidly to an asymptotic level that can be considered a theoretical constant corresponding to, say, L1 in Figure 4. The color of the letters simultaneously primes blue (color adjective), but this priming accumulates more slowly because we relatively rarely name the color of letters (see MacKay. 1987). This more slowly but systematically increasing priming level will exceed L1 by time ti for the YOUNG node in Figure 4, and by time t2 for the OLD node.
Cognitfort and Aging
227
Thus, because an activating mechanism always activates the most primed node in its domain, blue (color adjective) will be activated and the output wfll be error-free if the activating mechanism is applied after t l for the YOUNG node and after t2 for the OLD node. More specifically. if young and older subjects both adopt an error criterion corresponding to, say, L1 in Figure 4, rate of output will be faster for YOUNG nodes (inversely proportional to t l in Figure 4) than for OLD nodes (inversely proportional to t2). Turning to the accuracy side of the speed-accuracy trade-off (SAT) function, subjects will produce the prototypical Stroop error, "red" instead of 'blue" in this example, if they apply their activating mechanisms before ti for YOUNG nodes in Figure 4 and before t2 for OLD nodes. The substitution of "red" for "blue" in this example represents an error occuring at the lexical level within the semantic-sentential system, but the underlying principles can be generalized to apply to a wide range of error types within any arbitrarily chosen domain at all levels of the system. To better illustrate this generalizability, if nodes other than the correct node in a domain are labeled extraneous nodes, then the noise level or degree of priming arriving at the most primed extraneous node in the domain can in general be considered to assume a Gaussian distribution with mean Lo (the spontaneous or resting level) at any given point in time (see Figure 5). For example, the noise level represented as L1 in Figure 4 might exceed the spontaneous or mean resting level of other extraneous nodes in the domain by, say, 3 standard deviations because of priming received from 'other sources', here the nodes coding the orthography of the word "red'. In general, however, the noise level (priming of the most primed extraneous node) can be considered a random variable at any point in time, whereas priming for the correct node will increase systematically over time in the manner shown in Figure 4. As a consequence, the later the application of the triggering mechanism following onset of priming, the lower wfll be the probability of error up to the asymptote of a given priming function, In short, errors will trade off with speed in the theory and speed-accuracy trade off functions for YOUNG vs. OLD nodes should resemble the priming functions in Figure 4. a prediction that is readily subjected to experimental test (see e.g.. Wickelgren, 1977).
MacKay and Burke
228
h
n
b
Ll
N O I S E Rgure 5. The frequency distribution for noise (priming levels of the most primed extraneous node in a domain).
Time and errors are closely coupled under the NST: choosing a particular rate is equivalent to choosing a particular probability of error and vice versa. Priming transmission plays a universal role in determining what time-error criterion can be adopted because increased priming transmission causes a favorable shift in the absolute values that any time-error criterion can assume (see Figure 4). Also, many cognitive and motivational factors undoubtedly play a role in what time-error criterion an individual will in general select, and a given individual may alter their timeerror criterion depending on the situation or occasion. However, we can assume that people normally choose a conservative time-error criterion: a rate of output resulting in some acceptably low probability of error. COMMI'ITMENT LEARNING UNDER THE NST So far, we have been discussing the activation of information that is already represented in memory, the so called committed
Cognitton and Aging
229
nodes in the NST. Connections to a committed node are usually strong enough s o that the committed node can accumulate sufficient priming to enable activation when its activating mechanism (sequence node) is applied. For example, adults would normally have committed nodes representing already encountered lexical concepts such as "clumsy" and 'bear" in Figure 6. These committed nodes engage in a form of learning known as engrainment learning that provides the basis for effects of frequency or practice (repeated activation). Engrainrnent learning occurs when practice increases the linkage strength of connections between committed nodes in the model, and enables greater speed and accuracy in executing established behaviors. New learning, however, requires a dmerent type of node, the so called uncommitted nodes, and an additional type of learning mechanism. Unlike a committed node, a n uncommitted node receives connections that are so weak that it cannot accumulate enough priming to become activated when its activating mechanism is applied, even in young adults. Connections from uncommitted nodes are likewise so weak that they transmit too little priming for connected nodes to become activated when their activating mechanism is applied. Commitment is therefore a property of both nodes and of connections. New learning occurs to the extent that a n uncommitted node becomes committed, a process that requires prolonged activation and is called commitment learning. For example, no preestablished node represents the phrase "clumsy bear" in Figure 6, but only the uncommitted node labeled X. This is true for a person who has not experienced or processed this particular noun phrase before, or at least not to the extent required for permanent commitment learning. Thus when the NOUN PHRASE sequence node becomes engaged, none of the nodes in its domain have sufficient priming to become activated: inhibition of the sequence node occurs only when there is activation. As a result, there is sustained activation of the sequence node, which activates the binding node for noun phrases. (Each sequence node has a corresponding binding node that binds or commits nodes of that type.) Sustained activation also triggers orienting reactions, that
230
MacKay and Burke
NOUN PHRASE DOMAIN
1 ADJECTIVEDOMAIN
NOUN DOMAIN
6. A sample of lexlcal and noun phrase nodes and the sequence node and binding node for noun phrase nodes. Committed nodes are black and the uncommitted node for "clumsybear" is marked with an X
Figure
include behavioral components, such as the inhibition of ongoing activity (see MacKay. 1990). Bindings nodes operate a s follows. When activated, a binding node shuts down the inhibitory satellite of all content nodes to which it is connected, in the present example, all noun and adjective nodes. With inhibition shut off, nodes that happen to be activated at the time remain activated for a prolonged period of time, allowing priming to summate at the uncommitted node for a longer period. This summation of priming may eventually enable the uncommitted node to achieve sufficient priming so that it can become activated as the most primed in the domain when its activating mechanism is applied. Activating the uncommitted node in turn causes a slight but relatively long-term increase in linkage strength across its connections, resulting in "weak commitment" of the uncommitted node. Weak commitment is extremely fragile, however: unless weakly committed nodes become activated again within some relatively brief period of time. the
Cognition and Aging
231
increased strength of their connections can decay over time so that the node reverts to uncommitted status. Commitment Learning and the Transmission Deficit Hypothesis Strong transmission of priming is essential for the retention of new information or novel combinations of old information in the N S T account of commitment learning discussed above. If prolonged activation of the novel combination of committed nodes fails to deliver priming to the uncommited node in sufficient strength, activation and commitment of the uncommitted node cannot occur. This means that if aging reduces priming transmission, a s postulated under the Transmission Deficit hypothesis, then uncommitted nodes in older adults will fail to achieve even weak commitment, and commitment learning will fail to occur. As a result, memory for new information will suffer greater agerelated deficits than memory for old or committed information because commitment learning requires considerably more summation of priming than does retrieval of old information. Moreover, certain characteristics of commited nodes can compensate for an age-linked reduction in priming. In particular, practice involving already committed nodes is likely to increase with age and by increasing linkage strength, this additional practice would reduce the effect of an age-related transmission deficit. Further, the complex connections among committed nodes within the semantic system are likely to overcome an age-linked transmission deficit via convergent priming. For example, under the NST, associated concepts such as doctor and nurse are linked via connections to a large number of proposition and predicate nodes representing information such as "doctors and nurses work in hospitals", .wear white uniforms", .give injections", etc. Consequently, priming will converge on the committed nodes for such concepts from these multiple connections, enabling summation of priming that reduces the effect of a transmission deficit (see discussion of semantic priming effects and the single source factor below). 'I..
'I..
232
MacKay and Burke
Selective Impairment of Binding Nodes: An Alternative to the Transmission Deficit Hypothesis The Transmission Deficit hypothesis is not the only means of predicting age-related deficits in the learning of new information under the NST. The hypothesis that aging selectively impairs the functioning of binding nodes represents a viable alternative. Because binding nodes are essential for commitment learning, this "Binding Impairment hypothesis" makes the same predictions a s the Transmission Deficit hypothesis: age-related impairments will be most pronounced when new information or novel combinations of old information must be retained. Moreover, as MacKay (1990) points out, a Binding Impairment hypothesis is consistent with the pattern of amnesic effects that result when hippocampal and mediotemporal areas of the brain are damaged or lesioned. However, the Binding Impairment hypothesis faces a large number of unresolved theoretical issues that led us to abandon it in favor of the Transmission Deficit hypothesis as an explanation of age-related effects. One unresolved theoretical issue is why aging impairs the functioning of binding nodes but not the functioning of other closely connected nodes such as sequence and content nodes. Another unresolved issue concerns the explanation of aging effects in tasks that do not involve novel information, for example, the general slowing effects discussed above, and the TOT phenomenon [see below]. A final unresolved issue for a Binding Impairment hypothesis is the specification of precisely how the functioning of binding nodes could be impaired. For example, does aging influence how long a sequence node can sustain its activation so as to trigger its binding node? Or functionally equivalent under the NST, does aging somehow decrease the responsivity of binding nodes, thereby increasing the threshold that must be reached in order to activate a binding node? Or does aging selectively decrease how effectively a binding node inhibits its domain of inhibitory satellites, thereby interfering with the prolonged activation required for commitment? These various versions of the Binding Impairment hypothesis all share an additional drawback: they predict lapses of consciousness because binding nodes introduce prolonged
Cognition and Aging
233
activation, which corresponds to awareness in the NST (see MacKay, 1990). If age impairs the functioning of binding nodes, prolonged activation would fail to occur, so that awareness or consciousness of new information should diminish with age. However, neither diminished awareness nor disturbances of ongoing consciousness seem characterisitc of normal aging. COMMITTMENT LEARNING AND AGE DIFFERENCES IN MEMORY PERFORMANCE: A SELECTIVE REVIEW A s incorporated within the NST, the Transmission Deficit hypothesis predicts greater age-related impairments when commitment learning is involved. Thus, the theory accounts for one of the most striking characteristics of the memory deficit in older adults: the extent to which the deficit involves memory for new rather than old information (e.g., Light & Burke, 1988; Salthouse. 1985). For example, in evaluating the contribution of age and experience to chess performance, Charness (1985) concludes that age differences are seen when new information must be represented, but not when well-practiced information is involved. Further, relative to young adults, older adults improve more slowly with practice on tasks involving new learning, but their rate of improvement is at least as fast on familiar tasks (Welford. 1985). The NST uses a single principle, called here the "commitment learning principle", to account for the pattern of age-related memory deficits in a wide variety of experimental paradigms. Under the commitment learning principle, age deficits will increase with the amount of commitment learning required for accurate performance. To illustrate how the commitment learning principle can be applied to specific experimental situations. we analyze age effects in three well-known memory paradigms: encoding specificity tasks, episodic versus semantic priming tasks, and direct versus indirect memory tasks. We provide an account of these findings using a single principle rather than the variety of different mechanisms called for in the original studies.
234
MacKay and Burke
Encodlng Specificity, Aglng, and the Commitment Learning Principle One approach to investigating the representation of new memories has been to evaluate the effectiveness of different types of retrieval cues. According to the encoding specificity principle (Tulving & Thomson, 1973), effectiveness of a cue at retrieval depends on the extent to which it is encoded in the representation of the target material, Thus under the NST, encoding specificity depends on the strength of connections between nodes representing the retrieval cue and the target, that is. on how much priming can be transmitted between them. If retrieval cues require the formation of new connections to the target, that is, commitment learning, these cues should be more effective for young than older adults according to the commitment learning principle. If retrieval cues require only existing connections, that is. engrainment learning. these cues should be about equally effective for young and older adults. The commitment learning principle is consistent with the W i n g s of Craik and Simon (1980)who gave subjects sentences and tested for recall of a target noun (e.g.. "bear") after supplying as a retrieval cue either the original adjective from the sentence (e.g.. "clumsy")or the name of a superordinate category for the noun (e.g., "animal"). For the young but not the old, the adjective produced better recall than did the category (see Table 1). Similarly, Rabinowitz, Craik, and Ackerman, (1982) gave subjects word pairs that were either strongly associated (e.g., "king-queen") or weakly associated (e.g., "whiskey-water"; see Table 1). Retrieval cues for the second word in each pair were either the same word studied in the pair or a different word which varied in its strength of association to the target. For young but not older adults, the original word studied in the weak pair (Same-Weak)produced better recall than a different but strongly associated word (Different-Strong). Consistent with the commitment learning principle. the age decrement was about twice a s large in the Same-Weak condition compared to the Same-Strong condition in which retrieval cues depended on established connections.
Cognition and Aging
235
Table 1
Emunples of Stimuli and Resultsfiom Encodfqg Spec@Uy Studies
Craik & Simon (1980) Scudy: The highlight of the circus was the clumsy BEAR
Retrieval cue DiEerent (e.g., animal)
Original (e.g., clumsy) Young
.Ma
.32
Old
.22
.31
Rabinowitz, Craik & Ackerman (1982) Study: king-QUEEN (Strong);whiskey-WATER (Weak)
Retrieval Cue and Encoding Cue Identity:
Same
Different
Same
Different
Target-Retrieval Cue Relatedness
Strong
Weak
Weak
Stong
Young
.92a
.07
.85
.68
Old
.83
.03
.62
.63
Hess & Higgins (1983) Study: river bank shore Test Context Same
Different-Related (e.g.. money bank check)
Young
.92b
.80
Old
.71
.60
a Data are mean cued recall
Data are mean correct recognition
236
MacKay and Burke
Under the NST, the effectiveness of the adjective in the Craik and Simon (1980)study and of the weak associate in the Rabinowitz et al. (1982)study depended on commitment of new connections between the retrieval cues and their respective target words. In contrast, category cues in the first study and strong associates in the second study depended on engrainment learning that strengthens the already existing connections between cue and target. Thus older adults' apparent deficit in encoding specificity is best understood a s a n impairment in commitment learning relative to engrainment learning. This difference between commitment and engrainment learning is illustrated in Figure 7 . In order for clumsy to be an effective retrieval cue for bear, the uncommitted node labeled 'X' which represents the connection between clumsy and bear must become committed. However, the category name in the first study would. according to the model, be connected to the target word by an already committed node, in Figure 7 , bear Is an anfmd. Older adults did relatively better with this second type of retrieval cue because no nodes had to become committed for animal to call up bear. Regardless of the retrieval cue, memory that a word such as bear was presented in an experiment requires commitment learning under the NST. in the present example, an association between bear and some representation of the experimental context, simplified in Figure 7 as was presented1 An uncommitted node (X) represents this connection between target and experimental context, and this, according to the commitment learning principle, would explain the age decrement obtained by Rabinowitz et al. (1982)with the SameStrong retrieval cue. An alternative account of these findings postulates that age differences in encoding specificity reflect declines in "processing Such episodic knowledge requires a connection to nodes representing the context of the experience, and may be represented in the semantic-sentential system as, for example, a proposition that the concept "bear"was presented via computer terminal, or in a visual imagery system, as, for example, a visually based representation of the computer display or of the appearance of the experimental room. Representations for specific episodes are clearly varied and complex and may include information about time of occurence, location, modality, and other unique aspects of the episode (Burke & Light, 1981). However, an exhaustive analysis of episodic knowledge representations is not the goal of the present paper.
237
(ACTIVE DECLARATIVE)
@ I \
f?
BEARWAS
;', PRESENTED I ' I ' I ' I '
(NOUN PHRASE)
" I 1 I
'
" ..
BEAR IS ANIMAL
(VERBPHRASE)
I
\
NJIMAL
I
(NOUN DOMAIN)
Rgure 7. A sample of nodes representing "clumsy,""bear,""animal,""was presented," and their connections. Uncommitted nodes making new connections (designated by broken lines) are marked with an X and X. Established or cornmited connections are designated by solid lines. The content of a node is in captial letters and its sequential domain is in parentheses.
resources" in old age (e.g., Craik. 1983; Rabinowitz et al.. 1982). For example, Rabinowitz et al. (1982) concluded that older adults resorted to "general" encoding in the Same-Weak condition because limited processing resources prevented encoding of the specific context. This alternative account is weakened on a theoretical level by the amorphousness ofthe processing resources concept (see Light & Burke, 1988; Salthouse, 1988). and on an empirical level by the finding that older adults show context specific encoding when new
238
MacKay and Burke
information is not involved, as in, for example, a series of studies by Hew and his colleagues (see also Burke & Harrold. 1988). Subjects in Hess (1984)and Hess and Higgins (1983)studied homophones paired with one or two words related to one of the homophone's meanings. An example is shown in Table 1. Both young and older adults recognized homophones better with the original context words rather than with different context words related to the other meaning of the homophone target. This finding suggests that the original context words influenced encoding of the homophone meaning and thus is inconsistent with the hypothesis that older adults only engage in "general" encoding. With general encoding, homophone recognition should have been the same for the original and Werent context words: encoding at study would be independent of specific context and thus would involve one of the two general meanings of the homophone with equal probability (assuming they were equally available). Under the NST, however, the effect of the context words on the homophone should be unimpaired for older adults because the homophone and context words are semantically related and have pre-established connections that can be strengthened through engrainment learning. Older adults' recognition did in fact show a decrement compared to young adults' of about the same magnitude in both conditions because new connections to the experimental context were required in both conditions and this commitment process exhibits age-related impairment under the commitment learning principle. Finally, when subjects studied target words with a n unrelated context word, young but not older adults had better recognition of the target with the ortgfnal unrelated context word rather than a changed context word (Hess, 1984). Here contextual effects depend on establishment of new connections and the fact that the original context failed to improve older adults' recognition is consistent with the commitment learning principle. Semantic versus Episodic Priming Effects It is well established that a target word (e.g., doctor) can be processed more rapidly when it is immediately preceded by a
Cognitton and Aging
239
Table 2 Studies of Semantic h-imfng in Young and Older Adults
Authors
Task
Balota & Duchek, 1988 Bowles & Poon. 1985 Burke, White & Diaz, 1987
pronunciation (SOA) lexical decision lexical decision (SOA; Auto -Atten) lexical decision pronunciation lexical decision (Auto-Atten) lexical decision exical decision Stroop task lexical decision lexical decision (SOA; Auto-Atten) lexical decision lexical decision pronunciation
Burke &Yee. 1984 Cerella & Fozard, 1984 Chiarello. Church & Hoyer, 1985 G. a h e n & Faulkner, 1983
Howard. 1983 Howard, Lasaga & McAndrews, 1980 Howard, McAndrews & Lasaga, 1981 Howard, Shaw & Heisey, 1986 Madden, 1986 Madden, 1988 Nebes, Boller & Holland, 1986
Note: Studies are indicated that varied stimulus-onset-asynchrony(SON between prime and target and/or attempted to isolate automatic and attentional activation (Auto-Atten).
semantically associated word (e.g., nurse). This speeding of response time in, for example, lexical decision or pronunciation tasks has been called the semantic priming effect because it is believed that its source is a spread of activation from the prime word to representations that are associated in memory, temporarily increasing their accessibility (e.g., Collins & Loftus, 1975). A number of studies (see Table 2) have demonstrated that this "semantic priming" effect is at least as large in older as young adults. This holds for studies that were designed to detect age
240
MacKay and Burke
differences in the speed of semantic priming by varying stimulus onset asynchrony (SOA), the temporal intexval between prime and target. and in automatic versus attentional priming. The one exception to this is Howard; Shaw and Heisey (1986) who found significant semantic priming effects for young but not older adults at their shortest SOA, 150 ms. Other studies. however, have provided no evidence for age differences in speed of semantic priming (Balota & Duchek. 1988: Burke, White h Diaz, 1987). The fact that semantic priming effects remain constant with age is consistent with the commitment learning principle because semantic priming involves existing rather than new connections. Activation of the prime results in the spread of primlng via the many already established connections between related words, which can then be activated more rapidly. Under some conditions, however, there should be small age differences in priming of established nodes because the Transmission Deficit hypothesis postulates age-linked declines in rate or amplitude of priming. There are several reasons why within the NST such age differences are not seen in semantic priming studies. First. age deficits would be more apparent with single rather than multiple sources of priming (see below). Within the semantic system many connections link nodes to one another (see also Jones, 1985) and multiple sources of priming diminish the likelihood of demonstrating Transmission Deficit. Further, the speed and amount of priming transmitted depends on the linkage strength of the connections. Linkage strength increases with practice (see MacKay, 1982) and older adults have accumulated more practice than younger adults with common words such as doctor and nurse and their interrelations (the connections between them).2 According to the commitment learning principle, the engrainment learning process resulting from practice is relatively unimpaired by aging. Intuitively one might think that practice effects for common words would asymptote at some point with age, ruling out the possibility of further age-linked practice effects for such words as doctor and nurse. However, the effects described hem depend on the interrelations between words via connections to predicate nodes representing concepts such as, "workin hospitals"and "give injections". Unlike low level phonological nodes, these proposition and predicate nodes are extremely unlikely to acquire asymptotic levels of practice, even over the lifespan of older subjects.
Cognition and Aging
241
Consistent with this, the rate of improvement with practice for a variety of already established behaviors is similar across age (see Charness, 1985: Welford. 1985). Thus, older adults' greater linkage strength for common words and the connections between them would tend to offset an age-linked Transmission Deficit. However, even if transmission of semantic priming in fact varies with age, this paradigm may be insufficiently sensitive for demonstrating such a n age difference. Older adults typically have longer absolute response times than young adults, providing a longer time interval during which priming can spread from the prime to the target, even when the prime-target SOA is controlled. This longer intenral from prime presentation to target response could compensate for a slower rate of priming in older adults (see Howard, 1988a). A more accurate picture of the time course of processing may be gained by using a deadline procedure in which subjects are induced to respond at particular points in time after presentation of the target (cf. Wickelgren, 1977; McClelland, 1979). In short, an analysis of the semantic priming paradigm within the NST suggests that this paradigm is a dfmcult testing ground for the Transmission Deficit hypothesis. Consider now the analogous paradigm that has been developed to test priming effects when the connection between prime and target is episodic or newly formed, so called episodic priming. For example, Howard (1988a. 1988b).Howard, Heisey, and Shaw (1986) and Rabinowitz (1986) asked young and older adults to study word pairs (e.g.. card-water)or sentences (e.g., The dragon ate thefudge) and then to recognize words as old or new, presented one at a time. An "episodic priming effect" is said to occur when target recognition is faster following a word from the same pair or sentence (e.g., fudse following dragon],than from a different pair or sentence. Because episodic priming effects require the formation of new connections between word concepts, and because these connections are more difficult for older adults to establish under the commitment learning principle, the NST predicts an age-related deficit for this task. However, results from episodic priming studies generally show comparable episodic priming effects for young and older adults, but poorer recognition accuracy for older adults. How are these
242
MacKay and Burke
surprising results to be explained? First, as with semantic priming tasks, older adults typically have longer absolute response times than young adults and this greater interval between the prime and the response to the target could compensate for a slower transmission of episodic priming. Second, these results are probably a consequence of the methodological requirement of assessing episodic priming only when a target and its preceding prime word have both been correctly recognized. Within the NST, accurate recognition that a target word is "oldqdepends on the formation of a new connection between the target word and the experimental context. This new connection can be seen in Figure 8 , where uncommitted nodes marked X connect the "was presented" node with the targets dragon andfudge. Episodic priming within the NST depends on the formation of another new connection, this time between the prime and target, in Figure 8 , for example, via a n uncommitted node X connecting dragon (prime)withfudge (target). It is consistent with the NST that when a target has been studied sufficiently to be correctly recognized as having been presented, that is, to have a committed node (X) connecting it to the experimental context, then the target is likely to have been studied sufficiently to show episodic priming, that is, to have a committed node (X) connecting it to the prime. However, with brief study time there is some evidence for an age-related deficit in episodic priming. Under such conditions, Howard, Heisey, and Shaw (1986) observed episodic priming with two nouns from the same sentence for young but not older adults. Recognition accuracy for individual words decreased with study time for both age groups but showed no age difference. This suggests that, under these conditions, older adults have greater difTiculty forming new connections between words in the sentence (for demonstrating a n episodic priming effect) t h a n forming connections between the words and the experimental context (for demonstrating recognition that the word had been presented). Rabinowitz ( 1986), however, obtained even lower recognition accuracy than Howard et al.. with age differences in recognition but not episodic priming. This variation may reflect the subjects' strategies for learning the stimuli, that is. whether they emphasize
243
Cugnitlon and Aging
DRAGON AND FUDGE WERE PRESENTED
DRAGON WAS PRESENTED
Q
@..-.-. I
I
*.
I
I
%*.
I I
DRAGON
b ',
;
%* ;
).
, :'
I
(NOUN PHRASE) i DRAGON ANDFUDGE
i I
', ,, , \
,
I
:
'\
*. *..-
#'
I
1 I
I
#
I
#
I
I
''t
*-.
/ I : ; I
-..o , #
I
; %. ; *.
1
FUDGE WAS
I
: i (VERB i PHRASE) II
1 ',
WAS PRESENTED
FUDGE
Figure 8. A sample of nodes representing "dragon,""fudge.""was presented,"
and their connections. The uncommitted node making a new connection between "dragon"and "fudge"is marked with an X and between each word and "was presented"with an X. X'marks the uncommitted node connecting "was presented and "dragonand fudge."
more the association between words or the association with the experimental context. The commitment learning principle is consistent with the pattern of age daerences found in Howard's recall and recognition data, namely, age decrements in cued recall and paired recognition without age decrements in episodic priming and single word recognition (Howard, 1988b; Howard, Heisey & Shaw, 1986). Cued
244
MwKay and Burke
recall (i.e.. What noun occurred in the same sentence as the cue?) and paired recognition (i.e., Did both words occur in the same sentence?) each require a new connection between the node linking the fkst word (e.g., dragon] with the second word (e.g..fidge)),and the node for experimental context. This new connection is made through the node marked X" in Figure 8. Thus, cued recall and paired recognition in these studies required commitment of more nodes (Xand X')than did single word recognition (X) or episodic priming tasks (XI. Direct versus Indirect Memory Tasks A currently popular dimension for describing memory tasks is
the involvement or non-involvement of conscious recollection. Episodic priming tasks and repetition priming tasks that show effects of prior experience without seeming to require conscious recollection of those experiences have been called indirect memory tasks. Cued recall or recognition tasks that seem to require conscious recollection have been called direct memory tasks. The dissociation between performance on direct versus indirect memory tasks in normal and amnesic subjects has been viewed by some as evidence that separate and fundamentally different memory systems 0.e.. explicit and implicit memory, respectively) underlie performance on each type of task (N.J.Cohen, 1984; Tulving, 1985; see Schacter, 1987 for a recent review). Applied to aging research. this theoretical approach has raised the question of whether these hypothesized memory systems exhibit differential aging effects (Chiarello & Hoyer. 1988; Light & Burke. 1988; Mitchell, 1989). However, a single memory system suffices to explain age differences within the present theoretical approach. Further, our approach extends to a broad range of memory phenomena and is not limited, in principle, to explaining age differences in direct versus indirect memory tasks. Predictions of the commitment learning principle for indirect tasks parallel predictions of the NST for practice effects in general: repetition increases linkage strength of existing connections Ii.e., causes engrainment learning) and the rate of increase should be unaffected by aging. The larger age
Cognition and Aging
245
deficits obtained in direct than indirect memory tasks (e.g.. Howard, 198813: Light & Singh, 1987) reflect the greater number of new connections typically required in direct tasks. However, the commitment learning principle predicts age differences for indirect memory tasks insofar a s they require new connections. A similar analysis of direct and indirect memory tasks in terms of the degree to which new associations are involved has been suggested in research on amnesia. Shimamura and Squire (1989) have recently demonstrated. for example, that amnesics who showed preserved semantic and repetition priming, were impaired in priming involving new associations. Further, the size of priming effects for new associations was related to the degree of residual direct memory ability (see also Schacter & Graf, 1986b). Similarly, the commitment learning principle predicts agerelated decrements on indirect memory tasks when new learning is involved. Unfortunately, we know of only one study with data relevant to this prediction, Fry and Howard, cited in Howard (1988b). Subjects were required to complete a word stem (e.g., STAJ with the first word that came to mind after generating sentences using unrelated word pairs (e.g., QUEEN-STAIRS). When the word stem was presented with a word from the original pair, subjects used the target word to complete the stem more often than when a different word preceded. This effect, demonstrating a new connection between the two words, was the same for young and older adults, although older adults had poorer cued recall. However. under more impoverished or time constrained study conditions, the effect was obtained for young but not older adults. Thus, age deficits have been obtained in both episodic priming tasks and word stem completion tasks requiring new connections, under conditions that reduce learning. The age deficit in direct recall is consistently more pronounced because this task requires more new connections than do indirect tasks. Finally, even young adults show decreased effects of lexical context on word stem completion under study conditions that reduce new learning (e.g., Graf & Schacter, 1985; Schacter & Graf, 1986a).
246
MacKay and Burke
Summary and General Predictions
To summarize our discussion of commitment learning, we have outlined an explicit mechanism for new connection formation. and have argued that the deflcit in transmission of priming that proved useful in explaining other effects of aging is especially likely to impair the occurrence of commitment learning. Engrainment learning resulting from practice or repetition via pre-existing connections involves a diEerent mecharam in the theory. namely changes in linkage strength following activation, and is unaffected by aging. Our analysis of selected memory tasks in terms of the new versus old connections involved explained available results and provided a basis for new predicitons. All other factors being equal, the commitment learning principle predicts that varying the number of new connections required in a task will affect the size of age differences independent of distinctions between direct versus indirect memory or general versus specific encoding, or even efYortfu1versus automatic processes, insofar as these distinctions differ from the one between engrainment and commitment learning.
TOTS: AN EXCEPTION TO THE COMMI'ITMENT LEARNING PRINCIPLE? TOTS occur when a speaker retrieves the meaning of a familiar word, but is unable to retrieve its phonology. Such retrieval failures are of interest to the field of cognitive aging because they increase with age (Burke et al, 1989). and because they provide an important counterexample to the popular view that age-related declines are primarily in the ability to use or remember new information (fluid intelligence) (see. for example, Horn, 1982;Light, 1988). T W h indicate that even highly familiar information can be more dimcult to retrieve with age. However, TOTs do not contradict the commitment learning principle that age deficits increase with the amount of commitment learning required for accurate performance. Rather. both TOTs and the commitment learning principle can be explained as derivatives of the Transmission Deficit hypothesis: TOTs occur when already established
Ccylnttbn and Aging
247
connections for infrequently or non-recently retrieved words become weakened due to age and infrequent or unrecent activation, and transmit reduced priming to phonological nodes. As discussed below, this instantiation of the Transmission Deficit hypothesis provides a detailed explanation of how TOTs originate and are resolved in young and older adults. The Origin of TOTs Within the NST TOTs originate under the NST when a lexical node becomes activated, providing access to semantic information about the target word, but at least some of its connected phonological nodes remain unactivated because insufficient priming is transmitted to enable activation. a transmission deficit related to frequency and recency of use. For example, a TOT for uelcro would occur if the lexical node velmo(concrete noun) becomes activated but transmits too little priming to enable some of its connected phonological nodes to become activated (see Figure 9). FZequency and Recency of Use Why do most TOTs involve words that are rarely and/or not recently used (see Burke et al., 198911 The reason under the NST is that frequency and recency of activation influence the rate and amount of priming transmitted across a connection and connections to the phonological nodes of rarely and unrecently used words become so weak as to transmit insufficient priming to enable activation. A similar explanation holds for why object naming is faster for high than low frequency names in normal individuals (e.g., Huttenlocher & Kubicek, 1983; Oldfield & Wingfield. 1965) and aphasics (e.g., Tweedy & Schulman. 1982). and for why phonological errors are more common for low than high frequency words, both experimentally induced (e.g., MacKay. 19701 and spontaneously occuring (e.g., Stemberger & MacWhinney. 1986). The lexical node for a . high frequency word will have stronger connections to its phonological nodes. These stronger connections will tend to reduce phonologfcal errors, which occur when the correct node has acquired less priming than some other,
248
MMay and Burke
inappropriate node in the same phonological domain when the activating mechanism is applied (see Burke et al, 1989).
Resolution of TOR3 TOT words often pop up or come to mind spontaneously, without
conscious attempts at retrieval. However, because the weak connections that originally cause TOTS under the NST reflect low production frequency and unrecent use, there is no reason to expect these particular connections to recover spontaneously. Rather, a n externally delivered boost in priming to the appropriate phonological nodes is required for the target to pop into mind. Consistent with Yaniv and Meyer's (1987) account of incubation effects in both problem solving and language production, Burke et al. [1989) suggest that this boost may arise when the l D T target is no longer in mind and the critical phonological components occur accidentally during internal speech and/or everyday language comprehension. This unconsciously processed internal or external stimulus increases the activation levels of the relevant nodes so that the pop-up can occur. Persistent Alternates Persistent alternates refer to words that come repeatedly and involuntarily to mind instead of the target word, even though the subject rejects the persistent alternate as inappropriate. An example from Burke et al. (1989) is dacron, a word that came repeatedly to mind instead of the TOT target, uelcro. As in this example, persistent alternates are usually similar to the target word in sound and meaning, and share the same domain or syntactic class as the target, Persistent alternates also increase the time required for the spontaneous resolution of a TOT, even though subjects who report more alternates also tend to recall more information about the target such as its initial phoneme and how many syllables it has (see Burke et al.. 1989). The NST readily explains these characteristics of persistent alternates. By way of illustration, the sample of nodes in Figure 9 shows how the persistent alternate dacron might come initially to
Cognition and Aging
249
mind instead of the TOT target, wlcro. The lexical node for uelcro is connected with a set of predicates such as, say, "is made out of nylon'' and "is used as a fastener", some of which are shared with other words such as dacron. For some subjects, for example, w h o may share the predicate "is made out of nylon'' with the word dacron (see Figure 9). This shared predicate therefore transmits "top-down" priming to the lexical node for d a c r o n when uelcrdconcrete noun) becomes activated. Moreover, dacron will receive additional priming "bottom-up" from some of uelcro's phonological nodes. Recall that the TOT for uelcro occurs because of a transmission deficit: some of its phonological nodes receive too little priming to become activated, whereas others such a s the node for the consonant cluster cr do receive sufficient priming and in fact become activated. These activated phonological nodes will transmit strong priming bottomup to all lexical nodes containing these components, including, for example, dacron(concrete noun). Top-down and bottom-up priming therefore summates on dacron(concrete noun) which may become activated as the most primed node in the (concrete noun) domain when the activating mechanism for this domain is next applied. This summation-ofpriming factor explains why targets and alternates are phonologically and semantically similar and why alternates share the same domain (syntactic class) as the target: activation occurs by lexical domain and repeated attempts to retrieve the TOT target would yield (activate) alternates only in that domain. This summation-of-priming process also predicts that factors tending to increase the priming delivered to alternates, such as augmented strength of connections due to high frequency of use of the alternate, should increase the probability that the alternate will persistently intrude. Finally, subjects who report more persistent alternates also report more information about the target because transmission of priming influences both of these variables: retrieving partial information about a target depends on how much priming is transmitted, and so does retrieving a persistent alternate.
MacKay and Burke
250
F IS MADE OP NYLON
FBUSEDAS A FAsrENER
SEMANTIC SYSTEM DACRON (noun)
(streseed syllable)
(
W
VELCRO (noun)
M
syllable)
SYSTEM
(initial
EL
CR
(vowel
(initial
consowrt)
pup)
coruKn\ant)
V
0 (vowel)
Flgure 9.
Illustration of the spread of priming preceding retrieval of a persistent alternate (dacronl for the TOT (who).
How Age Influences TOTS Theories of cognitive aging must explain why resolution times increase with age whereas persistent alternates and information available about the TOT target decrease with age (see Burke et al.. 1989). All three phenomena are consistent with the Transmission Deficit hypothesis. If aging reduces the priming delivered to nodes
Cognftfonand Aging
251
in the phonological system. phonological information would be less primed, leading to reduced probability of activation and reduced information retrieval about the target, a s observed. Older adults would also retrieve fewer alternates (as observed) because bottom-up priming from phonological nodes and top-down priming from shared predicates would both be reduced. Finally. pop-ups would take longer for older adults (as observed) because more external input would be required to raise priming levels to a point where the target nodes could be activated. The Transmission Deficit hypothesis therefore explains how TOTs arise and are resolved in general, and how TOTS change with age. The Single-source Factor: Some New Predictions The Transmission Deficit hypothesis makes some clear predictions regarding conditions under which age differences are likely to be most pronounced, all other factors being equal. For example, age differences are likely to be pronounced when a node critical to a task receives priming from only a single source or connection within the network. Age-linked transmission deficits are very likely to affect performance in such a task because no other sources of priming will be able to offset the reduced priming across that critical connection. The single-source factor and the concept of a critical connection are best illustrated by examining the priming of phonological nodes during attempts to produce a TOT word. In general, higher level phonological nodes have only a single or critical top-down connection and so receive a single source of priming without the possibility of convergence or priming summation. A syllable node. for example, receives top-down priming from a single lexical node during production (see Figure 2). As Burke et al. (1989)point out, this limitation in the number of sources that can deliver priming to phonological nodes during language production may explain why phonological but not semantic information is inaccessible in TOTs: Reverse TOTs where someone can produce or recognize the sequence of sounds for a familiar word but cannot access its lexical node (so a s to indicate that the sounds represent a known word) are extremely unlikely in the theory: a lexical node receives too many
252
MacKay and Burke
sources of priming from phonological and other nodes to suffer a transmission deficit. And as predicted, such reverse TOTS have never been observed or reported. However, this single-source factor does not apply in the semantic priming paradigm, discussed above, a fact that may help explain why age differences are so small and difficult to obtain using this task. Nor does the single-source factor apply in the case of everyday word perception, where priming converges or summates bottom-up onto a single lexical node from many phonological nodes (see Figure 2) and perhaps also from nodes in other sensory and conceptual systems, For example, we can activate the node for dog(noun) on the basis of convergent priming arriving from the sight, sound, touch, and perhaps even smell of a dog, as well as from the process of imagining any of these characteristics. The many propositions that we have stored about dogs can also contribute convergent priming for activating dog(noun)(Jones, 1985). The single-source factor in production suggests an interesting new prediction: the asymmetry between the maximal rate of production vs. perception seen in young adults (MacKay. 1987)will be accentuated in the case of older adults. That is, young adults can perceive speech at a faster rate than they can produce it, an asymmetry due in part to the single-source factor in production: priming comes from a single source in production (via the divergent or one-to-many top-down connections in action hierarchies), but comes from several convergent sources in perception (via the summating or many-to-one bottom-up connections in perceptual hierarchies). The maximal rate of perception is determined empirically by presenting speeded or time-compressed speech using electromechanical devices that systematically sample and compress acoustic signals so as to provide a wide range of acceleration without introducing pitch changes. Young adults can perceive connected paragraphs accelerated in this way at rates up to 400 words per minute (about 20 to 30 ms. per phoneme), but their maximal rate of producing speech is much slower (see MacKay. 1987). The prediction under the Transmission Deficit hypothesis is that this asymmetry will be exaggerated in older adults: the ratio of
Cognition and Aging
253
maximal perception rate to maximal production rate will be greater for older adults than for younger adults. AGING AND ERROR DETECTION: F U m E R NEW PREDICTIONS
Some of the evidence cited earlier as consistent with the NST connection formation mechanism might also be consistent with other perspectives in the literature, for example, Hebb's distinction between type A versus type B intelligence, and the Cattell-Horn distinction between fluid versus crystalized intelligence (Horn & Cattell, 1966). As Salthouse (1985) points out, "age effects are generally greatest on tasks requiring the acquisition or transformation of information (sometimes referred to a s fluid intellectual activities), but are minimal to non-existent on tasks involving the retrieval or utilization of prior knowledge (p.2)". However, the NST specifies how old connections are used, and how and when new connections are acquired, whereas earlier concepts such as the transformation versus the utilization of information have never been clearly specified (see Light, 1987). Moreover, unlike the NST, these earlier formulations concerning new knowledge have difficulty with age-linked deficits involving wellestablished knowledge, for example, the TOT phenomenon. Nonetheless, the question arises whether connection formation processes in the NST predict any genuinely new and surprising phenomena that are not closely related to already established empirical phenomena. What follows is an answer to that question: a description of some new and untested predictions that follow logically from the Transmission Deficit hypothesis and the NST account of error detection, discussed below. The NST Account of Error Detection Explaining how slips of the tongue are detected requires no new mechanisms, and follows directly from the mechanisms for connection formation and awareness in the NST. As MacKay (1990) points out. three conditions are necessary for awareness under the NST:
254
MacKay and Burke
The Novelty Condition: Under the Novelty Condition, two or more nodes that have never been activated in conjunction before must become conjointly activated. For example, the expression 'pertinent novelty' illustrates the novelty condition if the speaker/listener h a s never encountered this particular combination of adjective and noun before. 1)
2) The Pertinence Condition: Under the Pertinence condition, the domains containing the novel nodes must be familiar and sequentially related. For example, the expression 'pertinent novelty' illustrates the pertinence condition if the speaker/listener knows that 'pertinent' is an adjective, that 'novelty' is a noun, and that adjectives followed by nouns make up noun phrases in English.
3) The Strength-of-Priming Condition: Under the Strength-ofPriming Condition, the novel and pertinent node(s) activated in error must deliver sufficient priming to an uncommitted node to trigger the binding nodes for prolonged activation (i.e., awareness of the error) and orienting reactions (that halt ongoing production and enable error correction). We argue below that this third condition provides the basis for an age-related impairment in error detection under the 'Ikansmission Deficit hypothesis.
Slips of the tongue automatically meet the novelty and pertinence conditions for awareness or detection: Slips always involve the conjoint activation of two or more nodes that have never been conjointly activated before (i.e.. novelty) and the domains containing the nodes activated in error are familiar and sequentially related 0.e.. pertinent) because errors do not alter the original [intended) syntax or sequential domains of an utterance. That is, as discussed above, errors occur under the N S T when a node fails to achieve greatest priming in its domain when the activating mechanism is applied, and some other "extraneous" node in the domain that receives more priming becomes activated. As a result, when slips occur, verbs interchange with other verbs, adjectives interchange with other adjectives, and nouns interchange with other nouns, a s when "a full tank of gas" is misproduced as, "a full
Cognition and Aging
255
gas of tank". However, words from different domains or syntactic categories almost never interchange in speech errors. For example, adjectives almost never interchange with nouns, a s in the hypothetical error, "a full tank of gas" misproduced as "a gas tank of full". This same generalization also holds for phonological domains and is known a s the sequential class regularity (MacKay, 1979;Fromkin. 1973;A. Cohen, 1967). As already noted, these mechanisms underlying the sequential class regularity also underlie the syntactic similarity phenomenon observed when I "targets (see above). persistent alternates substitute for ' However. novelty and pertinence are not sufficient for awareness, and speech errors often pass unnoticed because they fail to meet the third condition for awareness. That is, nodes activated in error often deliver insufficient priming to an uncommitted node to trigger the binding nodes for prolonged activation and awareness. In the case of young adults, the reason for the insufficient priming is that a large number of connections sometimes intervenes between a unit activated in error and the novel unit (uncommitted node) that it primes. Consider for example the three Spoonerisms or phonological transposition errors (from Motley, Baars, & Camden, 1983)in Table 3. When crawl space is misproduced as crawl srace, an uncommitted or novel phonological unit representing syllable-initial sr becomes strongly primed because there exists no committed node representing syllable-initial sr in English. The priming on this uncommitted node is strong because no other nodes intervene between the nodes activated in error (the nodes representing the consonants /s/ and /r/) and the uncommitted node. As a result, young adults are very likely to detect this error. Turning to the second example where dump seat is misproduced as sump &at, an uncommitted or novel lexical node becomes primed because there exists no committed node representing the word deat in English. However, the priming on this uncommitted node will be weak because a committed node representing the syllable &at intervenes between the nodes activated in error (the nodes representing the consonants /s/ and /d/) and the uncommitted node at the lexical level, so that young adults are less likely to detect this error. Turning to the third example, where tool carts is misproduced as
256
MatKay and Burke
Table 3 The Probability of Detection Predicted Under the NST for Three Spoonerisms or Phonologicql Transposition Errors.
Distance
Predicted Probability of Detection
crawl space -----> crawl scrace
0
high
dump seat
-----> sump deat
2
moderate
tool carts
-----> cool tarts
4.5'
low
Intended
Error
Note: Distance represents the number of connections intervening between a node activated in error and the first uncommitted node that it conjointly primes. The arrow stands for "was misproduced as". The asterisk indicates an ambiguity with respect to level (see text for explanation).
cool tarts in the sentence, 'They were moving tool carts down the assembly line", an uncommitted or novel proposition node becomes primed because the speaker is unlikely to have a committed node representing the proposition, "They were moving cool tarts down the assembly line." Thus, the priming on this uncommitted node will be extremely weak because about four other committed nodes intervene between the nodes activated in error (again the nodes representing /s/ and /d/) and the uncommitted node at the proposition level, s o that this error is very likely to pass undetected. As noted in Table 3, however, the level of the uncommitted node is ambiguous in this example: if the speaker lacks a committed node for the phrase, cool tarts, then commitment of a novel phrase node rather than a novel propositional node is called for.
Cognition and Aging
257
The New Prediciton: Age and the Detection of Speech Errors From the Strength-of-Priming condition and the examples above it is clear that probability of error detection is directly related to strength of priming. This means that a deficit in the transmission of priming will reduce the probability of error detection. This means that, all other factors being equal, errors will be detected with higher probability in young than in older adults under the Transmission Deficit hypothesis. CONCLUSIONS Although we have applied the NST and the Transmission Deficit hypothesis to only a few cognitive aging phenomena in the present paper, the theory cuts across paradigm-specific boundaries and seems applicable to a wide range of tasks, some of which are new or relatively unexplored within the field of cognitive aging, for example, time-compressed speech perception. Moreover, we have only spelled out some of the predictions that follow from the theory: there are others. For example, the NST predicts that effects of ambiguity on errors induced by delayed auditory feedback (see MacKay, in press) should decline with age, and so should "perception without awareness" or subliminal perception effects (see MacKay. 1990) and the verbal transformation effect (see MacKay, 1987). At this stage, however, testing or even spelling out all of these novel predicitons may not be as important as addressing some of the other goals underlying development of a theory: to organize and integrate the existing literature in the research traditions of the field, and to establish priorities for future research within those traditions (see Salthouse, 1988). ACKNOWLEDGMENTS The authors thank Donald Kausler. Thomas Hess and two anonymous reviewers for helpful comments on earlier versions of this paper, and Nancy Woolf for providing neurophysiological references bearing on the Transmission Deficit hypothesis. Portions of this chapter were presented at the Cognitive Aging Conference in Atlanta, April, 1988. Support for this research was
258
MacKay and Burke
provided in part by grant number AGO2452 from the National Institute on Aging. REFERENCES Anderson, J. R. (1983). A spreading activation theory of memory. Journal of Verbal Learning and Verbal Behaulor, 22, 261-295. Aston-Jones, G., Rogers, J., Shaver, R D., Dinan, T. G., & Moss, D. E. (1985). Age-impaired impulse flow from nucleus basalis to cortex. Nature, 318,462-464. Balota, D. A., & Duchek, J. M. (1988). Age-related differences in lexical access, spreading activation, and simple pronunciation. P~y~hology and Aging, 3, 84-93. Birren, J. E. (1965). Age changes in speed of behavior: Its central nature and physiological correlates. In A.T. Welford & J. E. Birren (Eds.), Behaulor, aging and the nervous system (pp. 191216). Springfield. IL: Charles C. Thomas. Bowles, N. L.. Obler, L. K., & Poon. L. W. (1989). Aging and word retrieval: Naturalistic. clinical and laboratory data. In L. W. Poon. D. C. Rubin & B. A. Wilson (Eds.), Everyday cognition in adulthood and late lge. Cambridge: Cambridge University Press. Bowles, N. L., & Poon, L. W. (1985). Aging and retrieval of words in semantic memory. Journal of Gerontology. 40, 71-77. Burke, D. M., & Harrold, R. M. (1988). Automatic and effortful semantic processes in old age: Experimental and naturalistic y approaches. In L.L. Light & D.M Burke (Eds.),Language, m and aging (pp.100- 116). New York Cambridge University Press. Burke, D. M. & Light, L. L. (1981). Memory and aging: The role of retrieval processes. Psychologtcal Bulletin, 90, 5 13-546. Burke, D. M.. White, H., & Diaz, D. L. (1987). Semantic priming in young and older adults: Evidence for age constancy in automatic and attentional processes. Journal of Experimental Psychology: Human Perception and Perfonnunce, 13, 79-88. Burke, D. M., Worthley, J. S. & MacKay, D. G. (1989). On the ttp of the tongue: Why do wordjhding impatrments increase with age? Manuscript submitted for publication. Burke, D. M., & Yee. P. L. (1984). Semantic priming during sentence processing by young and older adults. Developmental Psychology, 20,903-910. Cerella, J., & Fozard, J. L. (1984). Lexical access and age. Deuelopmental Psychology, 20,235-243.
Cognition and Aging
259
Charness, N. (1985).Aging and problem-solving performance. In N. Charness (Ed.), Aging and human performance (pp. 225-2591, Chichester: John Wiley & Sons. Chiarello, C., Church, K. L., & Hoyer, W. J. (1985).Automatic and controlled semantic priming: Accuracy. response bias, and aging. Journal of Gerontobgy, 40, 593-600. Chiarello, C., & Hoyer, W. J. (1988).Adult age differences in implicit and explicit memory: Time course and encoding effects. P~ychologyand A Q ~3,358-366. , Cohen, A. (1967). Errors of speech and their implication for understanding the strategy of language users. Zeitschr?ft f w r P h o ~ t f k22, , 177-181. Cohen, G . . & Faulkner, D. (1983).Word recognition: Age differences in contextual facilitation effects. British Journal of Psychology, 74,239-251. Cohen. G., & Faulkner. D. (1986). Memory for proper names: Age differences in retrieval. British Journal of Developmental P~y~hology, 4 , 187-197. Cohen. N. J. (1984). Preserved learning capacity in amnesia: Evidence for multiple memory systems. In L. R Squire & N. Butters (Eds.), Neuropsychology of memory (pp. 83-103). New York Guilford Press. Collins, A M.,& Loftus, E. F. (1975). A spreading-activation theory of semantic processing. Psychologfcd Reufew, 82.407-428. Craik, F. I. M. (1983). On the transfer of information from temporary to permanent memory. Philosophical Transactions of the Royal Society of London, B302.341-359. Craik. F. I. M., & Simon, E. (1980).Age differences in memory: The roles of attention and depth of processing. In L. W.Poon, J. L. Fozard, L. S. Cexmak. D. Arenberg, & L. W. Thompson (Eds.),New Directions in Memory and Aging: Proceedings of the George Talland Memorial Conference (pp. 95- 112). Hillsdale, NJ: Erlbaum. Decker, M.W. (1987). The effects of aging on hippocampal and cortical projections of the forebrain cholinergic system. Brain Research Reviews, 12,423-438. Dorfman. L.J. & Bosley. T.M. (1979). Age-related changes in peripheral and central nerve conduction in man. Neurology, 29, 38-44.
Fromkin. V. A. (1973).Introduction. In V. A. Fromkin (Ed.), Speech errors as lfngulstfc eutdence (pp. 11-45)Paris: Mouton.
260
MacKay and Burke
Graf,P., & Schacter. D. L. (1985). Implicit and explicit memory for new associations in normal and amnesic subjects. Joumal of Experimental Psychology: Learnfng Memory and Cognitton. 1 1. 501-518. Hess, T. M. (1984). Effects of semantically related and unrelated contexts on recognition memory of different -aged adults. Journal of Gerontology, 39,444-45 1. Hess, T. M., & Higgins, J. N. (1983).Context utilization in young and old adults. Journal of Gerontology, 38, 65-71. Horn, J. L. (1982). The aging of human abilities. In B. Wolman (Ed.), Handbook of developmental psychology (pp. 847-870). Englewood ClifTs. NJ: Prentice-Hall. Horn, J. L., & Cattell, R. B. (1966). Age differences in primary mental ability factors. Journal of Gerontology, 21, 2 10-220. Howard, D. V. (1983). The effect of aging and degree of association on the semantic priming of lexical decisions. Experimental Aging Research, 9,145-151. Howard, D. V. (1988a). Aging and memory activation: The priming of semantic and episodic memories. In L.L. Light & D.M Burke (Eds.), Language, memory and aging. (pp.77-99). New York: Cambridge University Press. Howard, D. V. (1988b). Implicit and explicit assessment of cognitive aging. In M. L. Howe & C. J. Brainerd (Eds.), Cognitive Development in Adulthood: Progress in cognfttve development research (pp.3-37). New York Springer-Verlag. Howard, D. V., Heisey, J. G.,& Shaw, R. J. (1986). Aging and the priming of newly learned associations. Developmental P~ychol~gy, 22, 78-85. Howard, D. V., Lasaga, M. I., & McAndrews, M.P. (1980). Semantic activation during memory encoding across the adult lifespan. Journal of Gerontology,35, 884-890. Howard, D. V., McAndrews, M. P., & Lasaga, M. I. (1981). Semantic priming of lextcal decisions in young and old adults. Journal of G ~ ~ ~ n t o l36, o g707-7 ~ , 14. Howard, D. V., Shaw, R S.,& Heisey, J. G.(1986).Aging and the time course of semantic activation. Journal of Gerontology, 41, 195203. Huttenlocher, J., & Kubicek, L. F. (1983).The source of relatedness effects on naming latency. Journal of Experimental PSyCholOgy, 9, 486-496. Jones, G. V. (1985). Deep dyslexia, imageability, and ease of prediction. Brain and Language, 24, 1- 19.
Cognition and Aging
261
Landfleld, P. W. (1988).Hippocampal neurobiological mechanisms of age-related memory dysfunction. Neurobiology of aging. 9, 571-579. Landfield. P. W. & Lynch, G. (1977). Impaired monosynaptic potentiation in in uitro hippocampal slices from aged memory deficient rats. Journal of Gerontology, 32, 523-533. Light, L. L. (1987).Review of A theory of cogntttue agtng by T.A. Salthouse. American Journal of Psychology, 100, 125-132. Light, L. L. (1988).Language and aging: Competence versus performance. In J. E. Birren & V. L. Bengtson (Eds.), Emergent theories of aging (pp. 177-213). New York Springer. Light, L. L., & Burke, D. M. (1988).Patterns of language and memory In old age. In L. L. Light & D. M. Burke (Eds.).Lungucige, memoy and agfng (pp. 244-271).New York Cambridge University Press. Light, L. L.. & Singh, A. (1987).Implicit and explicit memory in young and older adults. Journal of Experfmental Psychology: Learning Memory and Cognition, 13, 531-541. MacKay, D. G. (1970).How does language familiarity influence stuttering under delayed auditory feedback? Perceptual and MOtW Skilts, 30, 655-669. MacKay. D. G. (1979).Lexical insertion, inflection and derfvation: creative processes in word production. Journal of Psycholtrtgulstfc Research, 8. 477-498. MacKay, D. G. (1982).The problems of flexibility, fluency, and speed-accuracy trade-off in skilled behavior. Psychological Reviao, 89,483-506. MacKay, D. G. (1987).The organization of perception and action.: A theory for language and other cognitive skills. New York: Springer-Verlag. MacKay, D. G. (1988).Under what conditions can theoretical Psychology survive and prospex? Integrating the rational and empirical epistemologies. Psychological Review, 95, 559-565. MacKay, D. G. (1990).Perception, action, and awareness: A three body problem. In W. Prim. & 0. Neumann (Eds.), Relattonshtps between Perception and Action (pp. 269-303).Berlin: SpringerVerlag. MacKay, D. G. (in press). Errors, ambiguity and awareness in language perception and production. In B. Baars. (Ed.), The psychology of errors: Awareness, volition, and the structure of cognition New York Plenum. Madden, D. J. (1986).Adult age difTerences in visual word recognition: Semantic encoding and episodic retention. Experimental Agtrtg Research, 12, 71-78.
262
MacKay and Burke
Madden, D. J. (1988). Adult age differences in the effects of sentence context and stimulus degradation during visual word recognition. Psychology and Aging, 3, 167-172. McClelland, J. L. (1979).On the tfme relation of mental processes: An examination of systems of processes in cascade. Psychological Review. 86,287-330. McClelland, J. L., & Rumelhart, D. E. (1985).Distributed memory and the representation of general and specific information. Journal of Experimental Psychology: General, 114, 159-188. Mitchell, D. B. (1989). How many memory systems? Evidence from aging. Journal of Experfmental Psychology: Learnfng, Memory and Cognftion 15,31-49. Motley, M. T.,Baars, B. J., & Camden, C. T. (1983).Experimental verbal slip studies: A review and an editing model of language encoding. Communication Monographs, 50, 79-101. Nebes, R D., Boller. F., & Holland, A. (1986).Use of semantic context by patients with Alzheimer's disease. Psychology and Aging, 1,261-269. Oldfield, R C., & Wingfleld, A. (1965). Response latencies in naming objects. Quarterly Journal of Experimental Psychology, 17,273281. Rabinowitz. J. C. (1986).Priming in episodic memory. Journal of GeronMogy, 41,204-213. Rabinowitz, J. C., Craik. F. I. M.. & Ackerman, B. P. (1982).A processing resource account of age differences in recall. Canadian Journal of Psychology, 36,325-344. Rumelhart, D. E.. & McClelland, J. L. (1986).Parallel dfstributed processing. Explorations in the mfcrostructure of cognition: Vol. 1. Foundations. Cambridge, MA: MIT Press. Salthouse, T. A. (1982).Adult cognition. New York: Springer Verlag. Salthouse, T. A. (1985).A theory of cognitive aging. Amsterdam: North-Holland. Salthouse, T. A. (1988).Initiating the formalization of theories of cognitive aging. Psychology and Aging, 3,3-16. Schacter. D. L. (1987). Implicit memory: History and current status. Journal of Experimental Psychology: Learning, Memory and Cognition, 13.501-518. Schacter, D.L.,& Graf, P. (1986a). Effects of elaborative processing on implicit and explicit memory for new associations. Journal of Experimentat Psychology: Learning, Memory and Cognition, 12,432-444.
Cognition and Aging
263
Schacter, D.L., & Graf, P. (1986b).Preserved learning in amnesic patients: Perspectives from research on direct priming. Joumal of Clinical and Experimentclr Neuropsychology, 6, 727-743. Shfmamura, A. P.. & Squire. L. R (1989).Impaired priming of new associations in amnesia. Journal of Experimental Psychology: Learning, Memory and Cognition, 15, 721-728. Stemberger. J. P., & MacWhinney. B. (1986).Frequency and the lexical storage of regularly inflected forms. Memory & Cognftion, 14, 17-26. Streng. H. & Hedderich. J. (1982). Age-dependent changes in central somatosensory conduction time. European Neurology, 21, 270276. Tulving. E. (1985).How many memory systems are there? American Psychologist. 40. 385-398. Tulving, E., & Thomson, D. M. (1973).Encoding specificity and retrieval processes in episodic memory. Psychological Review, 80,352-373. Tweedy, J. R . & Schulman. P. D. (1982).Toward a functional classification of naming impairments. Brain and Language, 15. 193-296. Welford, A.T. (1985).Practice effects in relation to age: A review and a theory. Developmental Neuropsychology, 1 , 173-190. Wickelgren, W.A. ( 1977).Speed-accuracy tradeoff and information processing dynamics. Acta Psychologica. 41, 67-85. Yaniv, I., & Meyer, D. E. (1987).Activation and metacognition of inaccessible stored information: Potential bases for incubation effects in problem solving. Journal of Experimental Psychology: Learning. Memory. and Cognition, 13. 187-205.
