Advances in Psychology Volume 98 Imagery, Creativity, and Discovery: A Cognitive Perspective

ADVANCES IN PSYCHOLOGY 98 Editors:

G. E. STELMACH

P. A. VROON

NORTH-HOLLAND AMSTERDAM LONDON NEW YORK TOKYO

IMAGERY, CREATIVITY, AND DISCOVERY A Cognitive Perspective

ADVANCES IN PSYCHOLOGY

98 Editoi-s:

G. E. STELMACH P. A. VROON

NORTH-HOLLAND AMSTERDAM * LONDON * NEW YORK TOKYO

IMAGERY, CREATIVITY, AND DISCOVERY A Cognitive Perspective

Edited by

Beverly ROSKOS-EWOLDSEN Depai.tnient of Psychology University of Alabama Tuscnloosa, AL, U.S.A.

Margaret Jean INTONS-PETERSON Depurtnient of Psychology lndianu Uiii\vrsity Bloomingtnn, I N , U.S.A.

Rita E. ANDERSON Department oj'Psychology Memorial University St. John's, Neu;foundland, Canuda

1993

-

NORTH-HOLLAND AMSTERDAM LONDON * NEW YORK 'TOKYO

NORTH-HOLL AND ELSEVIER SCIENCE PUBLISHERS B.V. Sara Burgerhartstraat 25 P.O. BOX21 I , 1000 AE Amsterdam, The Netherlands

Library of Congress Cataloging-in-Publication Data

Imagery, creativity, and discovery: a cognitive prspectivdedited by Beverly Roskos-Ewoldsen. Msrgarer Jean Intons-Peterson Rita E. Anderson p. cm. - (Advances in psychology; 98) Based on papers presented Bf a conference held at Vandabilt University in May, 1991 Includes bibliographical references and indexes. ISBN 0-444-89591-4 (alk. paper) 1. Creative ability-congresses. 2 Imagery (Psychology)s. I. Roskos-Ewoldsen, Bevaly. II. Intons-Peterson,Margaret Jean. In. Anderson. Rita E. IV. Series: Advances in psychology (Amsterdam. Netherlands) ; 98. BF411152 1993 153.34~20

93-17150

CIP

ISBN: 0 444 89591 4 0

1993 ELSEVIER SCIENCE PUBLISHERS B.V. 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.. Copyright & Permissions Department, P.O. Box 521, 1000 AM 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. Thih booh is printed on acid-free paper. Printed in The Netherlands

V

TABLEOF CONTENTS Preface..

.....................................

vii

..............................

xi

List of Contributors

Imagery’s Role in Creativity and Discovery Margaret Jean Intons-Peterson

............

1

The Ins and Outs of Working Memory: . . . . . . . . . . . . . . 39 Overcoming the Limits on Learning from Imagery Daniel Reisberg and Robert Logie Images Are Both Depictive and Descriptive Deborah Chambers

...........

77

.......

99

Imagery, Reconstructive Memory, and Discovery Ira E. Hyman, Jr.

Mental Imagery: Fixed or Multiple Meanings? . . . . . . . . 123 Nature and Function of Imagery in Creative Thinking Geir Kaufmann and Tore Helstrup 6

The Ambiguity of Mental Images: . . . . . . . . . . . . . . . . . . 151 Insights Regarding the Structure of Shape Memory and its Function in Creativity Mary A. Peterson Discovering Emergent Properties of Images Beverly Roskos-Ewoldsen

. . . . . . . . . . . 187

Multiple Perspectives on Discovery . . . . . . . . . . . . . . . . . 223 and Creativity in Mind and on Paper Rita E. Anderson and Tore Helstrup

Table of Contents

vi 9

Mental Imagery and Creative Discovery Ronald A. Finke

10

Imagery and Discovery Stephen K. Reed

. . . . . . . . . . . . . . 255

..........................

11 Imagery, Creativity, and Discovery: Conclusions and Implications Beverly Roskos-Ewoldsen, Margaret Jean Intons-Peterson, and Rita E. Anderson

.287

. . . . . . . . . . . . . . . . .313

12

Author Index

..................................

329

13

Subject Index

..................................

337

vii

PREFACE Does creativity depend on imagery? Can discoveries be made in imagery? Einstein and many others (see Shepard, 1978), have claimed that their creations began first as images and only later were transformed into words. If imagery has precedence in creative discoveries, why do people sometimes have difficulty "seeing" information in their visual images that they readily see in perception? Do our expectations affect the way images are generated and maintained; do our expectations affect what can be discovered in or learned from images? How do creativity, discovery, and imagery intersect? These intriguing questions have long tantalized psychologists and others interested in creativity. Despite the extensive history of creativity, discovery, and imagery as phenomena in need of explanation, relatively little progress has been made in answering these queries until fairly recently. Now, however, recent theoretical developments, coupled with a growing data base, suggest that it is desirable to bring together relevant research, to assess theories, and to chart some courses for the future. In part, we were acting and reacting to the developments in the field; we thought it timely to juxtapose various views of imagery, creativity, and discovery. The above goals guided development of the book, but, like most efforts, the book evolved. It began with hallway conversations at the meetings of the Psychonomic Society in November, 1990, while some of us chatted about our research. Frustrated by the short time available at the meetings, we decided to reconvene at a conference to explore common themes emerging from a cognitive approach to imagery, creativity, and discovery. Bev Roskos-Ewoldsen assumed responsibility for organizing the conference. She contacted various people and institutions about sponsoring the conference, invited prospective participants and attenders, made the arrangements, and engineered an intellectually challenging and delightful conference. The conference was held at Vanderbilt University in May, 1991.

viii

Preface

Diverse opinions, paradigms, and results sparked our imagination. The result was an invigorating and vigorous sharing of experiences and views, as well as an agreement to share our experience with the world via a book. A number of themes emerged, and these are developed in the chapters in this book. One theme, of ancient but still current vintage, is that many individuals, universally acclaimed for their creativity, believe that their creative discoveries blossomed from their images. We could not ignore these impressions, often offered with compelling fervor. However, the experimental evidence seemed less persuasive than the anecdotal accounts. For example, college students, unselected for imagery or creative ability, failed to detect a second interpretation of the perceptually ambiguous duck-rabbit figure in their images, although they were able to do so as soon as they drew their images on paper (e.g., Chambers & Reisberg, 1985). Such results suggest that the images might constrain creativity and discovery, rather than expanding it, as the anecdotal reports had intimated. Needless to say, the original duck-rabbit findings drove a number of our contributors to the laboratory to pursue various explanations. The results of these explorations delivered still other themes. One of the additional themes is that expectations affect what c a n be detected in images. If subjects expect a particular orientation, it may be difficult for them to re-orient their images, thereby reducing the probability of making discoveries from their images (see chapters by Chambers, Peterson, and Reisberg-Logie). These expectations may be determined by linguistic or by perceptual (graphic, pictorial) information used to induce imagery. Moreover, the more cohesive, symmetrical, or psychologically "good1'an imaged pattern is, the more difficult it may be to dissemble and reassemble (see chapters by Peterson and Roskos-Ewoldsen),another theme. In brief, expectations may constrain the utility of imagery. Several ways to overcome these constraints are demonstrated. Chambers reports the use of instructions to view the image from a different orientation; Hyman and Peterson use hints designed to serve the same purpose. Kaufmann and Helstrup demonstrate greater facility with mental discovery in images among artists than is typical of college students unselected for imaginal or creative ability.

Preface

ix

Still another theme is that images can be combined mentally to yield composites. These composites may serve as media for the creation and discovery of new patterns (see chapters by Anderson and Helstrup, Finke, Intons-Peterson, Roskos-Ewoldsen). In the mental synthesis of component parts, constraining the kind of product to be discovered may actually increase the numbers of composites identified as creative by independent judges (Finke). Furthermore, patterns judged t o be perceptually good or cohesive are more difficult t o parse imaginally than those judged less good or cohesive, which suggests that imagining the known and predictable may be creatively counterproductive (Roskos-Ewoldsen). Practice with the mental construction of novel image composites may increase the production of creative composites in a second session (IntonsPeterson). Finally, compared to pure mental synthesis, the provision of drawing support increases the numbers of creative patterns by a surprisingly surprisingly small amount (Anderson and Helstrup). The cognitive influence permeated the conference, and it shapes the models presented in the chapters. All of the chapters offer or evaluate models of imagery, creativity, or discovery which have a distinctive cognitive flavor to them, as is demanded by the evidence. Some of the models attempt t o integrate perceptual and other sensory systems with imagery, creativity, or discovery (Reisberg and Logie). Others emphasize working memory (Anderson and Helstrup, Kaufmann and Helstrup, Reisberg and Logie, Roskos-Ewoldsen). As well, the need to examine cognitive models of imagery and creativity and discovery from other perspectives (i.e., evolution, development, neurophysiological, motivation) was raised (Anderson and Helstrup, Intons-Peterson). The book culminates with two chapters. Reed develops the conditions under which we are or are not likely t o find imagery useful. Roskos-Ewoldsen,Intons-Peterson, and Anderson summarize the current status of the field and identifies new directions t o pursue. To say more now would be to give away many of the creative images to be discovered in the chapters. We thank Vanderbilt University and Indiana University for their contributions to and financial support of the conference. Specifically, Randy Blake and Keith Clayton helped Bev Roskos-

X

Preface

Ewoldsen and me to organize the conference by guiding Bev toward knowledgeable contacts at Vanderbilt University. Also assisting were the Office of Research and the University Graduate School (Dean George Walker) and the College of Arts and Sciences (Dean Morton Lowengrub), both of Indiana University. More good luck was in store, for Elsevier Scientific Publishers agreed t o publish the book. In particular, I want t o state our appreciation of the excellent assistance of K. Michielsen and Erik Oosterwijk. Other acknowledgments are in order. Catherine Barnes and Vicki Blackwell are responsible for preparation of the manuscript. They and we often relied on Nathan Engle and Sheryl Mobley for their voluminous knowledge of computers and software as we prepared camera-ready copy. We are most grateful to them. Finally, we thank our families and colleagues for their support and forbearance.

Beverly Roskos-Ewoldsen Margaret Jean Intons-Peterson Rita E. Anderson July 1992

References Chambers, D., & Reisberg, D. (1985). Can mental images be ambiguous? Journal of Experimental Psychology: Human Perception and Performance, 11, 318-328. Shepard, R. N. (1978). Externalization of mental images and the act of creation. In B. S. Randhawa and W. E. Coffman (Eds.). Visual learning, thinking, and communication, (pp. 133-189). New York: Academic Press.

xi

LIST OF CONTRIBUTORS Rita E. Anderson Department of Psychology, Memorial University of Newfoundland, St. John's, Newfoundland Canada A1C 5S7 Deborah Chambers 1903 Gatewood Place, Silver Springs, Maryland 20903, U.S.A. Ronald A. Finke Department of Psychology, Texas A & M University, College Station, Texas 77843-4235,U.S.A. Tore Helstrup Department of Cognitive Psychology, University of Bergen, Sydneshaugen 2,5007 Bergen, Norway Ira E. Hyman, Jr. Department of Psychology, Western Washington University, Bellingham, Washington 98225,U.S.A. Margaret Jean Intons-Peterson Department of Psychology, Indiana University, Bloomington, Indiana 47405,U.S.A. Geir Kaufmann Department of Cognitive Psychology, University of Bergen, Sydneshaugen 2,5007 Bergen, Norway Robert Logie Department of Psychology, King's College, University of Aberdeen, Aberdeen AB9 2UB, United Kingdom

xii

List of Contributors

Mary A. Peterson Department of Psychology, University of Arizona, Tucson, Arizona 85721,U.S.A. Stephen K. Reed Department of Psychology, San Diego State University, San Diego, California 92182,U.S.A. Daniel Reisberg Department of Psychology, Reed College, Portland, Oregon 97202,U.S.A. Beverly Roskos-Ewoldsen Department of Psychology, The University of Alabama, Box 870348,Tuscaloosa, Alabama 35487-0348,U.S.A.

Imagery, Creativity, and Discovcry: A Cognitive Pcrspectivc B. Koskos-Ewoldson, M.J.Intons-Peterson and K.E. Anderson (Editors) 0 1993 Elscvier Science Publishcrs B.V. All rights rescrvcd.

1.

Chapter 1

IMAGERY'S ROLE IN CREATMTY AND DISCOVERY Margaret Jean Intons-Peterson Department of Psychology Indium University Bloomington, IN 47405 USA My central task in this chapter is to give a brief historical survey of what we know about the creative process. In preparation for it, I studied books, articles, manuscripts, and more books. It was an enlightening experience. I learned that, in 1926,Wallas told us that the creative process has four stages: preparation, incubation, illumination, and verification. In 1962, Hilgard's third edition of his famous text, Introduction to Psychdogy, told us the same: The four stages of creative thought are preparation, incubation, i l l h a t i o n , and verification. In 1991,65 years after Wallas's pronouncement, Solso delivers the same message. Psychologists seem to be astonishingly uncreative! Seriously, what do we know about imagery's role in the creative process, defined as the sense of bringing into being or as causing to exist, and in the discovery process, defined as making known or visible or of finding something for the &st time? I'm tempted to say, "Preciouslittle." But that is too pessimistic. We have made progress, progress that will be addressed in the chapters in this book. Although we may not have as extensive, compellingevidence as we would like, the topic has long piqued curiosity, and investigations of the creative process have a substantial history. One strongly held view is that imagery plays a critical role in creativity. For example, in his book,Imagery in Scientific Thought,Miller (1984) says, ' I . . . an ingredient essential to scientific research of the highest creativity is what Poincark described as 'our need of thinking in images"' (p. 222). Miller attributes to Aristotle the claim that "thought is impossible without an image" (1984, p. 263) and to Plato the view that

2

Margaret Jean Intons-Peterson

mental images are like impressions on a wax tablet which may be stored for later use. Much of the evidence for the role of imagery in creativity and discovery has been anecdotal, consisting largely of reports from creative persons about the processes contributing to their creative efforts (e.g., Arieti, 1976; Arnheim, 1969; Ghiselin, 1952; Miller, 1984, Roe, 1951; Shepard, 1978). These anecdotal observations make fascinating reading, and they offer insights into the role of imagery in both creativity and discovery, Nevertheless, the anecdotes do not answer some basic questions about the interplay of imagery, creativity, and discovery. For answers, we must turn to more systematic investigations. These investigations have taken two forms: correlational and experimental. The plan of this chapter is to chronicle some of the insights afforded by anecdotal observations,to identify some central issues, and to examine correlational and experimental evidence pertaining to these issues.

Anecdotal Observations Imagery has long been thought to contribute to creativity. For example, in his provocative chapter about the externalization of mental images, Shepard (1978)recounts instances in which chemists, physicists, biologists, musicians, artists, writers, architects, mnemonists, chess masters, mathematicians, geometers, and surgeons claim that images were central to their creative output. Some of the stories related by Shepard (1978)are well-known, such as Einstein’s realization, while imagining that he was traveling beside a beam of light, that the stationary spatial oscillation he %awl’ did not correspond to Maxwell’s equations for the propagation of electromagnetic waves, Kekule’s dream about a snake swallowingits tail, which led to his formulation of the benzene ring, Watson’s awareness of the structure of DNA from pairs of adenine residues that whirled before his closed eyes while napping, and Mozart’s claim that he could hear an entire musical work before recording any of it. Others are less well-known. While practicing golf swings in a dream, Jack Nicklaus discovered an error in the way he gripped his club. By correctingthis error, his game improved by 10 points almost immediately. The American sculptor James Surls reports imagining his sculptures. He is mentally able t o remove and add to various portions of a sculpture. In

Imagery's Role in Creativity and Discovery

3

a Regents' Lecture given at the University of California at Berkeley in 1976, Joan Didion described how "pictures in her mind" drove her novels. Clearly, these creative individuals believe that images inform, guide, and even illuminate their creative activities, corresponding to Wallas's third stage of illumination. These individuals also are noteworthy for being professionals in their respective areas, for having prepared themselves by virtue of their training and occupation. Most reports mention preoccupation with their topics. Thus, the reports lend credence to Wallas's (1926)fmt and second stages of preparation and incubation,as well. Finally, many talk about going over their creations or discoveries to be sure that they were correct and correctly transcribed, thereby supporting Wallas's fourth stage of verification. Poincar6 summed up the situation, when he said, "It is by logic that we prove, but by intuition that we invent" (Miller, 1984, p. 233). These delightful reports, and many others, are inspiring, but anecdotal, haphazardly or selectively collected. There are no checks to ascertain whether images preceded the creative moment or whether creative realizations unfolded in interaction with imagery. It is not clear how discovery occurs. That is, when does the creator become aware that a discovery has been made? How complete are most initial discoveries? Do they require substantial correction and modification? Most important, are images necessary for creative invention? This last question sparked a bitter debate in physics. Part of the lore of the field, Miller (1984) tells us, is that theoretical physicists should be able to intuit processes and solutions;they should be able to imagine them. Letters from Einstein, Bohr, Faraday, and countless others were cited as evidence. With the advent of the general and special theories of relativity and quantum mechanics, physicists began to experience difficulty in imagining processes with multiple dimensions. They found it hard, for example, to imagine simultaneous changes in time as a fourth (or more) dimension or to imagine something not previously experienced, although some people, including Einstein and Mozart, claimed to be able to do so. One camp of physicists, led by Heisenberg, argued that physicists would have to rely on mathematics because "there can be no directly intuitive geometrical interpretation because the motion of electrons cannot be described in terms of the familiar concepts of space and time" (Miller, 1984, p. 142). Fortunately, the work of Schrodinger,Dirac, Pauli, and Feynman showed otherwise. In effect, they returned visualization to a hallowed status in physics.

Margaret Jean Intons-Peterson

4

Nevertheless, it remains difficult to imagine simultaneously depth, movement, and time lapse. The ability to do so may be a characteristic that sets some gifted individuals apart from others. These anecdotal reports leave many questions unanswered, to be sure, but Shepard has extracted some challengingthemesfrom them. One set of themes has to do with characteristicsof creative individuals and the circumstances that may foster their creativity. Another set focuses on the interrelationships between imagery and creativity. Let us begin with the first set. Although these characteristics are not really our bailiwick, they are thought-provoking.

The Creative Person The characteristics that may aid the development of creativity may provide a framework for more systematic exploration of the imagerycreativity connection. Shepard (1978, p. 155) suggests, . that the genetic potential for visual-spatial creativity of a high order seems especiallylikely to be revealed andor fostered in a child (a) who is kept home from school during the early school years and, perhaps, is relatively isolated from age mates, as well, (b) who is, if anything, slower than average in language development, (c)who is furnished with and becomes unusually engrossed in playing with concrete physical objects, mechanical models, geometrical puzzles or, simply, wooden cubes. In addition, the inspiration to press relentlessly and concertedly toward the highest achievements that such a creativity makes possible may require the stimulus or model provided by a previous great thinker of a similar turn of mind. Thus Einstein may require his Maxwell and Maxwell his Faraday . . . At the same time, some of the factors that contribute to this kind of creativity may also carry with them an increased predispositionnot only toward some degree of dyslexia . . , but also toward the sorts of mental breakdowns, aberration, or even hallucinations that at one time or another afflicted several of the scientists we have mentioned, including Newton,

. .

Imagery's Role in Creativity and Discovery

5

Faraday, Cardan, Pascal, and Tesla . . . These are interesting speculations about highly gifted and creative individuals. They may or may not apply to a less select sample, who represent the general focus of this book. In brief, we are concerned with the imagery-creativity-discoveryconnectionsin typical human processing, although Anderson and Helstrup raise questions about some developmental aspects of imagery and both Kaufmann and Helstrup and Reed report research with trained architects, artists, or designers in their chapters in this volume.

Imagery's Role in Creativity More important for our purposes are reasons Shepard (1978) offers for the special effectiveness of mental imagery and spatial visualization in the creative process, his second set of themes. He identifies five reasons. The firstis that imagery and spatial visualization offer rich alternatives to the typical strictures imposed by language and traditional ways of thinking. As Shepard (1978, pp. 156)notes, "it seems reasonable that the most novel ideas and radical departures from traditional ways of thinking are not likely to arise within the very system ofverbal communication that is the primary vehicle for maintaining and perpetuating established ideas and entrenched traditions." The second reason is that the richness of imagery, and its relation to external sources, may suggest interactions and relations not fully preserved by language. Even an image whose structure parallels that of temporally bound verbal communication may permit more or different comparisons than descriptive language. Third, the nature of images makes them amenable for intuition and manipulation in ways that undoubtedly precede language development, both individually and evolutionarily. Our ability to avoid running into moving objects is an example. We must anticipate where the object willbe at some future time and arrange to avoid being there at that time. These abilities are displayed by all moving organisms, including nonhumans and preverbal children,whether they have functional use of a human language or not. Fourth, Shepard hypothesizes that images are more likely to engage affective and motivational systems than verbal productions. He argues that this is "why powerful emotions of fear, anger, and desire tend to be more strongly determined by the vividness with which one concretely

6

Margaret Jean Intons-Peterson

pictures the relevant object or event (a plane crash or a terminal illness,in the case of fear) than by the probability that one abstractly assigns to that event by verbal reasoning" (p. 157). A possible h a l factor is that the search for structural symmetries and invariances may be aided by the sensitivity of the visual system to symmetry. Drives toward symmetry strongly motivated the work of Einstein, Maxwell, and Watson, for example. Shepard's reasons for the special effectiveness of mental imagery in the creative process are addressed in one way or another in the chapters of this book. Shepard's reasons for the special effectiveness of imagery in creativity provide a usefil structure for systematicexploration of imagery in creativity and discovery. Let us begin this exploration by asking whether imagery contributes to creativity in principled ways.

The Imagery-Language Connection If, as Shepard suggests, one of the advantages of the use of spatial images is that images may escape some the constraints imposed by our customary and circumscribed use of language, we might expect verbal ability and imagery to be dissociable from each other. One test of this possibility is whether the two play similar roles in creativity. Evidence h m paper-and-penciltests offered by G a o r d (1967),Torrance (1966)and others typically yields positive correlations. These correlations are so persistent that Shaw and DeMers (1986, 1986-87) focused on children with intelligence test scores above 115 as more likely to show a relation between imagery ability and creativity than children with lower scores. Shaw and DeMers correlated scoreson imagery and creativity tests among fifih- and sixth-gradechildren. They used three of Torrance's (1966)tests to assess three aspects of creativity (fluency, flexibility, and originality). Each test contains two verbal forms and two figural (spatial) forms; thus, Shaw and DeMers assessed both verbal and imaginal creativity. They measured three aspects of imagery sometimes thought to be important (vividness,control, and memory), using Marks' (1973)Vividness of Visual Imagery Questionnaire (VVIQ),Gordon's (1949) Test of Visual Imagery Control (TVIC), and Shaw's Visual Memory Test (described in Shaw & DeMers, 1986, 1986-871, respectively. These tests were administered to 54 childrenwho scored above 115on a school-administered intelligence test and to 84 children who scored lower than 115.

Imagery’s Role in Creativity and Discovery

7

In general, creativity scores for originality and flexibility correlated more highly with imaginal vividness and control for the high-IQ group than for the comparison group, in concert with previous demonstrations of an association between intelligence and creativity. More interesting for our purposes is that the high-IQ group showed higher correlations for the figural (spatial) test of creativity with imaginal control (-61)than for the verbal creativity test (.49). The comparison group showed similar trends, but with lower correlations (.29and .18,respectively). Thus, some aspects ofvisual imagery seem to be more tightly associated with spatial creativity than with verbal creativity. However, the effects of verbal communication on spatial aspects of creativity cannot be unambiguously assessed from these data, for a number of reasons. The first is that the tests for spatial and verbal creativity correlate with each other, albeit nonsignificantly (.28, Shaw & DeMers, 1986-87). The second is that all of the measures used in the research contained a heavy verbal component. This heavy linguistic weighting militates against pure assessments of visuospatial ability. Indeed, it is perhaps surprising that the imaginal-verbal creativity correlations were not higher. The correlational approach has other hazards, including interpretation of the direction of the effects, cause-and-effectrelations, the psychometric properties of the various scales, and the absence of control over the use of imagery. Unfortunately, the tests also have questionable validity. Although some correlations have been reported between the creativity tests and individuals judged by teachers to be creative, no unassailable evidence has yet been produced. Further, paper-and-pencil tests of imagery have often been rather unimpressive predictors of imaginal performance (Paivio, 1986). For example, Peterson, Kihlstrom, Rose, and Glisky (1992) failed to find that either the VVIQ or the TVIC predicted imaginal performance. Self-report measures of imagery tend to be uncorrelatedwith performance, although they show modest associations with other self-report measures (Ernest, 1977). Occasionally more substantial correlations are observed (Finke & Shepard, 1986). Perhaps the most disturbing aspect of the correlational approach is that it does not examine the linkbetween imagery and creativity directly. That is, the assessments of each are administered successivelyand utilize different measures. Hence, there is no convincing evidence that components contributing to or invoked by the creativity tests also are involved in the imagery tests. In brief, we cannot assess how or if imagery

8

Margaret Jean Intons-Peterson

supports creativity from these data. An alternative way to determine the strength of the link between imagery and creativity is to manipulate the use of imaginal strategies by direct instruction. For example,Helstrup and Anderson (1991)encouraged their subjects to use either visual or verbal strategies to construct new objects in a mental discovery task. Subjects who were instructed to use visual strategies performed considerably better than those told to use verbal strategies,just as Shepard might have predicted. (SeeAnderson & Helstrup's chapter, this volume, for further details.) Unfortunately, the productions were not measured for creativity and hence, their results speak only to a link between imagery and discovery. Language might affect images, like percepts, by biasing subjects toward a particular interpretation. This interpretation may then be very difficult to dislodge, as Chambers and Reisberg (1985) showed. In their work, subjects were able to detect one version of an ambiguous figure (e.g., the ducwrabbit), but not the other form in their image, even though they were able to detect both versions from a subsequent drawing of their image. In this case, language may impose modification-resistant expectations on image generalization. Although the above evidence suggests that language may establish expectations for the shape and characteristics of an image, language also may facilitate expectations if it is used to challenge interpretations of images ( H p a n , this volume; Hyman & Neisser, 1991). Indeed, a number of authors of chapters in this volume (e.g., Chambers, Hyman, and Kaufmann & Helstrup, in particular) argue that language and description may influence visual depictions in myriad ways, ways that extend from interference to facilitation. As Chambers maintains, description shapes depiction. Hyman likens imagery construction to story construction. Imaginal gaps may be filled in, just as is true with stories. Kaufmann and Helstrup contendthat images are neither descriptionsnor mental pictures (images, visual depictions). Instead, they are hybrid mental concepts which embody both symbolic and perceptual properties. Furthermore, the opportunity to articulate may suppress imaginal transformations. Brandimonte, Hitch, and Bishop (1992) recently demonstrated that subjects were more adept at identifymgthe remainder of an image aRer a part had been subtracted mentally when they simultaneously engaged in articulatory suppression (saying "la,la")than when they did not. In other words, the suppression of articulation apparently aided the mental manipulation of the visuospatial material.

Imagery’s Role in Creativity and Discovery

9

The effect disappeared when the original images were difficult to name, as we might expect. With easily named images, subjects may rely at least partly on linguistic processes, unless discouraged from doing so by articulatory suppression. With difficult-to-name images, subjects presumably rely almost exclusively on imagery. Significant effects of articulatory suppression on imaginal subtraction occurred for images generated from long-termmemory,but not for images constructed in shortterm or working memory. These results suggest that language (naming) of images may reduce the effectiveness of imaginal manipulations. Note that the evidence suggests that although language may establish expectancies which, in turn, guide either the generation or the interpretation of images in inappropriate or overly constrained ways, the data considered above do not indicate that language activelyinterfereswith the generation of creative images. The research by Shaw and DeMers (1986, 1986-87) implies that visual imagery is more closely associated with spatial than with verbal creativity. These different relations could well have prompted the nonverbal-verbal distinctions drawn from his anecdotal reports that led Shepard (1978) to conjecture that the development of creativity may be fostered by some isolation from traditional language strictures and by the opportunities and commitment to absorb oneself in spatial activities (see the previous quotation from Shepard, 1978, p. 155).

Special Characteristics of Imagery In this section, I combine Shepard’s second and third reasons for the special effectiveness of mental imagery in creativity and discovery, the ideas that imagery‘s richness may deliver interrelationships less apparent in language and that the nature of images makes them particularly amenable to integration and manipulation. Anecdotal Reports Anecdotal reports are strongly supportive. One particularly striking example comes from the inventor Tesla, who claimed that, aRer a few weeks, he inspected his image of a physical machine (e.g.,the three-phase electrical distribution system, the self-starting induction motor) for signs of wear (Shepard, 1978). A hallmark of these reports is that the images

10

Margaret Jean Intons-Peterson

represented novel combinations or relations not previously experienced.

Assemblage and Interpretations of Images from Verbal Input Experimentalevidenceoffers a similarmessage ofnovelcombinations within images. For example, young adults are mentally able to construct, synthesize,and manipulate the componentsof an image that are described to them verbally (e.g., Anderson & Helstrup, in press, this volume; Denis & Cocude, 1989;Finkey1990, this volume; Finke, Pinker, & Farah, 1989 Finke & Slayton, 1988;Helstrup &Anderson,1991;Intons-Peterson,1981, 1984, this chapter; Klatzky & Thompson, 1975; Roskos-Ewoldsen, 1989, this volume). Finke et al. (1989) asked subjects to mentally combine and manipulate alphanumeric characters and geometric shapes to form an object defined by the experimenters. Their imagers were able to discover the experimenters’ pre-experimentally determined forms. Strictly speaking,these researcherswere documenting the contributionof imagery to the discovery of forms because the tasks were to identi@predetermined recognizable (i.e., nameable) shapes. In this research, subjects usually hear phrases or sentences that describe the components to be included in the image. Mental rotation or other transformations also may be specified by the experimenter. Subjects are able to identi@ the target image correctly from a set of alternatives or to draw the target accurately. Roskos-Ewoldsen moved toward the detection of both predetermined and emergent forms in her dissertation (1989; also see her chapter in this volume). She found that her observers were able to detect some novel combinations when figures had been imaginally constructed. These abilities also can be harnessed to produce useful, nameable objects (Finke, 1990,this volume)or recognizableobjects (Finke & Slayton, 1988). For example, Finke and Slayton’sundergraduates, unselected for intelligencetest performance or other abilities, were told to combine three components, drawn randomly from a set of 15 forms (circle, square, triangle, rectangle, horizontal line, vertical line, and the alphanumeric characters L, T, C, D,J, 8, X, V, and P), to form and name the object. The objects drawn and their names then were judged for correspondence on a scale from 1 to 5. Those judged to have high drawing-name correspondence (scores of 4 or 5) also were classified as creative or not. Finke and Slayton found that about 38% had high drawing-name correspondence and of these objects, 16%were judged as creative. These

Imagery’s Role in Creativity and Discovery

11

creative patterns were rarely predicted by a naive experimenter or by another group of 70 undergraduates. This latter group even failed to predict many of the creative responses they themselves gave when tested later on the imagery task. Apparently, syntheses emerge from imaginal processes that are not immediatelyobviousfrom a surveyof the component parts. Note a potential definitional problem with this task, however. Subjects were asked to imaginally construct an object that could be identified by a briefverbal description. This task may bias subjectstoward identifying something already known and labelable, a near contradiction in terms for the novelty often assumed to be essential to creativity. Fortunately, ample evidence now shows that navel images often are reported, thereby reducing objections to this aspect of the procedure. For example, Finke and Slayton’s subjects and the participants in my experiment generated items judged novel and creative by independent judges. In addition,Anderson and Helstrup (in press, this volume), using similar procedures,found that their subjects generated as many novel and creative items (again, as judged by independentjudges) when they relied solely on the mental combination of parts as when they were able to augment their mental efforts by being able to draw the combinations. Nonetheless, it seems reasonable to view this work as imaginal discovery, as Finke (1990) does in his book and in his chapter in this book. In his book, Finke (1990) extends these efforts, asking subjects to generate useful objects from three components and to then describethe use of the object. In some cases, subjects generated two-dimensional objects; in other cases, they generated three-dimensional ones. F’inke’s book (and his chapter, this volume) are replete with interesting and creative objects generated by his subjects. In general, the results were similar to those obtained in the Finke-Slayton (1988)research. All of this work provides compelling demonstrations of the abilities of unselected adults to synthesize and manipulate mentally components of images in experimental tasks that assume the use of imagery. Typically, the likelihood of using imagery has not been manipulated, except by using control groups to whom no mention is made of the use of imagery. That is, the technique does not permit evaluation of the frequency of production of objects judged creative as a function of the likelihood of eliciting imagery. It seems reasonable to assume that imagery was used because the subjects needed to translate the verbal labels (names) of the components into a spatial array (image?) of the

12

Margaret Jean Intons-Peterson

mental objeds) depicted in the drawings. UsingFinke's general approach as a springboard,Wendi Russell and I hypothesized that if imagery contributes to creativity, creative productions should increase with increases in the likelihood of using imagery, Our approach was to devise a slightly more difficult task which was undertaken in conditions designed to manipulate the induction of imagery. The task consisted of the presentations of four (rather than three) simple componentswhich the subjectswere to combine into a useful object. Three conditions were devised to manipulate the likely use of imagery: picture, printed word, and verbal. In the picture condition, a task that should be least likely to involve imagery, the subjects saw pictures of the components. They could look at these pictures from various perspectives and thus be able to integrate the componentsvisually without having to imagine them. The task should involve more imagery if subjects read the names of the four components, as in the printed word condition. In this condition, subjects would have to rely on the visual system, with which imagery is supposedly closely linked (e.g., Finke, 1990),to read, interpret, and then retrieve the meanings of the names from long-term memory of the components they had to form into an image to perform the task. Similarly, in the verbal condition, where subjects heard the names of the components, as they had in F'inke's (1990)experiments, the subjects would have to interpret the names, retrieve information from long-term memory and convert this information into an image. The verbal condition also conferred the advantage that potential "input interference" from having to look at the stimuli would be reduced; hence, the visual system might be more free to use imagery. Accordingly, we had three maingroups of 20 subjects each. One saw pictures of the components to be combined into a u s e l l object. Another saw the printed names of the components, and the third heard the names of the components. On each trial, the participants were given two geometric figures, sampled randomly from the set (square, triangle, circle, spring, line, and crescent) and two capital letters, sampled randomly from the set (C, K, L, N, S, and V). The two figures and two letters were seen as pictures, as the printed names of the components, or heard as the names of the components. Of course, the printed letters and their pictures are both pictorial, but the two other components differed in these conditions. The subjects were to construct a novel, but potentially usefid object from the four components. ARer they construcM the object,using as much time as they needed, they drew it on a response sheet and then wrote a

Imagery’s Role in Creativity and Discovery

13

brief description of its purpose. No drawingwas permitted until they drew the final object. Our predictions were that, if imagery contributes to creativity, subjects in the picture condition should produce fewer creative objects than subjects in the printed word condition, and subjects in the verbal condition should produce the most creative objects. It is possible, however, that subjects need to learn how to cope with thisnovel task. With the threecomponent task, Helstrup and Anderson (1991)found that performance improved from the first to the second block of trials regardless of whether subjects were able to augment their compositionsby drawing. Finke (19901,however, observed no evidence of practice or familiarity effects with his three-component task. To test the possibility of a practice effect over an extended delay, we brought all subjects back to the laboratory two weeks later. The subjects worked on seven problems during the first session and on fourteen,during the second session. Because seven of the problems presented during the second session duplicated the seven problems from the first session, this design enabled us to test subjects with both the same and Merent problems. This approach should indicate whether subjects who devise creative solutions to some problems continue to do so with other problems and whether subjects do or do not produce the same solutions to problems repeated later (but in a new randomized order). As with Finke and Slayton (1988),we had threejudges independently rate the correspondence between the drawings of the objects and their descriptions, using a 5-point scale that ranged from 1 =very poor correspondence to 5 = excellentcorrespondence. Ifa pattern was rated as having high correspondence (rated as 4 or 5), the judge then rated it for creativity on a 5-point scale. Patterns were considered to have high correspondence and to be creative if they were rated as 4 or 5 on both scales by at least two of the three judges. In Session 1, 130 of 420 (7 problems x 60 subjects) or 31%of the opportunities yielded corresponding drawingdescriptions (rated 4 or 5). Presenting the components as pictures produced about the same numbers of recognizable drawingdescriptions (39 of 140 opportunitiesor 27.9%) as printed words (30 of 98 or 30.6%)or as verbally presented words (48 of 140 or 34.3%). Apparently, when subjects first begin thisnovel task, the mode of presentation of the four components does not substantially &ect the percentages of recognizable objects produced. But were the objectsjudged to be creative? Of the 130drawings that

14

Margaret Jean Intons-Peterson

corresponded to the descriptions,43 or 33.1% werejudged as creative. The percentages of drawingdescriptionsjudged as creativewas about the same for pictures (12of 39 = 30.8%) as for printed words (15of 43 or 34.9%) and verbally presented words (16 of 48 or 33.3%). Hence, the mode of presentation, which presumably manipulated the level of imagery, did not yield differential numbers of objects judged as recognizable or as creative during Session 1. These results are depicted in Figures 1and 2. Although our subjects seemed to understand the task and to respond reasonably quickly, their drawings during the first session corresponded to the descriptions on only about one-third of the trials. Of these, slightly more than one-third of the productions were considered creative. These relatively low frequencies also were found by Anderson and Helstrup (in press) and by F'inke and Slayton(1988).For example, F'inke and Slayton's subjects produced 38.1% recognizable patterns, of which approximately 15% were considered creative. Thus, this level of productivity may be typical of people unselected for their imaginal or creative abilities on their initial exposure to the task. Let us consider Session 2 responses, obtainedtwo weeks after Session 1. These responses reflect some experience with the task. The Session 2 data were separated into the old and new sets. Overall, the percentages of descriptions that corresponded to the drawings declined from Session 1for both the old and new sets of Session 2. The decline was from 35% on Session 1 to 21.9% (92/420)for old sets and 24.3% (10W420)for new sets on Session 2. All three groups displayed this decline, as shown in Figure 1,but the drop was significant for only printed words. These results suggest that the presence of printed words may have a suppressingeffect, given enough exposure. I will return to the Session 1-Session 2 differences aRer considering our major focus, the effects of likely recruitment of imagery on the number of corresponding productions that were also judged creative. During Session 2, the proportions of productions judged creative increasedreliably as the assumed likelihood of elicitingimagery increased, and this result appeared for both old and new sets. The percentages of creative responses for the presentation groups, in order of assumed elicitation of imagery (picture,printed word, verbal presentation),were 12, 21,and 26 for the old sets and 6,11,and 31 for the new sets. Thus, in Session 2,but not in Session 1, more items were judged creative in conditions designed to elicit more imagery (the printed words and verbal presentation) than the condition expected to elicit less

Imagery$ Role in Creativity and Discovery

15

imagery (the pictures). Three samples of creative responses appear in Figure 3. Another attribute often ascribed t o creativity is its versatility, That is, different images may be created on separate occasions from the same or highly similar components, a concept similar to Anderson and Helstrup's (1989, in press; this volume) "productivity." Versatility was instantiated in the current paradigm by repeating the sets from Session 1 as part of Session 2. The principle of versatility predicts that, even though seven sets were identical on the two sessions, the drawings and descriptions created would vary. That was true, although subjects were told that some sets might duplicate those from Session 1. The percentages of the 140 drawings from Session 2 that were rated as very similar or identical to those from Session 1 were 5.7, 10.7, and 7.9 for the pictures, printed words, and verbal conditions. The comparable percentages of descriptions rated as very similar or identical were 7.1,8.6, and 5.7. ObviousIy, identical components elicited different drawings and descriptions on most of the trials, supporting the principle of versatility. Our subjects were likely t o construct different objects from the same sets of components, and presentation condition had essentially no effect on versatility of the creative output. The next issue focused on individual differences in creativity. Are some people more creative than others? We explored this issue by tallying the numbers of individuals who gave at least one creative response to the seven old items and t o the seven new items on Session 2 as a function of whether they had given at least one response judged creative in Session 1 (see Table 1). These tallies were conducted separately for the three conditions. The pattern was very clear for the picture and printed words groups. At most two of the subjects giving at least one creative response during Session 1 also made a response judged creative during Session 2, whether the patterns were old or new. Thus, for the subjects in the picture and printed words groups, we found no evidence for a transfer effect or for generality of creativity from one set of patterns to another. In contrast, 50% of the subjects in the verbal presentation group who gave creative responses during Session 1 also gave creative responses t o old patterns during Session 2, and 70% gave

Margaret Jean Intons-Peterson

16

Proportions of Correspondlng Drawings and Descrlptlor

?

?

? A

0

w

N

?

P I

'Ictu res

Wnted Words

Jerbal

Pictures

Printed Words

Verbal

Pictures

Printed Words

Verbal I

Figure 1 Proportions of drawings rated as corresponding to the descriptions for the three conditions of presentations of sets in Sessions 1and 2. For Session 2, half of the sets repeated Session 1's sets (old) and half were new.

17

Imagery’s Role i n Creativity and Discovery

Proportions of Corresponding Drawings and Descriptions Rated as Creative 0 ~~

0

0

A

ka

~

0

tc

0

i

I

Pictures Printed Words Verbal

Pictures Printed Words Verbal ~~

Pics. Printed Words Verbal

Figure 2 Proportions of drawing descriptions judged as corresponding as highly creative.

Margaret Jean Intons-Peterson

18

DescriDtion of I n t e n d e d

Fisure

‘s(

A letter

bubble-blower.

C

T h i s i s a f u t u r i s t i c microphone s t a n d i n which you speak i n t o t h e S t o amplify your v o i c e t h r o u g h t h e V shaped part of t h e device.

-

t h e V i e an antenna p i c k i n g up a n t s i g n a l s , and when it f i n d s o n e t h e C acts as a c l a w and stores them i n the triangle.

A a n t catcher

Figure 3 Samples of drawings of objects.

Imagery's Role in Creativity and Discovery

19

TABLE 1 Numbers of subjects who made a t least one response judged Creative (c) and Not Creative (N). Pictures (N = 20)

Printed Words (N = 20)

Verbal (N = 20)

Session 2 Old

Creative

C

N

C

N

C

N

0

7

2

8

5

5

4

9

2

8

3

7

Session 1 Not Creative

Session 2 New

Creative

C

N

C

N

C

N

2

5

0

10

7

3

0

13

2

8

2

8

Session 1 Not Creative

creative responses to new patterns. These results suggest that the consistency of giving creative responses may be enhanced by use of a condition that fosters imagery, our verbal presentation of words. All subjects were asked, "How did you go about constructing an object from the figures?" The responses generally divided into two categories. The first (imaginal) one involved mentally manipulating the components. The second (perceptual) one consisted of drawing figures and then trying to interpret the drawing. Answers to this question constitute a type of manipulation check. The imaginal strategy was claimed by fewer subjects in the picture condition (27%) than in the printed words (50%) or verbal (54%) conditions, as we had predicted, even though no subjects were told to imagine the components or the objects. In summary, consistent with the view that the use of imagery

20

Margaret Jean Intons-Peterson

aids creativity, our evidence suggests that during the second session with the task, imagery-encouraging conditions fostered creative synthesis more than an imagery-neutral condition. This effect required some exposure to emerge, for it was not apparent during the first session. Imagery-encouragingconditions had little differential effect on the numbers of the production of recognizable objects produced during either session. The most reasonable explanation for the absence of an effect on the production of recognizable objects from manipulations of imagery-inducing conditions is that participants in the printed words and verbal conditions had ample time to activate representations of the figural referents of all components in a manner similar to that perceptually available to participants in the picture condition. Now, consider the effect of sessions. There are two issues: Why did the percentage of descriptions corresponding to the drawings decline over the sessions where the percentage of productionsjudged creative increase? We originally proposed that experience with the task would aid generation of corresponding productions. Clearly, that was not the case. The changes in both measures are at odds with the failure of Finke (1990) to find practice effects. There are two major differences between our approach and his. The first is that we used a more difficult (four-component)task; the second is that our sessions were separated by two weeks, whereas no delay was interposed in the Finke (1990) research. The use of a more difficult task cannot be the sole answer because Helstrup and Anderson (1991) found an increase in creative productions from the first to the second half of a single session, even though they used the same kind of three-component task as Finke. The Helstrup and Anderson (1991) results also suggest that the length of the delay is not a critical factor. Consequently, we considered another possibility: Post-experimental reports of the subjects had suggested a motivational explanation. The participants complained that they became bored with the repetitiousness of the task. The two-week delay also may have contributed to a decline in motivation. Our subjects were using experimental participation as one way to fulfill some course requirements, and they may not have been happy about the requirement. This possibility could explain the decline in the generation of corresponding productions. An alternative to this rather uninteresting possibility is that

ImageryS Role in Creativity and Discovery

21

exposure to (and increasing familiarity with) the task suppresses Note that the decline in the generation of performance. corresponding productions from Session 1to Session 2 emerged for the new sets on Session 2, as well as for the old sets. Hence, exposure to or familiarity with the task may function as a suppressor. This explanation can be extended to address the differential effect of sessions on creative synthesis if we assume that the suppressing effect of familiarity is ameliorated by the use of a condition that strongly encourages the use of imagery, such as verbal presentation of the components. The partial release from suppression occurs, we speculate, because the use of imagery facilitates varying combinations of components even within a now familiar task. Specifically, the imagery-encouraging conditions (printed words and especially verbally presented words) appeared to buffer the substantial decline in recognizable and creative objects seen in all conditions during Session 2. This decline was particularly pronounced for the pictorial condition, which presumably required less imagery than the other conditions. These experimental results offer support for the notion that creativity is fostered by taking novel approaches to tasks. Another striking result was the presence of substantial variability both between and within subjects. No subject gave recognizable and creative responses to all sets. Further, the same subjects rarely produced highly similar objects to the same sets presented two weeks apart. This was true, even when subjects claimed to have used imagery systematically. Finally, our subjects appeared to have little difficulty mentally arranging and rearranging independent components. These components could have been easily dissembled because they were not part of a cohesive reference frame (see chapters by Anderson & Helstrup, Finke, Hyman, Kaufmann & Helstrup, Peterson, and Roskos-Ewoldsen in this volume, and Peterson et al., 1992). These results imply that Finke (1990,this volume) has chosen an effective imagery-inducing procedure, one that is likely to yield creative syntheses.

22

Margaret Jean Intons-Peterson

Interpretation of Perceptually Ambiguous Images Collectively, the research described above suggests a reasonably facile manipulation of mental images to create or discover novel For example, combinations; other evidence is contradictory. Chambers and Reisberg (1985,1991) have found that images of ambiguous figures cannot be reinterpreted (reconstrued). Their work and related research ask questions about the relation between imagery and sensation-perception. Doubts about the ability of subjects to discover components in their images motivated Reed (1974)to ask whether subjects are able to discern target parts within their images. He found that subjects were not very good at identifying parts of imaginal wholes. For example, in one experiment, subjects saw a pattern for one second, say a Star of David. This was followed by a blank field, and then a test figure that may or may not have been an embedded part (e.g., a triangle or a hexagon) of the original pattern appeared. Subjects judged whether or not the part was in the original pattern. They had no difficulty identifying the entire pattern, but did have problems with some parts. In this experiment, Reed found that 68% of the subjects reported using imagery to perform the task, 12% said that they used both imagery and verbal descriptions, and 20% claimed to use only verbal descriptions. Reed interpreted these results as indicating that images are "structural descriptions," a form of propositions. Additional work also was conducted (see Reed & Johnsen, 1975,for example), but the primary messages were that (a) there was substantial variability in the detection rates of parts of the images and (b) parts were not necessarily "visible" in images in the same way that they are in pictures. The contention that images may differ from perceptual counterparts may extend to reinterpretation processes, as well. For example, certain ambiguous visual forms tend t o reverse as they are viewed. One example is the ducWrabbit figure. Chambers and Reisberg (1985)argued that, if imagery supports the same kinds of processes as perception, reinterpretability should emerge with imaginal ''reversible" forms. That is, the same kinds of components should be discoverable in images as in percepts. In a series of experiments, they tried to obtain imaginal reversals of the ducW rabbit and other classical ambiguous figures. They failed each time.

Imagery's Role in Creativity and Discovery

23

However, when the subjects drew their images of the ambiguous figures, most immediately saw t8hereversed figure. Investigations of auditory images yielded similar results (Reisberg, Smith, Baxter, & Sonenshine, 1989). Peterson et al. (1992; also see Peterson, this volume), offer a reconciliation of Reed's (1974)and of Chambers and Reisberg's (1985) failure to find reconstrual of images and the previous research demonstrating the ability of students to create and manipulate their images mentally. Peterson et al. distinguish between two types of reversal. The first type, reference-frame reversals, entails a change in reference-frame, such as a top-bottom reassignment, in addition to a reconstrual of image components. The second type, reconstrual, refers only to reorderings of the components to form a new interpretation, or reconstrual. They note that reinterpretation of Chambers and Reisberg's ducWrabbit would involve both referenceframe reversals and reconstruals of the components. If, as Finke et al. (1989) suggested, reconstruals in imagery occur more readily than reference-frame reversals, then we would expect imagers to have more difficulty reinterpreting figures whose reversal depends on alterations of the reference frame (e.g., Chambers & Reisberg's duck/ rabbit) than reinterpreting figures whose reversal rests on reorganization of parts (e.g., Finke et al., 1989; Roskos-Ewoldsen, 1989, this volume). These arguments led Peterson et al., (1992) to substitute a familiarization stimulus whose reversal required both alteration of the reference frame and reconstrual of the components (Tinbergen's, 1948, goosehawk figure) for Chambers and Reisberg's Mach book. This substitution was made because the Mach book reversal depends on reference-frame reversal, but not on reconstrual of the interpretation of the components. This modified procedure yielded reference-frame reversals in images of the duckhabbit figure, although these reversals were rare compared to the frequent reconstruals of images. In addition to seeing, and then imagining, the entire duckhabbit stimulus, they also tested conditions in which the observers imaginally constructed the stimulus from "good" parts that were divided by minima in curvature or from "poorll parts that were divided in ways that did not correspond to minima in curvature. The images constructed from good parts were more likely to reverse than those constructed from poor parts.

24

Margaret Jean Intons-Peterson

These results suggest that imagery might be differentially sensitive to different kinds of mental manipulations, with imagery being generally more responsive to reconstruals and reinterpretations of parts than to changes in overall organization. Peterson et al. (1992,this volume) propose that, at an early stage, the structural description of a shape is not yet connected to an interpretation. This connection might not yet have been made in the case of recognition, or a new one might not yet have replaced a previous one, in the case of reversal. Shape memory is searched for a best-fitting representation to serve as the interpretation for the structure. Exhaustive searches of representations with similar reference frames precede searches of representations with other reference frames. Matches found in the same reference frames are reconstruals; matches found in other reference frames, occasioned by failures to find matches in the same reference frame, are reference-frame reversals. These search strategies deliver simple reconstruals more often than reference-frame reversals. Although these processes might explain why reconstruals dominate in imagery, they do not explain why reversals of shapes such as the ducWrabbit dominate perception. Peterson et al. (1992, this volume) argue that this difference might be explained by differences between perception and imagery. From this perspective, the details of a figure present during perception might subvert or eclipse reconstruals and direct the search toward representations with different reference frames. Imaginal searches are less likely to be deflected from same-reference-framesets because images contain fewer diverting details. A second explanation is that naming or other verbal directions about processing of visual images may overcome difficulties in interpretation. Hyman and Neisser (1991;also see Hyman,this volume) used a procedure similar to that of Chambers and Reisberg (1985). They reasoned that subjects would be more likely to reinterpret the image if they were asked questions that encouraged them (a) to refocus on the orientation of the image, (b) to make a categorical classification of the image, or (c) both. These questions would induce reinterpretations of the image, thereby allowing imaginal reconstrual. Their data supported this view. These results, and similar ones by Peterson et al. (1992,this volume) are intriguing because they reveal the role language can play in guiding the search

Imagery's Role in Creativity and Discovery

25

of visual images. Language, both that of the experimenter and selfimposed restrictions, can establish powerful expectations for the shape and components of an image. Language also may facilitate overcoming these linguistic strictures, if used adroitly and appropriately to alter initial expectations. Chambers, Hyman, and Kaufmann and Helstrup all use such findings to argue in their chapters that language and description may influence visual depictions. In brief, there can be no doubt that language constrains visual imagery. A third explanation is that the transitory nature of images requires their regeneration for an extended search. There may be time for a search for reconstruals to occur before the images must be refreshed, but not for searches of different reference-frame sets. If the image is not refreshed, for some reason or other, then the opportunity for a more extended search also is lost. Perception presumably is not as restricted under ordinary circumstances which permit viewing until a match is achieved. Of course, limited glimpses of the items might also produce more reconstruals than reference-frame reversals. A somewhat different, but complementary explanation has been offered by Reisberg and his colleagues (Chambers & Reisberg, 1992; Reisberg & Logie, this volume; Reisberg et al., 1989; Smith, Reisberg, & Wilson, 1992). According to their perspective, perceptions arise from actual stimuli and are subject to interpretation and reinterpretation, whereas . . . mental images in any modality have no existence outside our understanding of them, making the image and its comprehension inseparable. In perception, there is a physical stimulus, existing independent of the perceiver, which needs interpretation. However, in imagery, there is no free-standing icon waiting to be interpreted, and no interpretation is needed to learn what the image depicts" (Reisberg et al., 1989, p. 620). Their view is that "pure" auditory images are unambiguous, and the meanings of the images (the interpretations of the images) are immediately clear. Pure auditory images may be "enacted1'by subvocalizing the image. In this case, the auditory image will support more interpretations and yield results more like those of perception. Tests of this view have used both auditory and visual ambiguous stimuli. In the auditory case, traditional switching between, say, "stress" and ''dress" occurred when "stress" was repeated aloud.

26

Margaret Jean Intons-Peterson

When imagery subjects subvocalized, their performance approached that of the perceptual (aloud) group, as expected. In the visual case, Chambers and Reisberg (1992) argued that the initial interpretation of an image affects subsequent analyses of that image. For example, with ambiguous images, the initial interpretation may omit information not relevant to the interpretation. If so, imagers would not be able to detect deviations inconsistent with their interpretation. To evaluate this prediction, they asked subjects to imagine the duckhabbit figure while under directions to imagine a duck or a rabbit. Next, the subjects had to identify one of two pictures that most closely resembled their image. The idea was that if the subjects imagined a duck, they would interpret the original figure in terms of "duchess" and generate an image with a clear left side (the duck side of the ducWrabbit figure), or what they called the "duck face." In contrast, if the subjects imagined a rabbit, they would interpret the original figure in terms of "rabbitness" and generate an image with a clear right side (the rabbit side of the figure), or what they called the "rabbit face." The two pictures presented as a two-alternative forced-choice test showed the original figure paired with a figure with variations in either the duck face or the rabbit face. Subjects who had interpreted the figure as a duck and presumably had a clear image of the duck face should be able to correctly identify the original when it was paired with the figure with a slight modification of the duck face. The images of the ''duck'' subjects would be less distinct on the rabbit side. Hence, these subjects should not be able to distinguish between the original and a slight modification of the rabbit side. The opposite results would hold for subjects told to imagine a rabbit, The results supported these predictions, even when the imagers were told to reinterpret their image as a rabbit after having first imagined it as a duck, and vice versa. Apparently, the cognitive interpretations (construal) suggested to the subjects affected their images and their subsequent abilities to compare the original stimulus with a slightly modified one. Chambers and Reisberg conclude that their work demonstrated "that image construal guides what is and what is not included within the image." I agree with the general notion that image construal guides image generation, but not with their argument that their data tell us about what was not included within an image. Selection on a two-

Imagery's Role in Creativity and Discovery

27

alternative forced-choice task informs us that subjects are or are not able to differentiate between the two stimuli. When subjects are not able to differentiate, the task does not indicate why. Chambers and Reisberg postulate that the "backsides" of the images were deficient or even that the information was missing. These interpretations are not warranted. It may be that the subjects had substantial information about both the original stimulus and their images, but they could not distinguish between the two test stimuli. Or they might have forgotten the original and lost the image. In either case, we would expect chance-level performance. This issue becomes important because Chambers and Reisberg try to explain how subjects were able to "fill inttinformation when told to reconstrue the image. Chambers and Reisberg conclude that, as Kosslyn (1980, 1983)proposed, components may be missing from a pixel-like image in working memory that still is available in long-term memory. Before accepting this conclusion, I would want more evidence that items were, in fact, lost. One possibility is to use more complex figures, ones that allow for the inclusion or exclusion of specific components. Some of the reversible figures in Shepard's (1990)book, Mind Sights, are possibilities. Figures with or without specific features then would be presented for yesho identity judgments. This approach should disclose whether and what kind of featural information is lost from the images. Integrating over various findings provides some evidence for Shepard's second contention that the special richness of imagery may facilitate creativity and discovery. When we generate images from scratch, so to speak, by constructing them from individual components, we seem to build images with reference frames, both in the sense that the images are guided by our linguistic concepts and in the sense that the images have specific orientations, outlines, or configurations. Reference frames - the anchors of our perceptual world - appear to constrain our mental gymnastics and, perhaps, our creative discoveries. These topics emerge in most of the chapters in this volume. Specifically, imagery seems t o support more creative responses when unencumbered by simultaneous instructions to use a verbal strategy (Anderson & Helstrup, this volume; Helstrup & Anderson, 1991). Likewise, imaginal search appears to aid internal reorganization of the images (reconstruals; Hyman, this volume;

28

Margaret Jean Intons-Peterson

Roskos-Ewoldsen, this volume), particularly when reconstruals are interpreted literally (Kaufmann & Helstrup; Peterson, this volume). Imagery and its kin, pictures and diagrams, may foster efficient computations (Reed, this volume). Finally, images foster rich creative productions (Anderson & Helstrup, Finke, Intons-Peterson, all this volume) that exceed those sponsored by pictures or printed words, given some experience with the task as I described earlier in this chapter. Evolutionary and Developmental Precedence of Imagery over Language The use of spatial imagery or spatial knowledge to guide purposive, unimpeded locomotion has been noted by Mandler (1983). These investigators note that the ability to use spatial knowledge precedes language development (see Anderson & Helstrup, this volume). Similarly, monkeys respond to mental rotation in ways that resemble those of humans, although they do not speak in the usual sense (Georgopoulos, Lurito, Petrides, Schwartz, & Massey, 1989; also see chapter by Reisberg & Logie, this volume). This limited evidence supports Shepard's claim for imaginal precedence over language. The problem is obvious: These studies are handicapped by an inability to assess language in preverbal children and nonhuman animals. Motivational Aspects of Imagery Shepard argues that images are more likely to engage affective and motivational systems than verbal productions. This may well be true for creations in the arts and related fields, but evidence from psychology appears to be limited to two experiments. One experiment, the work I reported earlier in the chapter, suggested that boredom with spaced practice reduced the number of objects generated from four components, but this suppression was partly alleviated when the mode of presentation of the components encouraged the use of imagery. In other words, exposure to the task may suppress creative integration, unless countered by

Imagery's Role in Creativity and Discovery

29

encouragement to use imagery. Perhaps creativity is cultivated by adopting novel perspectives, such as those afforded by imaginal manipulations. Although these speculations intrigued us, the research itself, was not designed to investigate the motivational aspects of imagery. Neither was research by David Roskos-Ewoldsen and Jeffrey Franks (personal communication). Nevertheless, their work also suggests that imagery may engage affective systems. Their subjects judged 48 items for ease of imagining. Half of these items (24) then were included among 60 items presented for affective judgment, and the other half were included in another set of 60 items presented for animateness judgments. The items were counterbalanced across subjects. The assumption was that affectivejudgments would trigger affective reactions to the words, but animateness judgments would not. Finally, recognition of all 120 items was tested. Items rated as high in both affect and ease of imagining were recognized significantly more often than items rated as high in affect but less easy to imagine or items rated for animateness, regardless of the ease of imagining. These results do not disclose the underlying mechanisms, nor do they indicate that the subjects actively tried to imagine the items before the items were rated for ease of imagery. These caveats notwithstanding, the results implicate a positive connection between imagery and affect. Hence, this area stands as a challenge for the future.

The Search for Structural Symmetries Shepard's last possibility is that searches for structural similarities and invariances may be helped by visualization. Indeed, Roskos-Ewoldsen(1989,this volume) found that "good"constructions, which often were symmetrical, facilitated detection of emergent patterns. Hyman (this volume) reported that canonical (usually symmetrical) objects were easily reproduced both when the subjects drew them from memory and when the objects were visible, compared to simple or complex asymmetrical objects. The reverse was true when the task was to organize the objects into a single, cohesive whole. In the latter case, complex, asymmetrical objects were the easiest to organize, followed by the simple components, with

30

Margaret Jean Intons-Peterson

the canonical ones, the hardest. Reed (1974)also found symmetry to be beneficial. The notion may be extended to natural discontinuities, such as Peterson et ale's (1992,this volume) changes in curvature. The situation is expanded further by Roskos-Ewoldsen's finding of better detection of emergent components when the pattern constructed from components was poor (typicallyasymmetrical) than when the pattern was judged as good (typically symmetrical). The cohesiveness of good or symmetrical patterns may function like a reference frame in imagery, thereby retarding the necessary decomposition of the pattern for the detection of emergent elements.

Individual Differences and Training Images thus seem to play a special role in creativity and discovery, as Shepard surmised. This possibility raises the interesting question of whether the use of images in creative discovery can be taught or is the special province of a few gifted individuals. Evidence from Anderson and Helstrup (in press, this volume), Finke (1990,this volume), Intons-Peterson (this volume), Intons-Peterson and Roskos-Ewoldsen (1989),and Roskos-Ewoldsen (this volume) suggests that normal people can do it. Can they be taught to become more creative? Can creativity be modeled by computer simulation as a form of problem solving, as Simon and his colleagues maintained (Langley,Simon, Bradshaw, & Zytkow, 1987)? In their theory of scientific discovery, Langley et al. construe discovery as problem solving, saying, "A hypothesis that will be central to our inquiry is that the mechanisms of scientific discovery are not peculiar to that activity but can be subsumed as special cases of the general mechanisms of problem solving" (1987,p. 5). They devised computer simulation models of problem solving based on the assumptions that the human brain is an information processor "whose memories hold interrelated symbol structures and whose sensory and motor connections receive encoded symbols from the outside via sensory organs and send encoded symbols to motor organs" (1987,p. 8). In brief, problems are defined as problem spaces, and the task, as goal satisfaction. The solution uses operators (heuristics or algorithms) to search the space. Their treatment of imagery is to

Imagery's Role in Creativity and Discovery

31

assign a form of representation (called pictorial or imaginal) to an ''internal representation when the information it contains appears similar, and similarly organized, to the information in an external picture or diagram, and when the inferences that can be drawn from it rapidly and effortlessly are similar in kind to those that can be drawn immediately from a picture or diagram" (1987, p. 327). The simulation models show promise, but the authors have not yet extended their tests to people, so we cannot really evaluate this approach. It is important to note, however, that Langley et al.'s (1987) definition of imagery seems too narrow to accommodate the kinds of abilities already discussed. In his chapter in this volume, Reed describes another approach to training, the use of an imaginal extension of analogical problem solving to new situations. Bassok and Holyoak (1989)' Gick and Holyoak (19831, Hayes and Simon (1976), Hinsley, Hayes, and Simon (1977))Krueger (19761, Needham and Begg (1991), and Novick (1988) tested transfer from one problem solving situation to other analogous ones. The general finding is that the transferability of the solution has to be made quite obvious, say by repeating it, before subjects transfer appropriately. Given these discouraging results, we cannot be sanguine about the learnability of creativity.

Summary Shepard's insights serve as useful guides to the study of the contribution of imagery to creativity and discovery. To summarize quickly, although Shepard's first conjecture that traditional verbal modes of communication may constrain the use of visualization and imaginal thinking was not addressed directly, Helstrup and Anderson's (1991, see also this volume) data suggest that visual strategies produce more mental discoveries than verbal strategies. Moreover, language may channel the interpretations given to images and impede the detection of alternative views (see review in Reisberg and Logie's chapter, this volume). Taken collectively, the evidence suggests some independence between two types of creativity, one primarily verbal and the other, primarily spatial. Shepard's second and third suggestions were that images may

32

Margaret Jean Intons-Peterson

afford more concrete richness than verbal descriptions and that the spatial nature of images renders them particularly amenable to spatial intuition and manipulation. This appears to be true when images are being generated and when their reinterpretation involves simple reorderings or reconstruals of the relations of components. Image reinterpretation is more difficult, however, when overall reference frames must be altered. It may be that creative individuals have faster access to their images, that their images have longer durations, or that they are able to refresh or regenerate images more rapidly than less creative individuals. These characteristics also would affect discovery within images. In a t least partial contradiction to this view were our results and the results of Anderson and Helstrup and Finke (see chapters in this book) that objects judged as creative are produced only some of the time by some of the people and that the production of objects judged creative during one session was essentially independent of the production of objects judged creative during a subsequent session. Nevertheless, despite the decline over a two-week interval in the production of descriptionsjudged to correspond to the drawings, the proportions of responses judged creative increased. We proposed that these results reflect a tendency for familiarity to suppress performance on the task, an effect that may be countered by the use of imagery. Shepard’s fourth characteristic was that vivid mental images may be specially effective in enjoining affective and motivational systems. This tantalizing possibility remains a challenge. Finally, the search for symmetry seems useful, particularly when extended to include natural discontinuities. Its utility may be constrained, however, by the greater difficulty in dissembling and transforming cohesive patterns and those with intact reference frames than less cohesive patterns and those with no reference frames. Challenges remain. Are some people better imagers than others as measured by image production and manipulation, rather than by pencil-and-paper or self-report? Some observerdimagers may process their image generation and memory searches more rapidly than others. These individuals presumably would be faster at gaining access to reference-frame reversals than others, and might be considered as more adept at discovery and, perhaps more creative. This notion implies that people judged creative are faster or more

Imagery’s Role in Creativity and Discovery

33

likely to reverse figures than people judged less creative, a notion that has not been tested, to my knowledge. Another possibility is that more rapid discoverers or more creative individuals use strategies to focus their searches on different reference-frame sets earlier than less rapid discoverers or less creative people. This redirection of the searches could occur because the initial search of the same reference-frame set is not exhaustive but, rather, is treated heuristically as an initial, fast scan, which, if not productive, is quickly broadened. Some answers appear in the other chapters in this book. As a final challenge to the authors of these chapters and to our readers, I end with a quotation from Henri Poincark, reproduced from Miller (1974, p. 307): The genesis of mathematical invention is a problem that must inspire the psychologist with the keenest interest. For this is the process in which the human mind seems to borrow least from the exterior world, in which it acts, or appears t o act, only by itself and on itself, so that by studying the process of geometric thought, we may hope to arrive at, what is most essential in the human mind.

References Anderson, R. E., & Helstrup, T. (in press). Visual discovery in mind and on paper. Memory & Cognition. Arieti, S. (1976). Creativity: The magic synthesis. New York: Basic Books. Amheim, R. (1969). Visual thinking. Berkeley: University of California Press. Bassok, M., & Holyoak, K. J. (1989). Interdomain transfer between Journal of isomorphic topics in algebra and physics. Experimental Psychology: Learning, Memory) and Cognition, 15, 153-166. Brandimonte, M. A., Hitch, G. J., & Bishop, D. V. M. (1992). Influence of short-term memory codes on visual image processing: Evidence from image transformation tasks. Journal

34

Margaret Jean Intons-Peterson of Experimental Psychology: Learning, Memory, and Cognition,

18,157-165. Chambers, D., & Reisberg, D. (1985). Can mental images be ambiguous? Journal of Experimental Psychology: Human Perception and Performance, 11, 317-328. Chambers, D., & Reisberg, D. (1992). What an image depicts depends on what an image means. Cognitive Psychology, 24, 145-174. Denis, M , , & Cocude, M. (1989). Scanning visual images generated from verbal descriptions, European Journal of Cognitive Psychology, 1,293-307. Ernest, C. (1977). Imagery ability and cognition: A critical review. Journal of Mental Imagery, 2 , 181-217. Finke, R. (1990). Creative imagery. Hillsdale, N J Erlbaum Associates. Finke, R. A., Pinker, S., & Farah, M. (1989). Reinterpreting visual patterns in mental imagery. Cognitive Science, 13, 51-78. Finke, R.A.,& Shepard, R. N. (1986). Visual functions of mental imagery. In K. R. Boff, L. Kaufman, & J. P. Thomas (Eds.), Handbook ofperception and human performance (Vol. 2,chapter 37,pp. 1-55). New York: Wiley Interscience. Finke, R. A., & Slayton, K. (1988). Explorations of creative visual synthesis in mental imagery. Memory & Cognition, 16, 252-257. Georgopoulos, A. P., Lurito, J.,Petrides, M., Schwartz, A., & Massey, J. (1989). Mental rotation of the neuronal population vector. Science, 243, 234-236. Ghiselin, B. (1952).The creative process. New York: New American Library. Gick, M. L., & Holyoak, K. J. (1983). Schema induction and analogical reasoning. Cognitive Psychology, 15, 1-38. Gordon, R. (1949). An investigation into some of the factors that favour the formation of stereotyped images. British Journal of Psychology, 39,156-167. Guilford, J. P. (1967). The nature of human intelligence. New York: Scribner. Hayes, J. R., & Simon, H. A. (1976).Psychological differences among problem isomorphs. In N. Castellan, Jr., D. Pisoni, & G. Potts (Eds.) Cognitive theory, Vol. 2 (pp. 21-41). Hillsdale, NJ:

Imagery's Role in Creativity and Discovery

35

Erlbaum Associates. Helstrup, T., & Anderson, R. E. (1991). Imagery in mental construction and decomposition tasks. In R. H. Logie & M. Denis (Eds.), Mental images in human cognition (pp. 229-240). Amsterdam: Elsevier Science. Hilgard, E. R. (1962). Introduction to psychology (3rd ed.). New York: Harcourt, Brace & World. Hinsley, D., Hayes, J. R., & Simon, H. A. (1977). From words to equations. In P. Carpenter, M. A. Just, & P. A. Carpenter (Eds.), Cognitive processes in comprehension (pp. 89-106). Hillsdale, NJ: Erlbaum Associates. Hyman, I. E., Jr., & Neisser, U. (1991). Reconstructing mental images: Problems of method (Emory Cognition Project Report #19). Atlanta, GA: Emory University, Department of Psychology. Intons-Peterson, M. J. (1981). Constructing and using unusual and common images. Journal of Experimental Psychology: Human Learning and Memory, 7,133-144. Intons-Peterson, M. J. (1984). Faces, rabbits, skunks, and ducks: Imaginal comparisons of similar and dissimilar items. Journal of Experimental Psychology: Learning, Memory, and Cognition, 10,699-715. Klatzky, R. L. & Thompson, A. (1975). Integration of features in comparing multifeature stimuli. Perception & Psychophysics, 18,428-432. Kosslyn, S. M. (1980).Image and mind. Cambridge, M A : Harvard University Press. Kosslyn, S. M. (1983). Ghosts in the mind's machine: Creating images and using images in the brain. New York: Norton. Krueger, T. H. (1976). Visual imagery in problem solving and scientific creativity. Derby, CT: Seal Press. Langley, P., Simon, H. A., Bradshaw, G. L., & Zytkow, J. M. (1987). Scientific discovery. Cambridge, MA: MIT Press. Mandler, J. M. (1983). Representation, In P. H. Mussen (Ed.), Handbook of Child Psychology, Vol. 3 (4th ed.) (pp. 420-494). New York Wiley. Marks, D. F. (1973).Visual imagery in the recall of pictures. British Journal of Psychology, 64, 17-24. Miller, A. I. (1984). Imagery in scientific thought: Creating 20th

36

Margaret Jean Intons-Peterson

century physics. Boston: Birkhauser. Needham, D. R., & Begg, I. M. (1991). Problem-oriented training promotes spontaneous analogical transfer: Memory-oriented training promotes memory for training. Memory & Cognition, 19,543-557. Novick, L. (1988). Analogical transfer, problem similarity, and expertise. Journal of Experimental Psychology: Learning, Memory, and Cognition, 14, 510-520. Paivio, A. (1986). Mental representations. New York: Oxford University Press. Peterson, M. A., Kihlstrom, J. F., Rose, P. M., & Glisky, A.L. (1992). Mental images can be ambiguous. Memory & Cognition, 20, 107-123. Reed, S . K. (1974). Structural descriptions and the limitations of visual images. Memory & Cognition, 2, 329-336. Reed, S.K.,& Johnsen, J. A. (1975). Detection of parts in patterns and images. Memory & Cognition, 3, 569-575. Reisberg, D., Smith, J. D., Baxter, D. A,, & Sonenshine, M. (1989). "Enactedll auditory images are ambiguous; "Pure" auditory images are not. Quarterly Journal of Experimental Psychology, 41A,619-641. Roe, A. (1951). A study of imagery in research scientists. Journal of Personality, 19, 159-170. Roskos-Ewoldsen, B. (1989). Detecting emergent structures of imaginal patterns: The influence of imaginal and perceptual organization. Unpublished doctoral dissertation, Indiana University, Bloomington, IN. Shaw, G. A., & DeMers, S. T. (1986).The relationship of imagery to originality, flexibility and fluency in creative thinking. Journal of Mental Imagery, 10, 65-74. Shaw, G. A., & DeMers, S. T. (1986-87). Relationships between imagery and creativity in high-IQ children. Imagination, Cognition, and Personality, 6 , 247-262. Shepard, R. N. (1978).Externalization of mental images and the act of creation. In B. S. Randhawa & W. E. Coffman (Eds.), Visual learning, thinking, and communication (pp. 133-189). New York: Academic Press. Shepard, R. N. (1990).Mind sights. New York Freeman. Smith, J. D., Reisberg, D., &Wilson, M. (1992).Subvocalization and

Imagery’s Role in Creativity and Discovery

37

auditory imagery: Interactions between the inner ear and inner voice. In D. Reisberg (Ed.), Auditory imagery (pp. 95-119). Hillsdale, NJ: Erlbaum Associates. Solso, R. L. (1991). Cognitivepsychology (3rd ed.). Boston: Allyn and Bacon. Tinbergen, N. (1948). The study of instinct. Oxford: Oxford University Press. Torrance, E. (1966). Torrunce tests of creative thinking: Norms and technical manual. Princeton, NJ: Personnel Press. Wallas, G. (1926). The art of thought. New York: Harcourt Brace.

Acknowledgment

I wish to thank Wendi Russell and Dana Berey for their assistance with the experiment reported herein, and Rita E. Anderson and Beverly Roskos-Ewoldsen for their splendid editorial critiques, challenges, and encouragement.

This Page Intentionally Left Blank

Imagery, Creativity, and Discovery: A Cognitive Pcrspectivc B. Koskos-Ewoldson, M.J. Intons-Peterson and R.E.Anderson (Editors) 0 1993 Elscvier Science Publishers B.V. All rights rescrved.

39

Chapter 2 ~~

~~

THEINSA N D OUTS OF WORKING MEMORY: OVERCOMING THE LIMITSON LEARNING FROM IMAGERY Daniel Reisberg Department of Psychology Reed College Portland, OR 97202 USA

Robert Logie Department of Psychology University of Aberdeen Aberdeen, AB9 2UB,

UK

Boundaries on Learning from Imagery

A large quantity of research documents the communalities between mental images and pictures -in terms of what information images and pictures include, what information they omit, how that information is accessed, and so on. In general, images seem to share many functional properties with pictures, and, crucially for this volume, images seem t o share with pictures the potential for supporting discoveries, or for inspiring inventions, or for leading to problem-solutions. The evidence for this claim comes from several sources. The history of scientific discovery contains many reports of problem solutions suggested by imagery (e.g., Miller, 1986). Likewise, people commonly use imagery to anticipate visual appearance or spatial relations. Most persuasively, there is much laboratory evidence documenting that we c a n glean new, unanticipated information from images (e.g., Finke & Kosslyn, 1980; Finke & Kurtzman, 1981; Pinker & Finke, 1980; Richardson, 1980). Thus it seems obvious that we can and do learn from images, that images can surprise us or instruct us. Yet, in 1985, Chambers and Reisberg reported a robust failure to learn from imagery; taken at face value, their data indicated a sort of discovery that happens

40

Daniel Reisberg & Robert Logie

routinely from pictures, but which does not happen with imagery. Concretely, Chambers & Reisberg (1985) showed their subjects drawings of classical ambiguous figures, for example, the ducklrabbit. Subjects were asked to memorize these figures, then the drawing was removed, and subjects were asked to form an image of the figure they had just seen. Subjects were then asked to reinterpret that image if they could (e.g., to find the duck in the rabbit image, or vice versa). To help them in this task, subjects were given a variety of hints and suggestions. Then, if all else failed, they were given a blank piece of paper, asked to draw the form they had just been imaging; they then tried to reinterpret their own drawing. No subjects in these studies were able to reinterpret their images, Subjects routinely succeeded, though, a moment later, in reinterpreting their own drawings. At the very least, these results point to a contrast between images and pictures, with the latter being easily reinterpreted in this procedure, while the former were not. (For a related result, see Hinton, 1979.) But these results pose a puzzle. Given that subjects can learn from imagery, can make discoveries in their images, what went wrong in the Chambers and Reisberg procedure? Why were these subjects not making a (seemingly) simple discovery from their image? To put this question more broadly, why is imagery sometimes capable of supporting discoveries and creativity, while sometimes it is not? What defines the limits on what we can learn from imagery? Several factors are likely to be relevant here. For example, one relatively uninteresting factor will be image complexity. Some discoveries about pictorial information involve very complex pictures or diagrams, and these might be too complex to image clearly. For discoveries of this sort, learning from images would fail, while learning from the appropriate picture might succeed. Or, as a similar example, some discoveries about pictorial information involve subtle nuances of form, and depicting these in an image might strain the limits of image acuity. Again, this would be a case in which learning from images would be difficult, even though learning from the corresponding picture might go smoothly forward. These factors doubtless constrain learning in many circumstances, but there are also deeper limitations on what can be learned from, or discovered within, an image. By way of introducing these, let us review some classic data from perception. A wealth of

The Ins and Outs of Working Memory

41

Figure 1 Pictures of the Necker cube and the vasdprofiles figure are neutral with regard to three-dimensional organization, designation of figure and ground, and the like. But these pictures areperceived as having a specific organization, and this organization seems to govern phenomenal appearance, what the shapes will evoke from memory, and so on.

evidence makes it plain that the conscious, phenomenal percept contains much information not present in the retinal image. In Bruner’s (1957,1973) oft-quoted phrase, the percept goes “beyondthe information given’’in many ways. For example, the retinal image of Figure 1A has no three-dimensional organization; the retinal image has no designation of figure and ground, nor is it parsed in any way. The retinal image is simply a pixel pattern, nothing more. But the pattern-as-perceived does have a three-dimensional organization, does have specifications about figure and ground, is, in fact,

42

Daniel Reisberg & Robert Logie

perceived as having a particular orientation. These specifications about the perceptual organization are all present in the percept, but not in the stimulus. These specifications play important roles in perceiving. They have a strong influence on phenomenal form. They have a strong influence on judgments of similarity and what a shape is seen as resembling. And, thanks to various perceptual linkages, a term due to Hochberg and Peterson (1987),these specifications also govern how other aspects of the stimulus will be perceived. What is crucial for our purposes is that these specifications also govern what can be learned, or discovered, about a perceptual form. Likewise, these specifications govern what a perceptual form is likely to evoke from memory. There are many results pertinent to these claims, but the major evidence is reflected in the standard illustrations of any introductory textbook. For example, consider the form shown in Figure 1B. We can show this form to subjects, and arrange, perhaps via instructions, for subjects to perceive it as a vase. Sometime later, we show the same figure to subjects, but this time we lead them to perceive it as two profiles. Subjects will, in this setup, routinely deny that they have seen the figure before, even though this exact stimulus was under their view just minutes earlier, Apparently, it is not the familiar geometry which governs recognition or discovery, it is instead the geometry-as-understood,in this case, the geometry with a certain specification of figurdground organization. The same is true for orientation. If Figure 2A is shown to subjects, many fail t o recognize the familiar shape of Africa (Kanizsa, 1979;Rock, 1973). This remains true even if subjects tilt their heads by 90 degrees, setting the figure upright. What matters is subjects’ understanding of upright, not retinal position, and when the form is understood as having a top different from Africa’s top, the form is not recognized. A great many results make broadly similar points. In general, what can be gleaned from a percept seems bounded by the half-dozen previously mentioned appearance specifications. These specifications jointly provide a reference frame within which the stimulus geometry is interpreted. (For further discussion of the relevant data, see Reisberg & Chambers, 1991. The reference frame terminology is Peterson’s, although we are using the term in a slightly different

The Ins and Outs of Working Memory

43

Figure 2 Perceived orientationgoverns recognitionfor many forms, so that subjects routinely fail to recognize the (perceptuallygiven) sideways map of Africa. Reiaberg & Chambers showed that the same was true in imagery (using a sideways map of Texas);what mattered was subjects' understanding of upright, rather than orientation on the "imagemedium."

manner than she has; cf. Peterson, this volume.) Thus, in these terms, what matters for learning or discovery is the stimulus within its reference frame, and not stimulus geometry per se. If some target or discovery is consistent with stimulus geometry, but not consistent with how the perceiver understands the geometry, learning is unlikely to go forward. As we have seen, numerous results warrant these claims about perception. In addition, Reisberg and Chambers (1991)have argued that similar considerations apply to imagery. (Parallel data, and a closely related conception, can be found in Peterson, this volume.)

44

Daniel Reisberg & Robert Logie

On their proposal, images, like pictures and like percepts, are depictions. That is, images directly represent the layout of the represented material, they preserve proximity and between-part relations, and so on. We regard these points as amply documented in studies of mental rotation, image scanning, and so on (e.g., Kosslyn, 1983; Shepard & Cooper, 1982). At the same time, however, images, like percepts, but unlike pictures, are obligatorily accompanied by a perceptual reference frame and are only understood within this frame. Hence discoveries from images, like discoveries from percepts, are bounded by this reference frame. In imagery, as in perception, one will make discoveries about the depicted form only if the discoveries are compatible with both the depicted geometry and the imager’s specifications about how that geometry is to be understood. Evidence for this claim is detailed in Reisberg and Chambers (1991). The evidence is easy to describe, since the experiments are modeled after the standard textbook demonstrations already mentioned. For example, subjects in one experiment were shown a series of nonsense shapes. After subjects had encoded each shape, they were asked to image it, and then to rotate their image by a specified amount - sometimes 90 degrees clockwise, sometimes 90 degrees counterclockwise, and so on. The tenth figure in the series, presented with no special notice, was the form shown in Figure 2B. Subjects presumably encoded this form as one more nonsense shape, and they were then instructed to imagine the form rotated by 90 degrees, so that they were now contemplating an image of a correctly righted map of Texas. Subjects were then told that this shape resembles a familiar geographic form, and were asked to identify that form. In perceiving this form, subjects presumably understood the side topmost in the drawing a5 being the form’s top. The form was therefore imaged with a reference frame identifying this top. It seems likely that imaged rotation would not change this understanding, on the hypothesis that imaged rotation would, in this regard, be like rotation on the retina (Olson & Bialystok, 1983; Rock, 1973). Thus, after the rotation, the image was isomorphic with Texas (i.e., shares a depiction), but, because of this assignment of orientation, was understood differently. Geometrically, the image and Texas are closely related, but, psychologically, the image has the

The Ins and Outs of Working Memory

45

wrong reference frame, and consequently has a different phenomenal shape than Texas does, does not resemble Texas, and should not call Texas t o mind. Consistent with these claims, no subjects succeeded in discovering Texas in their images. Moments later, though, subjects were asked to draw a picture, based on their image, and more than half of the subjects were able to discover Texas in their own drawings. Again we see a case of learning failing t o occur with imagery, but succeeding from a picture, drawing attention to the functional contrast between these representations. More specifically, though, we here see a case of learning failing to occur despite the fact that subjects are imaging the ttcorrectll geometry. The learning seems to fail simply because subjects understand that geometry in an inappropriate way. There is a problem in interpreting these data, however. Perhaps the Texas form is too complex or too subtle to be imaged clearly. We know that subjects succeeded in memorizing the form their success with their own drawings tells us that. Thus subjects had a reasonably complete, reasonably veridical memory of the test form. But we might still worry about the adequacy of the image itself -whether it is complete enough, or specific enough, to support recognition. Given these concerns, subjects' failure t o discover Texas in their images might reflect a structural limitation on imagery, rather than (as alleged) an influence of the image's reference frame. To rule out this possibility, Reisberg and Chambers ran a second group of subjects, with one change in instructions. Rather than instructing subjects t o imagine the test form rotated by go", subjects were told directly to change their understanding of the form's top. In short, subjects were told directly how to change the image's reference frame. Thus, while the rotate instruction left subjects with an image isomorphic with Texas, but understood differently, the reassign-top instruction should leave subjects with an image both geometrically and psychologically congruent with the Texas outline. Consistent with our claims, many subjects in this latter group did discover Texas in their images. Thus the understanding of the image seems the crucial element in determining success in this task. If one changes the understanding, one changes the result. Reisberg and Chambers (1991) report a variety of other procedures which extend and confirm these claims. (See also

46

Daniel Reisberg & Robert Logie

Peterson, this volume. Reisberg & Chambers, 1991, also provide

Figure 3 Subjects imaging this form often discover the Arabic numeral if given a hint about changing the depiction's reference frame; subjects without this hint do not recognize the numeral.

discussion of several conceptual issues that bear on the generality of these findings.) As an example, subjects in one of the Reisberg and Chambers studies were shown the form in Figure 3, and asked to memorize the shape. Subjects had previously seen (and practiced memorizing) a series of training figures, each composed of the same black forms as the test figure, but in rearranged position. These training figures provided ample opportunity for subjects to learn the shapes of the black forms, and also encouraged subjects to pay special attention to the positions of each black shape. The training figures also served to emphasize the figural importance of the black forms, providing the appropriate set for viewing the test form (i.e., Figure 3).

The Ins and Outs of Working Memory

47

Once subjects had imaged the test form, they were asked what familiar object was depicted. For half of the subjects, no further information was provided. The other subjects were told to think of the figure as being the shadow cast by a solid object - in essence changing their understanding of figure and ground relations within the imaged form. The data show the expected pattern. No subjects in the no-hint group discovered the numeral 5 . These subjects were imaging objectively-familiar geometry, but they understood this geometry in the wrong way, and so no discoveries emerged. In contrast, many subjects in the hint group discovered the 5. When the image’s geometry and the image’s reference frame are consistent with those of the target, then discoveries routinely happen. To summarize this section, the Reisberg and Chambers results confirm, once again, that discoveries can be inspired by imagery, that one can find unanticipated forms in one’s own images. Thus images can serve a crucial role in learning and in problem-solving. At the same time, however, there appear to be strong boundaries on the learning and discoveries inspired by images. In a variety of procedures, we have shown that learning from imagery goes forward only if the image and target form share both geometry and reference frame. Learning from imagery does not take place if there is a mismatch between the image an.d target reference frames; this seems true even when subjects are imaging familiar configurations, configurations easily recognized with pictorial (rather than image) presentations.

A Rigid But Fluid Boundary We hasten to add, however, that this boundary on learning from imagery is a peculiar one. On the one side, the boundary seems relatively rigid. In our experiments, no subjects have discovered the target form in their images when the image and target were understood differently. Other studies have not replicated this zerolevel performance, and we will discuss this point in a moment (cf. Hyman, this volume; Kaufmann & Helstrup, this volume; Peterson, this volume). But all studies agree that there is a massive effect in the data. Discoveries from imagery happen almost universally when the discoveries are compatible with the image frame. Discoveries

48

Daniel Reisberg & Robert Logie

from imagery happen with much lower frequency (and, in our studies, never) when they are incompatible with the image frame. This is a strong effect, and it is this that leads us to describe the learning boundary as rigid. At the same time, however, this boundary on learning is easily moved. Subjects' specifications about an image - its figurdground organization, its orientation, and so on - are, after all, the subject's specifications. The specifications are set by the subject, are manifest in how the subject reads or understands his or her own image, and these specifications can be changed by the subject. We know that image specifications can be changed with a little prodding from the experimenter. This is crucial, for example, in the Reisberg and Chambers experiments (e.g., the reassign top experiment). This point is also clear in recent work by Brandimonte and Gerbino (in press), and by Hyman (this volume; Hyman & Neisser, 1991). In both of these latter studies, subjects were specifically instructed to change their understanding of an imaged form (e.g., the duckhabbit figure). Prior to this change (that is, when the image is understood "incorrectly"), discoveries from imagery rarely occurred. Subsequent to this change, discoveries routinely occurred. Peterson (this volume) has obtained similar effects using so called teaching examples as a way of conveying the required change in image specifications. Thus the overall pattern of these results resembles that of the Reisberg and Chambers findings with discoveries governed both by the imaged geometry and by how that geometry is understood, and with new discoveries becoming available once this understanding (the reference frame) is changed, For present purposes, though, these results remind us once again that, with instruction, an image's reference frame is easily open to change. In some ways, this pattern of results is peculiar. After all, nothing prevents a subject from deciding, on his or her own initiative, to change an image's reference frame. These changes would seem quite likely when subjects are specifically urged to explore their images, seeking new forms. Indeed, as we have seen, discoveries from imagery are much more likely if subjects change their image's reference frame. If, however, subjects did redefine their image's reference frame, this would undermine the impact of the experimenter-defined reference frame. As we have seen, though,

The Ins and Outs of Working Memory

49

the experimenter-defined reference frame has a huge effect on the data. We conclude, therefore, that these spontaneous redefinitions of the frame are surprisingly infrequent. Of course we need t o be careful here. We know that some subjects do make image discoveries that are incompatible with the image's initial frame (cf. Hyman, this volume; Kaufmann & Helstrup, this volume; Peterson, this volume); we presume that these are the subjects who have spontaneously redefined their own images - i.e., have changed the reference franies on their own, without instruction from the experimenter. But we are struck by how few of these subjects there seem to be. We read this as implying that, despite situational encouragement for this redefinition, subjects are either incapable or inept in this redefinition. That is, with experimenters' instructions, a clear majority of subjects can change their image's reference frame. Without the instructions, only a small minority makes these changes. A number of hypotheses are available for explaining this contrast, but this surely seems an issue in need of further research. It does seem appropriate, though, to sketch one way this research might unfold: Subjects can, of course, change an image's reference frame if they so choose. Moreover, they can change the reference frame in any way they choose: They can (for example) define any side of the form as the form's top; they can define the figure's configuration in depth in any of' a number of ways, and so on. This breadth of options may itself provide something of an obstacle for subjects: Their task suggests to them that a redefinition of the image will be useful, but subjects have no way of knowing which redefinition to choose, from the many that are possible. Conversely, the likelihood of subjects stumbling across the "correct" redefinition (i.e., the one that will lead to the target form) may be small, and this may be the reason why few subjects succeed (without some sort of aid). On this view, then, the duckhabbit or Texas procedures may be special in two senses. Not only clo these procedures require a change in reference frame, they require a specific change, if the sought-after form is to be discovered. Note, however, that one can design tasks without this second requirement - that is, tasks requiring a change in reference frame, but tasks for which any of a number of changes will serve. The implication of our claim is that creative, unexpected

50

Daniel Reisberg & Robert Logie

discoveries will be far more likely in tasks of this latter sort. (Finke & Slayton’s (1988) procedure may provide such a case; likewise the procedure employed by Anderson & Helstrup, this volume.)

Auditory Imagery and the Importance of Stimulus Support To return to our main agenda, we have now argued that learning from images is bounded, but that this boundary can be overcome. All subjects need to do to expand their imagery discoveries is to change the reference frame within which an image is understood. More specifically, subjects need t o understand the image within a reference frame that matches that of the target. Subjects seem easily able to do this with instruction, but do this infrequently without instruction. There is, however, something else subjects can do to escape the boundaries on learning from imagery. They can create a stimulus. In both the Chambers and Reisberg (1985) and Reisberg and Chambers (1991) studies, subjects were able to draw pictures, based on their images, and then t o make discoveries from these pictures that they had not made with the corresponding images. Why is this? We have argued elsewhere that some judgments, by their nature, require stimulus support. That is, these judgments can easily be made about a stimulus, but are not easily made with reference t o a mental representation (e.g., Reisberg, 1987; Reisberg, Wilson, & Smith, 1991). This leads to a question: If a judgment requires stimulus support, does that mean the judgment cannot be made with imagery? Some recent data suggest a surprising answer to this question. Most of the published data on imagery (and virtually all of the data in this volume) concern visual imagery. Yet claims about imagery are often framed broadly - suggesting that these claims should apply to imagery in modalities other than vision. In a recent series of studies, therefore, we set out t o examine whether claims made about visual imagery do in fact generalize to other imagery modalities. An early group of experiments (Reisberg, Smith, Baxter, & Sonenshine, 1989) was modelled after the Chambers and Reisberg (1985) studies, already described. Subjects were acquainted with

The Ins and Outs of Working Memory

51

auditory ambiguous figures, and then asked to image one of these figures. Once the image was in place, subjects were then asked whether they could reinterpret this image. As their ambiguous figures, Reisberg et al. exploited the fact that certain words, if repeated over and over aloud, yield a soundstream compatible with more than one segmentation. For example, rapid repetitions of the word life produce a physical soundstream that is fully compatible with segmentations appropriate to repetitions of life or of fly. These repetitions are usually perceived first as one of these words, then the other, then the first, changing in phenomenal form just as the Necker cube or duckhabbit do. This allows us to ask if imagined repetitions produce verbal transformations, just as heard repetitions do. Subjects in these studies were asked to imagine (or, in some conditions, actually heard) a voice repeating the word, stress, over and over. With an actual (perceived) stimulus, 100% of the subjects heard this soundstream transform into repetitions of dress, a construal fully compatible with the acoustical input. However, this transformation was rarely guessed by a control group, making the transformation a relatively clear indication of bonafide perceptual reversals. Subjects who imagined these repetitions often detected the stress-to-dress transformations. This seemed to indicate a sharp contrast with the data from visual imagery, in which subjects have routinely failed to reinterpret their images. However, subjects’ success in reversing auditory images turned out to depend on subvocalization. In one study, subjects were prevented from subvocalizing by having them chew candy, and this dropped reversals (i.e., detections of dress) to the level achieved by guessing subjects (25%). Reversals were also reduced (from 73% to 27%) when subvocalization was blocked by having subjects tightly clamp their jaws shut, firmly press their tongue up against the roof of their mouth, and hold their lips firmly shut. One might worry, though, that these manipulations (chewing, or clamping the mouth shut) do more than block subvocalization. For example, these manipulations might be generally distracting to subjects, and so disrupt performance for this reason. However, we know from various other studies that these manipulations do block covert speech and seem not t o have other effects, such as distracting

52

Daniel Reisberg & Robert Logie

the subjects (see, for example, Baddeley, 1986; Besner, Davies, & Daniels, 1981; Levy, 1971; Murray, 1968; Slowiaczek & Clifton, 1980; Wilding & White, 1985). We conclude, therefore, that the stress/ dress reversals really do depend on subvocalization, and so are blocked by these manipulations that prevent covert speech. But if stresddress reversals depend on subvocalization, what role does the inner speech play? Is the inner speech perhaps providing some kinesthetic cues, crucial for performance? Or is it possible that subjects subvocalize the to-be-imagined event, and then listen with some inner ear to find out what they have themselves produced? To address these questions, subjects in a further experiment were asked to perform this task while hearing an irrelevant message through headphones. This manipulation reduced the number of reversals from 73% to 13%, suggesting that subjects need both the inner voice and the inner ear to perform this task. The implication is that subjects are literally talking to themselves and then listening to this self-produced stimulus. Other studies show this t o be a widespread pattern in auditory imagery. For example, consider the following task (Smith, Reisberg, & Wilson, 1992). Subjects were visually presented with strings such as D 2 R and asked what familiar word or phrase would result if this string were pronounced out loud (detour). Some subjects were asked to perform this task while hearing auditory input through headphones, blocking use of the inner ear. Subjects were instructed to ignore this input, and it was in any event irrelevant to their task. Other subjects read the strings while repeating Tuh-Tuh-Tuhaloud, blocking use of covert speech. A third group of subjects performed with both the irrelevant auditory input and the concurrent articulation task (both inner ear and inner voice disrupted); a fourth group received neither type of interference. The results yielded the same pattern as the stresddress task. When subjects interpreted these strings with no interference (with auditory input absent and subvocalization possible) they were able to decipher 73% of the strings. But, denied the inner ear or inner voice, performance declined to 40% and 21%, respectively, and to 19% with both forms of interference present simultaneously. Again, it seems that subjects perform this task by speaking to themselves, and then listening to hear what they have said. Not all auditory imagery tasks show this pattern, however. For

The Ins and Outs of Working Memory

53

example, consider the following tasks. What familiar word would this produce if pronounced aloud: phluim? Or, would these sound alike if pronounced aloud: hejlhedge? These homophone judgments phenomenally seem to draw on auditory imagery and certainly depend on some representation of sound. But these judgments do not require subvocalization - that is, performance is not hurt if use of the inner voice is blocked (e.g., Baddeley & Lewis, 1981;Besner, 1987;Richardson, 1988;Wilding & White, 1985). For the sake of contrast, though, consider this task. Would these rhyme if pronounced aloud: days /maize? This task obviously resembles the homophone task just described. In both cases, subjects are shown a pair of letter strings, and must decide how these would be related, if pronounced aloud. Despite this similarity, the homophone and rhyme tasks yield rather different results. Homophone judgments seem not to depend on subvocalization, and performance is not compromised if subvocalization is blocked. However, the studies just cited also document that rhyme tasks do rely on subvocalization, and so performance in rhyme tasks suffers if subjects are prevented from using the inner voice. An account of this contrast was suggested early on by Besner (19871,and has recently been extended by Reisberg, Wilson and Smith (Reisberg et al., 1991;Smith et al., 1992). More generally, these authors have offered an account of why some judgments about sound seem dependent on subvocalization (and so are disrupted by concurrent articulation), while other judgments are not. We hasten to say that this account is speculative, since relatively few studies bear on this issue directly. Nonetheless, there is a pattern in the available evidence, and the pattern will call our attention back to issues raised earlier in this chapter. In a wide variety of tasks, one must create and then judge some mental representation of sound. For some of these tasks, one needs to do little more than this. That is, one can create these representations of sound as intact units, often drawing on some template, so t o speak, already in memory. One can then judge the representation without any further analysis -without any reparsing or reorganization. This is obviously true, for example, in the homophone judgments, because these require no analysis, only the activation of a single logogen in long-term memory. In cases of this sort, the inner voice seems not to be needed, and one can simply

54

Daniel Reisberg & Robert Logie

draw on pure auditory imagery. Other tasks require that one create a mental representation of sound, and then do some post-assembly operations on this representation - operations that might include segmentation, or some reparsing, or even mere maintenance. Tasks of this sort seem to require the support of the inner voice, and performance is disrupted if use of subvocalization is prevented. This pattern obviously fits with the cases we have described so far. The image reversal task (stress/dress) clearly requires that an auditory image be reparsed, and success in this task requires subvocalization. The D 2 R task likewise requires reorganization of an image, and also requires the inner voice. Similar considerations apply to rhyme tasks. In judging rhyme, one must re-segment the imaged sound, to cut away the word-initial sounds and judge the word-final sounds. Thus it is sensible that rhyme tasks require use of the inner voice. In contrast, homophone tasks require no postassembly analysis -there is no need for reparsing or reorganization. Therefore the inner voice is not required for homophone judgments. Finally, as one more example, findings indicate that concurrent articulation disrupts children’s spelling, even though it does not disrupt their reading (e.g., Kimura & Bryant, 1983). This again fits with the proposed pattern, on the assumption that spelling, but not reading, requires analysis of phonological codes, requires the creation of and then dissection of auditory images. (For further discussion of evidence compatible with these claims, see Besner, 1987; Reisberg et al., 1991; Smith et al., 1992.) This data pattern suggests an important role for subvocalization, and, in addition, this pattern draws us back to our earlier claims about stimulus support. In general, the key seems to be this. Some auditory tasks require judgments that are compatible with one’s initial understanding of a sound, while other tasks require judgments incompatible with this understanding. For tasks of the latter sort, subjects need to set aside their initial understanding to reanalyze or resegment the sound. This is precisely the case in which subjects need something like an actual stimulus, with an existence that is independent of the subjects’ thoughts and understanding. With a stimulus, the subject can take a neutral stance toward the to-be-judged sound, and so make new discoveries, in essence guided by the sound itself, rather than being guided by

The Ins and Outs of Working Memory

55

the sound-as-initially-understood. Within this context, subvocalization appears to provide a stimulus, albeit a covert one. In essence, subvocalization allows subjects to create an auditory event, and then to disown it, to hear the event as a stranger might. In this fashion, subvocalization can aid those auditory tasks requiring stimulus support.

The working memory system As we have now seen, many auditory tasks require a partnership between the inner voice and the inner ear, with subjects literally talking to themselves, and then listening to hear what they have pronounced. This partnership may seem familiar to many readers because a similar pattern can be found in the literature on working memory. It seems appropriate, therefore, to review that literature briefly; this will allow us t o make several points including one further advantage that derives from the use of subvocalization. It has long been known that there is a close relation between working memory and some sort of phonological or speech-based coding. For example, subjects in verbal short-term memory tasks frequently err by substituting phonologically similar items for the correct ones. Recall is also reduced if the to-be-remembered (TBR) items are phonologically similar to each other, even with visuallypresented materials (e.g., Baddeley, 1966; Conrad, 1964). These effects are generally attributed to phonological coding of the TBR items. The role of speech-based coding is also demonstrated by the so-called word-length effect, in which memory span for words that can be pronounced quickly is actually somewhat greater than span for slowly-pronounceable words (Baddeley, Thomson, & Buchanan, 1975). To explain these results, Baddeley and others (see, e.g., Baddeley, 1986; Vallar & Baddeley, 1984) describe a cognitive resource called the articulatory rehearsal loop. This loop involves two components, subvocal rehearsal and a phonological store, working in concert. For example, with visual presentation, subvocalization is used to load the TBR materials into the phonological store. The contents of the store decay, but subvocalization can be used to refresh the store’s contents.