This Page Intentionally Left Blank
Aging and Cognition. Knowledge Organization and Utilization
Thomas M . Hess (Editor) 0 Elsevier Science Publishers B.V. (North-HolZandl, 1990
CHAPTER SIX
SEMANTIC-MEMORY FUNCTION AND DYSFUNCTION IN ALZHEIMER'S DISEASE Robert D. Nebes
University of Pittsburgh
SUMMARY
Unlike the normal old, demented patients with Alzheimer's disease show severe deficits on many tests of semantic memory. It has been suggested that Alzheimer's disease causes an actual loss of semantic knowledge in these patients and that this constitutes one of the primary psychological impairments of the disease. Even the severe episodic memory problems found in Alzheimer patients have been attributed to their inability to encode the semantic features of stimuli. This chapter reviews the performance of Alzheimer patients on dmerent aspects of semantic memory as seen in tests examining their word finding abilities, and their knowledge of the semantic attributes, associates and category membership of concepts. The effect that semantic context has on cognitive processes such as lexical and semantic priming and memory encoding is also reviewed. Evidence is presented suggesting that semantic knowledge remains intact in Alzheimer patients, but can only be accessed and used under certain conditions. Memory problems are common both in normal older persons and in demented patients with Akheimefs disease. The memory impairment found in Alzheirner patients, however, is not only much more severe than that found in the normal old, but may also be more extensive, involving memory processes that are relatively
266
Nebes
spared by normal aging. In particular, it has been suggested that, unlike normal aging, Alzheimer's disease disrupts the structure and function of semantic memory. This impairment of semantic knowledge has been postulated to actually cause many of the other cognitive deficits seen in the dementia associated with Alzheimer's disease. The present chapter will review the literature in this area, and will examine whether the semantic deficits found in these demented patients reflect an actual loss of semantic information (a disruption of semantic structure) or whether they are due to an impairment in the patients' access to, or ability to use, what is essentially an intact semantic structure (for a more detailed review, see Nebes. 1989). We will show that in certain situations, Alzheimer patients can not only access appropriate semantic information, but can also use this information to facilitate their performance on a variety of cognitive tasks. Finally. since Alzheimer patients appear to be more impaired on some semantic tasks than on others, we will examine whether concepts such as automaticity, implicit memory,or environmental constraint might be useful in explaining this variability. The concept of semantic memory (memory for meaning) has been most extensively articulated by Tulving (1984). who distinguished it from what he called episodic memory (memory for events) on a variety of characteristics. Episodic memory is a record of unique events in an individual's experience, encoded in relation to a particular time and context. These memories make up an autobiographical record of events in an individual's Me. It is this type of memory that allows people to describe what happened on a television show they saw several weeks earlier, or to recall a list of words they heard in a memory test. In the case of remembering a word list, the subjects have seen those words thousands of times before: what they are remembering is the event - the fact that those words were seen at a particular time and in a particular context. Episodic memory is usually examined though tasks in which subjects are presented with words, pictures, stories, etc., and are later asked to recall or recognize this material. By contrast, semantic memory is thought to consist of a hierarchically organized body of knowledge involving facts, concepts, their meanings and associations, as well a s the rules for manipulating these symbols
S e m t C Memory inAlzhetmers Dfsease
267
and concepts. Information in semantic memory is organized conceptually without reference to the time and context in which it was learned. Semantic memory is typically assessed by testing an individual's knowledge of vocabulary. word associations, category membership, concept attributes, facts, etc. While there is strong evidence that most older persons show a significant impairment in episodic memory (Poon, 19851, their semantic knowledge appears to be well preserved (Howard, 1988). By contrast, it is commonly held (Bayles 8r Kasznlak. 1987) that one of the primary cognitive deficits in the dementia associated with Alzheimer's disease is an impairment in demented patients' knowledge of concept meaning. The evidence for such a semantic decrement is described below. SEMANTIC DEFICITS IN ALZHEIMER'S DISEASE Word Finding Problems Patients with Alzheimer's disease have a major problem generating specific words, such a s the names of objects, or concepts. This word finding deficit is evident early in the disease, both in patients' spontaneous speech and in tasks that require them to produce a speciflc word, such a s the name of a common object. The magnitude of this word finding problem is highly correlated with the overall severity of a patient's dementia. The spontaneous speech of demented patients is often tangential and relatively empty of meaning. They have difficulty generating the exact word needed in a particular context and thus, their speech contains few speciflc referents, being instead full of indefinite terms such as "stuff' and "things". An example given by Bayles (1982) is of a patient asked to describe the use of a bathtub, who says 'Well I see now, yeah, well you go into the bathtub and you don't have any problems here or there. You have got that over there though on that". It should be noted that, while empty of meaning, the speech of demented patients has relatively normal syntax and grammar. Alzheimer patients also show major deficits when asked to name common objects. While some studies (e.g., Barker & Lawson, 1968) claim that this anomia reflects a problem in perception (i.e., the patient misperceives the object and thus misnames it), most
268
Nebes
researchers now feel that the source of the naming problem is semantic. They argue that the patients' knowledge of the semantic features associated with a given concept is either lost or inaccessible and thus, the patient has difficulty matching the attributes of a presented stimulus with those of the appropriate semantic concept, Evidence for such a semantic deficit includes the fact that Alzheimer patients make primarily semantic errors when they misname objects (Bayles & Tomoeda, 19831, calling an item by the name of its superordinate category or by the name of another member of the same category (for example, calling a trumpet a musical instrument or calling a saw, a hammer). Alzheimer patients usually cannot recognize the name of the item they cannot name on sight (Flicker, Ferris, Crook & Bartus, 1987). suggesting that the semantic information for that particular item may be lost rather than being merely inaccessible. These patients have particular trouble recognizing the correct name of a pictured object if the distractors are other members of the same category (Huff,Corkin & Growdon, 1986). That is, if shown a hammer, they are much less likely to pick out the word "hammer". if the distractors are %aw" "screwdriver" and "chisel" than if the distractors are "comb" "shoe" and "truck". A common interpretation of these findings is that the patients have lost their knowledge of the specific features and attributes that are essential for distinguishing closely related concepts, although they retain enough semantic information to determine the object's overall category. Thus, when they misname an object they call it by its category name, or by the name of another object in that category. A similar explanation is given for the trouble demented patients have in naming a concept upon hearing its definition (Rissenberg & Glanzer. 1987). Another way to examine demented patients' word finding abilities is through tests of verbal fluency. Here, the subject is given the name of a semantic category (e.g., animals or words and is asked to generate as many beginning with the letter '3') examples as possible from that category within a given time period (60 to 90 seconds). Verbal-fluency tasks are very sensitive to early dementia, and show a strong correlation with dementia severity. Ober et aL(1986) found that even mildly demented subjects were
Semantlc Memory InAlzhetmers Dkease
269
severely impaired in the rate at which they produced examples from a category, with most of their responses occurring within the first few seconds. Alzheimer patients not only made fewer responses overall, but they had more perseverations and gave more noncategory responses. One important aspect of the results in the Ober et al. study is that while the demented patients gave many fewer responses than did the normals, they did not just generate the most typical (for semantic categories) or most frequent (for letter categories) items in a category. Rather, the demented patients' responses covered the entire range of response typicality and word frequency, That is, for a category such as birds, while they were more likely to say "robin" or "sparrow". like normals. they also produced some atypical examples such as "penguin". This suggests that the demented patients had not lost the less typical or less frequent words from their lexicon. and thus, the overall structure of their categories remained intact. Knowledge of Concept Meaning Another area of research on semantic abilities in dementia has investigated what these patients know about a semantic concept's attributes. associations and category membership. The most obvious way to determine what information patients have about a particular concept is to ask them by giving them a vocabulary test. Alzheimer patients do quite poorly on vocabulary tests, whether it is a verbal definition task, such a s the vocabulary subtest of the Wechsler Adult Intelligence Scale WAIS). or a recognition test, such as the Peabody Picture Vocabulary test. Their performance also steadily declines as their dementia progresses (Bayles h Kasmiak, 1987, pp 76-77). Thus, Alzheimer patients appear to have lost knowledge of word meaning, at least when this knowledge is assessed by their ability to carry out a directed search for the meaning of a word. Demented patients also have difficulty in the Similarities subtest of the WAIS, in which they have to say how two concepts are similar (e.g.. "How are a table and a chair the same?'). Instead of responding with an abstract generalization that relates the two concepts (e.g.. for table and chair saying that they are furniture), demented patients are more likely to give a concrete
270
Nebes
point of similarity (e.g.. 'They both have legs.") or to say how they differ (e.g., "You eat off one and sit on the other."). Thus. it appears that demented patients may be unaware of the nature of the semantic relationships between different concepts. Another approach used to examine concept meaning involves presenting subjects with a stimulus concept and asking them to generate associatims. Here, semantic structure is reflected in the nature of the associations produced. Most normal individuals tend to produce mainly paradigmatic associations, that is. items related to the stimulus concept in meaning (e.g.. items from same category, synonyms, or antonyms). Alzheimer patients not only produce many fewer paradigmatic associations than do the normal old, but they also give many more idiosyncratic responses, that is, words with no obvious relationship to the stimulus ( G e m , Shindler & Hier, 1984). This pattern of results has been interpreted a s showing an impairment in the network of semantic associations in Alzheimer patients, in that a stimulus word is not as likely to evoke a semantically-associatedword in these individuals. Salmon, Shimamura, Butters and Smith (1989) used an associative - priming paradigm to assess the integrity of semantic associations between ditrerent concepts in Alzheimer patients. In this procedure, subjects are presented with pairs of weakly related words and are asked to make some decision about them. Later, as part of what the subjects think is a totally dflerent task, they are given a series of stimulus words in a free association task and for each stimulus word are asked to say the first word that comes to mind. Embedded within the series of stimulus words are the first words from the previously presented word pairs. The question here is whether prior presentation of the two words a s a pair acts to prime (i.e.. activate) the associative link between them, thus making subjects more likely (in comparison to a control condition in which there is no prior presentation of the word pair) to emit the second word of the pair when they see the first word in the free association task. Salmon et al. (1989) found that Alzheimer patients were significantly less likely than normals to show such associative priming. That is, prior presentation of an semantically associated pair of words did not change the likelihood that the demented patients would emit the second word of the pair when shown the
Semantlc Memory
in Alzheimers Disease
271
flrst in a free association task. The results of a control condition embedded within the main task suggest that the superior associative priming seen in the normals is not due to their using a conscious recall strategy that is unavailable to the dememted patients. Salmon et al. argue that these results are evidence for an impairment in the network of semantic associations between concepts in Alzheimer patients. There have also been several studies that have directly examined Alzheimer patients' knowledge of the specific attributes that make up the meaning of a concept. Martin and Fedio (1983) showed photographs of common objects to Alzheimer patients and had them sort them into general categories, which the patients did quite adequately. Martin and Fedio then asked them direct questions about the photos. Demented patients could accurately answer questions about the superordinate and specific category of the objects (e.g., Is it man made? Is it a tool?), but were much less accurate in answering questions about the objects' specific attributes (e.g., Is it made of metal? Is it used to cut? Does it have moving parts?). They concluded from these results that while Alzheimer patients retain general information about a n item's category membership, they have lost their knowledge of, or access to, those semantic attributes that are necessary to determine concept meaning. Flicker et al. (1987) gave subjects 25 items and asked them both to name the objects and to pick out the eight objects they would use for a particular chore (e.g., cooking dinner). While the demented patients did have less trouble picking out objects needed for a given task than they did producing the objects' names, the patients were significantly impaired on both tasks. Flicker et al. concluded that demented patients have difficulty accessing specific attribute information about a concept from semantic memory. Use of Semantic Structure in Episodic Memory Finally, it has been suggested that a semantic impairment may be at least partially responsible for the deficit Alzheimer patients have in episodic memory. Weingartner et al. (1981) tested subjects' memory for lists of 20 words. One list (random) consisted of
272
Nebes
randomly selected words, while a second set (unclustered) consisted of 10 words from each of two categories all intermixed. A third list (clustered) was also made up of 10 words from each of two categories, but here, the words were presented grouped by category. Weingartner et al. found that the recall of normal persons improved if there was a semantic structure present in a list. Thus, their memory for the clustered list was better than for the unclustered list, which was in turn substantially better than for the random list. In contrast, the performance of the Alzheimer patients was equally poor on all three lists; the semantic structure of the lists apparently having no effect on their memory performance. Even when the lists were presented multiple times, demented patients showed no tendency to use the semantic structure inherent in the lists to improve their recall. The Alzheimer patients also did not show any consistent organization in their recall. That is, unlike normals, they did not recall items from the same category as a group, wen after having seen the same list SIX times. summary The most common interpretation of the studies summarized above is that while demented subjects retain information about the general category to which a concept belongs, they have either lost, or have difficulty accessing, information about the specific attributes and features of concepts. This semantic defect is seen to have widespread effects. It causes defective episodic memory because the demented patients do not encode semantic attributes. The patients have dimculty retrieving the names of objects or exact referents in spontaneous speech because they do not have available the specific information that would allow them to differentiate closely related referents and thus to chose the appropriate word. Their deficits in vocabulary and verbal comprehension tests could be explained along the same lines. There are however, other experiments suggesting that patients with Alzheimer's disease can access and use semantic information, including specific semantic features and attributes, at least under certain conditions.
Semantic Memory fn Alzhelmers Disease
273
PRESERVED SEMANTIC FUNCTION IN ALZHEIMER'S DISEASE Knowledge of Concept Meaning A recent study by Grober, Buschke. Kawas and Fuld (1985) suggests that knowledge of the specific attributes of concrete stimuli is actually preserved in demented patients, but that the organization of these attributes is disrupted. In this task, subjects were shown a target concept in the form of a word (e.g., "airplane"). They were then given a list of stimulus words (e.g., shoe, fly, pflot. desk, radar, milk, etc.) and asked to check off those words that were "related to" the target. Grober et al. found that demented patients responded correctly to 95% of the stimulus words, somewhat less than the 98% accuracy of the normal old, but stfll quite good. A second experiment, however, suggested that the demented patients were less aware than normals of the relative importance of the various attributes. Subjects were given word triplets, all of which were related to a single target concept, but which varied (according to norms) in their relative importance to the meaning of that target (e.g., for "airplane" - they might get "luggage", ''fly'. "radar"). Subjects were to pick out the word most important to the meaning of the concept first, and then the next most important word. Grober et al. found that demented patients were significantly worse than normals in discriminating among the attributes on the basis of their importance. They suggested that whfle Alzheimer patients have not lost their knowledge of concept attributes, the organization of these attributes is disrupted, the patients being relatively unaware of the relative importance of the various aspects of a concept's meaning. Such a disruption of attribute organization could produce many of the same cognitive deficits as an actual loss of those attributes. For example, if, in order to remember a concept, an individual must encode its semantic features, then any disorganization of these features could impair his or her episodicmemory performance. When given a word such a s "tiger" to remember, whfle a normal individual would encode distinctive features such as '*cat","stripes': and "large". a demented patient might encode less relevant attributes such a s "whiskers", "fur'; and "tail". which would not clearly specify the target concept. The
274
Nebs
demented patient's episodic memory would thereby suffer as a result of a semantic disorganization. One difficulty with the Grober et al. study is that it defined related attributes rather loosely, in that not only were distinctive physical features and actions involving the target concept given as related words, but also words that were just generally associated with the target (e.g.. for "airplane" the stimuli "sky" and luggage"). We recently carried out a study (Nebes & Brady. 1988) that differentiated between the various types of related words. Subjects were given a concrete noun as a target concept. They were then shown a series of ten stimulus words. We recorded the time it took the subjects to decide whether or not each stimulus word was related to the target. Of the five stimulus words that were actually related to a given target concept, one was the target's category, another a generally associated word, a third a prominent physical feature (a distinctive adjective or part), a fourth a function (a distinctive use or action involving the target), while the fifth was the target word itself a n identity relation. For example for the word "shirt" the category was "clothing", the associate was Yie", the action was "wear", the feature was "collar", and identity was "shirt". There were also five unrelated words interspersed among the related ones. If demented patients are differentially impaired with regard to the distinctive features and actions of concepts, then they should be slower and less accurate in making decisions about these types of stimulus words than about words involving a category, associate or identity relationship. The results are shown in Figure 1. While the demented patients were slower to make decisions than were the normals, the magnitude of this group difference was no greater for decisions about a physical feature or action than it was for decisions about a category or a general associate. In fact, the greatest difference in response time between normal and demented subjects was for decisions about the category of a target word. As in the Grober et al. (1985)study, the demented patients were very accurate in this type of decision, correctly identifying related words over 96% of the time. Thus, like Grober. we found no evidence that the field of semantic attributes, at least for these concrete concepts, was in anyway constricted or diminished in Alzheimer patients.
Semantic Memory
in Alzhelmers Dtsease
275
1150
m, /
1100
AD
1050
1000
'
950
900.
w Z
850
v)
g
800
v, Lu
u
750
700 650
w'
600
550
500
TARGET WORD TYPE Figure 1. The time (in milliseconds) that normal and demented subjects required to determine that various types of associated words are related to a given concept.
276
Nebes
Nor, was there any evidence for a specific impairment or loss of knowledge of the physical features and actions of concepts. In a study just completed, we attempted to directly test Grober et al.'s hypothesis that Alzheimer patients are unaware of the relative importance of the different attributes of a concept. Here, we used a procedure similar to that in the study just described. The subjects were given a target word followed by a stimulus word and they were to decide whether or not the two words were related. Unlike the previous study, the important stimulus variable in this experiment was not the nature of the attributes (i.e., physical feature, action, etc.) but rather, their relative dominance or salience with respect to the concept. We first carried out a nonning study in which we gave the target nouns to normal older persons and asked them to generate as many attributes as they could to each noun. We then selected high, medium, and low dominance attributes, based on the percentage of the norming sample who gave a particular response. High dominance attributes were those that over 45% of the subjects in the norming sample generated, while medium dominance attributes were generated by 20 to 30%. and low dominance attributes by less than 10%. For example, for the concept "elephant" the high, medium, and low dominance attributes selected were 'Yrunk", "ivory" and "memory". If Alzheimer's disease does disrupt the organization of concept attributes according to their importance or salience, then the time it takes Alzheimer patients to decide that a given attribute is related to a concept should not vary a s a function of the rated dominance of that attribute, while in normals it should. A preliminary analysis of the results (see Figure 2)shows that the time Alzheimer pattents took to decide that a given attribute was related to a target concept was aEected by attribute dominance to the same degree a s was that of both normal-old and normal-young individuals. The question then arises a s to why Alzheimer patients show sensitivity to attribute saliency in this task but not in that of Grober et al. (198517 One possible explanation is that in the Grober et al. task, the patients had to make a decision based on a direct comparison of the relative importance of multiple attributes (i.e., which of these three is most important). By contrast, in our study their knowledge of attribute saliency was derived indirectly from the
Semantic Memory in Alzhdmers Dlsease
277
1200
11so 1100
IOSO 1000
CI
3 3 8
8t2 m
/
/
/
9so 900
as0 100 750
700
bS0
600 $50
so0
!
High Medium Low ATTRIBUTE DOMINANCE
mure 2. The time taken by normal and demented subjects to decide that an attribute is related to a given concept, as a function of the relatlve dominance of that attribute.
relative speed with which they made a simple decision are these two words related or not. Thus, whether or not the semantic organization of concept attributes appears normal in Alzheimer patients may depend on how access to this information is tested. A similar explanation may account for why, in the Martin and Fedio (1983)study, Alzheimer patients were apparently unaware of the
278
Nebes
physical features and functions of objects. In the initial experiment in the Grober study and in our two studies, subjects had only to decide whether or not a concept and an attribute were related, while in the Martin and Fedio study they were asked a direct question about a concept (e.g., Is this item made of metal?) which required them to search for, and to make a decision about, a speciflc bit of information. Again. the manner in which demented patients' semantic knowledge is tested may be crucial in determining whether or not they succeed. There is also evidence that the structure and organization of semantic categories remains relatively intact in Alzheimer patients. We noted earlier that when demented patients are asked to generate members of a given semantic category, they produce many fewer items than do normals, but their responses contain low and medium tgpfcdity (dominant) items in roughly the same proportion as do the responses of normals (Ober et al., 1986). That is, their responses cover the whole range of typicality for members of a category, and thus, there is no evidence that they have lost the less typical category members. This was confirmed in a study from our lab (Nebes, Boller & Holland, 1986) in which we measured the speed with which subjects decided whether a particular word was a member of a given category. We varied the typicality of the items (i.e.. how commonly normals generate that item as a member of that category) and found that not only were demented patients quite accurate in making category decisions, but their knowledge of the relationship between an item and its category showed a similar structure to that of normals. That is. high typicality members were responded to faster than were low typicality members. While the demented patients took longer to respond than did normals, they were not disproportionately slower in responding to the less typical category members than to the more typical ones (see Figure 3). One line of evidence suggesting that the network of associations between semantic concepts remains at least partially intact in Alzheimer patients comes from studies using semantic priming. This approach is based on a model that views semantic memory as a hierarchical network of conceptual nodes. Presentation of a concept, for example by means of its name, activates not only its node, but also other concept nodes associated with it through a
Semanttc Memory hAlzhefmers Disease
279
1200
1100
h Demented
1000 m
W
c
-.-.-0, z
900 800
700 600 HIGH
DOMINANCE
LOW
FLgure 3. The time normal and demented subjects needed to determine that a given item is a member of a category, as a function of the category dominance of that item.
variety of relationships. This spread of activation briefly increases the accessibility of these related concepts, allowing them to be processed more rapidly if they occur soon after a concept to which they are related. In the semantic-priming paradigm, the time a subject takes to process a given stimulus is measured under two conditions: a) when the preceding stimulus (the prime) is semantically related to the item being processed, and b) when it is unrelated. Normals generally process a word faster if it is preceded by a semantically related concept (e.g.. they can process "doctor" faster if it follows "nurse" than if it follows "shoe"). Since this task does not require the subject to engage in any intentional search for, or use of, semantic associations, we felt it might be a good measure
280
Nebes
of the intactness of semantic associations in demented patients. We have carried out several priming studies. In the most recent, subjects were given a series of words and we measured the time it took them to respond to each word as a function of whether the preceding word (the prime) was semantically related to it or not. We used two difYerent tasks - word naming and lexical decision (Nebes. Brady & Huff, 1989). If there is an impairment in the network of semantic relationships between concepts in Alzheimer patients, then we would expect semantic priming to be reduced or eliminated in these individuals. However, as shown in Figure 4, if anything, the priming effect was larger in the demented patients than in the normal young or old, for both the naming and lexical-decision tasks. Thus, while demented patients may have difficulty explaining the nature of the relationship between a table and a chair (Similarities subtest of the WAIS), they are aware of this relationship at some level, since they process the word "chair" faster if it is preceded by "table" than if it is preceded by an unrelated word, such a s "mountain". How then can we reconcile these findings with the previously mentioned associative priming data of Salmon et al. (1988). One important difference between the semantic-priming and associative-prfmingparadigms is the demand for lexical search in the latter. In semantic-priming tasks, the effect of the prime is measured by a change in the speed or accuracy with which subjects process a presented word. By contrast, in associative-priming tasks, the effect of the prime is measured by a change in the likelihood that the subjects will emit a given word in a free association task. Therefore, successful associative priming requires more than just activation of a semantic node and the spread of this activation to related nodes; it requires lexical search. Thus, rather than dysfunctional semantic associations, an impairment in lexical search may be the cause of Alzheimer patients' fallure in associative priming. Use of Semantic Context and Information The performance of demented patients can also be facilitated by more complex semantic contexts such a s those provided by a sentence. In a semantic-priming task (Nebes et al., 1986). subjects
281
Semantic Memory InAlzhelmers Disease
1400
Lwmd
0 YOUNG
0 OLD
A
A L ZHEIMER
--
iaoo
l¶OQ
I100
1000
2
F
B 0
'0°' 100 '
700.
(00
F-- - - - - - _ _ y
900-
* - - - - - - - - - -- 0
400-
I
ASSbC.
UNA~SOC. TYPE OF PRIME
mure 4. Response time for lexical-declsion and word-naming tasks as a
function of whether the preceding word is semantically associated with the word being processed.
listened to a sentence in which the final word was missing. At the point in the sentence when the find word would have been heard, subjects instead saw a target word which they were to read aloud. The context of the sentence was either congruous with this visual word (e.g., the subject heard 'The child was born with a rare -'I
Nebes
282
and then saw "disease"), or was incongruous (e.g., 'When you go to followed by "disease"). These conditions were bed turn off the compared to a neutral condition in which the sentence provided no contextual information ('They said it was the -"). If demented patients are unable to use the semantic context provided by a sentence, then the speed with which they read the target word should be unaffected by the nature of the preceding sentence. It was not. A congruous sentence facilitated (i.e., speeded up) reading of the target word (in comparison to the neutral condition)just as much in the demented patients as it did in the normal old and young. The inhibition produced by an incongruous sentence (in comparison to the neutral condition) was even larger in the patients than in the normals. These results suggest that the semantic information present in the sentence must have been encoded and comprehended at some level in order for it to have affected the patients' behavior in this manner. Further analysis of the sentences showed that the priming effects produced by sentence context in this task could not be attributed merely to the presence in the sentence of a single word highly associated with the target. Instead, the effect of sentence context appeared to due to the meaning conveyed by the sentence as a whole. A companion task to the previous one examined the effect that sentence context had upon Alzheimer patients' ability to carry out a directed search of their lexicon (Nebes et al., 1986). Again, subjects heard an incomplete sentence and at the point where the final word would normally have occurred, they heard a tone. This was their signal to complete the sentence with a single word. In this task, subjects had to use the semantic-syntactic context of the sentence to guide their search through their lexicon for an appropriate word to complete the sentence. The sentences varied in the degree of selection constraint they imposed on this final word. Some sentences were very constraining, in that relatively few words would sensibly complete the sentence (e.g., "Father carved the turkey with while others were much less restrictive (e.g.. 'They went to a -."I, in that they could be completed by a large see the famous number of words. Normal young and older persons completed the high constraint sentences faster than they did medium or low constraint sentences. The demented patients were -I'
-.'I),
Semantk Memory inAlzheimrs Disease
203
disproportionally slower than the normals on the medium and low constraint sentences. In fact, the patients actually had a great deal of difficulty with any sentence that did not provide a high degree of constraint, often being unable to produce an ending within the 10 second time limit. The patients were much better. both in terms of accuracy and speed of response, if the sentence context was highly constraining. Thus, if anything, Alzheimer patients were more sensitive to the semantic context of the sentences than were normals. Therefore, the word-finding deficit previously shown for Alzhefmer patients was also evident in this task. but it could be ameliorated to a degree, if the patients were provided with a highly constraining semantic context that guided their search of their lexicon for the appropriate word. In a study still underway. we followed up on this last task by investigating whether the degree of constraint present in a sentence also d e c t e d the time subjects took to judge whether a final word made sense in the context of that sentence. The subjects heard a sentence in which the final word was presented visually. and they had to say yes or no as to whether this visual word provided a sensible completion to the sentence (e.g., "He mailed the letter without a stamp. vs. The wealthy child attended a private hand"). Again we varied the constraint of the sentences. We found that the decision time of Alzheimer patients was affected by sentence constraint to the same degree as was that of normals (see Figure 5). That is, like normal young and old subjects, demented patients detected an appropriate completion of a high constraint sentence faster than they did a low constraint sentence. Thus, they responded "yes" to "He scraped the cold food from his plate." faster than they did to 'The surface of the water was very smooth". These last three studies demonstrate that complex semantic context does affect the speed with which Alzheimet patients can: identify a word, make a decision about the contextual relevance of a word and search their lexicon for an appropriate word. Thus. whatever problems demented patients have on semantic tasks. their performance is sensitive even to complex semantic information conveyed by whole sentences.
Nebes
284
1500 1450
a,
1400 1350
/
1300
/
1250
/
A.D.
/
P ’
/‘
W‘’
1200 1150 n
3
1100
2
1050
U
F LLI
1000
v)
B v)
LLI
950 900
w
850 800 750 700 650
600
500
I
I
I
I
I
High Medium Low DEGREE OF SENTENCE CONSTRAINT Ftgwe 5. The time normal and demented subjects required to determine that a particular word appropriately completes a given sentence context.
285
Use of Semantic Context in Episodic Memory Finally, two studies from our lab suggest that the presence of semantic structure in verbal material can facilitate Alzheimer patients' episodic memory for this material, probably by improving encoding. The first study (Nebes et al., 1984) examined how accurately subjects could recall lists of words that differed in the amount of language structure present in the list. The lists varied in how closely they approximated English text, that is, the degree to which they conformed to the constraints of English syntax and semantics. The lists ranged from a zero-order approximation words selected totally at random (e.g.. day, angel, error, view, wire, park, rail, belt) - to lists with a moderate (fifth-order)approxtmation (e.g., to, ask, for, is. to, earn. our, living) in which there is some language structure, to actual text (e.g.. it, was, a, warm, day, in. July, when). The more closely a word list approximates a natural sentence, the easier it is for normals to recall it (Miller 81 Selfridge, 1950). Higher order-of-approximation lists are more likely to have embedded within them meaningful groupings of words (chunks) that allow subjects to encode the words in terms of well-learned language patterns. In the Nebes et al. (1984) study, task difficulty was individually equated for normal and demented subjects by first determining how long a list of random words had to be before a particular subject recalled only 25% of them. This list length was then used for all the varying approximations to text that were given to that particular subject. Obviously, since demented patients have a poor episodic memory, this meant that the lists given to them were shorter than were those given to the normals. However, we did not analyze the absolute number of words recalled in this task, but rather the percentage of a list that a subject recalled a s a function of order of approximation. Results showed that a s approximation-to-text increased, the percentage of the list recalled also increased, and this increase was virtually identical for normals and for demented patients. It thus appears that patients with Alzheimer's disease can use the language structure present in text to facilitate their memory for that text. However, there was a major confound in this task. As approximation to text increases, not only are word lists more likely to form semantically coherent phrases,
286
Nebes
but they are also more likely to conform to normal syntactic patterns. It is known (Marks & Miller, 1964) that both syntactic and semantic structure can facilitate recall. Since syntax is relatively well preserved in demented patients (Hier, Hagenlocker & Shindler, 1985). the improved memory of Alzheimer patients for the higher orders of approximation might be due to their use of syntactic rather than semantic structure. This confound was not, however, present in a followup study (Nebes. Brady & Jackson. 1989) in which subjects were given seven-word sentences that were either normal, bizarre (i.e.. they portrayed unlikely events), semantically anomalous 0.e. * they portrayed impossible events that violated semantic constraints), anagrams (i.e.. sentences whose word order was disarranged), or random word lists which had neither semantic nor syntactic structure. For example, a normal sentence was "Five girls swam in the shallow pool"; a bizarre sentence was 'Three bugs jumped over the rotten meat": a semantically anomalous sentence was Many carrots sang to the sickly ashtrays"; an anagram was "Ship near sunken two floated ducks the): while a random-word list was "Dentist town a the decided tender broken". Again, overall memory performance was equated between the normal old and the patients prior to testing. In this case, we varied the duration of an interference period following presentation of random-word lists until we found a duration that produced 25% recall accuracy. An interference period of this duration was then used with the various sentence types. Since the only difl'erence between the normal, bizarre, and semantically anomalous sentences is the amount of semantic structure present in the sentence (all of them have normal syntax). if demented patients cannot use semantic structure to improve their recall, but instead rely on syntactic structure, there should no dflerence in their recall accuracy for these three types of sentence. From the data shown in Figure 6 it is clear that the pattern of recall accuracy across the various sentence types was very similar in normal and demented subjects. Recall accuracy improved as the amount of semantic information in the sentences increased (i.e.. from semantically anomalous, to bizarre to normal sentences) and this was equally true in elderly normals and in demented patients. The mechanism by which semantic structure
Semantk Memory in Akheimers Disease
287
100
90
80
0
70
W
K K
80
0 0 I-
z
50
0
40
W
K
w n
30
20
DEMENTED
-e
NORMAL OLD
0-
--
10
NORMAL
BIZARRE
SEMANT. ANOMAL.
ANAQRAM
WORD LIST
TYPE OF SENTENCE
Flgure 6. Recall accuracy by normal old and demented subjects for sentences with varying amounts of semantic and syntactic structure.
facilitated memory perfonnance also seemed to be the same in normals and in demented patients. Semantic structure appears to provide a basis for subjects to organize a string of words into chunks (meaningful phrases), thus allowing more efffcient encoding of the words into memory. If we look at the number of words m a w up a recall "chunk" (defined as a sequence of correct words recalled in the same order a s they were presented). it is clear (Figure 7 ) that a s the semantic structure of the sentences increased, so too did chunk size and that this was equally true of both normal and demented subjects. Thus, demented patients do appear capable of using the semantic structure present in text to
Nebes
288
Y
z
o\
3
I
0
4
\
U
w
P 0
3
P
2
0 U
'\\ DEMENTED
0-
NO RM AL O L D
0-
--
\ \
\
01
NORMAL
BIZARRE
SEMANT. ANOMAL.
- --=a
ANAQRAM
WORD LIST
TYPE OF SENTENCE Flgure 7 . Mean size of the recall chunk (words recalled in the same order in which they were given) in normal and demented subjects' responses, as a function of the semantic and syntactic structure of the sentences.
improve their episodic memory and the mechanism by which semantic information af€ects their memory is the same as in normal individuals. The question now becomes why can demented subjects use semantic structure in these sentences, but not in the categorized word lists of Weingartner et al. (19811. One possibility is that Alzheimer patients benefit from the semantic information present in sentences, because reading connected text is such an overlearned automatic sklll in most persons. Demented individuals may fail to encode semantic information only if the situation requires that they self-initiate a novel coding scheme, such a s grouping words by
Semantic Memory in Alzhetmers Dlsease
289
membership in a semantic category. Demented patients may be able to effectively encode semantic information only to the extent that their encoding is induced and directed by the stimulus material itself through overlearned patterns of processing.
SUMMARY OF RESULTS There is no question that demented patients have trouble generating an appropriate word, both in spontaneous speech and in tasks that require them to name an item upon seeing its picture or upon hearing it deflned. However, it also appears that this wordfinding problem can be reduced if an adequate stimulus context is provided to guide or constrain the patient's search for a suitable word, a s in the sentence-completion task. In situations in which subjects must generate their own lexical-search strategy with minimal contextual guidance (e.g., a verbal-fluency task), the wordfinding performance of demented patients is deficient. As to what demented patients know about the meaning and associations of a particular concept, the answer appears to depend on just how you ask them. For example, if you test their knowledge of an item's superordinate category by asking them to generate as many items as possible from a given category (e.g., furniture), or to tell you how a table and chair are alike, they do poorly. If, on the other hand, you ask them to sort items by category, or to make a decision as to whether a given item belongs to a particular category, they do quite well. Similarly with associations. If asked to generate associates, demented patients are less likely than normals to come up with a paradigmatic (i.e., semantic) associate. This could indicate that such semantic associations are lost in Alzheimer patients. However, if you examine their knowledge of associations by indirect means, such as semantic priming, it is clear that the network of associations between the different concepts remains relatively intact. In terms of their knowledge of the attributes of a concept, when demented patients are directly asked about the specific physical features and functions of objects, they are often unable to correctly answer the question. However, if they are just asked whether or not a specific attribute is related to a given concept, they are very accurate. In all of these tasks, it appears that the
290
Nebes
situations that give the demented patients the most difficulty are those in which they must carry out a self-directed search of their semantic memory for a specific piece of information. The preciseness of the search seems crucial. The effect that semantic context has on episodic memory in demented patients also depends on the experimental situation. When the semantic structure present in a set of words consists only of the words being segregated into several categories, demented patients do not use the semantic structure to improve their memory encoding. When, however, the semantic structure is inherent in connected text, they do. Similarly. connected text also had a strong priming effect in demented patients, even greater than that found in normals. Since reading text is a highly overlearned skill for most persons in our society. the patients' use of the semantic information in text may be more automatic than it is in situations where they must first notice and then use the semantic structure in a more novel setting, such a s categorized word lists.
INTERPRETATION Given the variability in the test results described above, how are we to conceptualize the nature of the semantic deficit in Alzheimer's disease? First of all, is semantic information actually lost in Alzheimer patients or are their semantic impairments due to other cognitive deficits that limit their ability to access and appropriately use semantic information? Certainly many investigators (e.g., Bayles & Kaszniak, 1987:Salmon et al., 1986) feel that Alzheimer's disease produces a fundamental disruption of semantic information. To determine whether information is really lost or merely inaccessible is, of course, extremely difficult. However, to the extent that demented patients can access or use semantic information under some conditions but not under others, this would suggest that at least some semantic information remains intact in Atzheimer patients. The tasks on which demented patients appear most impaired are those that require them to carry out a directed search of their semantic memory for a specified piece of information (e.g., verbal fluency, object naming, generating associates, etc.). However, it is unlikely that we can attribute all
Semantic Memory in Alzheimers Dlsease
291
semantic problems in Alzheimer's disease to a reMeuu.2 denit, as the performance of demented patients can differ greatly between conditions that make similar demands on retrieval. For example, demented patients can answer direct questions about whether a particular object belongs to a certain category, but not similar questions about the object's physical properties and functions (Martin & Fedio. 1983). Another possibility is that rather than being lost, or relatively inaccessible, semantic information is disorganized (e.g., the semantic attributes of a concept no longer being organized according to their relative importance), leading to ineffective use of this information (Grober et al., 1985). However, there is also evidence that the overall organization of at least some components of semantic memory remains intact in Alzhiemer patients (Ober et al.. 1986: Nebes et a]., 1986). Many of the studies just reviewed seem to suggest that the presence or absence of a semantic deficit in demented patients depends greatly on the nature of the task. It may therefore be more useful to determine under what conditions Alzheimer patients can access and use semantic information and under what conditions they cannot. The question then becomes are there any untfying concepts that might explain why demented patients can apparently retrieve and use semantic information in certain tasks and not in others. For example. can we use the concept of automaticity? Jorm (1986) and Nebes et al. (1984) suggested that if the semantic operations involved in a task make heavy demands on attentional capacity, then Alzheimer patients will show a deficit. However, if access and use of semantic infomation is relatively automatic, they will perform normally. While this is a very appealing concept, there is no direct evidence for it in the literature. The concept of automaticity has been most clearly conceptualized in restricted experimental paradigms: in visual and memory search by Schneider and Shrif€rin (1977). in memory encoding by Hasher and Zacks (1979). and in semantic priming by Neeley (1977). Of the studies described in this review, only the semantic-priming procedures have any well defined criteria for determining whether subjects are using automatic or attentlon-dependent processes. We had predicted in our priming studies (Nebes et al., 1984; Nebes et al., 1986: Nebes et al., 1989a) that demented patients would show
292
Nebes
semantic priming resulting from an automatic spread of activation, but not priming dependent upon attentional processes. These two components of priming have been differentiated in a number of studies in normals through the presence or absence of priming inhibition (Neely, 1977) or through the use of a word-naming versus a lexical-decision paradigm (Seidenberg,Waters, Sanders & Langer, 1984). Unfortunately, the predictions that spring from these earlier studies with normals were not supported by the results with Alzheimer patients, If we accept the criteria that have been laid down for differentiating automatic from attention-dependent priming, it appears that both types of priming occur in Alzheimer patients. Thus, there is no strong evidence that demented patients can access and utilize semantic information only through automatic processes. Another theoretical construct recently advanced by Craik (1984) to explain the effects that normal aging have on cognition, might also be of use in understanding the pattern of semantic deficit in Alzheimer patients. Craik suggested that the amount of age deficit seen in a given task is a function of the degree to which the required cognitive operations are self-initiated versus being driven by the task and stimulus materials (i.e., the amount of environmental support provided). It may be that in tasks in which demented patients must self initiate and organize the retrieval and use of semantic Information, their performance is defective, while in tasks in which their processing is strongly constrained and guided by the external task situation, they perform fairly normally. The most obvious support for this view comes from the sentencecompletion task (Nebes et d., 1986) and its followup task. Here, the amount of constraint present in an incomplete sentence certainly had a major effect on demented patients' ability to complete the sentence with an appropriate word or to judge whether a presented word sensibly completed the sentence. It is possible to also categorize at least some of the other semantic tasks used with Alzheimer patients in terms of the amount of environmental support they provide (e.g., verbal fluency has little support, while category decision has a great deal). It is clear, however, that this conceptualization cannot explain the entire pattern of semantic test results in Alzhiemer patients. For example, an associative-priming
Semantic Memory in Alzheimers Disease
293
task does not seem to provide any less semantic guidance than does a semantic-prlming task using single words, and yet Alzheimer patients perform abnormally on the first and normally on the second. Another potentially useful concept is impZicit memory. As recently reviewed by Schacter (19871,implicit memory is evident when previous experiences facilitate performance on a task that does not require conscious intentional recollection of those experiences. This is contrasted to explicit memory, which does involve intentional recollection (e.g., typical recall and recognition tasks). There is a growing literature showing that even amnesic patients who are grossly impaired on traditional measures of explicit memory perform normally on measures of implicit memory (Graf, Squire & Mandler, 1984). Many of the semantic tasks that Alzheimer patients perform successfully, could qualify as measures of implicit memory (e.g., semantic priming). Even the superior performance that Alzheimer patients show in remembering sentences as compared to random words might involve implicit use of their knowledge of the semantic relationships expressed in the sentences. However, Alzheimer patients do not necessarily perform normally on all of the tasks that are presently classified as measures of implicit memory. For example, Alzheimer patients do not show normal associative priming (Salmon et al.. 1988). Therefore, the implicit memory concept alone cannot totally account for the pattern of semantic-memory performance in Alzheimer's disease. At present, there does not seem to be any one simple explanation for the variability in demented patients' performance on measures of semantic memory. We may have to look for combinations of factors (e.g., automaticity plus environmental constraint) or it may turn out that semantic memory is comprised of multiple components, some of which are impaired in patients with Alzheimer's disease, while others are spared. REFERENCES Barker, M. G., 81 Lawson. J. S. (1968).Nominal aphasia in dementia. British Journal of Psychfairy, 114, 1351-1356.
Nebes
294
Bayles, K. A.. & Kaszniak, A. W. (1987). Communication and cognition in normal aging and dementia. Boston: College -Hill press. Bayles. K. A., & Tomoeda, C. K. (1983). Confrontation naming in dementia. Brain and Language, 19.98- 114. Craik. F. I. M. (1984). Age differences in remembering. In L. R. Squire & N. Butters (Eds.). Neuropsychology of m y (pp. 312). New York Guilford Press. Flicker, C., Ferris, S. H., Crook, T.. & Bartus, R. T. (1987). Implications of memory and language dysfunction in the naming deficit of senile dementia. Brain and Language, 31, 187200.
Gewirth. L. R., Shindler, A. C., & Hier, D. B. (1984). Altered patterns of word associations in dementia and aphasia. Brain and Language, 21.307-3 17. Graf, P., Squire, L. R., & Mandler, G. (1984). The information amnesic patients don't forget. Journal of Experimental Psychology: Learning Memory and Cognition, 10, 164- 178. Grober. E., Buschke, H., Kawas, C., & Fuld, P. (1985). Impaired ranking of semantic attributes in dementia. Brain and Language, 26, 276-286. Hasher, L. & Zacks. RT. (1979). Automatic and effortful processes in memory, Journal of Experimental Psychology: General., 108, 356-388.
Hier, D. B.. Hagenlocker, K.,& Shindler, A. G. (1985). Language disintegration in dementia: Effects of etiology and severity. Brain and Language, 25, 117- 133. Howard, D. V. (1988).Aging and memory activation: The priming of semantic and episodic memories. In L. L. Light & D. M. Burke (Eds.), Lunguage, memory, and aging (pp 77-99). New York: Cambridge University Press. Huff, F. J., Corkin, S., & Growdon, J. H. (1986). Semantic impairment and anomia in Alzheimer's disease. Brain and Language, 28,235-249. Jorm, A. F. (1986). Controlled and automatic information processing in senile dementia: A review. Psychological Medicine. 16, 77-88.
Marks, L. E., & Miller, G. A. (1964). The role of semantic and syntactic constraints in the memorization of English sentences. Journal of Verbal Learning and Verbal Behavior, 3. 1-5. Martin, A., & Fedio, P. (1983).Word production and comprehension in Alzheimer's diesease: The breakdown of semantic knowledge. Brain andLanguage, 19, 124-141.
Semantic Memory
in Alzhetrners Disease
295
Miller, G. A, & Selfridge. J. A. (1950).Verbal context and the recall of meaningful material. Amer[cal Journal ofp~ychology~ 63, 176185. Nebes, R.D. (1989)Semantic memory in Alzheimer's disease. P ~ y ~ h o l ~ gBulletin., i ~ a l 106,377-394. Nebes. R. D., Boller, F., & Holland, A. (1986).Use of semantic context by patients with Alzheimer's disease. Psychology and Agfng. 1 . 261-269. Nebes. R. D.. & Brady. C. B. (1988).Integrity of semantic fields in Alzheimer's disease. Cortex, 24,291-300. Nebes, R. D., Brady, C. B., & Huff, F. J. (1989a).Automatic and attentional mechanisms of semantic priming in Alzheimer's disease. Journal of Clinical and Experimental Neuropsychology, 1 1 , 219-230. Nebes. R. D.. Brady, C. B., & Jackson, S.T. (1989b).The effect of semantic and syntactic structure on verbal memory in Alzheimer's Disease. Brain and Language, 36, 301-313. Nebes, R. D., Martin, D. C., & Horn, L. C. (1984).Sparing of semantic memory in Alzheimer's disease. Journal of A b n o M P ~ y ~ h o l ~93, g y ,321-330. Neely. J. H. (1977).Semantic priming and retrieval from lexical memory: Roles of inhibitionless spreading activation and limited-capacity attention. Journal of Experimental Psychology: General, 106, 226-254. Ober. B. A., Dronkers. N. F., Koss. E., Delis, D. C., and Friedland, R P. (1986). Retrieval from semantic memory in Alzheimer-type dementia. Journal of Clinical and Experimental Neuropsychology. 8.75-92. Poon, L. W. (1985).Differences in human memory with aging: Nature, causes, and clinical implications. In J. E. Poon, & K. Warner Schaie (Eds.), Handbook of the psychology of aging (pp. 427-462). New York Van Nostrand Reinhold Co. Rissenberg, M., & Glanzer. M. (1987).Free recall and word finding ability in normal aging and senile dementia of the Alzheimer's type: The effect of item concreteness. Joumal of Gerontology,42, 318-322. Salmon, D. P., Shimamura. A. P.. Butters, N.. & Smith. S. (1988). Lexical and semantic priming deficits in patients with Alzheimer's disease. Journal of Clinical and Experimental Neuropsychology. 10. 477-494. Schacter, D. L. (1987). Implicit memory: History and current status. Journal of Experimental Psychology: Learning, Memory, and Cognition, 13, 501-518.
296
Nebes
Schneider, W., & ShifMn, R.M.(1977). Controlled and automatic human information processing: I. Detection, search and attention. PsychoZogfcaZ Revfew, 84, 1-66. Seidenberg. M. S., Waters, G. S.. Sanders, M., & Langer. P. (1984). Pre- and postlexical 'loci of contextual effects on word recognition. Memory and CognUf.on, 12,315-328. Tulving, E. (1984). Precis of "Elements of episodic memory." The Behavioral and Brain Sciences, 7, 223-268. Weingartner, H., Kaye, W.. Smallberg, S.A., Ebert, M.H., Gilin, J.C., & Sitaram, N. (198 1). Memory failures in progressive idiopathic dementia. Journal o f A b m a l Psychologg, 90, 187- 196.
Aging and Cognition: Knowledge Organization and Utilization Thomas M. Hess (Editor) 0 Elseuim Science F.zLbZi~her~ B.V. (Abrth-Holland), 1990
CHAPTER SEVEN
A DEVELOPMENTAL APPROACH TO VISUAL COGNITION IN THE ELDERLY Matthew J. Sharps
California State University - Fresno
SUMMARY Significant age-related influences on visual cognitive abilities are commonly reported. I t is argued that methods derived from developmental theory and findings may be useful for the study of these influences. Such “ageby-task’’ methods involve the systematic and simultaneous manipulation of task-related and subjectrelated variables. Examples of the utility of these methods are drawn from the spatial perspective-taking and spatial memory literature. The application of this approach to problems of mental imagery, within a mental image rotation framework, is also demonstrated. A review of literature on mental image rotation and aging indicates that performance may depend upon a variety of subject- and task-related factors, including stimulus characteristics. A new experiment addressing some of these factors is reported. Subject age is shown to interact significantly with specific characteristics of the stimulus items employed. The results indicate that an age-by-task approach to problems of aging and mental imagery may lead to more comprehensive understanding of the factors influencing performance than would otherwise be possible. Visual cognition, or visual information processing, has been defined as the entire process by which human beings receive visual information and adjust their behavior on the basis of that
298
Sharps
information (Spoehr & Lehmkule, 1982). A variety of visual cognitive abilities are included within the sphere of activities involved in knowledge utilization (see Lachman & Lachman, 1980). including the use of visual characteristics of stimuli to organize encoding and retrieval, the ability to judge similarities and differences among stimulus items, and the ability to judge the appropriateness of stimulus organization to various real-world and experimental contexts (e.g., Biederman, 1977). An understanding of the principles underlying aging and visual cognition is therefore important for the development of knowledge utilization theory. Numerous studies have demonstrated age-related declines in specific areas of visual cognition. Such declines are accepted a s essentially axiomatic in a number of functional areas, including the ability to manipulate images in mental %pace" (e.g., Gaylord & Marsh, 1975).the ability to remember spatial relations (e.g.. Light & Zelinski, 1983;Pezdek, 1983).and the speed of cognitive functions (e.g., Birren. 1974). Cognitive declines with age have been documented so thoroughly that few investigators question their existence or importance; current research tends to address the severity or impact of given deficits, rather than their existence. As new information about aging and cognition accumulates, however. it is becoming more critical to focus attention on the abilities retained by the elderly, as well as upon the declines that occur with age. It is necessary to examine not solely what the elderly tend to do when compared with young adults, but also what they can do, given the appropriate circumstances. It is therefore important to develop and employ experimental methods which permit the analysis of the abilities retained throughout the lifespan. One set of methods which has proven to be highly useful for such analyses derives from the developmental theory of Gollin (e.g, 1965. 1981,1984,1985;Gollin & Saravo, 1970).Research designs in both developmental and nondevelopmental research often involve only one level of either subjects or tasks (Gollin & Saravo. 1970,p. 59). Subjects of different ages may be tested within only one task framework, or subjects of the same age may be tested using several tasks. Gollin proposes that more comprehensive information may be obtained through the use of experimental designs involving the
Visual Cognition
299
systematic variation of both subject and task characteristics within the same experimental framework, This age-by-task approach "mandates a research methodology that seeks to access the central processes available to organisms, to plot their polyphasic relations over ontogenesis, and to spec@ the successive configural characters of organism-environment patterns.. .The essential feature of those methods is the recognition of the constructive. organizational character of central processes, their patterned interrelations and variation during ontogenesis, and their species-specific qualitative and temporal properties" (Gollin, 1981, p. 246-247). It is especially important to note the phrase "central processes available to organisms" in the above. Within Gollin's research framework, the investigator selects task characteristics and organismic characteristics (such a s age) according to theoretical considerations, and varies them systematically in an age-by-task manipulation. If these variables are properly chosen, the investigator may obtain a n understanding not only of what individuals of a given age tend to do on a given task, but also of what they are able to do when task circumstances are reconfigured. It becomes possible to isolate specific characteristics of the task environment which are optimal or suboptimal for the performance of individuals of dflerent ages. Gollin's approach has been employed in a number of areas in comparative and developmental psychology, including the developmental analysis of learning in rats (e.g., Rudy, Vogt, & Hyson, 1984). the study of the development of conceptual-verbal mediational systems in young children (e.g., Gollin & Garrison, 1980). and other areas of child development research. A particularly pertinent example of these applications stems from research on visual cognition in children, specifically in the area of spatial perspective-taking. The original studies of spatial perspective-taking (Piaget & Inhelder, 1967. original study 1948) involved the "three mountain task." in which children were asked to examine an array composed of three papier-mache mountains which differed in several features. A doll was also placed at the array, but at a different orientation from that of the child. The child was then asked to choose the perspective which represented
300
Sharps
the doll's "view" of the array. Several kinds of perspective representation were used in these experiments. It was found that younger children tended to make "egocentric" responses, choosing perspectives which represented the view that they themselves saw, rather than choosing the view that represented the doll's true "viewpoint." F'rom these results, Piaget and Inhelder (1967) further observed that the ability to engage in nonegocentric spatial perspective-taking improved over the course of child development. According to Piaget and Inhelder. the possession of virtually no ability to take perspective (around age 4) gives way to an evident recognition of some need for perspective-taking. followed by actual attempts to separate viewpoints. Around the age of 9 or 10 years, children were observed to achieve a more complete understanding of spatial perspective-taking, and were in general able to employ perspective-taking abilities similar to those of adults. These findings gained broad acceptance, and were widely replicated in subsequent years (see Laurendeau & Pinard, 1970). However, further studies reflned these ideas. Spatial perspectivetaking abilities were shown not only to depend on the age of the child, but also on the characteristics of the tasks employed to test them (e.g., Borke, 1975; Fishbein, Lewis, & Keiffer, 1972; Gzesh & Surber, 1985: Huttenlocher & Presson. 1979; Masangkay, McClusky. McIntyre. Simms-Knight, Vaughn, & Flavell. 1974; Olson, 1970 Presson, 1980: Shantz &Watson, 1970). For example, Brodzinsky, Jackson, and Overton (1972). using a task framework similar to the three-mountain paradigm, showed that masking stimulus arrays during the child's attempt to judge perspective improved perspective-taking performance. Walker & Gollin (1977) found that masking of a dollhouse array "reduced egocentric errors among 4-year-olds, producing a pattern of responses more typical of that observed among older children" (p. 343). The complexities of the child's perspective and of the doll's line of regard (e.g., corner versus side views of dollhouse arrays) have also been shown to interact with the ages of the children studied (Schachter & Gollin, 1979; Walker & Gollin, 1977): younger children tend to have more difficulty with perspective-taking when dealing with more complex views, a s well as with doll positions from which the doll's "view" is more complex.
Visual Cognition
301
A later study (Gollin & Sharps, 1987) demonstrated that the availability of strategy, in the absence of any physical alteration of the stimulus materials, also influenced spatial perspective-taking in young children. When children were asked first to generate an egocentric view of a dollhouse array, and then to demonstrate the doll's view, young children were able to respond nonegocentrically. The salient findings of this research, for present purposes, may be summarized succinctly: spatial perspective-taking in children was not shown to be a function of subject age alone. Rather. it was shown to be an interactive function of subject age and task type. The performance of younger children, in whom refined spatial perspective-taking abilities were not expected, could be enhanced significantly, either by altering the physical characterstics of the stimulus array, or by altering the strategy with which the child approached the problem. These results further indicated that an investigator might also flmft the performance of younger children, by manipulating strategy or stimulus characteristics to make perspective-taking more difficult. This is a critical polnt. If experimentation on spatial perspective-taking had remained within the strict parameters of the traditional three-mountain task, it might have been assumed that young children were simply incapable of perspective-taking. If, on the other hand, research in this area had begun and ended with task frameworks which were optimal for the performance of younger children, the actual organismic changes that occur through the course of child development might never have been detected; currently, it might be believed that development produces no qualitative changes in the young child's perspective-taking abilities. The use of a wide range of tasks, and of systematic age-by-task analyses (e.g., Schachter & Gollin, 1979; Walker & Collin, 1977). however, has precluded these conclusions; it is now understood that spatial perspective-taking depends both upon age and task characteristics, rather than upon either alone. Precisely the same type of issue arises with research on the visual cognitive abilities of young and elderly adults. In general, experimental paradigms in this area have not systematically manipulated age as a variable against task types. This raises the same problem obsenred in studies of spatial perspective-taking: any
302
Sharps
given study, no matter how well-designed. provides only information about the performance of older people within the range of tasks examined. Without specific attention to task selection, critical issues of plasticity and of the availability of abilities are left unevaluated. We can best address these issues by including a spectrum of experimental tasks, selected systematically against the variables of interest, and administering this systematic series of tasks to samples drawn from across the adult lifespan. One area of aging research within which these developmental methods have been useful is the study of memory for spatial relations, Spatial memory is obviously of considerable importance to living independently, a concern to the majority of the aged. Much research on spatial memory has employed some variant of a paradigm in which a group of items (such as toy objects, pictures. letters or printed words) are arranged in an array in some sort of context (such as a room, a map. a matrk of boxes, or a model town). The m a y of items is presented to the research subject for encoding. After some period of time, the subject is asked to demonstrate memory for the locations, and in some cases the identities, of the items employed. The kinds and scales of the task contexts used in this work have been extremely variable. Reviewers (e.g.. Weatherford. 1985) have distinguished to some extent between large- and small-scale spaces, and Siegel, Herman, Allen, and Kirasic (1979) have shown that one may cautiously generalize between studies employing test spaces of dif€erent scales. However, no systematic evidence concerning the extent of such generalizability is presently available. A number of studies have been focused upon the developmental course of spatial memory across the adult lifespan. Perlmutter, Metzger, Nezworski, and Miller (1981)examined the memory of young and elderly adults for the locations of buildings on a schematic map. Age-related declines in spatial memory were observed: sixty-four year-olds recalled locations significantly less accurately than did twenty year-olds. Park, Puglisi. and Lutz (1982) examined young and elderly subjects' recall of the left-right positions of pictures presented singly or in pairs. Age-related declines were also observed in this study. Similar results were obtained by Park, Puglisi. and Sovacool (1983)in a study of
Visual Cognition
303
memory for line drawings and printed words located on a large card, and by Pezdek (1983)in a study of memory for items placed in a six-by-six square matrix. Other investigators also observed agerelated declines in spatial memory using schematic maps (Light & Zelinski. 19831, chess problems (Charness. 1981). urban landmarks (Evans. Brennan, Skorpanich. & Held, 1984). slides of a building interior (Weber. Brown, & Weldon. 1978), and the Tactual Performance Test of the Halstead Neuropsychological Battery (Moore, Richards, & Hood, 1984). These studies, taken in combination, seem to provide overwhelming evidence for agerelated decline in spatial memory. However, in a few studies, a somewhat different pattern emerged. McCormack (1982)found no difference in spatial memory between young and elderly subjects in a study of memory for the location of words printed in vertical arrangements or in quadrants on a large card. Waddell and Rogoff (1981) conducted a study in which middle-aged and elderly women were asked to remember the locations of small objects, which were placed either in an array of cubicles or in a realistic, small-scale panorama. Elderly subjects exhibited significantly poorer spatial memory performance than did middle-aged persons in the array condition, but did not do so in the panorama condition. This study provided evidence that the relationship between aging and spatial memory may not be fixed and determinative. Rather, the characteristics of the task employed may differentially influence the performance of young and elderly subjects. Results consistent with this conclusion were obtained (Sharps & Gollin, 1987a) in a study of memory for the locations of objects placed in constrasting contexts. This study explicity employed Gollin's age-by-task model. A large room was furnished with unconventional "structures," including cable-spools, piles of boards, piles of fabric, and so on. The unconventional context was intended to obviate the potential influence of the experiences of young and elderly subjects with different kinds of real-world spaces. A precise. black-and-white schematic map of that room was drawn. Subjects were asked to remember the locations of common objects placed either in the room or on the map. Within the map condition, elderly subjects performed at significantly lower levels
304
Sharps
on this task than did college-aged control subjects, a finding consistent with most of the literature in this area. However, within the room condition, the spatial memory of the elderly was significantly enhanced relative to that of the younger subjects. This indicated that spatial memory in young and elderly adults was in fact task-dependent; some characteristics of the task contexts employed differentially influenced the memory performance of the two age groups. A follow-up experiment was conducted, using table-sized arrays in which the structures were again blocks, piles of fabric, and so on. These arrays varied in the "visual distinctiveness" of component "structures," from a black-and-white schematic map on which the structures were merely line drawings, to a "painted model" on which structures were three-dimensional and differently colored. The "room" condition of the previous experiment was also replicated. It was found that elderly persons performed at significantly lower levels than did young subjects when the objects to be remembered were placed on the schematic map. However, the performance of the elderly improved significantly, both absolutely and relative to that of younger subjects, when color, dimensionality, or both were added to the spatial memory arrays (Sharps & Collin, 1987a). These results demonstrated that age-related deficits in spatial memory are more complex than might at first be apparent. It would appear that the age-related declines generally obsenred in spatial memory may be increased within more visually diverse task contexts, and reduced within contexts whose elements are less visually distinctive. Evidence is now emerging, both from the author's laboratory and from others (e.g.. Park, Cherry. Smith & Lafronza, in press), that these specific beneficial effects of contextual distinctiveness on the spatial memory of the elderly may be confined to specific procedures or populations. For example, no such effects are observed when items to be remembered are identifiably related by category (Sharps, 1990). However, when the procedures and samples of Sharps and Gollin (1987a) were employed, and an age-by-task analysis was conducted, spatial memory performance was shown to be a function both of age and task type, This indicates that studies of aging and spatial memory may benefit from specific consideration and systematic variation
Visual Cognition
305
both ofthe ages of the subjects employed and of the tasks used to test them. Similar benefits of the use of age-by-task methods may be observed in other areas of aging and visual cognition as well, including the study of mental imagery abilities and the influence of the aging process on these abilities. In recent years, the study of mental imagery has become a particularly active area, in part because of the interest intrinsic to the subject, and in part because mental imagery paradigms may be used to address a variety of issues, such as the rate at which cognitive processes occur. Many of these issues are critical in the psychology of aging, and numerous investigations of the relationship of the aging process to mental imagery performance have been made. However, the bulk of the investigations reported in the literature have not systematically evaluated the influences of age- and task-related factors against each other. Since Gollin's age-by-task methods have been shown to be of use in the closely-allied area of aging and spatial memory research, it was felt that such an approach to aging and mental imagery might also be productive. What follows is a brief review of some of the most important findings, for present purposes, in the area of aging and mental imagery. Special attention is given to the study of mental image rotation, one of the most productive research frameworks employed in the area. A new experiment on aging and mental image rotation, based on this review and employing the age-by-task method described above, is also reported. MENTAL IMAGERY: BACKGROUND AND PERTINENT FINDINGS
Scholarly attention to mental images dates back at least to the time of Aristotle. Mental imagery was a major focus at the advent of scientific psychology, as well: Wundt (e.g.. 1894) regarded an understanding of mental imagery as crucial to comprehension of the elements of consciousness which constituted the subject matter of the structuralist school. However, imagery fell from favor within the mainstream rather quickly. The study of imagery was especially shaken by the idea of "imageless thought." As is well known, Watson attacked the idea of "mind" in a seminal paper
306
Sharps
(1913). The process of thinking, in Watson's view, could most parsimoniously and accurately be considered without reference to the idea of images: within Watson's formulation, imagery was subsumed by subvocal thinking, The images which become subjectively evident when one thinks about specific experiences or events were. for Watson, the result of being able "to carry on a conversation about (events) either to ourselves. ,. or with someone else" (Watson, 1928,p. 77). Watson furthermore held that, since images could not be proven to exist. they were therefore either nonexistent or unimportant for a science which must focus on the objectively verifiable alone. As Kosslyn (1980)wrote, Watson "seems honestly to have believed that imagery was simply talking to oneself" (p. 455). The behaviorist perspective advocated by Watson dominated experimental psychology through the middle years of this century, and, not surprisingly, thoughout this period little empirical interest was paid to questions of mental imagery. However, with the rise of modem cognitive psychology in the latter half of the century, more references to the subject began to appear once again. For example, work by Paivio on imagery and verbal processes, culminating in the dual coding theory (see Paivio. 1971, 1986,for review discussions), focused attention on evidence for the existence of imagery a s a process functionally autonomous from. if interactive with, verbal cognition. There are major differences between the mental imagery research of the present time and that of the turn of the century. In part as a result of the behaviorist emphasis on strict scientific method, post-behaviorist researchers have paid far more attention to methodology than previously, as well as to the creation of more objectively verifiable methods of measuring performance on mental imagery tasks. One of the most important of these methods has been the use of mental chronometry, the measurement of how much time is required to perform a given cognitive manipulation. An excellent example of the use of mental chronometry is seen in the work of Kosslyn and colleagues (e.g., Kosslyn, 1973;Kosslyn. Ball, & Reiser, 1978)on the scanning of images. Kosslyn's work demonstrated that the time required to scan an image varies a s a function of the distance to be scanned. This research was criticized
VLsuaZ Cognition
307
on the grounds that results might have reflected the subjects' tacit beliefs or knowledge about scanning rates, rather than the process of scanning itself (e.g. Pylyshyn, 1981; Intons-Peterson, 1983). However, these criticisms were answered by subsequent research (e.g., Reed, Hock, & Lockhead. 1983) in which "tacit knowledge of scanning rates" could not reasonably have affected performance. The bulk of the available research seems to support Kosslyn's contention that subjects scan images in a manner analagous to that in which they scan percepts. This research has contributed substantively to the view that images are not merely manifestations or special cases of verbal activity. Another use of mental chronometry in the study of imagery is to be found in the work of Shepard and colleagues. In 1971. Shepard and Metzler presented a mental chronometry paradigm involving the mental rotation of images. The stimulus materials employed were pairs of abstract figures. each composed of ten white cubes, arranged in configurations involving different right-angle turns (see Figure 1). The two figures were either identical in configuration or were stereoisomers, and were rotated relative to one another. Angles of rotation were varied systematically. and images were rotated either in the picture plane or in depth. The pairs of figures were shown sequentially to adult subjects, who were asked to decide whether the figures were the same or were difTerent (stereoisomers). The reaction times of subjects in this experiment increased as a function of angular rotation, both for picture plane and depth rotation, providing evidence that subjects do in fact treat mental images a s though they are "real." The manipulation of mental images was shown to be a process that takes time, and subjects required more time for tasks requiring more manipulation. Further studies (Cooper & Shepard, 1973a,b),in which subjects saw rotated alphanumeric characters, and were required to determine if a given character was "normal" or a mirror image of the normal configuration, supported this hypothesis: the average time to make this determination increased with increasing angular departure from the character's normal orientation. However, when subjects were provided with information on the upcoming character and its orientation, reaction times dropped to about 400 milliseconds
308
Figure 1.
figures.