56

Daniel Reisberg & Robert Logie

In this manner, one rehearses by talking to oneself, and then listening to hear what one has just said. This cycle can be repeated indefinitely, so that material can be held in working memory for as long as one wishes. However, the capacity of the rehearsal mechanism itself appears to be around 2 seconds' worth of speech, although the time for material to decay completely fiom the phonological store, in the absence of rehearsal, remains to be clarified (see, e.g., Baddeley & Lewis, 1984;Vallar & Baddeley, 1982). This conception of working memory plainly relies on a partnership between the inner ear and inner voice, just as does our conception 'of auditory imagery. Moreover, this conception leads to a number of testable hypotheses about working memory, and these have fared well in the laboratory. In the interests of brevity, we will simply say that this conception does an excellent job of predicting (among other things) when phonological confusions will or will not occur, when the word-length effect will or will not be observed, and so on. Thus we regard the notion of the articulatory loop as well supported by evidence. (Much of the relevant evidence is reviewed in Baddeley, 1986, 1990. For recent reviews, see Baddeley, 1992; Della Sala & Logie, in press; Logie, in press.) We wish to take special note, though, of the advantages conveyed by use of the articulatory loop. Rehearsal via this loop maintains information near-at-hand, serving as a scratch pad for working memory. Crucially, the rehearsal loop provides this function with minimal cost. Bear in mind that rehearsal draws on the speech system, and of course many of the relevant aspects of speech are highly automatized. Thus the central executive of working memory is needed t o initiate rehearsal, but, within the duration of each rehearsal cycle, processing can proceed automatically, freeing the executive to work on other matters. Hence rehearsal becomes an important resource in the support of time-sharing in working memory. Communalities and Contrasts in Auditory and Visual Imagery

It will be useful at this point to summarize the argument so far. A mass of evidence tells us that imagery can inspire discoveries and

The Ins and Outs of Working Memory

57

inventions. It also appears that there are boundaries on what can be learned from an image. In particular, discoveries from an image must be compatible both with the actual depiction and with how that depiction is understood. Put differently, images seem to be understood within a reference frame, and discoveries from imagery depend on the image within that frame. This pattern has emerged in studies of both visual imagery and auditory imagery, although the pattern manifests itself in somewhat different ways in the two modalities. This boundary on learning from imagery has a peculiar status. On the one side, the boundary seems relatively rigid. Discoveries from imagery are extraordinarily common when the discoveries are compatible both with the depiction and the reference frame; discoveries are relatively rare when they are incompatible with either depiction or reference frame. At the same time, though, this boundary on learning is readily escaped. Given explicit instructions or other appropriate cues, subjects can change an image’s reference frame, and this changes what can be discovered within the image. Likewise, subjects have the option of externalizing the image, either by drawing a picture (for visual imagery) or speaking aloud (for auditory imagery). The resulting stimulus then exists quite independently of any reference frame, and so can readily be understood within a new reference frame. So far, this summary has emphasized the same points for visual and auditory imagery. But the evidence also indicates a point of divergence between the modalities. In auditory imagery, subjects have the option of creating a covert stimulus, via subvocalization. Thus subjects can make discoveries from (subvocalized) auditory images even though they fail to make the corresponding discoveries from visual images. As a concrete case, consider the various studies of imaged ambiguous figures. The Chambers 8z Reisberg (1985) subjects failed to reinterpret a visual image of an ambiguous figure, and, correspondingly, the Reisberg et al. (1989)subjects failed t o reinterpret a pure image of an auditory ambiguous figure. In both cases, subjects easily reinterpreted the figure once a stimulus was provided (a drawing in the former study and overt speech in the latter). The point of contrast, though, arises when subvocalization was allowed. With access to this covert stimulus, the Reisberg et al.

58

Daniel Reisberg & Robert Logie

subjects were able to reinterpret the auditory image. Thus subvocalization plainly broadens the range of what can be accomplished with auditory imagery. In fact, auditory imagery seems often to rely on subvocalization, and, in particular, to rely on a partnership between inner speech and hearing. Subjects do not need to be coached or instructed in how to use this partnership; it seems instead to be a natural and spontaneous strategy. We have already sketched an account of subvocalization’s role in terms of the stimulus support it provides. It also seems that subvocalization can provide a further advantage. As we have seen, rehearsal in working memory takes advantage of speech to create an efficient and (largely) effortless storage buffer. Thanks to the automaticity of speech production, complex time-sharing becomes possible, as the mere maintenance of information can be done with minimal attention. Given these advantages, then, it is no wonder that subvocalization is a commonly used, spontaneous strategy.

The Possibility of Other Rehearsal Loops Consider exactly what is needed to create a rehearsal loop in working memory. First, one needs a highly efficient efferent channel so that production can be automatized, permitting time-sharing. Second, one needs some coding scheme, so that one’s actions can serve as representations. Finally, this.efferent production must feed into some afferent channel, so that the covert production can be recorded. This feed into an afferent channel is, of course, essential for our claims about stimulus support, and is also needed for working memory rehearsal, to close the rehearsal loop. These three criteria are certainly satisfied by articulatory rehearsal. For this rehearsal, the activity of speech provides the efferent channel (the inner voice), the rules of language provide the coding scheme, and the inner ear provides the relevant afferent channel. But many activities other than speech are entirely automatized; other coding schemes can be located or invented; and, of course, other efferent channels might feed into different afferent channels. Thus it is possible that other rehearsal loops exist, or can be created, in addition to the articulatory loop we have been considering.

The Ins and Outs of Working Memory

59

All of this is plainly relevant to our main agenda. We have argued that imagery provides poor support for some discoveries, namely, those discoveries that require a reference frame change. This limitation can be overcome, however, by means of stimulus support. This support is generally provided by external stimuli, but can also be provided by self-created, covert stimuli. In the case of auditory imagery, covert stimulus support is created through subvocalization, creating a covert auditory event. We have now argued that there is nothing unique about this partnership between the inner voice and the inner ear, and other partnerships, other rehearsal loops, may exist. Thus other forms of covert stimuli, and other forms of stimulus support, may be created. This leads us to ask: what about visual imagery? Could visual images, like auditory images, be enacted via some sort of efferent inner scribe, analogous to the inner voice? If such enactment were possible, it would carry real advantages - providing stimulus support for visual judgments requiring this support, and allowing subjects to overcome the apparent limits on visual imagery function. The issue, therefore, is whether such enactment of visual images is possible. The available evidence already indicates that other rehearsal loops can be created, in addition to the standard loop of articulatory rehearsal. For example, evidence suggests that the deaf use a manual form of rehearsal, with the efferent channel in this case being hand movements, and the afferent channel kinesthesis (e.g., Bellugi, Klima, & Siple, 1975;Shand, 1982;see also Campbell, 1992). There also is evidence that the hearing population can be taught to use a different manual system to perform rehearsal (Reisberg, Rappaport, & O’Shaughnessy, 1984). But what about visual imagery, and visual rehearsal? Baddeley has explicitly proposed that a visuomotor buffer provides one of the slave systems used by working memory and has suggested that visual imagery relies on this buffer (Baddeley, 1986; Logie & Baddeley, 1990). What is the structure of this buffer? Does it rely on a partnership between motoric output and some sensory channel, just as articulatory rehearsal seems to? The initial answer to this question would seem to be that visual imagery does not have this structure. That is, visual imagery seems to follow different rules from those for auditory imagery, although, to anticipate the argument t o come, we will argue that the evidence

60

Daniel Reisberg & Robert Logie

for this point is unpersuasive. We spend the remainder of this chapter exploring the suggestion that visual imagery may, under some circumstances, rely on something like an inner scribe. From this base, we will return to the question of what discoveries can, and what discoveries cannot, be made about one’s own visual images.

The Inner Scribe and the Inner Eye We have considered several points of contrast between auditory imagery and visual imagery. For example, auditory images are spontaneously reversible, unless one takes steps to block subvocalization. Visual images seem not to be spontaneously reversible (at least for most subjects), unless one provides subjects with specific hints or cues. So in one case we have reversibility unless one prevents a strategy; in the other case, we have no reversibility (or infrequent reversibility) unless one provides some help. Moreover, the reversal of auditory images seems to rely on stimulus support, and not on specific information about the image’s reference frame. In contrast, instructions about reference frame seem crucial in the visual case. All of this surely makes it look like visual imagery and auditory imagery are governed by different principles. In the same spirit, auditory images seem to function like stimuli, at least when subvocalization is allowed. The point of our discussion earlier in this chapter was that visual images do not function like stimuli, self-produced or otherwise. Again, this would appear to speak against the idea that visual imagery relies on a partnership between an efferent channel and a sensory channel. These arguments do not settle the issue. The arguments indicate that visual images, in some procedures, with some sets of instructions, function rather differently from auditory images. More specifically, the data indicate that, in some procedures, with some instructions, visual images are not accompanied by motoric enactment. We could ask, however, whether visual imagery always shows this pattern. What we really need to ask, therefore, is whether visual imagery euer draws on motoric support, with the functional benefits that we have proposed. We turn now to evidence that may bear on these questions. It

The Ins and Outs of Working Memory

61

will be useful in what follows to distinguish several claims. First, we will argue that there is more than one species of visual imagery; these various forms can be distinguished both functionally and neurally. Second, we will argue that one species of visual imagery is more properly considered spatiaVmotoric imagery, not visual. For imagery of this sort, one images by planning, or mapping out, a series of possible movements; there may be little involvement in this case from visual mechanisms or the inner eye. Finally, we will discuss whether this motoric imagery can produce a covert stimulus which then feeds into the channels of vision. It is this last claim, of course, which is directly pertinent to our earlier suggestions about a partnership in visual imagery between an inner scribe and inner eye.

Motoric Influences on Visual Imagery

A variety of research paradigms are interpreted as reflecting the properties of visual imagery, including chronometric studies of image scanning, studies of imagery mnemonics, studies of various sensory effects produced by imagery, and so on. We do not quarrel with the claim that imagery is indeed involved in these various paradigms; the issue, though, is whether the same imagery is involved in each case. As we see it, there has been an unspoken, and unexamined, assumption made by many imagery researchers, namely that visual imagery is one thing, manifesting itself in slightly different ways in a wide range of phenomena. We see no reason to credit this assumption. To the contrary, a number of considerations speak in favor of distinguishing various species of this imagery. For example, some imagery effects seem truly visual in nature, and may literally involve mechanisms within the visual system (e.g., Finke, 1989). Other imagery effects seem more spatial in their character, and may not draw on the visual system in any way (see, for example, Baddeley & Lieberman, 1980; Carpenter & Eisenberg, 1978; Farah, Hammond, Levine, & Calvanio, 1988; Jonides, Kahn, & Rozin, 1975; Kerr, 1983; Logie & Marchetti, 1991; Marmor & Zaback, 1976; Paivio & Okovita, 1971; Zimler & Keenan, 1983). Likewise, some imagery phenomena are predictively linked to imagery self-report, whereas others are not (e.g., Dean & Morris, 1991; Heuer, Fischman, & Reisberg, 1986;

62

Daniel Reisberg & Robert Logie

Reisberg & Leak, 1987). Neuropsychological data also imply that we may need a fractionation of visual imagery (e,g., Della Sala & Logie, in press; Farah et al., 1988; Kosslyn, 1991). Thus, the domain of visual imagery may be internally diverse, and we may be misled if we fail to distinguish the various species of imagery. In addition, and crucially for our purposes, the data suggest that at least some sorts of visual imagery draw on motoric enactment, suggesting a possible role for the (hypothesized) inner scribe. The relevant evidence, once again, derives from the literature on working memory. One widely used paradigm in visual working memory is to study the temporary retention of visual patterns. For example, subjects might be instructed to visualize a grid, and then to place, mentally, to-be-remembered items within the grid. (“In the lower-left corner, place a J. One square to the right, place a K , , . I’ and so on - after Brooks, 1968.) Retention of this sort of material is disrupted if subjects are simultaneously required to make arm movements that involve following a moving target (Baddeley & Lieberman, 1980). The concurrent movement task has little effect, though, on retention of similarly structured verbal material. Moreover, retention of the spatial material is unaffected by a concurrent brightness judgment task, which presumably involves visual rather than spatial processing. This pattern of data seems to imply that the matrix is retained via some sort of spatiallmovement code, rather than by a visual code. A number of other studies (Logie & Marchetti, 1991; Morris, 1987; Quinn, 1988, 1991; Quinn & Ralston, 1986; Smyth & Pendleton, 1989) also have reported evidence for a link between the control of movement and temporary memory for spatial material. Other results show a different pattern. For example, Logie (1986) required his subjects to retain information by means of a visual mnemonic, and were disrupted by a concurrent visual task (concurrent presentation of irrelevant patterns). Similar findings have been reported by Johnson (1982), and by Matthews (1983). Phillips and his colleagues have also demonstrated that subjects can retain static visual patterns (e.g., Phillips & Christie, 1977; see also Logie, Zucco, & Baddeley, 1990). This seems to imply that visuospatial working memory can deal with visual, as well as spatiall movement, information. Could it be that working memory contains two separate components, one specialized for visual information, one for spatial/

The Ins and Outs of Working Memory

63

movement information? Some evidence suggests that this is the case (see also Logie, 1989, 1991). Logie and Marchetti (1991) asked their subjects to remember either static visual images, or a sequence of movements. In the former condition, subjects had to retain a static array of color patches. Performance in this condition was disrupted by presentation, during the retention interval, of a series of irrelevant pictures. There was no memory disruption associated with performance of an unrelated movement task during the retention interval. When subjects were asked to remember the spatial sequence in which a series of color patches was presented, however, the reverse pattern of disruption was observed: Interpolated arm movement disrupted memory for this sequence, whereas interpolated, irrelevant pictures did not. A related pattern was observed by Hadden (1991). In a study of individual differences, Hadden discovered that memory for an object’s color was correlated with subjects’ self-reported imagery vividness, but memory for the object’s size was not. Conversely, memory for size was correlated with a psychometric assessment of spatial skill, but memory for color was not. This kind of crossover interaction, in the Hadden data or those reported by Logie and Marchetti, seems to provide an experimental double dissociation between a system that can retain static visual images, and a system that can retain a series of spatial displacements. Similar distinctions - between visual and spatidmotoric representations -can be found within the visual imagery literature. For example, Engelkamp (1986, 1991) and others (e.g., Saltz and Donnenwerth-Nolan, 1981) have shown that images that are enacted are better remembered than those that are not. Engelkamp documents that asking subjects to mime the act of smoking a pipe, or even to imagine miming this act, leads to better memory retention than asking subjects to imagine a static image of someone smoking a pipe. That is, images which are enacted are remembered better than images which are not. One last line of research also speaks to our theme. Studies of movement control reveal many points of overlap between the mental representations of movement, and mental representations of spatial position. As an initial observation, note that the parietal cortex is plainly implicated in the control of movement (e.g., Georgopoulos, Kalaska, Caminiti, & Massey, 1982,19831,and is also responsible for

64

Daniel Reisberg & Robert Logie

determining where an object is relative to the body (Anderson, Essick, & Siegel, 1987; Butters, Barton, & Brody, 1970). Georgopoulos and his colleagues have, in fact, documented patterns of neural activity in the parietal cortex that precede voluntary movement; the most straightforward reading of this is that the parietal activity instantiates a motor plan, or motor intention, for th.at movement (e.g., Georgopoulos, Lurito, Petrides, Schwartz, & Massey, 1989). Indeed, Georgopoulos et al. have documented mental rotation in this neural activity, plainly linking the study of motor planning with the study of imagery. For example, a monkey in one procedure was shown a visual cue, signalling a direction of motion. The monkey’s task in the relevant trials was to move in a direction 90 degrees counterclockwise from the direction indicated by the cue. Immediately after the cue’s presentation, the neural activity showed a pattern appropriate for motion in line with the cue. The vector of this activity then rotated, as a linear function of time, to the desired direction. In effect, the monkey rotated a motor plan, passing, in an analog fashion, through the directions intermediate between that of the cue and that of the target direction. The parallels are clear between this result and human studies on mental rotation. The suggestion is that, in human mental rotation (Shepard & Cooper, 1982), it is a motor plan being rotated; likewise, in image scanning (e.g., Kosslyn, 1983), subjects are moving a motor intent (perhaps an intent to point) across imagined space. It is interesting to consider the implications of this for specific results within mental rotation or image scanning, but that is not our project here, We simply note that, if this extrapolation is correct, then it may emerge that procedures thought to explore visual imagery are in fact revealing motoric imagery, instead!

A Partnership in Visual Imagery The previous section made two simple points: that the domain of visual imagery is internally diverse, and that one of the species of visual imagery seems, in fact, to be motoric imagery. Although we have not reviewed the evidence here, we note that another species of visual imagery does seem to be more strictly visual - employing pathways within the visual system, and showing many of the

The Ins and Outs of Working Memory

65

functional characteristics of vision (see Finke, 1989, for a review). The question now to be asked is the relation among these species. To claim enacted visual imagery, or to claim that the inner scribe feeds information t o the inner eye, we need to ask how (or whether) these various species of imagery interact. Unfortunately, little evidence speaks to this point directly. The best we can do, therefore, is to discuss the plausibility of the partnership idea. Let us return briefly to the articulatory case, i.e., the partnership between inner speech and the inner ear. One might reasonably ask how this partnership came to be -i.e., why evolution has provided us with this link between covert speech and hearing. One plausible suggestion is that this link serves as a feed-forward connection, priming the channels of hearing for upcoming speech. This may provide a means through which the speech produced can be checked for correctness (e.g., Zivin, 1986). This function would be particularly important, for example, in acquiring new spoken vocabulary, because, in this case, the production is not yet automatized, making it crucial that one match the sound of one’s own spoken output with auditory input from another speaker. Consistent with this, young children’s ability to repeat aloud words spoken by an experimenter is closely related to the children’s ability to acquire vocabulary for the first time (e.g., Gathercole & Baddeley, 1989, 1990). Likewise an impairment of the inner ear or inner speech in neuropsychological patients causes considerable difficulty in acquiring foreign language vocabulary (Baddeley,Papagno, & Vallar, 1988; Papagno, Valentine, & Baddeley, 1991). As a related suggestion, this feed-forward, from speech production to audition, might allow the speaker to filter hisher own voice, facilitating attention to other auditory inputs. These various (hypothesized) advantages would not be unique to speech output, nor to hearing. One could imagine that parallel mechanisms might allow one to monitor one’s own movements, to make certain these were faithful to one’s intentions. In addition, we know that, for a number of purposes, these movements must be fed into the visual system. For example, perceivers easily distinguish between image displacement across the retina caused by their own motion, and displacement caused by motion in the environment. To make this distinction, the visual system needs information about motor actions. All of this is consistent with the kind of efferent plus

66

Daniel Reisberg & Robert Logie

afferent partnership we are proposing, with motor plans feeding into the visual system. For the sake of discussion, then, let us assume that an inner scribe does exist, with its (covert) output feeding into the visual system (the inner eye). How would this influence subjects’ performance on imagery tasks? First, note that the inner scribe would not influence a large number of tasks. As we have seen, the stimulus support provided by the scribe would be relevant only t o some species of discovery, and only to some manipulations of one’s own images. But the stimulus support provided by the inner scribe could facilitate many other imagery tasks - namely those requiring a change in the image’s reference frame. For example, some of the discoveries documented by Finke and Slayton (1988) seem to fall in this category. We would expect, therefore, that the rate of creative discoveries documented by Finke and Slayton might increase if subjects were encouraged to enact their images - i.e., to sketch out, with some covert motions, the imaged forms. Correspondingly, it seems possible that some subjects are, on their own initiative, using this enactive strategy, and so creative strategies might be rendered less likely if enactment were prevented. Similar arguments might be applied to reversals of the duck/ rabbit figure. In the Chambers and Reisberg (1985) results, no subjects succeeded in reversing their image of this figure, but several subsequent studies have shown non-zero rates of reversal. We earlier suggested that these success reversals may be due to some subjects’ spontaneous redefinitions of the image’s reference frame, and we have discussed the fact that such redefinitions must be possible, but seem (given the data) to be relatively rare. We can now supplement this suggestion with a further hypothesis, namely, that some subjects may help themselves, with the duckhabbit image, by spontaneously enacting this image (again: covertly sketching out some pattern of motions, providing stimulus support for the visual discovery). With this stimulus support, subjects could set their initial reference frame to the side, and re-perceive the target form, leading to new discoveries. This would be entirely in line with our auditory imagery data, in which subvocalization provided a covert stimulus, capable of supporting discoveries that were not possible with imagery alone. Indeed, this enactment of visual imagery may be more likely

The Ins and Outs of Working Memory

67

when other forms of image maintenance (or image support) are ruled out, or made more difficult. For example, Brandimonte and Gerbino (in press; also Brandimonte, Hitch, & Bishop, 1992)have argued that reversals of the ducWrabbit are more likely if subjects are blocked from verbalizing while they are encoding the image. This procedure may deprive subjects of a verbal label for the form, which may in turn make the image more difficult to maintain, which may, finally, encourage subjects to rely on enactment as a means of aiding image maintenance. With the stimulus support then provided by this enactment, subjects might be more likely to reverse this figure. The remarks in the previous few paragraphs are plainly speculative, but are fully consistent with the conception we are offering: We have argued that discoveries from imagery are bounded, in a fairly rigid way, by an image’s reference frame. This boundary can be overcome in either of two ways - either by subjects changing the reference frame (spontaneously or with instruction), or by subjects creating some stimulus support (rendering their original reference frame irrelevant). These are precisely the two hypotheses just offered, concerning subjects’occasional successes in reversing the ducWrabbit figure. Which of these proposals accounts for these successes, however, is a matter for further research. In the meantime, though, these proposals serve to illustrate our discussion of the inner scribe, and, specifically, the role it may play in supporting discoveries from imagery.

Enacted Imagery, Temporary Memory, and Discovery Our discussion has by now covered a lot of ground, and so it seems worthwhile to reiterate the key points. Plainly imagery can inspire creative discoveries, and plainly one can learn new things by contemplating one’s own images. In general, though, these discoveries seem far more likely if they are compatible with the image’s initial reference frame. Discoveries not compatible with an image’s reference frame seem to happen with relatively low frequency, and, in some procedures, not at all. It is this that leads us t o argue that learning from imagery is bounded, i.e., that there are limits on what one can discover from one’s own images. We have argued, however, that these limits on imagery function

68

Daniel Reisberg & Robert Logie

can be escaped through the use of stimulus support. In many circumstances, this stimulus support is provided by self-produced stimuli; these can either be overt - for example, a self-produced picture or sound - or covert, as it is in the case of subvocalization. In general, it appears that subvocalization conveys rather large functional advantages, both in terms of stimulus support and in terms of the time-sharing it makes possible. Nothing in these advantages derives specifically from speech itself, leading us to ask whether these same (or comparable) advantages can be obtained in other ways. In particular, we have asked whether enactment of visual imagery is possible, potentially enlarging the range of what can be learned or discovered from a visual image. In many circumstances, visual imagery seems not to be enacted, and this places bounds on imagery function. Is visual imagery euer enacted? This question is legitimized by the clear diversity within the domain of visual imagery; it seems perfectly possible that some visual imagery is enacted, possibly leading t o novel depictions, and concomitant discoveries, even if other cases of visual imagery are not. The possibility of enactment is further fueled by the fact that some cases of visual imagery rely heavily on motoric imagery. More specifically, it seems that some cases of visual imagery may in fact rely on planning mechanisms within the motor system. We regard the points just listed as well-rooted in data. But all of this still leaves open the question of enacted visual imagery, and, more specifically, the possibility of an inner scribe feeding into the inner eye. In the end, we can argue for the plausibility of this partnership, but plausibility is the best we can do right now. In closing this chapter, we wish to reiterate the crucial importance of all this for imagery discovery, and for creativity. As a concrete case, informal discussions with colleagues in chemistry suggests that a great deal of discovery at the molecular level relies on imagery, specifically, on mental rotation of molecular structures, to see if two structures can fit together or dock to form a new combined structure. This form of discovery seems to require a new understanding of the molecules (of their orientation, of their interrelation), and so seems to require a reference frame switch. Thus, on our account, some form of enactment of the image seems likely to promote such discoveries. We have offered similar words about the discoveries documented in the laboratory by Finke and

The Ins and Outs of Working Memory

69

Slayton. Finally, at a more mundane level, even the planning of a novel route around a town may require an enacted image, with the enactment in this case perhaps taking the form of an imagined walk along potential routes! These potential applications present interesting research opportunit,ies,and underscore, once again, the potentially large contribution from image enactment. Thus we end this chapter on a speculative note -we simply do not know if the proposed partnership, enabling enacted visual imagery, exists. We have indulged ourselves in this speculation, though, because we believe it leads into important and uncharted territory, and so we would argue that this partnership is an important focus for future research. As we have seen, this research may illuminate the nature of working memory, and also several theoretical issues (e-g., the role of stimulus support, and thus the contrast between mental representations and external stimuli). Above all, this research may illuminate the nature of imagery hnction, and, obviously, the issue of what limits there may be on this function. We believe that our exploration of these issues, within this chapter, has already clarified these points to some degree, and has, in fact, led us to new questions about imagery, about working memory, and about creativity. We believe that further exploration along these lines will continue to enrich our understanding of these domains.

References Anderson, R. A., Essick, G. K., & Siegel, R. M. (1987). Neurons of Area 7 activated by both visual stimuli and oculomotor behavior. Experimental Brain Research, 67, 316 -322. Baddeley, A. D. (1966). Short-term memory for word sequences as a function of acoustic, semantic, and formal similarity. Quarterly Journal of Experimental Psychology, 18, 362-365. Baddeley, A. D. (1986). Working memory. Oxford: Oxford University Press. Baddeley, A. D. (1990). Human memory: Theory and practice. Hove: Erlbaum. Baddeley, A. D. (1992). Is working memory working? The fifteenth Bartlett lecture. Quarterly Journal of Experimental Psychology,

70

Daniel Reisberg & Robert Logie

44A, 1-3 1. Baddeley, A. D., & Lewis, V. (1981). Inner active processes in reading: The inner voice, the inner ear and the inner eye. In A. M. Lesgold & C. A. Perfetti (Ed.), Interactive processes in reading (pp. 107-129). Hillsdale, NJ:Erlbaum. Baddeley, A. D., & Lewis, V. J. (1984). When does rapid presentation enhance digit span. Bulletin of the Psychonomic Society, 22,403-405. Baddeley, A. D., & Lieberman, K. (1980). Spatial working memory. In R. Nickerson (Ed.),Attention and performance (pp. 521-539). Hillsdale, N J Erlbaum. Baddeley, A. D., Papagno, C., & Vallar, G. (1988). When long-term learning depends on short-term storage. Journal of Memory and Language, 27,586-595. Baddeley, A. D., Thomson, N., & Buchanan, M. (1975).Word length and the structure of short-term memory. Journal of Verbal Learning and Verbal Behavior, 14, 575-589. Bellugi, U., Klima, E. S., & Siple, P. (1975). Remembering in signs. Cognition, 3, 93-125. Besner, D. (1987). Phonology, lexical access in reading, and articulatory suppression: A critical review. Quarterly Journal of Experimental Psychology, 39A, 467-478. Besner, D., Davies, J., & Daniels, S. (1981). Reading for meaning: The effects of concurrent articulation. Quarterly Journal of Experimental Psychology, 33, 415- 437. Brandimonte, M. A., & Gerbino, W. (in press). Mental image reversal and verbal recoding: When ducks become rabbits. Memory & Cognition. Brandimonte, M. A., Hitch, G . J., & Bishop, D. V. M. (1992). Influence of short-term memory codes on visual image processing: Evidence from image transformation tasks. Journal of Experimental Psychology: Learning, Memory, and Cognition, 18,157-165. Brooks, L. (1968).Spatial and verbal components of the act of recall. Canadian Journal of Psychology, 22, 349-368. Bruner, J. (1957). Contemporary approaches to cognition. Cambridge, MA:Harvard University Press. Bruner, J. S . (1973). Beyond the information given. New York Norton.

The Ins and Outs of Working Memory

71

Butters, N., Barton, M., & Brody, B. A. (1970). Role of the parietal lobe in the mediation of cross modal associations and reversible operations in space. Cortex, 6, 174-190. Campbell, R.(1992).Speech in the head? Rhyme skill, reading and immediate memory in the deaf. In D. Reisberg (Ed.), Auditory imagery (pp. 73-93). Hillsdale, NJ: Erlbaum. Carpenter, P. A., & Eisenberg, P. (1978). Mental rotation and the frame of reference in blind and sighted individuals. Perception & Psychophysics, 23, 117-124. Chambers, D., & Reisberg, D. (1985). Can mental images be ambiguous? Journal of Experimental Psychology: Human Perception and Performance, 11, 317-328. Conrad, R. (1964). Acoustic confusions in immediate memory. British Journal of Psychology, 55, 75-84. Dean, G.,& Morris, P. E. (1991).Imagery and spatial ability: When introspective reports predict performance. In R. H. Logie & M. Denis (Eds.), Mental images in human cognition (pp. 331-347). Amsterdam: North-Holland. Della Sala, S., & Logie, R. H. (in press). When working memory does not work: The role of working memory in neuropsychology. In F. Boller & H. Spinnler (Eds.), Handbook of neuropsychology, Vol. 8. Amsterdam: Elsevier. Della Sala, S., Logie, R. H., Marchetti, C., & Wynn, V. (1991). Case studies in working memory: A case for single cases. Cortex, 27, 169-191. Engelkamp, J. (1986). Motor programs as part of the meaning of verbal items. In I. Kurcz, E. Shugar, & J. H. Danks (Eds.), Knowledge and language (pp. 115-138). Amsterdam: NorthHolland. Engelkamp, J. (1991). Imagery and enactment in paired-associate learning. In R. H. Logie & M. Denis (Eds.), Mental images in human cognition (pp. 119-128). Amsterdam: Elsevier. Farah, M. J., Hammond, K. M., Levine, D. N., & Calvanio, R. (1988). Visual and spatial mental imagery: Dissociable systems of representation. Cognitive Psychology, 20, 439-462. Finke, R. (1989). Principles of mental imagery. Cambridge, MA: MIT Press. Finke, R., & Kosslyn, S. (1980). Mental imagery acuity in the peripheral visual field. Journal of Experimental Psychology:

72

Daniel Reisberg & Robert Logie

Human Perception and Performance, 6, 126-139. Finke, R., & Kurtzman, H.(1981). Mapping the visual field in mental imagery. Journal of Experimental Psychology: General, 110,501-517. Finke, R., & Slayton, K. (1988). Explorations of creative visual synthesis in mental imagery. Memory & Cognition, 16, 252-257. Gathercole, S., & Baddeley, A. D. (1989).Evaluation of the role of phonological STM in the development of vocabulary in children: A longitudinal study. Journal of Memory and Language, 28, 200-213. Gathercole, S., & Baddeley, A. D. (1990). The role of phonological memory in vocabulary acquisition: A study of young children learning arbitrary names of toys. British Journal of Psychology, 81,439-454. Georgopoulos, A. P., Kalaska, J. F., Caminiti, R., BE Massey, J. T. (1982). On the relations between the direction of twodimensional arm movements and cell discharge in primate motor cortex. Journal of Neuroscience, 2 , 1527-1537. Georgopoulos, A. P.,Kalaska, J. F., Caminiti, R., & Massey, J. T. (1983). Spatial coding of movement: A hypothesis concerning the coding of movement direction by motor cortical populations. In J. Massion, J. Paillard, W. Schultz, & M. Wiesendanger (Eds.), Neural Coding of Motor Performance. Experimental Brain Research Supplement, 7 , 327-336. Georgopoulos, A., Lurito, J., Petrides, M., Schwartz, A,, & Massey, J. (1989). Mental rotation of the neuronal population vector. Science, 243,234-236. Hadden, S. (1991). Investigation of self-reports of imagery vividness and the nature of memory for colors. Unpublished Bachelor’s Thesis. Reed College, Portland, OR. Heuer, F., Fischman, D., & Reisberg, D. (1986). Why does vivid imagery hurt colour memory? Canadian Journal of Psychology, 40,161-175. Hinton, G. (1979). Some demonstrations of the effects of structural descriptions in mental imagery. Cognitive Science, 3, 231-250. Hochberg, J., & Peterson, M. (1987). Piecemeal organization and cognitive components in object perception: Perceptually coupled responses to moving objects. Journal of Experimental

The Ins and Outs of Working Memory

73

Psychology: General, 116, 370-380. Hyman, I., & Neisser, U. (1991). Reconstruing mental images: Problems of method (Report #19). Atlanta, G A Emory Cognition Project, Emory University. Johnson, P. (1982). The functional equivalence of imagery and movement. Quarterly Journal of Experimental Psychology, 34A, 349-365. Jonides, J., Kahn, R., & Rozin, P. (1975). Imagery instructions improve memory in blind subjects. Bulletin of the Psychonomic Society, 5,424-426. Kanizsa, G.(1979). Organization in vision. New York: Praeger. Kerr, N. H. (1983). The role of vision in 'lvisual imagery" experiments: Evidence from the congenitally blind. Journal of Experimental Psychology: General, 112, 265-277. Kimura, Y., & Bryant, P. (1983). Reading and writing in English and Japanese: A cross-cultural study of young children. British Journal of Developmental Psychology, 1, 143-154. Kosslyn, S. M. (1983).Ghosts in the mind's machine. New York: W . W. Norton. Kosslyn, S. M. (1991).A cognitive neuroscience of visual cognition: Further developments. In R. H. Logie and M. Denis (Eds.), Mental images in human cognition (pp. 351-381). Amsterdam: North-Holland. Levy, B. (1971).The role of articulation in auditory and visual shortterm memory. Journal of Verbal Learning and Verbal Behavior, 10,123-132. Logie, R. H.(1986). Visuo-spat?ialprocessing in working memory. Quarterly Journal of Experimental Psychology, 38A, 229-247. Logie, R.H. (1989). Characteristics of visual short-term memory. European Journal of Cognitive Psychology, 1, 275-284. Logie, R. H. (1991). Visuo-spatial short-term memory: Visual working memory or visual buffer. In C. Cornoldi & M. McDaniel (Eds.), Imagery and cognition (pp. 77-102). New York: Springer-Verlag. Logie, R. H. (in press). Working memory in everyday cognition. In G. M. Davies & R. H. Logie (Eds.), Working memory in everyday life. Amsterdam: Elsevier. Logie, R. H., & Baddeley, A. ID. (1990). Imagery and working memory. In P. J. Hampson, D. F. Marks, & J. T. E. Richardson

Daniel Reisberg & Robert Logie

74

(Eds.), Imagery: Current developments (pp. 103-128). London: Routledge Logie, R. H., & Marchetti, C. (1991).Visuo-spatial working memory: Visual, spatial or central executive. In R. H. Lode & M. Denis (Eds.), Mental images in human cognition (pp. 105-115). Amsterdam: Elsevier. Logie, R. H., Zucco, G. M., & Baddeley, A. D. (1990). Interference with visual short-term memory. Acta Psychologica, 75,55-74. Marmor, G.S.,& Zabeck, L. A. (1976).Mental rotation by the blind: Does mental rotation depend on visual imagery? Journal of Experimental Psychology: Human Perception and Performance, 2,515-521. Matthews, W. A. (1983).The effects of concurrent secondary tasks on the use of imagery in a free recall task. Acta Psychologica, 53,231-241. Miller, A. I. (1986).Imagery in scientific thought. Cambridge, MA: MIT Press. Morris, N. (1987). Exploring the visuo-spatial scratch pad. Quarterly Journal of Experimental Psychology, 39A, 409-430. Murray, D. (1968).Articulation and acoustic confusability in shortterm memory. Journal of Experimental Psychology, 78, 679-684. Olson, D., & Bialystok, E. (1983).Spatial cognition. Hillsdale, N J : Erlbaum. Paivio, A., & Okovita, H.W. (1971). Word imagery modalities and associative learning in blind and sighted subjects. Journal of Verbal Learning and Verbal Behavior, 10,506-510. Papagno, C., Valentine, T., & Baddeley, A. D. (1991).Phonological short-term memory and foreign-language vocabulary learning. Journal of Memory & Language, 30,331-347. Phillips, W. A,, & Christie, D. F. M. (1977). Components of visual memory. Quarterly Journal of Experimental Psychology, 29,

117-133. Pinker, S., & Finke, R.(1980).Emergent two-dimensional patterns in images in depth. Journal of Experimental Psychology: Human Perception and Performance, 6, 244-264. Quinn, J. G. (1988). Interference effects in the visuo-spatial sketchpad. In M. Denis, J. Engelkamp, & J. T. E. Richardson (Eds.), Cognitive and neuropsychological approaches to mental

The Ins and Outs of Working Memory

75

imagery (pp. 181-189). Dordrecht: Martinus Nijhoff. Quinn, J. G. (1991). Encoding and maintenance of information in visual working memory. In R. H. Logie & M. Denis (Eds.), Mental images in human cognition (pp. 105-115). Amsterdam: Elsevier. Quinn, J. G., & Ralston, G. E., (1986).Movement and attention in visual working memory. Quarterly Journal of Experimental Psychology, 38A, 689-703. Reisberg, D. (1987).External representations and the advantages of externalizing one's thought. In E. Hunt (Ed.), The ninth annual conference of the cognitive science society (pp. 281-293). Hillsdale, N J Erlbaum. Reisberg, D., & Chambers, D. (1991). Neither pictures nor propositions: What can we learn from a mental image? Canadian Journal of Psychology, 45, 288-302. Reisberg, D., & Leak, S. (1987). Visual imagery and memory for appearance: Does Clark Gable or George C. Scott have bushier eyebrows? Canadian Journal of Psychology, 41, 521-526. Reisberg, D., Rappaport, I., & O'Shaughnessy, M. (1984).The limits of working memory: The digit digit-span. Journal of Experimental Psychology: Learning, Memory and Cognition, 10, 203-221. Reisberg, D., Smith, J. D., Baxter, D. A., & Sonenshine, M. (1989). "Enacted" auditory images are ambiguous; "pure" auditory images are not. Quarterly Journal of Experimental Psychology, 41A,619-641. Reisberg, D., Wilson, M., & Smith, J. D. (1991). Auditory imagery and inner speech. In R. Logie & M. Denis (Eds.), Mental images in human cognition (pp. 59- 81). Amsterdam: Elsevier. Richardson, J. T. E. (1980). Mental imagery and human memory. New York: St. Martin's Press. Richardson, J. T. E. (1988). Phonology and reading: The effects of articulatory suppression upon homophony and rhyme judgments. Language and Cognitive Processes, 2 , 229-244. Rock, I. (1973). Orientation and form. New York: Academic Press. Saltz, 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. Shand, M.(1982). Sign-based short-term coding of American Sign

76

Daniel Reisberg & Robert Logie

Language signs and printed English words by congenitally deaf signers. Cognitive Psychology, 14, 1-12. Shepard, R. N., & Cooper, L. A. (1982). Mental images and their transformations. Cambridge, MA. MIT Press. Slowiaczek, M., & Clifton, C. (1980).Subvocalization and reading for meaning. Journal of Verbal Learning and Verbal Behavior, 19, 573-582. Smith, J. D., Reisberg, D., & Wilson, M. (1992). The role of inner speech in auditory imagery. In D. Reisberg (Ed.), Auditory imagery (pp. 95-119). Hillsdale, NJ: Erlbaum. Smyth, M. M., & Pendleton, L. R., (1989). Working memory for movements. Quarterly Journal of Experimental Psychology, 41A, 235-250. Vallar, G., & Baddeley, A. D. (1982).Short-term forgetting and the Quarterly Journal of Experimental articulatory loop. Psychology, 34, 53-60. Vallar, G., & Baddeley, A. D. (1984). Fractionation of working memory: Neuropsychological evidence for a phonological shortterm store. Journal of Verbal Learning and Verbal Behavior, 23,151-161. Wilding, J., & White, W. (1985). Impairment of rhyme judgements by silent and overt articulatory suppression. Quarterly Journal of Experimental Psychology, 37A, 95-107. Zimler, J., & Keenan, J. M. (1983). Imagery in the congenitally blind: How visual are visual images? Journal of Experimental Psychology: Learning, Memory, and Cognition, 9,269-282. Zivin, G. (1986).Image or neural coding of inner speech and agency? The Behavioral and Brain Sciences, 9,534-535.

Imagery, Creativity, and Discovery: A Cognitive Pcrspective

B. Roskos-Ewoidson. M.J.Intons-Petersonand R.E. Anderson (Editors) 8 1993 Elsevier Science Publishers B.V. All rights reserved.

77

Chanter 3

BOTH DEPICTIVE AND DESCRIPTIVE IMAGES ARE

Deborah Chambers Department of Psychology North Dakota State University Fargo, ND 58102 USA Pictures or Propositions How are mental images represented? This question has been extensively debated, with proponents arguing that images are like either pictures or propositions (Kosslyn, 1981; Pylyshyn, 1981). The pictorialists argue that images depict what they represent in an analogue form, whereas the propositionalists argue that images are a description of a scene or object. These two choices have in the past been thought to be mutually exclusive - either images are like pictures or like propositions. One issue at the heart of this distinction is whether mental images have structural properties in common with actual physical objects such as pictures. That is, do mental images depict what they represent? If so, imagers should be able to reconsider their representations and find new interpretations of their images, similar to the way that perceivers find new interpretations of pictures. As can be seen in other chapters in this volume (see chapters by Hyman, Kaufmann & Helstrup, and Peterson), research indicates that imagers can reinterpret their mental images, suggesting that mental images do contain information that is subject to inspection and reconstrual. This similarity between mental images and pictures may make mental images a form of representation that is particularly well suited to creative discovery (see Anderson & Helstrup, this volume, and Finke, this volume).

78

Deborah Chambers

The research indicates that mental images share characteristics with physical pictures. But how far can we extend this comparison between images and pictures? At the extreme, we could argue that mental images, like pictures, are raw data structures that must be interpreted for imagers to apprehend their meaning. In this way, interpretations of images stand independently of the structural properties of images (or the depiction of images). However, our research (Chambers & Reisberg, 1992;Reisberg & Chambers, 1991) suggests that this characterization goes too far. We have found that mental images are both depictive like pictures and descriptive like propositions. Further, we find that both the depictive and descriptive properties contribute to what can be discovered from mental images, In fact, our research demonstrates that the descriptive aspects often control what is to be depicted within the image (Chambers & Reisberg, 1992). This chapter will examine the relation between the depictive and descriptive aspects of images.

Images Depict Imagers often claim that entertaining a mental image is just like perceiving a picture in the world. That is, when inspecting a mental image some feel that they are inspecting a representation that retains spatial layout and depth relation among objects, as well as size, color, shape, and texture. This experience of entertaining a mental picture is often so compelling that many imagers responding to Marks’Visual Inventory Questionnaire claim that their images are as clear and vivid as normal percepts (Marks, 1972). Experimental research supports the introspective report that images are similar to pictures. For example, research on mental scanning indicates that when subjects are asked to scan across an image, or rotate an image, zoom in on a n image to inspect detail, or pan back from on to make gross comparisons, their response times were all highly correlated with the time it takes to inspect pictures in the world. (For reviews of this large literature see Kosslyn, 1980, 1983;and Shepard & Cooper, 1982). This research indicates that images, like pictures, preserve the metric properties of space. Images have also been found to function much like pictures. Visual illusions such as prism-induced displacement (Finke, 1979)

Images are Both Depictive and Descriptive

79

were found to be effective when subject imagined the stimulus situations. Likewise, modality-specific interference effects between imagery and perception suggest that the two may be processed by the same systems (Segal & Fusella, 1970). Clearly, images and pictures are similar in many ways. This similarity is evident in both introspective reports and experimental results. At the same time, there are a number of convincing arguments that images must differ from pictures because images, unlike pictures, are mental representations. Because images are mental representations, whereas pictures are not, they will have properties in common with other representations, such as percepts, that pictures will not have. For example, mental representations are characterized by intentionality; that is, they are meaningful. As Bretano (1874/1973, p. 88) states, “Every mental phenomenon is characterized . . . by what we might call . . . reference t o a content, direction toward an object.” Therefore, at the very least, mental images must be meaningful depictions, describing a specific view of a scene or an object. Pictures, on the other hand, can often be ambiguous as to what they represent. To take an example from Fodor (19751, a picture of a woman with a protruding belly may represent a woman who is pregnant or a woman who is overweight. There is no way from the picture alone to determine what, it is meant to represent. While the picture is ambiguous, a person’s interpretation, or thought about the picture, is clearly one or the other. If images are to serve as a vehicle of thought, they, like percepts and other mental representations, must include meaningful, descriptive information. In other words, images must be determinant representations (Fodor, 1981). Therefore, if an image resembles a picture then it must, at the very least, have a caption that will identify what the image represents.

Images are Meaningful Our own work is a clear demonstration that images are meaningful representations, i.e., descriptive information must accompany depictive information. We have shown that mental images of classical ambiguous figures are very difficult t o reverse

80

Deborah Chambers

(Chambers & Reisberg, 1985). In these studies, subjects were asked to create a mental image of an ambiguous figure (such as Jastrow's duckhabbit; see Figure 1). After a brief inspection period, subjects were instructed that the picture they imaged had an alternative interpretation and were asked to inspect their image to discover the new view. Subjects were told that furating on the right or left side of their image would help them make this discovery. If subjects failed they were asked to draw their image on paper and to inspect their drawing for the alternative figure.

Figure 1 Jaatrow's duckhabbit. Test stimulusemployedby Chambers and Reisberg (1986).

All of our subjects failed to find a new figure within their images of ambiguous forms; however, they unanimously discovered the alternative within their drawings. That is, subjects who created an image of the duck of Jastrow's figure were unable to find a rabbit within their image (and vice versa), but were able to discover the rabbit within their drawing. We argue that this descriptive information constrains imaginal discovery (or reinterpretation). Pictures, unlike images, are free of this interpretive information and,

Images are Both Depictive and Descriptive

81

thus, can easily be reinterpreted. While our data indicated a zero rate of reversal, data collected by Hyman and Neisser (1991)and by Peterson, Kihlstrom, Rose, and Glisky (1992)have both shown approximately a 10% reversal rate (i.e., approximately 2 to 3 successes out of 20 subjects) under similar experimental conditions. Despite the discrepancy, we can safely say that it is very difficult to reverse images of ambiguous figures. These results present a clear difference between images and pictures. Images are clearly influenced by subjects' intentions to image a particular object. When viewing a picture of the duckhabbit figure, subjects are easily able to alternate between construals. However, the same is not true when inspecting a mental image of the figure. It appears that subjects' intentions limit what they can discover from their images. At first glance, the Chambers and Reisberg (1985)results appear to argue that subjects cannot discover a new form by inspecting their mental images. However, the imagery literature quite persuasively argues against this interpretation. In fact, there is evidence that subjects can inspect their images and discover new relations that were not anticipated at the time of the image's creation. For example, we can inspect our images and discover what a three-dimensional scene might look like from another viewing perspective (Pinker & Finke, 1980). We can also reorganize two letters of the alphabet and discover new shapes (Finke, Pinker, & Farah, 1989). Subjects who were instructed to image a "D" rotated counterclockwise by 90" and place a "J"in the middle of its spine often reported that their image resembled an umbrella. In addition, imagers can create objects that were originally unanticipated from a group of parts by combining and arranging them in imagery (Anderson & Helstrup, this volume; Finke, this volume; Finke & Slayton, 1988). For example, imagers might create a television from a square, a circle, and the letter V). These data demonstrate that subjects can make discoveries from their images, but the Chambers and Reisberg (1985)findings indicate that what can be discovered is limited by subjects' intentions to create an image of a particular object.

82

Deborah Chambers Constraints on Image Discovery

In searching for possible factors that might act to constrain image discovery, we turned to the perception literature. We reasoned that factors which contribute to the overall appearance of the object (i.e., our percepts of an object, which includes both descriptive and depictive information) may act to limit what can be discovered from an image, because visual images, by their very nature, represent the appearance of an object or scene (Reisberg & Chambers, 1991). For example, Rock (1973) has demonstrated that the way a perceiver specifies the orientation of an object contributes to the appearance of that object. Rock made this point by demonstrating that subjects often fail to recognize that Figure 2 is a drawing of Africa rotated 90" counterclockwise. Similarly, we have found that how imagers designate the top of an image limits what they can discover from their images (Reisberg & Chambers, 1991). In our experiments, subjects imaging Figure 3 were unable to discover Texas, even when they imagined rotating it 90" clockwise, which places it in its usual upright orientation. In sharp contrast, we found a dramatic reversal of our results after we altered the instructions. When we asked subjects to change their specification of the top of the figure (i.e., change how they understood the top), approximately half of our subjects were able to identify the Lone Star State from their image of the initially unidentifiable figure. Thus, how the imager specifies the top of the image appears to control what can be discovered from an image. Both Hyman and Neisser (1991) and Peterson et al. (1992) have found that even ambiguous figures, like the ducklrabbit, could be reversed in imagery if subjects were given clues leading them to change their understanding of the front and back of their images. Reisberg and Chambers (1991) found similar results when they manipulated imagers' understanding of the figure-ground relation within their images. The Reisberg and Chambers studies (1991) make it clear that image discovery is not entirely based on the pictorial (or depictive) properties of an image; rather, learning depends on both the depictive and the descriptive properties. If either of these aspects is incongruent with a particular representation, as in the case of the rotated Texas, the image simply represents some other object - in

Images are Both Depictive and Descriptive

83

Figure 2 Africa rotated 90" counterclockwise. After Rock (1973).

this case, an unidentifiable blob. It is obvious that if the depictive properties of our test figure were different (e.g., if the figure had a different shape) it would not resemble Texas. It is also true that if the perceiver (or the imager) attributes a different orientation or figure-ground relationship to the figure it would not appear to look like Texas. In other words, a percept "goes beyond the information given" by virtue of specifying a figure-ground organization for the perceived form, an orientation, a configuration in depth, and so forth. This specifying information is critical to what the percept will represent, what it will resemble, and what it will evoke from memory. The Reisberg and Chambers (1991) data make it clear that the same is true for images. (For a detailed discussion of these data, see Reisberg & Chambers, 1991). The essential claim being made here is that one's understanding

84

Deborah Chambers

Figure 3 Texas rotated 90"counterclockwise. After Reisberg and Chambers (1991).

of an image places boundaries on what one can discover from an image. Imagers' intentions to imagine, for example a duck, will lead them to create an image with a particular set of specifications, including figure-ground and depth relations, as well as a particular orientation (or, in Peterson's language, a particular reference frame; see Peterson chapter, this volume). These specifications will shape what the image resembles. Assuming that learning from images is based on appearance information, these specifications will also limit what can be learned from an image.

Images are Both Depictive and Descriptive

85

What is the Relationship Between the Depictive and Descriptive Aspects of Images? The previously described data tell us that images contain both depictive and descriptive information. Subjects’ success in recognizing forms in their image (with a change in the specifications of their images) attests to the fact that images do retain appearance information. Likewise, the failure to recognize an imaged form when it is accompanied by incorrect specifications attests to the fact that the descriptive information is critical. But what is the relationship between the descriptive and the depictive information? It is possible that the depictive and descriptive information exist independently, much like a picture and its caption. If the two are independent, altering one will not necessarily lead to the alteration of the other, although it will change how imagers understand their images. Alternatively, the depictive and descriptive information might be combined in a unified representation, such as a structural description. In this case, a change in the descriptive information would also lead to a change in the depictive information. Again, the perception literature gave us some clues as to the type of interaction we might expect. Tsal and Kolbet (1985) demonstrated that when entertaining a picture of the duckhabbit figure, a change in construal causes a corresponding change in the focus of attention. That is, Tsal and Kolbet demonstrated that when subjects are viewing the duckhabbit figure, attention (measured by increased sensitivity to detect a probe) is deployed to the area that corresponds to the perceiver’s designation of the face of the animal. They demonstrated that when subjects were fixating on the right side of the figure they were more likely to perceive a rabbit, and if they fEated on the left they were more likely to perceive a duck. If imagery operates like perception, we can expect that imagers, like perceivers, would be more likely to attend to areas of their image that represent the face of the intended animal. Based on studies of performance with imagery tasks, Kosslyn (1980,1983) has argued that attention plays a role in maintaining an image. Kosslyn has argued that images begin to fade as soon as they are constructed, but that one can maintain the image by a refreshing process linked to mental scanning or attention. Further, Kosslyn

86

Deborah Chambers

argues that the resources needed for this refreshing process are limited; therefore, if the image is large, or complex, only some parts of the image can be represented at one time. This leads to the possibility that images do not include all aspects of the form they were meant to represent; i.e., images are vague about some aspects of their depiction. Given Tsal and Kolbet's results concerning attention, it seems quite likely that what is included in the image and what is allowed to fade is influenced by imagers' intentions. In this way, the description would actually be shaping the depiction.