Sharps
A Shepard-Metzler item, requiring rotation in depth of cube
regardless of orientation. Subjects reported that, when provided with this information, they prepared for the rotated stimulus items by "imagining" normal versions of the characters rotated into the orientations provided, and were then able to test the match of character to orientation. The results of these studies, and of subsequent research employing a wide variety of stimulus materials and paradigm variations (e.g.. Cooper, 1975, 1976;Cooper & Podgorny. 1976; Corballis. 1986;Corballis, Nagourney, Shetzer, & Stefanatos. 1978; Corballis. Zbrodoff, & Roldan, 1976; Eley. 1982; Kubovy & Podgorny, 1981; Metzler & Shepard. 1974; Petrusic, Varro, & Jamieson, 1978: Steiger & Yuille, 1983;Tapley & Bryden, 1977: White, 19801, were consistent with subject self-reports of the rotation of images. Mental image rotation appears to be a real-time process, within which rotations through larger angles require more time than rotations of smaller magnitude.
Visual Cognition
309
The work of Shepard. Cooper and colleagues provided more than validative evidence for the real-time nature of mental imagery processes. The mental image rotation (MIR) paradigm also provided an excellent method of examining issues in the study of aging and mental imagery. A variety of studies (e.g.. Elias & Kinsbourne, 1974:Plude. Milberg, & Cerella, 1986;Salthouse, 1987) have focused on questions of imagery, spatial abilities involving imagery, and aging. Generally, some age-related declines in performance have been observed. For example, there is considerable evidence for declines in cognitive speed with advancing age (Birren. 1974).probably due to a general age-related slowing of the central nervous system (also see Cerella. 1985. and Salthouse, 1982, 1985). Specifically, Birren holds that the aging process is accompanied by a generalized, physiologically-mediated slowing of central nervous system functioning. The behavioral artifacts of this slowing are expected to include diminished speed of behavior and diminished efficiency of performance in other task contexts involving speed requirements. Mental image rotation procedures have been used to address both cognitive speed and the accuracy of decisions made under speeded conditions. The results have in general agreed with Birren's hypothesis. However, the results of these studies have not been entirely consistent. The most critical findings, for present purposes, are reviewed below. Gaylord and Marsh (1975)presented young (20year-old) and elderly (65-72year-old) male subjects with Shepard-Metzler cube figures on a tachistoscope. Depth rotation figures were employed. There were significant declines with age in both the accuracy of decisions and the speed at which they were made. Although several possible interpretations of these data could be made, these findings are consistent with Birren's (1974)hypothesis of generalized CNS slowing. Jacewicz and Hartley (1979)provided data which have been interpreted as conflicting with the work of Gaylord and Marsh, and with the speed-loss hypothesis. They presented young (22 year-old) and older (56year-old) English-speaking subjects with letters from the familiar Roman or from the Greek alphabet. In each trial, a letter was presented in its normal orientation. This was followed by the presentation either of the same letter or of its mirror image,
310
Sharps
rotated in the picture plane with reference to normal orientation. The experimental design allowed the separation of 'basic speed" (obtained from the latency of responses when two objects are presented at the same orientation) from the rate of actual image manipulation (rotation). The letters were rotated at angles varying from zero to 180 degrees. When the familiar Roman letters were employed as stimulus items, the performance of young subjects did not differ from that of the elderly, in basic speed or in rate of image rotation. When unfamiliar Greek letters were employed, the elderly were slower than the young subjects in their basic speed. However, the rate of manipulation did not differ with age. Decision accuracy did not dmer significantly between age groups. The results of this experiment were construed as failing to support the notion of generalized CNS slowing with age, and as a failure to replicate the work of Gaylord and Marsh (1975). These findings were strongly criticized by Cerella. Poon, and Fozard (19811, who employed the stimulus materials of Jacewicz and Hartley (1979) to test, in a standard mental image rotation task, a population more similar to that of Gaylord and Marsh (1975). It was noted that Jacewicz and Hartley employed elderly experimental subjects who were currently enrolled in college classes, and who were on average younger than those employed by Gaylord and Marsh. Cerella et al. (1981) tested elderly subjects ranging in age from 66 to 77 years, recruiting not from college classes but from Boston area housing projects. Although decision accuracy did not differ significantly between age groups, this study resulted in a n imposing 96% age-related decline in speed of response. Cerella et al. concluded that the failure of Jacewicz and Hartley to find age differences in their procedures resulted from their use of "unusually young and active" subjects (p. 621). Cerella et al. further concluded that their findings provided strong support for the generalized CNS slowing postulated by Birren (1974). It is important to note that subject characteristics were not manipulated in the work of Cerella et al. The idea that differences in the activity level and education of specific subject groups influences performance on imagery tasks remains hypothetical; controlled studies of specific subject-related variables would be necessary to test this hypothesis. However, considered together
Visual Cognition
31 1
with the studies of Gaylord and Marsh (1975) and Jacewicz and Hartley (19791, these results indicate that subject-related variables may be an important point for future study, and may have profound influences on cognitive functioning generally and mental imagery particularly. A variety of influences, including demographic variables, life-style, or health-related factors, may have contributed to differences in performance between the populations of Cerella et al. and Jacewicz and Hartley. Furthermore, Cooper (1982) has shown that there are important individual Werences in the cognitive style with which problems of mental image rotation are approached: different individuals tend to take either a n analytic or a wholistic approach to MIR problems, which influences performance significantly. In light of these results, it is clearly prudent within any given experimental framework to consider the possibility of subject-related influences on performance. Like Cerella et al. (1981). Berg, Hertzog, and Hunt (1982) provided strong support for generalized CNS slowing. "Palmer figures", which are stick-like configurations generated by connecting points on a 3 x 3 dot matrix (see Palmer. 1977). and which are rotated within the picture plane, were employed as stimulus items in a mental rotation task. Significantly increased reaction times with age were observed. However, no age differences were obsenred in error rates. The Primary Mental Abilities Figures Test (PMAF). a test of spatial ability, and the Nelson-Denny Vocabulary Test (NDV), a test of verbal ability, were also administered. Reaction times and the slopes of reaction time/rotational angle functions were significantly correlated with performance on the spatial abilities test. but not with performance on the verbal test. Subjects in this study were tested over four consecutive days. Significant effects of repeated exposure to the stimulus materials were observed: reaction times and the slopes of reaction time/rotational angle functions decreased for both young and elderly adults, although age differences were neither eliminated nor systematically reduced. The repeated exposure of the stimulus items in this study occured over a period of four days; this research did not address putative effects of long-term familiarity or practice
312
Sharps
on performance, Therefore, these results should not be construed as ruling out possible age-differential effects of longer-term familiarity. However, within the parameters of this study, the findings of Berg. Hertzog 'and Hunt indicated that age-related daerences in mental rotation reaction times did not result from the novelty or unfamiliarity of the stimulus items. As noted above. these results are consistent with the generalized speed-loss hypothesis of Birren ( 1974). Herman and Coyne (1980)addressed mental image rotation from a somewhat different perspective. They asked subjects in three age groups (20-, 60-, and 70-year olds) to engage in a spatial perspective-taking task, specifically the determination of the locations of target objects from imagined points of reference. Subjects were also asked to imagine the rotation of a n array of objects relative to their present position. Interestingly, no age difference was found in the rotation task. However, the spatial perspective-taking task produced better performance in the young than in the older subjects. This indicated that age differences in mental Imagery abilities were task-dependent. If the two tasks employed in this study had produced the same pattern of results for young and elderly subjects, this work might have been considered support for the idea of a unitary cognitive process of mental imagery, within which age-related deficits could be safely charted. Instead, the disparate pattern of results actually obtained indicates that the constellation of abilities which constitute mental imagery are just that: a constellation of cognitive abilities, rather than a unitary process. The findings of Herman and Coyne (1980)regarding task dependency also reinforce the idea that, even in the absence of evident subject-related factors such as those cited by Cerella et al. 11981).conflicts of results are not necessarily irreconcilable unless they result from precisely the same procedures. Evident discrepancies between the findings of different studies may arise as a function of differences in the experimental tasks employed. It is therefore important to test. systematically, the influences of various task types against the influences of aging within the general area of mental imagery. It should be pointed out that this
Visual Cognition
313
type of analysis is exactly what Collin's developmental, age-by-task approach is designed to accomplish. Krauss. Quayhagen. and Schaie (1980) used several tasks to address mental imagery abilities. Elderly subjects (aged 64 to 76 years) were tested with the standard, paper-and-pencil Spatial Relations Test of the Primary Mental Abilities (PMA) test (Thurstone. 1962, revised edition), with slide projector versions of the PMA test. and with a mental image rotation test involving "Attneave-Arnoult" (AA) stimuli. These stimulus items are irregulary shaped, two-dimensional, faceted solid black figures, generated and used by Cooper (e.g.. 19751, and derived by the method of Attneave and Arnoult (1956). Along with the speed of mental rotation, memory for the test stimuli was also evaluated. The results of the AA stimuli test indicated that memory for test stimulus items was important for performance at higher levels of MIR performance, although less so for subjects who performed more poorly. In other words, memory for items was a better predictor of performance in the upper portion of the performance curve than in the lower portion. It was also found that subjects experienced more difficulty with the paper-and-pencil PMA than with the projector versions: although the proportion of correct responses to the number of items attempted did not differ significantly between the PMA task types, the absolute number of correct responses was significantly higher in the projector conditions, indicating that performance was inhibited by the "mechanical barriers present in paper-and-pencil tasks" (p. 205). This study sewed as additional evidence that the manifestation of imagery abilities is task-dependent. Clarkson-Smith and Halpern (1983)tested three age groups of women (aged 18-28, 50-60, and 70-80 years of age) in a mental image rotation paradigm using "semi-abstract" figures, drawings representing a swan, a boat, a face or a witch on a broom. Meaningful or nonmeaningful labels were assigned to the figures. The effect of meaningful labels was to decrease errors in decision for older subjects, indicating that verbal mediation may be used to compensate for age-related deficiencies in spatial abilities. However, older subjects in this study did make more errors than did younger subjects. This finding is congruent with the results of
314
Sharps
Gaylord and Marsh (1975),who found that more errors were made by elderly than by young subjects when Shepard-Metzler cube figures were employed. However, these findings differ from others In the literature (e.g.. Berg et al.. 1982: Cerella et al., 1981;Jacewicz & Hartley, 1979). As discussed above, the results of these latter investigations indicated that young and elderly adults did not tend to differ significantly in the accuracy of their rotation-based decisions. The results of Clarkson-Smith and Halpern (1983) and of Gaylord and Marsh (1975),when compared with other research in the area, indicate that both the accuracy of mental image rotation decisions and the speed of mental rotation are influenced by task characteristics. Sharps and Gollin (1987b).using Shepard-Metzler cube figures rotated in depth, addressed the effect of speed requirements on the speed and accuracy of mental image rotation judgements in young and elderly adults. Mental image rotation tasks in general require subjects to work as fast as possible. In this study, young and elderly adults were asked to work either as rapidly as possible, a s accurately as possible, or to balance speed and accuracy. It was found that when asked to work a s quickly a s possible, elderly adults performed at the same rate of speed as young subjects, but made significantly more same/different judgement errors. When asked to respond accurately, without reference to speed, elderly adults responded with the same decisional accuracy a s young subjects, but required more than twice as much time on average to reach these decisions. When asked to balance speed and accuracy, the elderly responded with the same decisional accuracy a s younger persons, although, again, older individuals took longer on average to make these judgements. Interestingly, the elderly performed at the same level in both the accuracy and speed/accuracy conditions, although they used twice a s much time to arrive at decisions in the former condition. Subjective self-reports Indicated that older subjects used more caution in the accuracy condition; in the absence of speed requirements, they reported taking additional time to check their work. The results of this study are consistent with Birren's (1974) speed loss hypothesis, in that the elderly exhibited slower reaction times than the young when speed requirements were imposed.
Visual Cognition
315
However, since the elderly were substantially slower in the accuracy than in the speed/accuracy condition, but performed at the same level of accuracy in both of these conditions, part of the observed speed loss was probably due to simple caution. These findings may help to explain some of the discrepancies among results related to age-related speed loss in mental image rotation. Evidently, observed speed loss may result from a combination of factors, such a s caution and generalized slowing, rather than from one factor or set of related factors alone. These results add to the evidence that it is important to examine the possible sources of performance differences between subjects of different ages, rather than to assume that such differences reflect the influence of particular age-related variables. Herman and Bruce (1983) also addressed the question of age differences in MIR abilities, using Shepard-Metzler cube figures as stimulus items. However, instead of the tachistoscopic presentation often employed, Herman and Bruce used the paperand-pencil version of the Shepard-Metzler test developed by Vandenberg (e.g., Vandenberg & Kuse, 1978). In this version. each "problem" consists of an original figure to the left of a vertical line, and four figures to the right of the line. Two of the figures match the original but are rotated in depth between zero and 180 degrees. The other two are foils. Five of these problems are presented on each of three pages. Subjects are provided with sample problems, and are then instructed to select the two figures which match the original as quickly as possible. In this experiment (Herman & Bruce, 1983).young adults (mean age 25.3 years) performed significantly more accurately than did elderly adults (mean age 65.3 years). Herman and Bruce noted that this result differed from those of Berg et al. (19821, Cerella et al., (1981).and Jacewicz and Hartley (1979). However, age differences in accuracy were also observed by Gaylord and Marsh (1975).as well as by Wilson, DeFries, McClearn, Vandenberg, Johnson, and Rashad (1975). who employed the Vandenberg version of the test. The evident discrepancy in results among these studies was attributed by Herman and Bruce to stimulus factors; it was noted that the investigators who observed age-related declines in accuracy employed some variant of the original Shepard-Metzler cube
316
Sharps
figures, which involve the "visualization of a three-dimensional rotation from a two-dimensional figure" (p. 84). Those investigations which involved no such declines employed stimulus materials which required only the visualization of a twodimensional rotation from a two-dimensional figure. Herman and Bruce concluded that the greater "complexfty" of the cube figures was responsible for this discrepancy in findings. This is consistent with current theory: Cerella. Poon, and Williams (1980),in a survey of the literature, found that increasingly complex tasks in general result in increasing reaction times for older individuals. probably as a consequence of the "proportional slowing of the mental functions of the elderly" (p. 332). Cerella et al. (1980)also suggested that general CNS slowing should result in the slowing of all central cognitive processes approximately to the same degree. Therefore, the response times of older persons on any given task should increase as a function of the response times of younger individuals. The results of Herman and Bruce (1983)suggest that this "complexity hypothesis of Cerella et al. applies to not only the speed, but also the accuracy of spatial information processing" (p. 841. Since abundant literature suggests that mental image rotation is task-dependent, stimulus complexity would appear to be an important task characteristic to examine. Puglisi and Morrell (1986)conducted a study which was particularly pertinent to this issue. In a standard MIR procedure, young adults (mean age 21.7years) and elderly adults (mean age 70.9 years) were presented with pairs of three-dimensional toy cowboys, either the same or mirror images of each other, by means of a tachistoscope adapted for the presentation of three-dimensional objects. Significant declines in both reaction time and decision accuracy were observed with age. Also, the best fitting linear equation for these data provided slope and intercept values for the reaction time/rotational angle function which were consistent with the complexity hypothesis of Cerella et al. (1980). These findings fit well with the hypothesis of Herman and Bruce that the complexity hypothesis extends to accuracy as well a s to speed of response. Three-dimensional cowboys, rotated in depth, are complex stimulus items by comparison with Palmer or AA figures rotated in the picture plane, and subjectively appear more similar
VIsuul Cognition
317
in complexity to depth-rotated Shepard-Metzler cube figures. The data obtained from the cowboy stimulus materials were much more similar to those obtained with Shepard-Metzler depth items than to those derived from picture plane rotations. These data therefore supported the CNS slowing hypothesis of Birren (1974)and the complexity hypothesis of Herman and Bruce (1983). The results of studies reviewed thus far, when examined in combination, suggest that mental image rotation performance in young and elderly adults is task-dependent. The complexity of stimulus items appears to be one of the task-related factors which influence the performance of young and elderly subjects within a variety of MIR paradigms. What precisely constitutes "complexity" of stimulus items in a mental image rotation task? Several investigations have been directed toward the resolution of this issue. The complexity of AA figures, which rotate in the picture plane, is fairly easy to calculate. "Perceptual complexity depends strongly upon the number of points which determine inflections on the perimeter of the form" (Cooper, 1975.p. 23-24.Based on Attneave, 1957;Attneave & Amoult, 1956;Vanderplas & GaIvin. 1959). In other words, a star-shaped figure is more complex than a triangular figure, because more points and facets are present in the former. Cooper (1975).in a study of non-elderly adult subjects, found that whereas reaction time was a function of increasing rotational angle, the complexity of the figures employed did not influence reaction time, either in terms of the slope of the reaction time/rotational angle function or of its intercept. Cooper and Podgorny (1976)examined the mental rotation of AA figures, requiring subjects to discriminate previously learned versions of the figures from rotated distractor versions. Subjects were provided with advance information on the identity and orientation of the upcoming figure, and were instructed to prepare for the presentation of the test figure by rotating a mental image of that figure into the designated orientation. The time needed for this preparation was a function of the angular rotation of the test figure from the previously-learned orientation of that figure. However, figure complexity did not influence reaction time.
318
Sharps
These results differ somewhat from those of Hochberg and Gellman (1977). Using "crossbeamed" MIR stimuli (present author's designation), which somewhat resembled irregular telephone or telegraph poles, they found that reaction times varied not only with the angular rotation of the stimulus figures, but also with the presence or absence of "salient landmarks," prominent crosshatches or processes. Those figures with such landmarks had lower rotational angle/reaction time functions than those "in which these features have been made less distinguishable" (p. 25). The results were not attributed "to the greater complexity, per se. of the latter, inasmuch as Cooper (1975) found no such effects with the Attneave-Arnoult shapes" (p. 25). Rather, Hochberg and Gellman attributed these findings to the "relative inaccessibilities.. .of informative features" (p. 25). It would seem that such relative inaccessibilities might contribute to overall levels of complexity. However, whether or not inaccessibility is specifically related to complexity, the Hochberg and Gellman results demonstrate the influence of stimulus item configuration on mental image rotation. Several additional studies have clarified different facets of the complexity issue. Folk and Luce ( 1987) examined the rotation of AA figures, and found that although reaction times for dissimilar stimuli were not influenced by figure complexity. reaction times for more similar stimuli, with which discrimination should have been more difficult, increased as a function of stimulus complexity. These results were interpreted to mean that subjects rotated only incomplete images when stimuli were dissimilar, and that complexity effects resulted when subjects had to rotate entire images to come to their decisions. Bethell-Fox and Shepard (1988). using patterns of filled-in squares on 3 x 3 square matrices as stimulus items, obtained complexity effects on rotation times, although these effects were in general eliminated with practice. Yuille and Steiger (19821,using Shepard-Metzler cube figures as stimuli, found that figure complexity influenced the speed of mental image rotation. However, when featural redundancy was ignored in the generation of figures, the complexity effects were not obtained. In other words, additional complexity did not influence the speed of mental image rotation when the complicating features were not necessary to accomplish the rotation task.
Visual Cognition
319
At least two studies have addressed complexity effects in mental image rotation not from the standpoint of item complexity, but rather from the standpoint of rotational complexity. Jolicoeur. Regehr. Smith. and Smith (1985) employed Shepard-Metzler cube figures in a MIR task, either in their three-dimensional forms, rotated in depth, or with their internal line segments scrambled to form two-dimensional figures, rotated in the picture plane. Threedimensional rotation was found to require significantly more time than two-dimensional rotation for angles greater than 60 degrees, although this effect was not observed for smaller angles. Shepard and Metzler (1988) compared the results of tasks employing Shepard-Metzler cube figures, rotated in three dimensions, or AA figures, rotated in two dimensions. Either both forms were present, and one rotated "into" the other (as in the task used by Shepard 81 Metzler, 1971). or a single form was rotated into congruence with a previously learned original not actually present during rotation (as in the task used by Cooper, 1975). The rate of mental rotation was found to be significantly (about three times) higher for the singlestimulus than for the dual-stimulus task. Three-dimensional shapes required more initial encoding time than did twodimensional figures, although the results suggested that once threedimensional objects were encoded. their rotation could be imagined as rapidly as the rotation of two-dimensional shapes. This study, in agreement with the work of Jolicoeur et al. (1985). indicates that the dimensionality of the mental rotation required in a given task. as well as the characteristics of the stimulus items themselves, plays a powerful role in the determination of performance. The results of the studies considered above are summarized in Table 1. An examination of this literature indicates that the methods used to establish the relationship of rotational angle to reaction time, under different subject- and task-related conditions, have varied enormously. For example, some studies have reported only reaction times, whereas others have reported findings in terms of the slopes and intercepts of reaction time/rotational angle functions. These functions are particularly important. If the speed-loss hypothesis of Birren (1974)is correct, and much evidence suggests that it is. then different permutations of subject- and taskrelated factors might be expected to give rise to different
Table 1 Summary of Studies of Mental Image Rotation Discussed in
Authors and Year Shepard & Metzler. 1971 Cooper & Shepard. 1973 a,b Gaylord & Marsh, 1975 Jacewicz & Hartley, 1979 Cerella, Poon, & Fozard. 1981 Berg, Hertz~g.& Hunt, 1982 Herman & Coyne, 1980
Stimulus or Task Type
Shepard-Metzler Alphanumeric Shepard-Metzler Alphanumeric Alphanumeric Palmer Perspective-taking Array Rotation Krauss, Quayhagen, & Schaie, 1980 Several Clarkson-Smith & Halpern, 1983 "Semi-abstract" Sharps & Gollin, 1987b Shepard-Metzler Shepard-Metzler Herman &Bruce!. 1983 Wilson et. al. , 1975 Shepard-Metzler Puglisi & Morrell, 1986 Toy cowboys
the Text [in order of appearance)
Effect of Age
Effect of Stimulus Complexity
Factors of Special Importance
---Population? Population? Familiarity Task type Task type Memory for stimuli Meaningful labels Speed requirements Speed vs. accuracy
Table 1 (continuted)
Authors and Year Cooper, 1975 Cooper & Podgorny. 1976 Hochberg & Gellman, 1977 Folk & Luce, 1987 Bethell-Fox & Shepard. 1988 Yuille & Steiger, 1982 Jolicoeur, Regehr. Smith &smith, 1985 Shepard & Metzler. 1988
Stimulus or Task Type AA AA
"Crossbeam" AA Square matrices Shepard-Metzler
? = Uncertainty
* =Quallflcation; seetext
N N ? Y
Y* Y*
Shepard-Metzler
Y*
Shepard-Metzler
Y,
&AA
Y = Significant effect N = No detected significant effect
Effect of Age
Effect of Stimulus Complexity
Factors of Special Importance
---Informative figures Similarity of figures Practice Featural redundancy Dimensionality of rotation Dimensionality of rotation
322
Sharps
characteristic functions. These could theoretically be used to develop a systematic understanding of the interplay of the cognitive processes in question. Additional systematic analysis is needed to establish these functions, and the degree to which specifc processes are in fact reflected by characteristic functions. Even in the absence of such analysis, however, the following conclusions may be drawn from the findings considered:
There is abundant evidence that mental image rotation is a process which occurs in real time, and which requires more time to accomplish with increased angles of rotation. 1.
2. There are differences between the mental image rotation
abilities of young and elderly adults. For example, the elderly in general require more time than young adults to accomplish any given mental rotation task. There is much evidence in the literature to implicate a generalized central nemous system speed loss (Birren. 1974) in the mental rotation rate dnerence between young and elderly adults. 3. However, the performance of young and elderly subjects on mental image rotation tasks is in general task-dependent. Within different task frameworks, age-related differences in rotation rates are not always associated with age differences in the accuracy of judgements based on rotation. Factors which influence the rate and/or accuracy of mental rotation performance include the type of stimuli employed, the complexity of those stimuli. the dimensionality of rotation (depth versus picture plane), subject familiarity and practice with stimuli, speed requirements, and whether stimuli are to be compared with each other or with images of previously learned items. It is probable that this list is not exhaustive. The relationships of these various known influences on mental rotation to one another has not been systematically evaluated at the present time. The relationship of some of these factors (e.g., speed requirements) to the aging process has been partially clarified, but comprehensive systematic evaluations are currently lacking.
Visual Cognition
323
The task-dependency of mental image rotation, and the fact that some task characteristics (e.g., speed requirements) interact with aging to influence performance, indicate that mental image rotation probably involves a cluster of related abilities, which may be brought into play as task circumstances demand and as agerelated factors permit. Identification and analysis of these abilities would shed considerable light on the nature of mental image rotation and on mental imagery generally. Examination of these abilities with reference to the aging process should also allow much more specific definition and analysis of the imagery abilities retained and lost over the course of the adult lifespan. Systematic analysis of the results of a number of experiments, addressing various facets of these issues. will be needed to form a comprehensive picture of the factors and relationships involved. 4.
A preliminary experiment addressing some of these factors is
reported below. Since speed requirements have been shown to interact with age to influence performance (Sharps & Gollin. 1987b), and since stimulus complexity has been shown to influence performance on some, but not all, mental image rotation tasks (e.g.. Folk & Luce, 1987). the work presented below addressed the interactive influences of subject age, speed requirements, and stimulus complexity on the speed and accuracy of mental image rotation. Picture plane rotation of modified, two-dimensional Shepard-Metzler cube items was employed in the rotation task. The experiment was intended to provide preliminary data on the interactions of these factors, and also to demonstrate the usefulness of the age-by-task perspective in the study of aging and mental imagery. MENTAL IMAGE ROTATION IN YOUNG AND ELDERLY ADULTS: SPEED REQUIREMENTS AND STIMULUS COMPLEXITY It was previously shown (Sharps & Gollin. 1987b) that the mental image rotation performance of young and elderly subjects was differentially influenced by speed requirements: when speed was emphasized, the elderly were able to respond as quickly a s young subjects, but were significantly less accurate in their
324
Sharps
judgments. When subjects were specifically instructed not to emphasize speed at the expense of accuracy, the elderly were as accurate as the young in their judgments, but were significantly slower. These findings were demonstrated using standard, threedimensional Shepard-Metzler items, rotated in depth. Age differences in cognitive functioning may involve the effortfulness of the cognitive tasks in question. In several studies (e.g.. Craik, 1986).it has been shown that more effortful tasks (e.g.. free recall) are associated with relatively greater age-related deficits than less effortful tasks (e.g., recognition memory). This may be attributed to an age-related decline in general processing resources (e.g.. Craik, 1986: Norman & Bobrow, 1975). Such a decline in central processing resources might also be implicated in the performance of young and elderly subjects on mental imagery tasks. The processing of more complex stimuli, for example, would appear to be a more effort-intensive task than the processing of simple stimuli, and might therefore be expected to produce a greater age difference in performance. (This would only be expected, of course,If levels of complexity are not so great as to force subjects to employ entirely different strategies, such as partial rather than whole rotation of stimulus items). The rotation of simpler items should be less influenced by speed requirements, which hypothetically also tap such processing resources, than the rotation of more complex stimuli. Also, more complex stimuli should become more difficult to process with increasing angles of rotation, since both increased rotation and increased complexity presumably require increased processing resources. Complexity effects are not generally observed when twodimensional AA figures are employed a s stimuli [e.g., Cooper, 1975; Cooper & Podgorny, 1976;but see Folk & Luce, 1987).This gave rise to the hypothesis of Herman and Bruce (1983)that complexity effects are largely confined to MIR tasks involving threedimensional stimuli. However, given the processing-resource considerations discussed above, it seems probable that complexity effects may be obtained from the use of more complex twodimensional stimuli a s well. Since the use of Shepard-Metzler figures has in general resulted in the observation of complexity effects, it would appear that these figures may be intrinsically more
Visual Cognftbn
325
"complex" in some way than AA figures. Relatively complex twodimensional figures were required for the present experiment. Therefore, the two-dimensional figures used in the present experiment were modeled on Shepard-Metzler figures. The use of these stimuli provided a partial test of the Herman and Bruce hypothesis. Method The subjects of this experiment were 28 young adults (mean age 20.5 years, standard deviation 3.1 years) and 28 elderly adults (mean age 73.5 years, Standard Deviation 5.3 years). A questionnaire covering health and demographic factors was administered; all subjects reported being in good health, and no systematic demographic differences were observed between the young and elderly samples. The Vocabulary subscale of the Wechsler Adult Intelligence Scale was administered; the performance of the elderly adults in the study significantly exceeded that of young adults. 41.54)=14.18.pc.001. This finding is typical. A Snellen test of vision was also administered: all subjects exhibited visual acuity of 20/40 or better. These tests were intended as precautions against the influence on performance of population differences other than age per se. Although this battery of measures was by no means comprehensive, it was felt that major dmerences in important areas between young and elderly subjects would probably be detected by these instruments. The experiment employed a Gerbrands 3-field tachistoscope, reaction timing device and clock/counter. The apparatus was arranged so that the experimenter activated the presentation field of the tachistoscope and the timer simultaneously. The subject terminated tachistoscopic presentation and timer activity simultaneously by pressing another switch. The stimulus items employed were similar to Shepard-Metzler cube figures. being composed of squares arranged in series of rightangle turns. However. these figures were two- rather than threedimensional (see Figure 2). Rotation was to be accomplished in the picture plane. The figures were mounted on standard tachistoscope cards. Each item consisted of a rectangular black field on which
326
Sharps
were mounted two of these figures, either two representations of one figure, or representations of a figure and its mirror image. The figures in each item were rotated 20, 45. or 70 degrees from each other, all rotations being carried out in the picture plane. The figures also differed in complexity. As discussed above, it is difiicult to be precise concerning what actually constitutes stimulus "complexity." However, in order to decide whether figures were the same or were mirror images. subjects would have to examine the right-angle turns underlying the item configurations in making their decisions. Accordingly. complexity was defined in terms of the number of right-angle turns present in a given figure. The figures used were either "simple" (consisting of 10 squares with 2 or 3 turns), "intermediate" (10squares with 4 turns), or "complex" (10 squares with 5 turns). The procedure employed w a s similar to that of Shepard & Metzler (1971). Sample and practice items were administered to ascertain that subjects understood the task, which was to decide whether the pairs of figures were the "same" or "different" (mirror images). Subjects were asked to press the termination switch immediately on making this decision, and then to inform the experimenter of their decision. Subjects were either instructed to make their decisions a s quickly a s possible, or were instructed to balance speed and accuracy in making their judgements. This instructional manipulation was carried out between subjects. Results and Discussion There was no main or interactive effect of age or instructional condition on decision accuracy scores. This is interesting in light of previous similar research. using depth rotation of threedimensional Shepard-Metzler cube items, which provided evidence that speed requirements negatively affected accuracy of judgement for the elderly (Sharps and Gollin, 1987b). While previous experiments are obviously not directly comparable with the present one, given the use of difTerent stimulus materials, the discrepancy in the pattern of results obtained suggests that rotation in the picture plane may be "easier" than rotation in three dimensions. so that speed requirements have relatively little influence on rotation
Visual Cognltbn
327
Rgure 2. Examples of the two-dimensional figures, similar to ShepardMetzler figures, employed in the present experiment. From top to bottom, the figures are of simple @-turn).intermediate (4-turn). and high (5-turn) complexity.
328
Sharps
performance of the former type. This is in agreement with the work of Jolicoeur et al. (1985)and of Shepard and Metzler (1988). The results of this study regarding reaction times are summarized in Table 2. Mean reaction times, with standard deviations, are reported. Slopes and intercepts for the reaction time/rotational angles are not reported. An examination of Table 2 will reveal that, although greater rotational angles are associated with longer reaction times overall, in some cases reaction times for 45 degree angles were the same or even shorter than those for 20 degrees. This perturbation was unexpected, and may relate to the greater “familiarity” of 45 degree angles to subjects. The question is currently under investigation. However, for the purposes of the present study, slopes and intercepts for these functions over all would not be representative of the actual data. The effect of age on reaction time was significant, Fl1.52)=16.69. p < . O O l , Elderly subjects were significantly slower than young subjects across instructional conditions. This finding is in agreement with other research (e.g., Berg et al., 1982) supporting Birren‘s (1974) speed loss hypothesis. although, a s has been discussed above and elsewhere (Sharps & Gollin. 1987b).the degree of speed loss observed with advancing age appears in part to be a function of task condition. There was also a trend, albeit nonsignificant (PI 1,52)=3.41, p=.070),toward an effect of speed requirements on rotation rate: subjects appeared to work at higher rates within the speed than within the speed/accuracy condition. The trend was nonsigdcant and very weak, whereas in previous work speed requirements made a significant difference to reaction time (Sharps & Collin. 1987b). This can probably be attributed to the use of two-dimensional rotation, rather than rotation in depth, in the present experiment. These results are consistent with the hypothesis that twodimensional rotation, requiring less in the way of “processing resources,” may be relatively less influenced by speed requirements than more complex, three-dimensional tasks. This hypothesis However, will, of course, require further empirical evaluation. evidence is accumulating rapidly to the effect dependent upon a variety of stimulus-related factors, as well as upon procedural factors such as speed requirements, Hertzog and Vernon (1990).
329
Visual Cognition
Table 2 Mean Reaction nrnes (in seconds) of Young and Elderly Adults, with Standard Devfations. Rotational Angle
Complexity
M
SD
M
SD
20
3T
1.96
1.54
3.37
1.88
4T
3.09
1.63
6.04
3.26
5T
3.18
1.78
5.16
2.86
3T
1.90
0.98
3.35
2.37
4T
2.13
1.43
3.41
1.83
5T
2.45
1.40
4.43
2.84
3T
2.63
1.45
4.11
2.55
4T
3.66
2.32
6.29
4.18
5T
4.19
3.03
8.05
6.86
45
70
using two-dimensional stimulus items, recently observed that although both young and elderly respondents altered their reaction times in response to speed instructions, there was no interaction of age with instruction. In the present study, which also used twodimensional items, no significant main or interactive effects of speed requirements on rotation times or accuracy scores were observed. Both of these studies are at odds with Sharps and Gollin (1987b). in which main effects and interactions of speed requirements and age were produced for both reaction times and accuracy scores. However, recent work (in preparation) in the author's laboratory. employing three-dimensional stimulus items similar to those of Sharps and Gollin (1987b), has produced a pattern of results in agreement with the Sharps and Gollin study. These preliminary results are entirely consistent with the present thesis that mental image rotation performance depends upon a complex interplay of age, stimulus factors, and procedural demands.
330
Sharps
The effect of rotational angle on reaction time was significant, F(2.104)=26.41. pe.001. As has been found many times previously. greater rotational angles were associated with longer reaction times. However, there was no interaction of the influence of age with that of rotational angle. Although the elderly were in general slower than young adults, they were not placed at a relative disadvantage by increasing rotational angles. The performance of the elderly paralleled that of younger persons, the degree of required rotation notwithstanding. This indicated that increased rotation did not draw increased "processing resources." in the sense in which these have been discussed above. The interaction of complexity and rotational angle was significant, F(4.208)=5.73. pe.00 1. More complex items required more time to rotate through greater angles. This finding was consistent with the hypothesis advanced above, that more complex items may require more processing resources for rotation than simpler items. This further provides a qualification to the hypothesis of Herman and Bruce (1983) that complexity effects are more likely to be observed in three- than in two-dimensional rotation. Rotation involving the two-dimensional stimuli used in the present study exhibited significant complexity effects. This indicated that complexity effects on reaction time result from stimulus configuration as well as from the dimensionality of required rotation. Further research will be required to ascertain precisely why some types of two-dimensional stimuli are associated with complexity effects, while others are not. The work of Folk and Luce (1987).discussed above, is pertinent to this issue. The effect of figure complexity on reaction time was significant, F(2.104)=31.26, pe.001. More complex figures required more time for rotation than did less complex items. This was in agreement with most of the basic data for Shepard-Metzleritems, although not for AA items, again demonstrating the stimulus-dependency of M I R The interactive effect of age and figure complexity was also significant, F(2.104)=3.70.pc.03 (see Figure 3). This is consistent with the hypothesis of age-related depletion of processing resources: increased resource utilization, resulting from increased complexity of figures, was more deleterious for the performance of the elderly than of the young subjects. The interaction of age and
Visual Cognition
331
6000
5000
4000
2000
!
I
I
I
SIMPLE
INTERMEDIATE
COMPLEX
STIMULUS COMPLEXITY
mure 3. Mean reactfon tlmes of young and elderly subjects, plotted against the complexity of stimulus flgures employed.
rotational angle was not signiflcant. Neither was the three-way interaction of age, rotational angle and complexity. Although increased rotational angles did not negatively affect the performance of elderly persons relative to that of young adults, the complexity of stimulus items did have such an effect. This was consistent with the hypothesis that more complex stimulus items place a greater demand on available processing resources than simpler figures. Diminished central processing resources are associated with the course of aging (e.g.. Craik. 1986). Therefore, more complex stimulus items are more difficult to work with for the elderly than for young adults. Since increased rotational angles did not result in different performance trends between young and elderly adults, however, it seems that increased rotational
332
Sharps
requirements probably do not draw on processing resources to the same degree, or in the same way, as increased figure complexity. It is not known at present why this should be the case: current research is focused on this issue. An additional question which arises from this work is the nature of the "processing resources" to which reference has been made. The interactive effects of complexity x angle, and of age x complexity. were significant. If, however, these effects were due to the influence of the respective independent variables on some sort of unitary bank of processing resources, one would clearly expect the age x angle interaction, and the three way interaction of age, complexity. and rotational angle. to be significant. They were not. This indicates that, with reference to mental image rotation performance, there probably is no specific reserve of unitary "processing resources" drawn on in accordance with the needs presented by specific independent variables. Rather, it is logical to hypothesize that there are a variety of such resources, which become differentially important depending upon the types of subjects employed (young or elderly, in the present instance) and upon the constellation of independent variables with which those subjects are presented. The term "processing resources" is used here as a label of convenience, to describe a group of processing-related variables whose nature and function is not currently understood. The present study gives an indication of the way in which an "ageby-task" methodology may be used to obtain information on the ways in which these variables function in different experimental contexts, employing different subject populations. The present study also demonstrates a more concrete methodological point. In this work, if simple figures alone had been used as stimulus items, the magnitude of the observed main effect of age would have been relatively small. If only more complex stimuli had been employed, a large age-related deficit in the process of mental imagery might have been erroneously postulated. It should also be noted that, depending on the task employed (see Folk & Luce, 1987), if A A figures had been used instead of the figures actually used. it is probable that no complexity effects at all would have been observed (see Cooper, 1975; Cooper & Podgorny. 1976). Other stimulus materials, such as three-dimensional Shepard-
Visual Cognition
333
Metzler items, might have produced still different effects. The results of this study demonstrate that, in order to build a comprehensive model of the interaction of age. stimulus complexity, and rotational angle on mental image rotation, It is important to test these effects among a variety of systematicallyrelated stimulus types. The study of knowledge utilization is predicated upon valid knowledge of the ways in which information. including visual information. is processed. If erroneous or oversimplified models of image processing, "processing resources," or other variables critical to cognitive performance are constructed, adequate theories of the higher processes involved in knowledge utilization and actualization cannot be constructed. It is therfore critical for studies of imagery, of knowledge utilization, and of cognition generally that theories concerning the basic processes of visual cognition be formulated as accurately and comprehensively a s possible, with reference to the types of experimental and situational factors which impact on performance. and with respect to the influences brought to the research situation by different subject populations. The age-by-task perspective discussed in the present paper has been shown to be a useful instrument for the examination of these influences. Additional studies employing this perspective should clarify these issues further, a s the interactive effects of additional task-related and subject-related factors are systematically evaluated. SUMMARY AND CONCLUSIONS
The developmental method discussed in this chapter involves the systematic evaluation of age and task characteristics within the same experimental framework. This allows the investigator to examine influences on cognitive performance that would not otherwise be detected. Appropriate use of the age-by-task method may help to prevent the formulation of oversimplified or invalid conclusions about the cognitive abilities of the elderly, and may improve the accuracy of estimates of the effect that aging actually does or does not have on a given cognitive process. Furthermore, by varying task characteristics systematically along continua of
334
Sharps
theoretical interest, the investigator may be in a better position to assess the extent of the abilities avaftable to the elderly, as well as the performance tendencies that may be expected from elderly individuals within any single task context. The utility of developmental age-by-task methods for the study of aging and visual cognition has been discussed and demonstrated. The continued use of these methods to examine cognitive performance, across a broad spectrum of populations and paradigm types, should prove useful in the development of comprehensive theories of aging and visual cognition. ACKNOWLEDGMENTS
This work was supported by a Faculty Grant-in-Aid from the University of Wyoming. The author wishes to thank Thomas Hess of North Carolina State University, Karen Barrett of Colorado State University, and Narina Nightingale of the University of Wyoming, for their very helpful comments on an earlier draft of this chapter. REFERENCES Attneave, F. (1957). Physical determinants of the judged complexity of shapes. Journal of Experimental Psychology. 53. 22 1-227.
Attneave, F.,& Arnoult. M.D. (1956). The quantitative study of shape and pattern perception. Psychological Bullettn, 53, 452471.
Berg,C., Hertzog, C., & Hunt, E.
(1982). Age Merences in the speed of mental rotation. Developmental Psychology. 18, 95- 107. Bethell-Fox, C.E.. & Shepard, R.N. (1988). Mental rotation: Effects of stimulus complexity and familiarity. Journal of Experimental Psychology: Human Perception and Perfonnance. 14, 12-23. Biederman. I. (1977). On processing information from a glance at a scene. In S . Treu (Ed.), User-oriented design of interactive graphics systems. New York ACM. Birren, J.E. (1974). Psychophysiology and speed of response. American Psychologist, 29, 808-815. Borke, H. (1975). Piaget's mountains revisited: Changes in the egocentric landscape. Developmental Psychology, 1 1 , 240-243.
Visual Cognition
335
Brodzinsky, D.M., Jackson, J.P., & Overton, W.F. (1972). Effects of perceptual shielding in the development of spatial perspectives. Child D~v~ZOJNTE~~, 43. 1041-1046. Cerella. J. (1985). Information processing rates in the elderly. P ~ ~ ~ b l ~B~lletin. i c a l98,67-83. Cerella, J., Poon, L.W., & Fozard. J.L. (1981).Mental rotation and age reconsidered. Journal of Gerontology,36, 620-624. Cerella. J.. Poon, L.W.. & Williams, D.M. (19801. Age and the complexrty hypothesis. In L.W. Poon (Ed.). Aging fn the 1980s: Psychological &sues (pp. 332-342).Washington, D.C.: American Psychological Association. Charness, N. (1981).Visual short-term memory and aging in chess players. Journal of Gerontology, 36. 615-619. Clarkson-Smith, L.. & Halpern. D.F. (1983). Can age-related deficits in spatial memory be attenuated through the use of verbal coding? Expertmental Aging Research, 9, 179-184. Cooper, L.A. (1975).Mental rotation of random two-dimensional shapes. Cogntttue Psychology, 7, 20-43. Cooper, L.A. (1976). Demonstration of a mental analog of an external rotation. Perception and Psychophystcs, 19, 296-302. Cooper, L.A. (1982). Strategies for visual comparison and representation: Individual differences. In R.J.Sternberg (Ed.), Advances in the psychology of human intelligence, (uol I : pp. 77-124).Hillsdale, N.J.: Lawrence Erlbaum Associates. Cooper, L.A., & Podgorny. P. (1976).Mental transformations and visual comparison processes: Effects of complexity and similarity. Journal of Experimental Psychology: Human Perception and Performance. 2. 503-514. Cooper, L.A.. & Shepard, R N . (1973a).The time required to prepare for a rotated stimulus. Memory and Cognition, 1, 246-250. Cooper, L.A.. & Shepard. R.N. (1973b).Chronometric studies of the rotation of mental images. In W.G. Chase (Ed.), Visual information processing (pp. 75-176.) New York: Academic Press. Corballis, M.C. (1986).I s mental rotation controlled or automatic? Memory and Cognitton. 14, 124-128. Corballis, M.C., Nagourney, B.A.. Shetzer. L.I., & Stefanatos. G. (1978).Mental rotation under head tilt: Factors influencing the location of the subjective reference frame. Perception and P ~ y ~ h o p h y ~24, i c263-273. ~. Corballis, M.C.. Zbrodoff. J.. & Roldan. C.E. (1976).What's up in mental rotation? Perception and Psychophysics. 19, 525-530.
336
Sharps
Craik, F.I.M. (1986). A functional account of age differences in memory.In F. Klfx & H. Hagendorf (Eds.), Human memory and cognitive capabilities (pp. 409-422). Amsterdam: Elsevier Science Publishers. Eley, M.G. (1982).I d e n t w g rotated letter-like symbols. Memory and Cognition, 10.25-32. Elias, M.F.. & Kinsbourne. M. (1974).Age and sex Werences in the processing of verbal and nonverbal stimuli. Journal of get onto log^^, 29, 162-171. Evans, G., Brennan, P., Skorpanich. M.A., & Held, D. (1984). Cognitive mapping and elderly adults: Verbal and location memory for urban landmarks. Journal of Gerontology. 39,452457. Fishbein. H.D.. Lewis, S . , & Keiffer. K. (1972). Children's understanding of spatial relations: Coordination of perspectives. Developmental Psychology. 7 , 2 1-23. Folk, M.D., & Luce, R.D. (1987).Effects of stimulus complexity on mental rotation rate of polygons. Journal OJ Experimental Psychology: Human Perception and Perfmance, 13. 395-404. Gaylord. S.A., & Marsh, G.R. (1975).Age differences in the speed of a spatial cognitive process. Journal of Gerontology, 30,674678. Gollin, E.S. (1965). A developmental approach to learning and cognition. In L.P. Lipsitt & C.C. Spikes (Eds.), Advances in child development and behavior (Vol. 2;pp. 159-186). New York: Academic Press. Gollin, E.S. (1981).Development and plasticity. In E.S.Gollin (Ed.). Developmental plasticity: Behauioral and blologtcal aspects of Variattons in deuetopment (pp. 231-252).New York: Academic Press. Gollin, E.S. (1984). Developmental malfunctions: Issues and problems. In E.S. Gollin (Ed.), Malfomatfons of development: Bfologfcal and psychological sources and consequences (pp. 123). New York Academic Press. Gollin, E.S. (1985). Ontogeny, phylogeny, and causality. In E.S. Gollin (Ed.). The comparatfw deuelopment of adaptive skills: Evolutionary irnplfcations (pp. 1- 17). Hillsdale, N.J.:Lawrence Erlbaum Associates. Gollin, E.S.,& Garrison, A. (1980). Relationships between perceptual and conceptual mediation systems in young children. Journal of Experfmental Child Psychology. 30,325335.
Visual Cognition
337
Gollin. E.S..& Saravo, A. (1970). A development analysis of learning. In J. HelImuth (Ed.). Cognitive studies (Vol 1; pp. 5790.)New York: Brunner/Mazel. Gollin. E.S.. & Sharps, M.J. (1987).Visual perspective-taking in young children: Reduction of egocentric errors by induction of strategy. Bulletin of the Psychonornic Society. 25,435-437. Gzesh. S.M.. & Surber. C.F. (1985).Visual perspective-taking skills in children. Child Development, 56. 1204-1213. Herman. J.F., & Bruce, P.R. (1983). Adults' mental rotation of spatial information: Effects of age, sex. and cerebral laterality. Experimental Aging Research, 9,83-85. Herman, J.F.. & Coyne, A.C. (1980). Mental manipulation of spatial information in young and elderly adults. Developmental Psychology, 16, 537-538. Hertzog. C.. & Vernon, M.C. (1990). Age differences in mental
rotation task performance: The influence of speed-accuracy trade-offs. Manuscript submitted for publication. Hochberg, J.. & Gellman, L. (1977).The effect of landmark features on mental rotation times. Memory and Cognition, 5,23-26. Huttenlocher. J.. & Presson, C.C. (1979). The coding and transformation of spatial Lnformation. CognftfuePsychology. 11,375-394. Intons-Peterson, M.J. (1983).Imagery paradigms: How vulnerable are they to experimenter's expectations? Journal of Experimental Psychology: Human Perception and Performance, 9.394-412. Jacewicz, M.M.. & Hartley, A.A. (1979).Rotation of mental images by young and old college students: The effects of familiarity.
Journal of Gerontology. 34, 396-403.
Jolicoeur, P.. Regehr. S.. Smith, L., & Smith, G. (1985).Mental rotation of representations of two-dimensional and threedimensional objects. Canadian Journal of Psychology, 39. 100129. Kosslyn. S.M. (1973). Scanning visual images: Some structural implications. Perception and Psychophysics. 14, 90-94. Kosslyn, S.M. (1980).Image and mind. Cambridge, MA: Harvard University Press. Kosslyn, S.M., Ball, T.M., & Reiser, B.J. (1978).Visual images preserve metric spatial information: Evidence from studies of image scanning. Journal of Experimental Psychology: Human Perceptton and Performance, 4 , 47-60. Krauss, I.K.. Quayhagen, M.. & Schaie, K.W. (1980).Spatial rotation in the elderly: Performance factors. Journal of Gerontology, 35, 199-206.
338
Sharps
Kubovy. M., & Podgorny, P. (1981).Does pattern matching require the normalization of size and orientation? Perception and P ~ y ~ h o p h y ~30.24-28. ic~. Lachman, J.L., & Lachman, R (1980).Age and the actualization of world knowledge. In L.W. Poon, J.L. Fozard, L.S. Cermak. D. Arenberg, L.W. Thompson (Ed.), New dfrections fn memory and aging: Proceedings of the George A. Talland Memorial 11). Hfllsdale, N.J.: Lawrence Erlbaum Conference (pp.285-3 Associates. Laurendeau, M., & Pinard, A. (1970). The development of the concept of space in the child. New York: International Universities Press. Light, L.L.. & Zelinski. E.M. (1983). Memory for spatial information in young and old adults. Developmental P~y~hology. 19,901-906. Masangkay, A S . , McClusky, KA., McIntyre, C.S., Simms-Knight, J.. Vaughn, B.S..& Flavell. J.H. (1974).The early development of inferences about the visual percepts of others. Child Development, 45.357-366. McCormack, P.D. (1982).Coding of spatial information by young and elderly adults. Journal of Gerontology, 37. 80-86. Metzler. J., & Shepard, R.N. (1974).Transformational studies of the internal representation of three-dimensional objects. In R.L. Solso (Ed.), Theories in cognitive psychology: The Loyola Symposium (pp. 147-201.)Potomac. MD: Lawrence Erlbaum Associates. Moore, T.E., Richards. B., & Hood, J. (1984).Aging and the coding of spatial information. Journal of Gerontology, 39,2 10-212. Norman, DA., & Bobrow, D.E. (1975).On data-limited and resource limited processes. Cognitive Psychology, 7, 44-64. Olson. D.R (1970).Cognitive development: The chffdsacquisition of diagonality. New York Academic Press. Paivio, A. (1971).Imagery and verbal processes. Hillsdale. N.J.: Lawrence Erlbaum Associates. Paivio. A. (1986).Mental representations: A dual-coding approach. New York Oxford University Press. Palmer, S.E. (1977). Hierarchical structure in perceptual representation. Cognitive Psychology, 9,441-474. Park, D.C., Cherry, K.E..Smith, A.D., & Lafronza, V.M.(in press). The effect of distinctive context on memory for objects and their locations in young and older adults. Psychology and
Aging.
Visual Cognition
339
Park, D.C., PugUsi, J.T.. & Lutz, R (1982).Spatial memory in older adults: Effects of intentionality. Journal of Gerontology. 3 7 . 330-335. Park, D.C., Puglisi. J.T., & Sovacool. M. (1983). Memory for pictures, words, and spatial location in older adults: Evidence for pictorial superiority. Journal of Gerontology, 38. 582-588. Perlmutter, M., Metzger, R.. Nezworski, T., & Miller, K. (1981). Spatial and temporal memory in 20 and 60 year olds. Journal of Gerontology. 36, 59-65. Petrusic. W.M.. Varro, L.. & Jamieson. D.G. (1978).Mental rotation validation of two spatial ability tests. Psychological Research 40. 139-148. Pezdek, K. (1983).Memory for items and their spatial locations by young and elderly adults. Developmental Psychology. 19.895900. Piaget. J.. & Inhelder, B. (1967). The child's conception of space. New York W.W. Norton. Plude. D.J. Milberg, W.P.. & Cerella, J. (1986).Age differences in depicting and perceiving tridimensionality in simple line drawings. Eqerimental Aging Research. 12,221-225. Presson. C.C. (1980). Spatial egocentrism and the effect of an alternate frame of reference. Journal of Experimental Child P ~ y ~ h o l o g29.39 y. 1-402. Puglisi, J.T., & Morrell. RW. (1986).Age-related slowing in mental rotation of three-dimensional objects. Experimental Aging Research, 12.217-220. Pylyshyn. Z.W. (1981).The imagery debate: Analogue media versus tacit knowledge. Psychological Review, 88, 16-45. Reed, S.K.,Hock, H.. & Lockhead, G.R. (1983).Tacit knowledge and the effect of pattern configuration on mental scanning. Memory and Cognition, 1 1, 137-143. Rudy, J.W., Vogt, M.B..& Hyson. R.L. (1984). A developmental analysis of the rat's learned reactions to gustatory and auditory stimulation . In R. Kail & N.E. Spear (Eds.). Comparative perspectfues on the development of memory (pp. 181-208). Hillsdale, N.J.: Lawrence Erlbaum Associates. Salthouse, T.A. (1982). Adult Cognition. New York: SpringerVerlag. Salthouse, T.A. (1987).Adult age differences in integrative spatial ability. Psychology and Aging. 2. 254-260. Schachter, D..& Collin, E.S. (1979).Spatial perspective taking in young children. Journal of Experimental Child Psychology, 27. 467-478.
340
ShUlpS
Shank, C.U.,& Watson, J.S. (1970). Assessment of spatial egocentrism through expectancy violation. P s y chonornic Science, 18.93-94. Sharps, M.J.(1990).Spatial memory in young and elderly adults: The category structure of stimulus sets. Manuscript submitted for publication. Sharps, M.J.,& Gollin. E.S. (1987a).Memory for object locations in young and elderly adults. Journal of Gerontologg. 42, 336341. Sharps, M.J., 8t Gollin. E.S. (1987b).Speed and accuracy of mental image rotation in young and elderly adults. Journal of Gerontology, 42,342-344. Shepard, R.N.,8t Metzler, J. (1971). Mental rotation of three dimensional objects. Science, 171,701-703. Shepard, S.,& Metzler, D. (1988). Mental rotation: Effects of dimensionality of objects and type of task. Journal of Experimental Psychology: Human Perception and Performance, 14, 3-1 1. Siegel. A.W., Herman. J.F.. Allen, G.L.. & Kirasic. K.C. (1979).The development of cognitive maps of large- and small-scale space. Child Development, 50, 582-585. Spoehr, K.T., & Lehmkuhle, S.W. (1982). Visual tnformatfon processing. San Francisco: W.H. Freeman. Steiger, J.H., & Yuflle, J.C. (1983).Long-term memory and mental rotation. Canadian Journal of Psychology, 37,367-389. Tapley. S.M.,& Bryden, M.P. (1977). An investigation of sex differences in spatial ability: Mental rotation of threedimensional objects. Canadian Journal of Psychology, 31,122130. Thurstone, T.G. (1962). SRA primary mental abilittes. Chicago: Science Research Association. Vandenberg, S.G.,& Kuse. A.R. (1978).Mental rotations, a group test of three-dimensional spatial visualization. Perceptual and Motor Skflls, 47,599-604. Vanderplas, J.M.. & Gamin. E.A. (1959).The association value of random shapes. Journal of Experimental Psychology, 57, 147154. Waddell, K.J., 81Rogoff. B. ( 1981). Effect of contextual organization on spatial memory of middle-aged and older women. Dewlopmental Psychology, 17,878-885. Walker, L.D.,& Gollin, E.S. (1977). Perspective role-taking in young children. Journal of Experfmental-Child Psychology,-24, 343-357.
Vtsud Cognition
341
Watson, J.B. (1913). Psychology as the behaviorist views it. P S Y C ~Review, ~ C ~20, 158- 177. Watson. J.B. (1928). The ways ofbehauiorlsrn. New York Harper and Brothers. Weatherford. D.L. ( 1985). Representing and manipulating spatial information from different environments: Models to neighborhoods. In R. Cohen (Ed.). The development of spatlal cognitton (pp. 41-70). Hillsdale, N.J.: Lawrence Erlbaum Associates. Weber. R., Brown, L., & Weldon, J. (1978). Cognitive maps of environmental knowledge and preference in nursing home patients. Experimental Aging Research 4. 157- 174. White, M . J . (1980). Naming and categorization of tilted alphanumeric characters do not require mental rotation. Bulletin of the Psychonornfc Society, 15, 153-156. Wilson, J.R., DeFries, J.C., McClearn, G.E., Vandenburg. S.G. .Johnson, R.C., & Rashad, M.N. ( 1975). Cognitive abilities: Use of family data as a control to assess sex and age dflerences in two ethnic groups. International Journal of Aging and Human Development, 6.26 1-276. Wundt, W. (1894). Lectures on human and animal psychology. New York: Macmillan. Yuffle, J.C.. & Steiger, J.H. (1982). Nonholistic processing in mental rotation: Some suggestive evidence. Perception and P~y~hOphy~ 31.20 i c ~ ,1-209.