Images are Vague We are not the first to suggest that images can be vague about some aspects of their depiction. Based on introspective evidence, James (1880/1950), KoMra (cited in Arnheim, 1969) and Titchener (1926) all agreed that images rarely include all aspects of their depiction. For example, KoMra noted that one subject reported imaging a coin with no denomination, another reported imagine a train but was unable to tell whether it was a freight or passenger train. More recently, Slee (1980) has exploited this as a measure of image vividness, asking to what extent an individual's image included aspects that would be critical to a picture of an object. For example, if a subject images a cup placed on a table, does the image include the shape of the table, or its color? Slee has shown that subjects vary on the amount of detail they include in their images. Slee's studies also point to another difference between images and pictures. Slee's subjects often report that their image included details such as the placement of the cup and the box on the table, but failed to include other details such as the shape of the table. Obviously, a picture of this scene could not fail to include the shape of the table, but images can be noncommittal on aspects that are critical t o pictures. Dennett (1981) has argued that this "noncommittal" aspect of images is a factor that clearly distinguishes mental representations (e.g., images and percepts) from physical representations (e.g., pictures and photographs). This evidence converges on findings from perception indicating that percepts, like images, ''are not everywhere dense". For example, Rock, Halper, and Clayton (1972) demonstrated that when we

Images are Both Depictive and Descriptive

87

perceive objects we are often insensitive to many nuances of complex figures, especially when they do not change the global shape of the figure. They showed subjects a complex nonsense form for a few seconds, then removed the form and immediately presented two alternatives for a forced-choice recognition task. One of the alternatives was the original figure. The other test figure shared the global shape of the original figure, but differed in a small nuance of one section of the contour. Recognition performance was a chance levels, despite the lengthy study time and the zero retention interval. However, when the differing segments of the test figures were isolated, so that they were perceived as figures (rather than details of a larger shape), performance was quite good: subjects easily recognized the previously viewed form. From a somewhat different perspective, Hochberg (1981, 1982) similarly concludes that the "schematic map," which in his view constitutes a form percept, is not "everywhere dense." In several experiments, Hochberg (Hochberg, 1981,1982; Hochberg & Peterson, 1987)has demonstrated that the percept is dominated by "local cues'' within foveal view, and includes only a vague impression of information elsewhere in the form. Again, this argues that the percept does not include all details that are clearly visible within a stimulus, but is selective in important ways. Thus mental representations such as images and percepts are not literal or complete transcriptions of an object, but are instead selective: clear about some aspects of a scene or object, vague and indeterminate about others.

Does the Description Shape the Depiction? We hypothesized that what is depicted in an image and what is left vague depends on subjects' understanding of their images. In part, this understanding includes designations of orientation, figureground, depth relations, and so forth, and it also includes subjects' understanding of what the object is meant t o represent. To test this hypothesis we employed the diickhabbit figure and a paradigm similar to that used by Rock et al. (1972). Given Tsal and Kolbet's data and the arguments that indicate that images do no include all aspects of the imagined object, we

88

Deborah Chambers

reasoned that if subjects imaging the duck deploy attention differentially to the left side of the figure, then they probably have a more elaborated image of the duck’s head than the rabbit’s head. Hence, in a recognition test of slightly altered drawings of the ducWrabbit figure, subjects imaging a duck would be more likely to detect errors in the duck’s face, but would have a more difficult time detecting errors in the rabbit’s face. Just the opposite would be expected if subjects construed their image as a rabbit. In our experiments we asked subjects to image the duck or rabbit of Jastrow’s figure, then asked them which of the two pictorially presented shapes better resembled the imaged form. As in the study of Rock et al. (1972)’the test shapes differed only in small nuances of one section of the figure. Again, we hypothesized that imagery subjects, like Tsal and Kolbet’s perception subjects, attend to the area of their images depicting what they consider to be the face of the imaged animal. Consequently, subjects are likely to maintain this area of their images and to allow other areas t o fade. As a result, subjects should have a clear image of the animal’s face, but not a clear image of the rest of the figure. We expected, therefore, that subjects would easily detect departures from the original figure when the departures were located on areas that corresponded to what they considered to be the animal’s face, but do poorly otherwise. Thus, in our task we predicted that subjects imaging the duck would reliably choose the unaltered Jastrow figure if offered a choice between it (Figure 4a) and an alternative differing in the contour of the duck’s face (Figure 4b). (See Chambers & Reisberg, 1991, for a complete description of these studies.) Likewise, subjects imaging the rabbit would choose the unaltered figure if offered a choice between it (Figure 4a) and an alternative differing in the contour of the rabbit’s face (Figure 4c). Conversely, subjects would be less sensitive to the difference between the test stimuli when these differ on the “back”of the animal’s head. In the extreme, subjects would choose randomly between the original and the modified figure in this condition. It is possible that the modifications that we made to our test stimuli altered them in such a way to make them more or less prototypical of a duck or a rabbit. For example, the alterations made to the duck’s bill (Figure 4b) could potentially have produced a more

Images are Both Depictive and Descriptive

a)

Figure 4 Test stimuli employed by Chambers and Reisberg (1992): (a)Unmodified figure; (b) modification on the duck’s bill; (c) modification on the rabbit’s nose.

89

90

Deborah Chambers

rabbit-like figure, leading subjects thinking of their image as a duck to choose the original figure merely because it was more duck-like. Likewise, alterations made to the rabbit’s face may have produced a more duck-like figure, again leading subjects thinking of their image as a rabbit to choose the original figure. To investigate this possibility, we showed our figures to subjects who were led to expect either pictures of ducks or rabbits. They were asked to select among a set of test figures the most duck-like or rabbit-like figure. The results of these studies showed that the alterations made to our test stimuli did not significantly alter the prototypicality of the figures (for more details see Chambers & Reisberg, 1992). The main experiments employed a 2x2 design, with subjects either imaging the Jastrow figure as a duck or as a rabbit, then tested with a choice between the original and a stimulus distorted on the rabbit’s face. The results of these experiments were just as we predicted; Figure 5 shows the results from one of these procedures. However, we need to be cautions in interpreting these results. For example, it is possible that our results reflect a bias in perceptual encoding. that is, when subjects were asked to encode the figure they may have simply encoded the nuances of the attended side of the figure and failed to encode the nuances of the unattended side. Therefore, the results may be due to a bias in perceptual encoding, rather than the processes underlying the creation of an image. To eliminate this possibility, we repeated the above experiment with one critical change. After the subjects had encoded the figure and the test stimulus was removed from view, they were instructed that the figure could be viewed as a duck (for those who initially construed the figure as a rabbit). With this direct and explicit instruction, many subjects were able to reconstrue their image. Once this reconstrual was in place, subjects were tested in the same manner as in the above experiment. As shown in Figure 6 , subjects’ discrimination choices were completely predictable from the construal of the image they had in mind at the time of testing, and not from the construal they had in mind when initially viewing (and memorizing) the figure. Apparently, subjects who understood their image as a duck had a clear image of the duck’s face (as reflected in the accurate discrimination with the relevant test pair). However, when subjects changed their construal of the image they allowed this

Images are Both Dcpictive and Descriptive

91

Subjects chose between the original figure and one modified on the

90

40

-

* Duck'sface

1

Rabbit's face

1 Duck

Rabbit

Subject's Image Figure 5 Percent correct recognition. ("Correct" in this case means choosing the original form.) After Chambers and Reisberg (1992, Experiment 2).

portion of their image t o face or degrade, as revealed by their poor performance with the relevant test pair. By the same token, these subjects initially had only a vague image of the back of the duck's head; when the subjects changed their construal, they ''restored" this contour to make this section of the image precise. The data indicate that images, like percepts, can often be vague about some aspects of their depiction. Further, areas of vagueness can be predicted by knowing which construal subjects are imaging.

Deborah Chambers

92

Subjects chose between the original figure and one modified on the V

"1

Duck'sface

Rabbit's face

40

Rabbit Subject's Image AT TIME OF TEST Duck

Figure 6 Percent correct recognition. Experiment 4).

After Chambers and Reisberg (1992,

Tsal and Kolbet (1985) have shown that when subjects are viewing the duckhabbit figure as a picture, attention is deployed to the side of the figure that depicts the face of the animal that the subject is perceiving. Similarly, our data indicates that the face of the animal that subjects are imaging is more likely t o be fully articulated in the image than the back of the animal's head.

Images are Both Depictive and Descriptive

93

Discussion Our results demonstrate the importance of both the depictive and the descriptive aspects of images. Images, like pictures, are depictive in that they represent size, spatial layout, a specific viewing position, and so on. However, images are also descriptive, representing a specific view of a scene or object. This specificity is accomplished by organizational properties such as orientation, figureground relations, unit formation, configuration in depth, and by the aspects of a scene or object that are depicted in the image. It is clear from our data that both the depictive and descriptive aspects of images play a role in what is represented in images and, subsequently, what can be discovered from images. The Reisberg and Chambers (1991) results show that the description is inseparable from the depiction (see Reisberg & Logie, this volume). Subjects who have mentally rotated an imaged form into its usual orientation, but who describe the image as having a different top, simply do not recognize the form even though the rotated form is isomorphic with the target figure. For an image t o remind us of a new object both the description and depiction must be congruent (see Hyman, this volume; Peterson, this volume, Reisberg & Chambers, 1991; and Reisberg & Logie, this volume). In addition, the Chambers and Reisberg (1992) data argue that the descriptive and depictive aspects are interactive. The description appears t o dictate what is to be specified and what is to be left vague in an image. In this way, an image of Jastrow’s duck is simply different from an image of Jastrow’s rabbit, including different specifications of orientation and also depicting different aspects of the form. Therefore, it is not surprising that subjects do not easily reconstrue their images of the ducldrabbit figure because an image of a duck is simply different from an image of a rabbit. This chapter has focused on the limitations of learning from images. By using paradigms that demonstrate both failures and successes in image discovery, we have been able to reveal the importance of both the depictive and descriptive aspects of images. However, this focus on the limitations of image discovery should not overshadow imagers’ ability to learn from their images. Images, like all other forms of thought, can lead us to new ideas. Clearly, many

94

Deborah Chambers

of the chapters in this volume demonstrate that images can support creative discovery, Visual images represent the appearance of a scene or object and changes in the appearance specifications may lead imagers t o discover a new object within their images. As discussed above, Finke et al. (1989) demonstrated that through a series of transformations imagers could be led to discover unanticipated objects. For example, take the capital letter "B"and rotate it 90" to the left, place a 'Irldirectly below it, and remove the horizontal line, Many subjects who successfully followed these instructions discovered a heart within their images. Similarly, Hyman (this volume) and Peterson (this volume) demonstrated that a change in subjects' frames of reference led many to reconstrue their images of ambiguous figures. We have shown that the same is true with geographical figures (Reisberg & Chambers, 1991). Further we found that changes in subjects description of their images often led to changes in their imaged depictions (Chambers & Reisberg, 1992). In addition, subjects can create novel objects from a group of distinct parts in imagery (see Anderson & Helstrup, this volume; Finke, this volume). These creations are most likely made by creating new descriptions of the relationship among the parts (i.e., by joining the parts in novel ways, specifying a new frame of reference, and so on). In sum, when imagers alter their descriptions of their images, they will often find unanticipated objects within their imaged depictions. There are several ways of thinking about the relations between the descriptive and depictive aspects of an image. The first possibility is that subjects' understanding of their image is effectively part of the image itself. That is, the depictive and descriptive aspects are represented in a unified representation, such as a structural description. In this way, the depictive and the descriptive aspects are both on the scene, influencing what an image contains, and hence what can be discovered from an image. At the same time, we note that the present data can also be read somewhat differently. Rather than arguing that the construal accompanies (or is a part of) the image, one could claim that the construal of the image directs the creation of the image. Once created, the image could then be represented in some neutral form, perhaps as a pixel pattern within an imagery buffer much like that hypothesized by Kosslyn. This pixel pattern would be influenced by the imager's intentions, such that the pixel pattern created t o be a

Images are Both Depictive and Descriptive

95

duck image would include different material than one created to be a rabbit image. Nonetheless, the image, with that specific pixel patten, would function much as a stimulus - perhaps realized in some buffer in the visual system, processed through the normal channels of vision, and so on. This claim is consistent with the results of Chambers and Reisberg (1985, 1992). However, it is not clear how it fits the orientation and figure-ground results reported by Reisberg and Chambers (1991). In this work, subjects encoded a geographical figure in a novel orientation. Subjects were then asked either to imagine the form rotated, or to change their ttassignmentttof the image's top. They were then asked if their image reminded them of a familiar form. Many subjects in the "reassign" condition, but none in the "rotate" condition, recognized a new form. This is consistent with the unified representation view because the reassign condition should have led t o a revision in the image's description (i.e., the imagers' specification of top), while the rotate condition should not have (subjects can rotate a farm while maintaining the objectcentered specification of top). However, it is not clear how to understand these results in terms of pixel patterns because presumably both the rotated and reassigned conditions should have led to a new pixel pattern. Hence, we believe the full pattern of evidence favors a unified representation. Nonetheless, this is clearly a point on which further data are needed. These results also have important implications for the picture versus proposition debate. This debate described images as either purely pictorial or purely propositional. It is clear from the current results that neither is true. Images are truly a hybrid between depictions and descriptions.

References Arnheim, R. (1969). Visual thinking. Berkeley, CA: University of California Press. Brentano, F. (1973). Psychology from an empirical standpoint (A. Rancuello, B. Terrell, & L. McAlisters, Trans.). London: Routledge & Kegan Paul. (Original work published 1874) Chambers, D., & Reisberg, D. (1985). Can mental images be

96

Deborah Chambers

ambiguous? Journal of Experimental Psychology: Human Perception and Performance, 11, 317-328. Chambers, D.,& Reisberg, D. (1992). What an image depicts depends on what an image means. Cognitive Psychology, 24, 145-174. Dennett, D. (1981).the nature of images and the introspective trap. In N. Block (Ed.), Imagery (pp. 51-61). Cambridge, MA: MIT Press. Finke, R. A. (1979). The functional equivalence of mental images and errors of movement. Cognitive Science, 13, 51-78. Finke, R. A., Pinker, & Farah, M. (1989). Reinterpreting visual patterns in visual imagery. Cognitive Science, 13, 51-78. Finke, R. A., & Slayton (1988). Explorations of creative visual synthesis in mental imagery. Memory & Cognition, 16, 252-257. Fodor, J. (1975). The language of thought. New York Crowell. Fodor, J. (1981). Imagistic representation. In N. Block (Ed.), Imagery (pp. 63-86). Cambridge, MA: MIT Press. Hochberg, J. (1981).On cognition in perception: Perceptual coupling and unconscious inferences. Cognition, 10,127-134. Hochberg, J. (1982). How big is a stimulus. In J. Beck (Ed.), Organization and representation in perception. Hillsdale, NJ: Erlbaum Associates. Hochberg, J., & Peterson, M. (1987). Piecemeal organization and cognitive components in object perception: Perceptually coupled responses to moving objects. Journal of Experimental Psychology: General, 116, 370-380. Hyman, I. E., & Neisser, U. (1991). Reconstruing mental images: Problems of method (Emory Cognition Report #19). Atlanta, GA: Emory University. James, W. (1880/1950). The principles of psychology. New York: Dover. Kosslyn, S. (1980). Image and mind. Cambridge, MA: Harvard University Press. Kosslyn, S . (1981).The medium and the message in mental imagery: A theory. Psychological Review, 88, 46-66. Kosslyn, S. (1983). Ghosts in the mind’s machine: Creating and using images in the brain. New York Norton. Marks, D. (1972). Individual differences in the vividness of visual

Images are Both Depictive and Descriptiue

97

imagery and their effect on function. In P. Sheehan (Ed.), The function and nature of imagery. New York: Academic Press. Peterson, M., Kihlstrom, J., Rose, P., & Glisky, M. (1992). Mental images can be ambiguous: Parts, wholes, and strategies. Memory & Cognition, 20, 107-223. Pinker, S., & Finke, R. (1980). Emergent two-dimensional patterns in images in depth. Journal of Experimental Psychology: Human Perception and Performance, 6, 244-264. F'ylyshyn, 2.W.(1981).The imagery debate: Analogue media versus tacit knowledge. Psychological Review, 88, 16-45. Reisberg, D., & Chambers, I). (1991). Neither pictures nor propositions: What we can learn from a mental image? Canadian Journal of Psychology, 45,336-352. Rock, I. (1973). Orientation and form. New York: Academic Press. Rock, I,, Halper, F., & Clayton, T. (1972). The perception and recognition o f complex figures. Cognitive Psychology, 3, 655-673. Segal, S. J., & Fusella, V. (1970).Influences of imaged pictures and sounds on detection of visual and auditory signals. Journal of Experimental Psychology, 83, 458-464. Shepard, R. N., & Cooper, L. A. (1982). Mental images and their transformations. Cambridge, M A : MIT Press. Slee, J. (1980). Individual differences in imagery ability and the retrieval of visual appearances. Journal of Mental Imagery, 4 , 93-113. Titchener, E. B. (1926). Lectures on the experimental psychology of the thought-processes. New York: Macmillan. Tsal, Y., & Kolbet, L. (1985).Disambiguating ambiguous figures by Quarterly Journal of Experimental selective attention. Psychology, 37,352-373.

This Page Intentionally Left Blank

Imagery, Creativity, and Discovery: A Cognitive I’crspective B. Roskos-Ewoldson. M.J.Intons-Peterson and R.E. Andcrson (Editors) 0 1993 Elsevier Science Publishers B.V. All rights reservcd.

99

Chapter 4

Imagery, Reconstructive Memory, and Discovery Ira E. H y m n , Jr. Department of Psychology Western Washington University USA I began this mental imagery research in response to a set of experiments that I found unbelievable. Chambers and Reisberg (1985) found that people could not discover the alternative interpretation of an ambiguous figure by using mental imagery. In my original response to that work (Hyman& Neisser, 19911,I was interested in outlining some conditions that would facilitate reversals of ambiguous figures using images. I attempted to frame discovery using mental images as a question of the experimental methods employed, rather than as indicative of the underlying relationship between imagery and vision. In subsequent investigation, I have tried to d.iscern how manipulation of visual information using imagery is similar to other memory tasks, such as the reconstruction of a story. Phrasing the issue of imagery-based discovery in the language of reconstructive memory rather than the language of perception and perceptiodimagery equivalences provides an alternative view of the problem. Chambers and Reisberg (1985)asked people to view a figure, such as the duwrabbit figure, and to form a mental image of the drawing. The subjeds then attempted to discover the alternative interpretation from their mental image of the figure. Not one could do so by examining their mental images, but almost every one could when viewing their own drawing of the figure made after searching their images. This demonstrated that the failure to reinterpret the figure was not due to a lack of information. Chambersand Reisbergclaimed that images could not be reconstrued because mental imagery does not involve construal in the

100

Ira E. Hyman, Jr.

first place -- people know what they are visualizing. The only way to discover anything is to imagine a different image. Since the subjects in the Chambers and Reisberg experiments did not know what the other interpretation was, they could not change their mental image. Because I would have expected this to be a reasonably simple task, I was surprised by these results. Apparently, I was not the only person. Chambers and Reisberg (1985) themselves noted that a priori they expected people to be able to do this task. Furthermore, they continued to conduct research in this area and found that under some conditions people could find novel interpretations of mental images (Reisberg& Chambers, 1986,1991). In the midst of a series of meaningless shapes that subjects were to imagine, subjects were shown an outline of the map of Texas rotated 90"counter-clockwise. They did not recognize this form as Texas. When they were asked to rotate their mental image 90" clockwise they still did not recognize the figure. When, however, they were told to make the lef%side the top of the figure and that it was an outline of a familiar geographical form, about half of the subjects discovered Texas. Reisberg and Chambers suggested this was an example of subjects changing their understanding of an image and therefore changing the image itself, by reassigning the top of the form. With the new image, subjects were reminded of something known and this allowed many people to discover the form of Texas, In thisfashion, changing mental images could lead to discoveries through remindings but not through the process of interpretation that is used in vision. Finke, Pinker, and Farah (1989)were also challenged by the original Chambers and Reisberg results. They conducted a series of experiments in which subjects imagined, rotated, combined, and edited simple figures using mental imagery. Their subjects were able to discover new interpretations from their images. Finke and his colleagues argued that images do contain visual idormation that allows interpretation and reinterpretation. They resolved the differences between their results and those of Reisberg and Chambers by suggesting that classical ambiguous figures are unique: To reinterpret an ambiguous figure one must be able to perceive the whole figure at once. Images, however, are composed of dynamically fading and regenerating pieces, and the process of regeneration requires the resources of working memory. Because the whole ambiguous figure must be maintainedfor reconstrual to occur, and because imagery tends to fade and regenerate over time, reinterpreting ambiguous figures is difficult, to say the least. According to Finke et al.,

Imagery, Reconstructive Memory, and Discovery

I01

the difficulty of maintaining a whole figure for reinterpretation is beyond the limits of working memory. Finke et al.’s argument is interesting because it suggests that the differences between imagery and perception found in discovery studies may be explained by similarities between imagery and another cognitive processmemory. Most models of imagery discuss the reliance of visual imagesupon memoryprocesses. For example,Kosslyn,Pinker, Smith, and Shwartz (1979)stated that images are created from stored information and maintained in a visual buffer (orworking memory). Deterioration of stored information in long term memory causes an image to be less clear and detailed than the original percept. In addition, visual working memory holds a limited amount of information and must constantly be refreshed. Thus some constraints on the imagery system are dictated by memory limitations. Additional limitations on the imagery system, and hence limitations on the ability to make discoveries using imagery, may be due to the process of creating an imagewith a reconstructivememory system. Bartlett (1932) is known for his early studies concerning reconstrudive remembering of verbal material. He asked subjects to recall a short story and then looked at the errors people made, including both omissions and intrusions. He interpreted the pattern of errors as evidence for the influence of general knowledge structures on the recreation of a story. Bartlett referred to these general knowledge structures as schemata. Since Bartlett, there has been a great amount of research conducted on the reconstruction of verbal material (e.g., Bransford & Franks,1971;Hyman & Rubin, 1990;Jenkins, 1974;Kintsch &vanDijk, 1978;Mandler &Johnson, 19771,and, relatedly, on autobiographical memories (e.g., Barclay, 1986; Neisser, 1982). According to Bartlett’s reconstructiveview of memory, the recall of a sbry is guided by story details that are available from memory at the time of recall, story gist, knowledge of events like those described in the story, similar personal experiences, cultural understanding of narrative forms, and general world knowledge. honstructive story recall adds information to fill gaps in the story and in memory, and may also change details to fit with general knowledge structures. Although Bartlett is known primarily for his studies of story memory, he also investigated memory for pictures. In having people describe line drawings offaces, he found evidence of schema-basedreconstruction based on an analysis of both omission and intrusion errors. This finding is particularly interesting because of the claim that mental images are

102

Ira E. Hyman, JT.

similar to pictures (e.g., Kosslyn, 1981). Thus, there is reason to expect that the creation and maintenance of a mental image may involve reconstructive memory. Reconstruction of a visual memory as a mental image, like reconstruction of a story, would be based on visual details available at the time of recall, general visual form, knowledge of objects like the one being imagined, knowledge of the visual world, and cultural understanding of pictorial representation. The reconstructive creation and maintenance of an image should focus on the gist. In the case of images, as opposed to stories, the gist may be the general form. The reconstruction should also overwriteinconsistent details and fillin gaps in the remaininginformation. These additions will conform to the general form and to knowledge of the category of objects or events being imagined. The maintenance of the image will continue to emphasize the general form at the expense of details. Thus one can imagine a house without knowing the number of windows or the color of the trim. The reconstructive view of image construction suggests some constraints on discovery using mental images. Namely, discovering something about the gist or consistent details should be fairly straightforward, discovering something about less important details or something that contradicts the gist should be more difficult. This reconstructive view of imagery led me to two types of experiments. The first line of experiments (Experiments 1 and 2) investigated whether classical ambiguous figures can be reinterpreted using mental imagery (see Hyman & Neisser, 1991, for more detail). Others have also addressed this point (see Chambers, Kaufmann & Helstrup, Peterson, and Reisberg & hgie, all this volume). Our experimentswere modeled on the success of the Reisberg and Chambers (1986,1991) Texas experiment but used the duckhabbit and chefldog figures. We reasoned that subjeds’ images may have been limited by the gist of the figure they were imagining, until they were provided with instruction that allowed subjects to fill in details of the figure.The results of these experiments suggest that a view of image creation as a reconstructive memory task may be productive. The other line of experiments is based on ongoing research that more directly investigates the relation between reconstructivememory and mental imagery. I report one experiment (Experiment 31, conducted with Jeremiah Faries, that explored how the gist and details of visual information are remembered. This second area of research outlines some limitations that the

Imagery, Reconstructive Memory, and Discovery

103

reconstructivenature ofrememberingplaces on imagery-guideddiscovery.

Experiment 1 Method

If subjects who imagine one interpretation of the duwrabbit or chefldog figure are bounded, or limited, by the gist of the figure, their image may not include enough detail that is consistent with the alternate construal to be reinterpreted. If, however, subjects are encouraged to fill in detail, they may then be able to discover the other interpretation. Experiment 1was conducted to see if explicit instructions concerning the nature of the alternate construal .makea difference in the reversibility of an imagined figure. The explicit instrudions are assumed to cause the instantiation of additional information, thereby increasing the likelihood of discovering an alternative interpretation. There were two conditions, minimal and full information. Little or no reconshals were expected in the minimal information condition because this condition was similar to Chambers and Reisberg's (19851,and their subjects were not able to reconstrue their images. If'reconstruals are dependent on detail, then our fullinformationinstructions--thosethat encourage instantiationsof detail-should produce reconstruals. Subjects

Forty Emory University undergraduates were tested individually. They were assigned to one of two conditions that varied in the amount of instructions provided to help them reverse their images. Procedures

Subjectswere first familiarizedwith ambiguous figures. They were shown the Necker Cube, the Schroeder staircase, and the vasdfaces figures. Subjects had to indicate that they were able to see both If subjects experienced difficulty seeing both interpretations. interpretations they were provided with clues. Subjects were then told that the next part of the experiment would concern visual memory. They were told that they would be shown a line drawing for five seconds, that

104

Ira E. Hyman, Jr.

they should form a mental picture of the line drawing, and that they would later be asked to draw the figure from memory. Subjects were then shown either the ducWrabbit or the chefldog figure.'. Both figures were used for all subjects and order was counterbalanced. Orientation of the chefldog was also counterbalanced. Subjects were instructed to hold the image in their mind and then were asked if they had noticed, during presentation of the figure, that the figurehad two interpretations like the figures they had seen earlier. Data from subjects who said yes were discarded? Subjectswere then told that the figure did indeed have two interpretations and were asked to find the other interpretation. The information provided to subjects varied according to condition.

Minimal Information (n=20): This condition was modeled on the procedure of Chambers and Reisberg (1985). For the duckhabbit figure, subjects were advised that it might help to shift their focus across the image. For the chevdog, subjectswere advised to rotate.the image90"clockwise or counterclockwiseas appropriate with respect to whether the figure had been shown in the dog or chef orientation originally. Full Information (n=20): These subjects were given both orientation and categorical information about the alternative interpretation, modeled on information that was given to the subjects in the Reisberg and Chambers (1986, 1991) Texas experiment. For the duckhabbit figure, subjects were told to consider the back of the head they already "saw"to be the front of the head of a different animal. For the chevdog figure, they were told to rotate their image so that what they %awttas the

'I do not include illustrations of the duckhabbit figure because the figure is included in other chapters within this chapter. 2

About one quarter of the subjects were discarded because of previous familiarity with the duckhabbit figure. This minor problem arose because the duckhabbit figure was included in the text used in the introductory psychology class. However, the experiment was conducted before the figure was encountered in the course, and we were able to collect data from the students who had not looked ahead in the book.

Imagery, Reconstructive Memory, and Discovery

105

front would be the bottom of either the head of a man or an animal, as appropriate since the original presentation was counterbalanced to be in either the dog or chef orientation. Subjects succeeded by either naming the alternative interpretation or by providing a reasonable description. In keeping with the approach of Chambersand Reisberg(19851, other interpretationswere accepted-such as whale for dog,if the subject described the back end as appearing like a tail. All subjects were asked to draw the figure and those who had not reversed their mental image were asked to find the other interpretation in their drawing. The other ambiguous figure was then presented and the procedure repeated. Results

Very few of the subjects in the minimal information condition reversed their images, while approximately half of those in the full informationconditiondiscovered the alternativeinterpretation(see Table 1). It is important to note that even in the minimalinformation condition a few subjeds (one of 20 for the chtfldog and two of 20 for the ducklrabbit) were able to reverse their mental images. Although the numbers of successfbl discoveries in the minimal information condition are low, they contrast with the complete failure to reconstrue images shown by Chambers and Reisberg's (1985) subjects.

Experiment 2 The second experiment was a replication and expansion of the first experiment. The full information condition in Experiment 1 provided subjects with guidance concerning both orientation and category membership of the alternative interpretation. "he number of conditions was expanded in Experiment 2 to investigateboth orientation and category information separately as well as in combination. It was expected that providing only one piece of information would lead to a number of reversalsintermediate to the minimal and full information conditions. Because several subjects in Experiment 1were dropped due to familiarity with the duckhabbit figure, only the chefldog figure was used in this

Ira E. Hyman, Jr.

106 experiment.

TABLE 1

The number of subjects who reversed the figures using mental imagery

Ehperiment 1 Figure

Chef 1Dog

Reversed Image?

Yes

No

Yes

No

Minimal

1

19 (8)

2

18 (10)

Fd

8

12 (4)

11

9 (6)

DucklRabbit

Condition:

X2= 5.16, pe.02

Figure

Chef l Dog

Reversed Image?

Yes

No

Minimal

2

20 (5)

Orientation

3

19 (7)

ca%wY

2

20 (10

Full

8

14 (3)

X2 = 7.29,pc.01

Condition:

X2= 7.96,p<.05 Note: Parenthetical numbers indicate how many subjects later reversed the figure using their own drawings.

Method Subjects

The subjects were 88 Emory University undergraduates randomly assigned to one of four conditions and tested individually.

Imagery, Reconstructive Memory, and Discovery

107

Procedures The procedure from Experiment 1was generally followed here. The subjeds were first shown the Necker Cube, the Schroeder staircase, the vasdfaces, and the ducklrabbitto familiarize them with ambiguous figures. Subjectswere then told that the next part of the experimentwould concern memory. They were shown the chefdog figure for five seconds and asked to form and hold a mental picture of it. The subjects were asked to continue holding the image in their mind and then they were asked if they had noticed that the drawing had two interpretations. The data from any who said yes were discarded. The subjects were told that the figure did contain two interpretations and were asked to find the second interpretation. There were four instruction conditions. Minimal Information (n=22): As in Experiment 1, subjeds were given little guidance. They were advised to rotate the:ir image 90" clockwise or counterclockwise as appropriate for presentation orientation. For example, if the original orientation showed the dog, subjects were asked to rotate their image 90" clockwise so that the profile of the chef would be upright. Orientation Idormation (n=22):These subjectswere asked to rotate their image so that the "front" they "saw" would be the %ottom."

CategonJ Information (11=22): The 11subjects who had originally seen the figure in the dog orientation were told that the alternative interpretation was a man's head and were asked to describe it more completely. The other subjects were told that the alternative interpretation was an animal and were asked to name or describe the animal. Full Information (n=22): These subjects were given both orientation and categorical information about the alternative interpretation as in Experiment 1.

108

Ira E. Hyman, Jr.

A correct response was one in which a subject either correctly named the alternative interpretation or provided a detailed description of an interpretation. Some subjects noted, for example, that the man was "wearing a funny kind of hat," or that the animal "looked like a fetal pig with its legs stickingdown." Subjectsthen drew the figure and, if they had failed to reverse their image, attempted to reverse their drawing.

Results

As in Experiment 1, very few of the subjects in the minimal information condition were able to reverse their images, while nearly half of those in the f dlinformation condition were able to do so (see Table 1). Contrary to expectations, few subjects in the two intermediate conditions were able to discover an alternative interpretation. It appears that people will most readily discover the alternative interpretation of the chef7dog figure using imagery if they are provided with knowledge concerning how to orient the figure and what type of figure it will resemble. Discussion of Experiments 1 and 2 Based on the results of these fist two experiments, we can dismiss the claim that classicalambiguous figurescannot be reversed usingmental imagery. Under some conditions our subjects did reverse ambiguous figures using imagery. Even under conditions very similar to the original Chambers and Reisberg (1985)experiments we found that five to ten percent of the reversed their imagean important finding in light of the complete lack of reversals in their work. One interesting observationwas that the phenomenological aspectof discoveringthe alternative interpretation is similarwhen usingvision and imagery-subjects experienced an "ahha" type of reaction in both of the training trials, which relied on vision, and the test trials, which relied on imagery. Although records were not kept, many subjects made some expression such as "oh" or "now I see,"just before noting the alternative interpretation. These types of exclamationswere similarto the comments subjects made when viewing the familiarization figures and are perhaps similar to more dramatic expression accompanying discoveries, such as Archimedes shouting "Eurekaa'in his bathtub. Finding conditions in which subjects can regularly discover the

Imagery, Reconstructive Memory, and Discovery

109

alternative interpretation removes the need for Finke et a1.k (1989) theoreticalexplanationofthe qualitative differencebetween attempting to reverse images of classical ambiguous figures and reinterpreting other types of figures. More likely, the differences between Finke et al. (1989) and Chambers and Reisberg (1985)can be explained methodologically. It is not the difference between types of figures, but rather the difference between types of instructions. F'inke et al. gave detailed instructions to their subjects whereas Chambers and Reisberg gave limited instructions. When we provided detailed instructions and used the same figures as Chambers and Reisberg, we found results more in line with the work of Finke and his colleagues. Claiming that the differences in findings are due to methodological differences does not mean there are not interesting theoretical issues in this domain. Reisberg and Logie (this volume), for example, argue about the similarities and differencesbetween imagery and perception and make claims about the content of percepts and images based on this research domain. Peterson (thisvolume) also discusses the links between images and pictures and attempts to characterizethe differencesbetween the two in terms of the reversibility of figures,both imagined and seen. Both claim that such research will inform us about the types of discoveries people can make using mental images and the conditions under which such discoveries should be possible. While not disagreeingwith this perspective, I would like to use these data to consider the linksbetween imagery and a cognitive process other than perception: namely, reconstructive memory. The creation and maintenance of a visual image probably relies on the same memory processes that are involved in the recall and reconstruction of a story or an autobiographical experience. Reconstruction of a story builds on information from the story that remains available in memory, both the gist and the details. Reconstruction also relies on other information, including knowledge of events similar to the events being recalled, knowledge of the people involved, cultural biases in the manner in which stories are told, and other general information. Thus the story or event that is reported is a new constructionthat emphasizes the gist and includes consistentdetails that are available. The construction fills in gaps in the original story and gaps in memory, and may ignore or replace inconsistent information. I argue that mental images are created and maintained using similar reconstructive memory processes. Thus, in a fashion similar to the reconstruction of a story, a great variely of knowledge would go into the

110

Ira E. Hyman, Jr.

creation of a mental image: knowledge of the object being imagined, of details fi-om the particular object, of other items that look like it, of other objects categorically similar to it, and of how society typically displays visual information. The image should emphasize the gist of the object or event to be imagined. The gist is most likely the general form of the objecvevent. In addition, the reconstruction of an image, like that of a story, will fill gaps in details and either overwrite or ignore less important or inconsistent details. The result of the reconstructive process is that in the image, the general form, or gist, should be fairly clear and the details more vague. Discovering something about the gist, and to a lesser extent consistent details, shouldbe relatively easy because the image emphasizes the gist. Discovering information about other details should be more difficult because they may not be consistent with the gist of the image. Discovering something that contradicts the gist will be most difficult because it will require reworking the image. In the first two experiments, the gist of the duck was not enough to discover the rabbit. Drawing the figure helped because it forced the instantiation of details left vague in the image. Instructions helped because they constrained the possible ways of looking at the remaining visual information while suggesting alternative ways of organizing the details. The lack of facilitation when subjects were provided with either orientation or category information alone suggests that for people to successfully discover the other interpretation, they need to know how to restructure the available details and how to apply general knowledge structures to the image.

Experiment 3 Although the reconstructive language used in the discussion above may not differ theoretically from other descriptions, such as those looking at perceptiodimagerycorrespondence,it may suggest new ways of looking at the problem of discovery using mental images. With Jeremiah Faries of Northwestern University, I have been engaged in some new research that addresses the relation between reconstructive memory and imagery more directly than did the first two experiments. In the current experiment,we explored subjects’ ability to keep details in mind when they concentrated on the whole, a problem similar to the one with the ambiguous figures, where subjectsheld an image of the overall figure (gist)while trying to use

Imagery, Reconstructive Memory, and Discovery

111

the details of the imaged figure to discover an alternative interpretation. In our experiment we asked subjects to construct a coherent whole (i.e., gist)from pieces (i.e.,details) and then looked at their ability to recall both the completed figure (gist)and the constituent pieces (details). We chose recall as a measure, rather than frequency of reversibility or discovery, to probe the influence of reconstructive remembering on the contents of mental images. Nonetheless, we address both imagery and discovery because the pieces are presented so that the subjects need to use imagery to discover,through combination,the overall shape, or gist, of the figure. The final forms of the figures used were geometric shapes, such as squares, triangles, and rhombuses. The pieces presented were varied t o explore the effeds of complexity on the ability to discover the figure and on recall for the whole, or gist, and the pieces, or details. There were three types of pieces: canonical (e.g., square); simple (not easily labeled, but containing simple, straight inner edges); and complex (not easily labeled and containing irregular inner edges). We also varied whether the pieces were present or absent during drawing to investigate the effects of remembering on the recall of visual information.

Method Subjects Twenty-seven Northwestern IJniversityundergraduates participated in thisexperiment. The research was conducted in groups of four to seven subjects. Materials Thirty-six puzzles with three pieces each were created. All puzzles could be combined to create the well-known geometric shapes of square, rectangle, isosceles triangle, right, angle triangle, square rhombus, or rhombus. The pieces themselves were of three different types: canonical, at least two of the three pieces were themselves well-known geometric shapes (square, rectangle, etc.); simple, all pieces had only a few, very straight internal edges, but could not be easily labeled; or complex, the three pieces were created with complicated internal edges. Figure 1shows examples of the types of pieces and the completed figures for each set.

112

Ira E. Hyman, Jr.

Figure 1also shows another aspect of the puzzle pieces-the number of pieces that were presented in the same orientation with respect to one another was varied from all three in the same orientation to all three in different orientations.

Procedure Subjectswere presented with a series of three-piece puzzles. Each set of three pieces was shown for five seconds. Subjects were asked to memorize the pieces and discover the completed pattern. The subjects were then asked to draw the completed figure, with the internal lines representingthe distinct pieces. On half the trials this task was a memory task (i.e., the pieces were removed from the subjects’view a h r the initial five seconds) and on half it was not (i.e., the pieces were left in the visual field while the subjects completed their drawings). For each drawing subjects were asked whether or not they had discovered the completed figure during the viewing period. Design

Two within-subject manipulations are of interest with respect to reconstruction of visual information. The first is the type of pieces: canonical, simple, or complex. The second is whether the drawing of the completed figure was a memory task or not; on half of the trials the pieces were withdrawn after five seconds; for the other half the pieces remained in view during the drawing phase. The orientation of the pieces with respect to one another was also systematically varied, but this is of less interest to the issue of reconstructive memory for visual figures. Results This was primarily a study of memory for visual information. Discovery using imagery was involved for if subjects discovered the correct figure, they generally did so while the pieces were visually present by imaginally manipulating the pieces. We were concerned with two aspects of the subjects’ drawings: whether the completed figure was correct and whether the individual pieces were correct. Thus, each subject was given two scores in each of the 2 x 3 (memory condition x piece type) conditions. One score was the number of

Imagery, Reconstructive Memory, and Discovery

113

Figure 1 Sample puzzle pieces and completed patterns: a) canonical pieces; b) simple pieces; c) complex pieces.

114

Ira E. Hyman, Jr.

drawings with correct figures while the other was the number with correct pieces. Because the subjects were presented with six sets of pieces in each treatment condition, the scores ranged from zero to Six.

For the number of drawings with correct figures, a 2 x 3 analysis of variance indicated a main effect of memory condition, F(1,26)=18.89, p < . O l , a main effect of piece type, F(2,52)=10.27, p<.Ol, and no interaction. It was easier for subjects to draw the whole figure correctly when the pieces remained in the visual field than when they were forced to rely on their memory. In addition, as the pieces became more complex it was easier t o form the whole figure correctly. Table 2a shows the mean number of correct whole figures in each condition. For the number of drawings with correct pieces, a 2 x 3 analysis of variance indicated a main effect of memory condition, F(1,26)=230.13, p<.Ol, a main effect of piece type, F(2,52)=75.11, p<.Ol, and a significant interaction, F(2,52)=5.11,pc.01. When subjects were forced to rely on memory they drew fewer of the pieces correctly than when the pieces remained in vision (see Table 2b). Subjects drew canonical pieces more accurately than simple pieces and simple pieces more accurately than complex pieces. The interaction is explained by the mean score in the memory, complex pieces condition, which was nearly zero. Table 3 presents a combination of the two scores. For each condition, it shows the mean number of drawings with both the figure and pieces correct, the figure correct and the pieces incorrect, the figure incorrect and the pieces correct, and both the figure and pieces incorrect. As the pieces became more complex it became easier to discover the whole figure but more difficult to recall the pieces. Two additional observations about the drawings are, first, that subjects generally drew the figure before the internal lines that described the individual pieces. It is possible to draw each piece individually so that they fit together to form the whole figure, but we saw very few examples of this approach. Second, errors made when drawing internal lines from memory indicated that the details maybe known in only a vague way. When drawing internal lines for complex pieces, for example, many subjects drew squiggly lines inside the figure in approximately the shape of the pieces.

Imagery, Reconstructive Memory, and Discovery

115

TABLE 2 Number of drawings of correct figures and of correct pieces, Experiment 3

(a) The number of drawings with correct figures Piece Type Memory condition

Canonical

Simple

Complex

Visual

3.22

3.85

4.22

Memory

2.66

3.18

3.52

(b) The number of drawings with correct pieces

Memory Condition

Canonical

Simple

Complex

Visual

4.74

4.33

3.07

Memory

3.02

1.70

0.04

Note: Scores can range from 0 to 6.

Discussion The results of Experiment 3 are informative about the general process of discovery with visual information. Although discovery of the completed figure occurred while the pieces were visually present, subjects relied on imagery to rotate, move, and combine the pieces. In addition, more complex pieces were combined into the correct whole figure more readily. This is because complex pieces constrained the manner of their combination: the jagged edges had to be internal and the number of internal corners facilitated finding similar edges. In addition, canonical pieces may have been more difficult to combine than the simple pieces because they were known pieces that resisted reinterpretation. Discovery when information is not visually present is probably similar. The more the information constrains thinking toward a solution the more likely that solution will be discovered, just as complex pieces containing similar edges

Ira E. Hyman, Jr.

116

Drawings scored for correctness of figure and pieces by condition

Condition

Combined Drawing Rating Figure Correct Figure Incorrect Pieces Pieces Pieces Pieces Correct Incorrect Correct Incorrect

Canonical Visual

3.00

0.22

1.74

1.04

Memory

1.59

1.07

1.33

2.00

Visual

3.07

0.78

1.26

0.89

Memory

1.37

1.81

0.33

2.48

Visual

2.37

1.85

0.70

1.07

Memory

0.00

3.52

3.52

2.44

Simple

Complex

Note: Scores can range from 0 to 6.

constrain the possible figures that the pieces could form. Also, the more that information restrains thinking t o the current state, the less likely any solution will be discovered, as when canonical pieces are seen as complete, independent figures. This view is relevant to Roskos-Ewoldsen's (this volume) work concerning the properties of pieces that make them "good" for combining into wholes. With respect to memory for visual information, the results of this experiment suggests that reconstructive memory approaches, such as Bartlett's (19321,may usefully be applied. Subjects were able to recall and draw the whole figure across all piece type conditions, although they performed better with complex pieces. Subjects were not, however, able t o recall the pieces in all conditions. As pieces became more complex, subjects were less able to draw them from memory. Subjects recalled pieces accurately when the pieces were coherent or canonical. They performed less well when the pieces had no internal coherence but remained relatively simple. They almost never recalled pieces correctly when they were complex,

Imagery, Reconstructive Memory, and Discovery

117

in spite of the fact that they could correctly draw the whole figure better than in the other conditions, and could draw such pieces when the pieces were visually present. In essence, subjects were able to recall the gist-in this case the whole figure and the general shape of the pieces and were able to reconstruct details when the pieces made sense. When the pieces were complex, the errors showed that subjects were reconstructing based on the general shape.

Conclusion The first two experiments reported in this chapter demonstrated that mental images of classical ambiguous figures can be reversed, especially when subjects are given appropriate instructions. Thus the difference between Chambers and Reisberg's (1985)finding that classical ambiguous figures could not be reversed and Finke et a1.k (1989)finding that discoveries could be made with simple figures need not be explained as qualitative differences in the figures (Finke et al., 1989). Instead, the explanation is most likely due to the differences in instructions provided in the two experiments, Chambers and Reisberg provided minimal instructions whereas Finke et al. provided explicit instructions. When explicit instructions concerning how to manipulate an ambiguous figure were provided, subjects were able to discover an alternative interpretation. Further, it is likely that imaging relies on the visual system for its processing much as perceiving does. Therefore, a variety of discoveries about visual aspects of information should be possible with mental images. Looking at an image, however, is not the same as looking at a picture. Thus, not all tasks that can be accomplished when looking at a visual display will be completed with the same ease when one relies on mental images. A primary distinction between perception and imagery is the source of the information for the visual system: in perception that source is the world, while in imagery it is memory. This difference is important because relying on memory introduces schema-based distortions-both omissions and intrusions because recalling information from memory is often a reconstructive process. The types of errors will limit the ability to make discoveries with imagery. Experiment 3 investigated whether creating and maintaining

118

Ira E. Hyman, Jr.

an image is a reconstructive memory task in the same way that Bartlett (1932) viewed story recall as reconstructive. Memory for gist, which is considered to be the general form of the figure, was excellent in all conditions. Recall of details unrelated to the gist, in this case the specifics of the internal lines based on the original pieces, was affected by the complexity of the task with more complex pieces remembered less well than canonical or simple pieces. The errors made by subjects suggests a reconstructive memory process at work, and it seems reasonable to argue that an image is a product of reconstructive memory. This view of image creation differs from that of Kosslyn and his colleagues (Kosslyn, 1980;Kosslyn et al., 1979). In their view, image creation involves calling up a single store representation of the to-be-imagined object and a straightforward filling in of the visual buffer based on that information. Applying Bartlett’s (1932) approach to images, the creation of an image is a construction based on the general form of the figure (which is similar to the gist of a story), the remaining details of the figure, knowledge of similar objects (i.e., a visual schema), knowledge of the category of the object, and other general knowledge such as cultural norms for pictorial The remembered general form should be representations. emphasized in the image and will be a compromise between the information about the figure and general knowledge structures. Consistent details, such as the external lines and corners in Experiment 3, will be imaged well, t o the extent of the system t o maintain details in an image. Irrelevant details, such as the internal lines in Experimental 3, will be less explicit or ignored. Gaps in the original information or forgotten information will most likely be overwritten by a combination of the gist and general knowledge structures. Inconsistent details such as the bumps on the back of the rabbit’s head in Experiment 1(see also Chambers, this volume) may be the most difficult to maintain and may therefore be replaced by the gist and general knowledge structures. When someone draws a picture based on an image, the gist should again be emphasized. Consistent details should occur more frequently and accurately in the drawing because the constraints of the imagery system in maintaining details are not present to limit performance. Irrelevant and inconsistent details may not appear in drawings at all because they are not supported by the general frame of the reconstruction

Imagery, Reconstructive Memory, and Discovery

119

process. Discovery using images will depend on the relationship of the task to reconstructive memory. If the task emphasizes or is congruent with the gist or general form of the figure then it should be performed easily. If the task is based on consistent details then it should also be performed well as long as the limits of working memory are not exceeded. If the task is inconsistent with the gist, or is based on irrelevant or inconsistent details, it will be much more difficult to perform. This seems to be the case in attempting t o discover the alternative interpretation of ambiguous figures using mental images. Such tasks may require additional instructions about how to regroup details, which unspecified details to instantiate, and which general schemas to reference. Individual differences may also play a role in the more difficult tasks in our case, those with [complex pieces. People with high mental imagery ability may be better able to manipulate their images or better able to visualize more extraneous details than people with low imagery ability (see Kaufmann & Helstrup, this volume). Thus, they may make more discoveries with images and engage in image-based thought more frequently. People more familiar with the domain in question may also be able to make more discoveries. Their more complete schemas should guide explicit image construction more easily than those people of less familiar with the domain who would possess less complete schemata. This last issue of individual differences and expertise probably deserves special mention with respect to scientific discoveries made with mental images. Experts in any field have, by definition, excellent knowledge of the domain. This knowledge may enable them to form more complete images less effortfully, to manipulate the images easily, and t o have a better understanding of the constraints on manipulations. Thus creative insights may occur during thinking, playing, or dreaming since their schemas may allow images that are more complete. Another, related form of discovery involves the application of schemata from one domain t o a second domain. In this case, a scientist may visualize a problem in the form of pictorial representation or image that has previously been used for a different purpose. One famous example of this phenomenon is of a chemist visualizing a benzine ring as a snake biting its tail.

Ira E. Hyman, Jr.

120

In conclusion, I have attempted to frame my imagery research in terms of similarities between imagery and memory rather than between imagery and perception. This change in focus suggests different styles of experiment, of which Experiment 3 is a first step, and should provide a valuable addition to theories of mental imagery. This view is not suggested as an alternative theoretical approach, but rather as a means of expressing some limits on the creation of images that in turn will limit the amount and type of discoveries people can make.

References Barclay, C. R. (1986). Schematization of autobiographical memory. In D. C. Rubin (Ed.),AutobiographicaZmemory (pp. 82-99).New York Cambridge University Press. Bartlett, F. C. (1932). Remembering: A study in experimental and social psychology. New York: Cambridge University Press. Bransford, J. D., & Franks, J. J. (1971).The abstraction of linguistic ideas. Cognitive Psychology, 2, 331-350. Chambers, D., & Reisberg, D. (1985). Can mental images be ambiguous? Journal of Experimental Psychology: Human perception and Performance, 11, 317-328. Finke, R. A., Pinker, S., & Farah, M. J,(1989).Reinterpretingvisual patterns in mental imagery. Cognitive Science, 13, 51-78. Hyman, I.E., Jr., & Neisser, U. (1991). Reconstructing mental images: Problems of method. Emory Cognition Project Report, 19, Emory University, Department of Psychology. Hyman, E . E., Jr., & Rubin, D. C. (1990). Memorabeatlia: A naturalistic study of long-term memory. Memory & Cognition, 18, 205-214.

Jenkins, J. J. (1974). Remember that old theory of memory? Well, forget it! American Psychologist, 29 785-795. Kintwh, W., & van Dijk, T. A. (1978). Toward a model of text comprehension and production. Psychological Review, 85, 363394. Kosslyn, S. M. (1980).Image and mind. Cambridge, MA: Harvard University Press. Kosslyn, S. M. (1981). The medium and the message in mental

Imagery, Reconstructive Memory, and Discovery

121

imagery: A theory. Psychological Review, 88, 46-66. Kosslyn, S. M., Pinker, S., Smith, G. E., & Shwartz, S. P. (1979). On the demystification of mental imagery. The Behavioral and Brain Sciences, 2, 535-581. Mandler, J. M., & Johnson, N. S. (1977). Remembrance of things parsed: Story structure and recall. Cognitive Psychology, 9, 111-151. Neisser, U. (1982). John Dean’s memory: A case study. In U. Neisser (Ed.), Memory observed: Remembering in natural contexts (pp. 139-159).San Francisco: Freeman. Reisberg, E., & Chambers, D. (1986). Neither picture nor propositions: The intensionality of mental imagery. Proceedings of the Eighth Annual Conference of the Cognitive Science Society (pp. 208-222).Hillsdale, NJ: Erlbaum. Reisberg, D., & Chambers, D. (1991). Neither pictures nor propositions: What can we learn from a mental image? Canadian Journal of Psychology, 45, 336-352.

Acknowledgment

I would like to thank Angela Hanvood for her helpful comments concerning this manuscript. Also, I would like to thank Ulric Neisser for his help with Experiments 1and 2,and Jeremiah Faries for his aid with Experiment 3.

This Page Intentionally Left Blank

Imagery, Creativity, and Discovery: A Cognitive Pcrspective 13. Koskos-Ewoldson. M.J.Intons-Peterson and R.E. Anderson (Editors) 0 1993 Elsevier Science Publishers B.V. All rights rescrved.

123

Chapter 5

MENTALIMAGERY: FIXEDOR MULTIPLEMEANINGS? NATURE AND FUNCTION OF IMAGERY IN CREATIVE THINKING Geir Kaufmann and Tore Helstrup Department of Cognitive Psychology University of Bergen Sydneshaugen 2 5007 Bergen, Norway Highly creative people often report a preference for using imagery in their creative endeavors (Ghiselin, 1952; Hadamard, 1945; Miller, 1984; Shepard, 1978; cf. Intons-Peterson, this volume). More specifically, the history of scientific discovery contains many reports of new discoveries suggested by images (e.g., Miller, 1984). In line with the views expressed in these subjective reports, several contemporary cognitive theories of symbolic processing assign a special functional role in the less programmed aspect of problem solving. Degree of programming is held to be a major dimension for classifylng problems (e.g., Simon, 1977). A problem is programmed t o the extent that the problem-solver has a definite procedure to handle it. Problems are non-programmed when they are novel for the subject, unstructured or unusually complex (cf. Kaufmann, 1984a, 1989 for a more detailed discussion of the basic constituents of degree of programming in problems). In standard definitions of creativity, the property of task novelty is regarded as a prominent feature of the problem spaces characteristically dealt with in creative activities (cf. Kaufmann, in press; Raaheim, 1974, 1984; Sternberg, 1988).

Geir Kaufmann and Tore Helstrup

124

Morris and Hampson (1983,1990) have developed a model based on the thesis that the functions of consciousness are to monitor and control processing that does not proceed automatically. They argue that because imagery is a form of conscious representation, it should be particularly useful under novel task conditions. A potential limitation of their model is that it does not distinguish between the role of imagery and other forms of conscious representation, like verbalization, with regard to task novelty. Kaufmann (1980, 1985, 1988)put forward a theory of symbolic representation in problem solving. In this theory, the symbolic strategies of verbalization and visualization are also linked to low programmed tasks via the concept of the general functions of consciousness in cognitive processes. The theory also prescribes a division of labor between linguistic and imagery representations within the task novelty dimension. Specifically, it is argued that visual imagery is linked to the highnovelty end of the continuum through its capacity to access a set of simpler processes of a perceptual kind. These simpler processes are postulated to be needed in low programmed tasks, where computational processes in the form of rule-governed processes are difficult or impossible to perform. The theory is outlined in Table 1. The basic assumptions of the theory are founded on the empirically well-established problem solving principle that when a TABLE 1 Kaufinann’stheory of symbolic representations in problem solving.

coNBcIous

VERBAL

IrUAGrnAL

Computational Transformations (Rule-Governed Inferences)

Perceptual Simulations (Mental modeling)

REPRESENTATIONS Mode of Operation

Main information processing categories

UNDERLYING REPRESENTATIONS

Deductive reasoning

Inductive reasoning

PROPOSITIONAL

Perceptual comparisons

Perceptual anticipations

ANALOG

Mental Imagery: Fixed or Multiple Meanings?

125

task is low in programming, the individual has to resort to so-called weak (i.e., heuristic) methods (Newell, 1969; Simon, 1978). Weak methods are general, pragmatic strategies that may be applied to a wide range of problems. They are weak in the sense that they lack precision, and do not guarantee success. In contrast, strong (i.e., algorithmic) methods are precise, tailor-made to the situation, and grant a safe and fast solution of the task. It is argued that this principle may be transferred to the domain of representational methods. The basic thesis of the theory is as follows: A linguisticpropositional representational format is a strong one in the sense that great precision may be achieved in the form of explicit descriptions. It is easily and quickly manipulated, and it allows, in principle, the full range of computational operations to be actualized. In contrast, imagery is more ambiguous, sluggish and less easily manipulated. Besides, imagery mainly actualizes simpler cognitive operations of a perceptual kind, like anticipations and comparisons. Perceptual operations may, however, be useful and even necessary in low programmed task-environments, where computational operations are difficult or impossible to perform. Limitations on the use of computational operations may be due t o lack of rule-based Limitations of information on which to operate (novelty). computational operations may also result from strain on working memory due to a high information load (complexity). Finally, uncertainty as to which rule or procedure should be used may lead to computational dysfunctions (ambiguity). Images may be best described as perceptual-like mental models (cf. Sanford, 1985) that allow this perspective. Such perceptual mental models allow a transformation from computational to perceptual operations. More specifically, we have suggested that deductive operations may be converted into simple, quasi-perceptual comparisons, where certainty of judgment may be reached. The imagery parallel to inductive operations may be found in quasi-perceptual anticipations, where a future state of affairs may be imagined on the basis of a previous sequence of events (cf. Kaufxnann, 1990 for a more detailed discussion). In agreement with the theory, extensive reviews of the relevant experimental literature suggest strongly that imagery gains increasing importance in direct proportion to the degree of illstructuredness of the task. With highly novel, complex or ambiguous

126

Geir Kaufmann and Tore Helstrup

task environments, the subject tends to switch from a linguisticpropositional to an imagery-based strategy (Kaufmann, 1984b, 1990). The basic premise of the theory rests on the assumption that linguistic-propositionaland imagery-based representations are clearly distinguishable in terms of the kinds of operations that may be most readily executed through these various forms of symbolic modes of representation. Specifically, it is held that the sensory-perceptual properties of imagery have functional value. The level of processing of the task, among other conditions, may determine the functional usefulness of the perceptual properties of imagery. By implication, a major role is assigned to the processing required in the typically low programmed task environments characteristically dealt with in creative thinking. Imagery research has thus far mainly progressed along separate lines. We think it is important to make this distinction explicit and to spell out in what ways they are separate, and how they may intertwine toward a more unified and coherent understanding of the place of imagery in cognitive functions.

Nature of Imagery The central issues in this research category are intrinsic in the sense of turning inward to the structure and mechanics of imagery as such. The aim here is to unravel the composition of mental imagery and to elucidate the processes involved in imaging and how the processes interact. In this domain, we find classification schemes for different forms of imagery experiences (e.g., Ahsen, 1984; Horowitz, 1983; McKellar, 1957; Richardson, 1983). The work of Kosslyn (1980, 1983, 19871, aimed at clarifying the basic process components and the general mechanics of the imaging process, and Finke’s probing of the levels of equivalence between imagery and perception (Finke,1989), also belong to the conceptual category of research where the focus is on the nature and properties of imagery (cf. Anderson & Helstrup, this volume). Function of Imagery

In contrast, the questions belonging to this research category are extrinsic by being directed at the issues of how imagery processing may influence cognitive performances external to imagery.

Mental Imagery: Fixed or Multiple Meanings?

127

The goal is here to identify the functional role of imagery in such processes as learning, memory, thinking and problem solving by way of identifying the conditions where imagery is used, and is useful, in the processing underlying these functions. The pioneering research of Paivio (1971) clearly belongs to this category, and so does the previously mentioned work of Kaufmann (1979, 1980, 1985, 1988, 1990) and Morris and Hampson (1983, 1990) on the functional role of imagery in thinking and probllem solving. Considerable attention also has been given to the functions of imagery in affective processes (e.g., Horowitz, 1983; Sheikh, 1983; Singer & Pope, 1978). This category also includes attempts to understand how imagery can be used as a problem solving strategy in a wide variety of cognitive tasks (e.g., Helstrup, 1988a; Helstrup & Anderson, 1991). It seems well advised to observe the conceptual differences between these different fields of inquiry and to grant the legitimacy of being primarily concerned with one set of issues in imagery research rather than another. Signs of misplaced criticism and potentially fruitless debate as a result of failing to observe this distinction are apparent in the literature. This is seen when Kosslyn (1983) accuses Paivio’s theory of process inadequacy for not being able to deal with the question of the generation, transformation and inspection of images. But this is not the issue primarily addressed by Paivio’s theory. On the other hand, Kosslyn’s theory has come under attack for not saying much about the functional role of imagery in memory, learning and problem solving performances (e.g., Kolers & Smythe, 1979). Again it is important to realize that this issue is not the primary target of Kosslyn’s theory. To avoid confusion and futile straw man debates, it should be recognized that one field of imagery research concerns the nature and properties of the imagery phenomenon itself. Another is driven by its interest in imagery as a tool aiding processing in other psychological functions (cf. Anderson & Helstrup, this volume). This distinction does not imply that an iron wall divides the two categories of research. To the contrary, clarification of the nature and structure of imagery may :provide a basis for a finer grained analysis of the mediating functions of imagery. On the other hand, knowledge about the conditiolns where imagery has external functional value may provide important insights on the intrinsic nature and properties of imagery and the imaging process. In a more

128

Geir Kaufmann and Tore Helstrup

advanced scientific conquest of the imagery territory, we may envisage a unified theory that speaks both to issues of the nature of imagery and to issues of its external role in psychological functions. Publication of this book may signal that the time is ripe for making the move toward greater integration in the realm of imagery research. Chambers and Reisberg (1985, Chambers, this volume, Hyman, this volume, Peterson, this volume, Reisberg & Logie, this volume) have addressed a core question about the nature of imagery by focusing on the bounds of the functional equivalence between imaging and perceiving. Chambers and Reisberg (1985)favored a radical distinction between imaging and perceiving because, contrary to the case of pictures, images are inherently unambiguous and cannot be reconstrued. The general implication is taken to be that the essence of an image is the intentional description under which it is constructed. Thus, the sensory-perceptual aspect of imagery is to be regarded as epiphenomenal. The acid test of whether image reconstrual is possible or not is said to depend on the empirical answer to the question of the possibility of discovering an unanticipated, uncued configuration in an image. Yet, the answer to this question also speaks to the issue of the functional value of imagery processes. All of the major functional theories of imagery assume that imagery has a special role in cognitive processing due to genuinely functional perceptual properties of images. Thus, we see that the question of the possibility of genuine, uncued reconstrual in imagery pertains both to the issue of the nature of imagery and to the question of the functional role of imagery in cognition. If Chambers and Reisberg (1985) are correct in claiming that images are inherently unambiguous and cannot be reconstrued, a descriptionalist view of the nature of imagery is supported. Then one important premise for the alleged privileged role of imagery as a mode of representation in creative thinking processes has been undermined because it relies on the assumption that perceptual processes may be activated through imagery. In the sequel, we will first highlight the question of the nature of imagery and consider logical and empirically based arguments linked to the issue of the level of equivalence between imaging and perceiving. Then we will try to spell out the implications for the question of the functional role of imagery in the processes of

Mental Imagery: Fixed or Multiple Meanings?

129

creativity and discovery.

The Descriptionalist Theory The basic assumption beh.ind theories in this category is that the content of an image may be completely assimilated to the intentional description it is constructed under. No surplus meaning in terms of its sensory-perceptual content is allowed for.

The Logical Grammar of Imaging

It is interesting that the mental image has had a different fate in recent philosophical discussions than in the more empirically based psychological treatment. Although the pro-imagery stance seems to dominate in psychology (e.g., Anderson, 1990), an antiimagery view is expressed by philosophers. For example, Dennett's claim that the disposal of mental images would be ''a clear case ofgood riddance" (Dennett, 1969, p. 141), seems the more Contemporary representative of the philosophical position. philosophers try to establish the nature of mental concepts by charting the logical geography of these concepts (cf. Ryle, 1949). This is normally done through the Wittgensteinian procedure of probing the logical grammar of the concept in question (e.g., Wittgenstein, 1953). In the case of imagery, the aim of the analysis is to find the most appropriate conceptual analogies that can be used to express the facts about imaging. Ryle (1949)claims that expressions like ''seeing a mental image" have no valid use. Thus, the common sense view of imagery reflects a distorted view of the nature of the mental image. According to Ryle, an image is not an object. It lacks size, shape, temperature, cannot be given a specific location in space, and is not found existing anywhere. Since the concept of seeing logically requires an object to be seen, it follows that there is nothing to be llseenll in images. So imaging cannot be treated as a subspecies of perception. Armed with such arguments, Ryle (1949)goes on to suggest a dispositional analysis of imaging. Here imagery is conceived as a form of expecting to see something, except that the expectations involved in imaging are not fulfilled. Rather, it is like a rehearsal of them being

130

Geir Kaufmann and Tore Helstrup

fulfilled (cf. Ryle, 1949). This conception of imagery has been further developed in psychology by Neisser (1976,1978; cf. Kaufmann, 1980, 1986 for a more detailed treatment and critique). Still, there is definitely some sort of visual experience involved in imagery. This aspect of imaging is hardly properly accounted for in a purely dispositional analysis along Rylean lines. Shorter's (1952) influential Wittgensteinian analysis of the nature of imagery further pushes negative claims about imagery. According to Shorter, mental images are not seen, they are not objects, and therefore they do not exist (in the technical sense of not being independent objects). These counterintuitive claims are held to be forced upon us both by logical and empirical considerations. Expressions like "seeing a mental picture" have no logical use. Besides, it is held to be a matter of empirical fact that there is nothing we can literally see in imagery, not even in a lax and metaphorical sense of seeing. A basic premise of these conclusions is that there is a radical distinction between imaging and perceiving. The basis of this distinction is: whereas imaging is a doing, perceiving is a getting. This is because perceiving is a relational expression, in the sense that perception relates to an interpretively neutral object, whereas imaging relates to an intentionally constructed description that forms its essential content (see also Chambers & Reisberg, 1985;Dennett, 1982;Heil, 1982). The reality of the image is the thought of what the image is an image of. The conclusion follows that we cannot have an image of x without knowing that the image is of x, whereas we can perceive an object without knowing what the object is. Thus, "having an image" could not possibly be placed in the same logical category as "perceiving a picture." One important consequence of such arguments is that an image is totally fmed under its intentional description and cannot be ambiguous. Are the experiments designed by psychologists to examine the issue of image detection and image reconstrual to be regarded as a philosophically naive ploy with a foregone logical conclusion? This implication clearly seems not to be valid. The points that imaging is more of a doing than perceiving, and that images, at least in the paradigmatic case, are constructed under intentional descriptions both are well taken. These arguments do not, however, warrant the conclusion that the mental image is entirely determined by its

Mental Imagery: Fixed or Multiple Meanings?

131

descriptive content. On the contrary, the argument to this effect seems a clear case of begging the question, in the sense that it assumes as premise what is to be shown in the conclusion. In fact, the conclusion that the image is strictly commensurate with the thought of what the image is, seems to strip the mental image of its conceptual autonomy. Besides, by assimilating imagery strictly to thought, any special functional value of imagery as a self-contained form of cognition seems denied. Yet, imagery clearly seems to have some identity of its own. In imaging there is an added element of a quasi-perceptual nature beyond our control. Thus images may be partly like sensations in that they have duration and intensity besides intentional content. When we decide to summon up an image, say of a man, the resulting image may be of a man with a black coat, a red scarf and an oldfashioned hat. He may even have an umbrella! Who knows? Ordinary experience thus seems to suggest that an image may, and normally does, contain features that are not "mentioned" in the intentional description that the image was constructed under. In fact, this is what makes the image an image rather than just a ''pure'' thought. This feature of the imagery experience may be the legitimate basis for the expression "seeing1'in the case of images. Because some features of the image may come "as a surprise," and seems 'hone of our doing," imaging also may partly be a "getting" as in perception, and not exclusively a "doing." The strict line drawn between "imagery" and "perception"as separate conceptual categories of "doing" and "getting" also may be too simple. In cognitive theories of perception (e.g., Neisser, 1976),the active, constructive nature of perception is, indeed, emphasized. A theory of imagery has to do justice to the full range of the phenomenon it purports to explain. Since the getting (discovery) aspect of imagery seems not to be properly accounted for in the main stream philosophical account described above, it is important to explore this dimension of imagery. It needs to be determined what the discovery aspect implies for the nature and functional role of imagery. The depiction analogy of imagery metaphorically treats imaging as the equivalent of the process of drawing a picture. This analogy, championed by Shorter (1952)and Dennett (1969,1982)in philosophy, and Chambers and Reisberg (1985)in psychology, is clearly an improvement over a purely dispositional analysis as

132

Geir Kaufrnann and Tore Helstrup

advocated by Ryle (1949) and Neisser (1976, 19781, where maor parts of the imagery phenomenon seem to be categorized out of existence. Yet it may not be adequate. From the examples given above, and from the existence of phenomena like "unbidden images," hypnagogic imagery, and probably also eidetic imagery (cf. Marks & McKellar, 19821, it seems that the term depictive is not sufficiently descriptive! In these cases it seems more appropriate, as suggested by Flew (1956) and Hannay (19711, to use the concept of "seeing." The present analysis suggests that there is an important grain of truth in the old picture theory. The old version of the picture theory is probably not correct, nor is it completely wrong, as most Contrary to main stream philosophers seem to contend. contemporary philosophical claims and the dogmas of artificial intelligence-orientedcognitive psychology, we may have to admit that we do store some pictorial information in a rather detailed and pictorial way. There is no a priori, logical reason why this could not be so. Human cognition may not operate on the allegedly rational principle that we store only descriptive representations to which we attend. Neither do we have to go to the other extreme and claim that every minute detail of pictorial information gets stored. What is stored may be information that is potentially describable. Potentially describable information may include both visual and motoric information (cf. Engelkamp, 1990; Helstrup, 1988b). It is likely that the pictorial information that naturally comes along with the descriptive information may form the raw material that constructive imagination, and, by implication, creativity and discovery feed on. As psychologists we can return the criticism to the philosophers and claim that questions about the validity of expressions like "seeing an image" are, in the final analysis, empirical questions, to be answered by test and experiment.

The Pictorialist Theory Seeing with the Mind's Eye Many of the imagery models developed in psychology as frameworks for experimental studies of the nature of imagery seem to have been inspired by earlier philosophical theories of imagery in

Mental Imagery: Fixed or Multiple Meanings?

133

the Empiricist tradition (see Price, 1969). Imagery and perception are seen as species of the same root concept, and the similarity between and equivalence of perception and imagery is highlighted. In the traditional Empiricist theory of imagery, particularly as championed by Hume (cf. Hannay, 1971; Price, 1969), imaging is seen as a postperceptual activity. Here the distinction between imagery and perception is not a qualitative one. They only differ in degree of intensity, in terms of vividness and detail. The work of Kosslyn (1980,1983) may legitimately be classified as a sophisticated version of this kind of theory. In his account images are seen like pictures on a TV screen, and are then "worked over'' by an "interpretive mechanism." Yet, such imagery experiences may only cover a limited number of cases, where the getting aspect of the image is unusually prominent. In most cases of deliberate imaging, the doing aspect of imagery becomes prominent. Interpretation now is inherently linked to the image through the initial intentional construction. Also, as Hannay (1979) points out, the Kosslyn model does not seem to cover the other polar activity of imaging that takes place in dreaming. Here the imager is almost a passive observer to whatever the brain turns up, and there is no ''putting" and "finding" here (Hannay, 1971, p. 552). In his equivalence model of imagery, Finke (1989) pursues the same track as Kosslyn (1980, 1983). Finke defines mental imagery as ''the mental invention or recreation of an experience that in at least some respects resembles the experience of perceiving an object or an event, either in conjunction with, or in the absence of direct sensory stimulation" (Finke, 1989, p. 2). In practice, Finke takes the position that there is a close resemblance between perception and imagery. The conceptual point of departure for Finke's research program is formulated in the principle of perceptual equivalence. This means that "imagery is functionally equivalent to perception to the extent that similar mechanisms in the visual system are activated when objects or events are imagined as when the same objects or events are actually perceived" (Finke, 1989, p. 41). From a series of experiments, Finke (I9891 arrived at the conclusion that the correspondence between perceiving and imaging is indeed close, and may extend to an impressively low level of processing. Finke has been able to demonstrate that a strong degree of equivalence may obtain between perception and imagery. Yet,

134

Geir Kaufmann and Tore Helstrup

Finke's research program may be criticized for being biased by focusing mainly on similarities of imaging and perceiving. Searching selectively for differences between the two kinds of mental activities might lead one to the opposite conclusion (see Intons-Peterson & McDaniel, 1991,and Intons-Peterson & Roskos-Ewoldson, 1989,for further discussion on this point). Besides, his model seems to suffer from the same serious limitation as the Kosslyn model. It seems clearly to be a special model, where there is no qualitative principle that distinguishes perception and imaging. Finke, like his Empiricist philosophical predecessors, seems clearly to contend that the distinction between the two kind of phenomena is a matter of degree of intensity, in terms of vividness and detail in content. Thus Kosslyn and Finke seem to play the hand too heavily in the opposite direction of the descriptionalist theory. A paradoxical result is that the theories, supposedly most favorable to imagery, seem stranded in the same untenable position as the purist theory. The equivalence of imagery and perception is taken to such an extreme as to deny any distinction in terms of qualitative principles. Thus there is no conceptual basis for making the required distinction between imagery and sensation. Paradoxically, these models also are seen to deny the existence of imaging as an autonomous mental activity (cf. Lawrie, 1970,for a systematic philosophical treatment of the issue of distinguishing imaging from sensation).

A Suggested Solution and an Alternative Conceptual Model

It seems to us that both the descriptionalist and the pictorialist theories fail by pressing the case for their contrasting views toward their cherished extremes. Based on logical as well as empirical considerations, time should be ripe for conceding that images are not either descriptions or mental pictures. This is probably the point Fodor (1976)has in mind when he claims that as far as images go, there is an "indefinite range of cases in between paragraphs and photographs" (Fodor, 1976,p. 190). From our point of view, images are neither pure symbols, nor pure perceptual experiences, but a typical hybrid mental concept with both symbolic and perceptual properties. In principle, images should, then, be seen as located across the boundary between thought and sensation. Also, we agree with the view of functional images put forward by Bartlett (1921,

Mental Imagery: Fixed or Multiple Meanings?

135

1925,1927). Images normally put to use are far more sketchy and fleeting than most current models of imagery suggest. This is in line with the claim made by Anderson (1990)that most images may be spatial rather then visual or pictorial in nature. In Figure 1,we suggest an alternative conceptual model of the nature of the imagery phenomenon, that we believe captures its full range. Based on the arguments put forward above, we also indicate the relative frequencies of the various types of images along the relevant conceptual dimension with the polar activities of thought and sensation. As is seen from Figure 1,the model implies that images have intentional properties (in the sense of reference and abstraction), and perceptual-experiential properties (intensity and duration). The model thus clearly distinguishes images from both pure symbols and pure perceptual experiences, which is logically required if imagery is to be considered an autonomous mental activity. We believe that this conceptual model allows for a new and more refined analysis of the concept of imagery. It is intended to be used as a framework for constructing more specific, explanatory theories of the nature and function of mental imagery. If one grants that images have a getting (i.e., an autonomous perceptual layout) and a doing aspect (i.e., intentional content), the model explicitly implies that reconstrual of images is possible. This implication of the model seems patently invalidated by the results of a series of experiments performed by Chambers and Reisberg (1985). In five experiments, Chambers and Reisberg presented certain classical ambiguous figures, like the duck-rabbit figure (see Figure 2) to their subjects. The task given to the subjects was to internalize the figure as a mental image, inspect the image and try to detect the alternative meaning configuration. Despite variations in procedure and extensive prompting in five experiments, not a single subject was able to find the rabbit when they had initially seen the duck or vice versa. Reisberg and Chambers (1992)corroborated these results in a new series of experiments involving different ambiguous figures (also see Chambers, this volume; Reisberg & Logie, this volume). What is natural and easy in perception thus seems impossible in imaging. Are images, after all, conceptually different from percepts? Based on their experimental evidence, Chambers and Reisberg (1985)took the strong stand that images are radically different from

Geir Kaufmnn and Tore Helstrup

136

Frequency

of kind of imagery

i

THOUGHT

SENSATION

-

-

-

INTENTIONAL

EXPERIENTIAL

reference

-

intensity

abstraction

-

duration

Figure 1 Postulated frequency of kinds of imagery along a continuum ranging between the polar concepts of thought and sensation.

percepts. They claimed that images are "saturated" with our understanding of their form, and, indeed, that images have no

Mental Imagery: Fixed or Multiple Meanings?

137

existence independent of our understanding. That is: Images are exclusively determined by the intentional description under which they are constructed. However, in terms of our conceptual model, the pictures used by Chambers and Reisberg hardly represent paradigmatic cases of functional images. In its detailed outline, the duck-rabbit figure is more like the polar case of mental pictures. Besides, mental images are dynamic. To inspect a mental image in many respects resembles the construction of this image. Mental images, therefore, have sharp details in the focal area, and vague parts in the periphery (cf., Finke, 1989). The duck-rabbit figure is more like a picture than an image. Pictures are sharp in all details, and they differ from images in lacking unfocused vague points. Being able to form images with detailed picture-like qualities may thus require extremely high imaging ability. In relation to the image reconstrual task, this is a particularly relevant consideration, because successful performance may depend on being able to preserve crucial details of the pictures in question (cf. Chambers, this volume; Peterson, this volume). As shown by Kosslyn, 1980,mental images suffer considerable fading in the normal case. The mental picture aspect of imagery is, therefore, designated in our model as a low frequency instance of imagery. As such, our model implies that although reconstrual should be possible, it should be difficult under the Chambers and Reisberg conditions. Consequently, successful performance may depend on more specific instructions, as shown by Hyman (this volume). But such a procedure may be regarded as violating the Chambers and Reisberg premise that reconstruals should be uncued. Alternatively, our model suggests that successful, uncued performance may require superior visualization ability. Thus we decided to replicate the Chambers and Reisberg experiment using the same ambiguous figures. In contrast to Chambers and Reisberg, who used ordinary university students as subjects, we employed a group of art students, presumably highly skilled in this kind of task. Our hypothesis is that at least some of these subjects should be able to reconstrue imagined, ambiguous figures under uncued conditions.

138

G'eir Kaufinann and Tore Helstrup

Method Subjects

Thirty-three students (5 males and 28 females) at the National College of Arts and Design in Bergen, Norway, participated in the study. Subjects participated on a voluntary basis and were selfselected as high visuals who used visualization processes extensively in their academic and artistic work. Stimulus Materials and Instrumentation

Six slides of classical ambiguous figures were employed and projected on a normal sized wall screen. The following four served as practice figures in the order listed: Schroeder's staircase, Necker cube, mother-in-law and vase-face. There were two test figures: duck-rabbit and chef-dog (see Figure 2). A 10-item, Norwegian version of the Minnesota Form Board Test was employed as a measure of individual differences in visualization ability. In a student population Cronbach's alpha is estimated to be .74 (N = 370).

Procedure Subjects were tested individually. A four-step procedure was used. In Step 1 subjects were informed about the experiment in general terms. It was explained that rival theories were being tested, not the abilities of the subjects. This was done to reduce test anxiety and possible fear of failure, which turned out to be a special problem in this group. In Step 2, the subjects were given the Minnesota Form Board Test, also as a way of introducing the subjects to visualization activity. Step 3 consisted of the training session, where subjects were presented Schroeder's staircase, the Necker cube, mother-in-law and the face-vase figures under ordinary perceptual conditions. Care was taken to point out that such figures could change both in form and meaning. The experimenter ascertained that the subject had perceived the ambiguity, and that they had succeeded in reversing the figures in perception before proceeding to the next step. Step 4 was the actual test situation. The subjects were instructed that, as in the previous instances, they

Mental Imagery: Fixed or Multiple Meanings?

Figure 2 Stimuli used as training and test figures.

139

140

Geir Kaufmann and Tore Helstrup

would be presented with figures that were ambiguous and could shift from one meaning configuration to another. They were told that they would have 5 sec to inspect the figure, and that they were to form a visual image of the figure that was as precise and detailed as they could possibly achieve. To avoid misunderstanding, the subjects were required to recount the procedure to the experimenter. Following the presentation of the test figure, they were told to examine their mental image to see whether they could discover a meaning configuration that differed from the first one. They were told that it might be helpful to inspect the figure at different locations: to the left, to the right, top or bottom. Similarly, they were instructed to manipulate the figure and to turn it over in various positions in their mind. Time from figure exposure to the first answer was recorded. If the answer was given immediately following figure exposure, the subject was asked whether the alternative was seen in the image or in the figure during exposure. One subject reported having seen both alternatives in both figures and was discarded from the analysis. Of the remaining subjects, half were given got the duck-rabbit first, and the other half were given the chef-dog first.

Results Based on the findings reported by Chambers and Reisberg (1985)and Reisberg and Chambers (1992)our null hypothesis was that no reversals should take place. Table 2 shows the results for the duck-rabbit and chef-dog figures separately, and for both. With the duck-rabbit, 4 of 26 subjects (15.4%) were able to detect the alternative meaning configuration in the duck-rabbit image. Given zero as expected frequency, the observed frequency in Table 2 yields a Fisher’s exact probability of .0555 for the null hypothesis. The corresponding value for the chef-dog is .0559. When we expand the test to comprise two items, a statistically significant deviation from the null hypothesis is shown: 28% of the subjects were able to reverse either one or the other, or both (Fisher’s exact probability = .0048). Even using a strict reversal criterion, these results differ sharply from the predictions based on the Chambers and Reisberg theory.

Mental Imagery: Fixed or Multiple Meanings?

141

~

TABLE 2 Number and percentage of reversals with "correct"solution (strict criterion).

N

Number of Reversals

Percentage

Duck-rabbit

26

4

15.4

Chef-dog

29

4

13.8

One or both

25

7

28.0

Test Stimulus

There were 32 subjects; six saw the reversal in the duck-rabbit picture (32 - 6 = 26), three saw the chef-dog reversal (32 - 3 = 29), and seven saw the reversal in one or both of the test figures (32 - 7 = 25).

An interesting additional finding of the experiment is that several alternative interpretations were suggested. Many of these seem at least partly appropriate and indicate that most of the subjects are able to reconstrue at least part of the image. For instance, with the duck-rabbit figure, several subjects suggest fish as an answer. On the hypothesis that the lower part of the original image has faded, this seems a reasonable suggestion for an alternative meaning configuration. Table 3 shows the results when fish was accepted as a valid interpretation in addition to rabbit and duck. In the either-one-or-both category the percentage of valid reconstruals now is seen to increase t o 40% (Fisher's exact probability =.0003). Many other reversals were also suggested by the subjects. On a lax criterion with fish, hand, palette (duck-rabbit) and foot, kneeboots (chef-dog)counted as valid responses, Table 4 shows that with a lax criterion an impressive 76% of the subjects were able to perform a valid reconstrual of the imaged form of the test figure. No significant difference was found in spatial ability scores between those who could reconstrue and those who could not (6.9 vs. 7.0, respectively). It should be kept in mind, however, that the range

Geir Kaufinann and Tore Helstrup

142

TABLE 3 Number and percentage of reversals with fish included as valid answer.

Teat Stimulus

N

Number of Reversals

Percentage

Duck-rabbit

29

7

24.1

Chef-dog

29

4

13.8

One or both

25

10

40.0

There were 32 subjects; three saw the duck-rabbit reversal (32- 3 = 291, three saw the chef-dog reversal (32 - 3 = 29),and seven saw the reversal in one or both of the test figures (32 7 = 25).

-

of scores on the spatial test was very restricted in this group (normal student population mean = 5.0).

Discussion

At the very least, our results clearly show that imagery reconstrual is possible. On the lax criterion, imagery reconstrual seems, in fact, to be highly probable, and natural, at least for this selected group of subjects under the present test conditions. Thus, the strong position adopted by Chambers and Reisberg (1985), Reisberg and Chambers (1992)and Reisberg and Logie (this volume) seems clearly untenable. Chambers and Reisberg (1985)do not report evidence on alternative answers. Our evidence is consistent with the hypothesis that part of the image has faded, thereby losing some of the exact details of the original test figure. Because the crux of the matter is whether it is possible t o reconstrue images, we should allow for some image fading, and employ a more lax criterion than the absolute standard provided by the correct answers. From a general theoretical perspective, our results indicate that there is, indeed, a getting aspect of imagery. Images may contain extraintentional, perceptual content that can be used to create new

Mental Imagery: Fixed or Multiple Meanings?

143

TABLE 4 Number and percentage of reversals with alternatives fish, palette, hand (duck-rabbit) and foot, Knee-boots counted as appropriate responses, plus one uncertain response (lax criterion). N

Number of Reversals

Percentage

Duck-rabbit

29

12

41.4

Chef-dog

29

8

27.6

One or both

25

19

76.0

Test Stimulus

interpretations of the original image. In Reisberg and Chambers’ terminology, we fmd that the figure can change caption as well as vice versa. The latter possibility is the only one that is permissible on the latest version of their theory (Reisberg & Chambers, 1992). It is interesting that what is held to be logically impossible by many prominent philosophers seems, nonetheless, to be an empirically demonstrable fact of nature. Such findings seem to justify the arguments of a growing number of philosophers (e.g., Block, 1983;Goldman, 1986;Harman, 1977) to the effect that the analytic-synthetic dimension falls on a continuum. Thus philosophy of mind cannot be done exclusively in the armchair, but has to take account of the empirical findings established in psychological experiments. Also, the evidence points clearly to important differences between perceiving and imaging. Even in a highly skilled group of subjects, only 15.4% are able to detect the standard answer to the duck-rabbit figure, an answer that is very easily detected by most subjects in the perceptual condition. This aspect of the results may serve as a warning to those who are eager to stretch the analogy between perceiving and imaging too far. Undeniably, there are important differences between the two kinds of mental activities. We need some qualitative principle t o distinguish between imaging and perceiving. The property of intentionality distinguishing between imaging and perception in terms of doing and getting may be quite important, yet not absolute. Even if Chambers and Reisberg have

144

Geir Kaufmann and Tore Helstrup

overstated their case in their distinction between perceiving and imaging, they have, nevertheless, been able to demonstrate an important dimension of difference between to two types of activities. This brings us back to the issue of the functional role of imagery in creative thinking processes. On the positive side, our results show that uncued imagery reconstrual is possible. We may argue, then, that images do carry perceptual information that may cause image transformations. Such findings seem to corroborate the basic premise of the functional theories of imagery to the effect that imagery has special perceptual properties that can not be reduced to its purely intentional content. These perceptual properties may provide the image with such special functional properties in cognitive processing as prescribed by these theories. Still, one might argue that reconstruals very easily performed in perception are much more difficult in imagery. Many sweeping claims have been made in favor of imagery as a privileged mode of representation in creativity. The results from imagery reconstrual experiments may be taken to suggest that imagery may be an awkward mode of representation when reconstruals in thought are necessary. It seems to us, however, that this view is ill-founded for several reasons. Firstly, being able to perform reconstruals within the image may not at all be required in the extrinsic task of using imagery as an aid to perform reconstruals and make discoveries in thought. As an example, one might refer to Einstein’s development of the theory of relativity as reported by Shepard (1978). The basic idea first came to him when, at the age sixteen, he imagined himself traveling along side of a beam of light. He realized then that the imaginary stationary spatial oscillation did not correspond to anything that could be perceptually experienced. Neither did it fit in with Maxwell’s equations for the propagation of light. There is no reconstrual within the image here. Rather, imagery processes are being used to promote discoveries on a highly abstract level of thinking. In our model (Kaufmann, 19901,this would be an example of imagery-based perceptual anticipations substituting for inductive thinking under extremely low programmed task conditions. Creativity through imagery may also be achieved by bringing together separate images into the same mental space, as has been demonstrated in a number of interesting experiments performed by Rothenberg (1986;also see Rothenberg BE Sobel, 1980, 1981, and

Mental Imagery: Fixed or Multiple Meanings?

145

Sobel & Rothenberg, 1980). Rothenberg defines a new concept termed "homospatial thinking" that is essential to high-level creativity. In distinction to a simple combination or integration process, homospatial thinking consists of bringing together and radically transforming brought-together elements, entities or phenomena into a new organization. By way of its link to perception, imagery is held to be a particularly well-suited representational system for the integration and transforming processes required in homospatial thinking. Experimental evidence from a task devised by Finke (this volume; also see chapters by Anderson & Helstrup, Intons-Peterson) indicates an important role of visual imagery in the integration of initially unrelated elements of experience into new combinations. The experimental evidence cited here lends support to theories of symbolic representations that emphasize the potential of visual imagery as a vehicle for creative thought. However, it still remains to be seen if the integrative properties of imagery are superior to those of other representational forms, like a linguisticpropositional mode of representation. Secondly, it can be argued that the Chambers and Reisberg paradigm is highly artificial in many important respects and may lead one to a distorted view of the potential of imagery as a vehicle for reconstruals. The figures used are hardly representative of the fleeting and sketchy images that are likely to constitute the paradigmatic case of functional images (cf. arguments above and the conceptual model shown in Figure 1). The procedure of putting a totally extraneous figure in the head of the subjects and asking them to perform reconstruals of it may be very different from what occurs in ordinary thinking processes. Normally, the individual is in a stream of thought aiming for the solution of a problem when images may appear as part of the total thinking activity. Thus, in ordinary cognition, images may lend themselves much more easily to reconstrual, and may still be an important facilitating mechanism in creativity and discovery processes in thinking. Future research on the issue of image reconstrual, and its importance to the creative thinking process, should therefore aim at a closer simulation of the natural conditions people operate under when engaged in creative activities.

146

Geir Kaufmann and Tore Helstrup

References Ahsen, A. (1984). The triple code model for imagery and psychophysiology. Journal of Mental Imagery, 8,15-42. Anderson, J. R. (1990). Cognitive psychology and its implications (3rd ed.). New York Freeman. Bartlett, F. C. (1921). The function of images. British Journal of Psychology, 11, 320-327. Bartlett, F. C. (1925). Feeling, imaging, and thinking. British Journal of Psychology, 16,16-28. Bartlett, F. C. (1927). The relevance of visual imagery to thinking. British Journal of Psychology, 18,23-29. Block, N. (1982).Imagery. Cambridge, MA:MIT Press. Chambers, D., & Reisberg, D. (1985). Can mental images be ambigous? Journal of Experimental Psychology: Human Perception and Performance, 11, 317-328. Dennett, D. C. (1969). Content and consciousness. New York: Humanities Press. Dennett, D. C. (1982). Two approaches to mental images. In N. Block (Eds.), Imagery (pp. 87-107). Cambridge: MIT Press. Engelkamp, J. (1990). Das menschliche Geduchtnis. Gottingen, Germany: Hogrefe. Finke, R. A. (1989).Principles of mental imagery. Cambridge, M A : MIT Press. Flew, A. G. N. (1956).Facts and "imagination." Mind, 65,392-399. Fodor, J. A. (1976). The language of thought, Hassocks, Sussex: Harvester Press. Ghiselin, B. (1952).The creative process. New York New American Library. Goldman, A. I. (1986). Epistemology and cognition. Cambridge: Harvard University Press. Hadamard, J. (1945). The psychology of invention in the mathematical field. Princeton, N J : Princeton University Press. Hannay, A. (1971).Mental images: A defence. London: George Allen & Unwin Ltd. Harman, G. H. (1977). Thought. Princeton, NJ: Princeton University Press. Heil, J. (1982). What does the mind's eye look at? The Journal of Mind and Behavior, 3, 143-149.

Mental Imagery: Fixed or Multiple Meanings?

147

Helstrup, T. (1988a). The influence of verbal and imagery strategies on processing of figurative language. Scandinavian Journal of Psychology, 29,65-84. Helstrup, T. (1988b). Performed memory acts in loci contexts: Some data and some theoretical reflections. In C. Cornoldi (Ed.), Proceedings of the Second Workshop on Imagery and Cognition (pp. 289 -300). Padova, Italy: University of Padova. Helstrup, T., & Anderson, R,. E. (1991). Imagery in mental construction and decomposition tasks. In R. H. Logie & M. Denis (Eds.), Mental images in human cognition (pp. 229-240). Amsterdam: North-Holland. Horowitz, L. M. J. (1983). Image formation and psychotherapy. New York: Jason Aronson. Intons-Peterson, M. J., & Roskos-Ewoldsen, B. (1989). Sensoryperceptual qualities of images. Journal of Experimental Psychology: Learning Memory, and Cognition, 15, 188-189. Intons-Peterson, M. J., & McDaniel, M. A. (1991). Symmetries and asymmetries between imagery and perception. In M. McDaniel and C. Cornoldi (Eds.), Imagery and Cognition (pp. 47-76). New York: Springer-Verlag. Kaufmann, G. (1979). Visual imagery and its relation to problem solving: A theoretical and experimental inquiry. Oslo: Norwegian University Press. Kaufmann, G. (1980). Imagery, language and cognition. Oslo: Norwegian University Press. Kaufmann, G. (1984a). Can Skinner define a problem? The Behavioral and Brain Sciences, 7,599. Kaufmann, G. (1984b). Mental imagery and problem solving. In A. Sheikh (Ed.), International Review of Mental Imagery, Vol. 1 (pp. 23-56). New York: Human Sciences Press. Kaufmann, G. (1985). A theory of symbolic representation in problem solving. Journal of Mental Imagery, 9, 51-70. Kaufmann, G. (1988). Mental imagery and problem solving. In M. Denis, J. Engelkamp, & J. T. E. Richardson (Eds.), Cognitive and neuropsychological approaches to mental imagery (pp. 231-239). Dordrecht Martinus Nijhoff. Kaufmann, G. (1990). Imagery effects on problem solving. In J. T. E. Richardson, P. J. Hampson, & D. Marks (Eds.), Imagery: Current developments (pp. 169-196). London: Routledge &

148

Geir Kaufmann and Tore Helstrup

Kegan Paul. Kaufmann, G. (in press). The content and logical structure of creativity concepts: An inquiry into the conceptual foundations of creativity research. In S. G. Isaksen, M. Murdock, R. Firestien, & D. Trefinger (Eds.), Understanding and recognizing creativity. Nomood, N J Ablex. Kolers, P. A,, & Smythe, W.E. (1979). Images, symbols and skills. Canadian Journal of Psychology, 33, 158-184. Kosslyn, S. M. (1980). Image and mind. Cambridge: Harvard University Press. Kosslyn, S. M. (1983).Ghosts in the mind's machine: Creating and using images in the brain, New York Norton. Kosslyn, S. M. (1987). Seeing and imagining in the cerebral hemispheres: A computational approach. Psychological Review, 94, 148-175. Marks, D., & McKellar, P. (1982).The nature and function of eidetic imagery. Journal of Mental Imagery, 6, 1-124. McKellar,P. (1957). Imagination and thinking. New York Basic Books. Miller, A. I. (1984). Imagery in scientific thought: Creating 20th century physics, Boston, MA: Birkhaiiser. Morris, P. E., & Hampson, P. J. (1983).Imagery and consciousness. New York: Academic Press. Morris, P. E., & Hampson, P. J. (1990). Imagery, consciousness, and cognitive control: The BOSS model reviewed. In P. J. Hampson, D. Marks, & J. T. E. Richardson (Eds.), Imagery: Current developments (pp. 78-102). London: Routledge & Kegan Paul. Neisser, U. (1976).Cognition and reality. San Francisco, C A W. H. Wheels. Neisser, U. (1978). Anticipations, images and introspection. Cognition, 6,169-174. Newell, A. (1969).Heuristic programming: Ill-structured problems. In J. Aronsky (Ed.), Progress in operations research, Vol. 3 (pp. 363-413). New York: Wiley. Paivio, A, (1971). Imagery and verbal processes. New York: Holt, Rinehart & Winston. Price, H. H. (1969). Thinking and experience. London: Hutchinson. Raaheim, K. (1974). Problem solving and intelligence. Oslo:

Mental Imagery: Fixed or Multiple Meanings?

149

Norwegian Universities Press. Raaheim, K. (1984). Why intelligence is not enough. Bergen, Norway: Sigma Forlag. Reisberg, D., & Chambers, D. (1991). Neither pictures nor propositions: What can we learn from a mental image? Canadian Journal of Psychology, vol. ??, pages??. Richardson, J. T. E. (Ed.) (1983). Mental imagery in thinking and problem solving. In J. St. B. T. Evans (Ed.), Thinking and reasoning: Psychological approaches (pp. 197-226). London: Routledge & Kegan Paul. Rothenberg, A. (1986). Artistic creation as stimulated by superimposed versus combined-composite visual images. Journal of Personality and Social Psychology, 50,370-381. Rothenberg, A., & Sobel, R. S. (1980).Creation of literary metaphors by superimposed versus separate visual images. Journal of Mental Imagery, 4,77-91. Rothenberg, A., & Sobel, R. S. (1981).Effects of shortened exposure time on the creation of literary metaphors as stimulated by superimposed versus separated visual images. Perceptual and Motor Skills, 53,1007-1009. Ryle, G. (1949). The concept of mind. London: Hutchinson. Sanford, A. J. (1985). Cognition and cognitive psychology. London: Weidenfeld & Nicholson. Sheikh, A. A. (1983). Imagery: Current theory, research, and application. New York: Wiley. Shepard, R. N. (1978).Externalization of mental images and the act of creation. In B. S. Randawa & W. E. Coffman (Eds.), Visual learning, thinking, and communication (pp. 133-1891, New York: Academic Press. Shorter, J. M. (1952).Imagination. Mind, 61,527-542. Simon, H. A. (1977). The new science of management decision. Englewood Cliffs, N J Prentice-Hall. Simon, H. A.(1978).Information theory of human problem solving. In W. K. Estes (Ed.), Handbook of learning and cognitive processes (pp. 271-198). New York: Wiley. Singer, J. L., & Pope, K. S. (1978). The power of human imagination. New York: Penguin Press. Sobel, R. S., & Rothenberg, S. (1980).Artistic creation as stimulated by superimposed versus separated visual images. Journal of

150

Geir Kaufinann and Tore Helstrup

Personality and Social Psychology, 39,953-964. Sternberg, R. J. (1988). The nature of creativity. New York: Cambridge University Press. Wittgenstein, L. (1953).Philosophical investigations. Oxford: Basil Blackwell.

Acknowledgments The authors wish to express their thanks to the faculty and students at the National College of Arts and Design for their interest, support and willingness to participate in this study.

Imagcry, Crcaiivity, and Discovery: A Cognitivc Pcrspcctive R. Koskos-Ewoldson, M.J. Intons-Petcrsoii and K.E. Andcrson (Editors) 0 1993 Elsevier Science Publishcrs R.V. All rights reservcd.

151

Mary A. Peterson Department of Psychology University of Arizona Tucson, AZ 85721 Mental images must differ from pictures in some respects. Neither percepts nor mental images are pictures of physical objects. A critical question is how to characterize the ways in which mental images differ from pictures, and how to use those characteristics to learn about the mental structures underlying mental imagery and shape memory. Furthermore, to the extent that we find that the mental structures underlying shape memory have different characteristics from the mental structures underlying linguistic or propositional memory, we can begin to explore the extent to which these different mental structures can play different roles in creativity. Three lines of experimentation have indicated some of the ways in which pictures and images differ: First, Reed (1974;Reed & Johnsen, 1975) demonstrated that the embedded figures task could not be performed very well in mental imagery; imagery did not seem capable of supporting the repartitioning processes necessary for finding embedded figures. It should be noted, however, that those embedded figures that were least likely t o be found in an image were the same as those that were least likely to be found in a picture. In other words, Reed's experiments did not indicate a complete dissociation between imagery and perception. Second, Hinton (1979a) demonstrated that the memory image of a three-dimensional cube failed to specify its three-dimensional structure precisely. Hinton's subjects reported that they could

252

Mary A. Peterson

imagine a diagonal line connecting the lower left front corner of a cube to its upper right back corner. They also reported that they could imagine rotating the cube so that the imagined diagonal line became vertical. Nevertheless, most of Hinton's subjects were unable to report either the correct number or the correct location of the corners that would lie between the cube's top and bottom corners following this rotation. Third, Chambers and Reisberg (1985) demonstrated that observers who saw only one interpretation of Jastrow's (1900)duck/ rabbit when viewing a picture were unable to reverse their mental image from duck to rabbit (or vice versa'); this finding was replicated in other experiments using the Necker cube and the Schroeder staircase. Chambers and Reisberg demonstrated that their results were not attributable to image degeneration by showing that all of their subjects could find the alternative interpretation in drawings generated from their own images. Based on their evidence, Chambers and Reisberg (1985;see also Reisberg & Chambers, 1991; Reisberg & Logie, this volume) claimed that mental images differ from pictures in that mental images cannot be reconstrued; that is, they cannot be separated from their interpretations. The first two lines of research described above indicate that not all of the part relationships that can be found in a picture or an object are preserved in a mental image. All three lines of research described above clearly demonstrate that pictures and mental images are not isomorphic. These previous experiments do not, however, demand the broad and general conclusion proposed by Chambers & Reisberg (19851,that mental images are inextricably bound to their interpretations. Moreover, other experiments conducted by Finke, pinker, and Farah (1989)have showed that certain types of mental images can be reinterpreted. For example, observers in the Finke et al. experiments combined imagined letters according to a set of instructions and were able to discover that a particular combination of a "J"and a "D"formed an umbrella. How are we to reconcile the findings of Finke et al. (1989)with those of Chambers and Reisberg (1985)? Finke et al. proposed two untested solutions: (1)that the complexity of classical reversible

Henceforth, I will use the term "duck-to-rabbitreversals"to indicate rabbit-to-duck reversals as well.

The Ambiguity of Mental Images

153

figures might preclude their reversal in mental imagery, and (2) that the reversal of classical reversible figures might require processes that are available in perception but not in mental imagery. I argue that neither of these conjectures captures the critical determinants of imagery reversal. Rather, I propose that the route to understanding reversals in mental imagery lies in appreciating both the diversity of types of reversal and the mechanisms and processes of shape recognition.

Types of Reversal Perception theorists studying reversible figures have long known that reversals are not unidimensional (e.g., Price, 1969). In fact, perceptual reinterpretations come in a t least three types, as illustrated in Figure 1. First, there are reversals that entail a reference-frame realignment. By "reference-frame realignment," I mean the reassignment of the top-bottom and/or front-back directions in a figure. Hence, when I use the term ''reference frame" I mean the specification of the shape's top and bottom and/or of its front and back. I will not use the term to denote an axis of symmetry or an axis of elongation, as some have done (cf. Marr, 1982). Reversals of the Necker cube and Mach book figures shown in Figures la and l b are examples of reference-frame realignments. Second, there are reversals that entail a reconstrual of the parts of the figure, but no (or little) reference-frame realignment. For example, in a reversal of Fisher's (1976)snaiyelephant figure (Figure lc), the snail's shell becomes the elephant's head and ears, and the body and head of the snail become the trunk of the elephant. Both the front and back and the top and bottom of the snail and the elephant are the same; hence, a reversal between the snail and elephant interpretations does not entail a reference-frame realignment. Reversal of Hill's (1915) wifdmother-in-law figure (Figure Id) mostly entails part reconstrual, although some slight change in reference frame might occur as well. Henceforth, I will refer to reversals that entail no or little reference-frame realignment as "reconstruals." Note that reversal of the duckhabbit figure, shown in Figure 2a, entails both reconstrual and reference-frame

Mary A. Peterson

154

ED a

b

@ ..

... .

C

d

Figure I Reversals of a and b require reference-frame

Types of reversal. respecification. Reversals of c and d require part reconstrual only. Reversal of c requires figure-ground alternation. (From "Mentalimages can be ambiguous: Reconstruals and reference-fkame reversals," by M. A. Peterson, J. F. Kihlstrom, P. M. Rose, & M. L. Glisky, Memory & Cognition, 20, 109.1

The Ambiguity of Mental Images

155

realignment: the bill of the duck becomes the ears of the rabbit, and the front of the head of the duck becomes the back of the head of the rabbit. Third, there are reversals that entail a redetermination of figure and ground relationships and hence, a repartitioning of the shape's contour. In figure-ground reversals, the part structure of the figure changes from one minute to the next, as Hoffman and Richards (1985) demonstrated so elegantly. The Rubin (1915) vase/ faces stimulus, shown in Figure le, is an example of this type of reversal. When the white region is figure, the faces are seen, and the parts of a face can be identified along the white side of the border between the black and white regions of the stimulus. The vase is not seen when the white region is figure; the black region becomes an undifferentiated background against which the faces are viewed. On the other hand, when the black region is figure, the vase is seen, and the parts of a vase can be identified along the black side of the border between the black and white regions of the stimulus. The faces cannot be seen when the black region is figure; the white region becomes an undifferentiated background against which the vase is viewed. Note that the parts of the vase and the faces are seen in different regions of the stimulus, both as defined by color (side of border) and as defined by contour segment: The noses begin and end at contour minima of curvature defined from the white side of the border whereas the bowl of the vase, although roughly coextensive, begins and ends at contour minima of curvature defined from the b1,ack side of the border. Henceforth, I will call these reversals "figure-ground reversals"; in figureground reversals, both figure-ground structure and part structure changes (Hoffman & Richards, 1985). For the remainder of this chapter, I will concentrate on reconstruals and reference-frame realignments only and not on figure-ground reversals. Research in my laboratory indicates that only certain types of borders support true figure-ground reversals (e.g., Peterson & Gibson, forthcoming; Peterson, Harvey, & Weidenbacher, 1991). On the b,asis of these experiments, I do not expect that true figure-ground reversals can occur in mental imagery, although some have been reported (see Reisberg & Chambers, 1991; Reisberg & Logie, this volume). In what follows, I will describe a series of experiments in which we examined

Mary A. Peterson

156

whether mental images of classical ambiguous or reversible figures were categorically incapable of reversal, as implied by Chambers and Reisberg‘s (1985)claim, or whether some types of reversal, say reconstruals, are not more likely to occur in mental imagery than other types, say reference-frame realignments.