This Page Intentionally Left Blank
Aging and Cognition. Knowledge Organizaifon and UtUizatbn Thomas M. Hess (Editor) 0 Elsevier Science Ahlishers B.V. (North-Holland), 1990
CHAPTJ3R EIGHT
EXPERTISE AND AGING: LIFE IN THE LAB Neil Charness and Elizabeth A. Bosman
University of Waterloo
SUMMARY
Why do old adults perform s o poorly with typical laboratory tasks, yet do so well in professional and everyday performance? Further, why is there such a discrepancy between the ages at which peak performances are obtained in sports versus intellectual spheres? We explore the age/skill tradeoff, and suggest how "software", knowledge acquisition, can compensate for age-related declines in "hardware", cognitive architecture. We review research on several intellectual domains, emphasizing work conducted in chess and bridge, and contrast those studies with ones dealing with more physically demanding domains, typing and athletic performance. We outline the boundary conditions for compensation within each domain, and discuss the implications for theories of compensation. THE SKIU/AGE TRADEOFF
Murrell, an industrial gerontologist, once observed: "Anyone reading the results of the laboratory experiments could be forgiven for imagining that any person who achieves the age of fifty will have become a slow, forgetful. half-blind, half-deaf, palsied character of little use in industry. In fact, many older men and women hold down jobs with
344
Chamess and Bosman
complete satisfaction to their employer." (1965. P.449). Similarly. Talland remarked: "I am still puzzled by the contrast of the athlete who,
at thirty, is too old for the championship and the maestro, who, at eighty, can treat us to a memorable performance on the concert stage... Are our aged masters freaks of nature, paragons of self-discipline, or do they demonstrate the inadequacy of our present notions about the effects of age on human capacities?" (1965,P. 558). Are our in uffro laboratory experiments adequate indicators of
fn vim cognitive behaviofl Why is there such a great discrepancy between the performance of older people on laboratory tests of cognitive ability and their performance in their professional and everyday activities? There are four classes of easy, but inadequate answers. The first is that laboratory tests are uninteresting to older adults who don't particularly exert themselves. A variant is that older adults don't share the same cultural norms to do well on tests. The hypothesis is that age differences are due to problems with motivation. The second is that cross-sectional studies of age effects are confounded by cohort differences. Earlier born cohorts were disadvantaged at certain types of test-like tasks, (Schaie, 1983; Labouvie-Vief. 1985). The hypothesis is that age differences are due to cohort confounds. A third is that laboratory tasks don't successfully tap everyday types of behavior. The hypothesis is that tasks lack ecological validity. A fourth is that in cross-sectional research, samples of young and old are not representative and are positively biased for the young since most are college students. The hypothesis is that age differences are due to unrepresentative samples. The reasons to doubt the utility of these explanations are many. Tests of the motivation hypothesis have usually failed. When you
Expertise and Aging
345
reward young and old, the old do not increase their performance disproportionately compared to the young. Usually, young and old increase performance equally. (See. for example, Grant, Storandt & Botwinick, 1978 with the very age-sensitive digit-symbol substitution task). The argument also tends to lose force when the experimental tasks are ecologically sound ones being performed by professionals and serious devotees [e.g.. typing: Salthouse. 1984; bridge: Charness. 1979. 1983. 1987; chess: Charness. 1981a. 1981b. 1981~). A reason to doubt the cohort confound explanation is that
longitudinal research has shown similar trends to cross-sectional research when the time interval for the longitudinal comparison increases towards that of the typical cross-sectional comparison (Botwinick & Siegler. 1980). For instance, Schaie (1983, 1989) provides longitudinal data that show that decrements on PMA scales are observed at earlier chronological ages for greater time interval comparisons (e.g.. 28 year versus 14 year measurements). Further, time lag comparisons of young undergraduate students on nonsense syllable learning have shown few changes (Kausler, 1982). Older cohorts are sometimes advantaged compared to younger ones: see Schaie (1983) for the cases of word fluency and number skills. Further, Lehman (1953) has shown that the age of peak performance in professional achievement has not changed much over the past few centuries; if anything it has occurred earlier for more recent cohorts. Stability for the age of peak athletic performance is also the rule (Schulz & Curnow. 1988). As Kausler (1982) noted, the controversy over generalizing age differences from lab to life (ecological validity) may be based on false premises. If the same processes are utilized by subjects in "ecological" and "unecological" investigations of performance, valid inferences about processes are possible. Further, as Salthouse (1985) indicated, if you want to examine "pure" age effects, then looking at situations where knowledge could compensate for age changes is probably not a good idea: 'The basic premise of the selective expertise concept is that the relation between performance and chronological age will vary with the amount of
346
Charness and Bosman
experience people at each age have had exercising the relevant ability. An important implication of this proposal is that meaningful reflections of 'basic' or 'pure' aging processes must either be based on novice levels of performance, or great care must be taken to ensure that experience is equated across age groups." (P.114). Those working on psychometric intelligence measures have made much the same point. They argue that intelligence subtests that require unpracticed mental operations are probably the more valid ones to use when it is assumed that intelligence reflects native ability. (A nasty problem is that it may be the case that smart people are those who have had a great deal of useful experience. Few tasks, for them, will be truly unpracticed.) It is true that samples lack representativeness, suffering from the typical biases that volunteer samples are prey to (e.g.. Camp, West & Poon. 1989;Todd, Davis, & Cafferty. 1983). Nonetheless, in one cross-sectional study where a probability sample was obtained (Herzog & Rodgers, 1989). strong linear age decline was still observed in an ecologically sensible incidental memory measure. Further, the nature of the bias probably results in underestimates of age effects. Older adults who volunteer for experiments are generally much better educated and in better health than more representative members of their cohort. What then is the explanation for the consistent pattern of decline in laboratory cognitive performance with increased age? If we rule out experience-based explanations (disadvantaged cohorts), then maturational changes are plausible candidates. Specifically, our working hypothesis is that the "hardware" supporting cognitive operations is undergoing change, usually negative change. (Patrick Rabbitt phrased it rather colorfully at a Talland Conference one time as 'The brain rots".) Although it has been fashionable to argue against the "constant decrement'' view of aging, we want to claim that it is still the most parsimonious model for consideration. Two complementary investigative models for cognitive aging are those of information processing psychology and neuropsychological case investigation. The standard information
Expertbe and Aging
347
processing perspective on behavior is that humans act with limited rationality because of limitations in information processing capabilities (Simon, 1969). If you want to understand human capabilities. you need to find situations where behaviors depart from perfect rationality (ideal performance). The neuropsychologist examining a patient tries to find specific problems with behavior. Amnesics look perfectly normal in standard conversations, until you probe recent memory. Dyslexics are fine, until confronted with printed language. Older people look just like younger people, until you push them to the limits with complex tasks, particularly speeded ones. Baltes and colleagues (e.g.. Kliegl, Smith, & Baltes, 1986,1989)have termed the exercise "Testing the Limits". when looking at skill acquisition for a memory task by younger and older adults. Our hypothesis is that older adults can perform more than adequately in most real world tasks because they have acquired specific knowledge (declarative, procedural: Anderson. 1983)that allows them to compensate for declines in raw hardware capability. When tasks push people to their limits, however, age decline may be demonstrated even in expert performance. The famous debate between Lehman and Dennis about age of peak professional achievement hangs on this point (see Simonton, 1988). Although Dennis succeeded in finding cases where there was little decline in productivity over the lifespan by choosing modest achievements, Lehman could point to top achievements and find consistent evidence of a peak in the thirties. A recent study of the publishing behavior of North American psychologists (Horner, Rushton, & Vernon, 1986)also shows that for those whose output is highest initially (in the 20s age range) there is linear decline across the career. Moderate publishers give the usual inverted U-shaped function with peak performance in the 30s and 40s. rather than in the 20s. A similar linear decline trend is observed for the quality of teaching by psychology professors (Homer. Murray. and Rushton, 1989). Studies of skilled performers often show that older adults perform equally to young adults. How they maintain performance in the face of documented decline in various cognitive processes thought to support behavior is an intriguing question. The study of
348
Chamess and Bosman
compensatory mechanisms is just beginning (see Salthouse. 1989: 1990). When confronted with the wearying array of findings that show that older groups do less well than younger ones on nearly every task invented by experimental psychologists, the tendency is to assume that all older adults perform less effectively than all younger adults. Means tend to hide the extensive individual dmerences in performance in both groups. There was a striking example in our work on mental calculation (Charness & Campbell, 1988). A further look at individual data shows that a 71 year old was the second fastest individual at squaring two-digit numbers mentally in the flrst session. Not every older adult shows decline, as Schaie and colleagues have been at pains to point out with longitudinal data on psychometric abilities. A Theory of Tradeoff
How can knowledge compensate for declines in hardware? We want to draw on a n analogy with human and computer chess playing. Computer chess programs play excellent chess by doing incredibly extensive searches of parts of the game tree. (The game tree is the set of all possible legal moves from the initial position to the end of the game.) Master level programs such as "Deep Thought" can search millions of paths in seconds. In contrast, human chess players rely on extensive knowledge to control the size of the search tree, and as a result, may only look at 100 or so paths in ten minutes of search. Such knowledge ensures that the best human players can still beat the best programs (though human superiority may be challenged in the next few decades). The ability to recognize chess patterns and understand their meaning enables humans to search selectively. In its strong form. the knowledge-search tradeoff proposes that equivalent performance can be reached by fundamentally different programs. (See Newell, 1989,for a good discussion of preparation versus deliberation a s strategies for problem-solving by information processing systems). Camp (1989) gives the intriguing example of how someone can answer the information question on the WAIS about the distance between New York and Paris in two
J3xpertlse and Aging
349
ways. They could know the answer (memory retrieval) or could compute it from other facts (remembering how long Lindbergh's flight between the cities took, estimating the speed of the plane, and calculating the answer), as did one older adult he tested. Let's take the slowing hypothesis (e.g.. a s explicated by Salthouse. 1985, 1988) as a good first approximation theory of how aging affects information processing. We can use the slowing parameter of 1.6 - 2.0 (Cerella. 1985) to estimate an older person's reaction time from that of a younger person. (We can also estimate processing parameters like those in the Card, Moran & Newel1 (1983) cognitive model used to predict task completion times for routine cognitive tasks such as word processing. See Charness and Bosman. 1990). The strong prediction from a n age-slowing model is that if young and old adults follow the same internal program, the young will complete the task sooner than the old. But, if the two infomation processors are using different internal programs, all bets are off for who finishes first. For tasks like simple reaction time, or choice reaction time, we can be moderately confident that the programs being followed are quite similar for young and old. We are only moderately confident, since choice of a speed/accuracy criterion is a free parameter.) Thus, those types of tasks should yield reliable age dflerences. When it comes to more complex tasks (choosing a move in chess, typing an unfamiliar passage of text), we can no longer ensure that young and old use the same program. It is always possible for an older adult to use a more efficient program, run at a slower rate, to outperform a young adult who runs a less efficient variant at a faster rate. Knowledge is being equated here with the ability to construct, compile, or operate a very efficient program. A look at how experts manage to solve problems so successfully in their domains of expertise also shows that they rely on knowledge to minimize and sometimes even bypass search through the problem space (Ericsson & Smith, in press; Chi. Glaser & Fan, 1988). Some medical diagnosis and chemical analysis programs have also evolved in this direction. The triumph of knowledge over reasoning (search) for successful artificial intelligence has been
Chamess and Bosrnan
350
aptly phrased as the "Knowledge Principle" (KP) by Feigenbaum (1989):
"The Knowledge Principle states that a system exhibits intelligent understanding and action at a high level of competence primarily because of the specific knowledge that it can bring to bear: the concepts, representations, facts, heuristics, models, and methods of its domain of endeavor. A corollary of the KP is that the reasoning processes of intelligent systems are generally weak and are not the primary source of power" (P.179). Rephrasing the problem for the field of aging, we can begin to ask a different question than that motivating most of the work in the field so f a r (are there changes with age). Given that older adults can maintain very high levels of performance in the face of declines in basic information processing rates (the slowlng hypothesis), how do we adequately describe the knowledge that enables them to do it? This is of course the compensation question (see Salthouse. 1989; 1990). Human expertise spans a n enormous set of tasks. A typical dictionary of occupational classifications lists thousands of job types, and a quick perusal of the Guinness book of records indicates there are many other non-paid avocations that people are willing to devote enormous effort to. We'll review a sample of studies from two primarily intellectual (chess, bridge) and two primarily perceptualmotor domains (typing, sports). Much of the work concerning compensation has been conducted by equating young and old on some molar task (e.g., typing speed, chess ratingl and seeing how they vary on the component processes thought to underlie the task. When you select young and old to be equal, however, you do not know whether in the population as a whole there is no relation between age and skill. You may have selected only the fittest old and the less fit young. Fortunately, researchers have usually shown that expert older adults look fairly typical on domain-unrelated measures, namely, they do worse than younger experts on psychometric tests such as digit-symbol
Expertbe and Aging
351
substitution, and worse on typical experimental tasks such a s choice reaction time and memory for list items. SEMANTICALLY-RICH DOMAINS Simon (1979)introduced the term "semantically-rich" to characterize domains where a large knowledge base was instrumental for skilled performance. Semantic memory is conceptualized a s a long-term memory store containing symbol structures: networks of interconnected idea units. (See Chang. 1986.for a review of semantic memory theories.) The key to skill, in this view, is the possession of thousands of "productions". rules for interacting with domain situations. Such "if-then" rules (e.g.. if you have a headache, then take an aspirin) enable the expert to deal smoothly with new situations that can be mapped onto existing structures. For situations where thinking is required, even for an expert, productions guide search along the most effective paths. Experts often search forward from the initial state to the goal state (in physics, Simon & Simon, 1978;in medical diagnosis, Pate1 & Groen, 1986). Such knowledge-informed search processes contrast sharply with the backwards search that results from the means-ends style reasoning that characterizes novice behavior in these situations. The expert knows that in problems of a given type, if you solve for particular unknowns in a given sequence, the answer will soon be in sight. Not all expertise involves pulling knowledge out of a hat for each occasion. Allard and Starkes (in preparation) argue that often in "open" sports skills (e.g., baseball, tennis), the expert is someone who can carry out goals in a flexible fashion. In contrast, in "closed" sport skills, where the goal and its manner of implementation in the environment is relatively constant (figure skating. archery), the winner is the athlete who can carry out wellpracticed motor programs with minimal variation. Characterizing the knowledge store of the expert is still an ongoing concern (see Charness, in preparation, for an example of categorizing chess knowledge). Estimates of extent of knowledge, such as 50,OOO chess patterns (Simon & Gilmartin, 1973)have been widely cited. Given that humans take a relatively long time to add
352
Chamess and Bosman
new information to memory (5 - 15 seconds for rote memorization: Simon, 1974). it is clear that building such a knowledge base takes thousands of hours. When considering the factor of age, it is highly likely that age will be positively associated with experience and practice. That is, older adults will have had more opportunity to build up knowledge than younger adults. An example is vocabulary, where psychometric investigations typically reveal a positive correlation between age and word definition knowledge (e.g., Birren & Morrison, 1961). The tension between knowledge possession and its successful activation and management underlies the compensation argument. Chess Expert practitioners have done little to discourage the view that chess is the most intellectual of all games. Masters have long amazed amateurs with their ability to foresee distant checkmates. Work by psychologists in this century (see Charness, in preparation: Holding, 1985) has shown, however, that skilled perception is an important component of skilled play. It was also evident to those building chess programs that it was impossible to examine the consequences of all alternatives when planning which move to make. Rather, a limited search would have to take place. Humans are far more selective than machines, In a 10 minute problem-solving episode, where players are asked to find the best move, about 100 chess positions will be examined (Charness. 198lb). Current top-flight chess programs examine between 100,OOO and 1,000,000 positions per minute (Berliner & Ebeling. 1989). What enables the best humans to beat the fastest computers most of the time? The actual problem space of possible chess games is estimated at about positions. The fastest programs, now, and in the future, will never be able to examine all alternatives. The key is to look at the best alternatives. Humans achieve this by knowing which moves to examine, and not even generating most of the irrelevant moves. Computer programs use clever pruning and move ordering techniques to grow the smallest possible search tree,
Jbpertlse and Aging
353
while still looking at all required alternatives to a (timedetermined) fixed depth.
Determinants of Skill The major marker for knowledge in chess has been de Groot's (19651 brief recall task. If you show a chess player an unfamiliar middle game position for a few seconds, their ability to reproduce the position depends strongly on their skill level. Grandmasters can place nearly all pieces (about 24)perfectly. Beginners can place 4 or 5 pieces correctly. Miller's (1956)notion of a chunk can explain these differences. With structured positions. the Master can recognize clusters of pieces (7 f 2 chunks). When randomly arranged pieces are shown briefly, master and novice alike place only 4 or 5 pieces correctly (Chase & Simon, 1973a. 1973b). With random displays. there is little opportunity. even for a master. to recognize chunks larger than a single piece. These results have been taken to show that masters have a large vocabulary of familiar chess patterns (chunks) in semantic memory. Chase and Simon hypothesized that such patterns trigger move generation processes, enabling the master to look at the right alternatives. while the novice goes merrily down the wrong path, never even generating the correct move for examination. Recent experimental research suggests that the Chase and Simon theory is probably insufficiently specified, and others have emphasized the importance of delineating the mechanisms controlling search processes (Charness, in preparation; Holding, 1985). Nonetheless, pattern recognition is seen a s the key for guiding search, both for generating base moves at the root position for analysis, and for providing evaluative information at the leaf nodes of the search tree. Tournament chess games are always conducted with time limits. In top-level tournaments, players have, on average, 3 minutes per move (40moves must be completed in the first two hours). Research has shown that players typically generate a new position for analysis every 5 - 15 seconds (they seem to spend more time evaluating than generating moves). Thus, they will probably look only at about 20 moves, including analysis moves, in an
354
Charness and Bosman
average position over that time interval (Charness. 1981b). In short, skilled players must very quickly recognfie the right moves to look at. Chess, at top levels, is a game of time stress, precisely, the type of working environment that older adults are advised to avoid (Welford, 1985). Nonetheless, chess is also a game requiring extensive knowledge, something that older chess players are likely to have in greater abundance than younger ones. Elo (1965, 1978) showed that the careers of chess grandmasters parallel those of other peak performers (see Lehman, 1953; Simonton. 1988). Best tournament performances are achieved in the 309 (meanage of peak is 36)with declines thereafter. The nature of the decline is quite modest, about a half standard deviation from the peak in the 30s to the low point in the 60% the age at which most players give up active tournament participation. The relatively modest decline in chess performance for Grandmasters (0.5 SO1 contrasts sharply with peak performance trends mentioned earlier. Could possession of greater knowledge be compensating for older players' declining ability t o search quickly? Charness (1981a. 1981b. 1981c) explored how age and skill jointly determined the performance of chess players on a set of chess-related tasks. The strategy adopted was to equate players varying in age on skill level, using chess rating as the criterion measure for skill, and examine how the processes thought to support chess playing varied. The main sample consisted of 34 male chess players varying in age between 16 and 64. and in skill between 1283 and 2004 rating points. Players were selected such that age and skill were virtually uncorrelated (r = ,085). A substantial number of variables were examined, but this review wffl concentrate on two subsets: search characteristics and memory indexes. Chess search varied systematically with age and with skill (see Charness. 1981b). The more skilled the player, the farther ahead they projected the consequences of a base move (the first move for their side). Also, skilled players tended to examine more moves in total (both base moves and analysis moves). Older players tended to examine fewer moves than younger players of equivalent skill level, though their depth of search was equivalent. That is, though older players were capable of looking as far ahead
Expertise and Aging
355
(pushed search as deeply into the tree of move possibilities), they tended to end search sooner, having examined fewer alternatives. Nonetheless, the bottom line for chess is the strength of the selected move. On that measure only skill was a significant factor. Old and young alike chose moves appropriate to their skill level. In brief, age seemed to have little impact on the quality of the move selected, though it did exert an influence on the processes responsible for move selection. In fact, the curious picture arises that older players were more efficient than younger players, choosing an equally good move with less extensive search. Older players took significantly less time, on average, to choose their move. Depth of search, a measure showing skill but not age ef'fects, is a type of working memory measure. It defines the point at which a player can no longer retain accurate information about the projected changes to a perceived chess position (the physically present one) when moves are envisioned. Because of limitations in working memory, players must return to the base position and rework their analysis, a process termed "progressive deepening" by de Groot (1978). Depth of search is controlled by move generation processes (what are the possible moves), evaluation processes (knowing the value of a given position enables players to terminate a search episode), and memory processes (keeping track of moves generated and evaluated). Apparently, such highly-practiced processes execute as smoothly in the old as in the young. (Some idea of how practiced search behaviour can be is given in Charness, in preparation. The most active Canadian chess players, usually young ones, may exhibit an upper limit of a million positions examined during search over a year of tournament play.) A direct test of evaluation accuracy showed skill but not age effects. Players were asked to evaluate endgame positions (positions with few remaining pieces) as win. draw, or loss as quickly a s possible. Even with this speeded task. no age effects were evident. More direct tests of memory efficiency yielded a different result, however. An incidental memory test for the four chess problems that players were given, following interpolated tasks, showed that older players did more poorly than equally skilled younger ones. More skilled players recalled more pieces correctly than their less-
356
C h r n e s s and Bosman
skilled counterparts. By using the pause structure in recall, chunk boundaries were defined. As had been shown for intentional recall of chess information by Chase and Simon (1973a). skill was associated with recalling a position as a small set of large chunks. Here, however, there was an interaction between age and skill for incidental memory chunk size. The recall advantage accruing from greater skill declined with increasing age. By about 60 years of age, there was no longer any advantage to being more skilled at chess for recalling positions as small numbers of large chunks. A direct test of intentional memory for briefly presented chess positions (Charness. 198lc) showed similar age-related deficits. The finding of age-related memory declines in chess position recall has been replicated by Pfau and Murphy (19881. The declines observed for recall of chess information can be modeled successfully by assuming that older players are slower in the pattern recognition and chunking process. Using a stochastic version of W P , a chess recall simulation by Simon and Gilmartin (1973). Chamess (1988) generated reasonable fits to the intentional recall data reported earlier (Charness, 1981~). How older chess players compensate for declining memory ability (as measured with intentional and incidental chess memory tasks) is still unclear. It is possible that the time demands are not extreme enough under normal playing conditions to influence success in problem solving very strongly. Note that for humans, who search very modestly, what gets searched, rather than how deeply search proceeds. is the critical variable. Knowledge, rather than speed (except in time pressure situations), is critical to performance. It must be admitted, though, that speed does set some constraints, particularly in terms of the rate of knowledge activation. A particularly telling experimental point is that studies of semantic memory priming (speed of activation) have shown little change with age (e.g., Balota & Duchek, 1988).though mean reaction time differences still favor the young. Still, we have to explain why Elo found that the peak tournament performance for chess players comes in the 30s rather than in the 60s at a time when accumulated chess knowledge is probably at its peak.
Expertbe and Aging
357
Peak Chess Performance and Age As mentioned earlier, Elo (1965. 1978) examined the relationship between age and peak chess performance, showing a classical inverted-U function. A plot of that function appears in Figure 1 for relatfue performance. That is, scores are based on deviations from a player's peak performance, with age 21 set to zero. The number of players on which each point is based is: 10,51. 51, 48. 51, 50, 48, 41, 27, 14,and 9 for ages 18 through 65, respectively. Given the self-selection factors that operate, the amount of decline shown may be underestimated or overestimated because of selective attrition. Since players leave the pool for dmerent reasons (involuntary factors such as illness and death, as well a s voluntary ones). a rough test for selective attrition is to see whether the age at which players stop playing in tournaments is related to their peak absolute performance (do better players play longer) and the age at which they reach that peak (do those who peak at a young age leave earlier). Using Elo's (1965)data, players whose dropout age was artificially truncated were first eliminated (since the survey was published in 1965.recent players could not achieve true dropout ages). To estimate age at withdrawal we chose the first half decade with no rating. The regression of age at withdrawal on peak rating and year of peak rating (the interaction term made no difference) was not quite significant, p < .06,with peak age being the only significant factor. A simple regression of peak age on age of withdrawal was significant (n1.35)= 6.4,p < .02) for the 37 players:
Age-at-withdrawal= 36.3+ ,655* Age-at-Peak, SE = 6.64,$ = .16. That is. peaking later in life is associated with later withdrawal, though how skilled you are at that peak is not a factor. Filtering out players who died before age 60 only intensifies the relation between peak age and withdrawal: r (20) = .49. The framework of programs and architecture can be helpful in explaining age of peak performance. A serious player in their 30s has had about 20 years of experience (most start chess before the age
Charness and Bosman
358
0.8 1 0.6 0.4
SDChws 0.2 Rating 0
/
-0.2 -0.4
15
20
25
30 35
40 45
50
55
60
65
70
Age in Years
FYgure I. Longitudinal chess performance for 51 Grandmasters from Elo (1965). Rating is scaled to be 0 at age 21. The bars represent f 1 SE.
of 10). His or her cognitive architecture is still near its peak. Memory efficiency declines from the 20s and 30s in cross-sectional studies (e.g., Hertog & Rodgers, 1989). Psychometric abilities often do not show much decline until the 60s or 70s in longitudinal studies (e.g.. Schaie. 1983. 1989). Since chess does have time constraints, a younger player may occasionally generate a slightly better move than a more knowledgeable older player who can't activate the appropriate knowledge quickly enough. One slightly better move can be the difference between winning and drawing (or wen losingl a game of chess. That is. architecture may carry more weight than knowledge when speed is a significant factor in play. If this explanation is correct, you ought to predict that various forms of chess should show earlier peaks and declines. A reasonable ordering would be: speed chess (about 10 sec/move), tournament chess (about 3 min/move), then correspondence (postal) chess (3 days/move). with the latter showing later age decline. Unfortunately, there have only recently (1988)been systematically organized speed chess tournaments and speed chess ratings, so it is only possible to look at the latter two cases. Prior to the acceptance of the Federation Internationale Des Echecs, FIDE. in 1946 as the legitimate governing body for chess by most nations, world champions have been determined in
Expertise and Aging
358
inconsistent ways. If we restrict examination to the championships conducted by FIDE we find the expected pattern. The mean age when someone first becomes world champion in tournament chess is 29.9 years, (SD = 6.79). for the 8 world champions starting with Botvinnik (in 1951) and ending with Kasparov (in 1984). The mean age when a player first becomes world champion in correspondence chess is 45.7 years (SD= 5.80). for the flrst 11 world champions starting with Purdy (in 1953) and ending with Baumbach (in 1988). This 16 year difference is significant statistically: t(17) = 5.48. p c .001. There are many problems with this comparison, one being the very long time period for a championship match in correspondence chess (at least 3 years, versus several months for over the board play). Even if it is assumed that the correspondence chess player has won the championship by the mid-point of the three year period, there is still a significant difference between the age of attaining the top title for these two variants of chess. There may also be self-selection factors delaying when players take up correspondence chess. Usually the champions are players who started with tournament chess, never achieved world class results, and then took up correspondence play. Nonetheless, it is clear that being young is much less of a factor in becoming world champion in correspondence chess than in regular tournament chess. One significant difference between the two forms of chess, aside from time for analysis, is style of analysis. You must do all your analysis without physically moving pieces in over-the-board tournament play. You are permitted to move pieces during analysis of positions in correspondence play. Thus, working memory (keeping track of outcomes, visualizing future positions) is less stressed in correspondence play. During postal play the player can rely more on external memory (written notes). There is corroborative experimental evidence that time to choose a move is an important factor in quality of play. In a recent study, Calderwood, Klein, and Crandall (1988) have shown declines in move quality with speed pressure, with less-skilled players more hurt than highly skilled ones. A n extension to look at the age factor is needed. Such a possibility exists in tournament chess in future years, if so-called "active chess" (averaging 30 seconds/move)
360
Charness and Bosman
becomes rated independently of standard tournament chess. Comparisons between ratings under the two time limits ought to reveal whether age is a significant factor. Chess playing offers a rather unique environment in which to pursue questions of skill-age tradeoff. Older players, despite showing the memory declines anticipated for aging individuals, still maintain high levels of skill. The exact mechanisms for compensation have yet to be identified. Future research will need to focus on identifying the knowledge base that players possess. Studies will also need to test the extent to which speed constrains skilled performance. Bridge Whereas it is possible to have a chess program play in a masterful way without knowing much about chess beyond the rules of play. it is impossible to avoid embedding a great deal of knowledge in bridge-playing programs. Bridge, unlike chess, is a game of incomplete information. The problem-solver must deal with uncertainly about the exact layout of the playing cards held by the opponents and partner during the bidding phase, and is uncertain about half of the cards at the beginning of playing out a contract (whether declarer or one of the defence players). Players must deal with probabilities and constantly revise their predictions about card holdings that are unseen (count cards during play and make inferences: see Charness, 1989;Frey, Truscott & Kearse, 1976;Engle & Bukstel. 1978). Skill in playing out a contract involves knowing a great deal about specific card combinations and their proper play. Similarly, in the bidding phase, bidding systems demand that the player memorize specific card-bid combinations. Take the case of opening the bidding with "one no trump" in "Standard American" bidding. It is bid when a player has 16-18(sometimes 15-17)high card points (Ace = 4, King = 3, Queen = 2. Jack = 1) and the suit distribution (number of cards held in spade. heart, diamond, and club suits) is 43-3-3or 4-4-3-2, and occasionally 5-3-3-2where the 5-card suit is diamonds or clubs, There may be dozens of such to-be-learned bids in a typical comprehensive bidding system. Human and computer
Expertbe and Aging
361
alike must master a great deal of detail, since search is a n impractical way to proceed. Age and Skill Effects
Early work in bridge supported the notion that sktll depended on a large knowledge base. Skill in bridge was estimated via the logarithm of accumulated master points, self-rating. and perfoxmance on a bridge quiz in Charness (1979).The problem with master points as a measure of skill is that they can be won, but never lost, so they accumulate over a player's lifetime. If you play often enough, even at a modest level of skill, you can accumulate many points. Both Charness (1979)and Engle and Bukstel (1978) showed that memory for recently played bridge hands was a good predictor of skill level. Similar to the case in chess, Charness (1979)also showed that both incidental and intentional memory for bridge hands declined with age, and increased with skill. However, two criterion measures, bidding accuracy and play planning accuracy, increased with skill and were unaected by age. The Charness (1979)sample was very small (N = 20) and was not selected to ensure that age and skill were uncorrelated. When 45 players were selected such that age and skill were uncorrelated (r = -.06: Charness, 1983. 1987).similar results were obtained. Speed of bridge bidding increased with skill, but decreased with age. Bidding accuracy, though uniformly high, increased with skill, and was unaected by age. As was the case in chess, an interaction between age and skill was observed, here, for bidding speed. The greater a player's age, the less the advantage to being more skilled. By about age 60. there was no longer any bidding speed advantage assoclated with being more skilled. Similarly, when the stage of determining the honour card point count was investigated (Charness, 1987).again accuracy was related solely to skill, whereas latency (decomposed into slope and intercept components) was positively correlated with age (for the intercept) and negatively correlated with skill (for the slope). Unlike the earlier case, however. no interaction was observed between age, skill, and point count estimation speed. The somewhat artificial bidding task did generate patterns of effects for
362
Charness and Bosman
age that mirror other laboratory studies. Skill was associated with manipulating already encoded information efficiently (the slope parameter), whereas age was associated with difficulties in getting information into and out of the system (strong intercept effects). It is worth stressing that in normal bridge play there is no necessity to announce a bid or make a play as quickly as possible. (In fact, too rapid a bid may be unethical in the context of the player's usual bidding latency. Such a rapid response may convey the information that the player had such poor cards that there was nothing to think about.) Bidding accuracy is the "ecologically valid" measure of bidding acumen. In bridge, there is no premium on rapid responses. One of the hallmarks of weaker players is their impulsive play and lack of planning (Charness. 1989). Bridge play possibly represents a case where compensation may not be necessary for older players. Skill in bridge may be a form of power test: Do players have the knowledge to make the correct play or bid, or are they lacking the knowledge? Still, memory constraints can play an important role in a player's ability to keep track of events and plan future bids and plays. (Recall the negative age effects observed for incidental memory for played hands: Charness, 1979). There is little evidence of searching in bridge (Charness, 1989).though the laboratory tasks have not yet done an adequate job of capturing on-line problem solving. It can be expected that errors will occur more frequently in older players if their memory system is less reliable. The short duration tasks investigated so far may not be adequate to detect such age effects. Since there is no "gold standard' for measuring bridge skill, there is less of an opportunity to trace the age of peak bridge playing. Also, since bridge is a partnership game (play is by pairs and teams of four players), it will be difficult to disentangle ageperformance relationships if partners are not the same age. Investigation of age and skill effects in bridge is still in its infancy. MOTOR SKILLS The next two domains we'll review do have the benefit of "gold standards" for measuring skill. Typing speed is often measured in net words per minute, a combined speed/accuracy measure. It is
Expertise and Aging
363
easily enough measured with standardized test passages that job descriptions will require specific speeds (e.g., want ads appear specifying 60 wpm). Similarly, track and field and other athletic events are often administered under highly controlled conditions. Record times are measured quite accurately. Often there are agreed upon criteria for crediting performance even in team events (e.g., who is the winning pitcher when several participate in a baseball game). Motor skills have the further benefit that they tap into human performance at what are thought to be physiological levels. At least in theory, compensation ought to be more difficult the more a task depends upon hardware limitations. Yet, surprisingly, cognitive components appear to be important contributors to skill in these domains (e.g.. Allard 8t Starkes, in preparation).
Several cross-sectional studies have reported no age-related declines in typing speed (Bosman, in preparation: Salthouse. 1984) indicating that older adults are able to perform at very high speeds. For example, a typist whose speed is 80 words per minute (wpm) Is executing a keystroke every 150 ms. A less skilled typist whose speed is 40 wpm is executing a keystroke every 300 ms. That older typists are able to maintain these speeds is quite impressive given the age-related declines that are typically found on reaction time tasks. For example, in a recent study examining age differences in simple reaction time, Wilkinson and Allison (1989) reported a n increase in reaction time from 240 ms in the twenties to 280 ms in the sixties indicating that elderly adults are 1.17 times slower than young adults. For choice reaction time tasks elderly adults are approximately 1.6 to 1.7 times slower than young adults (Cerella, 1985). A critical question becomes why is there no age-related slowing in a complex task like typing when such declines occur on much less complex reaction time tasks. Research examining age differences in typing skill has tried to answer this question in several ways. One approach involves examining age difTerences in the molecular components underlying typing sku1 (Bosman. in preparation; Salthouse, 1984). Molecular
364
Charness and Bosman
components can be regarded a s the constituent parts, or building blocks, of skilled performance. They represent the simpler parts that are combined to produce the more complex skill. Examples of the molecular components underlying typing are finger tapping and choice reaction time. For samples in which there is no correlation between age and skill level, identifying age-sensitive molecular components makes it possible to infer which processes are agesensitive, and suggests where in the process of combining the molecular components older adults may be required to compensate if they are to maintain performance. Another approach involves trying to identify compensatory mechanisms that may be employed by older typists through a fine grained analysis of typing performance. In addition to varying in terms of typing speed, typists of different skill levels also vary on other descriptive measures of typing skill. The objective of this line of research is to identi@ age differences on these measures. Age Differences on Molecular Components
At least three categories of processing underlie typing performance: a) input processes which are concerned with reading the text; b) translation processes which operate to convert the letters in the text into the appropriate motor programs, and c) execution processes which are responsible for executing motor programs. Evidence regarding age-related changes in the molecular components underlying typing skill suggests that when age-related declines are observed they are located in the translation and execution stages.
Input processes Normal reading speed has been used as an index of the efficiency of the input processes involved in typing. Several studies have shown that there is little relationship between age and reading speed for typists: r = -.13,Bosman (in preparation); mean r = .02. Salthouse (1984). This suggests that typists of varying ages are probably equally efficient in perceiving the to-be-typed text. In addition, it is generally thought that the limit on typing speed is not
JZqertLse and Aging
365
associated with input processes. Even for typists whose speed is in excess of 1 0 0 wpm. reading speed is at least twice as fast as typing speed. (The Gufnness Book of World Records, 1989, gives top manual typewriter rates of 170 net wpm for one minute, 176 net wpm for five minutes, and 147 net wpm for one hour of typing. The rather severe penalty of 10 words lost per error was imposed in these contests.) Several studies have demonstrated that there is no relationship between reading speed and typing skill ( r = .lo, Bosman. in preparation: r = .13.Salthouse. 1984). Taken together these findings suggest that input processes are probably not the locus of age-related declines in typing. Tt.anslation processes
The Digit Symbol subtest from the Wechsler Adult Intelligence has been used as an index of the speed of perceptualScale (WNS) motor translation processes involved in typing. Digit Symbol performance is moderately correlated with typing skill ( r = .31, Bosman. in preparation: r = .53,Salthouse. 1984)suggesting that it is not strongly related to the translation processes involved in typing. However, the moderate correlation with typing skill, coupled with the pronounced age-related declines on Digit Symbol performance suggests that translation processes are one locus of age-related declines on the molecular components underlying typing. Choice reaction time (CriT) has also been used as an index of the speed of translation processes involved in typing. The available evidence suggests that the pattern of sktll and age merences in C W among skilled typists depends upon how similar the reaction time task is to typing. In one task that was quite dissimilar to typing, either an 'L'or a n 'R could appear on a screen, and the response was to press the key associated with each letter (Salthouse. 1984).As expected. the task was related to age, (mean r = .54).but showed only a slight relationship with typing skill (mean r = -.27). The latter finding suggests that this particular CRT task may not be representative of the translation processes involved in typing. Consequently, it is not clear if the age-related decline observed in
366
Chamess and Bosman
this task indicates a specific age-related difficulty with the translation processes involved in typing. This possibility received some support from the pattern of age and skill differences in a choice reaction time task more similar to typing (Bosman, in preparation). The task required the typist to sit with their hands in the home row position, and when a letter appeared on the screen they were to type the letter as quickly as they could. The results revealed significant effects of age (r = .51). and skill (r = -.65). Even more interestingly the results revealed a significant interaction between age and skill indicating age-related declines in reaction time for less skilled typists, but no age-related differences in reaction time for highly skilled typists. The implication is that for highly skilled older typists there is compensation at the level of molecular components.
Motor processes Speed of finger tapping has been used as an index of speed of execution processes. Tapping rates for a single finger are modestly related to typing skill (r = -.30.Bosman. in preparation; mean r = .32,Salthouse. 1984). and do not always show strong age-related declines ( r = .23,Bosman, in preparation: mean r = .68. Salthouse. 1984). However, tapping rates for alternate finger tapping be.. tapping a finger from each hand in alternation), are moderately related to skill, (mean r = -.43) and seem to show more consistent age-related declines (mean r = .46, Salthouse, 1984). This finding is consistent with studies examining novice-expert differences which have shown that expert typing speed is in part dependent upon the ability to overlap finger movements across and within hands. In contrast, the slower speed of novices is in part attributable to a n inability to effectively overlap finger movements (for reviews. see Gentner, 1983; Salthouse, 1986 ). Speed of alternate flnger tapping can be regarded as an index of the ability to overlap finger movements, and the finding of an age-related decline on this task suggests that older typists may have greater difficulty than younger typists in effectively overlapping finger movements. The research reviewed above suggests that there are no agerelated declines for input processes, but the results for translation
J3xpertLse and Aging
367
and motor processes are less clear. One difficulty with research examining age and skill dmerences in the molecular components underlying typing skill is that there is no consensus regarding which tasks are most representative of the underlying component processes. For example, Digit Symbol performance, and reaction times on a C W task unrelated to typing skill suggest age-related declines on translation processes. A choice reaction time task more similar to typing reveals an interaction between age and skill suggesting compensation by older skilled typists. Additional research is required to identify the most representative tasks before the pattern of age-related changes on the molecular components of typing can be accurately specifled.
Compensatory Mechanisms An age-related difference in typing skill that seems to function as a compensatory mechanism is the increase in the size of the eyehand span with age (Bosman, in preparation: Salthouse. 1984). A typist always reads a few characters ahead of the actual letter they are typing: and the eye-hand span refers to the number of characters between the character being read, and the character being typed. Although it can be measured by monitoring eyemovements, the eye-hand span is usually measured indirectly by varying text preview, &e., varying the number of characters in the to-be-typed text that can be seen while typing). If the amount a typist can read ahead is less than their usual eye-hand span, typing speed falls off sharply. This implies that the eye-hand span represents the amount of text the typist needs to have available in order to maintain typing speed. In addition. the size of the eyehand span varies with skill, being no more than 2 or 3 keystrokes for novice typists, and as much as 6-8 keystrokes for experts. The implication is that skill acquisition is associated with a n increased reliance upon advance preparation of keystrokes. Considered in conjunction with these findings, the age-related increase in the eye-hand span suggests that dder typists maintain typing performance by beginning preparation of keystrokes sooner than younger typists.
368
Chamss and Bosrnan
At present there are no definitive answers regarding how the eye-hand span enables older typists to compensate for age related declines. In part the answer depends upon the hypothesized function of the eye-hand span during typing. One suggestion is that the eye-hand span reflects the number of characters required to keep the translation and execution processes continuously supplied. That is, the eye-hand span reflects the demand for characters by the translation and execution processes (Cooper. 1983;Salthouse. 1984). Consistent with this interpretation is the finding that within a n individual typist the size of the eye-hand span varies with typing speed. As the individual's rate of typing slows their eye-hand span decreases in size, and vice versa. When typing speed slows there is a reduced demand for characters by the motor and execution processes, and consequently the eye-hand span decreases (Salthouse, 1984). Also consistent with this interpretation is the argument that the limit on typing speed is not associated with input processes, but rather with translation and motor processes. If this interpretation of the eye-hand span is accepted, it suggests that older typists require a larger buffer of characters if translation and motor processes are to be optimally supplied. Although the available evidence is not entirely consistent with this interpretation, it could be argued that older typists compensate for slower translation and motor processes by beginning preparation of keystrokes sooner, and consequently have a larger eye-hand span. Another possibility is that translation and motor processes are more variable in older typists, or more easily disrupted. To maintain continuous output older typists may require a larger buffer of characters so that they can recover quickly from any disruptions that would otherwise slow performance. The argument is that the larger eye-hand span of older typists does not function to compensate for age-related slowing, but makes it easier for older typists to compensate for an increased susceptibility to the breakdown of performance.
Expertbe and Aging
369
Conclusions Among the skills that have been examined by research focusing upon age differences in skilled performance, typing is unique because it is the only skill for which there is direct evidence indicating that older typists may be compensating for age-related declines. The results of the CRT task employed by Bosman (in preparation) suggests that highly skilled older typists are able to maintain performance on some of the molecular components of typing skill. The age-related increase in the eye-hand span suggests that greater advance preparation is another mechanism used by older typists to maintain performance. Future research should seek to define the boundary conditions of these two compensatory mechanisms in order to better understand their nature of operation. It is likely that there are additional compensatory mechanisms that have not yet been identified, and an objective of future research should also be to identify other compensatory mechanisms employed by older typists. Athletics
As the Talland quote made clear, the sports domain is critical for examining questions about skill and age. For no other human endeavor is the emphasis on youth so great. Recently. there has been a push to ensure a place for older athletes in competitive sports with the "Masters" categories. The age at which the masters events begins varies considerably. Ericsson (in press) noted that for swimmers, the youngest "old" swimmers category is 25-29years in Germany. For runners the earliest masters age category is 40. The official categories internationally. however, are 40 for men and 35 for women. Peak Performance and Age In a recent review, Schulz and Curnow (1988)have shown that the mean age of peak performance by athletes in some sports domains has remained remarkably constant over the past century. Over the same time frame, the levels of performance have continued
370
Chamess and Bosman
to improve, from as little as 10% to as much as 90% (the latter in shot put). They looked primarily at Olympic sports (track and field, swimming) as well as tennis, golf, and baseball. They also examined sex dltrerences. showing that the peak age of performance for females tends to occur a year to two years earlier than for males. These sex differences tend to align with physical maturation rates, which also peak one to two years earlier In females. Nonetheless, there were pronounced daerences in the peak ages depending on the sport. In events requiring explosive energy mobilization (short distance running and swimming, jumping, shot put), peak ages are in the 21-24 year range. In golf and baseball, where knowing what to do may be more important than doing something quickly. the peak ages (31 and 28 years, respectively) are later. Long distance events (running, swimming) have later peak ages (25-29 years of age) than short distance events. The one notable exception for the distance rule is female swimmers, where increasing distance of event is correlated with a younger peak age. The authors argue that biologically-related buoyancy factors may play an important role there.
Theories of Age and Peak Performance Theorists about age and peak athletic performance (e.g., Ericsson, in press: Hartley & Hartley, 1984, 1986; Stones & Kmma, 1982. 1984, 1985, 1986) have attempted to link peak performance age with biological and training/practice explanations. Stones and Kozma (1986) evaluated a number of models proposed to account for age trends across various athletic events (particularly track and field events). They concluded that their "power output relative to power available" model (POrPA) provided the best fits to reaction time, strength assessment, and athletic records. Their model assumes that the typical negative age trends for record performance will be steeper when the event requires a high amount of energy relative to the available supply. There are two sources of energy supply: anaerobic and aerobic. Anaerobic power depends on the ATP-PC and lactic acid metabolic systems. Aerobic power depends on oxygen metabolism, Power output demands in
J3xpertlse and Aging
371
many athletic events are related to two features: the strength required and the duration of the event. In short duration events, which rely mostly on anaerobic power sources, age-related decline will increase with strength demanded. Hurdles and high jump, events which have pronounced "explosive" strength demands, will show greater age decline than sprints. where power translation is probably smoother past the explosive start phase. In longer duration events where aerobic power becomes more important, such a s long distance running and swimming, age decline will increase with event length. Aerobic energy sources will be mobilized less effectively with increasing age. Such predictions were well-supported by the data. At first blush, such results seem paradoxical when contrasted with Schulz and Curnow's (1988)finding that peak age is later for long distance events. But peak age depends largely on training. Aerobic fitness and long-term changes in muscle fiber composition may require a longer perlod of intensive training for long distance events than for short ones. (Most sedentary young people can probably muster enough energy to run for 100 m, but many would not be able to maintain running speed for 1000 m.1 Once you are past peak age, the decline slope may still be greater for events demanding high power output relative to the power available. Underlying these predictions are assumptions about agestrength and age-oxygen utilization trends. Power depends in part on muscle mass (as well as neural enervation). Large muscles (legs, trunk) show greater age decline in strength than smaller muscles (arm, hand). Similarly. aerobic capacity shows strong linear age decline in sedentary individuals. Of course, both capacities (strength, aerobic capacity) show remarkable plasticity with respect to training effects. Nonetheless, peak capabilities may be age-sensitive, and skilled athletes may well be operating near asymptotic levels where such peaks matter. Perhaps a good analogy for age and peak performance effects is the case of a rechargeable battery: the lithium cell model. Power output is a function of the previous charging period (training), the freshness of the internal components (age), and the length of time that the battery is active (event duration).
372
Charness and Bosrnan
Trafntng Effects Ericsson (in press) focused most closely on training effects. To get around the problem of relying on Olympic performance records, since the athlete has only one opportunity every four years, Ericsson examined running went performances of the top 10 male athletes for a variety of countries in 1986. He found different age peaks in different countries (e.g., China had significantly younger peak competitors, because, Ericsson surmised. China only recently provided organized training). But, he confirmed the tendency for longer distance track events to be best performed by older runners. Ericsson also noted that there was very high variability among the oldest participants in Masters athlete programs, suggesting that training effects might be a factor. He cited a longitudinal study (13 years) by Letzelter which showed that German masters athletes in swimming showed no statistical declines in performance. Stability was attributed to frequency and intensity of training. To look at individual trends, Ericsson examined in detail three top calibre competitors (Nurmi: running, Hein: hammer throw, Jiixvinen: javelin). He found that personal best performances (27 years, 30 years, 27 years, respectively) occurred later than would be expected from Olympic records, though how much this may be due to secular trends (improvements in equipment and track conditions) is unknown. Further, it is difficult to generalize from only three cases. Strong age trends were quite evident when looking at yearly peak performances. Ericsson also considered the physiological evidence for peak athletic performance, focusing on runners for whom the most data is available. Aerobic power, ability to utilize oxygen. is a very important aspect of such sustained motor activity. Early on it was clear that features such as increased capillary supply to muscles, heart size, and percent of slow-twitch muscle fiber were important daerentiators of elite athletes and inexperienced athletes and nonathletes. Proportion of slow-twitch fiber was a strong predictor of long-distance running success for a mixed distance group of runners, though it was not a significant predictor within a group of elite long distance runners. The classification of fiber types, however, is controversial (Houston, 1978), s o statements about
Eqertise and Agtng
373
percent of slow-twitch fibers are best viewed a s heuristic. Nonetheless, such variables as fiber type, capillary supply, and heart size do tap into the anatomical underpinnings of aerobic power. It was also believed that such factors had high heritability coefficients (90%). (Athletes were born, not made.) Ericsson summarized later work which showed how intensive exercise was sufficient to promote these changes, including the change in composition of muscle fiber (fast-twitch being replaced by slow-twitch fiber). Later estimates of heritability were considerably lower. The training necessary to effect such changes was not shortterm. Evidently, reliable changes in muscle fiber composition are only observed with training periods in excess of 20 weeks. It usually takes about 10 years for runners to reach their peak. Further, only muscle groups that are exercised show changes. Ericsson cites a study by Tesch and Karlsson that shows that middle and long-distance runners have higher percentages of slowtwitch fiber in the legs than control subjects, but not in the back. Kayakers show the reverse pattern. Even in sports, however, cognitive factors can set limits. Work by Backman and Molander on miniature golf (Backman & Molander, 1986a. 1986b; Molander & BAckman. 1989) suggests that ability to handle cognitive arousal during competition diminishes for players in their fifties. When older players move from training to competition, their performance declines, whereas the opposite is true for younger players. Older players fail to show heart rate deceleration during concentration phases of putting, whereas younger players exhibit such deceleration, and it is greater during competition. Ericsson concluded that athletic performance can be maintained with minor decrements into the 60s, if there is continuous intensive training. He also notes, however, that with intensive training comes an increased risk of injury. Age may set the limits for performance, but these limits are rather generous ones for those who can avoid injury. It is in the area of athletic performance that the boundaries between hardware (architecture) and software (task knowledge) begin to blur. Practice and training significantly alter the
374
Charness and Bosman
hardware (muscle fiber, blood supply, heart size), though a s Ericsson noted, training cessation often leads to a reversal in muscle fiber composition. Here, as in other domains, it is only the specific task (and task-related hardware) that shows change with training (contrast kayakers and runners). The Salthouse and the Bosman studies show that typists only appear superior when the task maintains important elements of transcription typing. Although there is a hint of a generalized (genetic?) advantage for skilled performers (e.g., a slight tapping speed correlation with typing skill). it may be the by-product of training rather than the determinant of typing skill. THEORIES OF COMPENSATION If aging is associated with inefficiencies in cognitive and perceptual-motor processing, how do older adults maintain their performance? There are a number of possibilities. First, we may be making an unwarranted assumption that aging leads inevitably to inefficient functioning for every older adult. Longitudinal studies of psychometric performance suggest that little decline is noticeable in most people until the 60s and that for some, no decline is evident until the 70s or later (see Schaie. 1989). Nonetheless, lack of molar decline, as argued earlier, is not strong evidence against hardware deterioration. Given the remarkable improvements that can accrue from knowledge accumulation and task-specific practice, we have to wonder why there is no improvement in performance with increasing age. One possibility is that knowing too much could slow down retrieval. Even with parallel search, the more information that is in the system, the slower will be the retrieval process. Another explanation for the absence of a positive trend with age is forgetting. Since humans forget, it is also possible that a s people age. they forget old information at the same rate that they gain new information, leaving performance unchanged, as Salthouse ( 1989) suggested. A second possibility for maintained performance is that aging impairs some functions but selectively spares others. (Such an assumption underlies much of experimental research on cognitive
Expertise and Aghg
375
aging, since the hope is to show interactions with age across experimental conditions. The goal is to identify age-sensitive and age-insensitive processes.) The tasks that adults perform competently may, by chance, fall into the set of spared functions. If you accept standard information processing assumptions (e.g., Card et al., 1983).basic cycle times (such as those for the perceptual, cognitive, and motor processors) are thought to drive all operations. Aside from practice-mediated improvements, if the cycle time lengthens with age, all mental operations are expected to be slowed. Such slowing can be expected to hurt efficiency of performance quite generally. The age-slowing hypothesis falls out of these assumptions, and there is reasonably compelling support for this hypothesis. Thus, a selective sparing view must posit that, for instance, the motor and cognitive processors are hurt by aging, but not the perceptual one, thus leaving tasks that depend heavily on perceptual processes intact. No simple class of activities fits well with such a selective processor-sparing viewpoint. Older adults seem to be able to perfom well in motor domains as well as cognitive ones. A variant on the latter view is that practiced tasks are insulated from age-related deterioration, but novel ones that rely on slowed processing show age-related deficits. Ftybash, Hoyer, and Roodin (1986)called the results of practice "encapsulation". There is a substantial body of evidence that practice enables older adults to improve their performance drastically, even to levels well beyond those achieved by novice young adults. Some of the studies mentioned in this chapter flt such a perspective. Nonetheless, it is not yet clear how the complex processes that support everyday activities and more critically, paid employment, fit into this framework. Practice and Skill What processes improve with practice? There is a long history of finding limited transfer of training for practiced activities. Thus far, there is only weak evidence for a general cognitive toning-up function for general mental exercise. With training "near transfer" is commonplace, but "far transfer" is quite weak (e.g.. Baltes,
376
Chamess and Bosman
Sowarka h Kliegl. 1989). Unless you develop a great many different procedures to handle specific cases. simple practice may not improve a broad function such a s memorizing ability very much (see Ericsson, 1985,for a review of memory training). We will need a close examination of typical working environments to understand better the mbt of practiced and unpracticed activities. It may be the streaming of successive pieces of the processing chain that undergoes the greatest change with practice. Anderson ( 1982, 1987)used the terms "compilatfon" and "proceduralization" to describe these processes. Only about one third of the speed-up that Chamess and Campbell (1988)observed in a two-digit mental squaring task was due to faster execution of elementary arithmetic operations. Most of the speed-up was attributed to learning how to chain the subgoals together quickly and efficiently. T h e Fisk, McGee, and Giambra (1988)semantic categorization results indicate that older adults may have a difficult time developing automaticity. Such task organization processes (control processes) may be important mediators of age and skill differences. In short, we have a long way to go in developing adequate models of how Compensatory mechanisms operate in older adults. Even when we look at typing, the one clear-cut case where a compensatory mechanism has been isolated, intriguing questions arise. Why, for instance, don't younger typists avail themselves of the same buffering mechanism to type even faster than they already do? If buffering keystrokes is good for the skilled, and better for the old and skilled. why isn't it best for the skilled young? One argument is that perfomance would break-down if younger typists attempted to read further ahead in the text. Specifically, input processes would get too far ahead of translation and execution processes causing a break-down in performance. The assumption is that the size of the eye-hand span cannot be flexibly controlled by the typist, but as discussed previously, reflects the demand by translation and execution processes for characters. The implication is that the eye-hand span increases in size when text characteristics or skill acquisition increases the speed of execution processes thereby increasing the demand for characters. There is no reason to assume that increasing the she of the eye-hand span by reading further ahead than normal would increase the efficiency
Experttse and Aging
3?7
of execution processes, and as suggested, this may interfere with typing. We can also fall back on the argument that people usually satisfice (Simon. 1969) rather than optimize performance. After all, typists don't usually have to meet a 100 wpm typing speed criterion to be hired or even to be promoted once hired. In certain highly competitive environments. particularly in athletic endeavors, where optimization is the goal, we do see that virtually all the world record holders are young. CONCLUSIONS We want to argue that aging, at least with respect to cognitive tasks and some perceptual-motor ones, can be viewed as a process causing slowing in the rate of information processing. Such slowing leads to poorer performance on non-practiced operations, and likely slower acquisition of new skills, particularly those dependent on new knowledge structures. Countering age-related slowing in most real world tasks are the effects of practice. Practiced operations speed up for young and old alike. Practice can maintain the performance of older adults at high levels, with one proviso. The task to be accbmplished must be self-paced rather than externally paced. That Is. the input flow rate must be under the person's control, otherwise slowing in basic perceptual operations may limit performance. -ping is an example of a task that allows successful compensation, primarily because the maximal output flow rate (keystrokes) is considerably less rapid than input processing rates (reading speed). In sports domains, where output processes are slaved to unpredictable inputs (e.g.. tennis) such buffering is not possible, and age-related slowing effects cannot be overcome. Practice is not necessarily an easy cure for age-related slowing. Sometimes it takes a great deal of practice to produce the cognitive or physiological changes that support high level performance. One can draw the analogy between the "fast-twitch to slow-twitch" change in muscle fiber and the development of compiled programs (automaticity) for cognitive tasks.
378
Charness and Bosman
Still, with this viewpoint comes the admittedly extreme claim that virtually every skilled older person (age 65+)who performs at similar levels to younger adults is doing so via some compensatory mechanism. Further. when.it comes to peak performance in speedconstrained tasks, no old individual will outperform a comparably (Establishing trained young individual in that domain. "comparable training" is not going to be easy, though there are some interesting attempts at cataloguing practice with diaries (Ericsson, Tesch-Rlimer, & Krampe, in press.) The extreme claim of no old individual being better than the best young one derives considerable support from the field of athletics. Nonetheless, Ericsson (in press) noted that some of today's Masters athletes (using the 1979 Masters competitors' records) have achieved performances that equalled or bettered those of young athletes who won events in the 1896 Olympic games. Such record-surpassing or equalling performance only occurred for Masters athletes less than 65 years old, with the exception of the marathon. There were no cases, however, of Masters athletes equalling the 1896 World Records. Given improvement trends, even that barrier is likely to fall, if it hasn't done so already. Another potentially important factor in age decline is forgetting: a form of the "disuse" hypothesis captured in the quote "use it or lose it". Forgetting rates do not vary much with age {Charness & Campbell, 1988; Kriauciunas 1968; Salthouse & Somberg, 1982; Wickelgren. 1975). Admittedly, older adults have a longer time frame over which to forget information than their younger counterparts. particularly if they do not use the information. Some recent research by Salthouse, Babcock, Shovronek, Mitchell, and Palmon ( 1990) suggests, however, that disuse is not the explanation for the decline in spatial visualization ability experienced by skilled older architects. Concatenating these phenomena leads to some interesting predictions. First. there ought to be a great deal of plasticity in the performance of older adults, since it is unlikely that they have practiced a specific skill to asymptotic levels. Second. it will take them considerably longer to reach the same performance levels a s the young: if both practice a task, the gap will close slowly, and perhaps not at all (Fisk. McGee, & Giambra, 1988). (Neither the
Expertise and Aging
379
young nor the old may live long enough to hone a skill to the level of automaticity that brings them down to physiological limits of response.) Third, maximal age differences are expected in simple, poorly practiced tasks, where hardware sets limits on performance. Minimal age differences are anticipated in complex tasks demanding knowledge. when the old are more likely to possess the relevant knowledge than the young, and when they also have the time to prepare their response. Maximal differences can also be expected in complex tasks demanding the assembly of new performance programs, because of the slower learning rate of older adults. Life in the lab is difficult when it comes to aging research. We have to contend with an independent variable, age, that cannot be manipulated. We also run the risk that our tasks will not have much generalizability beyond the lab, Nonetheless, the study of expertise and aging holds some promise that when we look through the lab window, we can see behaviors that are life-like. Most of the time the environment (work or play) poses challenges that do not tax the capabilities of most aging adults too heavily. They resemble young people in their performance at the molar level. Digging a little deeper, you sometimes discover that molar equivalence is achieved in fundamentally different ways by young and old. As biology has amply demonstrated, a diverse set of mechanisms can cope successfully with the same environmental challenges. Aging research is at the early stage of establishing a taxonomy of compensatory mechanisms. It is a necessary first step if we hope to uncover the path that leads from laboratory theories to life practices. ACKNOWLEDGMENTS
This work was supported by grants from the Natural Sciences and Engineering Research Council of Canada, NSERC A0790, and BellNorthern Research. We thank Fran Allard, Tom Hess. and Tim Salthouse for helpful comments on earlier drafts.