Shape Recognition Theory and Its Implications for Mental Imagery Reversal Reconstruals Our expectation that reconstruals might occur in mental imagery was grounded in current theories of basic level shape recognition. In these theories, the shape representation medium is characterized as composed of a finite vocabulary of representational components. Were these components visible, they would correspond roughly to generalized cylinders (i.e., cross sections swept out in depth along axes orthogonal to the plane of the cross section) (e.g., Biederman, 1987;Binford, 1971, 1981;Marr & Nishihara, 1978). A componential analysis of a shape may operate at a number of different hierarchical levels (Marr, 1982). For example, at one level of analysis, a human arm and hand can be described by a single cylinder. At a more subordinate level of analysis, it can be described by three end-connected cylinders, one for the upper arm, one for the forearm, and one for the hand. At a yet lower subordinate level of analysis, the hand itself can be described by six cylinders (Marr, 1982;Marr & Nishihara, 1978). The components employed for basic level shape recognition may be those that can be fit between the minima of curvature identified along an object’s contour (Hoffman & Richards, 1985;Marr, 1977). I will focus on Biederman’s (1987)Recognition-by-Components (RBC) theory of shape recognition as a framework for this discussion because Biederman has taken the most explicit stand on the issues we find relevant to reversal of mental images. According to RBC, the vocabulary of representational components is quite small - 24 to 36 components, with a subset of these components used to represent any particular shape. If the 30,000 or so shapes we can potentially recognize are represented by this

The Ambiguity of Mental Images

157

small set of components, then each component must itself be capable of supporting thousands of interpretations (Biederman, 1987). According to Biederman, in any given shape, a single component is disambiguated both by the particular combination of other components to which it is connected and by their spatial inter-relations. Of course, features and details may contribute to further disambiguation. It seemed reasonable t o assume that mental images might be assembled from the same Components as those used in shape If that is the case, then it follows that representations. reconstruals of individual components might be quite likely, given that each component can support so many interpretations. Some proportion of these component reversals might find support in the particular arrangement of other components in the whole mental image and might, as a consequence, induce a reversal of the whole mental image. Hence, considering the question of image reversals within the context of shape recognition theory generates the prediction that reversals entailing reconstrual only should occur commonly in mental imagery.

Reference-Frame Realignments Inspection of the shape recognition literature affords a prediction regarding the likelihood of reference-frame realignments in mental imagery that likewise is quite different from that espoused by Chambers and Reisberg (1985;Reisberg & Chambers, 1991;Reisberg & Logie, this volume). A substantial body of shape recognition evidence shows that observers’ latency to name disoriented shapes increases linearly as the disparity between the shapes’ typical orientation and the presented orientation increases (e.g., Jolicoeur, 1985, 1988; Tarr & Pinker, 1989). Most investigators take the linear shape of the function relating disorientation to naming latency to indicate that a process similar to mental rotation precedes the recognition of disoriented shapes. Let me suggest another interpretation of the naming latency data: In the recognition process, descriptions of disoriented shapes may first be matched to shape representations whose canonical reference frame matches the viewer-centered or environmentcentered reference frame. If a good match cannot be found in this

158

Mary A. Peterson

set of representations, the shape description may be matched against sets of representations specified in other reference frames. The sets of representations searched first are those with reference frames similar to those of the input shape. The greater the disparity between the presented orientation and the shape's typical orientation, the larger the number of sets of representations searched prior to recognition. Recognition occurs when the description of the presented shape is matched to the appropriate structural representation, specified in the canonical reference frame. A finding that the latency for reference-frame reversals is longer than the latency for reconstruals would be consistent both with my interpretation of the naming latency data and with the mental rotation interpretation. Neither of these interpretations provides any foundation for predicting that reference-frame realignments would be impossible in mental imagery, however. Nevertheless, Chambers and Reisberg (1985;Reisberg & Chambers, 1991)have repeatedly demonstrated a failure to obtain referenceframe realignments in imagery. Of course, it is possible that certain aspects of Chambers and Reisberg's experiments limited the number of reversals they obtained. In particular, the demonstration figures used by Chambers and Reisberg to familiarize their subjects with the notion of reversibility (e.g., the Necker cube, the Mach book, and the Rubin figure-ground stimulus) all reverse in different ways than the ducklrabbit figure which served as their test figure (see Figure 1). Consequently, using them as demonstration figures may have induced incorrect strategies, thereby making ducWrabbit reversals less likely than if no strategies had been suggested or if a correct strategy had been suggested.

Overview

In what follows, I will describe two experiments examining imagery reversal that were drawn from a series of experiments conducted with John Kihlstrom, Patricia Rose, and Martha Glisky (Peterson, Kihlstrom, Rose, & Glisky, 1992). The experiments show (1) that reference-frame realignments can occur in mental imagery, but (2)that other types of reversal - reconstruals in

The Ambiguity of Mental Images

159

particular - are more common in mental imagery than referenceframe realignments, and (3) that the demonstration figures used are critical to whether or not; reference-frame realignments are obtained. We obtained these effects using two classic ambiguous figures: the Jastrow ducWrabbit figure, shown in Figure 2a, and Fisher’s snaivelephant figure, shown in Figure lc. In addition, I will describe two experiments that examine the structural versus semantic nature of the imagery reversal reports. Finally, I will discuss the implications of these findings for the structure of shape recognition systems and for the role of visual imagery in creativity.

Experiment 1 Subjects in this experiment (N = 116) were shown a full version of the ducWrabbit figure for 5 seconds and were asked to form a good memory image of it so that they would be able to draw it later. Following image formation, they were shown either (1)the goosehawk figure (Figure 2c), (2) the chefldog figure (Figure 2b), (3) the Rubin vasdfaces figure (Figure le), or (4) no demonstration figure. These demonstration figures were designed to provide the subjects with different types of implicit hints about reversal strategies. The goosehawk figure was adapted from Tinbergen (1948). Similar processes are involved in reversal of the goosehawk figure and the ducWrabbit figure, as can be seen in Figure 2: Both require reversing fronthack directions in the reference frame as well as reconstruing certain components. Accordingly, we predicted that if the demonstration figures serve to induce reversal strategies, subjects viewing the goosehawk figure should be more likely to reverse the ducWrabbit figure than subjects viewing the other demonstration figures. The chefYdog figure was one of the demonstration figures used by Chambers and Reisberg (1985). Reversal of the chevdog figure entails a reference frame reversal as well as a part reconstrual in that the top of the chefs hat becomes the hindquarters of the dog, as shown in Figure 2b. Because reversal of the chefldog figure entails a reference-frame realignment as well as part reconstruals, its presence as a demonstration figure might induce some subjects

160

Mary A. Peterson

V.C.

top, "rabbit" top. "duck" top

4 "duck" front

+

c "rabbit" front

top

V.C.

b.

V.C.

top, "eooao" top, "Iuwk" top

4

C.

Figure 2 The ducklrabbit figure a,the chef7dog figure b, and the goosehawk figure c with reference frame annotations.

The Ambiguity of Mental Images

I61

to use the correct strategy in reversing their image of the duck/ rabbit. Remember, however, that Chambers and Reisberg had used the chefYdog figure as a demonstration figure and had obtained no reversals of the ducWrabbit image whatsoever. Indeed, notice that reversal of the chef7dog figure entails a different type of referenceframe realignment than reversal of the ducwrabbit figure does. For example, the object-centered top is not the same as the environment-centered (or viewer-centered) top for either the chef or the dog interpretation, whereas the object-centered top of both interpretations of the duckhabbit figure corresponds to the environment-centered (viewer-centered)top of the stimulus. Notice also that reversal from the chef interpretation into the dog interpretation for the chef7dog figure requires a 900 respecification of the object-centered top, whereas reversal from the duck interpretation into the rabbit interpretation for the ducWrabbit figure requires a 1800 respecification of the object-centered front. Thus, the reversal strategies suggested by the chef7dog figure may not be completely appropriate for the duckhabbit figure. Consequently, subjects viewing the chefldog figure might be less likely than subjects viewing the goosehawk figure to reverse the duckhabbit figure. The Rubin vase/faces figure-ground stimulus was also among the demonstration figures used by Chambers and Reisberg (1985), whose subjects failed t o report any reversals of their mental image of the ducWrabbit figure. As noted earlier, reversals of figureground stimuli may follow different principles than those involved in reversals of the ducWrabbit figure. Thus, any strategies induced by figure-ground stimuli would be inappropriate for reversing the ducWrabbit figure. Therefore, we predicted that subjects in the figure-ground demonstration figure condition would be unlikely to reverse their image of the duckhabbit figure. In the condition in which no demonstration figure was shown, subjects counted backwards from 500 for durations that were matched to the durations subjects in the other conditions viewed the demonstration figures. Immediately following their exposure to the demonstration figures, subjects were told that the picture they had viewed originally (i,e., the Jastrow ducWrabbit figure) was (also) an

162

Mary A. Peterson

ambiguous figure and they were asked t o try to find an alternative interpretation in their mental image. Subjects who did not report a duck-to-rabbit reversal were then asked to attend first t o the left side of their mental image and then to the right side in an attempt to find an alternative interpretation. (These strategies were suggested to subjects by Chambers and Reisberg, 1985, as well.) Subjects who were still unable to report a duck-to-rabbit reversal were given the two explicit reference-frame hints used by Hyman (this volume; Hyman & Neisser, 1991). The first explicit hint contained no reference to animals; we simply asked the subjects to "Consider the front of the thing you were seeing as the back of something else." We refer to this hint as the "abstract" referenceframe hint. Subjects who failed to report a duck-to-rabbit reversal following the abstract reference-frame hint were given a second hint, containing category information as well as reference-frame information. They were asked to "Consider the front of the head of the animal you just reported as the back of the head of some other animal." We refer to this hint as the "conceptual" referenceframe hint. Hyman and Neisser had found that the conceptual reference-frame hint was more effective than the abstract reference-frame hint in prompting a reversal. At the end of the experimental session, we clarified any uncertainties about the interpretations subjects had named earlier by asking them to point out the parts of the interpretations they had named, as well as the fronthack and tophottom objectcentered directions. Our subjects offered many interpretations of their mental images other than, or in addition to, ducks and rabbits. Of course, some interpretations were more valid than others. We considered an interpretation to be valid if it met a two-part criterion: (1)The interpretation had to be of the entire mental image, not just a single part. This meant that interpretations like "nose" and Yhgers'' were not considered valid. (2) The interpretation had to name a particular shape depicted by the mental image, and not merely describe the image in general terms. This meant that interpretations like "shape" and "animal" were not considered valid. The valid interpretations were structural reconstruals of the mental image in that their componential structure matched that of the original figure. Examples of the valid interpretations are listed in

The Ambiguity of Mental Images

163

Table 1, along with a specification of whether their reference frame was considered to be the same as that of the duck or the rabbit interpretation of the duckhabbit figure. The componential structure of the ducWrabbit figure and of these valid interpretations is shown in Figure 3. It is clear that, at a basic TABLE 1 The proportions of subjects reporting each of the "valid" interpretations of the ducklrabbit mental image classified according to whether their reference frame is the same as that of the duck or the rabbit interpretation. Interpretations

- Reference Frame

Proportion

Rabbit

R

.94

Duck

D

.34

Hand making a peace sign (or shadow figure)

D

.14

Fish

R

.ll

Pair of scissors

D

.10

Deedantelope

R

.09

DOg/pupPY

R

.08

Profile of person with hair streaming behind head (or wearing a feather headdress)

R

.06

Note: The reference frame for the various interpretations is marked with a D if we considered the front to be the same as the front of the duck, and an R if we considered the front to be the same as the front of the rabbit. Interpretations within the same reference frame that followed one another were scored as reconstruals. Interpretations within different reference frames that followed one another were scored as reference-frame realignplents.

level, the components of all of the valid interpretations are similar.

Duck-rabbit Reversals. Reversals between the duck and rabbit interpretations of the mental image of the duckhabbit figure are shown in the top part of Table 2. The data in the table are grouped according to reversals obtained before the explicit

164

Mary A. Peterson

Figure 3 The central figure is the basic componential structure of the Jastrow duckhabbit figure and of a number of alternative interpretations. The interpretations shown above the basic structure are those specified in the same reference frame as the duck interpretation; they are from left to right, a duck, a pair of scissors, and a hand making a peace sign. The interpretations shown below the basic structure are those specified in the same reference frame as the rabbit interpretation; they are, from left to right, a rabbit, a fish, and a person with hair (or headdress) streaming behind.

The Ambiguity of Mental Images

165

reference-frame hints and reversals obtained after the explicit reference-frame hints. Note that some subjects in each group reported duck-to-rabbit reversals (i.e., reference-framerealignments) before the explicit reference-frame hint. Except in the figureground condition, the proportion of subjects reporting duck-torabbit reversals prior t o the reference-frame hint was significantly greater than expected on the basis of Chambers and Reisberg's experiments (3.07 Iz I 12.69, p c .001).2 In addition, the proportion of subjects reporting duck-rabbit reversals prior to the reference-frame hint was greater in the goosehawk condition than in the other conditions, zs 2 2.27, p s c .02, as predicted. Subjects in all conditions were more likely to report a duckto-rabbit reversal after the reference-frame hints than before the reference-frame hints, but the increase in reversals was statistically significant only in the figure-ground condition. We did not find, as Hyman and Neisser had, that the conceptual reference-frame hint prompted more duck-to-rabbit reversals than the abstract reference-frame hint, z < 1. Accordingly, the responses t o the two types of reference-frame hint are not reported separately here (for details, see Peterson et al., 1992). Despite the increase in duck-to-rabbit reversals following the reference-frame hints, the total proportion of duck-to-rabbit reversals remained larger among subjects who had viewed the goosehawk demonstration figure than among subjects in any of the other three conditions (z 2 1.94, p s < .03).

All Reference Frame Realignments. The middle part of Table 2 shows reference-frame reversals of any sort, including both duck-to-rabbit and other :reference-frame realignments (i.e.,

The tests of significance of these proportions were found in Bruning and Kintz (1977). When comparing our proportions against expected values, we could not use an expected value of 0, which is the correct value based on Chambers and Reisberg's (1985) experiments. Accordingly, we calculated an expected value of .02, using the following reasoning. Chambers and Reisberg tested a total of 55 subjects, not one of whom reported a reversal in imagery. I:f we suppose that the next subject might have reported a reversal, then we arrive a t .02as a conservative theoretical estimate of the predicted probability of reversal. Because our predictions regarding the presence of reversals in imagery were clearly unidirectional, we employed one-tailed tests for these comparisons. All other tests were two-tailed.

Mary A. Peterson

166

reversals between any two interpretations listed in Table 1 that are specified in different reference frames). When the proportions of all reference frame reversals in imagery were examined, reversal probabilities before the reference-frame hints were greater in the chefldog condition than they were when only duck-to-rabbit reversals were counted (.27 vs. .lo), z = 3.22, p < .002). There were similar trends

TABLE 2 Duck-to-rabbit reversals, all reference-frame realignments, and all reversals reported by subjects in Experiment 1. Questions

Goose/€Iawk

Chef/Dog

Figure-Ground

None

Duck-to-Rabbit Revereale Before RF hints

.35

.10

.06

.10

After RF hinta

.52

.23

.27

.24

All Reference-Frame Rea@nmente Before RF' hints

.38

-27

.24

.21

After RF hints

.66

.65

.52

.48

All Reversah Before RF hints

.69

.69

.65

.69

After RF hinta

.93

.83

.86

.89

Note: Reversals are cumulative. All reference-frame realignments include duckto-rabbit reversals as well as other reference-frame reversals. All reversals include all reference-frame realignments as well as reconstruals. RF = reference frame.

in the figure-ground condition (.24 vs. .06) and in the no demonstration figure condition (.21 vs. .lo), although these did not reach statistical significance. When all reference-framerealignments were considered, the differences among the conditions disappeared: A minimum of 21% of subjects in all conditions reported referenceframe reversals of their image of the ducWrabbit figure before the

The Ambiguity of Mental Images

167

explicit reference-frame hints were given (range 21% to 38%). Once again, the likelihood of reporting a reference-frame realignment increased following the reference-frame hints.

All Reversals. The bottom part of Table 2 shows all reversals (including reconstruals as well as reference-frame realignments). When reconstruals are counted, the likelihood of reversal prior t o the explicit reference-frame hints is significantly greater in all conditions than it is when only reference-frame realignments are counted, 2.36 I z I 3.67, p s c .02. Between 65% and 69% of all subjects reported a reconstrual of their mental image of the ducWrabbit figure before the reference-frame hints were given. In all conditions, the likelihood of reversal increased still further following the referenceframe hints. The results of Experiment 1 support the hypothesis that the type of demonstration figures used is critical to whether or not duckto-rabbit reversals are obtained in mental imagery: Reversals between duck and rabbit interpretations were most likely when reversal of the demonstration figure (the goosehawk figure) entailed the same types of reversal as did reversal of the experimental figure. Experiment 1 also suggests that the type of demonstration figures used may be less critical t o whether or not reconstruals or other reference-frame reversals are obtained: These types of reversals were equally likely in all demonstration figure conditions. In addition, Experiment 1demonstrates clearly that referenceframe reversals are possible in mental imagery: At least 20% of the subjects in all demonstration figure conditions (including the no demonstration figure condition) reported reference-frame reversals before the explicit hints were given (see the third row of Table 2). Finally, Experiment 1 shows that reconstruals are common in imagery: At least 65% of the subjects in all demonstration figure conditions (including the no demonstration figure condition) reported reconstruals before the explicit hints were given (see the fifth row of Table 2). Moreover, the occurrence of reconstruals was unaffected by the demonstration figures, as expected. Why did we obtain reversals of mental images when Chambers and Reisberg did not? One possibility consistent with our results is that the reversal strategies suggested by the set of demonstration figures used by Chambers and Reisberg (1985) (i.e., the Necker cube,

168

Mary A. Peterson

the Mach book, and the Rubin figure-ground stimulus) were sufficiently misleading so as to preclude duck-to-rabbit reversals, even though the more appropriate chefldog figure was presented later. Second, Chambers and Reisberg discounted responses other than "duck" and "rabbit," assuming that these other responses were not reports of true mental image reversals, but rather were simply responses to the demand character of the experiment, implicit in the experimenter's repeated questioning. Given that we obtained some duck-to-rabbit reversals, the second explanation is not sufficient to explain why we obtained reversals of mental images whereas Chambers and Reisberg did not. Moreover, I will show below (Experiments 3 and 4) that the valid interpretations of Experiment 1 are unlikely to be generated via non-structural routes. A third possibility is that Chambers and Reisberg's subjects were not sufficiently motivated to report reversals of their image, perhaps because they were recruited for the experiment while they were visiting the library for other reasons (our subjects received either course credit or monetary payment). Chambers and Reisberg not only recruited their subjects in the library, they conducted their experiments there as well (ours were conducted in a conventional laboratory environment). Perhaps implicit demands present in a library setting and not in a controlled laboratory setting conspired to lower the likelihood that Chambers and Reisberg's subjects would report reversals (e.g., subjects' desire to get back to what they were doing before being interrupted). Experiment 2 The valid interpretations of the duck-rabbit figure which we scored as reconstruals or as other reference-frame realignments in Experiment 1, while structurally consistent with the Jastrow ducW rabbit figure, were clearly not the designated interpretations for that figure. Although some observers do report these interpretations when viewing pictures, the proportions are quite small (see Experiment 3, below). We are left with the question of whether reversals of mental images necessarily differ in nature from reversals of pictures. To examine this question, we conducted another experiment,

The Ambiguity of Mental Images

V.C.

169

top, "snail" top, "elephant" top

4

l*snaill*and

"elephant" front +

a.

b. Figure4 (a) Fisher's snaiVelephant figure with reference frame annotations. (b) The simplified snaivelephant figure. (From "Mental images can be ambiguous: Reconstruals and reference-frame reversals," by M. A. Peterson, J. F. Kihlstrom, P. M:.Rose, & M. L. Glisky, Memory & Cognition, 20, 118.)

170

Mary A. Peterson

using a figure whose two dominant interpretations require reconstrual only, without any reference-framerealignment: Fisher’s snaiyelephant, shown in Figure 4a with reference frame annotations. If imagery reconstruals, which we found to be pervasive in Experiment 1, are similar to perceptual reconstruals, then reversal between the two designated interpretations of the snaiyelephant figure should occur commonly in imagery. Such a finding would indicate that imagery reversals do not necessarily differ in kind fromperceptual reversals. We began by constructing a simpler version of the snaillelephant figure (Figure 4b), which we thought might be easier t o hold in imagery than the original version. We tested whether our simplified version still supported snail-to-elephant reversals by asking perceiving subjects t o list as many objects-resembledby the picture as they could. This procedure also served as a means of identifying other valid reconstruals for the image a priori: Interpretations offered by more than 10% of these perceiving subjects were designated as valid interpretations for the image. As shown in the top part of Table 3, perceiving subjects reported both of the designated interpretations of our simplified snaiyelephant figure, and a number of other valid interpretations as well (e.g., bird, seashell, leaf, flower, helmethat). Alternations between any two of these valid interpretations constituted a reversal (with one exception: seashell did not count as a different response from snail and vice versa). In the imagery section of this experiment, 12 subjects viewed the snaiVelephant figure under conditions similar to those employed in the no demonstration figure condition of Experiment 1. As shown in the bottom part of Table 3, we found that 83% of our imagining subjects reported reversals between two valid interpretations of our snaiYelephant figure. Of these reversals, 60%were snail-to-elephant reversals. Thus, Experiment 2 demonstrates that reversals between the two dominant perceptual alternatives also dominate in imagery when those reversals entail reconstrual only, without reference-frame realignment. Thus, reversals of images do not necessarily differ in kind from reversals of pictures. Implications of Experiments 1 and 2

Taken together, the results of the first two experiments

The Ambiguity of Mental Images

171

TABLE 3 Proportions of perceiving subjects reporting the valid interpretations of the simplified version of the snaillelephant figure and the proportions of imagining subjects reversing between two valid interpretations in Experiment 2. Interpretations

-

Proportion of Subjects

Perceiving Sukects Elephant

73%

Snail'

69

Bird

31

Seashell'

20

Leaf

20

Flower

16

Helmemat

13

Imagining Subjects Snail/Elephant

.33

ElephanVSnail

.17

SnaiVLeaf

.17

SnaiVBird

I

08

ElephanWlower

-08

Total Reversals in Imagery

.83

'Alternations between these two interpretations were not counted as reversals. Note: A n additional 39 interpretations were offered by less than 10%of the viewers looking at the maillelephant figure.

demonstrate that reconstruals occur commonly in imagery, regardless of whether the most common type of perceptual reversal is a reconstrual (Figure 4b)or a reference-frame realignment (Figure 2a). Furthermore, the results of Experiment 1 indicate that referenceframe realignments are clearly possible in mental imagery even

172

Mary A. Peterson

without the administration of explicit or implicit hints, even though they are less common than reconstruals.

Perception Versus Imagery What remains to be explained is why reconstruals of the duck/ rabbit figure dominate in mental imagery, whereas reference-frame reversals dominate in perception. (We confirmed this latter assumption by showing the duckhabbit figure to perceiving subjects, and found that 93% of them reported duck-to-rabbit reversals.) In our view, reversals in both imagery and perception illuminate the structure and operation of shape memory. We argue that in both initial recognition processes and reversal processes, there exists a stage in which a structural description of a shape is not connected to an interpretation; the connection may not yet have been made in recognition, whereas it may have been severed in reversal. At this stage, we propose that all representations specified in the same reference frame as that initially assigned to the input shape are searched exhaustively. If a good match is found, recognition (or reversal) occurs. For any number of reasons (perhaps including fatigue of the best-fitting representation in reversal situations), a good match may not be found in this original search set. In the absence of a good match in the original set of representations, the search through memory representations is widened gradually to include representations specified in different reference frames. Thus, representations in memory are accessed in functional subsets according to their reference-frame specifications. A process such as this one could underlie the naming latencies found by Jolicoeur (1985, 1988) and others in shape recognition experiments. It can also account for our imagery results, and for the discrepancy sometimes obtained between the types of reversals found in mental imagery and in perception, as follows. Suppose that the individual components as arranged in the duckhabbit figure support a number of interpretations within the original reference-frame set. All but one of these interpretations within the original referenceframe set may be subverted by the surface details available on a picture that is continually in view. Consequently, these subverted reconstruals will not be seen when the picture is present. When

The Ambiguity of Mental Images

173

reversal opportunities occur for pictures that remain in view, the failure to find an alternative match in the original reference-frame set will precipitate a search through sets of representations specified in different reference frames, until finally a good match is found in a set of representations with a reversed frontback reference frame, producing a duck-to-rabbit reversal. What’s different about imagery reversals? Two things perhaps: First, it may be the case, as Kosslyn (1980) has argued, that fewer surface details are present in the image than in the picture. Hochberg (1968, 1981) makes a similar claim about the relationship between percepts and objects. (Or it may be the case that subjects can choose to ignore details in an image more easily than in a picture.) If indeed, fewer details are present in the image than in the picture, potential reconstruals within the original reference-frame set might not be precluded in mental imagery. Consequently, reconstruals would be more common in imagery than in perception. A second distinction between imagery and perception is required to account for our finding that reference-frame reversals are less common in imagery than in perception. This characteristic might be a consequence of the fact that it takes longer to complete a match between a structural description and a representation specified in a reference frame different from that of the input shape (e.g., Gibson, in progress; Jolicoeur, 1985,1988; Peterson, et al., 1991). Within the time frame required to complete this match, certain processes endemic to mental imagery may interfere with reference-frame reversals. For example, suppose that mental images must be regenerated periodically, as Kosslyn (1980; Reiser, Kosslyn, Farah, & Fliegel, 1983) has proposed. Suppose further that image regeneration is conducted within the original reference frame in which a mental image was specified. If the periods of image regeneration fall within the time required to conduct the search that culminates in a reference-frame reversal, then the image regeneration process might interfere with the search process, by resetting it, for example. Hinton’s (1979a & b; Hinton & Parsons, 1981) proposal that mental images are generated within a viewercentered reference frame is consistent with this interpretation. Reconstruals would not be affected by this re-setting procedure because, by definition, they are described in the same reference frame as the original interpretation. (Of course, the finding that at

174

Mary A. Peterson

least 21% of the subjects in Experiment 1 reported reference-frame reversals before the reference-frame hint suggests that the search process is not totally constrained by an image refreshing process.) In addition, Experiment 1 suggests that observers given implicit or explicit hints can intervene in this matching process, thereby causing a reference-frame reversal to occur.

Experiment 3 The prediction that reconstruals might be more likely to occur when fewer surface details are present need not be limited t o mental imagery conditions; it should extend to perception conditions as well. Accordingly, a test of this prediction was conducted in Experiment 3 with perceiving subjects. Subjects viewed different versions of the Jastrow duckhabbit and listed all the objects the picture could depict. One group (N = 43)viewed an intact version of the Jastrow duck/ rabbit figure. Other groups of subjects (N = 46/group) viewed schematic versions of the Jastrow figure that were missing certain details, either the eye and/or the rabbit’s nose. The results of this third experiment are shown in Table 4. Note first that perceiving observers reported many of the interpretations that had been reported by imagining subjects in Experiment 1 (see Table 1).Some of the interpretations were more likely to be reported by observers viewing schematic versions of the ducwrabbit figure than by observers viewing the intact ducWrabbit figure. For example, 24% of the observers viewing a schematic figure reported that the figure resembled a fish, whereas only 2% of the observers viewing the intact Jastrow ducWrabbit figure reported that it resembled a fish. In addition, some interpretations offered for the schematic ducwrabbit figure and for the mental image in Experiment 1were not offered by observers viewing the intact figure (e.g., deer; profile of a person with hair streaming behind; hand-held hair dryer3). These interpretations were offered by our imagining subjects, as shown in Table 1.

The “hand-heldhair dryer” interpretation was not offered by any of the imagining subjects in Experiment 1. This interpretation was offered by imagining subjects in other experiments reported in Peterson, et al. (1992),however.

The Ambiguity

of

Mental Images

175

Thus, the results of Experiment 3 are consistent with the claim that the non duck-to-rabbit reversals reported by imagining subjects were structural reversals. The same interpretations were offered by perceiving observers. In addition, there is some suggestion in the results of Experiment 3 that details present on the picture may preclude fitting certain interpretations to the picture, interpretations that can be fit when schematic versions of the picture are viewed.

TABLE 4 Interpretations offered by observers looking at the intact Jastrow figure and schematic versions.

Interpretations

Intact Jastrow Figure

-

Schematic Jastrow Figure

(%I

(%)

93

92

100

90

Fish

2

24

Hand making shadow figure or peace sign

9

8

Profile of person with hair streaming behind

0

4

Duck Rabbit

2

Deedantelope Pair of scissors Dog/PuPPY Hand-held hair dryer'

'This interpretation was not offered in Experiment 1 of this paper, but it was offered in other experiments. using similar methodologies (see Peterson, et al., 1992).

Experiment 4

We have argued that the reconstruals reported by our subjects

176

Mary A. Peterson

were structural reconstruals, in that they depend on the componential structure of the image (see Peterson, et al., 1992). But we have not shown that structural reconstruals differ from semantic associates that might be produced to mental images of ducks and rabbits, or to the corresponding concept. Empirical evidence regarding a distinction between structural and semantic reconstruals is relevant to our claim that imagery reversals and perceptual reversals do not necessarily differ in kind. Experiment 4 was designed to identify the semantic associates of the words ''duck' and "rabbit" and of mental images generated in response to these words. Subjects were asked to list all the objects that came to mind when they saw either the word "rabbit" ( N = 12) or the word "duck" (N = 12) printed on a page, or when they created a mental image of either a rabbit (N = 12) or a duck (N = 12) in response to the printed words. They were given one minute to generate these associations. Different subjects participated in each of these conditions. None of the subjects in this experiment was shown the Jastrow duckhabbit figure; hence, their mental images of ducks or rabbits were not expected to resemble the rather idiosyncratic Jastrow figure. We had two goals in examining the associations generated by subjects under these conditions. First, we were interested in seeing whether or not any of the interpretations we had considered to be valid reversals out of a duck (or rabbit) interpretation in Experiment 1were offered as semantic associates of ducks (or rabbits) under the conditions of Experiment 4. If so, that would undermine our assessment that those reversals of mental images of the idiosyncratic Jastrow duckhabbit figure were structural reversals rather than semantic reversals. Second, we were interested in seeing whether the associates generated to a mental image of an object were the same as those generated to the corresponding word. If not, that would suggest that the mental structures indexed by the activation of the mental image were different from those indexed by the activation of the lexical item. If associations generated to a mental image of an object can be shown to differ from associations generated to the name of that object, then Experiment 4 may serve to indicate one way in which the use of visual imagery can lead thought in creative directions. Subjects generated an average of eight associations per condition. Table 5 shows the average proportion of associations that

The Ambiguity of Mental Images

177

were duplicated across any two of the conditions. Notice first the overlap between items generated in response to ducks and rabbits. Across both word and image conditions, approximately 15% of the items generated in response to one of these stimuli was also generated in response to the other. Thus, very small proportions of associates were generated for both ducks and rabbits. The duplicated responses included items like "hunter," "grass," ''food,'' "baby," "white," and "animal." None of these responses would have been

TABLE 5 Proportions of associations duplicated across conditions.

Conditions

Rabbit

Duck Word

Image

Word

Image

1.00

.40

.14

.14

1.00

.16

.14

1.00

.34

Duck Word

Image

Rabbit Word Image

1.00

scored as a valid reversal in Experiment 1. Indeed, the only one of these items that was ever reported as an interpretation of an image in Experiment 1was "animal," ilnd in Experiment 1, ''animal'' was not accepted as a valid interpretation. Thus, it appears that subjects in Experiment 1were not simply generating semantic associates to their initial interpretation in response to the experimenter's continued prompts for alternative interpretations. The results of Experiment 4 show that taking a semantic route from duck does not lead one to a rabbit (or vice versa), nor to a pair of scissors, nor to a hand making a peace sign (see Table 1and Figure 3). The failure to

Mary A. Peterson

178

obtain these items under the conditions of Experiment 4 is consistent with the interpretation that the reconstruals obtained in Experiment 1 were structural reconstruals; that is, they were matches between representations in shape memory and the structural descriptions of the imagined shape. Two subjects in the word-duck condition and one in the imagerabbit condition reported "dog1'as an associate. Although "dog" was one of the valid interpretations in Experiment 1, none of the imagery subjects who initially saw the ducWrabbit figure as a duck

TABLE 6 Proportions of associations in various categories that were offered in the word and image conditions of Experiment 4.

Category

Image (N = 191)

Word ( N = 187)

Actions

.03

.04

Adjectives

.12

.06

Alternative name

.oo

.03

Color

.05

.06

Exemplars

.02

.05

Perceptual partslattributes

.15

.16

Scene Components

.28

.20

Semantic Associates

.24

.26

Sound

.01

.02

Superordinate

.04

.05

Miscellaneous

.05

.08

reinterpreted it as a dog. Thus, the fact that it was reported as a semantic associate to the word "duck1'in Experiment 4 is of no consequence for the results of Experiment 1. However, quite a few subjects in Experiment 1 reported "dog" as a reconstrual of mental images they had initially interpreted as rabbits. All but one of these

The Ambiguity of Mental Images

179

subjects also offered other reversals of their mental images, so the pattern of results obtained in Experiment 1would remain essentially unchanged were we to exclude "dog8' from the list of structural recon~truals.~ The second pattern to notice in the data shown in Table 5 is that less than half of the associates reported to an object and the corresponding name were the same. Indeed, 65-67% of the associates reported to the imagined objects were unique to the mental image condition ke., these items were not produced when the corresponding word served as the seed for the association responses). This finding suggests that the majority of the mental structures indexed by the activation of a mental image were different from those indexed by the activation of the corresponding lexical item. The associates offered to the mental images and to the names of the objects were classified into the categories shown in Table 6 . Even though the item overlap in the word and mental image conditions was low, it is clear that for the most part, the different items generated to mental images and to lexical items belong to similar categories, at least for the following categories: actions (e.g., "running," "swimming"), adjectives (e.g., "fast," ''silly''), colors (e.g., "white," "green"),perceptual features (e.g., "ears," "bill"),sounds (e.g., "quacking"), and, superordinate terms (e.g., "animal," "bird"). The degree of overlap is som.ewhat lower for exemplars (e.g., "jackrabbit," "Donald") and for the two major categories into which the associations were classified: semantic associations and scene components. Scene components were defined as concrete objects that might be found along with the critical item in a typical scene. Examples of objects scored as "scene components" are "lake," "tree," and ''park," Semantic associations differed from scene components in that semantic associations (1)'were not necessarily objects, and (2) would not tend to occur in a typical scene containing the named object. Examples of semantic associates for rabbit were "car," "chocolate,""Easter''; some examples of semantic associates for duck were "cartoon," "dinner," "orange." The data suggest that a larger proportion of semantic associations are indexed when words rather than mental images serve as the association cues, and that a larger

The one subject whose only reported reconstrual of his image was from "rabbit"to "dog"was in the no demonstration figure group.

180

Mary A. Peterson

proportion of scene components are indexed when mental images rather than words served as the association cues. These data are merely suggestive because the number of items sampled was so small, but they suggest an intriguing direction for future research. It is of note that previous investigators whose work has been taken as relevant to imaginal associative processing did not use an association paradigm like that used in Experiment 4. Instead, previous investigators either used a scanning paradigm to test the spatial structure of mental images of a production paradigm to study the structure of mental scripts (for a summary, see Clark & Paivio, 1987).

Implications of Experiments 3 and 4 The results of Experiments 3 and 4 support the notion that the interpretations of the ducWrabbit image that we scored as valid interpretations in Experiment 1were indeed structural reversals: In Experiment 3, perceiving subjects reported seeing many of those interpretations. Moreover, the fact that some of these interpretations were more likely to be offered by perceiving subjects viewing schematic stimuli in Experiment 3 is consistent with the proposal that one difference between perception and imagery may be the presence versus absence of details. Furthermore, the finding that the list of associations generated in Experiment 4 was largely different from the list of valid interpretations for the ducWrabbit figure provides additional evidence of the structural nature of the reversals and reconstruals in Experiments 1 and 2. Subjects in Experiment 4 were not asked to list other items resembled by their mental images, but rather, to list other things they thought of while they observed their mental image. Experiment 4 yielded an additional interesting finding: A majority of the associations generated when mental images of objects served as seeds for the association process were different from those generated when the name of the object served as the seed. Beyond their obvious implications for theories of mental imagery and shape recognition, the results of the four experiments discussed here have implications for theories of creativity.

The Ambiguity of Mental Images

181

Implications for Creativity

An idea is called creative when it represents a novel solution to a problem, regardless of whether the problem itself is ordinary or extraordinary. The results of the experiments reported here suggest that employing visual images in problem solving situations might produce solutions quite different from those produced by simply thinking about the problem, using propositional memory structures (Intons-Peterson, this volume; Shepard, 1978). In particular, it is now clear that images can be reconstrued, and that the reconstruals entail a structural reinterpretation of the form of the mental image. They are not simply semantic associations indexed via a conceptual route; the structural reconstruals of mental images of the idiosyncratic Jastrow duckhabbit figure were different from the semantic associates generated to the words duck and rabbit and to mental images of ducks and rabbits generated from long-term shape memory. The ability to reconstrue the structure of a mental image may provide an avenue for creative solutions. One possibility is that structural reconstruals may be the source of problem solutions which might otherwise appear to be "remote analogies" (for discussion, see Weisberg, 1988). Consider, for example, Kekulk's description of how he discovered the ring structure of benzene: My mind's eye, sharpened by repeated visions of similar art, distinguished now greater structures of manifold form: long rows, sometimes more closely fitted together, all twining and turning in snake-like motion [italics mine]. But look! What was that? One of the snakes had seized hold of its own tail, and the whole form whirled mockingly before my eyes. (Rothenberg, 1979, p. 396) In propositional terms, snakes may be considered only remotely analogous to atoms, but Kekule's description clearly states that the structure of the mental image was momentarily "snake-like." Thus, the structure of the mental image may have been reconstrued as a snake, which in turn, facilitated KekulB's discovery. This is a clear example of how structure-dependent reinterpretations can propel

182

Mary A. Peterson

thought in new and creative directions. Furthermore, the finding that the associations generated in response to mental image cues were largely different from those generated in response to the names of the same objects suggests another route through which mental images can serve as the source of ideas that are extraordinary, when ordinary is defined on the basis of associations generated to the lexical items. Note that the subjects in the experiments reported here did not possess exceptional powers of imagery, unlike those Shepard (1978) has discussed. Therefore, these experiments suggest that novel solutions can be produced by ordinary people who attend to their mental images in the course of trying to solve a problem. I do not mean to imply that a creative approach t o problem solving necessarily involves the use of mental images. My claim is only that adding mental imagery to one’s repertoire of problem solving strategies might increase the likelihood of finding creative solutions. This proposal is consistent with the view that one’s potential for creativity increases with one’s knowledge relevant t o the problem domain (Weisberg, 1988). It has been shown that it is not sufficient for the problem solver simply t o possess the relevant knowledge, however; it’s relevance t o the problem at hand must be apparent as well (for discussion see Weisberg, 1988). Similarly, Experiment 4 implies that knowledge associatively encoded with a structural representation may not be available when the corresponding lexical representation is activated; instead the structural representation must itself be activated via mental imagery in order for the knowledge associatively encoded with it to be reported in an association task. Therefore, it seems plausible that mobilizing mental imagery along with propositional knowledge may suffice t o increase the size of the knowledge base operative in a given problem solving situation; this in turn may lead to superior solutions. Increasing the number of individuals working on a problem improves the chances of reaching a superior solution (Maier, 1950). Increasing the number of mental structures employed may operate in a similar fashion.

References Biederman, I. (1987). Recognition by components: A theory of

The Ambiguity of Mental Images

183

human image understanding. Psychological Review, 94, 115-147. Binford, T. 0. (1971). Visual perception by computer. In Proceedings of the IEEE Conference on Systems and Control. Miami, FL, December. Binford, T. 0. (1981). Inferring surfaces from images. Artificial Intelligence, 17, 205-244. Bruning, J. L., and Kintz, B. L. (1977). Computational Handbook of Statistics (2nd ed.). Glenview, IL: Scott, Foresman and Company. Chambers, D., & Reisberg, D. (1985). Can mental images be ambiguous? Journal of Experimental Psychology: Human Perception and Performance, 11, 317-328. Clark, J. M., & Paivio, A. (1987). A dual coding perspective on encoding processes. In M. A. McDaniel & M. Pressley (Eds.), Imagery and related mnemonic processes: Theories, individual difference, and applications; (pp. 5-33). NY:Springer-Verlag. Finke, R. A., Pinker, S., & Farah, M. J. (1989). Reinterpretingvisual patterns in mental imagery. Cognitive Science, 13, 51-78. Fisher, G. H. (1976). Measuring ambiguity. American Journal of Psychology, 80, 541-557. Gibson, B. S. (1992). Representations of shape in memory. Unpublished doctoral dissertation. University of Arizona, Tucson. Hill, W. E. (1915). My wife and my mother-in-law. Puck, November 6, 11. Hinton, G. E. (1979a). Some demonstrations of the effects of structural descriptions in mental imagery. Cognitive Science, 3 , 231-250. Hinton, G. E. (1979b). Imagery without arrays. Behavioral and Brain Sciences, 2, 555-556. Hinton, G. E., & Parsons, L. M. (1981). Frames of reference and mental imagery. In Long & Baddely (Eds.), Attention and Performance, IX (pp. 261-277). Hillsdale, N J Erlbaum. Hoffman, D. D., & Richards, W. A. (1985). Parts of shape recognition. In S. Pinker (Ed.), Visual cognition (pp. 65-96). Cambridge, MA: MIT Press. Hochberg, J. (1968). In the mind’s eye. In R. N. Haber (Ed.), Contemporary theory and research in visual perception. NY:

184

Mary A. Peterson

Holt, Rinehart, and Winston. Hochberg, J. (1981).On cognition in perception: Perceptual coupling and unconscious inference. Cognition, 10, 127-134. Hyman, I. E., & Neisser, U. (1991). Reconstruing mental images: Problems of method. Emory Cognition Project Technical Report #19,Emory University. Jastrow, J. (1900). Fact and fable in psychology. Boston and New York: Houghton, Mifflin. Jolicoeur, P. (1985). The time to name disoriented natural objects. Memory & Cognition, 13,289-303. Jolicoeur, P. (1988). Mental rotation and the identification of disoriented objects. Canadian Journal of Psychology, 42, 461-478. Kosslyn, S . (1980). Image and mind. Cambridge, MA: Harvard University Press. Kosslyn, S., Reiser, B. J., Farah, M. J., & Fliegel, S. L. (1983). Generating visual images: Limits and relations. Journal of Experimental Psychology: General, 112,278-303. Maier, N. R. F. (1950).The quality of group decisions as influenced by the discussion leader. Human Relations, 3, 155-175. Marr, D.(1977). Analysis of occluding contour. Proceedings of the Royal Society of London B , 197,441-475. Marr, D.(1982). Vision. Ny: Freeman. Marr, D., & Nishihara, H. K. (1978).Representation and recognition of the spatial organization of three-dimensional shapes. Proceedings of the Royal Society of London B , 207, 187-217. Peterson, M. A., & Gibson, B. S. (forthcoming). Shape recognition contributions to figure-ground organization in three-dimensional displays: Necessary conditions. Peterson, M. A., Harvey, E. R., & Weidenbacher, H. (1991). Shape recognition contributions to figure-ground organization: Which route counts? Journal of Experimental Psychology: Human Perception and Performance, 17,1075-1089. Peterson, M. A., Kihlstrom, J. F., Rose, P. M., & Glisky, M. L. (1992). . Mental images can be ambiguous: Reconstruals and referenceframe reversals. Memory & Cognition, 20, 107-123. Price, J. R. (1969). Studies of reversible perspective: A methodological review. Behavior Research Methods and Instrumentation, 1, 102-106.

The Ambiguity of Mental Images

185

Reed, S. K. (1974).Structural descriptions and the limitations visual images. Memory & Cognition, 2 , 329-336. Reed, S . K.,& Johnsen, J. A. (1975). Detection of parts in patterns and images. Memory & Cognition, 2 , 569475. Reisberg, D., & Chambers, D. (1991). Neither pictures nor propositions: What can we learn from a mental image? Canadian Journal of Psychology, 45, 288-302. Rothenberg, A. (1979). The emerging goddess. Chicago: University of Chicago Press. Rubin, E. (1915/1958). Figure and ground. Reprinted in D. C. Beardslee & M. Wertheimer (Ed. & Trans.), Readings in perception (pp. 194-203). New York: Van Nostrand. Shepard, R. N. (1978).Externalization of mental images and the act of creation. In B. S. Randhawa & W. E. Coffman (Eds.), Visual learning, thinking, and communication (pp. 133-189). NY: Academic Press. Tarr, M., & Pinker, S. (1989). Mental rotation and orientationdependence in shape recognition. Cognitive Psychology, 21, 233-282. Tinbergen, N. (1948). The study of instinct. Oxford: Oxford University Press. Weisberg, R. W. (1988). Problem solving and creativity. In R. Sternberg (Ed.), The nature of creativity (pp. 148-176). Cambridge University Press.

Acknowledgments The preparation of this chapter was supported by Grant BNS909100 to M. A.Peterson from the National Science Foundation and the Air Force Office of Scientific Research. I thank John Kihlstrom for his thoughtful comments on a previous version of this chapter and Pat Rose, Martha Glisky, Deirdre Avery, and John Kihlstrom for their help in conducting these experiments.

This Page Intentionally Left Blank

Imagery. Creativity. and Discovcry: A Cogpitive Pcrspectivc B. Roskos-Ewoldson, M.J. Intons-Peterson and R.E. Anderson (Editors) 0 1993 Elsevier Science Publishers B.V. All rights reserved.

187

Chapter 7 ~

~

DISCOVERING EMERGENT PROPERTIES OF IMAGES Beverly Roskos-Ewoldsen Department of Psychology University of Alabama Tuscaloosa, AL 35486 USA This chapter serves as a bridge between the chapters concerned with image reinterpretation and those that discuss creative discoveries. It fits easily with the earlier chapters, as this chapter presents another line of evidence that images can be reinterpreted. With respect to the next set of chapters - those discussing creative discoveries - this chapter presents a basic model of creativity and discovery, discusses previous research and anecdotes in terms of the model, and describes two experiments on emergent property recognition within images to elucidate the discovery aspect of the model. I begin with anecdotal evidence of the use of imagery in creativity and discovery. Next, I defme creativity and discovery and present a basic model of the creativity and discovery processes. I address the necessary conditions for imagery's use in creativity and discovery, and describe two experiments investigating the connection between imagery and discovery. The chapter concludes by evaluating the current research in terms of imagery's use in creativity and discovery, and by suggesting future research avenues.

The Use of Imagery in Creativity and Discovery Shepard (1978,1988) and Shepard and Cooper (1982)provide anecdotal evidence that imagery can be an important part of the

188

Beverly Roskos-Ewoldsen

creativityldiscovery process. Three of their examples regarding the analogical nature of imagery, and its use in creativity and discovery, are especially thought-provoking. In one example, Shepard and Cooper (1982, pp. 5-7) tell of James Surls, a sculptor, who reports that ''he manipulated the image around in his mind. He saw it tumbling and rolling, took an arm off, put an arm on ..." (reported originally by Samuels & Samuels, 1975). A second example is of Einstein, who said that he "'very rarely' thought in words [reported by Wertheimer, 19451, and that his particular ability did not lie in mathematical calculation either, but in 'visualizing ...effects, consequences, and possibilities"' (reported by Holton, 1972). Einstein indicated also that "for him this 'visualizing' consisted primarily of 'more or less clear images which can be "voluntarily" reproduced and combined"' (reported by Einstein, in Hadamard, 1945). The final example is provided by one of the authors (i.e., Shepard), who states that "just before 6:OO am on [November 16,19681 and in the absence of any noticed precursors, [I] experienced a spontaneous kinetic image of three-dimensional structures majestically turning in space" (Shepard & Cooper, 1982, p. 7). Most of the readers of this chapter will recognize this example as a precursor to Shepard's famed mental rotation studies. These examples were chosen because each has components that highlight ideas introduced and discussed in this chapter. Most important, all involved dynamic processes. Sometimes there was a stable overall structure that was preserved as the structure was mentally rotated (Shepard; some of Einstein's reports). Sometimes the example included a distinction between parts and wholes (Surls; Einstein). In the reports that included references to parts, the parts were being reorganized, or "played with." Also, these introspections suggest the idea that consequences and possibilities of a particular organization were realized. In the following section, I try to develop these ideas about dynamic processes, parts, wholes, and consequences into a model of possible stages of creativity and discovery. Later, I present two experiments that investigate properties of the discovery process.