380
Charness and Bosman
REFERENCES Allard, F. & Starkes, J. L. (in press). Expertise in human motor performance. In K. A. Ericsson & J. Smith (Eds.) Studies of Expertfse: Prospects and Lirnfts. Anderson, J. R. (1982). Acquisition of cognitive skill. Psychologlcal Review, 89.369-406. Anderson. J. R. (1983).The archttecture of cognftfon.Cambridge, Mass.: Harvard University Press. Anderson, J. R. (1987).Skill acquisition: Compilation of weakmethod problem solutions. Psychological Review, 94,192-210. Balota, D. A. 81 Duchek, J. M. (1988).Age-related differences in lexical access, spreading activation, and simple pronunciation. P ~ y ~ h o l o &Aging, gy 3.84-93. Baltes, P. B., Sowarka, D.. & Kliegl, R. (1989). Cognitive training research on fluid intelligence in old age: What can older adults achieve by themselves? Psychology and Agfng, 4 , 2 17-221. Berliner, H. & Ebeling.C. (1989).Pattern knowledge and search: The SUPREM architecture. ArtiJicfal Intelffgence,38, 161-198. Birren, J. E., & Morrison, D. F. (1961). Analysis of the WAIS subtests in relation to age and education. Journal of Get~ntology,16.363-369. Bosman, E. A. (in preparation), Compensatory mechanfsms employed by older typists. Doctoral Dissertation, University of Waterloo. Botwinick. J. & Siegler. I. C. (1980).Intellectual ability among the elderly: Simultaneous cross-sectional and longitudinal comparisons. Deuelopmental Psychology, 16,49-53. Backman, L. & Molander, B. (1986a). Adult age differences in the ability to cope with situations of high arousal in a precision sport. Psychology and Aging, 1 , 133-139. B&krnan. L. & Molander, B. (1986b). Effects of adult age and level of skill on the ability to cope with high-stress conditions in a precision sport. Psychology and Aging, 1,334-336. Caldemood. B., Klein. G. A, & Crandall, B. W.(1988). Time pressure, skill, and move quality in chess. American Journal of PSyChobgg, 101,481-493. Camp, C. J. (1989).World knowledge systems. In L. W.. Poon. D.C. Rubin. & B. A. Wilson (Eds.) Everyday cognftfonin aduZthood and late Zge. (pp. 457-482).Cambridge: Cambridge University Press.
Expertise and Aging
381
Camp. C. J.. West, R. L., & Poon, L. W. (19891. Recruitment practices for psychological research in gerontology. In M. P. Lawton & A. R. Herzog (Eds.), Special research methods for gerontology (pp. 163-189). Amityville. N.Y.: Baywood. Card, S. K.. Moran. T. P. & Newell. A. (1983). The psychology of human-computer Interaction. Hillsdale, N.J.: Erlbaum. Cerella, J. (1985). Information processing rates In the elderly. Psychological Bulletin, 98(1), 67-83. Chang. T. M. (1986). Semantic memory: Facts and models. Psychological Bulletin. 99, 199-220. Charness, N. (1979). Components of skill in bridge. Canadlan Journal ofPsychology. 33. 1-16. Charness, N. (198la). Aging and skilled problem solving. Journal Of Expepmental P S Y C ~ ~ @ JGeneral, Y: 110, 2 1-38. Charness, N. (1981b). Search in chess: Age and skill difllerences.
Journal of Experimental Psychology: Human Perception and Performance, 7. 467-476. Charness, N. (1981~). Visual short-term memory and aging in chess players. Journal of Gerontology, 36. 615-619. Charness, N. (1983).Age, skill, and bridge bidding: A chronometric analysis. Journal of Verbal Learning and Verbal Behavior, 22, 406-4 16. Charness, N. (1987) Component processes in bridge bidding and novel problem-solving tasks. Canadian Journal of Psychology, 41,223-243. Charness, N. (1988). The role of theories of cognitive aging: Comment on Salthouse. Psychohgy and Aging. 3, 17-21. Chamess, N. (1989). Expertise in chess and bridge. In D. JSlahr and K. Kotovsky (Eds.) Complex information processfng: The impact of Herbert A. Simon (pp. 183-208). Hillsdale, N.J.: Erlbaum. Charness, N. (in preparation). Expertise in chess: The balance between knowledge and search. In K. A. Ericsson & J. Smith (Eds.) Studks of expertise: Prospects and limits. Charness, N. & Bosman, E. A. (1990) Human factors and design for older adults. In J. E. Birren & K. W. Schaie (Eds.) Handbook of the Psychology of Aging (3rd ed.. pp. 446-463). New York: Academic Press, Charness, N. & Campbell, J. I. D. (1988).Acquiring skill at mental calculation in adulthood: A task decomposition. Journal of Experimental Psychology: General. 117,115-129. Chase, W. G. & Simon, H. A. (1973a).The mind's eye in chess. In W. G. Chase (Ed.) Visual information processing. (pp. 215-2811, New York: Academic Press.
382
C h a m s s and Bosman
Chase, W. G. & Simon. H. A. (1973b). Perception in chess. Cognitive P~y~hology. 4, 55-81. Chi, M. T. H.. Glaser, R. & M. J. Fan- (Eds.) (1988).The nature of expertise. Hillsdale, N.J.: Erlbaum. Cooper. W.E. (1983).Introduction. In W.E. Cooper (Ed.). @gnitf.ue aspects of skilled typewriting (pp. 1-38).New York SpringerVerlag. de Groot. A. D. (1978).Thought and choice in chess (2nd ed.). The Hague: Mouton. Elo, A.E. (1965). Age changes in master chess performances. Journal of Gerontology, 20.289-299. Elo, A.E. (1978).The rating of chessplayers past and present. New York Arco. Engle. R. W.& Bukstel. L. (1978).Memory processes among bridge players of differing expertise. Amerlcan Journal of Psychology. 91,673-689. Ericsson. K. A. (1985).Memory skill. Canadian Journal of P~y~hology, 39,188-231. Ericsson, K. A. (in press). Peak performance and age: An examination of peak performance in sports. In P. B. Baltes & M. M. Baltes (Eds.). Successful aging: Perspectfues from the behavioral sciences. New York Cambridge University Press. Ericsson. K. A. & Smith, J. (Eds.) (in preparation). Studies of Expertise: Prospects and Limits. Erlcsson, K. A.. Tesch-Rdmer, C. & Krampe. R. Th. (in press). The role of practice and motivation in the acquisition of expertlevel performance in real life: An empirical evaluation of a theoretical framework. In M. J. A. Howe (Ed.) Encouraging the development of exceptional abilities and talents. Leicester: The British Psychological Society. Feigenbaum. E.A. (1989). What hath Simon wrought? In D. Klahr and K. Kotovsky (Eds.) Complex information processing: The impact of Herbert A. Simon (pp. 165-182). Hillsdale. N.J.: Erlbaum. Fisk, A. D., McGee, N. D., & Giambra. L. M. (1988).The influence of age on consistent and varied semantic-category search performance. Psychology and Aging, 3.323-333. Frey, R L.,Truscott. A. F. & Kearse. A. L. (Eds.) (1976).The ofiial encyclopedia of bridge. New York Crown. Gentner. D. R. (1983).Keystroke timing in transcription typing. In W.E. Cooper (Ed.). Cognitive aspects of skilled typewrittng (pp. 95 - 120).New York: Springer-Verlag.
J3xpertke and Aging
383
Grant, E. A., Storandt, M. & Botwinick. J. (1978).Incentive and practice in the psychomotor performance of the elderly. Journal ofGerontology. 33,413-415. Guinness Book of World Records (1989).New York Bantam. Hartley. A. A. & Hartley, J. T. (1984).Performance changes in champion swimmers aged 30 to 84 years. Experimental Aging Research, 10, 141-147. Hartley. A A & Hartley. J. T. (1986).Age differences and changes in sprint swimming performances of master athletes. Experimental Aging Research, 12,65-70. Herzog. A. R. & Rodgers. W. L. (1989).Age differences in memory performancc and memory ratings a s measured in a sample survey. Psychobgy and Aging, 4. 173-182. Holding, D. H. (1985). Thepsychology ofchess skffl. Hillsdale, N.J.: Erlbaum . Homer, K. L.. Murray, H. G., & Rushton. J. P. (1989). Relation between aging and rated teaching effectiveness of academic psychologists. Psychology and Aging. 4. 226-229. Homer, K. L., Rushton, J. P., & Vernon, P. A. (1986).Relation between aging and research productivity of academic psychologists. Psychology and Aging, 4 . 319-324. Houston. M. E. (1978). The use of histochemistry in muscle adaptation: A critical assessment. Canadfan Journal of Sport Sciences, 3, 109-118. Kausler. D. H. (1982).Experimental psychology and h u m aging. New York Wiley. Kliegl. R.. Smith, J. & Baltes. P. B. (1986).Testing-the-limits, expertise, and memory in adulthood and old age. In F. Klix and H. Hagendorf [Eds.) Human memory and cognitive capabilities (pp. 395-407). Amsterdam: Elsevier Science Publishers. Testing-the-limits and the Kliegl, R., Smith, J. & Baltes, P. B. (1989). study of adult age differences in cognitive plasticity of a mnemonic sktll. Developmental PsycFtologg, 25,247-256. Kriauciunas, R. (1968). The relationship of age and retentioninterval activity in short-term memory. Journal of Gerontology, 23. 169-173. Labouvie-Vief, G. (1985). Intelligence and cognition. In J. E. Birren & K. W . Schaie (Eds.) Handbook of the psychology of aging (2nd ed., pp. 500-530).New York Van Nostrand Reinhold. Lehman. H. C. (1953).Age and achievement. Princeton, N.J.: Princeton University Press. Miller. G. A. (1956).The magical number seven, plus or minus two: Some limits on our capacity for processing information. Psychological Review, 63, 81-97.
384
Charness and Bosman
Molander, B. & Backman. L. (1989).Age differences in heart rate patterns during concentration in a precision sport: Implication Journal of Gerontology: for attentional functioning. Psychological Scfences, 44, P80-87. Mumell. K. F. H. (1965).Human performance in industry. New York: Reinhold. Newell, A. (1989).Putting it all together. In D. Klahr and K. Kotovsky (Eds.) Complex information processing: The impact of Herbert A, Sfmon(pp. 399-440).Hfllsdale, N.J.:Erlbaum. Patel, V.L. & Groen. G. J. (1986). Knowledge based solution strategies in medical reasoning. Cognfffve Scfeme, 10,91 - 1 16. Hau, H. D. & Murphy, M. D. (1988).Role of verbal knowledge in chess skill. American Journal of Psychology. 101.73-86. Rybash, J. M.,Hoyer, W.J.. & Roodin, P. A. (1986).Adult cognitfon and agtng. New York Pergamon. Salthouse, T. A. (1984). Effects of age and skill in typing. Journal of Experimental Psychology: General, 13,345-371. Salthouse, T.A. (1985).A theory of cognftfue aging. Amsterdam: Elsevier Science Publishers. Salthouse, T. A (1986).Perceptual, cognitive, and motoric aspects of transcription typing. Psychologkal Bulletfn. 99, 303-319. Salthouse, T.A. (1988). Initiating the formalization of theories of cognitive aging. Psychology and Aglng, 3.3-16. Salthouse, T. A. (1989). Aging and skilled performance. In A. Colley & J. Beech (Eds.) The acqufsftion of cognitive skills (pp. 247264).Chichester. U.K.:John Wiley & Sons. Salthouse, T. A. (1990).Cognitive competence and expertise in aging. In J. E. Birren & K. W. Schaie (Eds.). Handbook of the psychology of agfng (3rd ed., pp. 310-314).New York Academic Press. Salthouse, T. A, Babcock, R. L., Skovronek, E..Mitchell, D. R D., & Palmon, R. (1990). Age and experience effects in spatial visualization. Developmental Psychology, 26, 128-136. Salthouse, T. A. & Somberg, B. L.. (1982).Skilled performance: Effects of adult age and experience on elementary processes. Journal of Experimental Psychology: General, 1 1 1. 176-207. Schaie. K. W. (1983).The Seattle longitudinal study: A 21-year exploration of psychometric intelligence in adulthood. In K. W. Schaie (Ed.) Longftudlnal studfes of adult psychological development (pp. 64-135). New York Guilford Press. Schaie. K. W. (1989).The hazards of cognitive aging. The Gerontologlst, 29, 484-493.
Expertise and Aging
385
Schulz. R. 81 Cumow, C. (1988).Peak performance and age among superathletes: Track and field, swimming. baseball, tennis, and golf. Journal of Gerontology: Psychological Sciences, 43, P113120. Simon, D. P. & Simon, H. A. (1978).Individual differences in solving physics problems. In R. Siegler (Ed.), Children's thinking: What dewlops? (pp. 325-348).Hillsdale. NJ: Erlbaum. Simon. H. A. (1969).The sciences ofthe a@ticial. Cambridge, M. A.: MI" Press. Simon, H. A. (1979).Information-processing models of cognition. Annual Review of PsychotOgy, 30, 363-396. Simon, H. A & Gflmartin, K. (1973). A simulation of memory for chess positions. Cognitive Psychology, 5,29-46. Simonton, D.K. (1988). Age and outstanding achievement: What do we know after a century of research? Psychological Bulletfn. 104.251-267. Stones, M. J. & Kozma, A. (1982). Cross-sectional, longitudinal, and secular age trends in athletic performances. Experimental Aging Research, 8, 185-188. Stones, M. J. & Korma, A. (1984).Longitudinal trends in track and field performances. Experimental Aging Research, 10, 107-1 10. Stones, M. J. & Kozma, A. (1985).Physical performance. In N. Chamess (Ed.) Agtng and human performance. (pp. 261-291). Chichester, U.K.:John Wiley & Sons. Age trends in maximal physical Stones, M. J. & Kozrna, A. (1986). performance: Comparison and evaluation of models. Experimental Aging Research, 12,207-215. Talland, G.A. (1965).Initiation of response, and reaction time in aging, and with brain damage. In A. T. Welford and J. E. Birren (Eds.) Behaufor, aging, and the nervous system Ipp. 526-561). Springfield, Illinois: Charles C. Thomas. Todd, M., Davis. K. E., & CaEerty, T. P. (1983).Who volunteers for adult development research?: Research findings and practical steps to reach low volunteering groups. International Journal of Aging and Human Development, 18. 177-184. Welford. A. T. (1985).Changes of performance with age: An overview. In N. Charness (Ed.) Aging and human performance (pp. 333-369).Chichester, U.K.: John Wfley & Sons. Wickelgren, W. A. (1975).Age and storage dynamics in continuous recognition memory. Developmental Psychology. 1 1 , 165-169. W i h o n . R.T..& Allison, S.(1989).Age and simple reaction time: Decade difllerences for 5,325 subjects. Journal of Gerontology: Psychological Sciences, 44(2),P29-P35.
This Page Intentionally Left Blank
Aging and Cognition: Knowledge Organization and Utilization Thomas M. Hess Editor) 0 Elsevier Science Publishers B.V. North-Hollandl, 1990
CHAPTERNINE
LEVELS OF KNOWLEDGE UTILIZATION IN VISUAL INFORMATION PROCESSING William J. Hoyer Syracuse University
SUMMARY
The aim of this chapter is to consider the interactive effects of aging and experience on knowledge utilization at several levels of analysis. Age-related and knowledgerelated changes in visual information processing operate at a range of levels of encoding and access beginning with the information available to the observer in the visual array. It is suggested that there are several ways in which knowledge and the processes of aging interact to aff'ect visual information processing performance. First, there are age-related and experience-related changes in the extent to which stimulus constraints enable object recognition. Second, the possession of prior knowledge affects the extent to which particular stimulus factors trigger or prime the' semantic aspect of visual object recognition. Third, the amount and type of prior knowledge affects information selection (e.g., target/distractor filtering) and the semantic processes (e.g., object naming) associated with visual object recognition. Age-related deficits are mainly associated with the selective processing of available information and with access to the semantics- of visual object recognition. That these aspects of perceptual processing are (or seem) less impaired when older adults possess specific prior knowledge relevant to the visual domain suggests that the skilled visual information processing performance of older adults depends on the extent to
388
Hoyer
which knowledge utilization processes have become inpenetratable to normal age losses as a function of domain-relevant experience. The aim of this chapter is to provide a description of developmental change (and regularity) in visual cognition in terms of age-related and experience-related changes in the processes that affect the speed and efficiency of knowledge utilization. Much of the research in cognitive aging has been directed at documenting agerelated changes in the speed and diciency of general encoding and retrieval processes. These findings provide a base relevant to describing age-related and experience-related changes in the processes involved in knowledge utilization. However, little attention has been given to examining the effects of what is processed on encoding and retrieval. It has long been observed that older adults can efficiently encode and retrieve some kinds of information (e.g.. in mastered domains), even under highly demanding conditions and despite deficits in what appear to be requisite processing abilities. Perhaps one of the most challenging issues in the study of cognitive aging is to explain how processing of some kinds of information is maintained against a background of age-related losses in the basic visual cognitive processes. Rabbitt ( 1977) posed the problem as follows: In view of the deterioration of memory and perceptual motor performance Wth advancing age, the right kind of question may well be not "why are old people so bad at cognitive tasks" but rather "how, in spite of growing disabilities. do old people preserve such relatively good performance?" (p. 623). It is a fact that cognition does not show an orderly pattern of change with advancing age. Within- and across-individuals, some aspects of visual cognitive performance improve, some are maintained at roughly the same level, and some decline with aging. Visual cognitive performance depends on a variety of subject characteristics (age, experience) and stimulus factors that interactively determine what and how information is processed. Visual knowledge must be functionally defined relative to the observer. The familiarity, salience, meaningfulness, affective
Knowledge Utilization
389
value, and complexity of the information for a particular subject determine what and how such information is processed. In addition, age-related differences in the speed and efficiency of the basic processes required for the encoding, selection, and retrieval of visual information need to be considered. Central questions in the study of aging and knowledge utilization in visual perception pivot on the relations between the efficiency of pick up of visual information and the efficiency of utilization of acquired knowledge. In complex visual processing tasks. the individual's prior knowledge pertaining to the visual array and the individual's abilities to apply relevant knowledge to task demands, determine performance. A complete description of age-related and experience-related processes affecting performance involves the examination of the processes of knowledge utilization at a number of levels of analysis. Consideration should be given to the processes that are involved in: a) pick up, selection. and organization of stimulus information; and b) selective activation, access. and retrieval of prior knowledge related to the stimulus information. A complete theory of the developmental aspects of knowledge utilization will require specification of the interacting effects of the stimulus characteristics and observer characteristics that determine performance. Emphasis in this chapter is on the analysis of age-related changes in the high-level visual perceptual processes associated with knowledge utilization. Processes at this level are presumed to operate on perceptual units that correspond to objects, and the parts and features that comprise objects, However, it is important to keep in mind that age-related declines in the basic visual processes associated with acuity, contrast sensitivity. retinal eccentricity. accommodation, and adaptation probably aflect the efficiency of higher-level processes [e.g., Fozard, 1990; Kosnik, Winslow. Mine, Rasinski, & Sekuler. 1988).
LEVELS OF ANALYSIS OF KNOWLEDGE UTILIZATION Fodor and Pylyshyn (1988) have noted that the study of cognitive architectures involves the investigation of causal relationships at many diiferent levels of analysis. To quote Fodor
390
Hoyer
and Pylyshyn (1988, p. 9): there is a scientific story to be told about quarks: and a scientific story to be told about atoms: and a scientific story to be told about molecules ... And the story that scientists tell about the causal structure that the world has at any one of these levels may be quite different from the story that they tell about its causal structure at the next level up or down." The implication of these remarks is that analyses of cognitive aging must specify the level of analysis that is at issue. The point is particularly relevant to the investigation of age and experience factors in knowledge utilization because very different kinds of methodologies and theories are involved at different levels of analysis. Types of visual knowledge range from that which is directly available in the visual array to that which gives meaning to what is perceived. Clearly, some aspects of visual processing are directly related to the quantity of information available in the visual array, and on this basis, some researchers have argued that visual processing is most usefully studied at a quantitative level that is free of higher order influences involving memory and inference (Gibson, 1966; McArthur. 1982). According to Gibson, for example, utilization of visual knowledge involves the detection and registration of invariant properties in the optical array. At this level, visual knowledge exists within the myriad of physical and photometric processes involving light and surface reflectance. Indeed, work by Marr (1982). Pinker (1984), Ullman (1984). and Zucker (1986) has shown that visual knowledge can be specifled in terms of measurable characteristics of the structure of visual arrays. It has also been shown that there are detectors and input analyzing systems at the retinotopic level that are highly specialized for responding to certain kinds of featural information (e.g., Todd & Reichel, 1989). Even these early input analyzing systems are influenced by experience across relatively brief exposures (e.g.. as in light adaptation) and by relatively long term experience (e.g., acquired recognition skills, a s in reading). Thus, there is a vast amount of visual knowledge utilization that is fast. ubiquitous, and derived directly from input; and this level of visual knowledge utilization operates prior to the activation of the processes of conscious recognition and object naming. At later levels of knowledge utilization, visual processing entails activation 'I...
Knowledge Utilization
391
of codes and networks of knowledge representations that are associated with recognizing, identifying, categorizing. and otherwise giving meaning to information in the visual array. Recently, Schacter (in press) and Tulving and Schacter (1990) have discussed some the factors that are related to activation of distinct types of codes. Higher-order knowledge utilization processes begin with or draw on the information that is available in the visual array. It is generally assumed that observers bring their prior knowledge to bear on the interpretation of objects and other aspects of the visual world. Consistent with a constructivist view of visual perception, higher-order knowledge serves to enable the individual to go beyond the information given (e.g.. via inferential and constructive processes). According to Neisser (1967). the observer acts a s a "paleontologist who carefully extracts a few fragments of what might have been bones from a mass of irrelevant details and 'reconstructs' the dinosaur" (p. 94). In addition to the service of filling-in and going-beyond. higher order knowledge utilization serves to constrain the selection of what is processed. Zucker (1986) pointed out that there may be an infinite variety of visual signals that could give rise to the same visual knowledge. but that problems of recognition are solved or at least simplified by exploiting constraints in the form of high level semantic knowledge about objects and scenes derived from past experience. As already mentioned, it seems useful to distinguish between visual knowledge that is part of information pick up and various kinds of declarative, representational, and semantic knowledge that provide the conscious experiences of visual perception. Beck (1966) asked young adult subjects to rate the perceptual similarity of an upright and a tilted T and of an upright T and a backward L. The upright T and tilted T were rated as more similar by the subjects, perhaps because readers are experienced at seeing letters in varying orientations. However, analyses of texture grouping revealed that the upright T and the tilted T were more differentiated by the visual system than were the upright T and the backward L. Thus,what appears similar at one level of analysis may be distinct at another level.
392
Hoyer
AGE-RELATED LIMITATIONS IN KNOWLEDGE UTILIZATION Although researchers can seldom pinpoint the specific mechanisms responsible for gains and losses in visual cognitive performance, it is possible to provide a rough summary of the kinds of processes that are most age-sensitive and the conditions under which age-related declines are most likely to be found. It is relatively well-established that rate and efficiency of a variety of higher-order encoding and retrieval processes account for a substantial amount of the interindividual variability in visual cognitive performance (e.g., Cerella. 1985. 1990;Salthouse. 1985). Age-related declines are frequently reported when performance involves visual search, localization. and selectivity (e.g.. Madden, 1986;Plude & Hoyer, 1981, 1985. 1986;Sekuler & Ball, 1986). mental rotation and spatial visualization (e.g., Berg, Hertzog. & Hunt, 1982;Salthouse, Babcock, Skovronek, Mitchell, & Palmon, 1990),memory search and retrieval (e.g., Salthouse & Somberg. 1982;Wickens, Braune. & Stokes, 1987).and stimulus selection under conditions of uncertainty (e.g., Hoyer & Familant. 1987; Nissen & Corkin. 1984;Rabbitt. 1982). Craik (1983)suggested that adult age dmerences in memory can be. explained in terms of the extent t o which self-initiated constructive operations are required. Craik pointed out that agereductions in memory performance are most evident in retrieval situations that require substantial self-initiated effort. By extension, some knowledge utilization tasks provide minimal environmental or contextual support and require a high degree of self-inititated activity, whereas other tasks are substantially context-bound and do not seem to involve effortful retrieval. Studies of semantic priming and aging typically show negligible age differences in the extent to which words activate networks of related words and meanings (Chiarello, Church, & Hoyer, 1985; Howard, McAndrews. & Lasaga, 1981). However, there are agerelated declines in the efficiency of retrieval of contextual information (e.g., Moscovitch, Winocur, & McLachlan, 1986).and in the extent to which certain information is accessible or available through priming (e.g.. Clancy & Hoyer. 1990).Recently, Tulving and Schacter (1990)provided evidence to suggest that memory involves
Knowledge Utilization
393
the operation of functionally independent (or dissociated) processing systems. Some findings in the area of memory aging have been interpreted along these lines (Light & Singh, 1987; Mitchell, 1989;but see Chiarello & Hoyer, 1988).
HOW DOES KNOWLEDGE UTILIZATION ENHANCE VISUAL COGNITIVE PERFORMANCE?
The benefits of knowledge utilization involve making emcient use of the information available in the visual array. Earlier. it was suggested that prior knowledge and experience have their effects at several levels of analysis. At an early level, information in familiar scenes is probably in a form that can be readily picked up without appealing to representative or inferential processes. Early detection is relatively automatic: the subject has little volitional control over what is selected, even though aspects of detection are influenced by prior knowledge. There seem to be few if any agerelated limitations at the level of parsing common or familiar visual arrays; efficiency of early detection is age-invariant and asymptotic. given sufficient perceptual experience. At this level, processing is merely a matter of detecting contextual constraints, a relatively resource-free, age-invariant process. At higher levels, processing involves access to prior knowledge in the form of schemas, and there is some evidence to suggest no decline in the contextualization of input and retrieval processes with aging (Azmitia & Perlmutter. 1988: Hess & Slaughter, in press). Object recognition is strongly contextually-based, object recognition is improved when the object is presented as part of coherent scene (e.g.. Biederman, 1981:Biederman. Blickle, Teitelbaum, & Klatsky, 1988;Boyce, Pallatsek & Rayner. 1989:Friedman. 1979).and the meaning of pictorial displays is picked-up in a glance, within 150 milliseconds (Boyce et al., 1989;Loftus, Nelson, & Kallman. 1983). At intermediate levels. where processing involves activation of memory and inference processes that give meaning and interpretation to visual scenes, there are age-related declines. However, the efficiency of these processes remains unaffected (or less affected) by age-related insult within highly practiced domains. I suggest that prior knowledge does not affect the pick up of
394
Hoyer
information from the visual array; rather, it affects the higherorder, resource-limited aspects of processing, which would ordinarily be slower with aging and without the benefit of prior knowledge. The benefits of knowledge utilization in some kinds of visual tasks may be associated with the attentional factors of preparation and selectivity. Suppose you are trying to find a contact lens that dropped onto a multi-colored carpet. If you know some of the characteristics of a contact lens, such prior knowledge will serve to simplify the task of finding the object. The benefits of prior knowledge in visual selective attention can be inferred from studies showing improvements in performance under conditions in which the opportunity to expect particular stimulus information is increased (e.g.. Hoyer & Familant. 1987;Posner. 1980).The time needed to access or use preparatory information is age-related and practice-related. Also. preparatory information (e.g., primes) can trigger access to different semantic codes at different rates as a function of age (Mitchell, 1989)and experience (Clancy & Hoyer, 1990).Along these lines, Posner and his colleagues (e.g., Posner, Inhoff, Friedrich. & Cohen, 1987)have provided some evidence that different aspects of attentional selectivity are controlled by different neural mechanisms. However, it should be noted that examination of the effects of age-related and knowledge-related differences in the time-course of selectivity and preparation involves consideration of a range of factors at a number of levels of knowledge utilization. First. the factors affecting preparation and selectivity (i.e., longer SOAS, more practice trials, consistent mapping) may yield improvements in some aspects of performance (e.g.. automatic access to targets in memory), but other aspects of performance (e.g., alertness) may show deterioration (i.e., altertness decrement) a s a function of the same factors depending on the task and individual differences in the time course of these effects. Second, the relation between performance benefits and costs and the readiness to select particular targets depends on the individual's level of skill at the task. Third, performance may be relatively domain-specific because the processes of attentional selectivity and preparation become dedicated to expected stimulus-response mappings. Fourth,
Know ledge Uttlbzatlon
395
performance may vary as a function of age and skill differences in strategy if, for example, there are reliable age-related and skillrelated individual differences in response criteria, stimulus sensitivity, or speed-accuracy trade-offs. All of these issues involve the extent to which preparation and selectivity affect different component processes at different levels of knowledge utilization. depending on a variety of age-related and experience-related individual difference variables. One of the best studies of the effects of age and skill was conducted by Salthouse (1984). He found that typing speed was uncorrelated with age for skilled typists, despite the expected negative correlations between age and psychomotor speed. On the typing task, it seemed that older skilled typists compensated for general psychomotor slowing by using a strategy of looking farther ahead within the window of available text; apparently, looking further ahead gave more time to plan the execution of keystrokes. Salthouse's findings bear on the domain-specific nature of skilled performance, since skilled typists did not effectively employ the "look ahead" strategy on other domain-general tasks. In another study, Clancy and Hoyer (1988)examined the effects of age and skill on the capacity demands of visual recognition performance. Although previous investigators had examined dflerences between novices and experts in the visual analysis of radiographic displays (e.g., Lesgold, Rubinson. Feltovich. Glaser, Klopfer & Wang. 1988; Myles-Worsley. Johnston, 8t Simons, 19881, Clancy and Hoyer selected the domain of medical laboratory diagnostics because there was a relatively wide age range of equivalently skilled individuals in this professional field. The experimental task involved visual detection of microbiological materials, and expert subjects were medical laboratory technologists with a substantial amount of knowledge and experience in microbiological analysis (Le., 5-20 years). It was expected that younger and middle-aged control subjects would be slower and less accurate under dual-task conditions when skilled visual recognition was the primary task compared with agematched experts. As predicted, skilled obsewers showed a typical pattern of age-related decline in domain-general visual processing. but only minimal age-related reductions in domain-specific visual
396
Hoyer
detection. Although these results suggest the importance of knowledge in accounting for maintenance of domain-specific performance in older adults, the findings r a s e questions about how to explain the benefits of knowledge. In this study, the beneficial effects of prior knowledge on visual recognition can be interpreted in several ways, First, it can be suggested that if skilled observers know what to see or look for in famillar displays, then compared to less skilled observers, less processing time (and/or capacity) is required for handling nonsalient infonnation. That is. compared with novices, it seems that skilled observers use different (or fewer) processing components to perform the same complex task. The typically observed deficits in the computational aspects of recognition, identification, search, selection, and localization are somehow made more efficient. these systems are either circumvented or replaced. or these systems have sufficient lead time to "run" without noticeable deficit. Further, there may be reduced susceptibility to distractor information in skilled domains. Skilled older observers may be able to effectively "see through" the noise available in familiar displays, even though older adults are generally more prone to noise distractors when searching unfamiliar displays (e.g.. Cremer & Zeef. 1987). It may be that the disadvantageous effects of noise are at minimum for experts, because of the benefits of attentional selectivity and preparation, which attenuate the typical pattern of age-related differences found when familiarity or preparation effects are not available. It is also reasonable to suggest that visual recognition performance in skilled domains involves a criterion adjustment at the encoding or identification stages of visual representation. Familiar context may affect the observer's criteria for reporting object identification on the basis of partial stimulus information. Particular features of the display may operate as primes to facilitate the encoding and stimulus identification processes through network activation. Thus, skilled performance in this domain may involve increased use of contextual cues and constraints available in the display, especially for the older experts.
Knowledge Utilizatfon
397
KNOWLEDGE AT THE LEVEL OF OBJECT FEATURES
Emphasis can be given to the importance of both feature-driven and object-driven processes in accounting for age-related and knowledge-related differences in visual cognitive performance. There is compelling evidence to suggest that visual objects are processed in terms of their parts or features. Purely feature-driven theories of visual cognition suggest that the observer first encodes structural features and then somehow organizes these features into patterns or objects (e.g., Biedeman. 1987: Julesz. 1984; Marr. 1982; Pinker, 1984). This data-driven view focuses on the ability of the visual processing system to "pick-up" and analyze the essential features or parts of objects, referred to as "primitives", and the ability of higher-order processing systems to represent and construct objects in the world in terms of such features. Presumably, a collection of primitives combine to give rise to phenomenologically real (or identifiable) objects. Anne Treisman's research on early feature analysis is exemplary of some of the work in this area. Treisman and Gormfcan (1988), for example, used visual search asymmetry data as evidence to suggest that the visual system automatically codes such elementary features a s color, contrast, brightness, texture, line ends, blobness. tilt, and curvature. Treisman and Gormican also suggested that there are temporary recognition networks (stored descriptions of objects), and that these relatively transient, knowledge-based networks serve to handle variations in a wide range of stimulus factors (illumination, blur, noise, object quality). Another line of research bearing on the role of structural features in object recognition gives emphasis to line arrangement principles and simple grouping features that serve to define the basic units le.g.. blocks, cylinders, cones) that comprise everyday objects (e.g.. Biederman, 1987; Lowe, 1985; Marr, 1982; Ullman. 1984). From this perspective, the computational requirements of visual perception involve the composition of features into parts and shapes. and the exploitation of knowledge-based uniformities at first approximation to recognition. According to Hoffman and Richards (1984). the visual system first parses object boundaries and extracts critical stimulus features. Second, parts of the object
398
Hoyet
are arranged hierarchically and organized on the basis of certain rules and regularities. Third, inferences are made based on featural configurations such that specific featural arrangements trigger memory search and activate specific memory representations. Based on research findings already discussed, the observer's relevant knowledge or experience affects object recognition by improving (or simplifying) the computational efficiency of these steps at the feature processing level. Biederman's ( 1987)recognition-by-components (RBC)theory is representative of the view that everyday objects are composed of basic geometric units. He proposed that the edges of any twodimensional image are segmented at regions of concavity into simple volumetric components (geons)such as cylinders, cones, and cubes. Although Biederman stressed the importance of information at the vertices, at areas of cotermination, and at the regions of convexity and curvature, as sources of first-order information about the organization of the features of objects, he argued that objects containing the same features are distfnguished on the basis of specific relations among the geons of an object. It is the organization of geon components that is matched to representations in memory. Thus, FU3C theory predicts the conditions under which specific types of contour deletions prevent object recognition. If essential features of the object are unavailable (i.e., by the removal or omission of the nonaccidental properties of collinearity or curvilinearity, RBC theory predicts impaired object recognition. In support of this hypothesis, Biederman and Blickle (cited in Biederman, 1987)reported that if edge-based deletions of contour were made at regions of concavity in such a way that collinearity or curvilinearity would not fill-in gaps, then the object was unrecognizable (nonrecoverable). Figure 1 illustrates the differences between recoverable and non-recoverable deletions at three levels of fragmentation as used in Biederman and Blickle's study. They found that when deletions were made at midsegment. and the visual properties were not constrained, the missing information was readily filled-in and the object was identifiable. When deletions corresponded to the removal or omission of the nonaccidental properties of collinearity or curvilinearity, object recognition was prevented, even at very long exposure durations (5
Knowledge UtiIlzatfon
399
NONRECOVERABLE
mure 1. Examples of Biederman's (1987)recoverable and nonrecoverable objects
seconds) and even under several conditions of familiarity. Biederman concluded that the most critical information was provided by the primitives of the object. and when these components were removed, object recognition was virtually impossible. A recent study by Frazier and Hoyer ( 1990)examined adult age differences in the type and amount of information necessary for object recognition using Blederman's (1987) RBC theory as a framework. The purpose of this study was to determine if there were age-related differences in the use of invariant features and contextual features in the recognition of incomplete objects. Consistent with Biederman and Blickle's findings. the type of features removed from the objects was a significant factor for younger adults. However, older adults were substantially slower
400
Hoyer
and less accurate at recognizing objects under fragmented conditions, regardless of the type of fragmentation. Results of two studies by Salthouse (i.e.. Danziger & Salthouse, 1978:Salthouse & Prill, 1988)also suggest that there is a variety of factors affecting adult age differences in object recognition. Salthouse and Prill (1988)investigated the extent to which age differences in the perception of incomplete objects is related to differences in the effectiveness of deriving implications, testing implications, and evaluating hypotheses regarding perceptual information. Deriving implications from partial information in a display required that subjects locate useful regions of the stimulus. Subjects were asked to locate the most informative regions of figures. Two measures of informativeness were investigated, absolute location and location relative to a fixed position within the display. Participants were asked to compare a display of a specific complete part of an object with an incomplete version of an object, and judge whether the part could have belonged to the figure. In another condition, subjects were shown two displays containing incomplete parts of a n object, and were to decide if both parts came from the same original object. This condition was designed to examine age differences in testing and evaluating hypotheses. Although it was not possible to attribute age differences in perceptual closure to a particular component, Salthouse and Prill (1988)suggested that each component was age-sensitive and contributed to the age deficit in object processing (see also Danzinger & Salthouse. 1978). KNOWLEDGE AT A REPRESENTATIONAL LEVEL
Although sometimes it seems that researchers are just now discovering the influence of knowledge in perceptual analysis, some of the first scientific work in the area of visual perception emphasized the relationship between knowledge and perception. In 1867,for example, Helmholtz took the position that we perceive the objects or events that most likely fit the sensory pattern that we are trying to interpret (Hochberg, 1968). In recent years, there has been a large amount of research demonstrating the effects of top-down knowledge on the selection of features and on the semantics of
Knowledge Utlltzzation
401
object recognition (e.g.. Hoyer & Plude. 1982: ltrersky & Schiano, 1989; Ullman, 1984). The current view is that top-down visual routines subserve the analysis and selection of particular information (e.g., specific shape properties, figure boundaries, and feature locations) that enable recognition (e.g., Ullman, 1984). Intuitively, given the veridicality of object recognition, one is practically forced to accept the position that visual cognition is highly inferential and dependent on prior knowledge. contextual factors, expectations, and so on. At one level, recognition of objects involves matching descriptions of what one sees with representational knowledge of what one has already seen. It seems likely that the processes involved in selecting, encoding, and integrating featural information about objects, and the processes involved in accessing object-specific knowledge or representations, are the main loci of age-related and experience-related differences in visual cognition. Practically, the computational processes at the detection level of visual object recognition must be largely automatic. Imagine the computational demands of recognition if every time one looked at an object one had to construct it from its features and parts, as well as its orientation, size, color, angle of view, luminance, and degree of occlusion. Such constraints suggest the principles that involve the use of knowledge in object recognition. First, it is necessary to assume that object recognition does not depend solely on the efficiency of processing featural details. Object recognition does not require processing of evey feature of an object prior to recognition. Knowledge probably serves the function of filling-in when featural information is incomplete or when there is missing information. The observer's knowledge and expectancies serve to simplify the demands of object processing. Second, top-down knowledge allows for constancies in object recognition despite featural variations associated with orientation, size, color, degree of occlusion, luminance, and so on. Given these principles, older observers may come to rely more on knowledge-based expectancies and constancies in familiar situations, as feature-integration processes become less eflicient and more time-demanding. More broadly, age-related gains and losses in various aspects of cognition have been characterized in terms of increased domain-
402
Hoyer
specificity and reduced plasticity of knowledge access (Baltes. 1987; Hoyer, 1987;Hoyer, Rybash, & Roodin, 1989;Rybash, Hoyer, & Roodin, 1986).Despite the well-known age-related declines in some of the components of infobnation processing, performance seems relatively unaffected by aging in skilled domains. Skilled performance can be maintained or extended more easily in some domains than in others. For example, performance in fast-paced types of work or sport eventually declines with age, even for experts (e.g.. Backman & Molander, 1986).yet skilled performance in cognitive domains that allow more time for planning, knowledge utilization. and retrieval is more likely to be maintained with age. Performance in skilled domains is associated with the efficiency of access to domain-specific knowledge and/or to the utilization of contextual cues or constraints available in familiar tasks. Specialized knowledge can facilitate the encoding and/or retrieval of complex visual patterns in a variety of ways. In studies of game domains, for example, Chase and Simon (1973)and DeGroot (1966)reported that chess masters could recall the positions of approximately 25 pieces from a game after having seen the display for 5 seconds, while a novice player could remember the locations of only about one-quarter as many pieces after the same exposure. However, even though experts were able to recall large amounts of domain-specific information, experts were no better than novices on domain-general measures of memory. The flnding that experienced individuals can recall at a glance far more specialized information than novices has also been reported in other domains such as the analysis of radiographic displays (e.g.. Myles-Worsley. Johnston, & Simons, 1988). CONCLUSIONS
It has been suggested that age-related and experience-related changes interact at several levels to determine task performance. Although most cognitive aging researchers would probably agree with this position, some of the problems associated with vedying (and unverifying) this approach should be mentioned. First, it is often difficult to isolate and independently measure age-related and experience-related variables. Clancy and Hoyer ( 1988) and
Knowledge Utiltzatfon
403
Salthouse (1984) were able to independently sample the subject variables of years of experience and age: however, experience is usually cumulative and age-correlated, and there can be substantial practical dFfficulties in obtaining representative samples of these variables (see Salthouse. 1990). Despite such practical dimculties, it is still surprising that so little research effort has been directed to the examination of the characteristics (e.g., experience, knowledge) that individuals bring to cognitive tasks, given the amount of experience-related interindividual variability typically obsenred in cognitive aging studies. Another problem is that observed, age-related declines in the speed or efficiency of visual encoding and access are generally attributed to age-reductions in the capacity, rate, or allocation of resources, but such claims must be considered unsubstantiated and unverified, without an independent measure of capacity, rate, or resource. However, careful analysis of the structure and organization of the visual array can provide a valid measure of the m o u n t and salience of visual knowledge that can be extracted from the array. Finally, it is suggested that researchers should take into account what is known about age-related changes in brain and sensory function when generating hypotheses about age-related changes in visual information processing. For example, Plude and Hoyer (1985, 1986) and others (e.g., Sekuler & Ball, 1986) have demonstrated that there are substantial age-related deficits in visual search performance, and relatively minor age dmerences in visual filtering 0.e.. object recognition and identification without search or localization requirements). These findings are consistent with recent work in the neurophysiology of vision suggesting that information about "what" and "where" is processed separately in the visual system (see Cowey, 1985). It has been shown that the neurons that represent an object's shape have relatively large receptive fields, and these neurons respond when the stimulus falls anywhere within a wide range of positions. Recently, Mishkin (e.g., Mishkin, Ungerleider, & Macko, 1983) has reported that different visual systems are responsible for locating objects and identifying objects within displays. One system is involved in the recognition of objects and features thereof, and enables application of
404
Hoyer
previously acquired knowledge to the task of object recognition. Once one has recognized an object, one h o w s its function and some of its attributes. A second (and presumably independent) processing subsystem enables the location of objects within the visual field. In this case, the task is not to encode features or objects in order to access relevant semantic knowledge, but the task is to locate "where" things are. The reported age-related deficits in localization, in contrast to no age differences in identification processes, may be interpreted a s suggesting differential aging of particular neurophysiological subsystems that subserve visual processing. ACKNOWLEDGMENTS Preparation of this chapter was supported by NIA grant AGO604 1 to WJH. REFERENCES Azmitia, M., & Perlmutter, M. (1988).Age differences in adults' scene memory: Knowledge and strategy interactions. Comprehensive Gerontology, 2, 75-84. Bi.ichan, L.,& Molander. B. (1986).Effects of adult age and level of skill on the ability to cope with high-stress conditions in a precision sport. Psychology and Aging, 1,334-336. Baltes. P. B. (1987).Theoretical propositions of life-span developmental psychology: On the dynamics between growth and decline. Developmental Psychology, 23, 61 1-626. Beck, J. (1966).Effect of orientation and of shape similarity on perceptual grouping. Perception and Psychophysfcs, 1, 300-302. Berg, C., Hertzog, C. K.. & Hunt, E. (1982). Age Werences in the speed of mental rotation. Developmental Psychology. 18.95-107. Biederman, I. (1981).Do background depth gradients facilitate object identification? Perceptton, 10, 573-578. Biederman, I. ( 1987). Recognition-by-components:A theory of human image understanding. Psychological Review. 94, 1 15147. Biedeman, I., Blickle. T. W., Teitelbaum, R. C., & Klatsky. G. J. (1988).Object search in nonsense displays. Journal of Experimental Psychology: Learning, Memory, and Cognition, 14,456-467.
Knowledge Ut11Lza t ion
405
Boyce, S. J.,Pollatsek. A..& -per. K. (1989).Effect of background information on object identiflcation. Journal of Experimental Psychokgy: Human Perception and Perfxmmce, 15, 556-566. Cerella, J. (1985).Information processing rates in the elderly. Psychologfcd Bulletin. 98,67-83. Cerella, J. (1990).Aging and infomation processing rate. In J. E. Birren & K. W. Schaie (Eds.), Handbook of the psychology of ugmg (3rd ed. , pp. 201-221).New York: Academic Press. Chase, W.G.. & Simon, H. k (1973).Perception in chess. Cognitive Psychology. 4. 55-81. Chiarello, C., Church, K., & Hoyer. W. J. (1985).Automatic and controlled semantic priming: Accuracy. response bias, and aging. Journal of Gerontology,40, 593-600. Chiarello, C., & Hoyer. W.J. (1988).Adult age dlfferences in implicit and explicit memory. Psychofogy and Aging, 3,358-366. Clancy, S . M..& Hoyer, W. J. (1988).Effects of age and skill on domain-specific search. In V. L. Pate1 & G. J. Groen (Eds.), Proceedings of the tenth conference of the Cognitive Science Socfety (pp. 398-404).Hillsdale, NJ: Erlbaum. Clancy, S. M.. & Hoyer. W. J. (1990,April). Age and skU in prfmfng. Paper presented at the Cognitive Aging Conference. Atlanta, GA. Cowey, A. (1985).Aspects of cortical organization related to selective attention and selective impairments of visual perception: A tutorial review. In M. I. Posner & 0. S. M. M a r i n (Eds.), Attention and performance XI (pp. 41-62).Hillsdale. NJ: Erlbaum. Craik. F. I. M. (1983).On the transfer of information from temporary to permanent memory. Philosophical Damactions of the Royal Society of London, B302, 341-359. Cremer, R., & Zeef, E.J.(1987).What kind of noise increases with age? Journal of Gerontology, 42. 515-518. Danziger, W. L., & Salthouse, T. A. (1978). Age and the perception of incomplete figures. Experimental Aging Research, 4,67-80. DeGroot, A. (1965). Thought and choice in chess. The Hague: Mouton. Fodor, J. A., & Qlyshyn, 2. W. (1988).Connectionism and cognitive architecture. Cognitton, 22, 3-71. Fozard, J. L. (1990). Vision and hearing in aging. In J. E. Birren & K. W. Schaie (Eds.),Handbook of the psychology of aging (3rd ed. pp. 150-171). New York Academic Press. Frazier, L. D., & Hoyer, W. J. (1990). Components of object recognition: Adult age dffferences. Unpublished manuscript, Syracuse University.
406
Hoyer
Friedman. A. (1979).Framing pictures: The role of knowledge in automatized encoding and memory for gist. Journal of Experimental Psychology: General, Z 08. 3 16-355. Gibson, J. J. (1966).The senses considered as perceptual systems. Boston: Houghton-Mifflin. Helmholtz, H. von. (1962).neatbe on physiofogfcal optics: Vol3. New York Dover (original work published in 1867). Hess, T. M.,& Slaughter, S . J. (in press). Schematic knowledge influences on memory for scene information. Developmental Psychology. Hochberg, J. (1968). Perception. Englewood Cliffs, NJ: Prentice Hall. Hoffman, D. D., & Richards, W. (1985).Parts of recognition. Cognition, 18,65-96. Howard, D. V., McAndrews, M. P., & Lasaga, M. I. (1981).Semantic priming of lexical decisions in young and old adults. Journal of Gerontology, 36, 707-714. Hoyer, W. J. (1987).Acquisition of knowledge and decentralization of g in adult intellectual development. In C. Schooler & K. W. Schaie (Eds.), Cognitfuefunctiontng and soctal structures ouer the lfe course (pp. 120-141).Norwood, NJ: Ablex. Hoyer, W. J.. & Familant, M. E. (19871.Adult age dHerences in the rate of processing expectancy information. C o g n ftiue Development, 2, 57-70. Hoyer, W. J., & Plude, D. J. (1982).Aging and the allocation of attentional resources in visual information processing. In R. Sekuler, D. Kline, & K. Dismukes (Eds.). Aging and human ulsualfunction (pp. 245-2631,New York Alan R. Liss. Hoyer, W. J., Rybash. J. M., & Roodin. P. A. (1989). Age-related cognitive change as a function of knowledge access. In M. L. Commons, J. D. Sinnott, F. A. Richards. & C. Armon (Eds.), Adult development: Vol. 1 . Comparisons and applications of developmental models (pp. 293-305). New York Praeger. Julesz. B. (1984).Toward an axiomatic theory of preattentive vision. In G. M. Edelman. W. E. Gall, W. M. Cowan (Eds.), Dynamfc aspects of neocorticat function (pp. 585-612).New York Neurosciences Research Foundation. Kosnik. W.. Winslow, L.,Kline, D. W.. & RasinsM. K., & Sekuler, R (1988).Visual changes in daily life throughout adulthood. Journal of Gerontology: Psychological Sciences, 43, P63-P70.
Knowledge Utilization
407
Lesgold. A.. Rubinson, H..Feltovitch. P.. Glaser. R, Klopfer, D., & Wang. Y. (1989).Expertise in a complex skUk Diagnosing X-ray pictures. In M. T. H. Chi. R Glaser. & M. J. Farr (Eds.). The nature of expertfse (pp. 31 1-342).Hillsdale, NJ: Lawrence Erlbaum Associates. Light, L. L.. & Singh, A. (1987).Implicit and explicit memory in young and older adults. Journal of Experimental Psychology: Learning, Memory, and Cognftfon.13, 531-541. Loftus. G. R., Nelson, W. W.. & Kallman. H. J. (1983).Differential acquisition rates for different types of information from pictures. Quarterly Journal of Experimental Psychology, 35. 187-198. Lowe, D.G.(1985).Perceptual organlzation and ukual recognition. Boston: Kluwer. Madden, D. J. (1986).Adult age differences in the attentional capacity demands of visual search. CognftfueDeuelopment, 1. 335-363. Marr. D. (1982). Vision. San Francisco. CA: Freeman. Computer vision and perceptual psychology. McArthur. D. J. (1982). PsychologfcalBulletin, 92. 283-309. Mishkin, M., Ungerleider, L. G., & Macko, K. A. (1983).Object vision Trends fn and spatial vision: T w o cortical pathways. neurosciences, 6,414-417. Mitchell, D. B. (1989). How many memory systems? Evidence from aging. Journal of Experimental Psychology: Learning, Memory, and Cognition, 15,31-49. Moscovitch. M., Winocur, G., & McLachlan. D. (1986).Memory a s assessed by recognition and reading time in normal and memory-impaired people with Alzheimer's Disease and other neurological disorders. Journal of Experimental Psychology: General, 115, 331-347. Myles-Worsley, M., Johnston, W. A.. & Simons. M. A. (1988).The influence of expertise on X-ray image processing. Journal of
Experimental Psychology: Learnfng, Memory. and Cognitfon. 14,553-557. Neisser. U. (1967). Cognitive psychology. New York Appletion Century Crofts, Nissen. M. J., & Corkin, S. (1984). Effectiveness of attentional cueing in older and younger adults. Journal of Gerontology, 40. 185-191. Pinker, S . (1984). Visual cognition: An introduction. Cognition 18, 1-63.
408
Hoyer
Plude. D. J.. & Hoyer. W. J. (1981).Adult age differences in visual search as a function of stimulus mapping and processing load. Journal Of Gm~ntology,36, 598-604. Plude, D.J., & Hoyer, W.J. (1985).Attention and performance: Identifying and localizing age deficits. In N. Charness (Ed.), Aging and human performance (pp. 47-99). London: Wiley. Aging and the selectivity of visual Plude, D. J., & Hoyer, W. J. (1986). information processing. Psychology and Aging, 1. 1-9. Posner, M. I. (1980).Orienting of attention. Quarterly Journal of Experimental Psychology, 32, 3-25. Posner. M. I.. Inhoff. A. W., & Friedrich. F. J. (1987).Isolating attentional systems: A cognitive anatomical analysis. PSyChoblology, 15, 107-121. Rabbitt, P. M. A. (1977).Changes in problem solving ability in old age. In J. E. Birren & K. W. Schaie (Eds.), Handbook of the psychology of aging (2nd ed., pp. 606-625).New York: Van Nostrand Reinhold. Rabbitt. P. M. A, (1982).Breakdown of control processes in old age. In T. M. Field. A Huston, H. C. Quay, L. Troll. & G. Finley (Eds.). Review of human deuelopment (pp. 540-550).New York Wiley. Rybash, J. M., Hoyer, W. J.. & Roodin. P. A. (1986). Adult cognition and aging: Developmental changes in processing, knowing, and thfnking. Elmsford. New York Pergamon Press. Salthouse, T. A. (1984). Effects of age and skill in typing. Journal of Experimental Psychology: General. 1 13, 345-371. Salthouse, T. A. (1985).Speed of behavior and its implications for cognition. In J. E. Birren & K. W. Schaie (Eds.), Handbook ofthe psychology of aging (2nd ed.. pp. 400-426).New York: Van Nostrand Reinhold. Salthouse, T. A. (1990).Cognitive competence and expertise in aging. In J. E. Birren & K. W. Schaie (Eds.). Handbook of the psychology of aging (3rd ed., pp. 310-319):New York: Academic Press. Salthouse, T. A., Babcock, R. L., Skovronek, E., Mitchell, D. R. D., & Palmon, R. (1990).Age and experience effects in spatial visualization. Developmental Psychology, 26, 128-136. Salthouse, T.A., & Prill. K.A. (1987).Effects of aging on perceptual closure. American Journal of Psychology. 101. 217-238. Salthouse, T.A.. & Somberg, B. L. (1982). Skilled performance: The effects of adult age and experience on elementary processes. Journal of Experimental Psychology: General, 1 1 1 , 176-207.
Knowledge Utilization
409
Schacter, D. L. (in press).Perceptual representation systems and implicit memory: Toward a resolution of the multiple memory systems debate. In A. Diamond (Ed.), Development and neural bases of higher cognitiue functions. New York Annals of the New York Academy of Sciences. Sekuler, R., & Ball, K. (1986).Visual localization: Age and practice. Journal of the Optical Society of America. 3, 864-867. Todd, J. T.,& Reichel. F. D. (1989).Ordinal structure in the visual perception and cognition of smoothly curved surfaces. Psychological Review. 96,643-657. Treisman, A.. & Gormican, S. (1988).Feature analysis in early vision: Evidence from search asymmetries. Psychological Reulew, 95.15-48. Tulving, E.. & Schacter. D. L. (1990).Priming and human memory systems. Sctence, 247.301-306. Tversky, B., & Schiano, D. J. (1989).Perceptual and conceptual factors in distortions in memory for graphs and maps. Journal of Experimental Psychology: General. 1 1 8 . 387-398. Ullman, S.(1984). Visual routines. Cognition. 18, 97-159. Wickens, C. D.. Braune, R, & Stokes, A (1987). Age merences in the speed and capacity of information processing: 1. A dual-task approach. Psychology and Aging, 2.605-614. Zucker, S . W. (1986).The diversity of perceptual grouping. In M. &bib 8t A. Hanson (Eds.), Vision, brain, and cooperative computatton (pp. 1-29). Cambridge: MIT Press.
This Page Intentionally Left Blank
Aging and Cognition: Knowledge Organization and Utilization Thomas M. Hess (Editor) 0 Elsevier Science Publishers B.V. (North-Holland), 1990
CHAPTER TEN
AGING AND EVERYDAY COGNITNE ABILITIES Steven W.Cornelius Cornell University
SUMMARY
In this chapter, recent work on everyday cognitive abilities in adults and the elderly is reviewed. Despite substantial theory and data that have been generated in the study of intellectual aging, there has been a recurrent concern that traditional paradigms for assessing intellectual abilities are insensitive to cognitive skills adults use in adapting to the demands of everyday life. In the first section, we examine how this concern has been paralleled by recent studies of people's implicit theories of intelligence. Implicit theories refer to people's ideas and beliefs about intelligence and how it changes across the life span. Findings show that people have distinct concepts of everyday and academic intelligence, and they believe that facets of everyday intelligence become increasingly salient, especially in middle age and late adulthood. In the second section, research on adults' selfconceptions of intelligence and conceptions of ability tests is discussed. The findings underscore the multidimensional nature of everyday intelligence and highlight the distinct importance of practical and social cognitive abilities in conceptions of everyday intelligence. In the third section, two tasks recently developed to assess adults' strategic knowledge about solutions to problems in everyday and stressful situations are described. Results show differences and similarities between performance on everyday and traditional ability
412
Cornelius
tests and suggest the potential for growth in everyday cognitive abilities during adulthood. Research on intelligence has been a focal point of scientific inquiry in the psychology of aging for more than 70 years (e.g., Dixon, Kramer, & Baltes, 1985:Woodruff, 1983). Initially, much of the research focused on age-related declines in functioning. However, more work has appeared during recent decades which qualifies conclusions about general and universal decline. Indeed, it now appears that whether intelligence declines, remains stable, or increases with age depends on which aspect of intelligence is examined. Findings that have emerged from research on intellectual aging suggest that intelligence is a multidimensional construct, and cognitive changes during adulthood involve a n interplay between decline. stability, and growth (e.g.. Denney. 1982; Labouvie-Vief, 1985;Staudinger. Cornelius, & Baltes, 1989). A number of researchers have argued that the characteristics of tests differ for measures that do and do not show age-related decrement (e.g., Cornelius, 1984). One common observation is that ability tests that show increments or are stable with age involve more familiar content or skills than ability tests that exhibit age decrements with increasing age. For example, Anastasi (1968) suggested that 'Whether intelligence test scores rise or decline with increasing age in adulthood depends largely on what experiences the individual undergoes during those years and on the relationships between these experiences and the functions covered by these tests" (p. 294). Likewise, Botwinick (1977)argued that verbal ability tests apparently measure "the manipulation of familiar material in familiar ways" whereas nonverbal or performance ability tests measure "the manipulation of unfamiliar materials, perhaps in unfamiliar ways" (p. 588). More recently, Denney (1982)proposed that abilities that are relevant to adults' experiences and "optimally exercised" may remain relatively constant in adulthood, whereas "unexercised abilities" may decline as a function of disuse and lack of practice. These perspectives raise fundamental questions about the nature of intelligence in adulthood and its assessment. More than thirty years ago, Demming and Pressey (1957)issued a critique of
Everyday Cognlttue Abflftfes
413
traditional methods of ability assessment in research on intellectual aging. They argued that tests commonly used in research were inappropriate for the assessment of intellectual functioning in adulthood because many tests were adapted to a large extent from instruments designed for children and youth to predict academic performance. They suggested that adult ability assessment should focus on the skills and knowledge that adults use to function in everyday living. Since their critique, other gerontologists (e.g., Schaie. 1978; Scheidt, 1981: Willis & Baltes, 1980) have reiterated concern about traditional measures with special attention devoted to issues of external and ecological validity. During the recent decade, there has been growing interest in everyday cognitive abilities a s evidenced by an increasing number of investigations examining everyday or practical problem solving, practical intelligence, a s well as cognitive skills in particular domains of everyday functioning (e.g., Cavanaugh, Kramer, Sinnott, Camp, & Markley, 1985: Frederiksen. 1986; Kuhn. Pennington, & Leadbeater. 1983; Poon, Rubin, & Wilson. in press; Sinnott, 1989a: Sternberg & Wagner, 1986). Much of the recent work h a s originated from contextual models of intellectual development (e.g., Berg & Sternberg. 1985: Ceci. 1990: Dixon & Baltes, 1986; Labouvie-Vief. 1985). Although there is some diversity in how contextual theorists define intelligence, contextual models seem to concur in viewing intelligence a s consisting of the mental activities that enable individuals to adapt successfully to their environment (e.g.. Berg, 1989; Labouvie-Vief. 1985: Staudinger et al.. 1989; Sternberg, 1985a). The purpose of the present chapter is to review some recent work in the study of everyday intelligence and cognition in adulthood and old age. As is the case for other domains of psychological inquiry, researchers have adopted both implicit and explicit approaches in the study of everyday intelligence. In the first section, we address the question of what intelligence is by examining people’s implicit theories or mental representations of the concept of intelligence. In doing so, our aim is to elucidate the attributes that people believe characterize intelligent functioning with special attention to such functioning in everyday contexts.
414
Cornellus
Moreover, implicit theories of adult intellectual development highlight the central role of everyday intellectual competence in defining intelligence during the latter part of the life span. In the second section, research conducted in our laboratory examining everyday cognitive abilities in adulthood and old age from an implicit approach is reviewed. This research has focused on adults' self-conceptions of everyday intelligence and its relation to objective test performance, and the correspondence between people's ideas about the characteristics associated with exceptional performance on ability tests and their relation to psychologists' conception of everyday intelligence. In the third section. an explicit approach to the assessment of everyday cognitive abilities is described. Results from two experiments we have performed to examine everyday problem solving and adults' knowledge about adaptive responses to stressful situations are discussed as illustrative of two key facets of everyday intelligence. The findings demonstrate differences and similarities between everyday cognitive abilities and traditional measures of verbal and problem-solving abilities and emphasize the potential for growth in everyday cognitive abilities during adulthood. IMPLICIT THEORIES OF INTELLIGENCE
Implicit theories are concerned with people's ideas and beliefs about psychological functioning, and such theories reflect how people construct and mentally represent psychological concepts. Explicit theories, on the other hand. involve scientists' constructions of concepts to describe or explain psychological phenomena. Implicit and explicit theories differ in the type of data one views as the basis for examining psychological functioning. In implicit theories, the data of interest focus on people's communications about their beliefs regarding the psychological construct of interest. By contrast, explicit theories are based on people's performance on tasks or tests that are presumed to measure psychological functioning. Implicit theories are important for understanding psychological constructs because explicit theories frequently are derived from implicit theories of a construct. Particularly at
Everyday CognUiw Abilities
415
initial points in inquiry, implicit theories provide a framework for the development of explicit theories because they aid in articulating some basic conceptual dimensions for investigation (e.g., Sternberg, 1985a). For example, in the domain of personality, people’s implicit theories have been useful in developing a taxonomy of dimensions (e.g.. McCrae & Costa, 1985). From a contextual perspective, implicit theories are significant because they help clarify how people in a given sociocultural context define a particular psychological construct. Prototypic Conceptions of Everyday and Academic Intelligence The concept of intelligence is frequently used in social discourse to describe a major component of human personality and functioning (e.g., McCrae & Costa, 1985). Natural language provides a means for communication about intellectual characteristics with descriptors available for describing desirable (e.g., logical, knowledgeable. wise) and undesirable (e.g., confused, absentminded, forgetful) aspects of cognitive functioning (e.g., Heckhausen, Dixon, & Baltes. 1989). People’s judgments about intelligence can reflect different foci depending on whether a person evaluates one’s own or another person’s abilities (e.g.. Kihlstrom & Cantor, 1984). In either case, the judgment one reaches about how intelligent an individual is may depend on the extent to which an individual’s behavior resembles an ideal or prototypic concept of an intelligent person (e.g.. Neisser. 1979). From this perspective, different individuals might be viewed to be of comparable intelligence but could resemble an ideal prototype along different dimensions. Despite the importance of the concept of intelligence in self and interpersonal evaluations, relatively little systematic research has been conducted on the content or structure of people’s prototypic conceptions of intelligence. Recently, Stemberg and his colleagues (Sternberg, 1985b: Sternberg & Berg, 1987; Sternberg, Conway. Ketron. & Bernstein. 1981) reported a series of investigations exploring people’s implicit theories of intelligence. In their initial work (Stemberg et al. 1981).people were surveyed about behaviors they thought were indicative of intelligence. Afterwards, experts
416
Cornellus
(i.e, psychologists conducting research on intelligence) and laypersons (Lealadults without formal training in psychology) rated the extent to which .each of these behaviors characterized their ideal concepts of "everyday intelligence" and "academic intelligence". Sternberg et al. (1981) found some important differences between these concepts of intelligence that appear to reflect a sensitivity or awareness to the context-specific nature of intelligent functioning, Experts and laypersons endorsed similar ideas about the behaviors characterizing these concepts, but both groups attributed different characteristics to everyday and academic intelligence. For example, when the profiles of attributes characterizing each of these concepts were examined. experts' and laypersons' judgments were quite similar about behaviors that were most and least characteristic of everyday intelligence (r = .81) as well as academic intelligence (r = .89).They also made a meaningful distinction between conceptions of intelligence in everyday and academic settings. The profiles for everyday and academic intelligence showed only a moderate degree of similarity within each group (for experts, r = .43: for laypersons, r = .45). Thus, the behaviors that people believe are most characteristic of a n intelligent person in everyday settings dif€er from those they think characterize exceptional intelligence in an academic context. Differences in the content of people's mental representations of everyday and academic intelligence were clarified by factor analyses of the ratings for these concepts. Analyses of experts' ratings for everyday intelligence revealed three dimensions involving Practical Problem-Solving Ability (e.g., "seeks explanations and causes", "sees individual elements in their overall context", "perceives implied assumptions and conclusions"), Practical Adaptive Behavior (e.g., "able to cope with crises", "acts in a practical manner", "sizes up situations well"), and Social Competence (e.g., "displays common sense". "able to cope with everyday environment", "has good intuitions"). By comparison, analyses of experts' ratings for academic intelligence indicated dimensions of Problem-Solving Ability (e.g., "solves problems well". "makes good decisions". "gets to the heart of problems"). Verbal Ability (e.g., "is verbally fluent", "reads widely".
Everyday Cognitive Abilities
417
"communicates ideas well"), and Motivation (e.g., "displays dedication and motivation in chosen pursuits". "displays persistence", "studies hard'). These results suggest that the ability to solve problems is an important dimension that underlies concepts of both everyday and academic intelligence. The relative salience of other types of behaviors differs for these two concepts. In particular, experts believe that being able to adapt successfully to crises and to engage in competent interpersonal functioning are more characteristic of an ideally intelligent person in an everyday context, whereas verbal abilities and heightened motivation to acquire knowledge are more salient for intelligent functioning in academic settings. Implicit Theories of Adult Intellectual Development Because there is a shift in the contexts that adults encounter a s they complete their formal education in youth and then experience transitions in work and family roles, people's ideas about the nature and meaning of intelligence might be expected to change in ways that reflect such transitions. Thus, an important question is whether people believe that intelligence is manifest differently as individuals move through the life course. In the present section, three types of evidence are reviewed: (a) adults' beliefs about changes in psychological functioning during adulthood, (b) prototypic conceptions of intelligence across adulthood, and (c) adults' perceptions of changes in their own cognitive abilities. In general, findings from this research show that people believe that abilities reflective of everyday intelligence assume increasing importance, particularly in middle age and later life. Results emerging from research on people's conceptions of psychological change in adulthood indicate that adults perceive adult development to involve multidirectional changes in which both gains and losses in functioning coexist (e.g., Heckhausen & Baltes. 1989; Heckhausen et al.. 1989). Heckhausen et al. (1989) asked young, middle-aged, and older adults to rate an extensive set of psychological attributes on three dimensions: (a) the extent to which an attribute increased with age, (b) the desirability of the
418
Cornelius
attribute, and (c)the ages at which change was expected to begin and end. The dominant finding was that people perceived an increasing number of losses (i.e.. increases in undesirable attributes) and a decreasing number of gains (Lee,an increase with age in desirable attributes) for most domains of psychological functioning across adulthood. For example. there was a strong negative correlation (r = -.68)between the perceived desirability of attributes and the expected age of onset. Thus. increasing mean ages for the beginning of developmental change were associated with fewer and less desirable changes and more and more undesirable changes. Moreover, results of this study showed substantial consensus among adults from different age groups especially about their expectations of the desirability and timing of change. Focusing only on the results for the intellectual characteristics examined in this study, the dominant flnding indicated that people believe that increases in undesirable attributes (e.g., forgetful, slow, overcautious) begin in later adulthood, whereas increases in desirable attributes (e.g., intelligent, logical, knowledgeable) begin in early adulthood. There were some clear exceptions to this general trend, however. Heckhausen et al. (1989)also found that people believe that some desirable intellectual attributes--for example, being experienced, knowledgeable about human nature, or wise--increase with age in adulthood. In addition, people expected increases on such attributes to be most likely to begin in middle age and to continue until some time in old age. Evidence from the investigation of implicit theories of adult intellectual development also shows that people think everyday intellectual abilities become more salient in adulthood. Sternberg and Berg (1987)examined people's prototypic concepts of intelligence for early adulthood, middle age, and late adulthood. In their study, adults. who ranged in age from 26 to 85. rated how likely it would be for individuals of average and exceptional intelligence at ages 30. 50, and 70 to display behavioral characteristics that had been nominated as indicative of adult intelligence. Data were factor analyzed for each target age. Results revealed three factors for each target age. The factors for target age 30 were: (a) Novelty in Problem Solving, (b) Crystallized
Everyday Cognitive Abilities
419
Intelligence, and (c) Everyday Competence. The factors for target age 50 were: (a) Novelty in Problem Solving, (b) Everyday Competence, and (cl Social Competence. The factors for target age 70 were: (a) Composite Fluid and Crystallized Intelligence, (b) Everyday Competence, and (c) Cognitive Investment. The findings suggest that the behavioral attributes people use to characterize exceptional intelligence for middle-aged or older adults differ to some extent from those characterizing intellectual functioning for younger adults. In particular, Everyday Competence was a significant dimension underlying people's prototypic concepts for each target age, but the variance accounted for by this factor was greater for older than younger prototypes. Substantively, the dimension of Everyday Competence was identified by behaviors involving adapting well to the environment, being perceptive about people and things, and adjusting to difficult life situations (Sternberg & Berg. 1987). Everyday competence may assume increasing importance in people's ideas about intellectual functioning for older ages because of the overlap between people's concepts of intelligence and wisdom (e.g., Sternberg, 1985b). In particular, intelligent functioning in middle or late adulthood may be manifest by a person who has acquired from experience an elaborate system of knowledge about life. Such knowledge may be reflected in an appreciation of uncertainties that occur in life course development and a sensitivity to contextual influences that moderate and shape pathways during adult life Ie.g., Baltes & Smith, in press). If this knowledge is implemented, it may be reflected in sound, effective. or practical judgments about the conduct of life particularly during transitions involving normative or non-normative developmental tasks (e.g., Smith, Dixon. & Baltes. in press). In addition to ideal conceptions of adult intellectual development, research examining older adults' retrospective evaluations of changes in their own intellectual functioning suggests that people perceive an improvement in everyday cognitive abilities. Williams. Denney. and Schadler (1983)interviewed older adults and asked them whether they had experienced changes in their memory and problem-solving abilities. They found that the majority of adults believed that their memory abilities had
420
Cornelius
declined with age. By contrast, they also found that most older adults believed that their abilities to think, reason, and solve problems had improved with age. The perceptions expressed by older adults contrast shaiply with much research on problem solving that has consistently indicated age-related declines in performance on a number of problem-solving tasks (e.g., Denney, 1982; Reese & Rodeheaver, 1985). Further probing, however, revealed that the majority of older adults referred to their ability to solve everyday and financial problems rather than to the types of problems used by researchers in traditional problem-solving tasks. Similar evidence was obtained by Birren (1969) in an interview study of adults about decision making. He found that adults perceived changes in the strategies they used in making decisions in their work lives (also see Streufert & Streufert, 1978). The interviews indicated that changes in solving everyday problems might occur because people defined their problems dLfferently resulting from the additional experience they had accumulated from early to middle adulthood. For example, many respondents noted that problems had become easier to deal with because the intensity of their emotional reactions to situations had diminshed. In addition, the internews revealed that people perceived that their goals had changed and might continue to change in the future. Furthermore, adults expressed the belief that their strategies in making decisions were influenced by efforts to compensate for psychological changes accompanying aging. In combination, research on people's implicit theories of intelligence reveals four key findings. First. adults with and without formal training in psychology hold similar ideas about the meaning of intelligence. Although there are some daerences in the content and structure of these concepts among various groups, there is also a rather striking degree of agreement about the attributes people believe are indicative of exceptional intellectual functioning. Second, people's conceptions of intelligence are multifaceted and reflect multiple dimensions of intellectual functioning. Problem-solving and verbal abilities emerge a s salient dimensions in experts' concept of academic intelligence, and these dimensions seem to correspond well to abilities that have been identified in several explicit psychological theories, such as
Everyday Cognitive Abilities
421
dmerential theories of intelligence (e.g., Cattell, 1971; Horn. 1982; Thurstone, 1935). Third, people's conceptions of intelligence also appear to be richer in scope than the domains of abilities represented in most explicit theories of intelligence (however. see Gardner, 1983; Sternberg. 1985a). In particular, adults accord greater recognition to practical and social cognitive abilities especially in their beliefs about everyday intelligence. Practical intelligence is marked by a n interest and awareness in the immediate environment and world at large, the ability to appraise situations well, and skill in devising means to attain one's goals. Social competence is associated with effective interpersonal behavior, acknowledgment of one's errors, engaging in socially appropriate behavior. and displaying concern about social issues. Finally, practical and social cognitive abilities appear to be central to people's ideas about changes in adult intellectual competence, especially changes in functioning during middle age and later adulthood. These ideas are reflected in the expectation that there is a potential for growth extending into old age in everyday intellectual abilities, and older adults' perception of improvements in their ability to solve practical problems and make decisions. EVERYDAY COGNITIVE ABILITIES: AN IMPLICIT APPROACH Implicit theories of intelligence. especially findings concerning people's prototypic conceptions (Sternberg et al.. 1981).influenced our initial efforts to examine everyday cognitive abilities. This research was directed toward two major issues. One issue has focused on the possible growth of everyday cognitive abilities in adulthood suggested by people's implicit theories a s well a s theoretical models of adult cognitive development (e.g., Dixon & Baltes. 1986: Labouvie-Vief, 1982; Schaie. 1977/ 1978). The second issue centered on the question of similarities and dflerences between everyday cognitive abilities and traditional abilities. In the following sections, we describe the results of two experiments investigating everyday intelligence from an implicit approach. In the first study, we examined adults' self-conceptions of intelligence
422
Conzellus
and their relation to performance on standard ability tests. In the second experiment, an investigation of the kinds of ability tests people believe are indicative of everyday intelligence is reported. Experiment 1: Self-conceptions of Intelligence Neisser (1979) argued that a person may be considered intelligent to the degree that his or her behavior corresponds to attributes that characterize a prototypic or ideal concept of intelligence. Following this orientation, Sternberg et al. (198 1) examined the relation between adults' self-conception of behavior and experts' prototypes of everyday and academic intelligence. They found that people's self-conceptions resembled both prototypes. but they were more similar to experts' characterization of everyday than academic intelligence. In addition, performance on a verbal measure of intelligence was more highly related to people's self-conception of academic intelligence (r = .56) than to their self-conception of everyday intelligence ( r = .45). The prototype resemblance paradigm was adopted in our work (Cornelius et al., 1989) to explore adults' self-conceptions of intelligence. The experiment had two major objectives. The first was to determine whether the correspondence between adults' beliefs about their abilities and prototypic concepts of academic and everyday intelligence were related to age and education-two factors that have shown considerable importance in the analysis of objective test performance. The second objective was to investigate relations between performance on standard ability tests and adults' self-conceptions of their intellectual functioning. Based on arguments (e.g., Willis & Baltes. 1980)that traditional tests are less representative of everyday than academic kinds of cognitive abilities. it was expected that objectively measured performance would be less related to adults' self-conception of everyday intelligence. In the study, young, middle-aged, and elderly adults between the ages of 20 and 89 were administered an abridged version of the questionnaire compiled by Sternberg et al. (1981). Participants rated how characteristic each behavior was of themselves on a seven-point scale (1 = highly uncharacteristic, 7 = highly
Everyday Cognitive Abilities
423
characteristic). Self-concept measures of everyday and academic intelligence were then obtained by computing correlations between each individual's response pattern and experts' profile of ratings for prototypes of everyday and academic intelligence. The possible range of scores was from - 1.OO to + 1.OO with higher positive scores indicating closer resemblance between an individual's self-concept and the pattern of qualities characterizing an ideal prototype. In general, participants' self-conceptions were more similar to experts' concept of everyday intelligence (M= .31.SD = .2 1) than academic intelligence (M= .18, SD = .21). Individual differences in self-conceptions of intelligence were quite large. however: For everyday intelligence, scores ranged from -.21 to .71.and for academic intelligence, scores ranged from -.22 to .62.
Relations between self-conceptionsof intelligence, age, and education. To address the first objective of the experiment. analyses were conducted to examine the relation between age, education, and the self-concept measures of everyday and academic intelligence. The findings are depicted in Figure 1. Mean scares on the prototype similarity measures of everyday and academic intelligence are shown for three age groups (Young adults, M age = 25.6 years; Middle-aged adults, M age = 49.4 years: Old adults, M age = 70.7 years) and for adults with lower (High school diploma or less education) vs. higher (More than a high school diploma) levels of education. The findings shown in Figure 1 emphasize that the extent to which age and education are related to self-conceptions of intelligence differs for self-concept measures of everyday and academic intelligence. In particular, age was not related to selfconceptions of everyday intelligence, whereas self-conceptions of academic intelligence declined with increasing age. In addition, self-conceptions of everyday intelligence were somewhat higher for adults with higher levels of education, but self-conceptions of academic intelligence were much higher for adults with higher levels of education.
Cornelius
424
0.4,
IYOUNG MIDDLGAGEI)
HOLD
EVERYDAY ACADEMIC SELF- CONCEPTION OF INTELLIGENCE
1 LOW EDUCATION 1 HIGH EDUCATION
EVERYDAY ACADEMIC SELF- CONCEPTION OF INTELLIGENCE
Figure 1. Self-conceptions of everyday and academic intelligence by age and educational background.
Eueryday Cognitive Abilities
425
Regression analyses performed with age and education as predictors of self-conceptions of intelligence clarified these findings. Results showed that these variables accounted for less variance in self-conceptions of everyday intelligence (R2=.12) than academic intelligence (R2= .34).For everyday intelligence, age was not a signlficant predictor [D = .02) but education was significant [D = .34,p < .01). For academic intelligence, both age CD = -.30.p < .01) and education (D = .54,p < .01) were significant predictors. In addition to the prototypic resemblance measures, a factor analysis was conducted to examine dimensions underlying people's self-conceptions of intelligence. Four factors were extracted with eigenvalues greater than 1. The first factor was indicative of problem-soluing ability. It included high loadings on items associated with objectivity, quality of problem solving, and task persistence. The second factor was associated with uerbal intelligence and knowledge. It was loaded highly by behaviors indicative of communicative competence, verbal facility, and curiosity. The third factor comprised behavioral qualities associated with practical intelligence and social competence. It was loaded highly by variables indicative of common sense, acceptance of social conventions, ethical judgment, and interpersonal competence. The fourth factor represented facets of motivation It was loaded highly by variables such as self-appraisal. goalorientation, and achievement. The relationship between age and self-conceptions of intellectual functioning differed depending on the dimension examined. Figure 2 displays the main results. There was an increase with age in self-assessments of practical intelligence ( r = .32). By contrast, participants' age displayed negative but insignificant relationships with self-concept measures of problem solving (r = -.08), verbal intelligence (r = -.04), and motivation ( r = .02). Self-assessments on these dimensions also showed differential relationships with participants' educational background. Education was positively correlated with self-assessed problem solving (r = .25) and verbal intelligence (r = .32).but it was not related to self-assessed practical intelligence (r = -.04)or motivation ( r = .06).
Cornelius
426
.2 4
w
.
0 .I. U
c n .
cr:o
0
i - . -.14
L
4
w
L
.
-.2.
SOLVING
. -.34
INTELLIGENCE YObG
-
MIDDLE-AGED A G E
PRACIlCAL INTELLIGENCE
8 MOTIVATlON
&
G R O U P
FYgure 2. Assessments of problem solving, verbal intelligence, practical
intelligence, and motivation by age.
These findings are partially consistent with previous literature suggesting that people perceive an increase in everyday cognitive abilities during adulthood. Self-assessments on the dimension of practical intelligence showed an increase with age. However, age was not related to the prototype similarity measure of everyday intelligence. These findings may occur because prototypes of intelligence are multifaceted and reflect several dimensions. Correlations between prototypic measures and self-ratings (i.e., factor scores) on specific dimensions showed that self-conceptions of everyday intelligence were positively correlated not only with self-ratings of practical intelligence and social competence ( r = .60) but also self-ratings of problem solving ability ( r = .54). verbal intelligence and knowledge ( r = .50),and motivation ( r = .38).By comparison, self-conceptions of academic intelligence were unrelated to self-ratings of practical intelligence (r = -.lo), but positively related to ratings on the dimensions of problem solving ( r = .64). verbal intelligence ( r = .61)and motivation (r = -21).
Everyday Cognitive Abilities
427
Relations between self-conceptionsof tntellfgence and ability test performance In order to examine the relation between self-concept measures of intelligence and objective test performance, seven ability tests were administered to participants. Nonverbal tests included measures of inductive reasoning (Letter Series. Blieszner et al., 1981; and Letter Sets. Blieszner et al., 1981) and logical classification (Figural Relations Matrices, Plemons, Willis & Baltes. 1978). Verbal measures of analogical reasoning (Verbal Analogies: Guilford, 19691, vocabulary [Verbal Meaning: Thurstone, 1962). practical judgment (Comprehension subtest of the WAIS, Wechsler, 1955). and social judgment (Social Situations test, Horn & Cattell. 1966)were also administered. Table 1 lists correlations between objective test performance and self-concept measures of intelligence. The first two columns show that objectively measured abilities are significantly related to self-concept measures of everyday and academic intelligence, but the correlations are consistently higher with academic intelligence, The only exception to this general pattern occurs for the Social Situations test which shows a higher correlation with self-conceptions of everyday intelligence than it does with academic intelligence. The correlations between objective performance and self-conceptions are comparable with those reported in other research (e.g.. Cornelius & Caspi. 1986: McCrae & Costa, 1985; Sternberg. 1985b; Sternberg et al., 1981; Willis & Schaie, 1986). In the third through sixth columns of Table 1, correlations between ability performance and self-ratings on specific ability dimensions are listed. As expected, self-rated problem-solving ability was positively related to performance on each test. Likewise, self-rated verbal intelligence was significantly related to performance on five of the seven ability tests. Objectively measured performance, however, was not related to self-ratings of practical intelligence or motivation. Many of the ability tests included in this study were not designed to assess practical abilities or social competence. It was surprising, however, that self-assessments of practical intelligence
Table 1 Correlations Between Self-conceptions of intelligence and Ability Performance Ability Test performance
Self-Concept Variable Everyday Academic Problem Verbal Practical intelligence intelligence solving intelligence intelligence motivation
Letter Series
.28**
.46**
.22*
-18
-.16
-.12
Letter Sets
.w*
,42**
.26**
.22*
-.08
-.03
Matrices
.42**
.52**
.37**
.27**
- .07
Verbal Analogies
.37**
-55.'
.37**
.30**
-.15
-.0f3
Verbal Meaning
.41**
.49**
.39*
.36**
.o1
-.10
Comprehension
.36**
.a**
.37**
.30**
.05
-.01
Social Situations
.29**
.18
.19*
.06
-01
-.04
**
pc.01
* pc.05
-08
Everyday Cognltfue Abilltles
429
were unrelated to performance on either the Comprehension or Social Situations test. In the Comprehension test, an examinee is required to respond to a question (e.g.. Why are child labor laws needed?), including in the response all information that is needed to answer it adequately. Wechsler (1958) suggested that the Comprehension test assesses practical judgment, common sense, and the ability to evaluate past experience. In the Social Situations test, an examinee is presented with an interpersonal dilemma and chooses from among four alternative responses the one that provides the best solution. The Social Situations test has been described a s an indicator of the ability "experiential evaluation". which is characterized by maktng good judgments and using social conventions in making decisions (e.g.. Horn & Donaldson. 1980). If these interpretations of the tests are correct, the negligible relations between performance and self-rated practical intelligence could result because adults do not accurately assess their practical intelligence. Perhaps the feedback a person receives about performance in everyday situations is too ambiguous for a n individual to form accurate estimates of their practical abilities. It is also possible, however, that these tests tap only some attributes identified with adults' self-conceptions of their practical intelligence. Therefore, we conducted an experiment to examine people's conceptions of the qualities assessed by ability tests and their relation to implicit theories of intelligence. Experiment 2: Implicit Theories of Intelligence and Ability Tests Neisser (1979)proposed that omnibus scales of intelligence (e.g.. Stanford-Binet, WAIS) may be relatively successful in the assessment of intelligence because different tests included in the battery capture different facets of intelligence. Any ability test, by itself, is unlikely to be a n adequate indicator of intelligence. Intelligence is a multidimensional construct, and a single ability test is likely to assess only some of the qualities associated with the construct. For example, the prototypic conception of academic intelligence emphasizes both verbal and problem-solving abilities. Performance on tests of both kinds of abilities may provide a reasonable approximation to academic intelligence. Indeed, in
430
Cornelius
research on adult intelligence (Schde. 1979).a composite index of educational aptitude has been formed from performance on verbal comprehension (i.e.. Verbal Meaning) and inductive reasoning tests (Lea,Letter Series). The purpose of Experiment 2 was to investigate people's beliefs about the qualities characterizing exceptional performance on ability tests designed from an explicit approach and their relation to experts' prototypic conceptions of everyday and academic intelligence. Based on findings from research examining the relations between objective test performance and self-conceptions of intelligence, it was expected that people's characterizations of most ability tests would display greater correspondence to a prototypic conception of academic intelligence, because most ability tests emphasize problem-solving or verbal abilities rather than practical intelligence or social competence. Therefore, in addition to traditional tests of fluid (i.e.. problem solving) and crystallized abilities, tests we have developed to assess everyday cognitive abilities were also included a s stimuli in the study. Tests of fluid abilities included Letter Series (Blieszner et al.. 1981).Letter Sets (Blieszner et al., 1981) and Figural Relations Matrices tests (Plemons et al.. 1978). Two tests involved verbal reasoning (Verbal Analogies: Guilford. 1969) and verbal comprehension (Verbal Meaning; Thurstone. 1962). Finally, four tests included content relevant to practical intellectual abilities: the Everyday Problem Solving Inventory (Cornelius & Caspi. 1987), Stressful Situations Questionnaire (Cornelius & Rose. 1987).Social Situations test (Horn & Cattell. 1966),and Comprehension subtest of the WAIS (Wechsler. 1955). The format of these tests is described in more detail elsewhere (Cornelius et al., 1989). A sample of college students was randomly assigned to rate one of the nine ability tests. They were told to read the instructions for a test and were allowed t o examine practice and test problems included in it. Identifying information about the test (e.g.. the title) was removed. After inspecting the test, participants were instructed to characterize the qualities they believed would be displayed by a person performing extremely well on the test using items from an abridged version of the questionnaire developed by Sternberg et al. (1981)on people's conceptions of intelligence. Subjects rated
Everyday Ccgnitlw Abilities
431
exceptional performance for a test on these items using a sevenpoint scale (1 = Very Uncharacteristic. 7 = Very Characteristic). Three major findings emerged from analyses of these data. First, contrary to our expectations, ratings for most of the ability tests displayed greater similarity to the prototype of everyday than academic intelligence. Prototypicality measures were computed to assess the degree of similarity between ratings of the ability tests and prototypes of everyday and academic intelligence. Prototypicality measures were obtained by correlating each participant's ratings of a test with the pattern of ratings characterizing experts' concepts of these two types of intelligence. Figure 3 displays mean scores on the prototype similarity measure for each test. The prototype of everyday intelligence showed its largest correspondence with two tests developed in our work to assess practical and social abilities (i.e., Everyday Problem Solving Inventory, Stressful Situations), followed by traditional measures of practical and social judgement (Le.. Social Situations, Comprehension) and fluid ability tests (Letter Series, Letter Sets, Matrices), and least correspondence with verbal ability tests (e.g., Verbal Analogies. Verbal Meaning). By contrast, attributes associated with academic intelligence showed the greatest resemblance to ratings of tests of verbal ability, followed by ratings of fluid ability tests, and the least resemblance to ratings of tests involving practical abilities. A second key finding was that different characteristics of intellectual functioning were attributed to exceptional performance on different ability tests. The data were first factor analyzed to identify a parsimonious set of dimensions underlying judges' evaluations of exceptional performance on the test stimuli. Four factors were extracted with eigenvalues greater than 1. The factor solution that we obtained was similar to results of factor analyses of people's prototypic conceptions of intelligence (Sternberg et al., 1981) and people's self-conceptions of intelligence described in the previous section. The first factor was associated with attributes indicative of practtcal problem-solufng ability. It included high loadings on items associated with decision making, reasoning, and quality of
Cornellus
432
EVERYDAY INTELLIGENCE
0 ACADEKIC INTELLIGENCE
-.1 EPSI
STRS SOCS COW LSER LSET MAT VANL VMNG T E S T
S T I M U L U S
F'fgure3. Resemblance between prototypes of intelligence and test rating profiles. Test titles are abbreviated EPSI (Everyday Problem Solving Intentory), STRS (Stressful Situations), SOCS (Social Situations), COMP WAIS Comprehension), LSER (Letter Series), LSET (Letter Sets), MAT (Matrices),VANL (Verbal Analogies), and VMNG ('Verbal Meaning).
The second factor was indicative of verbal intelligence and knowledge. It was loaded highly by behaviors involving reading, verbal facility, and achievement. The third factor comprised behavioral qualities associated with social Competence and practical Intelligence. It was loaded highly by variables indicative of interpersonal competence, ethical judgment, and giving advice. The fourth factor represented several indicators of cognitfue style. It was loaded positively by variables such as curiosity, thoughtfulness, and style of problem solving. Comparisons between tests rated on these four dimensions revealed three major patterns (see Figure 4). Tests of practical abilities (Lee, Everyday Problem Solving Inventory. Stressful Situations, Social Situations, Comprehension) were evaluated highly on behaviors indicative of practical problem solving a s well a s social competence. Fluid ability tests (i.e.. Letter Series, Letter Sets, Matrices) were judged most highly on characteristics associated with practical problem solving ability. Finally, verbal ability tests (Verbal Analogies, Verbal Meaning) were rated highly problem solving.
Everyday Cognitive Abilities
433
6.0.
i ii f f
5.5.
5.0.
4.5.
4.0.
L
1.o
PRACIICAI, PROBLEM SOLVING A VERBAL INTELLIGENCE I
EPSI
eCOGNITIVE STYLE
I
I
STRS
SOCIAL COMPETENCE
SOCS COMP LSER LSET T E S T
1
MAT
I
VANL VMNG
S T I M U L U S
Flgure 4. Ratings of test stimuli on dimensions of intelligence (after Cornelius et d.,1989). Test titles are abbreviated: EPSI (Everyday Problem Solving Inventory). STRS (Stressful Situations), SOCS (Social Situations), COMP ( WAIS Comprehension), LSER (Letter Series), LSET (Letter Sets), MAT (Matrices),VANL (Verbal Analogies),and VMNG (Verbal Meaning).
on facets of verbal intelligence and knowledge. Overall. the pattern of ratings appears to coincide with researchers' interpretations of the tests and suggest that different tests may tap different features of intellectual functioning. Finally, a third major finding was that a combination of tests yielded a better prediction of experts' prototypic concepts of intelligence than any single test. A profile was computed for each test by averaging judges' ratings on each item in the questionnaire. Intraclass correlation coefficients were computed to obtain a n index of the reliability of these profiles. There was a high degree of interjudge similarity in ratings (M intraclass coefficient = .85) with intraclass correlations ranging from .79 to .92 for the different tests. Regression analyses were then performed using the profiles of ratings for the ability tests as predictor variables. Criterion
434
Cornellus
variables in these analyses were the proffles of ratings for experts' conceptions of everyday and academic intelligence. Results showed that the pattern of characteristics associated with everyday intelligence was best predicted by the rating proffles of four tests: Everyday Problem Solving Inventory 1p = .37). Stressful Situations 1p = .25). Letter Series Cp = .28), and Verbal Meaning [D = .24, all p's c .Ol). Ratings on these four tests accounted for 73 percent of the variance (R = .85. p < .001). Academic intelligence was predicted best by ratings of the Verbal Meaning ID = .67) and Letter Series tests (J.3 = .25. p's c .001). and together these tests accounted for 61 percent of the variance (R = -78.p c .OOl). In summary, the findings of this experiment highlight similarities and differences between concepts of everyday and academic intelligence and their relation to people's perceptions of ability tests. These conceptions of intelligence are distinguishable because different facets of intellectual behavior are weighted differently in people's ideas about these concepts. Experts emphasize both verbal and problem-solving abilities in their concept of academic intelligence. and people's conceptions of tests of these abilities provide good predictors of this prototype. These abilities are also significant in the prediction of experts' concept of everyday intelligence. However, independent of these, tests of everyday problem solving and coping with stressful situations also contribute to the prediction of the prototype of everyday intelligence. In line with Neisser's (1979) argument, the results suggest that a combination of ability tests may be required to assess adequately the multifaceted attributes that people believe are characteristic of intelligence. Similar to our findings examining self-conceptions of intelligence and ability test performance, these results suggest that verbal and problem-solving abilities are important aspects of both everyday and academic intelligence. These abilities provide only a partial account of everyday intelligence, however, and need to be supplemented by measures that specifically tap features of social competence and practical intelligence (cf. Willis & Schaie. 19861. The lack of attention to these dimensions in explicit approaches to the study of intelligence may have resulted from the focus on
Everyday Cognittue Abllfties
435
academic performance as the primary criterion for evaluating the validity of ability tests. EVERYDAY COGNITIVE ABILITIES: AN EXPLICIT APPROACH In this section, two experiments that were conducted to examine performance on measures of everyday cognitive abilities are discussed. Based on findings concerning experts' prototypic conceptions of everyday intelligence (Sternberg et al., 1981).these measures were designed to assess two major facets of practical and social cognitive abilities-everyday problem solving and knowledge about adaptive behavior.
Experiment 3: Everyday Problem Solving in Adulthood and Old Age
Phase 1 The first phase of our research was devoted to the development of an instrument to assess everyday problem solving. The Everyday Problem Solving Inventory (Cornelius & Caspi. 1987) was constructed using a behavior-analytic model for assessing competence (Goldfried & D'Zurilla, 1969;also see, D'Zurilla & Nezu. 1982). The basic elements consisted of specifying problematic situations, enumerating possible responses to them, and evaluating the perceived efficacy of the responses. The Everyday Problem Solving Inventory consists of 48 hypothetical problem situations sampling six content domains- problems an adult might experience as an economic consumer, in deaIing with complex or technical informatfon. in managing a home, in resolving interpersonal conflicts with one's family members, in resolving conflicts with friends, or in conflict resolution with coworkers. The majority of situations in the Everyday Problem Solving Inventory are similar to ones included in situational taxonomies (Scheidt & Schaie, 1978).inventories of unpleasant events (Lewinsohn & Talkington, 19791,and hassles in everyday living (Kanner. Coyne, Schaefer, & Lazarus, 1981). A
436
Cornelius
sample problem from the information domain and alternative responses to it follow.
was returned because you misinterpreted the instructions on how tofill it out. A. Obtain more information on how to complete the form correctly. B. Try to figure out on your own what was wrong. C. Ask someone to fill out the form for you. D. Blame the company for not making the instructions more clear.
A complicated form you completed
For each situation, four possible response modes were devised based on previous studies of practical problem solving and coping with real-life stressors (e.g.. Billings & Moos, 1981; Denney & Palmer, 1981; Lazarus & Folkman. 1984). h-oblem-focused action (Response A to the sample problem) involves self-initiated overt behaviors to deal directly with a problem and its effects. Cognitive problem analysis (Response B) refers to intrapsychic or cognitive efforts to manage one's subjective appraisal of a situation, to understand it better, to solve the problem through logical analysis, or to reinterpret the situation from a dmerent perspective. Passfvedependent behavior (Response C) includes attempts to avoid or withdraw from a situation, the absence of self-initiated behavior to alter a situation. or actions involving dependence on another person to solve the problem. Finally, auoldant thinking and denial (Response D) includes attempts to control the meaning of a situation through cognitive avoidance, denial of the situation or of one's personal responsibility in it, selective attention to things other than the situation itself, or attempts to manage one's affective reaction through the suppression of one's emotions. Because contextual models of competence (e.g., Goldfried & D'Zurilla, 1969; Scheidt & Schaie. 1978: Sternberg. 1985a) suggest that particular modes of response are likely to be more effective in some situations than in others, we wanted to examine people's judgments about the quality of the responses for each of the problems in the inventory, Thus, in the initial phase of research, a sample of judges rated how effective they believed responses would
Everyday Cugnitiue Abilities
437
be as solutions to the problems described in the Everyday Problem Solving Inventory. In making their judgments, judges were instructed to consider factors such a s the likelihood that the response would solve the problem, that it would prevent or minimize future problems, and that it would enhance the person's self-esteem (see Fisher-Beckfield & McFall, 1982). Judges were given the inventory and rated each response on a 5-point scale ( 1 = Extremely ineffective or poor solution, 5 = Extremely effective or good solution). Analyses were performed examining the effects of situation and modes of response on judgments of response efficacy. In general, problem-focused action and cognitive problem analysis were judged to be more effective than either passive-dependent behavior or avoidant thinking and denial. These analyses, however, also revealed significant interactions between situation and mode of response for each domain of problems. In fact, the pattern of ratings among the four response modes diaered depending on the situation examined. For example, in one consumer problem in which a person did not have enough money to buy a gift for a friend's birthday, problem-focused action (M= 4.6) was rated as most effective, followed by passive-dependent behavior (M = 2.6) and cognitive probIem analysis (M = 2.4), and avoidant thinkfng and denial (M = 1.1) was rated as least effective. In another consumer problem describing a person who did not have sufficient time to purchase items prior to a vacation trip, however, cognitive problem analysis (M = 4.6) was fudged to be the most effective response, followed by avoidant thinking and denial (M= 3.9)and passive-dependent behavior (M= 3.7).and problem-focused action (M= 2.6)was rated as least effective. In line with contextual models of competence, these findings suggest that an important skill in solving everyday problems may be an individual's sensitivity to situational differences and flexibility in responding to situational nuances. A heuristic strategy for solving everyday problems that involves reliance on a general response mode (e.g., problem-focused action) may prove optimal in many situations, but it may be maladaptive in particular situations. Given the heterogeneity of problems that occur in any domain of everyday life, it seems unlikely that a
438
Cornellus
particular response mode would be uniformly effective across diverse situations (cf. Baltes & Smith,in press: Berg, 1989;Feifel & Strack, 1989;Mischel, 1984). Phase 2
In the second phase of the research, the Everyday Problem Solving Inventory was administered to young, middle-aged, and older adults. The second phase was conducted to examine age differences in everyday problem solving performance and relationships between performance o n the inventory and traditional tests of verbal and problem-solving abilities. In administering the inventory, participants were instructed to imagine themselves in the situations described and to rate the likelihood that they would act in each of the four response modes listed after each situation. Ratings were made on a 5-point scale (1 = Definitely would not do, 5 = Definitely would do). It is important to emphasize that in this format, participants did not simply select one response to describe how they would respond. Rather. because solutions to everyday problems often involve multiple actions, participants rated each possible response. Thus,for each problem domain, participants made four ratings for each of eight situations, yielding a total of 32 responses. Scores on the inventory were then derived by computing a correlation between each participant's ratings and judges' average ratings of response efficacy obtained in the first phase of the research. Separate correlations were obtained for each domain of problems, and an overall problem-solving score was computed across all problem domains. In addition, participants were given tests of verbal comprehension (i.e., Verbal Meaning; Thurstone. 1962)and inductive reasoning (i.e.. Letter Series; Blieszner et al., 1981). Table 2 shows correlations between performance on traditional verbal and problem-solving tests and the Everyday Problem Solving Inventory. Several aspects of these data are notable. First, the overall score on the Everyday Problem Solving Inventory was signiflcantly correlated with performance on Verbal Meaning and Letter Series tests. These correlations are similar to each other and
Everyday Cognitive AbUftles
439
Table 2
Correlations Among Verbal Meaning. Letter Series, and Everyday Problem Solving Performance ~~
Test variable
1
2
3
4
5
6
7
8
9
1. Verbal Meaning ---.37 ---2. Letter Series 3. Everday Problem Solving .27 .29 ---Total Score .30 .34 .86 ---4. I d o r m at io n .30 .26 .80 .57 --5. Family .24 .22 .86 -70 .62 ---6. Work .18 .30 .87 .68 .65 .72 ---7. Friend .16 .12 .82 .63 .64 .67 .68 ---8. Consumer 9 ~. . Home .12 .18 .81 .64 .63 .66 .6!5 .59 ---Note: Correlations are significant with a value of .18 (p < .05)and .23 (p < .01).
to the correlation between scores on the Verbal Meaning and Letter Series tests. Second, performance on the Verbal Meaning and Letter Series tests show similar patterns of correlations with scores for each problem domain. The highest correlation is with performance on problems involving information, and the lowest correlations are with consumer or home management problems. Third, correlations among scores from different problem domains are moderate to large. These correlations indicate that individual differences in performance are fairly consistent across the different domains of problems examined. Moreover. correlations among scores for the diMerent domains of everyday problem solving are much larger in magnitude than are their correlations with traditional ability tests. Similar to the results we obtained for people's conceptions of exceptional performance on ability tests, these findings suggest that the Everyday Problem Solving Inventory may assess facets of intellectual functioning
440
Cornelius
that are somewhat independent from traditional measures of verbal and problem-solving abilities. Regression analyses with age, education, and sex as predictors of performance on these measures provided further support for a differentiation among them. The findings showed that these demographic characteristics were less highly related to performance on the Everyday Problem Solving Inventory ( R = .24) than on the Verbal Meaning (R = .55) or Letter Series tests ( R = 50). For the Everyday Problem Solving Inventory, only the linear component of age was a significant predictor U3 = .19. p < .05). For the Verbal Meaning test, significant predictors included the linear component of age @ = .16, p c .05) and education @ = .53, p c .001). For the Letter Series test, the linear @ = -.31) and quadratic components of age 10 = -.32) and education U3 = .30,all p's c .001) were significant predictors. The regression lines displayed in Figure 5 show the predicted values of performance on the Everyday Problem Solving Inventory, Verbal Meaning, and Letter Series tests as a function of age. Scores on the three dependent variables were transformed to T scores to facilitate a graphic display of the results. The regression lines show similar developmental functions for the Everyday Problem Solving Inventory and Verbal Meaning test. Both functions show a modest increase with age. By contrast, the function for the Letter Series test shows an increase from ages 20 to 30.relative stability from ages 30 to 50, and then a decrease from ages 50 to 70. One explanation for multidirectional age differences in cognitive abilities has centered on the differential familiarity of different ability dimensions (e.g.. Cornelius. 1984). A similar hypothesis has been proposed to account for age differences on different measures of practical abilities. For example, Cardner and Monge (1977) found that maximal performance on a test of knowledge of modes of transportation was achieved by 30 year olds. whereas peak performance on a test involving knowledge about death and disease was achieved by 60 year olds. Changes in everyday cognitive abilities may be dependent on a person's present or prior experience with similar problems or on a person's degree of interest in a particular domain of everyday life (e.g.. Camp,
441
Eueryday Cognltfue AbIZltIes
EVERYDAY PROBLEM SOLVINGINVENTORY VEREALMEANJNGTEST
1 LElTER SERIES TEST
42 40
20
30
40
50
CHRONOLOGICAL
60
70
A G E
Flgure 5. Developmental functions for different ability tests (after Cornelius & Caspt, 1987). (Regression lines display predicted levels of performance as a function of chronological age. Predicted performance is measured in standardized Tscores.)
Doherty. Moody-Thomas. & Denney, 1989; Denney. in press: Gardner & Monge, 1977). In a separate Everyday Situations Questionnaire, participants rated how frequently they experienced the situations described in the Everyday FYoblem Solving Inventory using a 5-point scale (1 = Never, 5 = Frequently). Separate problem familiarity ratings were computed for each domain, and a total score was obtained across all problems. Analyses of age differences in problem familiarity showed that young adults (M = 2.82)reported experiencing the problems a s frequently as middle-aged adults (M= 2.76). but older adults (M= 2.59) reported less familiarity with them than younger age groups. Thus, although there was a n increase with age In performance on the Everyday Problem Solving Inventory, selfreported familiarity with the problems decreased with increasing age. Likewise, analyses examining age differences in problem familiarity and problem-solving performance for each of the six domains showed little convergence. Moreover, correlations between familiarity ratings and problem-solving scores for the
442
Cornellus
separate domains displayed little correspondence ( f s ranged from -.22 to .08). These results and those of other studies (e.g., Berg, 1989;Camp et al., 1989) indicate that familiarity with everyday problems is unrelated to people's skill or preference in solving them. There may be several explanations for these findings including the possibility that the indices of problem familiarity that have been examined may be insensitive to the characteristics of people's experiences that are most influential in everyday problem solving (see Cornelius & Caspi. 1987). Further research that distinguishes between different aspects of experience (i.e.. recent familiarity. vicarious experience, cumulative experience) will be required to clarify the role that each of them may play in accounting for changes in everyday problem solving abilities during adulthood. Experiment 4: Conceptions of Coping Although research on everyday problem solving is an important step in gerontological research on everyday intelligence, the problems investigated have typically sampled "hassles" adults experience in everyday life rather than problematic situations that seriously tax a person's adaptive capabilities. However, research on prototypic conceptions of intelligence suggests that a n ability to "cope with crises" (Sternberg et al., 1981) or "disastrous situations" (Sternberg & Berg, 1987)is an important characteristic associated with everyday intellectual competence in adulthood (cf. Flavell, 1970). Likewise, recent work emphasizes the importance of examining everyday cognitive abilities in situations involving a n integration of emotion and cognition (e.g.. Baltes & Smith, in press: Blanchard-Fields, 1986: Kramer. 1989: Labouvie-Vief, HakimLarson, & Hobart, 1987;Staudinger, 1988).Thus,we initiated work (Cornelius & Rose, 1987) on the development of a Stressful Situations Questionnaire to investigate the facet of practical adaptive behavior. The specific objective in this experiment was to examlne people's knowledge about strategies they thought would be adaptive in resolving stressful situations. Research relevant to this inquiry h a s emphasized both normative and individual difference perspectives in people's
Eoeryday Cognitive Abilities
443
knowledge about adaptive behavior. On the one hand, several studies suggest that there is a normative concept of adaptation. This is reflected in a rather high degree of consensual agreement among people about strategies that would be more or less effective in solving problem situations. For example, Sternberg and Soriano (1984) presented subjects with stories describing conflict situations arising in three domains: interpersonal. interorganizational. and international relations. Subjects rated how effective seven different strategies would be in resolving each conflict situation. Their findings showed consistent profiles of efficacy ratings for these strategies both within and across the domains examined. Likewise, McCrae and Costa (1986) investigated the perceived effectiveness of 27 coping mechanisms for stressful situations involving losses, threats, or challenges. Subjects rated whether the mechanisms had been effective for solving the problem or for reducing distress. Their results showed similar patterns of ratings for both of these criteria as well as consistent patterns across different categories of stressors. On the other hand, evidence also shows that there are systematic individual differences among people in their knowledge about effective strategies. Sternberg and Soriano (1984) reported that individual differences in personality and intellectual characteristics were predictive of differences among people's judgments about conflict resolution strategies. For instance, need for deference and inductive reasoning were negatively related to judgments about the efficacy of using physical coercion or economic pressure to resolve conflicts, but these variables were positively related to judgments about the efficacy of responses involving waiting to see whether the situation would improve or accepting and making the best of a situation. Likewise, McCrae and Costa (1986)examined the relation between personality and adults' use of coping mechanisms that were judged to be more or less effective. They found that neuroticism was correlated with the use of less effective coping strategies, whereas extraversion was correlated with the use of more effective strategies. The initial phase of our research with the Stressful Situations Questionnaire was conducted to examine whether adults' conceptions of effective coping may differ across adulthood and the
444
Cornelius
possible role of cognitive and personality variables in accounting for individual variability in their prototypical judgments about effective responding to stressful situations. Young (Mage = 30.5 years), middle-aged (Mage = 49.1 years), and older (M age = 68.9 years) adults were administered the Stressful Situations Questionnaire. and traditional tests of intellectual abilities and personality. The Stressful Situations Questionnaire includes 16 hypothetical situations involving losses, threats, and challenges. For each situation, responses modelled after strategies in the Ways of Coping checklist (Lazarus h Folkman, 1984) were listed as possible ways the character in the situation might respond to it. Subjects judged the effectiveness of each coping response as a solution to the problem using a five-point rating scale (1 = Very ineffective, 5 = Very effective). An example of a problem from the Stressful Situations Questionnaire and its associated response items follow.
Helen ONeUl went to the Department of Motor Vehicles to renew her driver's license. Much to her dismay, she failed the required eye examination even when wearing her corrective lenses. Helen became very upset at not being allowed to drive anymore. S h e was afiaid that she would lose a lot of the independence she was accustomed to due to the necessity of having to rely on other people for transportation to andfiorn places.
A. B. C. D.
Helen could feel badly that nothing can be done to improve her vision. Helen could wish that she could get over her fear of becoming dependent on others for transportation. Helen could phone her sister, Phyllis. and talk to her about her concerns about losing independence. Helen could try to accept her fear of becoming dependent but not let it prevent her from keeping up with her daily activities.
If adults' knowledge about coping is organized around a dominant conceptual prototype, profiles of ratings between
Everyday Cognitive Abilities
445
different judges should be highly similar. When we compared profiles of ratings averaged across various subgroups of judges in the sample, the findings showed a high degree of congruence. Simple correlations were computed between the patterns of ratings by subgroups to explore consistency in judgments a s a function of the age, educational background, and sex of judges. Comparisons between the pattern of ratings by young and middle-aged judges (r = .97), young and old judges (r = .93), and middle-aged and old judges (r =.95) revealed high consistency. Likewise. there was high correspondence between ratings by judges who had a high school diploma or less education and judges who had received more education beyond a high school diploma ( r = .97). Moreover, comparison of ratings by male and female judges displayed high similarity (r = .97). Finally, an alpha coefficient was computed to obtain an estimate of the reliability of judges' ratings across all possible split halves of judges. This coefficient was also extremely high (a = .99) suggesting that the profiles of ratings were quite similar even for arbitrarily formed subgroups of judges. Even though various subgroups in the sample displayed high consistency in their prototypicality judgments about coping. judgments of individuals did not uniformly correspond to the prototype of the group as a whole. A prototype similarity score was obtained by computing a correlation between each individual's profile of ratings and the profile of ratings averaged across all other judges in the sample. These scores could range between -1.00 and + 1 .OO with larger positive values indicating greater similarity between an individual's judgments and those of the rest of the group. The average prototypicality score was relatively high (M= .69). but there was also a wide range of individual daerences (SO = ,171. Analyses were performed to determine whether individual differences in adults' knowledge about coping were systematically related to intellectual and personality variables. Participants were administered standard tests of social intelligence (George Washington Social Intelligence Test: Moss, Hunt, Omwake, & Woodward, 19491, inductive reasoning (Letter Series: Blieszner et al., 1981).and verbal comprehension (Verbal Meaning; Thurstone. 1962). a s well as measures of Extraversion and Neuroticism
446
Cornelius
(Eysenck Personality Inventory; Eysenck & Eysenck. 1968). Relations between prototypicality judgments, intellectual performance, and personality are shown in Table 3. Simple correlations presented in the first column of Table 3 indicate that prototypicality judgments were positively correlated with performance on intellectual ability tests. As expected, the largest correlation occurred with the measure! of social intelligence, and judgments were associated to a lesser extent with performance on traditional tests of problem-solving and verbal abilities. In addition, prototypicality judgments about coping were negatively correlated with personality dimensions, in particular neuroticism. Because one might suspect that judgments about effective coping simply reflect a tendency toward socially desirable responding, the correlation between prototypicality judgments and responses to the Lie Scale of the Eysenck Personality Inventory was also examined. The correlation was negative but not significant. A regression analysis was performed with scores on the intellectual and personality measures as predictor variables, and the prototype similarity score a s the criterion variable. In combination, the intellectual and personality variables accounted for 46% of the variance in prototypicality judgments about coping ( R = .68, p c .OOl). Standardized regression coefficients from this analysis are listed in the second column of Table 3. Significant predictors included performance on the social intelligence and inductive reasoning tests and self-reported neuroticism and extraversion. Thus, individuals whose judgments about effective coping resembled the judgments of the group as a whole tended to display better performance on measures of social intelligence and problem solving and reported themselves to be more emotionally stable and introverted. These findings are compatible with both normative and individual difference approaches to the study of people's knowledge (i.e., beliefs and ideas) about adaptive behavior. Adults' judgments about effective solutions in resolving stressful situations appear to reflect a system of strategic or procedural knowledge that is quite similar across various social groups. In accord with other research (e.g., McCrae & Costa. 1986; Sternberg & Soriano, 1984). these findings suggest that people's conceptions of effective coping are
447
Everyday Cugnitiue Abilities
Table 3
Relations between Prototypicality Judgments and Intellectual and Personality Tests Test Variable
Social Intelligence Letter Series Verbal Meaning Neuroticism Extraversion Lie Scale
n
*
Simple correlation (r)
Standardized regression coefficient (p)
.59***
.47***
.35***
.18* .09 16*
a***
-.27**
-.12 -.14
-.
-.17* -.11
p<.Ool p < .01
p < -05
organized around a conceptual prototype. In general, responses involving approach strategies were judged to be more effective than responses involving avoidance strategies (cf. Roth & Cohen. 1986). For example, responses involving rational action, taking things one step at a time, or changing oneself to deal with the situation were rated a s being quite effective solutions in the situations presented, whereas responses involving intellectual denial, withdrawal, and passivity were rated a s being quite ineffective solutions. These judgments, of course, are likely to be speciflc to the problem situations examined. Further research is underway to explore differences in judgments as a function of both situation and response modes. The findings of this experiment also reveal significant individual differences among adults in the extent to which their judgments resembled a prototypical concept. Moreover, both intellectual and personality variables were significant in predicting individual variability. Additional evidence demonstrating the importance of individual differences in people's knowledge has been reported in research examining prototypicality
448
Cornelius
judgments about categories of personality traits and concrete objects (Beck, McCauley, Segal, & Hershey. 1988) Thus, the findings of this experiment and other research (e.g., Ceci & Liker. 1986; Hoyer. 1987; Wagner & Sternberg. 1986) underscore the importance of individual variability in adults' knowledge about various domains of everyday life. SUMMARY AND CONCLUSIONS In cognitive aging research, issues of ecological and external validity have been a recurrent concern reflecting skepticism about the appropriateness of traditional ability tests for the assessment of adult intelligence. Despite criticisms that traditional tests are inappropriate for assessing adult intelligence, little progress was made until recent years in addressing the more basic issue of the meaning of everyday intelligence and differentiating it from traditional or academic conceptions of intelligence. By linking implicit and explicit approaches, the research reviewed in this chapter clarifies the distinction between these concepts and presents some initial advances toward the assessment of everyday cognitive abilities. Implicit and explicit approaches are complementary means for understanding intellectual functioning. Some similarities emerge in comparisons of implicit and explicit approaches. Research based on an implicit approach has consistently identified problemsolving ability, verbal intelligence, and practical intelligence or social competence as salient dimensions. These dimensions emerge in analyses of people's prototypic conceptions of intelligence, their self-conceptions of intelligence, and their conceptions of exceptional performance on various ability tests. Problem-solving and verbal abilitles also seem to correspond well to ability dimensions that have been identlfied in research from an explicit approach. such as differential theories of intelligence (e.g., Cattell, 1971: Horn, 1978, 1982;Thurstone, 1935). Everyday and academic intelligence are multifaceted constructs encompassing a number of dimensions of intellectual functioning. Differences between these concepts occur because various dimensions of functioning are differentially associated with them.
Everyday Cugnitiue Abilities
449
In examining self-conceptions of intelligence. our findings revealed that people's self-assessments on dimensions of problem solving and verbal intelligence were positively related to prototypic resemblance measures of everyday and academic intelligence. However, self-ratings on a dimension indicative of practical intelligence and social competence were positively related to everyday intelligence but unrelated to academic intelligence. Similar results emerged from our analyses of people's conceptions of exceptional performance on ability tests. Ratings of problemsolving (i.e.. Letter Series) and verbal ability (i.e.. Verbal Meaning) tests emerged a s significant predictors of experts' prototypes for both everyday and academic intelligence. In addition to these, however, tests of practical abilities involved in solving everyday and stressful problems were significant predictors of everyday intelligence. These results suggest that although problem solving and verbal abilities are important for both kinds of intelligence, they are not the whole story. They provide only a partial account of everyday intelligence and need to be supplemented by measures that specifically tap features of practical intelligence and social competence. Differences also emerge in comparing implicit and explicit theories of intelligence. Practical and social cognitive abilities are important facets of everyday intelligence, but such abilities have received far less attention than verbal or problem-solving abilities in most explicit psychological theories of intelligence. Nevertheless, practical and social cognitive abilities have been recognized as salient features of intelligence in some recent explicit theories (e.g.. Cantor & Kihlstrom, 1987: Gardner. 1983; Sternberg. 1985a). Moreover, these abilities have been at the center of theoretical propositions about aspects of intelligence that may be acquired in adulthood. For example, Schaie (1977/1978)proposed three stages of adult cognitive development involving the application of one's intelligence to real-life decisions: Baltes and his colleagues (e.g.. Dixon & Baltes. 1985; Staudinger et al., 1989) explored the potential for growth in the pragmatics of intelligence with special attention to adults' knowledge about the pragmatics of life (Baltes & Smith, in press); and, Labouvie-Vief (1982, 1'985) outlined structural transformations in cognition that are sensitive
450
Cornelius
to pragmatic and contextual constraints on logical thinking (also see, Kramer, 1983:Sinnott. 1984,1989b). Findings from our research are generally consistent with these propositions. One finding clearly emerging from Experiment 1 as well a s research on implicit theories of adult intellectual development is the belief that practical abilities increase with age. In addition, our data showed that performance on the Everyday Problem Solving Inventory increased with age with older adults performing better than younger adults. Other research (e.g.. Camp et al., 1989:Capon, Kuhn. & Carretero, 1989;Demming & Pressey, 1957: Denney & Palmer, 1981;Denney. Pearce & Palmer, 1982; Gardner & Monge, 1977;Labouvie-Vief, et al., 1987)has found a variety of developmental trajectories for performance on measures of practical abilities with most studies indicating an improvement in performance from early adulthood through middle age. Dtfferences among studies do occur in the age at which maximal performance is attained and in the direction and magnitude of changes after middle age. Although evidence from our research and other studies (e.g., Berg, 1989:Camp et al.. 1989)has been unable to demonstrate clearly that differential familiarity or experience with problems accounts for age differences, further research using more refined assessments of adults' experience with practical problems is necessary before firm conclusions can be reached. An additional focal issue in our research and other studies has been the question of the relationship between performance on problem s o w , verbal, and practical ability measures. In general, correlations between performance on measures of practical and other intellectual abilities tends to be moderate to low (e.g., Camp et al.. 1989;Ceci & Liker. 1986 Ford & Tisak. 1983;Frederiksen. 1986 Wagner & Sternberg, 1986:see however, Willis & Schaie. 1986).In our studies, correlations of practical ability scores with a n inductive reasoning test have ranged between .12 and .35.and correlations with a verbal comprehension test have ranged between .12and .43. In addition, there is also evidence supporting the convergent validity of different assessments of practical abilities. For example, correlations among scores for different domains of problems sampled in the Everyday Problem Solving Inventory
Everyday Cognllfue Abilities
451
ranged between .57and .72(see Table 2). Likewise, prototypicality judgments about coping were most highly correlated with performance on a test of social intelligence (r = .59).Thus, these initial findings provide empirical support for a differentiation among practical, verbal, and problem-solving abilities. They suggest that practical ability tasks developed in our research assess different aspects of intellectual behavior and functioning than traditional ability tests. Further research using multiple indicators of practical, verbal, and problem-solving abilities is required to investigate the multidimensional structure of these abilities. The correlational evidence we have obtained provides limited support for a view of practical intelligence and social competence as aspects of crystallized intelligence (e.g., Dittmann-Kohli 81 Baltes. in press). However, there are some similarities between everyday cognitive abilities and traditional measures of crystallized intelligence. Both tend to show increases with age (see e.g., Figure 5) or remain relatively stable during much of the adult life span. In addition, there is some conceptual similarity between the characterization of crystallized intelligence as knowledge of one's culture and findings showing substantial consensus among judges about the effectiveness of different strategies in solving everyday problems (Cornelius & Caspi, 1987) and stressful situations (Experiment 4; also see. McCrae & Costa, 1986; Sternberg & Soriano, 1984).Judgments of response efficacy do not appear to be highly idiosyncratic but seem to reflect widely-shared beliefs. In this sense. strategic or procedural knowledge about effective resolutions to everyday problems may be similar to the declarative or factual knowledge assessed by many traditional tests of crystallized abilities. Even so, different factors may influence the acquisition of practical knowledge about everyday life and the declarative knowledge sampled in Crystallized ability tests (e.g.. Hoyer, 1987). For example, in our research with the Everyday h-oblem Solving Inventory. formal education showed a substantial relation to performance on the Verbal Meaning test but was not related to everyday problem solving. Thus, it is possible that the knowledge required for everyday problem solving is not part of the
452
cornezius
acculturation process that is fostered in academic settings and typically associated with superior performance on ability tests identified with crystallized intelligence (cf. Wagner & Sternberg. 1986). In conclusion, contextual models of adult intellectual development and concern about the ecological and external validity of ability tests have provided an impetus for reexamining traditional conceptions of intelligence. Traditional conceptions have focused on abstract problem-solving and verbal abilities. These abilities appear to be fundamental to successful performance in academic settings, and they also may be important to effective functionlng in everyday contexts. However, research on everyday intelligence requires a broader scope that extends beyond these abilities. Further attention needs to be devoted to the questions of how adults acquire. represent. and utilize knowledge about the pragmatics of everyday life. ACKNOWLEDGEMENTS Preparation of this chapter was conducted while the author was a Visiting Scientist at the Max Planck Institute for Human Development and Education, Berlin, Federal Republic of Germany. Generous support by the Institute and discussions with colleagues are gratefully acknowledged. REFERENCES Anastasi, A. (1968). Psychologfcal testing (3rd ed.). New York Macmillan. Baltes, P. B., & Smith. J. (in press). Toward a psychology of wisdom and its ontogenesis. In R.J. Sternberg (Ed.), Wisdom Its nature. orfgins, and deueloprnent. New York Cambridge University Press. Beck, L.. McCauley. C., Segal, M..& Hershey. L. (1988).Individual differences in prototypicality judgments about trait categories. Journal of Personality and Social Psychology, 55,286-292. Berg, C.A. (1989). Knowledge of strategies for dealing with everyday problems from childhood through adolescence. DeueZopmentaZ P~y~hotOgy, 25,607-6 18. Berg, C. A , , & Sternberg, R. J. (1985). A triarchic theory of intellectual development during adulthood. Developmental Review, 5.334370.