Discovering Emergent Properties of Images

I89

Defining Creativity and Discovery A good starting point for defining terms is the dictionary. Webster’s New Universal Unabridged Dictionary defines creativity as “artistic or intellectual inventiveness.”The term create is defined as “to bring into being from nothing, cause to exist; to produce, bring about, give rise to.” Compare the definition of creativity to that of discovery. Webster‘s defines discovery as “a discovering; the unraveling of a plot in a poem, play, etc.,” and discover as “to find out, learn existence of, realize; to reveal, disclose, expose, uncover.” Basically, creativity involves production or invention and discovery involves revelation, realization, or uncovering. Contrast these definitions with current thoughts on creativity and discovery, especially those of cognitive psychologists. Finke (1990),for example, says that creativity is the pathway to a solution: I’ , . . creativity should not be thought of in terms of the specific products of the creative act, but rather, as the way one engages in creative exploration” (p. 168). He continues by saying, “In this sense, creativity is using the things we create, not creating the things we use” (p. 168). Finke alludes t o a definition of discovery by stating that, “Before one can make creative discoveries in imagery, it must be possible to recognize meaningful shapes or patterns that ‘emerge‘ when images are formed” (p. 7). Further, he speaks of invention as a product of ”imagining interesting combinations of parts, and then ‘recognizing’ useful applications of the resulting imagined forms” (p. 39). Finally, Finke discusses in depth how one should judge whether a particular discovery or invention is creative (i,e,,judging its practicality and originality), and he makes a distinction between creative combinations of the given parts and creative interpretations of the pattern. Based on these passages, it is clear that Finke is describing creativity and discovery at a different level than the dictionary meanings of the words. According to my reading, Finke uses “creativity” and “creative” to refer to aspects of the creative act, of which discovery is a part. Specifically, Finke’s writings emphasize what he considers to be the creative process (i.e., generation and exploration), and the criteria used t o judge creativity (i.e., practicality, originality). Anderson and Helstrup (this volume) also write of creativity and discovery in this way, although they use the

190

Beverly Roskos-Ewoldsen

terms construction and interpretation to describe aspects of the discovery process (see also Helstrup & Anderson, 1991). Both Anderson and Helstrup, and Finke contend that creativity can occur during either the constructiodgeneration or the interpretation/ exploration portions of discovery. Presumably, the presence of creativity is implicated by the presence of an invention (combination, interpretation) that is judged to be creative. Though these descriptions of creativity and discovery processes are intuitive, I propose that we take a more narrow view of these two processes, a view that reflects the dictionary definitions. Namely, I suggest that the term “creativity” be restricted to the cognitive processes involved with invention, in the sense of bringing into being, producing, or combining, and “discovery,”to the processes involved with uncovering or interpreting. Although dictionary and psychological definitions do not always coincide, in this instance the dictionary definitions provide a useful framework for studying the cognitive processes involved in creativity and discovery. Accepting the current definitions of creativity and discovery as combination and interpretation, respectively, requires aspects of the “discovery” tasks used by Anderson and Helstrup (this volume; Helstrup & Anderson, 19911, Finke (this volume; 1990; Finke & Slayton, 1988),and Intons-Peterson (this volume) to be reformulated with the new definitions in mind. For example, in the “discovery” task participants were provided with three or four alphanumeric characters or simple geometric shapes, and they were asked to assemble the parts to form a recognizable figure or useful object. The key assumption has been that participants are discovering figures by generating, constructing, and combining the parts in various ways, and searching or exploring the combination for a recognizable figure or object). My point is that the combining aspect of the task involved creativity, and discovery occurred only after the parts were combined. Thus, according to my view, the “discovery” task should not be labeled as such; rather, the task should be labeled as a “creativity and discovery” task because it involves the cognitive processes involved in both creativity (combinational play) and discovery (recognition or interpretation). Differences in describing tasks arise in part from our acceptance of the “science as discovery” metaphor. Roald Hoffman, a Nobel laureate and professor of chemistry, addresses this issue: “In

Discovering Emergent Properties of Images

191

describing what they do, scientists have by and large bought the metaphor of discovery, and artists that of creation. The cliche ‘uncovering the secrets of nature’ has set, like good cement, in our minds” (Hoffman, 1990, p. 14). In his field of chemistry, Hoffman claims that the discovery metaphor is inadequate because chemists create molecules that do not exist naturally. Further, Hoffman states that, “The synthesis of molecules puts chemistry very close to the arts. We create the objects that we or others then study or appreciate. That’s exactly what writers, composers, visual artists . . . do“ (p. 15, italics added). Finally, Hoffman puts the building of theories not in the domain of discovery, but in the domain of creativity: “In the building of theories and hypotheses, even more than in synthesis, the act is a creative one. One has t o imagine, to conjure up a model that fits often irregular observations. There are rules ....There are hints of what to do....But what one seeks is an explanation that was not there before, a connection between two worlds” (p. 15). When we consider creativity in this light, we see that creativity involves the building of something that did not exist before. I argue that the combination aspect of the “discovery”task, discussed above, involved putting the characters together in ways that were unusual, or in new ways, and therefore should be considered as the creativity aspect of the task. Imagining or drawing a “C,” a circle, and a V“ as a double-scoop ice cream cone certainly involved unusual combinations and orientations of the circle and the letters “V“ and “C,”regardless of the final recognition of the ice cream cone (which, by the way, was not judged to be creative). In contrast, the discovery aspect of this task consists of the recognition or interpretation of a particular combination of the parts as something familiar or useful, regardless of how the particular combination came to be. Support for the Current Definitions

Support for the definitions of creativity as combining and discovery as interpretation comes from three sources. First, I call attention to the distinctions of Finke (this volume) and Anderson and Helstrup (this volume; Helstrup & Anderson, 1991). Finke distinguishes generation from exploration, and Anderson and Helstrup differentiate construction and interpretation. In both cases, the authors argue that generatiodconstruction and exploration/

192

Beverly Roskos-Ewoldsen

interpretation require or rely upon different processes, In fact, when Finke and Slayton (1988)asked their participants to identify their strategies for combining shapes into recognizable forms, nearly 75% of the participants indicated two separate phases, a combinational one in which they combined the parts to form a figure and an interpretation phase during which they considered possible interpretations of the figure. Finke (1990)pointed out that these reports correspond to his subjective impressions of how to complete the task efficiently: ‘I . . . the most efficient strategy is to imagine combining the parts in various interesting ways and then mentally ‘seeing’ if anything meaningful emerges - what Einstein had referred to as ‘combinational play’ in his own thinking. . . . This ‘combinational play’ strategy is also remarkably successful when trying to discover creative inventions” (p. 27). In my opinion, it behooves us to take advantage of these natural distinctions when we try to define creativity as separate from discovery. Finke (this volume) currently describes generation and exploration as part of the creative process. Helstrup and Anderson (1991;Anderson & Helstrup, this volume) consider construction and interpretation as part of the visual discovery process. However, equating creativity as constructing (generating, combining) and discovery as interpreting (exploring)affords the study of creativity per say, rather than (or in addition to) studying the creativeness of a person, object, or interpretation. A second area of support, related t o the first one, is the ease with which “preinventive forms” can be constructed. Finke (this volume; 1990,Chapter 6)defines preinventive forms as the products of combinational play. These forms are shapes and structures that seem potentially useful, in a general sense. Preinventive forms may later be interpreted as particular objects, according to the dictates of the task. For example, subjects may, upon presentation of a set of parts, generate a specific form without knowing how the form will eventually be used. Once the parts are combined into a preinventive form, the subjects can then interpret the preinventive form as some kind of invention. Finke’s subjects were able to produce a preinventive form on nearly 100% of the trials. Compare this figure with the subjects’ ability to interpret these forms as practical inventions. When subjects generated their own preinventive forms they were able to interpret them as practical inventions on 33%

Discovering Emergent Properties of Images

193

(120/360) of the trials. If subjects were provided with a form, they identified a practical invention on 19% (69/360) of the trials. The discrepancy between generating (nearly 100%) and interpreting (1943%)suggests that different types of mental processing occur during the combinational play and the interpretation of the form. A third and final area I discuss in support of my distinction comes from Finke and Slayton’s data (Finke & Slayton, 1988) specifically, from subjects’ failure to predict a resulting pattern based on the parts alone. Let me begin this argument with the premise that discovery involves either interpreting an existing form or uncovering properties of an existing form. If one accepts this premise, then one must conclude that subjects should have been able to discover the resulting figure from its constituent parts. The finding that subjects were not able to predict the figure from the parts suggests that something other than discovery must have been occurring. Finke calls this other process “combinational play”; I contend that combinational play is creativity.

Implications of the Current Definitions Defining creativity as a combinational process and discovery as an interpretation process provides a new perspective on their understanding; namely that they are two different mental processes. In most tasks, disentangling creativity from discovery is difficult because the two processes interact during the completion of the tasks. In this respect, I agree with Helstrup and Anderson (1991; Anderson & Helstrup, this volume) when they say that the processes of construction and interpretation may in principle be independent and sequential, but, more likely, they interact such that a particular construction may lead to an interpretation, which may lead to another construction, and so on. Nevertheless, I believe that creativity can be studied separately from discovery. Finke’s (1990; this volume) successful separation of combinational play and interpretation indicates that dissociation is possible. Acceptance of these definitions carries with it an implicit assumption that there are real, observable differences between the creativity and discovery processes. It becomes our challenge to figure out what these differences are, if they do exist. How do “cognitive” variables, such as perceptual or semantic characteristics of the parts,

194

Beverly Roskos-Ewoldsen

or other variables, like personality or intelligence-related characteristics, influence creativity (isen, combinational play)? Which variables influence discovery (ie., interpretation)? Actually, we know a fair amount about the discovery (interpretation) process (see chapters in this volume by Chambers, Hyman, Kaufmann, & Helstrup, and Peterson). We know somewhat less about the creativity process, as I have defined it (see chapters in this volume by Anderson & Helstrup, Finke, Intons-Peterson, and Reed; see also Sternberg, 1988). Throughout the remainder of this chapter, and especially in the conclusion section, I address these factors.

A Model of Creativity and Discovery Cognitive Components of Creativity and Discovery

Consider again the examples of Surls, Einstein, and Shepard. These anecdotes are interesting and thought-provoking, but they are self-reports, and provide no more than a hint of the cognitive processes involved in creativity and discovery. Fortunately, others have addressed the possible differences in cognitive processes that occur during the construction and interpretation phases. Helstrup and Anderson (1991; Anderson & Helstrup, this volume) discuss construction and interpretation in terms of automatic and controlled processes. Construction is considered to be a controlled process. When a configuration is assembled, one may manipulate the elements or configuration in effortful ways (e.g., zoom in, rotate, insert, place elements side-by-side). Helstrup and Anderson argue that more automatic, gestalt-like figure formation processes occur at the same time as the controlled processes of construction. These automatic processes presumably contribute independently to the construction of a configural unit. The interpretation processes, according t o Helstrup and Anderson, are considered as partly controlled and partly automatic. They argue that a controlled deliberate feature analysis of the emerging figure may involve or give rise to a search for meaningful fragments, or a “free association” of interpretations to the figure or its elements. Presumably the free association involves some automatic processes. Finke (this volume; Finke, Ward, & Smith, in press) proposes

Discovering Emergent Properties of Images

195

the "Geneplore" model of generation and exploration - processes similar to Helstrup and Anderson's construction and interpretation. During the first, generative phase, mental representations (preinventive structures) are constructed. Properties of the preinventive structures, such as novelty, ambiguity, and the emergence of unexpected features, are explored during the exploratory phase to interpret the structures in meaningful ways. The model in Figure 1 incorporates the writings Finke et al., and Anderson and Helstrup. Before discussing the processes of creativity and discovery, one must first recognize that creativity and discovery require information. Information, labelled "data" in Figure 1, can be many different things - lines, patterns, ideas, methods, paradigms, theories, and so on. How one decides which information to focus on is an interesting question, but will not be discussed extensively here. Instead, the subsequent processing of the chosen data is considered. The first step in the creativity/discovery process is the creativity phase, and it involves combining the chosen data in new ways. During this stage, a creator or problem solver is combining information that has never been or is not typically put together in this way. Here, mental manipulations and transformations of the data occur; this creativity stage is dynamic and generative, involving not only the data but also knowledge of existing, potential, and impossible relations among the data. One may contend that seeing data in new, creative ways without having to combine information from different sources is part of the creativity stage. I propose that seeing data in creative ways, or more accurately, perceiving previously unnoticed information within existing knowledge structures (or perceptual structures), is more of a discovery or interpretive process than a creativity process. That is, the structure is already present, and the task of discovery involves an exploration of the existing structure for emergent properties, including new or unusual interpretations or reconstruals of the material. In a perceptual sense, all or part of the figure may become ground and ground may become figure. The discovery process, then, involves an interpretive search for emergent properties of the particular combination of information. If a solution is found during this discovery stage, the answer becomes the output of the process. If no satisfactory solution is

196

Beverly Roskos-Ewoldsen

7 Creativity

Combinational play; Manipulation

Discovery Interpretation; Search for emergent properties

i4 Solution

Figure 1 A model of creativity and discovery. The model represents a task that involves both creativity and discovery processes.

Discovering Emergent Properties of Images

197

discovered, the process continues by looping in one of two ways. The first is to start again with the data and organize them anew, thus returning to the creativity stage. The second loop involves staying at the discovery stage, but continuing the search for emergent relations. The model of creativity and discovery, as I have outlined it, is geared toward the visual-spatial realm. However, the model can be generalized easily to other realms, such as music, math, or language. As I mentioned earlier, the data can be many things and can include words, formulas, or musical notes. Creativity involves combinational play, and discovery, an interpretation of a particular combination. At this point, caveats are in order. First, I do not mean to say that putting together (i.e., combining) information always implies creativity. If one were instructed how to construct a figure from the data, I would not say that creativity had occurred. On my view, creativity requires combinational play. Second, as I have mentioned before, judging whether something is creative or not is a different matter than creativity. Judging the creativeness of a solution involves some perception of unusualness, cleverness, and for "inventions," practicality (see Finke, this volume). These judgments provide an indication of which organizations are more "fruitful" in terms of discovery, and of which discoveries, given a particular organization, are the more unusual, interesting, and (sometimes) practical. A third and final caveat involves the use of imagery in creativity and discovery. There is no mention of imagery in this model of creativity and discovery. The reason for this omission is that imagery is not necessary for creating and discovering. Rather, the use of imagery is a special circumstance of the creativity and discovery processes. Exploring the conditions in which imagery can be used in creativity and discovery requires an understanding of the critical components of the creativity and discovery processes, as discussed above, and an investigation of the properties of images that allow imagery's use in the processes of creativity and discovery.

Necessary Conditions for Imageiy's Use in Creativity and Discovery According to the model o f creativity and discovery outlined above, two necessary conditions must be met for imagery t o be used

198

Beverly Roskos-Ewoldsen

in the creativity/discovery process. First, one must be able to create in the mind’s eye- to manipulate and transform information imaginally, or “[take] an arm off, put an arm on,” in the words of the sculptor Surls. In other words, imaginal creativity involves the ability to rotate, alter size, shift position or orientation, or otherwise manipulate information imaginally. Much work has shown that visual and auditory information, both internally and externally generated, can be imaginally manipulated (e.g., Finke, 1989, 1990; Kosslyn, 1980; Shepard & Cooper, 19821, though arguments remain concerning the nature of the underlying representation of mental images and mental transformations of information (e.g., Block, 1981; Tye, 1991). Recent experiments reported by Finke (this volume; 1990; Finke, Pinker, & Farah, 1989; Finke & Slayton, 19881, Anderson and Helstrup (this volume; Helstrup & Anderson, 1991), and Intons-Peterson (this volume) show that subjects can creatively combine familiar alphanumeric characters and simple shapes characters that are not normally thought of in combination with one another - into known shapes. A second necessary condition is that one must be able t o discover within the realm of the mind‘s eye, to interpret, reinterpret, or recognize properties of imaginal information. Finke (this volume; 1990) has shown that subjects can discover practical uses for line drawings of unfamiliar objects, especially if the line drawings were created by the subjects themselves. However, whether an image can be reinterpreted to discover properties or alternative meanings is controversial. The answer seems to depend partly on the definition of the term “reinterpret.” This term can refer to discovering relations that are not explicitly encoded in the parts used to construct an image, or it can refer to the semantic interpretation of an image, and the ability to reinterpret the image. The latter position is adopted by Reisberg and Chambers (e.g., Chambers & Reisberg, 1985; Reisberg, Smith, Baxter, & Sonenshine, 1989; Chambers, this volume; Reisberg & Logie, this volume). Their argument is that an image is inherently unambiguous, just as a percept is unambiguous. In their studies, perceptually ambiguous stimuli were used to show that an image cannot be reconstrued. Subjects either created a visual image of an ambiguous figure and attempted t o %eelt a second interpretation of the figure (Chambers & Reisberg, 1985) or they imagined hearing a word repeated

Discovering Emergent Properties of Images

199

continuously until they detected a reorganization of the word (e.g., hearing "fly" from the repetition of "life," Reisberg et al., 1989). Participants could reconstrue neither the visual image, nor the "pure" auditory image above chance. On the other hand, if ''reinterpret" can refer t o perceiving information not explicitly encoded in the parts of an image, then images can be reinterpreted (e.g., Finke & Slayton, 1988; RoskosEwoldsen, 1989). In these experiments, subjects either imaginally constructed parts or were shown parts, and imaginally synthesized the parts to form a pattern. Subjects then reported interpretations of the imaged pattern (Finke & Slayton) or responded whether they could see new parts within the pattern (Roskos-Ewoldsen). In either case, subjects were recognizing properties of the pattern that were not encoded in the original parts of the pattern. Needless to say, questions remain regarding this controversy. On the surface, experiments investigating the semantic reinterpretability of an image appear t o differ in many respects from those that examine the detection of properties of constructed patterns. Chapters of this book authored by Chambers, Hyman, Kaufmann and Helstrup, Peterson, and Reisberg and Logie address some of these questions. On my view, the experiments are directly related because they involve discovery -of alternative meanings and of emergent properties. I now describe in more detail recent empirical work that establishes the use of imagery in creativity and discovery and illustrates the proposed model. Keep in mind that these experiments have been discussed by their authors primarily in terms of discovery.

Imagery, Discovery, and the Recognition of Emergent Properties of Patterns Imaginally Manipulating Known Shapes In one set of experiments (Finke & Slayton, 1988) participants received three familiar alphanumeric characters or simple geometric shapes (e.g., D, 8, V, circle, rectangle); they were instructed to close their eyes and mentally manipulate and assemble the parts into a recognizable, easily named figure. The parts could be increased or

200

Beverly Roskos-Ewoldsen

decreased in size and moved to new locations and orientations, but their shapes could not be altered (i,e,, bent). Overall, recognizable patterns were reported on 38.1% and 44.3% of the trials in Experiments 1 and 2, respectively, and the vast majority of subjects were able to report two or more recognizable patterns (of eight trials). The products provide convincing evidence of our ability to imaginally play with parts until we can recognize a figure from a particular combination. My favorite example is of a croquet scene which was made from the set (T, P, 8). The T, rotated 150 degrees to the right, was the mallet. The P, reversed, enlarged, and rotated 90 degrees to the right, formed the ground and the hoop. The 8, rotated 90 degrees, form two balls on the ground. The point of introducing this study here is to show that subjects can combinationally play with given parts. In this case they imaginally played with the parts (see also Anderson & Helstrup, this volume; Helstrup & Anderson, 1991; Intons-Peterson, this volume). In their minds' eyes, these participants were organizing, arranging, putting together and rearranging the data (i.e., characters). This activity I have labelled as the creativity process. The assessment of whether a particular combination forms a recognizable shape is what I refer to as the discovery process. Thus, students in this study not only manipulated the parts, but also searched for a recognizable interpretation of a particular grouping.

Reinterpreting Imaginally Constructed Patterns Another set of experiments (Finke et al., 1989) focused on the discovery of recognizable lkonsequencesll of guided imaginal synthesis. Participants were asked to superimpose one familiar character (e.g.,letter, number, or simple geometric shape) on another and describe any new features or patterns they could detect from the mental superimposition. They were told to report as many emergent features as they could discover. As an example, a subject might have been asked to superimpose a capital letter H onto a capitol letter X. The subject might describe a butterfly, a tilted hourglass, a triangle, or the capital letter M. The M and the triangle exemplify an interesting phenomenon. Each requires a structural, perceptual reinterpretation of the individual letters. To recognize an M, one has to reinterpret the X

Discovering Emergent Properties of Images

201

and H. Specifically, the X and H need to be parsed into smaller elements and then reinterpreted structurally to form a new figure and background -namely, the M and everything else (see Figure 2). This ability to recognize new overall forms and figure-ground configurations is part of the discovery process, I argue, since it involves the identification or interpretation of a particular configuration within an existing figure.

H+X=BB Figure 2 An illustration of the reinterpretation required for an emergent property to be discovered.

Viewing Finke's work in light of the proposed model highlights the need to reexamine experiments that reportedly investigate the processes involved in discovery. The model also provides guidelines for designing research on creativity and discovery. In particular, there are two ways to study creativity and discovery. One is to dissociate creativity from discovery processes by introducing variables that influence one process and not the other. For example, one might vary factors that, are related to creativity (i.e., combinationalplay or manipulation) and not related to discovery (i.e., detecting alternative meanings or searching for parts within

202

Beverly Roskos-Ewoldsen

patterns), or vice versa. The other way to examine the processes of creativity and discovery is t o control or eliminate one while focusing on the other. This second strategy is what I have chosen to use in my work. Below, I present two experiments that control the creativity process by guiding the participants in the construction of a pattern; my focus, therefore, is the discovery process. In particular, I investigate how easily one can discover a part within a whole pattern. I test discovery by presenting possible parts of patterns which have been constructed imaginally by the participant. The response is a simple “yes” or “no.” On the surface this procedure may not fully capture the discovery process, as the participant is not free to discover any and all parts or properties that emerge from the construction. However there are three main arguments for the use of this procedure. First, the test parts represent forms that could have been discovered by the participants. To the extent that subjects recognize the emergent forms, we can conclude which forms were more likely than others to have been discovered. Second, the patterns and parts used in the following experiments were unfamiliar to the participants (see Figure 3), which made the procedures used by others (e.g., Finke & Slayton, 1988;Helstrup & Anderson, 1991; Intons-Peterson, this volume) inappropriate to adopt. In the previous studies, participants described the resulting or emergent forms; here, description was difficult because of the abstract, unfamiliar stimuli. Presenting possible parts circumvented this problem. Third, and finally, with the present procedures I can vary the way the pattern is constructed, as well as aspects of the test part, and perceptual and organizational properties of the pattern to be constructed. The type of test part and the perceptual organization of the pattern are the main variables incorporated in the experiments that follow.

Discovering Emergent Properties of Imagined Patterns The perceptual organization of parts and wholes. Discovering emergent properties of a constructed form involves, in the present task, the recognition of parts within wholes. In previous experiments investigating the recognition of parts within wholes, the figure is usually presented as a whole, and the intent of the study is to measure how the goodness of a part influences its recognition; the

Discovering Emergent Properties of Images

203

goodness of the part refers to how “good”the part is in relation to the whole. For example, some researchers find that good parts of a pattern are more easily detected (i.e., discovered) than poor parts of the pattern, as measured by recognition accuracy or response latency (Kolinsky, Morais, Content, & Cary, 1987;Palmer, 1977;Reed, 1974; Reed & Brown, 1979;Reed & Johnsen, 1975). From my perspective, two questions arise from these experiments: (a)what happens with imaginal discovery when the pattern is constructed from two or more parts, and not presented as a whole; and (b) is imaginal discovery influenced by perceptual characteristics such as the goodness of a part, regardless of its relation to the whole, and the goodness of the pattern? These questions are partially answered by research finding that the organization of an imagined whole- that is, the way we perceive the imagined whole - influences what is easy t o “see” within the imagined whole (Hinton, 1979;Kosslyn, 1980;Reed, 1974; Reed & Johnsen, 1975;Thompson & Klatzky, 1978). All argue that parts are more accurately and quickly identified when they are congruent with the perceived organization of the whole, than when they are not. Thompson and Klatzky provide the most direct test of this argument. Thompson and Klatzky’s (1978, Experiment 3) research participants were presented with parts to be mentally synthesized into a whole, and subsequently tested with either the whole, an “old” part (Lee, one that matches a part to be synthesized), or an “emergent” part (i.e., a part that emerges from the synthesis). Response latencies for yes/no judgments showed that whole forms were recognized faster than old parts, which were recognized faster than emergent parts. We can conclude from these results that when the whole is constructed from two or more parts, discovery (i.e., recognition of emergent parts) is more difficult than simply recognizing a previously seen part. My questions still remain, however. What happens with imaginal discovery when both the parts and the whole are imagined? Furthermore, how does the perceptual organization of the parts and the whole influence imaginal discovery, if at all? To address these questions, I report two recently completed experiments. They were designed to investigate the influence of perceptual organization on discovery, or more specifically, on the recognition of emergent

204

Beverly Roskos-Ewoldsen

properties of imagined and seen patterns.

Two experiments. The first experiment addressed two related questions: (a) can emergent properties be discovered within imaginally constructed patterns; and (b) how easily does imaginal discovery occur, compared to the recognition of imagined parts used to construct the pattern? Upon establishing in the first experiment that imaginal discovery can occur, the factors that might influence imaginal discovery could be investigated in a second experiment. Perceptual organization was chosen as a possible influence on discovery. The perceptual organization of a form can refer to many aspects of the form; here, it is meant as the “goodness” of a form, where “good” forms tend to be better organized than “poor” forms (e.g., Attneave, 1955; Garner & Clement, 1963; Hochberg & McAlister, 1953;Peterson, Rawlings & Cohen, 1977). “Goodness” was incorporated into the design of the second experiment in two ways: goodness of the parts and goodness of the patterns. The methodology for the experiments was adapted from previous experiments, some using familiar parts (Finke et al., 1989;Finke & Slayton, 1988)and others using unfamiliar parts that were presented to the subjects to be mentally synthesized (Intons-Peterson, 1981, 1984;Klatzky & Thompson, 1975; Palmer, 1977;Reed & Brown, 1979;Thompson & Klatzky, 1978). In the present experiments there were two types of construction, imaginal and perceptual. Participants in the imagery condition saw neither the parts nor the pattern; rather, they imagined both according to the experimenter’s instructions. Perception participants saw both the parts and the patterns because they drew both, with the experimenter’s guidance. Before the experiment began, subjects memorized a set of 16 lines in various locations and orientations (Figure 3); each line had a number associated with it. The experimenter guided construction by calling out numbers and the subjects imagined (drew) the lines that corresponded to the numbers. Patterns were constructed by first imagining or drawing three lines to form half of the pattern (a 3 line part), then imagining or drawing three more lines to form the other half (a second 3 line part), Finally, they imagined or drew both parts together to form the overall pattern. After the pattern was constructed, subjects were tested with possible parts of the pattern. “Old”test probes matched

Discovering Emergent Properties of Images

205

either the first or second 3 line part of the pattern. “Emergent” probes matched neither the first nor the second part of the pattern; rather, they emerged from combining the two parts. “Noncomponent” probes were not a part of the pattern and were included to avoid response bias. Performance on emergent test probe trials is of primary importance because it reflects the discovery process. A typical trial proceeded as follows. The subjects formed the first 3 line part of the pattern, either by imagining the part within a frame of nine dots or by drawing it within a similar frame, and were instructed to remember the shape of the part. Then they constructed the second 3 line part, again, either imaginally or by drawing it within a frame. Once the second part was formed, the subjects recalled the first part from memory and combined it with the second part, either imaginally or by drawing the first part onto the frame containing the second part, to form the entire pattern. At this point the experimenter presented one of three test probes (old, emergent, noncomponent). Test probes were placed by the experimenter immediately below the frame containing the imagined or drawn pattern. As the test probe was presented the experimenter triggered a timing device. Timing was terminated when the subjects pressed the key labelled “yes” if they thought the test probe was within the pattern they constructed, or the “no”key if their decision was negative. Both accuracy and response latency were recorded, using the timing device, because it was unclear whether the manipulated variables would affect one, the other, or both. After a short delay, the next trial began and a new pattern was constructed and tested. In each experiment, patterns were chosen from a population of 50 patterns. Each pattern in the population could be parsed in at least two ways; each parsing consisted of two three-connected-line parts. A separate group of subjects rated the population of patterns and their three-line parts on three dimensions: ‘‘goodness”of the patterns; “goodness” of parts within their patterns; and ”distinctiveness” of the patterns. In another set of tasks completed by these same subjects, the three-line parts were treated as whole patterns and were rated for their “goodness” and “distinctiveness” as three-line “patterns.” Another group of participants was provided with the patterns and were instructed to divide the patterns into

Beverly Roskos-Ewoldsen

206

Pattern

Part Set B

Part Set A

B1

A1

A2

LI:

* . *

82

. . * *A Test Probes Old

LI: @ * *

Emergent

\

Noncomponent

:E

a

Figure 3 Original lines used to construct patterns; sample pattern and parts; sample test probes. The test probes labelled " o l d and "emergent"are for Part Set A only.

Discovering Emergent Properties of Images

207

parts as they saw fit. Of these ratings, the most relevant for the present experiments are the “goodness” of the patterns and the “goodness” of the threeline parts (treated as whole “patterns”). Stimuli in Experiment 1 were chosen to ensure that the patterns and their parts were of approximately equal goodness; eight patterns were selected based on this criterion. Perceptual organization, or “goodness”was controlled rather than manipulated in the first experiment because one must know if and how easily imaginal discovery occurs before one manipulates characteristics that may influence imaginal discovery. Experiment 2 stimuli were chosen to ensure a difference between good and poor patterns, and good and poor parts. The resulting stimuli were eight patterns rated as “good”and eight patterns rated as “poor,” In addition, each good and poor pattern had two parsings - one parsing contained “good” parts and the other contained “poor” parts. The results from the first experiment were straightforward. As a reminder, Experiment 1 was conducted to verifj7 that emergent properties can be discovered within imaginally constructed patterns, and t o compare the difficulty of discovery t o the recognition of the imagined parts. Figure 4 shows that emergent properties of imagined patterns were recognized on 70%of the trials, compared t o old parts, which were recognized on 83% of the trials. Thus, imaginal discovery can occur, although it is more difficult than recognizing parts used during construction. Response latencies corroborate this conclusion, as emergent probe latencies were 1.5 times longer than those for old probes (Figure 5). Discovery within drawn patterns tells a different story. Emergent probes were recognized as easily as old probes, in terms of both accuracy and response latency. Perceptual discovery appears t o be relatively easy, at least as measured in this experiment. The issue that guided the design of Experiment 2 was whether imaginal discovery was influenced by the perceptual organization of the stimuli. The goodness of the pattern and the goodness of the parts were between-subjects variables. All three types of test probes (old, emergent, noncomponent) were used, but the discussion presented here focuses on emergent probe recognition because we are most interested in the influence of “goodness”on discovery. Before I present the results of the second experiment, let me

208

Beverly Roskos-Ewoldsen

Figure 4 Experiment 1: Percent correct recognition as a function of construction group and type of test probe.

Discovering Emergent Properties of Images

209

Figure 6 Experiment 1: Response latency for correct responses as a function of construction group and type of test probe.

210

Beverly Roskos-Ewoldsen

digress and speak of the expected effects of pattern goodness and part goodness. Notice in Figure 4 that there is an overall performance difference between imagery and perception, with imagery being more difficult. If we accept a limited capacity cognitive system, the difference is understandable. Imaginally constructing parts and patterns requires more processing capacity because the construction is completely internally generated. Drawing necessarily provides an external aid and therefore requires less processing capacity. If we are using a task that taxes a limited capacity system, any savings in the construction of the parts and pattern may result in more processing capacity available for discovery. Palmer (1977) showed that mentally synthesizing good parts of a pattern took less time than synthesizing poor parts of the same pattern. In addition, Attneave (1955) and Hochberg and McAlister (1953) have argued that good figures tend to be compact or contain redundant information, and may therefore be easier to maintain than poor figures. Consequently, goodness of parts may reduce the processing load required to imaginally construct the parts and resulting pattern, leaving more processing capacity, within a limited-capacity system, to discover an emergent part. Thus, discovery should be easier when patterns are constructed with good parts, rather than poor parts. This should occur more so for imaginal patterns than for drawn patterns. In contrast, pattern goodness may have a different effect on discovery. Set aside for now the construction of a pattern and focus on the constructed pattern. A good pattern tends t o be internally coherent (see Garner & Clement, 1963; Peterson et al., 1977). By this I mean that the parts of a good pattern cohere, or "chunk" and consequently lose their separateness. They no longer are perceived as two parts; instead, they are perceived together as the whole pattern. Once this coherence occurs, discovering emergent parts of the good (coherent) pattern may be more difficult than discovering emergent parts of the poor patterns. Poor patterns tend to be less coherent and may afford discovery to a greater extent than good patterns. Thus, discovery should be more difficult when the overall pattern is good, relative to when the pattern is poor. Furthermore, pattern goodness may affect discovery more so with drawn patterns than with imaginal patterns, for the simple reason that drawn patterns are more defined than imaginal patterns.

Discovering Emergent Properties of Images

211

I mention again that I am primarily interested in the effects of perceptual organization on discovery. Therefore, I present accuracy and response latency results for emergent test probes only. Because there were no higher-order interactions of pattern and component goodness, the results for pattern goodness and part goodness are shown separately (Figures 6 and 7). Beginning with the effects of part goodness, Figure 6 shows that, for the imagery group, recognition of emergent properties was more accurate when patterns were formed with good parts than when they were formed with poor parts. In the perception group, recognition accuracy for emergent properties did not differ for good and poor parts. Effects of part goodness were not revealed in response latencies for either the imagery or the perception groups (Figure 7). The accuracy finding, however, is in agreement with predictions based on the extent to which a limited-capacity processing system is being taxed. Imaginally constructing parts and patterns is more taxing than drawing them. Therefore, with the imagery group, good parts may have served to conserve processing capacity, allowing more processing to be available for discovery. Easing processing limitations may not be as important for the perception group because the task is relatively easy. The effects of pattern goodness differ from those of part goodness. In the imagery group, emergent probes were recognized equally well when good and poor patterns were formed. Pattern goodness affected response latency, however, with emergent probes being recognized faster when the pattern was poor, rather than good. In the perception group, recognition accuracy was higher when poor patterns were formed, compared to good patterns. No effects of pattern goodness were revealed in response latencies. In summary, pattern goodness influences both imaginal and perceptual discovery, although the effects are revealed in different measures. In either measure, however, the direction of the effect was consistent: discovery was faster or more accurate when poor patterns were formed, compared t o the formation of good patterns. Constructed patterns that are “good” tend to be internally coherent and less likely to encourage reinterpretation, and consequently discovery, than “poor” patterns, which tend to be less stable internally.

212

Beverly Roskos-Ewoldsen

Figure 6 Experiment 2: Percent correct recognition of emergent probes as a function of construction,part goodness, and pattern goodness.

Discovering Emergent Properties of Images

213

Figure 7 Experiment 2: Response latency for correct responses to emergent probes as a function of construction group, part goodness, and pattern goodness.

214

Beverly Roskos-Ewoldsen

Conclusions and Implications This chapter began by presenting a basic, descriptive model of the cognitive components of creativity and discovery. The model defines the creativity process as combinational play and manipulation. Discovery is equated with the process of uncovering or interpretation. The reader may have noticed that the proposed model is mute regarding the modality of form construction, and it does not make predictions regarding organization, nor does it take into account the information processing limitations of a system. Rather, the model provides a useful framework within which cognitive, perceptual, and other factors may be explored in either the discovery or the creativity domains. In the current experiments, perceptual organization or “goodness” of the stimuli was chosen as a possible influence on the discovery process. Results from the present experiments suggest that the perceptual organization of the stimuli influences discovery both when the patterns are drawn and when they are imaginally formed, though the effects for each are revealed in different measures, and the effects of part goodness differ from those of pattern goodness. The implications of these results are discussed in more detail below.

Perceptual Organization Influences Imaginal Discovery Since we are interested particularly in the role of imagery and its use in discovery, the findings of the imagery group are noteworthy. Discovery, or more precisely the recognition of emergent properties of patterns, was better when the parts used to construct the pattern were “good” (Lea,more organized), rather than “poor.” In a limited-capacity system that is being taxed (for example, by the formation of an imaginal pattern), any factor that eases processing load in one part of the task (e.g., formation) may contribute to an increase in processing capacity available for another part of the task (e.g., discovery), resulting in better performance on that second part of the task. Furthermore, discovery within imaginally constructed patterns was faster when the perceptual organization of the pattern was “poor,”rather than “good,”a result similar to the perception group’s accuracy, This finding reflects the difficulty of finding an embedded

Discovering Emergent Properties of Images

215

part within a coherent or well-organized whole. These findings have important implications for the general study of imaginal discovery. One such implication is that discovery, defined by its interpretation processes, benefits when the "data" (i,e., parts) are coherent, or well-organized, but together form a combination that contains elements that have weak, or looselyorganized, ties t o one another. Consider one of Finke et al.'s (1989) tasks. Their subjects successfully synthesized alphanumeric and geometric parts and named figures that resulted from the synthesis. In that experiment the parts (data) were familiar and tended to be "goodttforms, compared to abstract line patterns. In addition, the shapes were not strongly associated with one another. Both of these properties probably contributed to the subjects' abilities to detect emergent properties of the synthesized pattern, based on the findings of the present research. Does Perceptual Organization Influence Imaginal Creativity?

The question remains as to whether the organization of stimuli influences imaginal creativity. My guess is that it does. Some familiarity with the parts may aid combinational play or manipulation of the "data" (parts). Parts that are familiar or coherent (e.g., perceptually) may be easier to move about imaginally than parts that are less familiar or coherent. More generally, the ability to rearrange or otherwise manipulate information during the creativity stage depends on a relatively thorough knowledge of the information to be combined. Using the language of this chapter, this knowledge should be well organized, or "good." On my view, creativity is more likely to occur when the data are coherent, because the manipulation of the information is easier than when the data are opaque or ill-formulated. Future Research Directions

Future research should focus on several questions. First, to what extent does familiarity with the parts or whole affect creativity and discovery? Familiarity appears to facilitate discovery. When parts are familiar, as they are in Finke's studies (Finke, this volume; Finke et al., 1989; Finke & Slayton, 1988; see also Anderson & Helstrup and Intons-Peterson, both this volume), subjects can

216

Beverly Roskos-Ewoldsen

creatively combine the parts to form recognizable figures. In the experiments presented in this chapter, good parts aided imaginal discovery. As it happens, the good parts often resembled known patterns (i.e., U, chair, backwards Z), whereas the poor parts did not. In contrast, novel parts or orientations hinder discovery. For example, Hinton (1979)provides several demonstrations that people misperceive relations within an imagined three-dimensional object (a cube), when the imagined object is mentally rotated to a new, unusual orientation. Presumably the properties of the cube are known. But when the imagined cube is in its new, unfamiliar orientation, normal relations (i-e.,parallelity, right angles) become distorted. At the other extreme, stereotyped or overlearned parts or patterns may also interfere with discovery. People may fail to combine parts imaginally in an unusual way, or fail to discover emergent properties of images because they fixate on only one way of seeing the parts or the pattern. This phenomenon is often referred to as "rigidity" in the problem solving domain (e.g., Duncker, 1945; Luchins, 1942). Another question t o be addressed concerns the effects of practice. Though practice is related t o familiarity - the more one practices the construction of a pattern, the more familiar the pattern becomes - practice and familiarity can be separated. Practice can be conceptualized in three ways: constructing many different patterns; constructing the same pattern repeatedly, using the same parts for each construction; and constructing the same pattern repeatedly, using different parts for each construction. The effects of these types of practice on discovery of emergent properties of images in unknown (but see Intons-Peterson, this volume, for data and thoughts on some aspects of this issue). A third question for future research has two parts. First, t o what extent are easily or previously identified parts necessary for manipulation and interpretation? The stimuli used by Chambers and Reisberg (1985)and Rock, Wheeler, and Tudor (1989)did not have easily identifiable parts, and their subjects were not able to reinterpret an image of an ambiguous figure (Chambers & Reisberg) or imagine a three-dimensional wire figure from an alternative viewpoint (Rock et al.). In contrast, subjects using Finke's procedures were given parts to combine into wholes; these subjects

Discovering Emergent Properties of Images

217

were able to discover emergent properties in their images. Second, how does the reinterpretability of an image that is generated part-bypart compare with one that has been retrieved? Retrieved images may be more difficult to reinterpret than images that are purposefully constructed with identifiable parts, simply because the imager has more explicit knowledge of the relations within a pattern when it is constructed than when it is retrieved (see Finke, this volume). A final issue to be addressed is the study of creativity. There are few experiments that investigate the cognitive "mechanics" (i.e,, mental precursors, conditions, processes) of creativity, beyond those based on the creative person's psychology (see next section). We need to investigate Finke's (this volume) paradigm for inducing creativity in the laboratory further, and introduce converging methodologies to address the cognitive bases of creativity adequately.

Individual Differences, Creativity, and Discovery The study of individual differences in creativity and discovery demands attention. What underlies creative ability and the use of imagery in creativity and discovery? Sternberg and others (see Sternberg, 1988)have approached creativity by attempting to identify variables in the environment and the individual that are associated with creative individuals. In this vein, Sternberg and Lubart (1991) introduced an investment theory of creativity, which posits six resources for creativity: intellectual processes; knowledge; intellectual style; personality; motivation; and environmental context. Initial tests of the investment theory provide partial support (Lubart & Sternberg, 1991). However, in the imaginal discovery domain, Kaufmann and Helstrup (this volume) did not find any relationship between scores on a version of the Minnesota Form Board and success in their reconstrual task. Obviously, more empirical work in this area is needed.

Summary This chapter began by distinguishing creativity from discovery by arguing that they are different "stages" of the creativity/discovery

21 8

Beverly Roskos-Ewoldsen

process, and then presented a descriptive model based on the distinction. According to the model, creativity involves combining, manipulating, transforming, or otherwise rearranging information in new or unusual ways. Discovery is conceptualized as the uncovering of previously unnoticed information within existing knowledge structures, or the (relinterpretation of those structures. Necessary conditions for the use of imagery in creativity and discovery were identified. First, one must be able to represent and manipulate information imaginally. Second, one must be able t o discover emergent properties of a particular combination of information. Two experiments focusing on the second condition were presented. Though the proposed model is silent regarding specific influences of particular factors, it provides a framework within which to explore those factors that may influence either the discovery process or the creativity process, or both. Subjects in the present experiments imagined two parts of a pattern separately and then imaginally combined them to form the pattern. They were subsequently tested with possible parts of the pattern; one possible test probe consisted of parts that emerged from the synthesis of the parts. Recognition accuracy of and response latency for the emergent test probes is assumed to reflect the ability to discovery emergent properties of an imagined pattern. The perceptual organization of the stimuli was varied to test their effects on discovery of emergent properties of imaginally constructed patterns. The results indicated that imaginal discovery was better when the parts were "good" or more organized, rather than "poor." On the other hand, "poor" patterns, rather than "good" ones, resulted in quicker discoveries. Implications of the present research were discussed and proposed in the final section of the chapter. Though we know a fair amount about the creativity and discovery processes, we still have far to go before we truly understand them.

References Attneave, F. (1955). Symmetry, information, and memory for patterns. American Journal of Psychology, 68, 209 -222. Block,N.(Ed.)(1981).Imagery. Cambridge, MA: MIT Press. Chambers, D., 8z Reisberg, D. (1985). Can mental images be

Discovering Emergent Properties of Images

219

ambiguous? Journal of Experimental Psychology: Human Perception and Performance, 11, 317-328. Duncker, K.(1945).On problem solving. Psychological Monographs, 58,(5, Whole No. 270). Finke, R.A. (1989).Principles of mental imagery. Cambridge, MA: MIT Press. Finke, R.A. (1990).Creative imagery: Discoveries and inventions in visualization. Hillsdale, NJ: Erlbaum. Finke, R. A., Pinker, S.,& Farah, M. J. (1989).Reinterpretingvisual patterns in mental imagery. Cognitive Science, 13,51-78. Finke, R. A,, & Slayton, K. (1988). Explorations of creative visual synthesis in mental imagery. Memory & Cognition, 16, 252-257. Finke, R. A., Ward, T. B., & Smith, S. M. (in press). Creative cognition: Theory, research, and applications. Cambridge, MA: MIT Press. Garner, W. R., & Clement, D. E. (1963). Goodness of pattern and pattern uncertainty. Journal of Verbal Learning and Verbal Behavior, 2,446-452. Hadamard, J. (1945). The psychology of invention in the mathematical field. Princeton, NJ: Princeton University Press. Helstrup, T., & Anderson, R. E. (1991). Imagery in mental construction and decomposition tasks. In R. H. Logie & M. Denis (Eds.), Mental images in human cognition (pp. 229-240). Amsterdam: North Holland. Hinton, G. (1979).Some demonstrations of the effects of structural descriptions in mental imagery. Cognitive Science, 3,231-250. Hochberg, J., & McAlister, E. (1953). A quantitative approach t o figure "goodness." Journal of Experimental Psychology, 46, 361-364. Hoffman, R.(1990).Creation and discovery. American Scientist, 78, 14-15. Holton, G. (1972). On trying to understand scientific genius. American Scholar, 41,95-1.10. Intons-Peterson, M. J. (1981).Constructing and using unusual and common images. Journal of Experimental Psychology: Human Learning and Memory, 7, 133-144. Intons-Peterson, M. J. (1984). Faces, rabbits, skunks, and ducks: Imaginal comparisons of similar and dissimilar items. Journal

220

Beverly Roskos-Ewoldsen

of Experimental Psychology: Learning, Memory, and Cognition, 10,699-715. Klatzky, R. L., & Thompson, A. (1975). Integration of features in comparing multifeature stimuli. Perception & Psychophysics, 18,428-432. Kolinsky, R., Morais, J., Content, A., & Cary, L. (1987). Finding parts within figures: A developmental study. Perception, 16, 399-407. Kosslyn, S. M.(1980).Image and mind. Cambridge, M A : Harvard University Press. Lubart T. I., & Sternberg, R. J. (1991,June). An investment theory of creativity: Evidence for a confluence of resources and risk taking. Paper presented at the meeting of the American Psychological Society. Washington, DC. Luchins, A. S. (1942). Mechanization in problem solving. Psychological Monographs, 54, (Whole No. 248). Palmer, S. E. (1977). Hierarchical structure in perceptual representation. Cognitive Psychology, 9, 441-474. Peterson, L. R., Rawlings, L., & Cohen, C. (1977). The internal construction of spatial patterns. In G . H. Bower (Ed.), The psychology of learning and motivation (pp. 245-2761, New York: Academic Press. Reed, S. K. (1974). Structural descriptions and the limitations o f visual images, Memory & Cognition, 2,329-336. Reed, S.K.,& Brown, J.L. (1979).Temporal organization of pattern structure. Memory & Cognition, 7, 205-213. Reed, S.K.,& Johnsen, J. A. (1975).Detection of parts in patterns and images. Memory & Cognition, 3, 569-575. Reisberg, D., Smith, J. D., Baxter, D. A., & Sonenshine, M. (1989). "Enactedtt auditory images are ambiguous; "Pure" auditory images are not. Quarterly Journal of Experimental Psychology, 41A,619-641. Rock, I., Wheeler, D., & Tudor, L. (1989). Can we imagine how objects look from other viewpoints? Cognitive Psychology, 21, 185-210. Roskos-Ewoldsen, B. (1989). Detecting emergent structures of imaginal patterns: The influence of imaginal and perceptual organization. Unpublished doctoral dissertation. Indiana University, Bloomington, IN,

Discovering Emergent Properties of Images

221

Samuels M.,& Samuels, N. (1975).Seeing with the mind's eye. New York: Random House. Shepard, R. N. (1978).Externalization of mental images and the act of creation. In B. S. Randhawa & W. E. Coffman (Eds.), Visual learning, thinking, and communication (pp. 133-190). New York: Academic Press. Shepard, R. (1988). The imagination of the scientist. In K. Egan & D. Nadaner (Eds.), Imagination and education (pp. 153-185). New York: Teachers College Press. Shepard, R. N., & Cooper, L. A. (1982). Mental images and their transformations. Cambridge, MA: MIT Press. Sternberg, R. J. (Ed.) (1988).The nature of creativity: Contemporary psychological perspectives. New York: Cambridge University Press. Sternberg, R. J., & Lubart, T. I. (1991). An investment theory of creativity and its development. Human Development, 34,l-31. Thompson, A. L., & Klatzky, R. L. (1978). Studies of visual synthesis: Integration of fragments into forms. Journal of Experimental Psychology: Human Perception and Performance, 4,244-263. Tye, M. (1991). The imagery debate. Cambridge, MA: MIT Press. Wertheimer, M. (1945).Productive thinking. New York: Harper.

This Page Intentionally Left Blank

Imagery, Crcativity, and Discovery: A Cognitive Perspective B . Roskos-Ewoldson, M.J. Intons-Peterson and R.E. Anderson (Editors) 0 1993 Elscvier Science Publishcrs R.V. All rights rescrved.

223

Chapter 8

MULTIPLEPERSPECTIVES ON DISCOVERY AND CREATMTY IN MIND A N D ON PAPER Rita E. Anderson c-pf!* Memonal Unzversity St. John's, Newfoundland Canada AlC 5S7

Tore Helstrup

DeParhnent0f-p-

University of Bergen Sydneshaugen 2 5007 Bergen, Norway

Repeated demonstrations that visual imagery is implicated in tasks that require the manipulation of visual-spatial information, such as mental rotation, have tempered the old debates about whether visual imagery uses distinct internal representations and processes. As well, the work on creative inventions (Finke, 1990;this volume) is difficult t o explain without reference to the use of mental imagery. The focus of this chapter is on visual discovery tasks where subjects use mental imagery t o find parts embedded within wholes (Reed & Johnsen, 1975), to decompose whole patterns into their component parts (Helstrup & Anderson, 1991), and to compose recognizable patterns from parts (Anderson & Helstrup, in press; Finke, 1989, 1990; Helstrup & Anderson, 1991). Our research is based on the task developed by Finke and Slayton (1988). In this version of the visual discovery task, subjects hear the names of three parts randomly drawn from a pool of 15 parts consisting of simple geometric shapes and alphanumeric characters. During a brief work period, the subjects try to create a recognizable pattern by combining the parts using only mental imagery. In general, subjects are remarkably successful. In Finke and Slayton's original studies, randomly selected undergraduates came up with recognizable or "good" patterns (as judged by three independent judges) on approximately 40% of the trials. And even though the subjects were not instructed to generate creative

224

Rita E. Anderson & Tore Helstrup

patterns, approximately 16% of the good patterns were classified as creative by at least two of three judges. In these experiments, as in several other studies reported by Finke (1990),care was taken to ensure that the results could not be explained by experimenter-bias or sophisticated guessing by subjects (cf. Intons-Peterson, 1983). In this chapter, we discuss some of our experiments on visual discovery. But first, we wish to introduce the levels of analysis framework and use it to examine visual imagery and the visual discovery task from a variety of perspectives. Our goals here are twofold: to provide additional evidence in support of the reality of non-propositional mental representations and to suggest a fruitful way to coordinate research efforts on creativity and visual discovery.

Levels of Analysis In an attempt to interrelate cognitive and ethological research, Anderson (1986,1989) modified the levels of analysis framework advocated by Tinbergen (1963)for the study of animal behavior. The levels of analysis framework is a promising way to organize information about many cognitive phenomena. The framework is convenient because it systematically organizes known information about a given phenomenon from a number of perspectives. It is effective because the organization invites comparison across levels, thereby providing constraints on old interpretations and suggesting new testable hypotheses. In addition, the levels framework readily reveals a lack of experimental effort in important areas. What is the levels of analysis framework and how might it be applied to the question of creative visual discovery? Anderson (1986, 1989)identified three main classes of questions and explanations, as can be seen in Table 1. Ultimate questions ask why an organism behaves as it does from the perspectives of evolutionary history and adaptive significance. Hypothetical-Constructs or cognitive questions define the problem in cognitive terms and focus on the processing of stimuli, most often from an information processing perspective’. Anderson’s original application of Tinbergen’s levels of analysis framework to cognition was developed independently of Bruce’s (1985)application to the study of memory. In Tinbergen’s and Bruce’s scheme, cognitive explanations are treated as mechanisms and do not occupy a separate level.

Multiple Perspectives on Discovery and Creativity

225

This is the level most cognitive psychologists work at most of the time. Proximate questions ask how a process or behavior comes to be; they focus on the developmental history of the individual and the neurological, physiological, or hormonal mechanisms that produce the behavior. TABLE 1 A levels of analysis approach to the study of visual imagery and discovery.

ULTIMATE (WHY) QUESTIONS Why is visual imageiyldiscovery the way it is? Evolutionary History: What is the likely evolutionary history of visual imagery and visual discovery processes? What are the evolutionaqdecologicalconstraints on visual imagery? Adaptive Significance: What is visual imagery and visual discovery good for? What problems does it solve or create? What individual differences relate to success in visual imagery and visual discovery tasks? HYPOTHETICAL-CONSTRUCTS QUESTIONS What is visual imageryldiscovery?

What are the operating characteristics of the hypothetical structures and processes involved in visual imagery and visual discovery? Formal Computational Models - Kosslyn (1980) Intuitive Spreading (Principles) - Finke (1989) What are the conditions that promote effective discovery in mental imagery? PROXIMATE (How) QUESTIONS How does visual imageryldiscovery work?

Developmental History: How do the structures and processes of visual imagery change across the life span? Are there age differences in fluency of combinatonal play or interpretation? Mechanisms: What are the neurological, physiological, and hormonal mechanisms that support performance in visual imagery and visual discovery tasks?