Everyday Cognitiue AbUUies
453
BilIings. A. G., & Moos, R H. (1981).The role of coping responses and social resources in attenuating the stress of life events. Journal of Behavioral Medicine, 4, 139-157. Birren. J. E. (1969). Age and decision strategies. In H. T. Blumenthal (Ed.), Interdisciplinary topics in gerontology (Vol. 4:Decision maldng and age, pp. 23-36). Basel: S.Karger. Blanchard-Fields, F. (1986).Reasoning in adolescents and adults on social dilemmas varying in emotional saliency. Psychology and Aging, 1,325-333. Bliesmer. R., Willis, S. L., & Baltes, P. B. (1981).Training research in aging on the fluid ability of inductive reasoning. Journal of Applied Developmental Psychology, 2, 247-265. BoMnick, J. (1977).Intellectual abilities. In J. E. Birren & K. W. Schaie (Eds.). Handbook of the psychology of aglng (pp. 580605).New York Van Nostrand Reinhold. Camp, C. J., Doherty, K., Moody-Thomas. S., & Denney, N. W. (1989). Practical problem solving in adults: A comparison of problem types and scoring methods. In J. D. Sinnott (Ed.), Everyday problem solving: Theory and applications (pp. 2 1 1228). New York Praeger. Cantor, N., & Kihlstrom. J. F. (1987).Personallfy and social intelligence. Englewood Cliffs, NJ: Prentice-Hall. Capon. N., Kuhn, D., & Carretero, M. (1989).Consumer reasoning. In J. D. Sinnott (Ed.), Everyday problem solving: Theory and applfcatfons (pp. 153-174).New York Praeger. Cattell, R. B. (1971).Abffltles: Their structure, growth. and action. New York Houghton Mifflin. Cavanaugh. J. C.. Kramer, D. A., Sinnott. J. D., Camp. C. J.. & Markley, R. P. (1985).On missing links and such: Interfaces between cognitive research and everyday problem-solving. Human Development, 28, 146-168. Ceci. S. J. (1990).On intelltgence...more or less: A bfo-ecological treatise on intellectual development. Englewood Cliffs, NJ: Prentice-Hall. Ceci. S . J.. & Liker. J. (1986). Academic and nonacademic intelligence: An experimental separation. In R J. Sternberg & R K. Wagner (Eds.), PracticaZ intelligence: Nature and orfgins of competence in the everyday world (pp. 119-142). New York Cambridge University Press. Cornelius. S. W. (1984).Classic pattern of intellectual aging: Test familiarity, difficulty. and performance. Journal of
[email protected]. 39.20 1-206. Cornelius, S. W., & Caspi. A. (1986).Self-perceptions of intellectual control and aging. Educatlonal Gerontology, 12, 345-357.
454
Cornelfus
Cornelius. S. W., & Caspi, A. (1987).Everyday problem solving in adulthood and old age. Psychology and Aging, 2, 144-153. Cornelius, S. W., Kenny. S., & Caspi, A. (1989).Academic and everyday intelligence in adulthood: Conceptions of self and ability tests. In J. D. Sinnott (Ed.), Everyday problem solving: Theory and applicatfms (pp. 191-210).New York Praeger. Cornelius. S. W., & Rose, J. H. (1987,November). Concepts of eflective coping: Theme and variations. Paper presented at the meeting of the Gerontological Society of America, Washington, D.C. Demmtng, J. A., & Pressey, S.L. (1957).Tests "indigenous" to the adult and older years. Journal oj' Counseling Psychology, 4 , 144-148. Denney, N. W. (1982).Aging and cognitive changes. In B. B. Wolman (Ed.), Handbook ofdeuebprnenta2 psychology (pp. 807827). Englewood CWs. NJ: Prentice-Hall. Everyday problem solving: Denney. N. W. (in press), Methodological issues, research findings, and a model. In L. W. Poon, D. C. Rubin, & B. A. Wilson (Eds.), Everyday cognition in adulthood and late lfe. New York Cambridge University Press. Denney, N. W.. & Palmer, A. M. (1981).Adult age differences on traditional and practical problem-solving measures. Journal of Gerontology, 36, 323-328. Denney. N. W.. Pearce. K. A., & Palmer, A. M. (1982). A developmental study of adults' performance on traditional and practical problem-solving tasks. Experimental Aging Research, 8,115-118. Dittmann-Kohli, F . , & Baltes, P. B. (in press). Toward a neofunctionalist conception of adult intellectual development: Wisdom as a prototypical case of intellectual growth. In C. Alexander, E. Langer, & R. Oetzel (Eds.), Hfgherstages ofhuman development. New York Oxford University Press. Dixon. R A.& Baltes, P. B. (1986).Toward life-span research on the functions and pragmatics of intelligence. In R J. Sternberg & R. K. Wagner (Eds.). Practical intelligence: Nature and orfgfns of competence in the everyday world (pp. 203-235).New York: Cambridge University Press. Dixon. R. A. Kramer, D. A.. & Baltes, P. B. (1985).Intelligence: Its life-span development. In B. B. Wolman (Ed.), Handbook of intellfgence: Theories. measurements, and applications (pp. 301-350).New York Wiley. D'Zurilla. T. J.. & Nezu, A. (1982).Social problem solving in adults. Advances in Cognftiue-Behavioral Research and Therapy, 1 , 201-274.
Everyday Cognitlw Abilltfes
455
Eysenck, H. J., & Eysenck, S. B. (1968). Manual: Eysenck Personality Inventory. San Diego: Educational and Industrial Testing Service. Feifel. H., & Strack, S. (1989). Coping with conflict situations: Middle-aged and elderly men. PsychorOgy and Aging, 4 , 26-33. Fisher-Beckfield, D. & McFall, R. M. (1982). Development of a competence inventory for college men and evaluation of relationships between competence and depression. Journal of Consulting and Clinical Psychology, 50, 697-705. Flavell. J. F. (1970).Cognitive change in adulthood. In L. R. Goulet & P. B. Baltes (Eds.). Life-span developmental psychology: Research and theory (pp. 247-253).New York Academic Press. Ford, M. E.. & Tisak. M. S. (1983).A further search for social intelligence. Journal of Educational Psychobgy, 75. 196-206. Frederiksen. N. (1986). Toward a broader conception of human intelligence. In R. J. Sternberg & R. K. Wagner (Eds.), Practical intelligence: Nature and origins of competence in the everyday world (pp. 84-1 16). New York Cambridge University Press. Gardner. E. F.. & Monge. R. H. (1977).Adult age differences in cognitive abilities and educational background. Experimental Aging Research 3,337-383. Gardner. H. (1983). Frames of mind: The theory of multiple intelligences. New York: Basic Books. Goldfried, M. R.. & D'Zurilla, T. J. (1969).A behavioral-analytic model for assessing competence. In C. D. Spielberger (Ed.), Current topics in clhical and community'psychology (Vol. 1, pp. 151- 196).New York: Academic Press. Guilford. J. P. (1967). The nature of human inteuigence. New York McGraw-Hill. Guilford. J. P. (1969).Verbal analogies test, I. Beverley Hills, CA: Sheridan Psychological Services. Heckhausen, J., & Baltes. P. B. (1989).Perceived controllability of
expected psychological change across adulthood and old age. Manuscript submitted for publication. Heckhausen, J.,Dixon, R A., & Baltes, P. B. (1989).Gains and losses in development throughout adulthood a s perceived by different adult age groups. Developmental Psychology. 25. 109-121. Horn, J. L. (1978).Human ability systems. In P. B. Baltes (Ed.), Life-span development and behavior (Vol. 1; pp. 2 11-256). New York: Academic Press. Horn. J. L. (1982).The aging of human abilities. In B. B. Wolman (Ed.). Handbook of deueloprnental psychology (pp. 847-870). Englewood Cliffs, NJ: Prentice-Hall.
456
Cornelius
Horn. J. L., & Cattell, R B. (1966).Age differences in primary mental ability factors. Journal of Gerontology, 21,2 10-220. Horn,J. L., & Donaldson. G. (1980).Cognitive development in adulthood. In 0. G. Brim, Jr., 8t J. Kagan (Eds.). C o n s t a m and change in human development (pp. 445-529).Cambridge, MA: Halvard University Press. Hoyer. W. (1987).Acquisition of knowledge and decentralization of g in adult intellectual development. In C. Schooler & K. W. Schaie (Eds.), Cognftfuefunctioning and social structure over the Zffe course (pp. 120-141).Norwood. NJ: Ablex. Kanner. A. D., Coyne, J.C., Schaefer, C., & Lazarus, R S. (19811. Comparisons of two modes of stress measurement: Daily hassles and uplifts versus major life events. Journal of Behavioral Medicine, 4 , 1-39. Kthlstrom, J. F.. & Cantor, N. (1984).Mental representations of the self. Advances in Experimental Social Psychology. 17, 1-47. Kramer, D.A (1983).Post-formal operations? A need for further conceptualization. Human Development. 26, 106-1 16. Kramer, D. A. (1989). A developmental framework for understanding conflict resolution processes. In J. D. Sinnott (Ed.), Everyday problem solufng: Theory and applications (pp. 133-152).New York Praeger. Kuhn, D.. Pennington, N., & Leadbeater. B. (1983).Adult thinking in developmental perspective. In P. B. Baltes & 0. G. Brim, Jr. (Eds.), Life-span development and behavior (Vol. 5 , pp. 157-195). New York Academic Press. Labouvie-Vief, G. (1982). Dynamic development and mature autonomy: A theoretical prologue. Human Development, 25. 161-191. Labouvie-Vief, G. (1985).Intelligence and cognition. In J. E. Birren & K. W. Schaie (Eds.), Thepsychology ofagtng (2nd ed., pp. 500530). New York: Van Nostrand Reinhold. Labouvie-Vief, G., Haldm-Larson, J., & Hobart, C. J. (1987).Age, ego level, and the life-span development of coping and defense processes. Psychology and Agfng, 2,286-293. Lazarus, R.S.,& Folkman. S.(1984).Stress, appraisal. and coping. New York Springer, Lewinsohn, P. M., & Talkington, J. (1979). Studies of the measurement of unpleasant events and relations with depression. Applied Psychological Measurement, 3, 83-101. McCrae, R. R.,& Costa, P. T., Jr. (1985). Updating Norman's "Adequate Taxonomy": Intelligence and personality dimensions in natural language and In questionnaires. Journal of Persondfty and Social Psychology, 49, 710-721.
Everyday Cognltiue Abilities
457
McCrae. R. R.. & Costa, P. T., Jr. (1986).Personality, coping, and coping effectiveness in an adult sample. Journal of Personality, 54,385-405. Mischel, W. (1984).Convergences and challenges in the search for consistency. American Psychologist, 39, 351-364. Moss, F. A., Hunt. T., Omwake, K. T., & Woodward, L. G. (1949).The
Social Intelligence
Test, George Washing ton Series.
Washington, D.C.: Center for Psychological Services. George Washington University. Neisser, U. (1979).The concept of intelligence. In R. J. Sternberg & D. K. Dettennan (Eds.), Human intelligence: Perspectfues on its theory and measurement (pp. 179-189).Norwood, NJ: Ablex. Plemons. J. K. Willis, S. L.. & Baltes. P. B. (1978). Modifiability of fluid intelligence in aging: A short-term longitudinal training approach. Journal of Gerontology. 33. 224-231. Poon. L. W.. Rubin, D. C.. & Wilson, B. A (Eds.) (in press). Eueyday cognition in adulthood and late life. New York: Cambridge University Press. Reese. H. W., & Rodeheaver. D. (1985).Problem solving and complex decision making. In J. E. Birren & K. W. Schaie (Eds.), The psychology of ugfng (2nd ed.. pp. 474-499). New York: Van Nostrand Reinhold. Roth. S.. & Cohen. L. J. (1986).Approach. avoidance. and coping with stress. American Psychologlst. 41. 813-819. Schaie, K. W. (1977/1978).Toward a stage theory of adult cognitive development. International Journal of Aging and Human Development, 8, 129-138. Schaie, K. W. (1978). External validity in the assessment of intellectual development in adulthood. Journal of Gerontology, 33, 695-701. Schaie, K. W. (1979).The Primary Mental Abilities in adulthood: An exploration in t h e development of psychometric intelligence. In P. B. Baltes & 0. G. Brim,Jr. (Eds.). Life-span development and behavior (Vol. 2 , pp. 67-115).New York Academic Press. Scheidt, R. J. (1981). Ecologically-valid inquiry: Fait accompli? Human Development. 24,225-228. Scheidt, R. J., & Schaie. K. W. (1978).A situational taxonomy for the elderly: Generating situational criteria. Journal of Gerontology, 33, 848-857. Sinnott, J. D. (1984).Posfformal reasoning: The relativistic stage. In M. Commons, F. Richards. & C. Armon (Eds.),Beyond formal operations (pp. 288-315).New York: Praeger.
458
Cornelius
Sinnott. J. D. (Ed.) (1989a). Everyday problem solving: Theory and applicatfons. New York Praeger. Sinnott, J. D. (1989b). A model for solution of ill-structured problems: Implications for everyday and abstract problem solving. In J. D. Sinnott (Ed.), Everyday problem solving: Theory and applications (pp. 72-99). New York Praeger. Smith, J., Dixon. R. A,. & Baltes, P. B. (in press). Expertise in life planning: A new research approach to investigating aspects of wisdom. In M. L. Commons, J. D. Sinnott, F. A. Richards. & C. Armon (Eds.). Beyond formal operations ZZ: Comparisons and applications of adolescent and adult developmental models. New York Praeger. Staudinger. U. (1988). The study of lffe review: An approach to the
inuestigation of intellectual development across the lffe span.
Berlin: Free University and Max Planck Institute for Human Development and Education. Staudinger, U. M.. Cornelius, S. W., & Baltes. P. B. (1989). The aging of intelligence: Potential and limits. Annals of the American Academy of Political and Social Sciences, 503.44-58. Sternberg. R J. (1985a). Beyond IQ: A triarchfc theory of human fntellfgence. New York Cambridge University Press. Sternberg, R.J. (1985b). Implicit theories of intelligence, creativity, and wisdom. Journal of Personality and Social Psychology, 49. 607-627. Sternberg, R. J., & Berg, C. A. (1987). What are theories of adult intellectual development theories of? In C. Schooler & K. W. Schaie (Eds.), Cognitive functionfng and social structure ouer the lffecourse (pp. 1-23). Norwood, NJ: Ablex. Sternberg, R J., Conway. B. E.. Ketron, J. L., &Bernstein, M. (1981). People's conceptions of intelligence. Journal of Personality and S ~ ~ iPaSlY C ~ O ~ O41,3745. ~Y. Sternberg, R. J., & Soriano, L. J. (1984). Styles of conflict resolution. Journal of Personality and Social Psychology, 4 1 . 37-55. Sternberg, R. J., & Wagner, R. K. (Eds.). (1986). Practical intelligence: Origins of competence fn the everyday world. New York Cambridge University Press. Streufert, S., & Streufert, S. C. (1978). Behavior in the complex environment. Washington, D. C.: Winston. Thurstone, L. L. (1935). The vectors of the mind. Chicago: University of Chicago Press. Thurstone, T. G. (1962). Primary mental abilities for Grades 9-12 (Rev. ed.). Chicago: Science Research Associates.
Everyday Cugnittue Abilities
458
Wagner, R. K., & Sternberg, R. J. (1986). Tacit knowledge and intelligence in the everyday world. In R. J. Sternberg & R. K. Wagner (Eds.), Practical intelligence: Nature and origins of competence in the everyday world (pp. 51-83). New York: Cambridge University Press. Wechsler. D. (1955). Manual for the Wechsler Adult InteUigence Scale. New York Psychological Corporation. Wechsler, D. (1958). The measurement and appraisal of adult intelligence (4th ed.). Baltimore: Willams & Wilkins.. , Williams, S. A., DeMey, N. W., & Schadler. M. (1983). Elderly adults’ perception of their own cognitive development during the adult years. International Journal of Aging and Human Development. 16, 147-158. Wfllis, S. L.. & Baltes. P. B. (1980).Intelligence in adulthood and aging: Contemporary issues. In L. W. Poon (Ed.), Aglng in the 1980s: Psychological issues (pp. 260-272).Washington, D.C.: American Psychological Association. Willis. S. L.. & Schaie, K. W. (1986). Practical intelligence in later adulthood, In R. J. Sternberg & R. K. Wagner (Eds.), Practlcal intelligence: Nature and origins of competence in the eveyday world (pp. 236-268).New York: Cambridge University Press. Woodruff, D. S. (1983).A review of aging and cognitive processes. Research on Aging, 5, 139-153.
This Page Intentionally Left Blank
Aging and Cognition: Knowledge Organization and Utilization Thomas M. Hess Bditor) 0 Elsevier Science Publishers B.V. North-Holland. 1990
CHAPTERELEVEN
AFF'ECT, INDIVIDUAL DIFFERENCES, AND REAL WORLD PROBLEM SOLVING ACROSS THE ADULT LIFE SPAN Fredda Blanchard-Fields Louisiana State University
Cameron J,Camp University of New Orleans
SUMMARY
Recent thinking in the area of adult cognitive development suggests that with adulthood comes an integration and consistency in reasoning across cognitive and dective domains. Given this theoretical framework and current research supporting this notion, we decided to extend this research into the realm of real world problem solving. The literature is replete with studies manipulating experimenter-imposed variables assumed to affect problem solving competence in older adults. However, until recently, relatively little effort has focused on the sensitivity of the dependent measures used to capture possible developmental differences in how the individual structures reality and approaches everyday problems. Thus, quantitative measures (i.e., number of solutions, etc.) may obscure qualitative differences inherent in the solutions themselves. Given these issues, two studies were conducted which used tasks that require individuals to either generate their own interpretations of problem situations varying in emotional saliency or endorse various problem solving strategies. In addition to qualitative differences in
462
Blanchard-Fields and Camp
problem solving, we examined individual difference indices such a s global problem solving ability, openness to experience, and social desirability, as possible moderators of variability in performance. The results from these studies will be discussed in terms of recent theories of adult cognitive development. From both a theoretical and methodological standpoint, researchers have become extremely interested in assessing everyday cognitive functioning in adulthood and aging (Cavanaugh, Kramer, Sinnott, Camp, & Markley. 1985; Sinnott, 1989;Poon. Rubin. & Wilson, 1989). In particular, a number of researchers have focused on adult developmental implications for everyday problem solving a s evidenced in Sinnott's (1989)current edited book on the subject. Along these lines, there is growing empirical evidence that adaptive changes in cognitive style in adulthood are most evident in socio-emotional or everyday types of situations, especially when the way the individual perceives and structures the problem situation is taken into consideration (Adams, 1986;BlanchardFields. 1986;Camp, Doherty. Moody-Thomas, & Denney. 1989: Dixon & Baltes. 1986;Labouvie-Vief. in press; Luszcz, 1989;Sinnott, 1989). In other words, whether cognitive change is observed may not simply be a function of how many more "correct" solutions to everyday problems an individual generates (i.e.. number of solutions generated, Denney. 1982; Denney. Pearce, & Palmer, 1982).but more a function of how the individual restructures the problem space of the everyday situation (Arlin. 1984. 1989;Sinnott. 1989). The purpose of the present chapter is to discuss everyday problem solving in a n adult developmental context with respect to a) conceptualizations and current research examining age differences in everyday problem solving, b) current theories on adaptive cognitive changes in adulthood, c) methodological issues directed at how we measure the effectiveness of solutions, and d) a brief review of current research conducted by the coauthors on everyday problem solving from adolescence through older adulthood.
Problem Solving
463
CONCEPTUALIZATIONS AND RESEARCH ON EVERYDAY PROBLEM SOLVING There is a long history of research examining age dmerences in problem solving (Botwinick, 1984; Giambra & Arenberg, 1980: Kausler. 1982;Rabbitt, 1977;Salthouse. 1982). Typical fhdings demonstrate more effective problem solving in college youth a s compared to older adults. However. the majority of these studies are grounded in traditional conceptions of cognitive maturity in adulthood and thus. reflect supposedly "context-free" measures of problem solving (Denney, 1982;Hartley & Anderson, 1983:Kausler, 1982;Rabbitt. 1965)and overly formalistic youth-oriented criteria of problem solving competence, such a s formal operational criteria (Kuhn, Pennington, & Leadbeater, 1983:Labouvie-Vief, 1984). The use of these formalistic or context-free measures of problem solving implies that they are valid indicators of adaptive skills across multiple contexts in adulthood. Alternatively, they may only represent adaptive skills most appropriate in a n academic context. Researchers have recently questioned whether these types of problem solving tasks adequately reflect a complete picture of adaptive cognitive functioning in adulthood (Dittmann-Kohli & Baltes. in press: Dkon & Baltes, 1986:Labouvie-Vief 1985;Reese & Rodeheaver, 1985). Effective problem solving on these more traditional tasks may be lirnlted when referring to specific adaptive outcomes in adulthood, for example, in socio-emotional or healthrelated domains (Heppner & Anderson, 1985;Labouvie-Vief. 1977: 1985;Luszcz, 1989). In addition, some have argued that the skllls necessary to perform abstract problem solving tasks are not well practiced in middle-aged and older adults (Denney, 1989;Camp et al., 1989). Thus, researchers have undertaken the task to identify and measure types of problem solving skills required for effective functioning during adulthood (Denney, 1989:Reese & Rodeheaver, 1985;Sinnott, 1989). The issue of ecological validity with respect to problem solving tasks resulted in a movement towards the use of more meaningful and realistic materials in problem solving studies, that is, changing the content of traditional problem solving tasks to conform to content more relevant in adulthood (Capon & Kuhn,
464
Blanchard-Fields and Camp
1979; Harber & Hartley, 1983; Heyn. Barry, & Pollack, 1978). Results of such studies reflected an attenuation of the age dfierence findings demonstrated in more traditional problem solving studies. However, a negative linear relationship between age and problem solving performance still remained (Arenberg, 1968; Capon & Kuhn. 1979; Denney & Palmer, 1981, Hartley. 1981; Hayslip & Stems, 1979). Other researchers developed completely new types of problems, similar to those that adults use in their everyday lives (Cornelius & Caspi. 1987; Charness, 1981a, 1981b, 1983; Denney, 1989; Denney et al.. 1983; Krauss, 1989). For example, Denney and her colleagues studied practical problem solving across different age groups in adulthood (ages 20-79 years), They developed age-relevant problems for young, middle-aged, and older adults. Results indicate that young and middle-aged adults performed best on problems relevant to their own age group, whereas older adults did not vary as a function of the age-relevance of the problems. Denney interprets this finding in terms of a biologically-determined decline in the maximum potential of cognitive processing. This decline begins after early adulthood. Although experience may compensate for this deficit during the major portion of adulthood, for example, through practice, it is in the older adult years that the decline is sufficient to interfere with even frequently practiced abilities Denney. 1989). Cornelius and Caspi (1987) also investigated everyday problem solving with different results. They constructed an inventory to assess practical problem solving in everyday situations. The inventory consists of six different content domains in which a n adult might experience a problem situation including consumer issues, home management, interpersonal conflicts with family members, conflicts with friends, conflicts with coworkers, and dealing with technical information. Possible responses were enumerated for each of these situations (including problem-focused action, cognitive problem analysis, passive-dependent behavior, and avoidant thinking and denial) and were evaluated by judges, ranging in age from 24 to 72 (with and without formal training in psychology), as to their perceived efficacy. The inventory was then administered to adults ranging in age from 20 to 78 years.
Roblem Soluing
465
Examination of age differences revealed that effective perfonnance on the problem solving inventory (as defined by the judges) increased with age, whereas performance on more traditional problem-solving tasks declined after middle age. In order to assess the effect of progressive levels of ecological validity in experimental stimuli, Camp, Doherty, Moody-Thomas, and Denney (1989)conducted a study in real world problem solving. The study will be presented in some detail to elaborate types of stimuli and theorizing currently found in real world problem solving research. Young, middle aged, and older adults (N=114)were asked to describe four problems which they had solved in their everyday lives. For the first two of these problems, there were no qualifications or constraints placed upon problem content. These will be referred to as participant-generated open ended problems. Two more problems were then elicited. Participants were told that these problems should deal with interpersonal (or socfal) topics. Thus, approximately four hundred real world problem solving vignettes were obtained. along with attempted solutions to these problems. Here are some examples of the types of problems and solutions that were generated: My sister accused me of taking money from my mother. I told her 1 didn't do it. but she didn't believe me, so I quit talking to her and don't have anything to do with her. My daughter kept taking my car, so I bought another car My girlfriend and I were always fighting, so we decided to stop seeing each other for a while. My new neighbor's kids began egging my house. I talked to the kids, I talked to their parents, and then I talked to the police. Nothing helped. So, I bought a dozen eggs and egged my neighbor's house, which finally solved the problem. For solutions to these problems, participants rated the goodness of their solutions (i.e. generated self-rated solution efficacy). Next, four experimenter generated problems which had been used in
466
Blanchard-Fields and Camp
previous research by Nancy Denney were presented. An example of such a problem is: Suppose that one evening you go to the refrigerator to get something cold to drink. But when you open the refrigerator up, you notice that it is not cold inside, but rather it is waxm. What would you do? Participants generated a solution to each of these four problems, and again rated the efficacy of their own solutions. In addition, solutions both for problems from participants' lives and experimenter-generated problems were rated for efficacy by experimenters. Efficacy was based on a 5-point rating scale a s follows: (1)
A solution that did not deal with the problem source directly and did not immediately remove the problem.
(2) A solution that dealt with the problem source indirectly. but did not immediately remove the problem.
(3) A solution that dealt with the problem source directly, but did not immediately remove the problem. (4)
A solution that dealt with the problem source directly. but only partially removed the problem.
(5) A solution that dealt with the problem source directly and immediately removed the problem. (1 = Least Effective: 5 = Most Effective)
Thus, for young, middle aged, and older age groups, comparisons could be made of solution efficacy for self-generated versus experimenter-generated problems as rated by participants versus experimenters. In addition, for problems from participants' lives, comparisons could be made for open-ended (unconstrained) versus interpersonal/social (constrained) problems.
Problem Solving
467
To classify the problems from participants' lives, a categorization scheme with four major components was devised: (a) interpersonal -- for example, marital difficulties: (b)impersonal -- dealing with other people, but not on a personal basis. for example, dealing with a repairman or other service provider: (c) personal/self -- for example, physical health: and (d) neutral -for example, dealing with a fie. Age was not related to the number of personal or neutral problems reported. Older adults reported fewer interpersonal problems than younger age groups, and more impersonal problems. This was true even when we specifically requested problem content to be interpersonal/social. There are several reasons why these results might have been obtained. Older adults may have been less comfortable discussing interpersonal problems in a n experiment: older adults may have been less able to distinguish interpersonal from impersonal problems: older adults may have been less likely to be confronted with situations involving family or work problems on a daily basis after children have left the home or after retirement; or older adults may have viewed interpersonal situations as less problematic than younger adults. For whatever reason these results were obtained, similar results were also obtained by Cornelius and Caspi (1987). They reported that older adults in their real world problem solving study reported encountering problems dealing with family, friends, and work less often than younger age groups, though no age dflerences were found for problems involving home management, acquiring information, or consumer issues. For self-ratings of solution efficacy, younger and middle aged adults rated their solutions to experimenter-generated problems a s being more efficacious (i.e. better) than the solutions they had used for problems from their own lives. Interestingly, older adults viewed their solutions to both types of problems a s equally efficacious. We also compared the average ratings for the participant-generated problems versus the experimenter-generated problems by age level and found that age level was not significant. For experimenter ratings of solution efficacy, solutions to experimenter-generated problems were rated a s better than solutions to real world problems. This was true for all age groups.
468
Blanchard-Fields and Camp
Thus, experimenter ratings were in agreement with the ratings of younger and middle-aged adults for these two types of problems. The self ratings of older adults did not reflect the pattern of experimenter ratings for solution efflcacy across problem types. For participant-generated unconstrained versus participantgenerated constrained (social-interpersonal)problems. adults of all ages rated solutions for both types of problems a s equally efficacious. When experimenters rated solutions to these same problems, solutions to interpersonal/social problems were rated as less effective than solutions to other types of problems. No age effects were found for any of these measures. Three important points are illustrated by these results: a) real world problem solving measures can be critically influenced by the world from which the problems are derived -- that of the experimenter versus that of the participant: b) what constitutes a good solution to a problem may depend on who is making such a decision, the experimenter or the individual who tried the solution and had to live with the consequences; and c) older adults were NOT found to be less capable at solving problems than other age groups, though age did influence the types of problems people reported. Older adults in this study were less likely to report interpersonal/social problems than younger adults. There were several reasons offered as to why these results were obtained. The explanation that older adults may find such situations less problematic fits recent theorizing in this research area. Mature adults may be less likely to catastrophize such situations, or they may be better able to coordinate the logical and emotional components of interpersonal problems. There is a recent view of adult cognitive development which supports the contention that with adulthood comes an integration and consistency in reasoning across cognitive and affective domains, that is, the ability to coordinate rational, analytic modes of thinking with more contextually-embedded thoughts and feelings (Labouvie-Vief. in press). In the next section, we will explore such theoretical frameworks and current research supporting this notion with respect to everyday problem solving.
Aoblem Solving
469
ALTERNATIVE THEORETICAL FRAMEWORKS FOR EVERYDAY PROBLEM SOLVING
As discussed above, approaches to everyday problem solving have ranged from a) changing the content of traditional problem solving tasks to b) including more meaningful and relevant material to c) developing age-relevant tasks for the elderly and other age groups. Attention is now being redirected from simple content variation of tasks towards utilizing dserential structures of tasks to insure ecological validity. Manipulations simply varying the content of the task rest on the premise that the appropriate standard of comparison for problem solving effectiveness is based on youth-oriented criteria. Problem solving tasks grounded in this orientation tend to be well-structured in that they demand a single correct solution to the problem situation. An important avenue that has been generally overlooked in such tasks is the examination of thought processes used by individuals to solve ill-structured or ambiguous tasks. These tasks require individuals to generate their own interpretation of a problem situation rather than rely on the structure imposed by others. Accordingly. such variables in reasoning as the ability to deal with conflict and uncertainty are manipulated to index reasoning differences. Arlin (19841,for example, proposes that the more mature thinker will redefine the problem space by accepting inherent uncertainties. This process would include problem finding and a delineation of dimensions needed to resolve these ambiguities. It is argued here and elsewhere (Kuhn et al., 1983; Sinnott. 1985; Tversky & Kahneman, 1980)that such a task structure is more representative of those problems encountered in everyday life. Other researchers suggest that successful problem solving can be conceived in the context of adaptive cognition where cognitive, affective, and physical characteristics interact or are embedded in cultural values, attitudes, and sociocultural institutions (LabouvieVief. 1985;Sebby & Papini. 1989). As a result, criteria unique to adaptive cognition in adulthood need to be specified in general, and for everyday problem solving in adulthood more specifically. For example, Schaie (1978)suggests that the well-structured, single criterion problems may exist for youth in academic contexts.
470
Blanchard-Fields and Camp
However, multiple criteria situations are more prominent in the problems of everyday living faced by older adults. There are several researchers who have attempted to specify these criteria from various theoretical perspectives. Labouvie-Vief (in press) suggests that adaptive problem solving entails a) an awareness that the veracity of the solution to a particular problem. embedded in the ambiguity of everyday living, is relative to differing perspectives and goals of the individual: b) an awareness of the logical uncertainty of solutions; c) an awareness of the role of self in interpreting problems: and d) a recognition that logical systems are embedded in a socio-cultural matrix. This perspective is in concert with the approach Dittmann-Kohli and Baltes (in press) take with respect to wisdom. They suggest that wisdom involves a n individual's ability to exercise good judgment in problems of uncertainty and ambiguity, the need to take a contextual perspective in defining and solving problems, and the use of relativistic and reflective thinking in recognizing variation among both individual perspectives and multiple "correct" solutions to problems. Sinnott (1989)elaborates on the concept of relativistic thinking with the inclusion of self-referential thinking processes used in solving everyday problems. She discusses the central controlling role of beliefs, intentionality, and creativity in problem solving. In other words, she calls attention to the importance of emotions and other psychosocial factors involved in solving problems. Effective problem solving in this context involves a n "openness" to perspectives and solutions in order to meet the adaptive demands of one's environment in addition to an organizational structure to maintain continuity. Finally, Kramer ( 1989) suggests that the individual's world views will influence the problem solving style of the individual. She argues that more mature adults will take a more "dialectic" perspective on the world and, thus, make more interactive or integrative attributions in assessing a problem situation, that is, placing conflict in a broader relationship and cultural-historical context. Therefore, in situations of interpersonal conflict (e.g., marital disputes), individuals who are more dialectic in their thinking will tend to make more relationship and situational
Problem Solving
471
causal attributions, as opposed to allocating blame to the individual alone. With respect to these conceptualizations of cognitive maturity beyond adolescence. Blanchard-Fields ( 1986;1988)has conducted several studies examining cognitive developmental differences between youthful thinkers and mature thinkers in more socioemotional domains. For example. in a recent study, BlanchardFields ( 1986) found that adolescents resolved interpretive discrepancies in interpersonal dilemmas low in emotional saliency as well as younger and middle-aged adults. However, for emotionally salient problems, adolescents were less effective at deriving good resolutions than their adult counterparts. These findings relate to the inability on the part of the youthful thinker to integrate and regulate both cognitive and affective systems. The less mature cognitive system may be less stable in a more differentiated or ambiguous emotional context. Given these limitations of youthful thinking, an emotionally salient context may be more disruptive for younger than for older thinkers, thus affecting performance. With increasing cognitive maturity, affective content may pose less of a disruption on performance. The above theoretical approaches to problem solving stem from a cognitive developmental orientation to intelligence or cognitive styles. These relativistic or dialectic styles of thinking can result in more effective problem solving ability for everyday types of tasks. However, it should be noted that it is not likely that aI1 or perhaps even most people develop such increased problem solving ability (e.g.. wisdom) as they age (Kimmel, in press). There is evidence to suggest that college students as well as adults do not display consistent formal operational ability (Brainerd, 1979;Day, 1978;Keating, 1980:Neimark, 1979). Likewise, post-formal styles of reasoning are not achieved by all or even a majorlty of the mature adults assessed (Kimmel. in press; Kramer & Woodruff, 1986;Labouvle-Vief, In press), or may not be applied consistently within the same individual. I t may be easier to deal with the problems of others at an advanced level than to do so with one's own problems (see Camp et al., 1989).Labouvie-Vief (in press) points out that we can specify a n ideal prototype of post-formal styles of thinking. Mature thinking is characterized by movement towards
472
Blanchard-Fields and Camp
this optimal level of thinking and self-regulation. Whereas, some adults will move in this direction, many adults are more likely to deviate from this ideal. As of now there is a need for more research in this area in order to reach any firm conclusions about such changes in cognitive style and its effect on problem solving. In light of the relatively sparse empirical research in this area, another, more traditional experimental approach to the issue of "ecological validity" in everyday cognitive processes has been proposed by Banaji and Crowder (19891. They argue that ecologically valid methods do not ensure generalizability of findings, and in fact, might compromise and obscure legitimate findings in the literature. In other words, as the context-specificity or ecological validity of an experimental design is increased, the external validity or generalizability of the results may decrease. Their contention is that everyday memory is best studied in a tightly controlled lab setting. This allows "context-free" analyses to be conducted, with context-free findings and principles emerging. Such findings and principles are deemed more likely to generalize to a large number of different settings, compared to research conducted in an everyday memory context. They support this contention by pointing to a) the multiplicity of uncontrolled factors in naturalistic contexts which undermine generalizability to other situations, b) the fact that information found in naturalistic studies can be demonstrated more precisely in a laboratory context, c) the fact that no new principles of memory have been discovered through everyday methods, and d) the fact that laboratory results are controlled by the same principles as everyday phenomena. We agree that scientific rigor in research is necessary for the study of psychological phenomena, and that laboratory findings can serve as compelling analogues to everyday cognitive processes. However, it is the purpose of this chapter to take a slightly different tack to the "ecological validity" issue in adult problem solving. Rather than exclusively focusing on the naturalistic "method" of gathering data, for example, gathering field data, (as described by Banaji and Crowder). we are also interested in discovering more ecologically valid "dependent variables" which are sensitive to changes in problem solving strategy which may occur as one grows older. The research we discuss represents a compromise in that we
Problem Solufng
473
use more everyday cognitive dependent measures in the context of a laboratory-based experimental technique for gathering the data. This point will be further elaborated in the following section. METHODOLOGICAL ISSUES
There are a number of salient issues in the problem solving literature that need to be addressed. First, it was pointed out earlier that the problem vignettes selected for studying age Werences need to be relevant to the sample age groups that are assessed. In particular, rather than experimenter-imposed problems, the most relevant problems are those generated by the subjects themselves. Therefore, as in the Camp et al. (1989) study, we administered subject-generated problems from previous studies to participants in the real world problem solving studies discussed below. Second, the literature is replete with studies manipulating experimenterimposed variables assumed to affect problem solving competence in older adults. However, until recently, relatively little effort has focused on the sensitivity of the dependent measures used to capture possible developmental differences in how the individual structures reality and approaches everyday problems. The measures used often have been primarily quantitative in nature, such as the number of solutions generated, the number of steps to arrive at a solution, or the lack of redundancy in responding. Alternatively, both responses and criteria for efficacy of the solution have been generated by experimenters. These types of measures may obscure qualitative differences inherent in the solutions themselves. Camp et al. (1989)used problem solving efficacy to capture the subject's perspective on problem solving competence. However, such a scale may be too global to delineate the qualitatively different form and content of the respondents' solutions. Therefore, we adapted a scoring scheme that reflects recent thinking on cognitive maturity beyond adolescence in the studies discussed below. A third goal of our program of research was to generally include a complete range of subjects from adolescence throughout older adulthood (as seen in Study 1). It is important to chart differences
474
Blanchard-Fields and Camp
in problem solving across a wide enough age range so as to assure that if responses of older adults are different from those of younger age groups it can be demonstrated. In this way, differentiation
between pre-adult maturity as well as "regressive" patterns and progressive adult responses can be demonstrated with greater precision. By contrast, previous research either looked at adolescence through middle-aged adulthood or young adulthood through older adulthood. Fourth, given the issue of dependent variable sensitivity, we feel it is necessary to use problem solving tasks requiring the individual to generate her or his own interpretation of the problem situation. Accordingly, such individual difference variables a s reasoning ability, problem solving approach, personality characteristics, and so on, may better index problem solving differences than age. The current literature suggests that age may not serve as the best marker of developmental differences, given the increased interindividual variability found in adulthood. A number of researchers have found other candidates that are better predictors of developmental differences such as ego development (Blanchard-Fields. 1986). social cognitive level (Blanchard-Fields & Irion, 19881, moral development (King, Kitchener. Wood, & Davison, 1989). and cognitive-personality style (Adams, 1986). Therefore, we included global and specific measures of general problem solving approaches, openness to experience. and social desirability, a s possible mediators and/or indices of variability in problem-solving performance. At this point we would like to present some selected results from initial analyses of two problem solving studies we are conducting. We will conclude with implications of these research projects. ONGOING RESEARCH
Consider the following situation. A woman bought dresses for her daughters that turned out to be
clearly defective. She tries to return them to the store, but the manager will do nothing about it. What should she do? Think about how you would try to solve this problem.
Problem Solving
47s
Now, rate on a scale from 1 to 5 haw familiar you are with this situation. A 1 means that you are not familiar and a 5 means that you are extremely familiar with this situation. Next, rate on a scale from 1 to 5 your level of emotional involvement with this type of problem. A 1 means that you have no emotional involvement with this type of problem and a 5 indicates that you are extremely emotionally involved with this situation. Emotional involvement is defined a s the degree to which the individual finds the problem solving vignette emotionally salient and arousing. Here is another problem. Please use the same rating procedures described above. A person discovers that his or her spouse is having an affair with someone else while the person is away at conventions. This has been going on for several conventions. What should the person do? Again, rate your familiarity with the situation and your level of emotional involvement. These two problems are representative of stimuli we have been using in one of several research projects dealing with real world problem solving. The types of problems presented above were varied with regard to their level of emotional saliency, and presented to adolescents. young adults, middle-aged, and older adults. In our first study, participants were asked to generate their own solutions to these problems in written form. In the second study, solutions representing different problem solving styles were presented and participants rated the perceived utility of each. In this study, we examined the relationship between controllability of the problem situation and problem solving efficacy from young adulthood through older adulthood. study 1 In this study we have data from approximately 320 subjects, representing about 80 members from each of four age groups: adolescents (14-17 years). young adults (25-35years). middle-aged adults (45-55 years), and older adults (65-75 years). These were community-dwelling individuals interviewed in their homes. All
476
Blanchard-Fields and Camp
were screened for the ability to read. write, and follow directions. Health information was also recorded. All were administered the WAIS-R Information subtest as a rough measure of verbal ability and world knowledge. The means (and standard deviations) for adolescents, young,middle aged, and older adults were 15.2 (5.4), 21.2 (4.1). 20.6 (4.81, and 20.3 (5.1). respectively for this measure, which produced a significant effect, F (3. 283) = 22.8, p < .001. Student Newman-Keuls post hoc analyses revealed that all adult groups were comparable in performance to each other, and a s might be expected had higher scores than adolescents. Mean years of education (and standard deviations) for adolescents, young, middle aged, and older adults were 10.6 (1.5). 15.3 (2.6). 14.8 (2.6), and 13.7 (2.9). respectively, producing a significant effect, F (3,283) = 52.0. p < .001. Student Newman-Keuls post hoc analyses revealed that adolescents had significantly less education than any adult group, and that among the adult groups older adults had signincantly less education than young or middle aged adults. The Roblem Solving Inventory (PSI) of Heppner and Peterson (1982) was administered a s a global measure of problem solving ability. It is a measure of global problem-solving efficacy resulting in scores on three problem solving dimensions: problem solving confidence, approach-avoidance style, and personal control. It contains items such as: "When confronted with a problem, I consistently examine my feelings to find out what is going on in a problem situation." Or, "I try to predict the overall results of carrying out a particular course of action." No age effects were found for this measure, indicating that age groups were roughly comparable in their self-assessment of global problem solving competence. This finding also replicates that of Heppner and Peterson (1982). who claimed that their scale had no age bias in adulthood. scale of We also administered the Openness to Experience (0-E) McCrae and Costa (1985). This 48-item measure represents one of five global scales of the Neuroticism, Extraversion, and Openness to Experience Personality Inventory (NEO)(Costa & McCrae, 1985) and includes six subscales: actions, values, ideas, feelings, fantasy, and aesthetics. Age was not related to total score for the O-E scale, nor
Problem Solving
477
to the subscales of actions, values, ideas, or feelings. However, we found a significant, monotonically decreasing use of fantasy from adolescence to old age. Interestingly, Costa (personal communication, 1988)reports that these results might be related to recent findings in which aging is related to a decline in the use of fantasizing as a coping mechanism for dealing with problems. Next. we examined the ratings participants gave to our specific problems, such as the "returning the dresses" problem. (For purposes of this chapter we will not examine the familiarity ratings.) We had presented 15 such problems. As noted earlier, these problems represent actual problem situations generated in the Camp et al. (1989)study by participants ranging from youth through older adulthood. They were divided into 3 levels of emotional saliency - low, medium, and high - with 5 problems at each level. We wanted to determine if participants' emotionality ratings agreed with our levels. This was the case. All age groups rated the "low" level problems a s lowest in emotionality: the "medium" level problems as "medium." and "high" items highest in emotional involvement. No age differences in emotional involvement were found for low and medium items. However, older adults indicated less emotional involvement than other age groups for the high emotion problems. Again, we are reminded of the idea discussed earlier that older adults view social/interpersonal problems as less problematic than other age groups. This might be due to developmental differences in dealing with the emotional component of such problems. This finding is also supported by the research of Diener, Sandvik, and Larsen (1985).They found that older adults report lower d e c t intensity than younger adults, both positive and negative. They explain these results in terms of lowered autonomic arousal, habituation to life experiences, and lowered emotional response as an adaptive mechanism to cope wiih life experiences. Finally, we administered the Marlowe-Crowne Social Desirability Scale (Crowne & Marlowe, 1960)to all participants. It was not correlated with either emotional involvement ratings or Openness to Experience. In addition to quantitative assessments, we examined qualitative dmerences in the solutions participants generated. The
478
Blanchard-Fields and Camp
qualitative scoring scheme was developed based on Labouvie-Vief et a1.k (1987) conceptualization and assessment of sources of stress. This approach applied the logic of Loevinger's scheme of assessing ego level to the causal evaluation of stressful situations. Briefly, it suggests that less mature individuals will locate the source of stress as being external to self and ignore the subjective interpretation of others involved in the conflict situation. In contrast, more mature individuals wffl perceive the source of stress as having differing subjective interpretations between individuals even though an external event may have been the impetus for the stressful situation. This will result in a more interactive approach where an evaluation of the self and others' standards and values is undertaken resulting in a consideration of multiple causes and their interactive relationship. Labouvie-Vief et al. (1987) found a linear relationship between their source-of-stress levels (nine levels) and age (rangingfrom 10 to 77 years). Based on this theoretical rationale, our scoring scheme consists of four categories or styles of responding, ranging from lower to higher levels of considering the interactive relationship of multiple causes in solution generation. They include: 1. Cut and dry, purely instrumental and/or coercive responses.
There is no attempt at an interactive approach and no involvement with another party or perspective. Examples of solutions to the infidelity problem from this category are: Divorce her/him. Shoot her/shoot him/ shoot him and her. 2. Responses that are still instrumental in nature and
somewhat coercive. Some attempt is made at recognizing that the problem involved another person. However, only lip service is given to it. For example:
Talk it over with her and make her decide what she wants. Confront the wife. T-ry to resolve conflicts. If there is a
medical problem, divorce her. 3. There is no coerciveness in these responses. We call it the "discuss and fix it" style. It is a conceptually simplistic style where respondents recognize multiple perspectives but do
Problem Solving
479
not recognize them as complex and involving the subjective interpretations of the parties involved. For example: Confront spouse and talk to try and find out what caused the problem and try to work out the problem. 4. At this level the complexity of the problem as well as the solutions is discussed. A concerted effort is made to
consider the interactive nature of the problem situation and emotional factors involved in the situation. For example: Face his wife with the issue -- If the marriage can still be saved, provided both parties are agreeable, then they should seek a marriage counselor for help. This would require give and take from both, and a lot of cooperation to make it work. Each of the 15 problems received a score reflecting these 4 problem solving styles (ranging from 1 to 4 a s delineated in the scoring scheme above). Problems were scored by three raters. Each protocol was scored by two raters, Coding discrepancies were resolved by the third rater. Interrater reliability ranged from .95 to .98 using Pearson Product-Moment correlations. A total score for the 5 problems contained in each of the three levels of problem solving emotional saliency described above was computed for each subject (i.e., separate totals for the 5 problems in the low emotionally salient level, the 5 problems in the medium emotionally salient level, and the 5 problems for the high emotionally salient level). It was predicted that older adults would generate more interactive solutions (level 4) to problems than younger adults, especially for the high emotionally salient problems. A 4 X 3 mixed model analysis of variance was conducted with four levels of age group as the between subjects factor and three levels of problem solving emotional saliency as the within subjects factor. There were significant main effects for age group, F (3.283) = 5.45, p < .001.and level of problem solving emotional saliency, F (2. 566) = 52.20, p c .0001.A significant interaction between age group and level of problem solving emotional saliency was also demonstrated, F (6. 566) = 3.35, p c .01. Simple main effects
480
Blanchard-Fields and Camp
analyses revealed no significant age differences for the low emotionally salient problem solving style scores. However, there were significant main effects for age for the medium emotionally salient problem solving style scores, F ( 3 , 286) = 7.58, p < .001. and for the high emotionally salient problem solving style scores, F (3, 286) = 4.11,p < .01. Student Newman-Keuls post hoc analyses revealed a similar pattern of age differences associated with each of these significant findings. Means (and standard deviations) are presented in Table 1. Post hoc analyses revealed that young adults' medium level emotionally salient problem solving scores were significantly higher than both adolescent and older adult scores, whereas the middle-aged adults scored significantly higher than the older adults. There were no significant differences between young adults and middle-aged adults. Young adults scored significantly higher than older adults on high level emotionally salient problem solving scores. Although there were age differences in response style, it may not be the most powerful marker of differences in problem solving. As mentioned earlier, researchers have found that age may not be the best marker of developmental differences (Blanchard-Fields, 1986; King et a]., 1989). Given this notion, further analyses were conducted using three individual differences measures (problem solving efficacy, i.e., PSI total score: social desirability: and total openness to experience score) to predict problem-solving style. Multiple regression analyses regressing problem solving scores on the Problem Solving Inventory, social desirability and openness to experience scales separately for low, medium, and high problem solving scores yielded nonsignificant results. Thus, these individual difference measures were not predictive of problem style. Future research should additionally include more developmentally oriented individual difference measures a s markers of problem solving response style (e.g. ego level, see Labouvie-Vief, Hakim-Larson. & Hobart, 1987). These findings are also suggestive of the ability on the part of young and middle-aged adults to be more differentiated in their style of responding. When the problem was emotionally salient, they were less cut and dry. considerlng the conceptual complexity of the situation. However, when the problem evidenced relatively
Problem Solving
481
Table 1 M e a n s land Standard Deviations) of Problem Solving Scores as a Amction of Age Group and Level of Emotionality Problems Level of emotionality LOW
Adolescents Young Middle Older
1.2 1.3 1.3 1.3
(.2)
(.3) (.3) (3)
Medium
1.4 1.6 1.5 1.3
(.4) (.4)
(.3) (.3)
High
1.4 1.6 1.5 1.3
(.4)
(.5) (.4)
(.3)
Note. 1 = purely instrumental/no interaction recognized: 2 = complexity and interaction recognized.
little emotional saliency their response style was more cut and dry and/or instrumental in nature. Contrary to our expectations, it appears that adolescents and older adults are not a s well differentiated in their responding. Blanchard-Fields' ( 1986)results are similar to the current findings in that she assessed the ability to resolve interpretive discrepancies in adolescent, young adult, and middle-aged adult samples and found that middle-aged adults demonstrated greater social reasoning ability in emotionally salient contexts. However, she did not include an older adult sample. There are both methodological and theoretical issues that may account for our findings. especially the fact that older adults responded with less interactive problem solving styles. First, there was a restriction of range problem, with the majority of scores across all age groups reflecting the first two categories of responding (cut and dry; and moderately instrumental). Perhaps the problems were not effective in eliciting a wider range of responses among participants. An alternative scoring scheme, assessing problem solving strategies defined by the coping literature (i.e., problem-focused coping, avoidance coping, passive-
482
Blanchard-Fields and Camp
dependent coping, etc.) may be another useful means for scoring the data and may yield more variability in responding. Alternatively. the medium for generating responses (hand written essays) may not have been conducive to eliciting a more comprehensive sampling of the participants' solution to the problems, especially among older adults. Audio tape recordings with interview prompts may yield more comprehensive data. Future research will need to address these issues. The question still remains as to what accounts for the similarity in responding (as defined by the scoring scheme totals) between adolescents and older adults. I s it simply a matter of regression, proposed by Denney and her colleagues? Or, could the cut and dry/instrumental style of solving these problems be attributed to other factors? For example, Luszcz (1989)suggests that cognitive psychology may offer insight into the nature of everyday problem solving in the form of schemata. In this case, schemata represent old solutions based on past experience imposed on the new problem situation. Over time, experiences are accumulated in order to provide well-instantiated heuristics whereby problem solving strategies are automatically invoked. The use of such schemata may result in cut and dry solutions with little explanation or elaboration. This may account for the older adult style of responding given the years of accumulated experience in old age. This is not to rule out the use of schemata in problem solving in other age groups. However, in addition to accumulated experience, problem solving behaviors of older adults could be compounded by automaticity due to a reluctance or inability to effectively engage in responses requiring high levels of cognitive effort (Salthouse. 1988;Denney. 1989). Further empirical research is needed to test this notion out in all age groups. Study 2 In this study we examined the relationship between locus of control and everyday problem solving efficacy in the following three age groups: 208 young adults (25-35years); 195 middle-aged adults (45-55years); and 134 older adults (65-75years). These too were cornmunity-dwelling individuals interviewed in their homes.
Problem Solving
483
All were screened for the ability to read. write, and follow directions. Health information also was obtained. All subjects were administered the Problem Solving Inventory of Heppner and Peterson (1982) as described above, the Everyday Problem Solving Inventory, Form A (EPSI) developed by Cornelius and Caspi (1987).and Levenson's (1981) Internality, Powerfulness of Others, and Chance Scale (IPC). For the purposes of this chapter we would like to focus on age differences found on the global measure of problem solving efficacy and the domain-specific assessment of everyday problem solving as assessed by the EPSI. A modified version of the WAIS-R Vocabulary subtest (Jastak & Jastak, 1964) was administered as a measure of verbal ability. Twenty of the forty original items were administered, though items were still scored in the same way. Means (and standard deviations) for the young, middle aged, and older adults were 24.5 (8.2).26.5 (9.6).and 23.5 (8.9).respectively for this measure, which produced a significant effect, F(2. 534) = 5.2, p < .01. Student Newman-Keuls post hoc analyses revealed that middle aged adults were significantly higher in their scores than younger or older adults. Mean years of education (and standard deviations) for the young, middle aged and older adults were 14.1 (2.2).14.2 (2.7).and 12.2 (3.0), respectively. producing a sigriificant effect, F(2. 534) = 26.7, p c .001. Student Newman-Keuls post hoc analyses revealed that older aged adults were significantly Ibwer in their years of education than younger or middle aged adults. Once again, no age differences were found for the Problem Solving Inventory (i.e., global problem solving efficacy). This replicates our findings in Study 1. The EPSI (Form A) assesses problem solving skills in three content domains (consumer problems, managing a home, and resolving conflicts with friends). For each of the 24 problem situations presented there are four ways of responding to the problem: problem-focused action, cognitive problem analysis, passive-dependent behavior, or avoidant thinking and denial. Participants rated the efficacy of each of these four problem solving styles a s applied to each of the 24 problem items using a 5-point Likert scale.
484
Blanchard-Fields and Camp
EPSI (Form A) data were analyzed in two ways. The first method was the same one used by Cornelius and Caspi (1987). As stated above, there were 24 problems presented, with 8 problems drawn from each of three problem domains -- Friends, Home, and Consumer Issues. After being presented a problem. participants rated each of four different solutions to the problem, representing the four problem solving strategies described above. Thus, each participant generated 96 problem solving ratings -- a rating for each of 4 solutions for each of 8 problems from each of 3 domains (4 x 8 x 3). Cornelius and Caspi (1987) had shown these problems and solutions to a large number of adults of different ages and had obtained the group average rating (consistent across age groups) for each of these 96 solutions. This set of 96 average ratings was considered to be the "normative" set of ratings which defines the "optimal" rating for each solution. In a subsequent study. adults of dHerent ages were presented the EPSI. For each participant, her or his 96 ratings were compared with the 96 "optimal" ratings. This was done by correlating the 96 ratings of the individual with the 96 optimal ratings. This yielded a Pearson product-moment correlation, which was the participant's "score" for problem solving ability. A high correlation represented a high correspondence between a participant's ratings and the normative group's ratings, and thus would indicate a "high level of problem solving ability. A low correlation would represent a low level of problem solving ability. We refer to this approach as our "Quantitative Analyses." We also analyzed the data in a new way. The 5-point Like& scale used to rate solutions generated scores ranging from "1," indicating that the participant definitely would not do this, to "5."indicating that the participant definitely would use the solution. We summed the ratings given for each of the four types of solutions across problems. For example, all of the ratings associated wlth PassiveDependent solutions were summed into a single score. This yielded a Total Passive Dependent scored based on the sum of the ratings for the 24 ( 1 per problem) passive dependent solutions. In addition, a separate Passive Dependent score was created for each of the three problem domains. Each domain score was comprised of the sum of 8 ratings, since there were 8 problems in each domain. Therefore,
Problem Solvlng
485
we generated a Total Passive Dependent Score which represented how likely the participant was in general to use this form of problem solving. In addition, a Passive Dependent Score was generated separately for the domains of Friends, Home, and Consumer Issues. This same approach was taken to generate summed rating scores for the other three styles of problem solving. Thus, we were able to determine how likely individuals felt they would be to try to use each of the different problem solving strategies. We refer to this approach as our "Qualitative Analyses." For the Quantitatlue Analyses, univariate analyses of variances assessing age group differences were conducted on four problem solability scores (determined by the correlational calculations described above). These included total correlation scores (across the three content domains) and correlation scores for each of the three content domains (friend-correlations, home-correlations. and consumer- correlations). There were significant age differences for problem solving ability combined across domains (total correlations), with F (2, 536)= 8.76,p < .001;problem solving ability within the friend domain (friend-correlations) with F (2,536) = 11.19,p < .001;problem solving ability within the home domain (home-correlations) with F (2. 536)= 9.67,p < .001: and problem solving ability with the consumer domain (consumer-correlations) with F (2. 536)= 5.17,p < .01. Student-Newman Keuls post hoc analyses revealed that both young and middle-aged adults scored significantly higher on problem solving ability than older adults for total problem solving scores. Total correlation means (and standard deviations) for young. middle-aged, and older adults were .46(.16), .49 (.14), and .41 (.18), respectively. This pattern was replicated for both the friend and home domains. However, middleaged adults scored significantly higher than young adults for the consumer domain. There was no significant difference between older and middle-aged adults in this domain. Overall. these findings support the results of Study 1 with the middle-aged adults achieving higher problem solving ability scores than the older adults and no significant differences between middle- aged and young adults (except for the consumer domain). In line with a number of research findings, the present study demonstrates an increase in practical problem solving ability from
486
Blanchard-Fields and Camp
early adulthood through middle age (Denney & Palmer. 1981; Denney et al., 1982). These results do not support Cornelius and Caspi's (1987) finding that overall problem solving efflcacy linearly increases from young adulthood to older adulthood. However, they found no age differences in the home and friend domains. This discrepancy in results could be accounted for by simple design differences: a) we have four times as many subjects participating in this study, b) there are large geographic differences in the samples (northeast versus southeast), c) educational differences, and d) Cornelius and Caspi included adults in their late 50's and early 60's in their older adult sample. For the Qualitative Analyses described above. subjects' solution endorsement ratings were analyzed in a 3 x 3 x 4 mixed-model analysis of variance. The between-S factor was age group (young, middle aged. and older) and the within-S factors were problem domain (friend, home, consumer) and problem solving strategy (problem-focused action, cognitive problem analysis, passive/ dependent behavior, and avoidant thinking/denial). Significant effects were found for age, F (2,534) = 4.6, p < .01, problem domain, F ( 2 , 1068)= 312.4, p < .001, and problem solving strategy, F (3. 1062) = 866.40, p < .001. Signfficant effects were also found for the interactions of age group x problem solving strategy, F (6. 1602) = 4.77, p < .001, problem domain x problem solving strategy, F(6.3204) = 5.87, p < .001. and age group x problem solving strategyx problem domain, F(12. 3204) = 5.87. p < .001. Older adults gave slightly higher endorsements overall ( M = 3.2, S.D. = .3) than young (M= 3.1, S.D. = .2) or middle aged ( M = 3.1. S.D. = .2)adults. Endorsement ratings for the solutions provided in the consumer domain were lower (M= 3.0,S.D. = .3)than for the domains of friend (A4 = 3.2. S.D. = .3) and home (M = 3.2. S.D. = .3). Finally, the problem solving strategies of problem-focused action and cognitive problem analysis were rated a s more frequently used than passive/dependent behavior and avoidant thinking/denial. Means (and standard deviations) for these strategies were 3.6 (.5),3.5 (,5), 2.8 (.5),and 2.7(-5).respectively. Univariate analyses of variance for age differences were once again conducted for each of the four problem solving strategies (problem-focused, cognitive analysis, avoidance, and passive-
Problem Solving
487
dependence) combined across domains and separately within each of the three domains (friend, home, consumer). The analysis for scores combined across domains revealed no significant age differences in the use of problem-focused or cognitive analysis problem solving strategies. However, there were significant age differences for avoidance strategies, F(2, 536) = 7.89, p c .001. and passive-dependent strategies, F (2, 536) = 9.47. p c .001. Student Newman-Keuls post hoc analyses (means are presented in Table 2) indicated that older adults endorsed more avoidance problem solving strategies than either young or middle-aged adults. Post hoc analyses also indicated that older adults endorsed more passive-dependent problem solving strategies than either young or middle-aged adults, and young adults endorsed more passivedependent problem solving strategies than middle-aged adults. The within domain analyses for problem solving style are summarized in Tablcz 3. Cognitive analysis problem solving strategies revealed no significant age differences. For problemfocused strategies there were no significant age effects for the friend and home domains. However, there were significant age differences within the consumer domain, F (2. 536) = 12.91, p c .001. Means (and standard deviations) for young, middle-aged, and older adults are presented in Table 2. Student Newman-Keuls analyses revealed that older adults endorsed significantly more problem-focused strategies than the young and middle-aged adults, and middle-aged adults endorsed more problem- focused strategies than young adults. For avoidance problem solving strategies there were no significant differences for the consumer domain. However, there were age differences for both the friend, F (2.536) = 7.00. p < .001. and home, F (2,536) = 12.04,p c .001, domains. Student NewmanKeuls analyses revealed a similar pattern for each domain. Older adults endorsed more avoidance strategies than both the young and middle-aged adults (see Table 2). Finally, there were no significant differences for the consumer domain when assessing passive-dependent problem solving strategies. Age differences were found for both the friend. F (2. 536) = 9.46, p c .001, and home, F (2, 536) = 4.74. p c .01. domains. Student Newman-Keuls analyses revealed that older adults
488
Blanchard-Fields and Camp
Table 2 Means (and Standard Deviations) of Subjects' Endorsement Ratings of Problem Solutions as a lWtction of Age Group, Problem Domain, and Problem Solving Strategg
Problem domain Problem solving strategy
Friends
Home
Consumer
Young adults (n=208) Problem-focused action Cognitive problem analysis Passive/dependent behavior Avoidant thinking/denial
3.5 (.5) 3.6 (.5) 3.1 (.5) 2.6 (.a
3.1 3.2 2.8 2.8
15) (.5) I.5) (.5)
(.5) (.5) (.5)
3.6 3.7 3.1 2.6
(.5) (.5)
(.5) (.5)
3.2 3.2 2.7 2.8
(.5) (.5) (.6) (.5)
(.5) (.6) (.6] (.6)
3.5 3.6 3.2 2.9
(.6) (.6) (.6) (.6)
3.3 3.2 2.8 2.8
(.5) (-6) (.6) (.6)
3.9 3.7 2.6 2.5
(.5) (.5)
3.8 3.7 2.5 2.5
3.8 3.6 2.7 2.7
(.4)
(.5)
Middle aged adults (n=195) Problem-focused action Cognitive problem analysis Passive/dependent behavior Avoidant thinking/denial
(.4)
Older adults (n=134) Problem-focused action Cognitive problem analysis Passive/dependent behavior Avoidant thinking/ denial
Note. 1 = Certainly would not use the solution: 5 = Certainly would not use the solution.
489
Problem Solving
Table 3 Summary of Age Eflects for Qualitative Analyses by Problem Solving Strategy and Problem Domain
Problem solving strategy Cognitive problem domain
Problemanalysis
Passivefocused
Friend Home Consumer
NS NS NS
NS NS 0 > Y.M
Avoidance / dependent action
Denial behavior
0 > Y.M 0 > Y,M
Y.0 > M
NS
NS
0 >Y,M
Note. NS = Nonsignificant difference: Y = Young adults;M = Middle aged: 0 = Older adults
endorsed more passive-dependent strategies than both the young and middle-aged adults within the friend domah (see Table 2). For the home domain, older adults endorsed significantly more passive-dependent strategies than middle- aged adults (see Table 2). A different picture emerges with respect to age differences in problem solving ability when the "qualitative analysis" is considered. The curvilinear relationship between age group and problem solving ability is no longer evident. Instead there were either no age differences or a linear relationship between age and amount of problem solving strategy endorsement. In general, the two more instrumental and proactive problem solving strategies, problem-focused and cognitive analysis. showed no age dflerences. However, age differences became more apparent in the two emotional management strategies (avoidance and passivedependence) with older adults endorsing more of these strategies than young and middle-aged adults. The domain-speciflcity of responding revealed that older adults use more problem-focused strategies within the consumer domain,
490
Blanchard-Fields and Camp
more avoidance responding within the friend and home domains. and passive-dependent strategies within the friend and home domains as compared to the young and middle-aged age groups. This may indicate a more differentiated approach to problem situations in that older adults use more diverse strategies in handling everyday problems. Blanchard-Fields has found support for this notion in several studies she has conducted. For example. older adults are better able to differentiate the cause from the outcome of a stressful event and in turn allocated self blame only when they felt responsible for the cause (Blanchard-Fields & Robinson, 1987). This was not so for more youthful individuals. This adaptive form of selective coping to situations perceived a s internal or external on the part of older adults is further supported by another study conducted which demonstrated the ability of older adults to effectively choose between emotion- or problem-focused coping depending on the appraised controllability of the situation (Blanchard-Fields & Irion, 1988). There are other implications of the finding that older adults are capable of using problem-focused and cognitive analytical strategies (as evidenced by a lack of age differences or increased use of these strategies). These Rndings are in contrast with the results of Study 1 as well as other research (Denney. 1989)that suggest that older adults "lack" the ability to engage in cognitively complex strategies. With respect to these findings it should be further noted that there were no age differences for all problem solving strategies on the majority of the consumer-oriented problems. These types of problems could, theoretically, be construed a s low in emotional saliency. Age differences do appear on both the home and family situations which can be construed a s higher in emotional saliency. In these situations older adults engage in more avoidance and passive-dependent strategies. The question remains, do older adults deliberately opt for more "defensive" and emotionally regulating responses given the demands of a problem situation or lack the ability to use more cognitively oriented strategies in this context? Given that older adults endorsed cognitively oriented strategies in the consumer domain and there were no age differences on these strategies, perhaps they possess a broad
Problem Solving
491
repertoire of problem-solving and coping strategies and choose to use more "defensive" styles if the situation warrants it, that is, more highly emotionally-laden contexts (home. family). Some research suggests that the ability to passively accept noncontrollable events is related to more effective adaptation (Irion. 1988; Rodin & Langer. 1977; Rothbaum. Weisz, & Snyder, 1982). Future research needs to address the flexibility-in-responding issue in order to gain a more complete picture of problem-solving changes in adulthood and aging. CONCLUSIONS AND IMPLICATIONS
Both Studies 1 and 2 represent parallel outcomes such that low emotionally salient problems (consumer-related in Study 2) yielded no age differences in problem solving style. However, in high emotionally salient problems, there was Increased variability in responding between age groups with older adults opting for more cut and dry (Study 1) and/or emotional regulating responses (Study 2). On the surface the lower level responding of older adults in Study 1 and "defensive" responding in Study 2 may appear as a production deficit, yet at another level it may represent a deeper awareness of when to "avoid." "passively" accept, or use more cut and dry responding in a situation which cannot be controlled. In conclusion, if we assume that traditional problem solving measures have been relatively abstract, with low ecological validity and/or low emotional saliency, then developmental/age differences would only emerge on measures sensitive to information processing/mechanistic manipulations. However, when presented with stimuli of high emotional saliency, ambiguity, and ecological validity, developmental differences can be documented which are related to qualitative changes in the way the individual perceives and structures everyday problems. Therefore, in order to attain a more complete picture of developmental variation in problem solving, both research perspectives are needed in order to delineate when age differences are due to a production deficiency on the part of older adults and when these differences can be attributed to flexibility in responding. As Rybash, Hoyer, and Roodin (1986) noted,
492
Blanchard-Fields and Camp
developmental differences can exist concurrently with regard to information processing decline and increased sophistication in approaches to solving real world problems, such as integrating both logic and emotion in their approaches. Both types of developmental changes can exist in parallel and may even compliment each other. In our approach, we were interested in the notion that mature thinkers are more differentiated in their responding. Given this research goal, it could only be evidenced if manipulations and measures were used which highlight these differentiations. REFERENCES Adams, C. (1986). Qualftattue changes In text memory from adolescence to mature adulthood. Unpublished doctoral dissertation, Wayne State University. Arenberg. D. (1968). Concept problem-solving in young and old adults. Journal of Gerontology. 23,279-282. Arlin. P.K. (1984).Adolescent and adult thought: A structural interpretation. In M.L.Commons, F.A. Richards, & C. Armon (Eds.). Beyond formal operations: Lute adolescent and adult cognitive development (pp. 258-271).New York Praeger. Arlin, P.K. (1989).The problem of the problem. In J.D. Sinnott (Ed.), Everyday problem soluing: Theory and applications. (pp.229-237). New York Praeger. Banaji, M. & Crowder, R. (1989). The bankruptcy of everyday memory. American Psychologist, 44. 1185-1193. Blanchard-Fields, F. (1986).Reasoning on social dilemmas varying in emotional saliency: An adult developmental perspective. PSyChOlogY and Aging, 1,325-333. Blanchard-Fields. F. & Irion. J.C. (1988).Coping strategies from the perspective of two developmental markers: Age and social reasoning. Journal of Genetic Psychology, 149, 141 - 151. Blanchard-Fields. F. & Robinson, S. (1987).Age differences inthe relation between controllability and coping. Journal of Ger~ntologY, 42,497-501. Botwinick. J. (1984).Problem solving: Forming concepts. In J. Botwinick, Aging and Behavior (pp. 274-293). New York: Springer. Brainerd. C.J. (1979).Piaget's theory of intelligence. Englewood Cliffs, NJ: Prentice-Hall.
Problem Solvtng
493
Camp, C.J.. Doherty. L.. Moody-Thomas. S.& Denney, N.W. (1989). Practical problem solving in adults: A comparison of problem types and scoring methods. In J.D. Sinnott (Ed.), Everyday problem soluing: Theory and applications (pp. 211-228). New York Praeger. Capon, N. & Kuhn. D. (1979).Logical reasoning in the supermarket: Adult females' use of a proportional reasoning strategy in an everyday context. Developmental Psychology. 15,450-452. Cavanaugh, J.C., Kramer. D., Sinnott, J.D., Camp, C.J. and Markley. RP. (1985). On missing links and such: Interfaces between cognitive research and everyday problem-solving. H m Deuelopment, 28. 146-168. Charness, N. (1981a). Aging and skilled problem solving. Journal of Experimental Psychology: General, 1 10. 21-38. Charness, N. (1981b).Search in chess: Age and skill differences. Journal of Experimental Psychology: Human Perception and Performance, 7 , 467-476. Charness, N. (1983).Age, sM11, and bridge bidding: A chronometric analysis. Journal of Verbal Learning and Verbal Behaulor, 22, 146-416. Cornelius, S.W. & Caspi, A. (1987). Everyday problem solving in adulthood and old age. Psychology and Aging, 2, 144-153. Crowne. D.P. & Marlowe. D. (1960). A new scale of social desirability independent of psychopathology. Journal of Consulting Psychology. 24,349-354. Day, M.C. (1978). Adolescent thought: Theory, research, and educattonal impftcations. Report prepared for the National Institute of Education. Denney. N.W. (1982).Aging and cognitive changes. In B.B.Wolman (Ed.). Handbook of developmental psychology (pp. 807-827). Englewood Cliffs. New Jersey: Prentice-Hall. Denney. N. W.( 1989). Everyday problem-solving: Methodological issues, research findings and a model. In L.W. Poon, D.C. Rubin. & B.A. Wilson (Eds.), Everyday cognition in adulthood and later lffe. New York Cambridge University Press. Denney, N.W. & Palmer, A.M. (1981). Adult age differences on traditional and practical problem-solving measures. Journal of Gerontology, 36.323-328. Denney. N.W.. Pearce, K.A.& Palmer, A.M. (1982).A developmental study of adults' performance on traditional and practical problem-solving tasks. Experimental Aging Research 8. 115118. Diener. E., SandE. & m e n . RJ. (1985).Age and sex effects for emotional intensity, Developmental Psychology, 21 , 542-546.
494
Blanchard-Fields and Camp
Dittmann-Kohli, F.. & Baltes. P.B. (in press). Towards a neofunctionalist conception of adult intellectual development: Wisdom as a prototypical case of intellectual growth. In C. Alexander & E. Langer (Eds.), Beyond formal operations: Alternative endpoints to human development. New York Oxford University Press. Dixon, R. & Baltes, P. (1986).Toward life-span research on the functions and pragmatics of intelligence. In R. Sternberg & R. Wagner (Eds.). Practical intelligence (pp. 203-234).New York Cambridge University Press. Giambra, L.M. & Arenberg, D. (1980).Problem solving, concept learning, and aging. In L.W. Poon (Ed.). Aging in the 1980's: Psychological issues (pp. 253-259). Washington, D.C.: American Psychological Association. Harber, K. & Hartley. A. (1983). Meaningfulness and problemsolving performance by younger and older adults. Experimental Aging Research, 9,93-95. Hartley, A.A. (1981).Adult age daerences in deductive reasoning processes. Journal of Gerontology,36,700-706. Hartley. A.A. & Anderson, J.W. (1983). Task complexity and problem-solving performance in younger and older adults. Journal Of G ~ o n t o b g y3.8(1),72-77. Hayslip, B. & Sterns. H.L. (1979).Age differences in relationships between crystallized and fluid intelligences and problem solving, Journal of Gerontology, 34,404-414. Heppner, P.P & Anderson, W.P. (1985).The relationship between problem-solving, self-appraised and psychological adjustment. Cognitive Therapy and Research, 9,415-427. Heppner, P.P. & Petersen. C. (1982). The development and implications of a personal problem-solving inventory. Journal of Counseling Psychology. 29. 66-75. Heyn, J.F., Bany J.R. & Pollack, R.H. (1978).Problem solving as a function of age, sex. and the role appropriateness of the problem content. J3xpZomtoryAging Research, 4 . 505-5 19. Irion. J. (1988). A test of two models of controllability in institutionallzed older adults. Unpublished Doctoral dissertation, Louisiana State University. Jastak, J. & Jastak. S. (1964).Short forms of the WAIS and WISC vocabulary subtests. Journal of Clinical Psychology. 20, 167199. Kausler, D.H. (1982).Experimental psychology and human aging. New York John Wiley.
Problem Solving
495
Keating, D.P. (1980). Adolescent thinking. In J. Adelson (Ed.), Handbook of adolescent psychology (pp. 21 1-246). New York: Wiley. Kimmel, D.C. (in press). Adulthood and aging: An interdisciplinary, developmental view. New York: John Wiley & sons,Inc. King, P.M..Kitchener, K.S.,Wood, P.K. & Davison. M.L. (1989). Relationships across developmental domains: A longitudinal study of intellectual. moral and ego development. In M.L. Commons, C. Armon. F.A. Richards & J. Sinnott (Eds.), Adult
development: Comparisons and applfcations of developmental models (Vol. 1. pp. 57-72).New York Praeger. Kramer. D.A. (1989). A developmental framework for understanding conflict resolution processes. In J.A. Sinnott (Ed.). Everyday problem solving: Theory and applications (pp. 133-152).New York Praeger. Kramer. D.A. & Woodruff,D. (1986). Relativistic and dialectical thought in three adult age groups. Human Deuelopment, 29. 280-290. Krauss, I.K. (1989).Testing cognitive skills with playing cards. In J.D. Sinnott (Ed.), Everyday problem solulng: Theory and applfcations (pp. 238-250).New York Praeger. Kuhn, D., Pennington. N.. & Leadbeater, B. (1982).Adult thinking in developmental perspective. In P.B. Baltes & 0. Brim (Eds.), Life span development and behavior (pp. 157-195). New York: Academic Press. Labouvie-Vief. G. (1977).Adult cognitive development: In search of Merrfll-Palmer Quarterly of a1ternative interpretations. Behavioral Development, 23, 227-263. Labouvie-Vief, G. (1984).Logic and self-regulation from youth to maturity. In M.L. Commons, F.A. Richards, & C. Amon (Eds.), Beyond formal operations: Late adolescent and adult cognitive dewlopment (pp. 158- 180).New York Praeger. Labouvie-Vief. G. (1985).Intelligence and cognition. In J.E. Birren & K.W. Schaie (Eds).. Handbook of the psychology of aging (pp. 500-530).New York: Van Nostrand Reinhold. Labouvie-Vief. G. (in press). Modes of knowledge and the organization of development. In M.L. Commons, C. Armon. F.A. Richards & J. Sinnott (Eds.). Beyondformal operations: 2.
The development of adolescent and adult thinking and perception. New York: Praeger. Labouvie-Vief. G., Haldm-Larson. J.. & Hobart, C.J. (1987).Age, ego level. and the life-span development of coping and defense processes. Psychology and Aging, 3,286-293.
496
Blanchard-Fields and Camp
Levenson. H. (1981). Differentiating among internality, powerful others, and chance. In H. Lefcourt (Ed.), Research with the Zocus ofcontrd construct (pp. 15-63).New York Academic Press. Luszcz. M.A. (1989). Theoretical models of everyday problem solving. In J.D. Sinnott (Ed.). Everyday problem solving: Theory and applications (pp. 24-39).New York Praeger. McCrae. RR & Costa, P.T., Jr. (1985).Openness to experience. In R Hogan & W.H. Jones (Eds.),Perspectives in personality: Theory, measurement, and interpersonal dynamics (pp. 145-172). Greenwich, CT: JAI Press. Neimark, E.D. (1979).Current status of formal operations research. H m Development. 22.60-67. Poon. L.W.,Rubin. D.C.. and B.A. Wilson (1989).Everydug cognition in adulthood and later ZLfe. Cambridge University Press. Rabbitt, P. (1965). An age-decrement in the ability to ignore irrelevant information. Journal of Gerontology, 20, 233-238. Rabbitt. P. (1977).Changes in problem solving ability in old age. In J.E. Birren & K.W. Schaie (Eds.). Handbook of the psychology of agtng (pp. 606-625).New York Van Nostrand Reinhold. Reese, H.W. & Rodeheaver, D. (1985).Problem solving and complex decision making. In J.E. Birren & K.W. Schaie (Eds.), Handbook of the psychologg of aging (pp. 474-499). New York: Van Nostrand Reinhold. Rodin. J. & Langer, E.J. (1977). Long-term effects of a controlrelevant intervention with the institutionalized aged. Journal of Personality and Social Psychology, 35, 897-902. Rothbaurn, F., Weisz. J.R, & Snyder. S.S. (1982).Changing the world and changing the self: A two-process model of perceived control. Journal of Personality and Social Psychology. 42, 537. Rybash, J.M., Hoyer, W.J., & Roodin. P.A. (1986).Adult cognition and aging: Developmental changes in processing, knowing, and thinktng. New York Pergamon Books. Inc. Salthouse. T.A. (1982). Decision making and problem solving. In T.A. Salthouse, Adult cognition (pp. 83-102). New York Springer. Salthouse, T.A. (1988).The role of processing resources in cognitive aging. In M.L. Howe & C.J. Brainerd (Eds.), Cognitive deuehpment in adulthood (pp. 185-239).N.Y.: Springer-Verlag. Schaie, K.W. (1978). External validity in the assessment of intellectual development in adulthood. Journal of Gerontology, 33,695-701.
Problem Solutng
497
Sebby, RA. & Papini, D.R.(1989). Problems in everyday problem solving research: A framework for conceptualizing solutions to everyday problems. In J.A. Sinnott (Ed.),Everyday problem solving: Theory and applications (pp. 55-71). New York Praeger. Sinnott, J.D. (1985). A model for solution of fll-structured problems: Implications for everyday and abstract problem solving. Paper presented at the Gerontological Society Annual meeting. Sinnott, J.D. (1989). Everyday problem solving: Theory and applications. New York Praeger. Tversb, A. & Kahneman. D.(1980). Causal schemas in judgments under uncertainty. In M. Fishbein (Ed.), Progress in social psychology (pp.117-128). Hillsdale. NJ: Erlbaum.
This Page Intentionally Left Blank
Aarpnson D.. 47.48,51,80,81
Abelson,R P..94.103.104,105,
118 Abeideen, J. S., 77 h , B. P.,114. 170,234,235 m,c.,73,74,174,462,474 J.W., 97,104,106,107,108 Alberhn, S. A, 113, 147 Akxam&r, R A.53 Auard,F., 351,363 Allen, G. L,302 AUlger, G. M., 53 Allison, S., 363 AMn.M., 43,78,116 Anashi, A,412 Andawn, J. R, 219,347,376 Anderson, J. W..463 h-n, P.A.76.103.119.124, 127 Andemon, R C., 51,107,108,i24 Anderson, W.P.,463 Arenberg, D., 463,464 Arh, P. K,462,469 Amould,D.. 113 m u k M. D., 313,317.318 AstonJones, G..223 Atkinson, R C., 20 Attlg. M. S.,80.114 Attneave, F.,313,317,318 Aanitia, M., 393
Baars,B. J.. 255 Babcock R L.378.392 Backman L,3,4,6,7,8,9,10, 11,
12, 14, 15, 16, 17,20,21,26,373, 402 Baddeley,A D., 172, 175 Baker-Wd, L.,98 Balcerak L. J.. 174. 203 Baldwin,M..80.114 m.K,392.403 Ball,T. M..306 mob, D. A, 239,240,356 P.B.. 99,101, 102, 147,347, 375, 402,412,413,415,417, 419, 421,422,427. 438,442, 449, 451, 462,463,470 Banaji. M.. 472 Bandura,A, 174,181.186,200. 201,202 Barker, M. C., 267
wb,
Barrett.T.R.57.99,llO. 111
Bany. J. R. 464
Bartlett, J. C.. 115 BartIett, F. C., 94, 104 Bartus,R T.. 268 Bayles, K. A. 267,268,269,290 Bean,G., 13,20 Beck J., 391
Beck L,448
Bell-
F. S., 118 Ekhnont, J. M..164 Behore, S. M., 113 Bern. S. L,129 Beg. C., 311,312,314,315,320, 328,392,413,415,418,419,438, 442,450 Berliner, H.. 352 Bernstein. M..415 J. M.. 173,174.175, 186, 201,202 Bethell-Fax C. E., 318, 320 Biederman, I.. 298,393,394,398, 399 Billings, k G., 436 B h n . J. E.214.298,308.310, 314,317.319.322.328.352.420 Bisanz, G. L., 51 Bjorklund, D. F., 98 Black, J. B., 94.103, 145 Blanchard-Fields, F.,58, 442. 462, 471.474,480,481,490 BIickle. T. W.. 393,398.399 BIieszner, R. 427,430,436,445 Boatwright, L. Ic, 172 Bobrow. D. E.. 324 Bollex-,F., 239, 278 Borke, H., 300 Borkowskl, J. G.. 164,201 Bosky, T. M., 223 E. A. 349,363,364365, 366,367,369.374 Boswell, D. A, 74 Botwinick J.. 24, 165, 170.345, 412,463 Bower, G. H., 118 Bowles, N. L., 221, 239 B o p . S. J.. 393 Bradbum. N. M.. 176 Brady. C. B.. 274,280,286 Brained, C. J., 471
w,
m.
500
Author Index
Bransfod, J. D., 79,80,97, 106, 114 Braune, R, 392 Brennan, P.,303 Bressie, S.,6 Brewer, W.F., 94,97,104,105,107 Brigham, M. C..166,167,168,170, 171,201 Britton, B. K., 101 Broadbent, D.E.,180,196 Brodzinsky, D.M.. 300 Brrrwn.A L..94,164 B m . L,303 Brown, S.C.,1 8 4 Bruce, P.R,170,315,316.317, 320,324,330 Bqden, M. P.,308 Buk~tel,L,360,361 Burke, D.M., 24,34,113,215,223, 244, 233,236,237,238,239,240, 246,247,248,250,251 Buschke, H.,273 B&ey, N.,3,5,6,12,22,115 Buttemeld, E.C., 170 Butters, N.,270 Butterworth, B.,51
Byrd. M.,40.55,80
Cafferty, T.P., 346 Calderwd, B.. 359 Camden, C. T..255 Camp. C.J., 162. 170,346,348, 413,440,442,450,462,463,465, 471,473,477 Campbell, J. I. D., 348,376.378 Cantor, N.,415,449 Capon, N.,450,463,464 Capps. J. L,72.80 card s.R,349.375 Caxpenter, P.A,47.49.51 Carr.M., 164 Carretern, M., 450 Carroll, J. D.,175 Caruso,M.J.,1 6 6 -pi, A,427,430,435,442,451, 464,467,483,484,486 Cattell. R B.,99,215,253,421. 427,430,448 -I.@. J.C.. 3. 115,162.163, 168.169,170.173.174.175,180,
184,186,187,188,195, 199,413, 462 Ceci. S.J.,94,98,413,448,450
cerella,J., 239,309,310,311,312, 314,315,316.320,349.363,392 Chabot, R J., 42,43 ChafEn, R.3,174 Chalom, D.,20 Chandler, M.J., 72 Chang, T. M.,351 Cham-, N.,99,102,233,241, 303,345,347,349,351,352,353, 354,356,360,361,362,376,378, 464 Chase, W.G.,97,353.356.402 Chein. I., 3 0 9 Cherry, K. E.,304 Chi,M.T.H.,94,98,349 Chiareh, C.,239,244,392,393 Chiesi, H.L.,95 Christie, R,181 Chrosniak, L.D.,5 Church, K. L,239.392 Chncy. S. M.,392,394,395,402 ClarkSon-Smith. L.313,314,320 Cohen,A,255 Cohen, G.,3,4,6.13,15,34,40,55, 75,77,78,80, 114.115, 221,239 Cohen. L J.,447 Cohen. N.J.,244 a h e n , R L, 2,3,4.6,7,8, 10,13, 14,16,19.20.23.26 Cole, K. D..172 collfns,A. M.,239 Conway, B.E.,415 Cooney, T.,42,43 Cooper, L.A,307,308,309.313, 317,318,319,320,321,324,332 Cooper, P.F.,180.311 Cooper, W.E.,368 Corballis, M.C.. 308 Corkin, S.,268,392 Cornelius,S.W..102.412.422, 427,430,433,435,440.442,451,
464,467,483,484,486
Casta, P.T.,Jr., 415,427,443,446,
451,476,477 Cowey, A,403 m e , A C.,170,312,320 Coyne,J. C.,435 CraikF.I.M., 10.2594, 101,111,
114.136,147.165,168.168,170,
171.187,234,235.236,237,292. 324,331, 392 Crandall, B.W.,359 Cremer, R,396
Author I n d e x Cmnbach, L. J., 178 C m k T., 268 C&ez R G., 94.472 Cmwne, D. P., 477 Curnow, C., 345,369,371 Daneman. M., 51 Danzinger, W.L., 400 Davldson, H., 162 Davis, K. E.. 346 Davis, RT., 4, 5.8, 11 D a w n , M. L,474 Day,M. C., 471 de G m t , k D., 353,355,402 Decker,M. w., 223 DeFries, J. C., 315 Demmlng. J. A,412,450 Demeby, D. M., 173 Denney, N. W., 412,419,420,436, 441,450,462,463,464,465,466, 482,486,490 Derman,H.,5 Deutsch. F., 24.26 DeyloE D., 3 Maz, D. L,239,240 Dick,M. B., 3,7 Diener. E., 477 Dinan, T. G., 223 Dittman-Kddi, F., 99,451,463, 470 DfXMl. D.. 15 Dkon, R A. 24,40,41,55,56,57, 50,68.99,110,111,115,162,163,
168,169,173,174,177,178,180, 182,184,185,188.189.191.194, 195,196,201,412,413,415.419, 421,449,462,463 Doherty. K.. 441,462,465 Donaldson. G., 429 Do*, J.. 75, 122, 129 Donnenwerth-Nolan,S., 12 Donovan, J. E., 180 Doollng.D. J., 106,109 Doren. B., 80,114 Dorhin, D., 47 DorEman, L. J., 223 Dam% M.. 74 Drevenstedt. J., 175 Duchek, J. M., 239,240,356 Dudley, W. N., 24 DZurilla. T. J., 435.436
Ebehg, C.. 352
501
Ekstmm, B. R, 5 Eley,M. G., 308 Ellas. M. F.. 309 Elo, A. E., 354,356,357,358 EngIe, R W.. 360,361 Epstein, W., 171 Erickson, R C.. 3 Ericsson, K. A, 349,369,370,372. 373,376,378
Evans, G., 303 Eysenck H. J.,446 Eysenck M. W., 165 Eysenck, S. B., 446
%milant., M. E.. 392.394 Flan,M. J., 349 Faulkner, D., 3.4, 6, 13. 15, 75,77, 78, 80, 221,239 Fedio. P., 271,277.278.291 Feifel, H., 438 Feigenbaum, E. A, 350 Feldman, J., 199 Feltovitch, P.. 395 Ferrls, S. H.. 268 Fisher-BecMield, D., 437 Fisk A. D., 376,378 FYncher-Kiefer. R,95.96. 101 M e . T. J.. 80.114 Fishbein,H. D., 300 Fltzgerald, P., 180 Flannagan.D. A,125 f i d . J . H..163,164,300,442 Fletcher, C. R , 49, 50 Flicker, C., 268,271 Fbdor, J. A.389 Fok M. D.. 318,321,323,324,330. 332 Folkman, S., 436,444 Ford, M. E.. 450 Fmard. J.L.8.14,24,26,111.165, 239,310,320,389 Francis, N. W.. 110 Fz-anks.J. D.. 79.80, 114 L. D.. 399 hderiksen, N., 413.450 French, J. W., 5 FreundJ.S..4.5,6,11.13.15 Fky. R L.,360 Fnedman,A,97,106,107,138, 139, 142,393 Friedrich. F. J.. 394 Fromkin,V. A. 255 W,P., 273
%.
502
Author Index
Fullerton, A. M.. 77 Gabfiesheski, A. S.,1 6 6 Galambos,J.k,94,103,145,147 Gallaher, D.,172 Gardner, E.F.,440,441 Gardner, H., 421,449,450 Garrison, A, 299 Garvin, E.A, 317 Gaylod, S.A., 298,309,310,311, 314,320 Cellman, L,318,320 Gentner, D.R.366 Geraxd. H. B.,109 Gernsbacher, M..115 Gewirth, L R,270 Clambra, L M.,376,378,463 Gibson, J.J.,390 Gflewald, M.J.,45,54.115,162. 172,175,177,179 Gilmartin, K., 351.356 Glanzer, M.,7,47,268 Glaser, R,349,395 Glenberg. k M.,171 Clover, J. A. 3 GoldMed. M.R,435,436 G o b , E.S..298,299,300,301, 303,304,305,314,320,323, 326, 328,329 Gonnan, M.,115 Gormican, s.,397 G-, J. G.,3, G d , P.. 245,293 Graesser, A. C.,48,80.97.104, 106,107,108,109.118,119,120, 121 G m t , E.A, 345 Greene, B . k , 107,125 G m n e , F.,48 Greene, T. R, 95 Grober, E.,273,274,276,278,291 Gxuen, G. J.,351 Grossman.J. L., 165 G m r , D.,202 Gmwdon, J. H.. 268 Grubs,L, 171 Grueneisch, R,171 Guider, R L,172 Guilford, J. P..427,430 Guttentag, R E.,3,4,6,12,13 Gzesh, S.M.,300 Haberlandt, K. F., 49,80
Hagen, J. W., 94 Hagenlocker. K., 286 Hakamt, M. K., 4,5,7,9, 11 Haldm-Larson, J.,442,480 Hall, S.E.,43,78,1 16 Halpern, D.F.,313,314,320 Hamilton, D.L.. 108 Hamm,V.,80,113 Hanges, P.J.,53 Hanley-D~nn,P.. 110 Harber, K., 464 HarkerJ. O.,41,102 Harman. H. H., 5 Harris, J. E.. 172,175 Hamold, R M.,238 Hartley,kk, 309,310,311,312, 315,320,370,463,464 Hartley, J. T., 41,102,370 Hasher, L,80,97,101,104,106, 107,108,113.114,117, 147,291 Hashtroudi, S..5,13 Hastie, R,107,128,130,133 Hayslip, B.,464 Heckhausen, J.,415,417,418 Hedderlch, J.,223 Held, D.,303 Hebey, J. G., 112,239,240-243 Helmholtz, H. von, 400 Helstrup,T., 3,10,12,16.20 Heppner, P.P.,463.476,483 Herman, J. F.,302,312,315,316, 317,320,324,330 Heron, A, 1 1 1 Henmann, D.J.,3,171. 173,174, 175,185,188,195 Hershey, L.,448 Hertzog, C..162.168,173. 174, 175,178,179,180,181,182,185, 188,189,190,191,194,195,196, 199,201,202,311,312,320, 328, 392 Herzog. A. R, 346,358 Hess,T.M., 103,113,114,115, 116,117,120,121,122,123,124, 125,126,129,130,131,132,133, 136,138,141,144,145,146, 147, 148,235,238,393 Heyn, J. F., 464 Her, D.B..270,286 Higgins, J. N.,114,115,235,238 Hinchley, J. L.170 Hhrichs, J. V.,172 Hobart, C.J., 74,442,480
Author Index Hochberg, J., 318,320,400 Hock, H.. 307 Hofhan, D. D., 397 Holding, D. H., 352.353 Holland. A. 239.278 Hood,J.. 303 Horn, J. L,99,215,246,253,421, 427,429,430,448 Homer, K L.,347 Houston, M..E..372 Howard, D. V.,112.117,239,240, 241,242,243,245,266,392 Howell, S.C., 110 Hoyer, W. J., 94.101,102.239,244, 375,392,393,394,395,399, 401, 402,403,448,451.491 H d , F. J.. 268.280 H&cka, I. M.,165 Hdtsch, D. F.,24,26,40,41,56, 57, 58,99,110,111,112,115,162, 163,165.168,169~174.175,178.
179,180,181,182,184,185,188, 189,191,193,194,201 Hunt, E.. 311,312,320,392 Hunt, R R.3.4.6.12.13 Hunt, T..445 Huttenlocher, J., 247,300 Hyson. R L,299 Imhoff,k w., 394 Inhelder, B.,94,299,300 Intons-Peterson. M.J., 307 Irlon, J. C.,474,490.491 Jacewicz, M. M.. 309.310,311,314, 315,320 Jackson, D. N.,181 Jackson, J. P.,300 Jackson, M.D., 47 Jackson. S. T.,286 Jamieson, D. G.. 308 Jarvella, R J., 47,48 Jastak, J.,483 Jastak S.. 483 Jelalian, E.,170,186 Jenkins, J. J.. 93,94,95 Jessor. R.180 Johnson. M.K.. 5.97.106 Johnson, N. S., 57.58, 78,97,116. 138 Johnson. R C., 315 Johnston, W.A,395,402 Jolicoeur, P.,319,328
503
Jones, G. V..240.252 Jongeward, R H.,94 J o m , A F., 291 Judd, C.M..180 Julesz, B.,397 Just M.A.47.49 Kahn, R L,172 Kahneman.D., 469
Ka& R V.,94 Kallman, H. J., 393 Kanfer, R, 201 Kanner, k D., 435 Kaplan,B., 47 Kaplan,M.,174 Kaxdash, C.k M.. 107,125 A W.,267,269,290 Katz, L,108 Kausler.D.H.4, 5,6,7,8,9, 11, 12, 13,14,15,20,23,25, 115,130, 162,165,345,463 Kawas,C.,273 Kean, M.C.,3,7 Kern.A. L.360 Keating. D.P.,471 Keenan, J. M.,49 &!Her, K.,300 Kemper, S.,41.55 km. J. In, 415 Kiely, J.,80 Kihlstrom, J. F.,415,449 D. C.,471 King, P.M..474,480 Khsbourne, M.,1 1 1,309 Kintsch, W.,47.48,50, 51,52.71, 96.97.108.182 Kirasic, K. c..302 Kitchner, K. s., 474 Klatsky,G.J., 393 Klein, G.A.359 Kliegl, R,102,347,376 m e , D. W..389 W e , G.E.,68 Klopfer, D.. 395 Koeske,R D., 98 Kohout, F. J., 172 Kosnk, w.. 389 Kosslyn. S.M.,306 Kowalski. A. H., 44 Kowalski, D. J., 97 Krmna, A,370 Kramer, D.A. 162,170,412,413. 442,450,462,470,471
Author Index
504
Krampe. R Th.,378 I(rauss. I.
EL, 313,464
Wll. R.6 Kriauciunas, R, 378 Kubicek, L. F., 247 Ku-, M., 308 Kucera, H., 110 Kuhn, D., 413,450,463,464.469
Kuse,A. R,315
LaBerge, D.. 47 9JAouvie-Vief, G., 72, 73, 74. 101, 344.412,413,421,442.449.450. 462,463,489,470,471,478.480 L~~MEU J.IL, , 162,164.170.201. 298 Lachman, M.E.,,170, 174. 181. 186,188 195,202,203 Lachman,R,106,162,164,170, 201,298 Lafronza,V. M.,304 L&u, C.J.,77 Lair,c. v.. 130 Landfleld. P. W.,223 Langer, E., 173, 199,200 Langer, E. J., 491 Langer. P., 292 Lamen, R J., 477 Lasaga M.I., 239, 392 Lamndeau, M., 300 Lawson. J. S, 267 Lmarus.R S., 435,436.444 Leadbeater. N., 413.463 Lea R,202 Lehman, H.C.,345,354 Lehmkule, S. W.,298 Leirer, V.. 108 Lemke,J.H.,172 Lesgolf,A,395 Levenson, H., 181,Levhe, R, 109 Lewinsohn, P.M.,435 Lewis.s.. 300 Ilchty.W..4.5,6,8, 10, 11. 12, 13, 15.20,21,23,23 Light, L L.24,34,45,54,72,74.76. 80,103.113,114,119, 124,127. 148,215,233,236,237.244.245, 246,253,298,303,393 Liker, J., 448,450 List J . k , 110 Lockhead, G. R, 307 Loewen, E. R, 169, 175
Mus, E.F.,239
LDftUck G. R,138.393 bmbadi, L,76.112
hd,F.M.,52 Lovelave,E.A,170.186 hwe, D.,76, 112 Lowe,D.G., 397 hce, R D.,318,320,323,324,330, 332
Luszcz, M.A,462,463,482 htz, R,302 Lynch, G., 223 Ma~Iky,D.G., 214,215,219,223, 225,230,232,233,240,247,252, 253,255,257 Macko. K. A,403 Mackworth, N. H., 138 Macwhinney, B.,247 Madden, D. M., 78,239,392 Mandel, R G., 57,s. 78, 116 Mandler, G., 293 Mandler, J. M., 78,97,104,105, 108,137,138,139,141, 144 Manelis, L, 50 Mantini-Atkinson, T., 10, 13, 16, 23 Markley, R P,162,170,413,462 Markman, E. M., 105 Marks,L E.,286 MarlrUs, H..175,199 Marlowe, D.,477 Marr, D., 390,397 Marsh, G. R, 170,186,298,309, 310,311,314,320 Martin, A,271,277,278,291 Martln, J., 75, 116 Martinez, D. R, 4 Masangkay, k S., 300 Masani, P. A,111 McAndrews, M. P., 239,392 McArthur, D. J., 390
McCauley, C., 448 McClearn, G. E., 315 McClelland, J. L.47,216,219,241 McClusky, K. A, 300 McCormack P.D..303 M-, R R, 415,427,443,446. 45 1,476 M m , R M., 437 McCee, N.D., 376,378 McIntosh, J. L,110 McInW, C. S., 300
Author Index McKoon. G.. 49 McLaclan, D., 392 Meehl, P. E., 178 Messick S., 178 Metzger. R, 302 Metzlm, D.. 307.308,309,314,315, 316,318.319,320,321,324, 325. 326,327,328,330,333 M q w , B. J.F..42.43, 55 Mgrer, D. E., 248 Micell, L.. 114 Milberg, W. P., 309 Miller, G. A, 285,286,353 Miller, J. R, 48,50.51 Miller. K.. 302 Mis~hd.W.. 438 Mi-, M., 403 Mitchell, D. B., 16,26,244.393, 394 Mitchell, D. R D., 378,392 MiS. A, 76, 128 Mobley, M.. 196 Molander. B., 373.402 Monge. R H., 440,441,450 S., 441,462,465 MOC~Y-T~OXIMS, Moon, W. H., 130 Moore, M., 80, 113,172 Moon?,T. E., 303 Mooe, R H., 436 M o m . T.P., 349 Momll. R W., 24,316,320 Monis, C., 171 Monlson. D. F.. 352 Morton, K. R, 162 Moscovltch, M., 392 MOBS. D. E., 223 Moss, F. A. 445 Motley, M. T.,255 Murphy, B., 109 Murphy, M. D., 165,166,167,170, 356 Murphy, N. Z., 184 M m y , H. G., 347 M m . K. F. H.,343 Myles-Worsley, M., 395, 402 Nagourney. B. A, 308 Nakamura. G. V.. 94.97.104,105. 107,108. 109.118.119,120, I21 Naus, M. J.,98 Nebes. R D., 239,266,274,278, 2&0.282.285.286.291.292 Nee@,J. H., 291.292
505
Neimark E. D.. 471 Neisser, U., 2, 175, 391, 415,422, 429,434 Nelson, K., 118 Nelson. T. 0.. 170 Nelson, W. W.. 393 Newell. A, 348,349 Nezu. A. 435 Nezworski, T..302 Niederehe. G., 173 Nllsson. L-C., 3,4,6,7, 9, 10, 11, 12, 14,15.16,20.21 Nissen,M. J.,392 Norgaard, L, 42,78 Norman. D. D. A, 324 Nonts.M.P.,3.5.6,8.10,12.13. 15.21 Navak C. A, 41 Novick M. R, 52 C. A, 115 No-
ober,B. A. 268,269,278.291 obler, L K.. 221 O m M. W.. 172 Oldfleld, R C., 247
Olson. D. R. 300 Olson, K.,42.78 Omwake, K T., 445 Omstein, P.A,98 Overton. W.F..300 Padgett, R J.,3, 5, 6, 10, 12, 13.21, 22,115 m o ,A. 306 Pallatsek.A,393 €% ,her A M., 436,450,462,464, 486
Paher, S. E.,138, 311 Palmon, R, 378,392 Papini. D. R, 469 park D. C.. 24.302.304 Parker, R E., 138, 139.141 Parks, K. R, 180 Patel. V. L.351 Paulhus,D., 181 m e , K. A, 450.462 R c k V., 170 Pena-Paez. A. 44 Pennington. N..413,463 Rrhutter, M., 3.16,26.162,163. 164,165,169,170, 174,182,188, 199,302.393 Person, D., 174
506
Author Index
Rmon, D. C., 199.200 Petersen, C.. 476,483 Petersen, M.. 10, 13, 16, 23 Petros, T.V.,42,43,56, 78 Petrusic,W. M., 308 Pezdek, K., 114,115,298,303 pfau. H,D., 356 Phil@~,P.L.4.5.7, 11, 14 Plaget, J., 94,298,300 Ward, A, 300 Pinker, S,, 390, 397 Pitchert, J.W., 51,107,108,124 Pkmons, J. K,427,430 Plude, D. J.. 309,392,403.403 Podgomy, P., 308,317,321,324, 332
Pollack, R H., 464 pboIlLW.,2,3,14,24,34,44,54, 63,76,111.112,165,168,175, 180.184,221.239.266,310,316,
320,346,413,462 Popkin, S. J., 165, 172 Posner, M. I., 394 Post, T. A, 95 bssey, S. L, 412.450 Pmdey, M.. 164,166,167,168, 170,171,201 Presson, C. C.. 300 Prill, K.A, 400 PugUsi, J. T., 24, 302, 316, 320 Pylyshyn, 2.W., 307,389
Quayhagen. M.. 313,320
k k m ,B. J., 103.306 Reynolds, G. S., 39 Rice, G. E., 42,43, 55 Richards, B., 303 Richards, W.. 397 ma, J. R, 49 Rtp,L J., 147.176 Rissenberg. M.. 268 Ritchey, G. H., 138 Robinson, S., 490 Rodeheaver, D., 420,463 Rcdgers, W. L,346,358 Fbdh J., 491 Roenker, D. L,184 Rogers,J., 223 Fbgers, M., 3 %@aB., 26, 136, 137, 141.303 Roldan, C. E., 308 IboCia P. A,94,375,402.491 Rse, J. H., 430,442
mth,s.,447
Rothbaurn, F., 491 Royer. J. M., 107, 125 Rubin, D. C., 46,56,57,58,71,413, 462
Rublnson, H.. 395 Rudy, J. W.,299 Rumelhart D. E., 100,104,106, 216,219
Rundus, D., 20 Rushton, D. K.,347 E. B., 199 J. M., 94.101,102.375,
m, 402,491
Rabbitt, P. M. A,388,392.463 RabinowItz,J. C., 101, 114, 130,
Salmon, D. P., 270, 271, 280. 290,
136,147,165,166,168,170,171, 234.235 236,237,241,242 RdoJ€, L S.. 181 Ranktn, J. L., 115 Rashad,M. N., 315 Rasinskt, K.,389 Ratcliff, R, 49 B-, H. H.. 3.5.6, 10, 12. 13.21. 22.115 Rayner, K,393 Read, S. J., 103, 145 Rebok,G. W., 43.78.116, 174,203 Reder, L M . , 75, 116, 124,146,147 Reed, s.K., 307 F2eese. H.W., 420,463 Fkgehr, S., 319,321 Rekhel, F. D., 390
293 Salthouae, T.A. 37.102,214.215. 221,222,225,233,237,253. 257. 308,345,347,349,350,363,364, 365,366,367,368,374, 378,392. 395,400,403,463,482 saltz. E., 11, 12, 15 Samuels, S. J.. 47 Sanders, M., 292 Sanders, R E.. 165, 166,167 Sander,S. P., 5,6,7,8,26 Sandvik E., 477 Sanocki, T.,171 sarmo, A, 298 Saylor, L., 196 Scarbomugh, H. S.. 47,48,51.80. 81
Author Index
Schachter,D., 300,301 Schacter, D. 392
L,244,245,293,391.
Schadler, M., 419 Schaefer, C., 435 Schaie, K W., 99,313,320,344. 345.358.374.413.421.430.434. 435,436,449,450,468 Shank R C.,103,104,105, 106, 118 Scheidt, R J.,413,435,436 Schd. D.A. 73,101 Schiano, D. J.,401 Schleser, R, 172 schmitt, F. A, 165,166,167 Schneider, N. J., 80 Schneider,W.. 163.169.291 schroeder, R,5, 6, 7, 8, 26 Schulenberg.J.,173 Schthan. P. D.. 247 Schulz R. 345,369,371
Sebby, R A.469 Segal. M.,4-48 Sehulster,J. R, 173, 197,199 Seidenberg,M. S.,292 Sekuler,R, 389,392,403 Sekldge, J. A, 285 Shank. C. U.,300 S-, N. E.,147 Shatp, M.J..301.303,304.314, 320,323,326,328,329
!Shaver, R D..223 Shaw, R J.,112, 169, 171, 187,239, 240,241,242,243
Shepard, R N.,307.308.309.314, 315,316.318,319,320,321,324, 325,326,327,328,330,332 Sherman-Brown,S., 99,110, 111 Shetzer, L. I., 308 Shevell. S. K, 176 S m ,R M., 20.291 Shimamura,A P.. 245.270 Shindler,A G., 270,286 Siegel,A. W., 302 Siegler,I. C., 345 Sigmundi, R,171
Simms-Knight,J.. 300 Simon, D. P., 351 Simon. E..234,235,236 Simon,E. W.. 40.41, 115 Simon, H. A, 97,347, 351,353, 356,377,402 Simons, M.A, 395,402
507
Shnonton, D. K.. 347.354 Singh, A.245.393 Sinnott, J.D.,102,162,413,450, 462,463,469,470 Skorpanich,M. A, 303 skavronek, E.. 378,392 slaughter.S. J.. 114.129.138,141, 145,148,393 Smith,A D., 24,77,115,165,304 smith,D.A, 97.109 smith,G., 319.321 Wth,J.. 347,349,419,438,442. 449 smith,L,319,321 smith,S.. 270 smith,S. W.,43,55,78, 116 smith.W. R, 43.78.116 Snyder, S. S., 129,491 Somberg, B. L, 378,392 Soriano, L,J.. 443,446.451 Sovaml, M.,302 Sowark D., 376 Spllich, G.J.,43, 55.61.95,96, 109,115
sgoehr,K. T.,298
Squire, L R,245,293 SIUII, T.K., 128,175,199 Starkes, J. L,351,363 Staudinger,U. M.. 412,413,442, 449 Stefanatos, G.,308 Stelgff,J. H., 308.318.321 Steinberg,E. S., 170 Stemberger,J. P., 247 stembeg, R J.. 201,413,415,416, 418,419,421,422,427,430,431,
435.436442.443.446.448.449. 450,451,452
Stems, H. L,464
Stewart M.,20
Stine,E.kL,34,44,54.55,62,63, 77.78,81.83.112.114
Stine,W. W., 53
stokes, A, 392 Stones, M.J., 370 Storandt, M., 24,345
strack s., 438 Streng, H., 223 Streufert,S., 420 Streukt, S. C.,420
Sulln. R A, 109 Sunderland,A. 172,173,175,184 Surber, C. F.,300
508
Author Index
Surber, J. R,44 'pabor. L,42,43,78
Talland, G.A,344,369 Talkhgton, J., 435 Tapley, S.M.,308 Tak,C. S.,132,136,146,147.148 TaUb,H.A,68,111 Teitelbaum, R C.,393 Tesch-Romer, C.,378 Tesser,A. 101 Thompson.C. P.,184 Thompson.L.W., 115,162.172 Thornson, D. M.,234 Thorndike, E.L,110 Thorndyke, P.W., 107 Thronesbery, C., 170 Thustone, L. L,421,448 Thustone, T. G.,313,430,438,445 Till, R E.,112,113,115 T l k , C.S.,162 Timme.V.. 3 Tlsak M. S..450 Todd, J.T., 390 Todd, M.,346 Tomoeda.C.El, 268 "kat, N.J.. 165 Treisman, A,397 'hyena, J. C.,97 lhtter, S. D.,170 lhscott, A F.,360 llhing, E.,234,244,266,391,392 Tun. P.A,44,58,78 l b n e r , A,48 'Rersky, A, 35,469 TVersky,B.,400 l k e d y , J. R,247 U h a n , S.,390,397,401 Ungerleider, L G.,403 Usala, P.,182,196
VanAlstine. J., 182 vanDijkT.A, 47,48,50,51.52.71. 96.108.182 Vandenberg, S.G.,315 Vandennaas, M.0.. 75,122.129 Vanderplas. J. M.,317 vano, L.308 Vaughn,B.S., 300 Veit, C.. 181 Verdonlk, F.,174 Vernon, M.C.,328
Vernon, P. A,347 Vesonder, G.T., 95 Vogt, M. B.,299 von Eye,A,40,115 Voss, J.F.,51,95 Waddell, K.J., 26,136,137.141, 303 Wagner. R K,413,448,450.452 Walker, C.R,147 Walker, L D.,300,301 Wallace. R B.,172 Wallsten, S.M.,117,131,146 Walsh. D.A,41,79,80,102,112. 113.114 Walsh, K K,165 Walsh-Sweeney, L.,3,14 wang, Y.,395 W m , J.E.,181 Waters, C.S.,292 Watldns, S.K,99,110,111 Watson, J. B.,305,306 Watson, J. S.,300 Watts, K,172 Weatherford, D.L,302 Weaver, S. L,174,203 Weber, R.303 Wechsler, D.,427,429,430 Weingartner, H.,271,288 Weisz, J.R,491 Weldon, J., 303 Welford, A. T., 233,241,354 Wellman. H. M.,162,la, 199,200 West, R L,3,5,6,8, 10,12,13,15, 21,172,173,184,346 White, H.,239,240 White, M.J.,308 Wble, C.,75,116 Wckelgren, W. A,241,378 Wckens, C.D.,392 Wlldnson, R T., 363 Williams. D. M.,316 willlams. S.A,419 Willis. S. L.,413,422427,434,450 Wilson. B.A,413,462 Wilson, J. R,315,320 WhgMd. A,34,44,51,54.55.62. 63,76.77,78,112,247 winocur, G.,392 Winslow, L.,389 Woll, S.B.,97 Wood, P.IC,474 Woodruff, D.S., 412,471
Author Index Wootiwad. L G., 445 Worden P.E.,99,110,111 Worthlqr. J. S.,223 Wrlghk M..57,110 Wundt W..305 Wurf. E.,176,199 WiR, S.,176,1 9 9 Yaniv, I., 248 Yee, P.L,113,239 Yekrrvich,F.R,107,147 Yoder,C.,173 Yde,J. C.,308,318,321 zacks. RT.,80,101,113. 114.117, 147,291 Zadny, J.. 109 Z d k S. H..172,173,174 Zbmdoff, J.,308 M, E J., 396 zellnsld,E. M.,45,54,76,80,113, 115.127,128,162,169,172,173, 174.175,177,179,184,298.303 Zeman, B. R,98 ZLlclvx, s. 390,391
w.,
This Page Intentionally Left Blank
SUBJECTINDEX abstraction 106, 108. 114 activation 219-225,228-229. 231233,239.240.246-249,251,254. 255 adaptive cognition 469 adaptive problem solving 470 age-by-task approach 299,304305,323.333-334 Alzheimer's disease 265-267,272, 273,276,285,290,291,293 ambiguous tasks 469 amnesia 245 aphasics 247 athletics 369-374,378 automaticity 266,291,293 beliefs 162,172-177,180,181, 182,184,185-188.190.192,195-
203 bottom-up processing 100,106 bridge 345,350,360-362
categorization 1 17,146 chess 345.348-361 cognitive activities 4-12,15,20. 22,23 cognitive speed 309 cognitive maturity 463.471.473 commitment learning 215.228238,240,241,243-246 compensation 350,352,360,362, 363,366,367,374-378 complexity hypothesis 316,317 comprehension 47.51.60.69.7779,83 conceptually driven retrieval 103, 116,117,119. 123 c o n h a t o r y factor analysis 179 construct validity 177,178.182 convergent validity 178,179, 180 declarative knowledge 391 delayed auditoly feedback 257 dementia 266-269 developmental theory 298 discourse processing 36,37,47, 50,52,54, 55,67,72-75,7!3-81, 83.85 discriminant validity 177.180, 181 distinctiveness 74,75.77,80,84
domain speciflcity 394-396,402 ecoIogid validity 2.3.102.344. 345,413,463,465,469,472.491 emotional saliency 471,475,477, 479,481,490,491 encoding speciflcity 215,233-236 engrainment learning 215.229, 234.236,238.240,244.246 environmental support 392 everyday cognitive abilities 4 13, 440, 414,419,421,426,430.435, 442 448,451 everyday problem s o h g 462464,468,469,482,463 experience 388-389.393-394. 401-403 experllse 345.349-351.379,395396.402 fhnilia.rIty 6-10,13,23-25 feature-driven processes 397
"gating" technique 78 general encoding 238 grammatical information 216 imagined actions 12, 14 impression formation 129. 146 Individual dmerences 95-97 integration 81,83-84,106-108, 112-114,117,122. 123,127,130, 138. 142,145-147 intelligence 412445,446,448, 449,451,452 academic 415-417,420,422427,429431,434,448,449 crystallized 99,253 everyday413.414.416,417.421, 422.423.425-427,431,434.
435,442,448,449,452 fluid 222,246 imp~icittheories of 415,429 mechanics of 99,100,102 pragmatics of 99-147 interference 127,129,130,132
knowledge influences 94-95,102104.149 knowledge structures 94.98,102103,137
512
Subject Index
knowledge uwiZaUon 388-390, 392-395.402
levels analysis 47,50,53.54,57, 72
levels effect 38,49, 50. 52-57, 72, 83
lexical-decision 239,280,281, 292
linkage strength 219-222,229231,240-241,244,246
list length 4,6-8, 10, 12-13, 26 meaningfihess 99, 108, 110-112 memory Complaints 172-174, 176. 177. 181,196 direct 215,233,244-246 discourse 35.37,47 episodic 266,267,271,272,274. 285,288,290 event 5.6, 10, 12, 13.21.23 for new idormauon 231 i n b c t 244,245 monitoring 163, 166.168.170171,187,195 strategies 164-165, 168-169, 179 taskpredictions 162.170, 185. 188 penson 128 scene 136-139, 148 t em antic 266,271,278,290291,293 tUrt 169. 184-185, 188-190, 192193.195 worldng 47.60- 63.69.72.78,
80.85 mental chronometry 306-307 mental image rotation 305,308319,322-323, 329, 332-333
mental imagery 305-306.309, 311-312, 323-324,332
node structure theory (NSq 2 14222,225,228-229.231-236.242, 248,253-254
object cuea 3, 4, 6, 16-19, 22, 25 object-driven processes 397 object naming 247 online measures 49 online reading 81 organization 3. 5-6, 14, 17-22, 24, 35,50,55,63,67,72,82-83, 94, 103, 105, 112, 137, 138, 139, 141142 orthographic system 216
peak chess performance 357 perception 222,257 perception and production of language 215 phonological system 216, 251 phonemic information 78 preparation in visual processing 394 prior knowledge 391,396,401 practice 219,222,229,231,233, 240-241,244,246
predictive validity 177, 182, 184185
presentation time 5-8. 10 priming 214-215,214234,238257,270-271.278-280,282,289293 problem soMng 413-414, 418, 420,425-426.430-442,446,449451 problem-solving style 480 processing resources 48, 58-60. 69.74-75.85.214.237 production deficiency 164-166, 167,201 production (speech) 216, 219-220, 222, 248, 251-254 prose characteristics 62
metamemory 162-203 metamemory questionnaires 175178,179,182.185,197,200,203 molar equivalence - molecular decomposition 37 motor cues 5, 11-17, 19, 21-23
real world 465 real world problem solving 465.
new versus old learning 2 15
recognition-by-components
qualltative age dmerences 34-38. 47,54,72,74,77.80.83-84
467-468, 473,475 (RBC) t h e ~ r 398-399 y
nzheard 14,20.22,24
Subject Index
relative memorability 56-58,6163,67,69-73,823 representational knowledge 391 response style 74,75 schema(s) 393 and memory 104,110-149 processing characteristics 106109 Structure 104- 105 scripts 77,103-105,108,118-129, 136,144,146 selectivity in visual processing 394,400-401 self-efficacy 163. 174-175,176179,180-181,185-188, 190, 192, 193,195,196,197-203 self-inhibition 219-220 semantic attributes 271-272,274, 291 semantic information 247. 251 semantic knowledge 391-404 ~klll343,345,347,350-351,353356,360-367,369,374.375-376, 378-379 Slowing 214-215,225,232 social cognition 199 solution efncacy 465-468 speech errors 255,257 speed requirements 309,314. 322-324,326.328-329 Spoonerisms 255-256 stimulus characteristics 301 S h o p task 226,239 structural regression model 180, 184,190 structural support 77.79.85 subject performed tasks (SITS) 3-5,7.8,10-11,14-17,20, 23,26 subliminal perception effects 257
task appraisal 187-188,197-198
theories of compensation 374377 theory in cognition and aging 214 theory of language and memory 214 theory of tradeoff 348 tip of the tongue 216 top-down analysis 400-401 top-down processing 100 trade off 227 training effects 371,372
513
transmission deflcit 215-216, 221-225,231-233.240-241, 246247.249-254.257 typing 345,350,362-369,374. 376-377,395
unit of analysis 56,73 verbal fluency268,290,292 verbal transformation effect 257 visual cognitionlproceselng 297299.305.333-334.388-390.395. 397,403 visual cognitive abilities 298,301 visual perception 391-397 visual (object) recognition 395399,401,403-404 watched activities 12-13 word-by-word reading 81 word list recall 1 8 4 word flnding 267-268
This Page Intentionally Left Blank