226

Rita E. Anderson & Tore Helstrup

The Hypothetical-Constructs (Cognitive) Level Because the cognitive level is most familiar to cognitive psychologists, it is appropriate to begin here. Note, however, that Anderson (1986, 1989) placed the cognitive level between the ultimate and proximate levels to highlight the need for cognitive explanations to be sensitive to what is known at the other levels of analysis. That is, Anderson argued that the operating characteristics of the hypothetical structures and processes proposed by cognitive psychologists to explain behavior should be consistent with what is known about the phylogenetic and developmental history of the behaving organism, the utility of these structures and processes for solving problems in current environments, and the nature of the mechanisms that produce the actual behavior. Some cognitive questions focusing on the operating characteristics of visual imagery and visual discovery are outlined in Table 1. More specific questions might focus on the time course of image generation and regeneration or the amount of detail that could be maintained in a visual image. Most cognitive information concerning the nature of imagery (cf. Kaufmann & Helstrup, this volume) would be organized within the hypothetical-constructs level. Within this level, different approaches to the study of visual imagery are possible, as outlined below. Of the theoretical approaches to research on visual imagery, two have been used particularly often: a formal computational stance that yields models of the sort proposed by Kosslyn (1980) and an empirically-driven intuitive-spreading stance that aims for the establishment of broad and general principles independent of the specifics of any model (Finke, 1989). Needless to say, there are clear costs and benefits resulting from adoption of either; moreover, significant points may be overlooked when one stance is pursued to the exclusion of the other. Most often, our research is driven by a combination of stances acting in sequence or in concert. Regardless of approach, psychologists operating at the hypothetical-constructs level tend to ask basic questions about the cognitive contours of visual imagery and creativity. What is visual imagery? What is creativity? What are the operating characteristics of the hypothetical cognitive structures and processes involved in visual imagery and visual discovery? Can individuals make visual discoveries using

Multiple Perspectives on Discovery and Creativity

227

mental imagery alone? Can visual images be re-interpreted? What are the conditions that promote effective discovery in mental imagery?

The Ultimate (Why) Level Questions from the ultimate level focus on why visual imagery is the way it is or what it is good for. Consider the evolutionary perspective first. Although the evolutionary sequence in Figure 1 is highly improbable, the cartoon highlights the issue of evolutionary change and provides a bit of entertainment to those of us interested in mental imagery. Evolutionary questions would focus on the likely evolutionary history of the processes and structures involved in visual imagery and creativity with respect to our human, primate, mammal, and vertebrate lineages. For example, some aspects of visual discovery may reflect constraints imposed by the evolutionary If our history of human visual information processing. representational system reflects the centrality of primate vision, as suggested by Lachman and Lachman (19791, many of our creative discoveries might have to be formed through visual representations. These visual representations were most likely internal visual representations, because the skills and tools needed t o externalize our internal representations (e.g., use of pencil and paper for drawing or writing) probably did not exist until fairly late in our evolutionary development. For example, although an early took-maker may have visualized potential in a rock or bone, it is unlikely that the individual would have previously worked out the concrete details of the tool on paper. Our evolutionary history probably imposed constraints on what could be done in mental imagery. Among others, these constraints likely reflect the ecological conditions of the times (Shepard, 1984, 19891, the limits and capacities of the relevant perceptual system (Finke, 1989) and those of the body (Parsons, 1987), and perhaps, the socio-cultural milieu of early humans (cf. Marshack, 1988). Now consider questions of adaptive significance. A behavior that has adaptive significance is one that is (has been) subject to selection pressure such that those individuals who can (could) do it are (were) more likely to survive and reproduce than those who can (could) not do it. Of interest here is the utility of visual imagery and

Rita E. Anderson & Tore Helstrup

228

&

Brains.

((9 by

Sidney Harris -American Scientist magazine.)

visual discovery in current ecological times: What is it good for? What problems does it solve or create? (See Kaufmann & Helstrup, this volume, for a slightly different approach t o the question of adaptive significance.) There is an obvious advantage accruing to organisms who can mentally represent the locations of objects in their environment. Research on cognitive mapping suggests that this ability must be available to all organisms who forage or roam from a fxed home site, from bees and birds to wolves and humans (Hazen, 1983;Heth & Cornell, 1985). The ability to rearrange perceived objects in an internal representation has enormous value. Consider, for instance, the advantage to the driver on a four-lane highway who correctly estimates who will be in what lane when. In general, the ability to manipulate spatial information mentally may be preliminary t o

Multiple Perspectives on Discovery and Creativity

229

success in visual discovery, as addressed here. Finke (19901,Neisser (19871,and Shepard (1978,1984,1989) have also been impressed by the apparent spatial nature of many thought processes. "Anticipatory images" (Neisser, 1976), "cognitive maps'' (Tolman, 1948) and "mental models" (Johnson-Laird, 1983) allow people to model current reality and anticipate future realities. If we can mentally rearrange worlds, can we not mentally rearrange simple shapes in the laboratory? Evolutionary thinking also leads to a focus on diversity within a population. From a biological perspective, variability in performance is the norm (see Bruce, 1985). A focus on variability translates directly to questions concerning the relation between individual differences in some cognitive structure or process and success in visual imagery tasks. Historically, most research on visual imagery has focused on individual differences - most often with little success. More recently, efforts have been made to identify which processes contribute t o individual differences in visual imagery performances (Kosslyn, Brunn., Cave, & Wallach, 1984). More germane to the present topic, recent evidence collected in Bergen, Norway, suggests a moderate correlation between performance in our visual discovery task and measures of spatial ability (a Norwegian version of the Minnesota Form Board), but not with verbal ability (the vocabulary test of the WAIS). The Proximate (How)Level

Turning now to the "howl' questions, we first focus on children's development of visual imagery and creativity. What are the developmental histories of the cognitive structures and processes that support visual imagery and visual discovery? Are there developmental milestones? As Shepard (1978)points out, internal representations based on visual-spatial information may be developmentally prior t o internal representations based on language. However, with the exception of some work reported by Kosslyn (19801, few psychologists outside the Piagetian tradition (see Mandler, 1983, for a review) have examined the development of One notable and relevant exception is a imagery per se. developmental study of decomposition, the ability to find parts within figures (Kolinsky, Morais, Content, & Cary, 1987). Although

230

Rita E. Anderson & Tore Helstrup

perceptually good parts were readily detected by preschool children, primary-school children, and unschooled adults, only primary-school children were able to detect deeply embedded, perceptually poor parts beyond chance levels. Apparently, the analytic postperceptual process involved in finding perceptually poor parts in figures does not simply develop with age, but is enhanced by exposure to formal schooling. Whether there are developmental trends in visual discovery tasks that require subjects to compose whole patterns from parts remains to be investigated. Finally, consider the mechanisms that support visual imagery and creativity. Research from this perspective might focus on the effects of known neurological pathologies on visual imagery tasks. work, as well Such work is well represented by Farah’s (1984,1988) as by Kosslyn’s more recent research (1987,1991). We focus on Farah (19841,who examined 37 case studies in the literature reporting loss of imagery. The pattern of deficits and preserved abilities was consistent with a computational model consisting of several independent components: specific deficits were associated with the image generation process, the visual memory system, and the image inspection process. Other research at this level might focus on differential patterns of positron emission tomography (PET) during task performance in normal individuals (see Raichle, 1990,for a clear introduction to this technology). A third source of interesting ideas may stem from connectionist models, assuming that they do succeed in functioning as an interlingua between cognitive and neuropsychological researchers (Kosslyn, 1987;Shepard, 1989). In summary, a levels of analysis perspective on questions of visual imagery and visual discovery organizes the existing data in a coherent manner and identifies areas in need of further research. Equally important, the levels of analysis framework reminds us that cognition has a biological basis, a fact often ignored in this computationally sophisticated world. In particular, the evolutionary perspective highlights the critical need for organisms to represent and process spatial information about the world mentally, a need that perhaps set the stage for the creative use of visual imagery several millennia ago (see Shepard, 1978;also Intons-Peterson, this volume). As cognitive psychologists,we have been careful to rule out the effects of skilled guessing or anticipation by subjects and experimenter bias as explanations for our results. The evolutionary

Multiple Perspectives on Discovery and Creativity

231

perspective suggests that we should not be surprised to find that people can make creative discoveries through the use of mental imagery alone.

A Model of Visual Discovery Tasks Helstrup and Anderson (1991) based the process model shown in Figure 2 on Farah’s (1984) componential analysis and applied the model t o Finke and Slayton’s (1988) basic visual discovery task. Helstrup and Anderson assumed that at the beginning of each new trial, the specified parts are loaded into the memory store, as visual representations or verbal descriptions or both. Construction processes are used to combine these parts in the visual buffer according to the rules of combination for the task. Each configuration is analyzed by the interpretation process. If the configuration is found to have all relevant parts and the interpretation is judged to be adequately described by a brief verbal label, a decision will be made to report the construction. Otherwise, the process begins anew.

Figure 2

A model of visual discovery [from Helstrup & Anderson, 19911.

As suggested by the work of Farah (1984),we view construction and interpretation as independent processes that may interact t o produce recognizable patterns. For example, a partial construction may suggest an interpretation which then guides subsequent construction, and so on. The notion of interacting processes is consistent with ideas and data provided by Finke (1990). In our conception, the controlled processes of goal-orientedconstruction and interpretation are augmented by more automatic perceptual

232

Rita E. Anderson & Tore Helstrup

processes. Low-level gestalt-like figure formation processes (Pinker, 1984)and low-level fragmentation processes (Biederman, 1987)are hypothesized to contribute to the construction (and decomposition) of a configuration. In contrast to the controlled processes, these processes are not resource-demanding, nor do they function under conscious control. For instance, a recognizable pattern may be formed by the automatic construction of one figure, while the subject is intentionally trying to combine the parts in a different way, or an unexpected interpretation may be suggested by the automatic fragmentation of a configuration. These postulated low-level processes may also provide a partial explanation for difficulties in reinterpreting images, as well as for reports of difficulties in recombining or re-interpreting the initial image in a visual discovery task. That is, if relatively automatic low-level processes are biased toward maintaining good gestalts, good configurations may be difficult to reparse.

In Mind and On Paper Cognitive Resources

The ability of any system to operate effectively depends upon access to adequate resources. In terms of the above model, resource limitations of the controlled processes or the visual buffer could compromise performance in the visual discovery task (see also Reisberg & Logie, this volume). One question from a resource perspective concerns the effectiveness of mental imagery in the construction of recognizable figures in the Finke and Slayton (1988) visual discovery task. If their subjects could produce recognizable patterns on 40% of the trials, might they not be even more productive if they could use pencil and paper as a tool t o overcome the capacity limits of the mental imagery system? In the set of experiments to be discussed below, we include an external representation condition (drawing) against which to evaluate the effectiveness of the use of internal representations alone (mental imagery). Note that the internal-external comparison is between the use of purely mental procedures and the use of mental procedures augmented by pencil-and-paper support.

Multiple Perspectives on Discovery and Creativity

233

As reported by Anderson and Helstrup (in press), Experiment 1was modeled after Finke and Slayton’s (1988)first experiment with the exception that during the construction period, subjects used mental imagery alone on some trials and drew or doodled with a pencil on paper on other trials. At the beginning of the experiment, subjects were told that we were interested in how people create new patterns by using mental imagery with and without external penciland-paper support. When using mental imagery alone, the subjects were to close their eyes and try to combine the parts mentally to form a recognizable pattern; when using pencil-and-paper support, subjects were also to draw or doodle. Subjects were shown the pool of 15 basic parts which consisted of simple shapes and alphanumeric figures (e.g., circle, square, the capital letters T, C, the number 8, etc.). On each trial, three randomly selected parts were presented to the subjects; they then had two minutes to construct one recognizable pattern. Subjects had t o use all three parts in the construction of the pattern; the pattern could have been anything - letters, numbers, objects, familiar shapes or symbols - as long as it could be recognized from the brief description. Although the size, position, and orientation of the parts could be varied, the shape of the part could not be changed. At the end of the two-minute construction period, the subjects first wrote out a brief description of the pattern and then drew it. The scoring conventions developed by Finke and Slayton (1988) were adopted for these experiments. Three judges, blind as to representation condition, but knowledgeable as to which parts were used to compose the pattern, independently rated the correspondence between each description and pattern on a scale from 5 (easy to identify) t o 1 (impossible to identify). All patterns receiving an average correspondence rating of at least 4.0 were classified as good correspondence patterns. Judges also classified each pattern as creative or not; no explicit definition of creativity was provided. A creative pattern was defined as one that had been classified as creative by at least two of the three judges. Finally, the number of part transformations needed to create each pattern was counted. The transformation counts included size (one or two parts larger or smaller than the third), rotation (of each part less than or greater than 90 degrees from normal), mirror imaging of a part, and the embedding of a part in one or two parts.

234

Rita E. Anderson & Tore Helstrup

Our expectation was that provision of external support would enhance the production of recognizable patterns compared to production based on the manipulation of internal representations alone. This expectation was based, in part, on suggestions in the empirical literature that perceptual tasks may be completed faster or more accurately than imagery tasks (Finke, 1989;Roth & Kosslyn, 1988; Weber & Harnish, 1974). To the extent that percepts are easier to interpret than images (chapters by Chambers, Hyman, Kaufmann & Helstrup, Peterson, Reisberg & Logie, this volume; Chambers & Reisberg, 1985;Finke, Pinker, & Farah, 1989;Reisberg, 1987),a pattern may also be easier t o interpret as a drawing than as an image. From a theoretical perspective, we expected that because the parts and compositions were externalized when drawing, the maintenance demands on the cognitive resources during construction and interpretation would be considerably lower than when all aspects of the task were carried out mentally (Kosslyn, 1980;Reisberg, 1987). In consequence, more cognitive resources should be available t o facilitate planning and construction when drawing than when using mental imagery alone. The results of Experiment 1 replicated Finke and Slayton’s (1988)report that untrained subjects could make discoveries by using mental imagery alone. Contrary to our expectations, there were virtually no differences between working with and without external support. Subjects produced patterns that were easy to identify from the verbal description (good correspondence patterns) on 32% of the mental imagery alone trials and 34% of the drawing-support trials. Nor were there any differences in the mean rated correspondence between the verbal descriptions and the drawings, the transformational complexity of the patterns, or the number of creative patterns. The only evidence to suggest a benefit of external support was a significant reduction in failures to produce any pattern on a trial from 13% in the mental imagery alone trials t o 6% in the drawingsupport trials. The results of the first experiment suggested that when the task was to generate one recognizable figure in each two-minute trial, mental imagery was as effective when it was used alone as when it was used with pencil-and-paper support. Would the same be true if the task were to produce multiple patterns? If subjects have more cognitive resources at their dispasal when using external support,

Multiple Perspectives on Discovery and Creativity

235

they may be able to produce more patterns or higher quality patterns than when using mental imagery alone. In Experiment 2, Anderson and Helstrup (in press) required subjects to produce and describe as many patterns as possible. 1)uring a three-minute construction period, subjects were to combine the parts until they came up with a recognizable pattern. Then they were to describe the pattern on the response sheet and return immediately to the construction task. At the end of the construction period, the subjects drew all patterns they had previously described. In Experiment 2, subjects were more likely to produce at least one good pattern per trial with external support than without (69% vs 49%,respectively); they were also less likely to fail to produce any patterns per trial with external support than without (9% vs 21%, respectively). These data replicated the apparent greater difficulty of producing any patterns using internal representations alone found in Experiment 1. Further examination of the data revealed that external support led to greater productivity per trial than reliance on mental imagery alone, as measured by the mean numbers of patterns and of good correspondence patterns per trial. However, being able to draw did not lead to proportionately more good correspondence or creative patterns than using mental imagery alone. The probability of producing a good correspondence pattern or a creative pattern, each conditionalized on total productivity, did not differ significantly across representation conditions; nor did the mean correspondence ratings or the mean transformational complexity of the patterns. In summary, providing external support increased the quantity, but not the quality, of patterns produced per trial. Helstrup and Anderson (1991)took a slightly different approach to the question of resources. They examined the effect of representation condition (internal vs. external) on composition and decomposition tasks, using stimulus materials that had previously been rated as more or less difficult to interpret or construct. The composition task was similar to the task used in Experiments 1and 2 above. On each trial of the decomposition task, subjects were presented one of the previously rated patterns; they were required to suggest an interpretation for that pattern and then to decompose the pattern into its component parts. The decomposition task proved to be easier than the construction task. As in Experiments 1 and 2 above, no-pattern trials were more frequent in the internal

236

Rita E. Anderson & Tore Helstrup

representation condition than in the drawing-support condition. However, the rated difficulty of the stimulus materials did not affect performance in either task, either as a main effect or in interaction with representation condition. Although none of the above experiments placed undue demands upon the resources available to perform the construction task, working with external support did lead to fewer no-pattern trials and greater productivity than working with mental imagery alone. However, because these effects are not always found (see Experiment 3 below; also Anderson & Helstrup, in press, Experiment 3), the advantage of working with external pencil-and-paper support appears to be rather small and statistically unstable in these construction tasks. In general, we are impressed by the ability of randomly selected North American and European subjects to produce new patterns using only internal representations. That these representations are probably visual is suggested by the research literature (Finke, 1989, 1990; see also Peterson, this volume), as well as by the results of a final experiment described by Helstrup and Anderson (1991). As in our previous experiments, subjects worked with or without external drawing support. In addition, separate groups of subjects were instructed to use either verbal or visual imagery strategies during the construction period. Although representation condition did not affect performance, strategy instruction did. Subjects produced acceptable patterns on approximately 70% of the trials when using visual imagery strategies and 40% of the trials when using verbal strategies; they failed to produce any pattern on 4% and 13% of the trials, respectively. This experiment raises, but does not answer, the question about the extent to which performance is due to controlled or more automatic figure construction and fragmentation processes. The fact that performance can be influenced by strategy instruction is noteworthy and perhaps can be used t o assess the relative roles of controlled and automatic processes during the construction and interpretation of novel patterns.

Expectations Taken together, the results of Experiments 1 and 2 suggested that mental imagery was remarkably effective for composing and

Multiple Perspectives on Discovery and Creativity

237

interpreting recognizable patterns. These results were contrary to the belief, shared by experimenters and subjects alike, that using pencil and paper to help construct patterns should be easier or more effective than using mental imagery alone. If subject expectations influenced performance, we might have underestimated the power of mental imagery. That is, because the subjects in Experiments 1 and 2 had complete instructions about both representation conditions prior to any experimental trials, their expectations might have suppressed their performance. This expectancy-based explanation of imaginal performance is at odds with the following results. As reported by Finke (1990), Neblett, Finke, and Ginsburg (1989) found that the subjects who used mental synthesis were at least as effective as other subjects who were able to manipulate parts that were provided as transparencies. As well, the results of the experiments reported by Helstrup and Anderson (1991) where representation condition was manipulated between subjects were remarkably similar t o those reported for Experiment 1. Working with external support led t o slightly more good-pattern trials and significantly fewer no-pattern trials than working with purely internal representations. In our third experiment, which was designed explicitly to examine the effects of expectations on performance, no differences in productivity were observed. As in Experiment 1 and those reported by Helstrup and Anderson (1991), the task in Experiment 3 was to produce one pattern on each two-minute trial. To estimate the time needed t o produce a pattern using each representation condition, subjects started a timer at the beginning of each trial. As soon as they had written down the verbal description of the pattern, they stopped the timer and recorded the elapsed time on a response sheet. The subjects then drew their pattern, as before. In contrast to Experiment 1, the subjects were told about the mental imagery condition only after completing the first block of five trials in the drawing-support first condition or vice versa. After completing all experimental trials, the subjects were given a rating sheet containing 1-3 sample patterns and their descriptions; these examples had been selected from previous experiments t o illustrate each point on the 5point rating scale, which ranged from 5 (easy to identify) to 1 (impossible to identify) from its verbal description. The subjects rated the correspondence of each of their patterns to its verbal

238

Rita E. Anderson & Tore Helstrup

description using this scale and decided whether each of their own patterns was creative or not. Subsequently, three judges used the rating sheet with examples to rate the patterns as before. The basic results are summarized in Table 2. Subjects took significantly less time to construct and describe a pattern when using internal representations only than when also using pencil-and-paper support. Apparently, our mind is sometimes more agile than our hands (see also Roskos-Ewoldsen, 1989). As in Experiment 1 (this chapter), there was absolutely no difference between representation conditions in terms of the number of good correspondence pattern trials, the mean correspondence ratings made by the judges, the transformational complexity of the patterns, or the number of creative pattern trials. In fact, the only difference between the two experiments was that in Experiment 3, there was no difference in the number of no-pattern trials. No significant effects of test block or order were found in the analysis of all data, and the results were identical when only first block performance was analyzed. Hence, subject expectations did not affect overt performance in measurable ways. However, the subjects’ ratings reflected a bias in favor of external representations. The correspondence ratings made by the subjects tended to be higher for those patterns produced with the assistance of external support than for those produced without external support; consequently, the number of subject-defined good correspondence patterns was significantly greater in the external than in the internal representation conditions. Very similar results were found in an experiment patterned after Experiment 2 (this chapter) wherein subjects were instructed to produce as many patterns as possible in a three-minute work period using either internal or external representations (Anderson 8z Helstrup, in press, Experiment 3). As in the present Experiment 3, subjects were not informed about the other representation condition until they had completed the first block of trials. At the end of the second block of trials, subjects rated the correspondence between the drawings and their verbal descriptions. The results revealed virtually no differences in overt performance as a function of representation condition when assessed by objective means (e.g., blind judges). Nor did expectations influence actual performance. Subjects, however, again revealed their biases, giving significantly

Multiple Perspectives on Discovery and Creativity

239

TABLE 2 Mean performance measures in Experiment 3.

Representation Conditions Measure (%Max) Mean time (sec) to verbal description Mean trials with any pattern (Max=5)

Internal

External

48.55

65.36

*

4.30

(86%)

4.25

(85%)

Mean good-pattern trials

1.30

(26%)

1.25

(25%)

Mean poor-pattern trials

3.00

(60%)

3.00

(60%)

.70

(14%)

.75

(15%)

Mean no-pattern trials Mean transformational complexity (Max=13)

2.83

3.11

Mean j-correspondence rating (Max=5)

3.35

3.41

Mean s-correspondence rating (Max=5)

3.41

3.66

Mean s-good pattern trials (Max=5)

1.75

(35%)

2.45

(49%)*

At least 2 judges

.30

(6%)

.55

(11%)

At least 1judge

1.30

(26%)

1.15

(23%)

Subject-judged creative

2.55

(51%)

2.50

(50%)

Mean creative pattern trials (Max=5)

Note:

* indicates that the difference is significant beyond the .05 level.

~~

~

higher correspondence ratings to patterns produced with external assistance.

Creativity In contrast to Finke (Finke & Slayton, 1988; Finke, 1990, this

Rita E. Anderson & Tore Helstrup

240

volume) and Intons-Peterson (this volume), we classified all patterns and their descriptions as creative or not, regardless of their correspondence ratings. In fact, a substantial number of poor correspondence patterns were classified as creative (see Figure 3 for some examples of creative patterns from Experiment 3). Across the experiments presented in this chapter where creativity was assessed, 50 of the 113 patterns classified as creative by at least two of the three judges were poor correspondence patterns. Perhaps with more time and work, the poor correspondence patterns would be more recognizable, as well as creative, to outside observers.

INTERNAL GOOD

POOR

EXTERNAL

GOOD

POOR

A

B

C Figure 3 Sample good and poor correspondence patterns produced in Experiment 3 under internal and external representation conditions that were judged as: (a) creative by at least two judges and the subject, (b) creative by only the subject, and (c) noncreative by all judges and the subject. *Note Therewerenopooramwpo&nce patterns in this cell. This pattern was classified as Creative by at least two judges, but not by the subject. Because this subject had classSed a pmvioua highway Bcene as creative, it seemed reasonable to use this pattern to represent the type ofpattern that might be foundin this cell.

How might the provision of external support affect creativity? Several arguments (though little data) can be advanced t o support

Multiple Perspectives on Discovery and Creativity

241

the hypothesis that more creative patterns should be produced with external support than without. The first argument is that the mind is more likely to generate prototypical images than novel variations; in other words, the mind is likely to remain mired in tradition and meaning. Drawing, as in doodling, might allow subjects more opportunities to create a novel combination accidentally. A second argument, related to appearances, is that patterns may not be as well formed in imagery as when drawing and hence, their potential may not be recognized as easily in an image as in a drawing. The third argument is that the choice of which pattern(s) to report is made more easily on paper than in the mind. This argument may be related to the notion that drawing is more productive than the use of mental imagery alone. Unfortunately, convincing evidence is lacking for all three arguments. In fact, although creativity was significantly affected by representation condition in only one of our experiments, it was often somewhat higher in the external than in the internal representation condition when the strict criterion (agreement of at least two judges) was used. For our current purposes, we focus on the data from Experiment 3 shown in Table 2. Although almost twice as many patterns were classified as creative in the external (11%)than in the internal representation (6%) condition, this difference was not significant. In contrast to the condition-blind judges, our subjects classified many more of their products as creative, and they classified equal numbers of external and internal productions as creative. When the criterion was relaxed so that a creative pattern was defined as a pattern that had been classified as creative by at least one judge, there was no hint of an effect of representation condition; however, even with this relaxed criterion, the proportion of patterns classified as creative by at least one judge did not reach the level of subject-defined creativity. The Eye of the Creator vs. the Eye of the Beholder Differences between the ratings made by the condition-blind judges and the subjects (Experiment 3) raise questions concerning the relationship between the creator and the beholder. Is the correspondence between a pattern and its interpretation the same for

242

Rita E. Anderson & Tore Helstrup

the observer as for the creator? Is creativity in the eye of the creator, the eye of the observer, or both? Because the judges rated the drawings blindly with respect to representation condition and subjects, their ratings should be unbiased. However, the quality of the actual drawing, defined as irrelevant by the experimental procedures, may nevertheless have influenced their ratings, such that poor quality drawings may have received lower correspondence ratings and been less likely to be judged creative than high quality drawings. In addition, external judges may not fully appreciate the potential of the pattern (cf. Finke, 1990). Because subjects know what concepts they have generated, they may be better able to judge the correspondence between their description and their drawing of the "ideal" concept. And, to the extent that creativity is based on process rather than product (Finke, 19901, subjects should have more information about the processes of construction and interpretation that went into the composition of the pattern than the external judge. On the other hand, because their ratings are not "blind" with respect to representation condition, the subjects' ratings may be biased by their beliefs as to the efficacy of internal and external representations.

Correspondence Ratings

To assess the agreement between the correspondence ratings made by the judges and the subjects, we first looked at the proportion of ratings where the external judges and the subject agreed that the correspondence between the verbal description and the pattern was either good or poor (e.g., the number of patterns receiving a mean rating of at least 4.0 from the judges and a rating of 4 or 5 from the subjects plus the number of patterns receiving a mean rating of less than 4.0 from the judges and a rating of 3 or less from the subjects divided by the total number of patterns). As shown in Table 3, overall agreement in the two experiments was acceptable, though a bit low. When the criterion was relaxed to include all patterns receiving a rating of 4 or 5 by at least one condition-blind judge, the subjects and judges seemed to be in much closer agreement. In neither case did representation condition have a measurable effect. We also examined agreement on good correspondence patterns

Multiple Perspectives on Discovery and Creativity

243

only as defined either by the judges or the subjects. A total of 51

TABLE 3 The proportion agreement between the external judges and the subject as to: (A) the correspondence (e.g., good or poor) between a pattern and its verbal label and (B)the creativity of the pattern. See the text for further details of scoring.

Representation Conditions Measure

Experiment

zn&n&d

zcukmu!

!lWd

A. Correspondence ratings. Strict criterion:

Lax criterion:

Experiment 3

.78

*

55

.67

Anderson & Helstrup (in press, Experiment 3)

.67

.70

.69

Experiment 3

.86

.81

.E34

Anderson & Helstrup (in press, Experiment 3)

.83

.83

.83

All patterns

.43

.52

.47

J-defined

.67

.91

.82

S-deiined

.08

.20

.14

All patterns

.52

59

.56

J-defined

.69

.83

.76

S-defined

.35

.38

.37

B. Creativity judgments (Experiment 3). Strict criterion (at least 2 judges): Lax criterion (at least one judge):

patterns (26 and 25 from the internal and external representation trials, respectively)received a mean correspondence rating of at least 4.0 from the three judges, while 84 patterns (35 and 49 from the internal and external representation trials, respectively) received a rating of 4 or 5 from the subjects. Only 39 patterns were rated as good by both judges and subjects. Thus, on the basis of the strict criterion, subjects and judges in Experiment 3 agreed on 76% of the

244

Rita E. Anderson & Tore Helstrup

judge-defined good correspondence patterns, but on only 46% of the For good subject-defined good correspondence patterns. correspondence patterns defined by the more lax criterion, the respective values were 90% and 75%. Parallel results were found in Anderson and Helstrup’s (in press) third experiment. The subjectcreators appear to see more correspondence between their verbal descriptions and drawings than do the judge-observers. Consequently, creators are more likely to agree with beholders that a particular pattern is easily recognizable than vice versa. Creators apparently produce much that is not recognizable to beholders.

Creativity Judgments Table 3 summarizes the agreement between the judges and the subjects with respect to creativity decisions. The proportion of creativity judgments where the judges and subjects agreed that a pattern was creative or not was low, whether based on the strict criterion (at least two judges) or the lax criterion (at least one judge). With few exceptions, the patterns that judges found to be creative according to the strict criterion were also identified by subjects as creative (88%). On the other hand, only 14% of the subject-defined creative patterns were seen as creative by the judges! A similar pattern was observed for the lax criterion. Figure 3 shows a sample of creative and non-creative patterns as classified by judges and subjects. Data such as these are potentially disturbing. How can we study creativity if creator and beholder do not agree? We seem to assume that they do. For instance, although Finke (1990)used external judges to identify creative discoveries, he frequently discusses his subject’s judgments, leaving the impression of agreement. In fact, Finke and Slayton (1988)intentionally developed a very conservative classification procedure, one that perhaps is desirable for the initial study of creative imagery. By relying on externally defined criteria (whether conservative or lax), we can maximize our hit rate, as well as ensure objectivity. On the other hand, from the subject’s perspective, our misses could be many and perhaps serious. The low agreement between subjects and judges may reflect difficulties in classifying the patterns. If so, agreement should be

Multiple Perspectives on Discovery and Creativity

245

increased by the use of a rating scale containing sample patterns, similar to the one developed for the correspondence ratings. Data from a recently completed experiment suggest that, although the use of a rating scale did increase agreement between subjects and judges, it remained lower than agreement for correspondence ratings. If replicable, the apparent failure of the provision of standards to increase the agreement for creativity judgments to the levels of agreement observed for correspondence ratings suggests that subjects and judges may be basing their creativity judgments on different types of information, as outlined below. Recall that our model of the visual discovery task postulates distinct construction and interpretation processes. If during the construction phase, conscious, controlled figure-construction processes are accompanied by an unexpected interpretation suggested by a more automatic low-level fragmentation process, the pattern may be experienced by the subject as creative. In a similar vein, controlled interpretive attempts may trigger an unexpected construction, which is again experienced as creative by the subject. Perhaps these are the events that underlie the "Ah Ha" experience when a construction comes together (Finke, 1990). In contrast, when both construction and interpretation develop consciously, in parallel, the result may be unsurprising to the subject and evaluated as noncreative -the end result of an effortful process. External judges, however, will not be aware of the mechanisms or processes of discovery. They will most likely focus on the product and may or may not see any of the patterns as creative. As the above analysis suggests, we need to develop procedures to identifjr the information used by creators and by beholders when they assess the creativity of a pattern or process (also see Roskos-Ewoldsen, this volume). Drawing as Externalized Imagery One benefit of having subjects draw while working out patterns is that the actual doodles can be used t o examine the independence of composition and interpretation. In Experiment 3,where the task of the subject was to produce one pattern per trial, subjects produced an average of 3.9doodles per successful trial; 1.8 of those doodles had known interpretations (i.e., they had been produced, interpreted, and

246

Rita E. Anderson & Tore Helstrup

reported by other subjects in the same or different experiments), one of which was chosen by the subject. On unsuccessful (no-pattern) trials, subjects averaged 8.5 doodles and 2.4 of these had known interpretations. Hence, subjects do produce numerous combinations, only some of which are interpreted. In Anderson and Helstrup’s (in press) third experiment, 60 doodles, in addition to the 160 chosen patterns, had known interpretations that were not reported. Because subjects in this experiment were instructed to produce and report as many patterns as possible, these data provide strong evidence that subjects do not interpret all combinations. Hence, construction processes may operate independently of interpretation processes, although they do not need to. We also examined the number of repetitions of chosen patterns. In Experiment 3, 65% of the reported patterns were repeated 1-4 times while doodling. In Anderson and Helstrup’s (in press) third experiment where productivity was emphasized, approximately 40% of the chosen patterns were repeated. These high repetition rates suggest that subjects may be refining an interpreted pattern or one that seems to afford interpretation. Because subjects rarely provided more than one interpretation for the same construction, the repetitions are unlikely to reflect attempts to generate a new interpretation for the same construction. The problem, of course, is that we lack information about how subjects interpret the patterns, either on the first or subsequent times they draw them. The fact that subjects can repeatedly construct the same pattern and still not provide an interpretation of it, provides additional support for the independence of construction and interpretation processes. Finally, we examined the doodles for evidence of systematic variation of the parts to create novel patterns. With the exception of trials in which subjects drew one part in more than one place on a two-part base figure, there was no evidence of systematic variation. Because the mind seems to be faster than the hand in this task (i.e., subjects took longer to generate a pattern when working with than without pencil-and-paper support) and the mind presumably directs the hand, subjects may have performed much of the systematic variation or planning before they actually doodled on paper. If so, there would be little or no overt evidence of systematic variation in the doodles. These data raise the interesting question of whether the same

Multiple Perspectives on Discovery and Creativity

247

types of events occur when subjects are relying solely on mental imagery to construct novel patterns. Do subjects construct mental patterns with known interpretations, yet fail to see the other interpretations? Sometimes yes, as suggested by many authors in this volume who discuss the problems of reinterpreting images. Do subjects recognize potential in their images and then refine them? Presumably yes (Finke, 1990, this volume). Do subjects try out various combinations of parts in their images? Presumably yes, otherwise where do the new constructions come from? Note that these questions parallel the arguments given in the previous section for the belief that creative patterns will be produced more often with than without external support,. Our tentative answers t o these questions suggest that the same types of events occur when subjects rely solely on mental imagery to create novel patterns as when they can also use external support.

Conclusions In summary, we have explored the efficacy of the use of mental imagery alone for the production of novel patterns, relative t o the use of mental imagery augmented by external drawing support. We find mental imagery to be considerably more proficient and productive than suggested by many current theories of cognition. Many people seem to believe that being able to use tools, such as pencil and paper, will enhance performance, either in terms of greater productivity, better correspondence, or greater creativity. Most of the reasons given for this belief appear to be resource-based, though other reasons suggest that external representations allow for higher probabilities of accidental discovery or greater ease of reworking a product. Despite beliefs, however, there was very little evidence that the use of tools make a major difference within the range of variation tested in these experiments. That tools are potentially important was suggested by: (a) the tendency for fewer no-pattern trials when external support was present; (b) the generally consistent, though nonsignificant differences, in favor of external support conditions for mean correspondence ratings, the number of good correspondence patterns, and the number of creative patterns; and (c) the evidence from the doodling analyses showing efforts to refine particular

248

Rita E. Anderson & Tore Helstrup

compositions. Presumably, as the task becomes more resource demanding, these effects will become more prominent. The bottom line, however, is that our ability to engage in mental visual discovery is much more potent than many, though certainly not all, cognitive psychologists have traditionally been willing to admit. As should be evident, our experiments on visual discovery in mind and on paper are typical of research conducted at the hypothetical-constructs (cognitive)level, the level of explanation that uses hypothetical mental processes and structures to explain behavior. Our experimental questions flowed from a general model of the visual discovery task in which specific mental structures (memory store, visual buffer) and processes (construction, interpretation, decision) were assumed to operate on mental representations (verbal, visual); our experimental data were used to make inferences about the operation of these hypothetical structures and processes. We close by posing some questions at the other levels of analysis, questions that should complement our understanding of the hypothetical cognitive structures and processes involved in visual discovery and creativity. First, consider how ultimate questions (evolutionary history/ comparative study and adaptivdfunctional significance) might be used to expand our understanding of visual discovery in general. We begin with the evolutionary history/comparative perspective. Attempts to formulate an evolutionary model of creativity (e.g., Findlay & Lumsden, 1988) have generally failed to address the role of mental representations (e.g., visual imagery) within specific models of discovery and creativity. Marshack (1988),however, does suggest that the extensive capability of the vision-oriented primates and early hominids for creativity and discovery depends, at least in part, upon their capacity for cognitive mapping, and presumably their ability to represent and manipulate visual-spatial information mentally. Within the context of a model of visual discovery, we might now ask whether the archaeological record can be used to determine the relative importance of construction and interpretation processes throughout primate evolutionary history. Has the ability to externalize images affected creativity in different domains or different primate species? Fagen (1988) argues that current conceptions of discovery and innovation are almost identical to those of animal play. Could we develop studies of play in nonhumans,

Multiple Perspectives on Discovery and Creativity

249

perhaps with and without objects, to provide a phylogenetic perspective on discovery processes, with and without external support? From the functional perspective, we might ask whether individual differences in working memory or the visual buffer relate to differences in construction rates or interpretation speed. Are individuals who are good at combinatorial play also good at interpreting configurations? Are there individual differences in the effect of external support on performance in the visual discovery task? Does practice affect construction, interpretation, or both? Can people learn to be more productive or creative in these types of tasks? Tentative answers to some of these questions appear throughout this volume (in particular, see the chapters by Finke, Intons-Peterson, Kaufmann & Helstrup, and Roskos-Ewoldsen). Second, consider some proximate questions (developmental influences and the design features of the neurophysiological mechanisms of behavior) that might be asked about our model of visual discovery. Developmental questions would focus on age differences in the various processes and structures. Are there age differences in the controlled processes of construction and interpretation; if so, are changes in the two processes synchronous? Do the characteristics of the visual buffer change with age? If they do, does the provision of external support in visual discovery tasks show age effects? Does the probability of producing a creative construction change with age? Questions about mechanisms include whether different structures of the brain support the operation of the two processes. For instance, would a carefully constructed sequence of studies using PET scan technology provide evidence about the localization and the time course of the construction and interpretation processes in visual discovery? On an even more speculative note, do creative constructions produce the same blood flow patterns as ordinary constructions? We obviously have much to learn about visual discovery and creativity. The fact that we are able to pose the above questions and the many chapters in this volume suggest that we are on the verge of learning much more about the creative powers of the nonverbal and verbal mind.

250

Rita E. Anderson & Tore Helstrup

References Anderson, R. E. (1986). Cognitive explanations and cognitive ethology. In W. Bechtel (Ed.), Integrating scientific discippline (pp. 323-336). Dordrecht: Martinus Nijhoff. Anderson, R. E. (1989). Cognitive ethology and the study of human cognition. In M. D. Topping, D. C. Crowell, & V. N. Kobayashi (Eds.), Thinking across cultures: The third international conference (pp. 245-253). Hillsdale, N J Erlbaum. Anderson, R. E., & Helstrup, T. (in press). Visual discovery in mind and on paper. Memory & Cognition. Biederman, I. (1987). Recognition-by-components: A theory of human image understanding. Psychological Review, 94, 115-147. Bruce, D. (1985). The how and why of ecological memory. Journal of Experimental Psychology: General, 114,78-90. Chambers, D., & Reisberg, D. (1985). Can mental images be ambiguous? Journal of Experimental Psychology: Human Perception & Performance, 11, 317-328. Fagen, R. (1988).Animal play and phylogenetic diversity of creative minds. In C. S. Findlay & C. J. Lumsden (Eds.), The creative mind (pp. 79-82). NY: Academic Press. Farah, M.J. (1984). The neurological basis of mental imagery: A componential analysis. Cognition, 18,245-272. Farah, M.J. (1988). Is visual imagery really visual? Overlooked evidence from neuropsychology. Psychological Review, 95, 307-3 17. Findlay, C. S., & Lumsden, C. J. (1988).The creative mind: Toward an evolutionary theory of discovery and innovation. In C. S. Findlay & C.J, Lumsden (Eds.), The creative mind (pp. 3-55). Ny: Academic Press. Finke, R. A. (1989). Principles of mental imagery. Cambridge, MA: Mll' Press. Finke, R. A. (1990).Creative imagery: Discoveries and inventions in visualization. Hillsdale, NJ: Erlbaum. Finke, R. A., Pinker, S., & Farah, M.J. (1989).Reinterpreting visual patterns in mental imagery. Cognitive Science, 13, 51-78. Finke, R.A., & Slayton, K.(1988). Explorations of creative visual synthesis in mental imagery. Memory & Cognition, 16,252-257.

Multiple Perspectives on Discovery and Creativity

251

Hazen, N. (1983).Spatial orientation: A comparative approach. In H. L. Pick, Jr. & L. P. Acredolo (Eds.), Spatial orientation: Theory, research and application (pp. 3-37). Ny:Plenum Press. Helstrup, T., & Anderson, R. E. (1991). Imagery in mental construction and decomposition tasks. In R. H. Logie & M. Denis (Eds.), Mental images in human cognition (pp. 229-240). Amsterdam: North Holland. Heth, C. D., & Cornell, E. H. (1985). A comparative description of representation and processing during search. In H. M. Wellman (Ed.). Children’s searching: The development of search skill and spatial representation (pp. 215-249). Hillsdale, NJ: Erlbaum. Intons-Peterson, M. J. (1983).Imagery paradigms: How vulnerable Journal of are they to experimenters’ expectations? Experimental Psychology: Human Perception and Performance, 9,394-412. Johnson-Laird, P. N. (1983). Mental Models. Cambridge, MA: Harvard University Press. Kolinsky, R., Morais, J., Content, A., & Cary, L. (1987). Finding parts within figures: A developmental study. Perception, 16, 399-407. Kosslyn, S . M. (1980).Image and Mind. Cambridge, MA: Harvard University Press. Kosslyn, S. M. (1987). Seeing and imagining in the cerebral hemispheres: A computational approach. Psychological Review, 94,148-175. Kosslyn, S. M. (1991). A cognitive neuroscience of visual cognition: Further developments. In R. H. Logie & M. Denis (Eds.), Mental images in human cognition (pp. 351-381). Amsterdam: North Holland. Kosslyn, S. M., Brunn, J., Cave, K. R., & Wallach, R. W. (1984). Individual differences in mental imagery ability: A computational analysis. Cognition, 18, 195-243. Lachman, J. L., & Lachman, It. L. (1979). Theories of memory organization and human evolution. In C. R. Puff (Ed.), Memory organization and structure (pp. 133-193). Ny: Academic Press. Mandler,J. M. (1983).Representation. In P.H. Mussen (Ed.), Handbook of child psychology, Vol. IZI (4th ed.) (pp. 420-494). Ny:Wiley. Marshack, A. (1988).The species-specific evolution and contexts of the creative mind: Thinking in time. In C. S. Findlay & C. J.

252

Rita E. Anderson & Tore Helstrup

Lumsden (Eds.), The creative mind (pp. 116-119). Ny: Academic Press. Neisser, U. (1976). Cognition and Reality. San Francisco: W . H. Freeman. Neisser, U. (1987). A sense of where you are: Functions of the spatial module. In P. Ellen & C. Thinus-Blanc (Eds.), Cognitive processes and spatial orientation in animal and man. Vol II. Neurophysiology and developmental aspects (pp. 293-309). Dordrecht: Martinus Nijhoff. Parsons, L. M. (1987). Imagined spatial transformations of one’s hands and feet. Cognitive Psychology, 19, 178-241. Pinker, S.(1984).Visual cognition: An introduction. Cognition, 18, 1-63. Raichle, M. (1990). Images of the functioning human brain. In H. Barlow, C. Blakemore, & M. Weston-Smith (Eds.). Images and understanding (pp. 284-296). Cambridge: Cambridge University Press. Reed, S.K., & Johnsen, J. A. (1975). Detection of parts in patterns and images. Memory & Cognition, 3,569-575. Reisberg, D. (1987).External representations and the advantages of externalizing one’s thoughts. Program of the Ninth Annual Conference of the Cognitive Science Society, Seattle, Washington. Hillsdale, N J Erlbaum. Roskos-Ewoldsen, B. (1989). Detecting emergent structures of imaginal patterns: The influence of imaginal and perceptual organization. Unpublished doctoral dissertation, Indiana University, Bloomington, IN. Roth, J. D., & Kosslyn, S . M. (1988). Construction of the third dimension in mental imagery. Cognitive Psychology, 20, 344-361. Shepard, R. N. (1978).Externalization of mental images and the act of creation. In B. S.Randhawa & W. E. Coffman (Eds.). Visual learning, thinking, and communication (pp. 133-189). Ny: Academic Press. Shepard, R. N. (1984). Ecological constraints on internal representation: Resonant kinematics of perceiving, imagining, thinking, and dreaming. Psychological Review, 91,417-447. Shepard, R. N. (1989). Internal representation of universal regularities: A challenge for connectionism. In L. Nadel, L. A.

Multiple Perspectives on Discovery and Creativity

253

Cooper, P. Culicover, & R. M. Harnish (Eds.), Neural connections, mental computations (pp. 104-1341, Cambridge, MA: MIT Press. Tinbergen, N. (1963).On aims and methods of ethology, Zeitschrift fur Tierpsychologie, 20, 410-429. Tolman, E. C. (1948). Cognitive maps in rats and man. Psychological Review, 55, 189-208. Weber, R. J., & Harnish, R. (1974).Visual imagery for words: The Hebb test. Journal of Experimental Psychology, 102,409-414.

Acknowledgments This research was supported in part by funds made available by the Challenge 1990 and 1991 program from the Department of Employment and Immigration Canada. We especially wish to thank Lisa Dillon for her diligent and enthusiastic work on various experiments and Christine Arlett, Ron Finke, Peggy Intons-Peterson, and Bev Roskos-Ewoldsen for thoughtful comments on earlier versions of this chapter.

This Page Intentionally Left Blank

Imagery, Creativity, and Discovery: A Cognitivc Perspcctive R. Koskos-Ewoldson, M.J. Intons-Peterson and R.E. Anderson (Editors) 0 1993 Elsevicr Science Publishers B.V. All rights reserved.

255

Chapter 9 ~

MENTALIMAGERY AND CREATIVEDISCOVERY Ronald A. Finke Department of Psychology Texas A & M University College Station, Texas 77843-4235 There is considerable evidence that mental imagery plays a fundamental role in human cognition (Finke, 1989;Finke & Shepard, 1986; Kosslyn, 1980;Pinker, 1984;Shepard & Cooper, 1982). A number of studies have shown that mental images can be used to retrieve information (e.g., Kosslyn, 19751, transform objects and patterns (e.g., Shepard & Metzler, 19711,and solve spatial problems (e.g., Cooper, 1990). Yet few studies have explicitly addressed the role that visualization plays in the creative process, at least from the perspective of modern cognitive psychology and cognitive science. This is somewhat puzzling. One would think that creative imagery would be a natural topic of interest to cognitive scientists, given the many anecdotal accounts of the role that imagery supposedly played in scientific discovery (Miller, 1984; Shepard, 1978, 1988). For example, Einstein described having imagined the consequences of traveling at the speed of light, which led him to the concept of special relativity. Kekul6 made his fundamental discovery in organic chemistry after having imagined a snake coiled so as to represent the molecular structure of benzene. Faraday claimed to have visualized lines of force that emanated from electric and magnetic sources, resulting in the modern conception of electromagnetic fields. More recently, the physicist Feynman claimed t o have used visual images in thinking about interactions among elementary particles, which led to the development of "Feynman diagrams." There are also numerous anecdotal accounts of the role that

256

Ronald A. Finke

imagery plays in creative literary and artistic achievement. One artist, who became blind as an adult, claimed to use visualization to paint landscapes (Finke, 1986). Writers frequently comment on the role that imagery plays in suggesting new ideas for stories (e.g., Ghiselin, 1952; Koestler, 1964). Architects have reported using imagery to explore new designs for buildings and other structures. Such accounts are certainly suggestive of the importance of visualization in creative thinking and discovery. However, one needs to go beyond anecdotes and introspective reports, to develop empirical methods for studying creative imagery under controlled laboratory conditions. Perhaps the previous lack of availability of such methods, combined with a general impression that creative imagery is an unscientific topic, has resulted in its neglect in traditional areas of cognitive psychology.

A Cognitive Approach to Creative Discovery This chapter begins by outlining a general, cognitive approach to the study of creative thinking and discovery, which will serve as a framework for investigating creative imagery. Both methodological and evaluative issues are considered.

Methodological Considerations The basic research strategy is to combine the experimental methods of cognitive science with the opportunity for subjects to make new discoveries. I mean "discoveries" in a literal sense; not just those intended or expected by the researcher. Rather, subjects are encouraged to generate ideas that neither they nor the researcher may have ever had before. This can be accomplished by using a variety of novel methods in combination with standard research techniques. Allowing for creative opportunities is a very different approach from that normally taken in cognitive research, where the usual procedure is to measure accuracy or response time to verify a set of predictions (e.g., Anderson, 1990; Shepard & Cooper, 1982). Although these more traditional paradigms have their usefulness, they are of somewhat limited value in the study of creative cognitive

Mental Imagery und Creative Discovery

257

processes, where one is less concerned with predicting exactly the specific responses a person might give or scoring the responses as 'korrect'l according to some criterion. On the contrary, highly creative responses are likely to be very difficult to predict (e.g., Finke & Slayton, 1988). Among the advantages of this approach is that a researcher can avoid many of the problems associated with demand characteristics and experimenter bias. Although there may be general demands to think more creatively, these rarely translate into particular expectations for responding in some appropriate way. Such studies are therefore relatively immune to the most common types of artifacts one encounters in cognitive research (e.g., Intons-Peterson, 1983). Although it is important to allow for creative opportunities, at the same time, sufficient constraints and controls are needed to have a scientific study. For this reason, subjects are not simply told to try to be creative in some completely unconstrained way. Rather, the types of components used in the creations, the ways in which the creations can be interpreted, and the time at which the interpretive constraints are imposed should be restricted. These constraints can then be related to the likelihood that a person will generate a creative idea. The idea that imposing constraints can help one to study creative thinking might seem paradoxical, in that when a person is no longer free to respond in particular ways, the person might be less creative. However, certain types of restrictions ought to enhance creativity. For example, by limiting the components used in an imagined construction, or the range of possible interpretations, one would be forced to think in original and resourcehl ways. Thus, imposing such restrictions would not only place the creative act under experimental control, it would also tend to encourage creative thinking. In general, whether a particular restriction enhances or inhibits creativity can be determined empirically. Another useful technique is to suspend or delay the application of expert knowledge. This is done in the spirit of trying to avoid conventional mental sets. A subject might be asked to generate a mental image of a structure without initially knowing what purpose the structure is supposed to have. An interpretive category, such as "furniture" or "appliances" can then be specified, and the resulting

258

Ronald A. Finke

product can be compared to those generated when the interpretive category was known in advance. Such methods can help to clarify the extent to which ambiguous mental images can be reinterpreted (Chambers, this volume; Chambers & Reisberg, 1985;Reed, 1974), and the extent t o which one must prepare for creative exploration (Intons-Peterson, this volume). A general procedure that will be used in the following studies is to have subjects mentally synthesize novel objects using a set of randomly chosen object parts. The mental images that result often exhibit emergent properties which can lead to new and unexpected discoveries (Finke, Pinker, & Farah, 1989;see also Roskos-Ewoldsen, this volume). This procedure is in the spirit of previous recommendations for engaging in "combinational play" as a creative strategy (e.g., Getzels & Csikszentmihalyi, 1976).

Evaluative Considerations The resulting products are evaluated for creativity by having judges rate the originality and practicality of inventions, and the originality and sensibility of abstract ideas. It does not matter whether the judges are always correct in regarding something as original, practical, or sensible; it is more important that the same criteria are applied across all experimental conditions, and that one counts as "creative"only those ideas or inventions for which there is reasonable agreement among the judges (e.g., Amabile, 1983). This raises the issue of the reliability of the judges' ratings. High reliability should be expected when experts in a field serve as judges, and the creative products pertain only to that particular field. However, in studies where the creative ideas can pertain to any number of interpretive categories, as in the studies to be reported in this chapter, the preferred method is t o use judges who are not necessarily experts, but who are practiced in evaluating products and ideas across a wide range of categories. In this case, because the reliabilities have been lower (generally in the range of .5 t o .6),one would want to classify on the basis of consensus, where the judges would have to agree on most of the ratings. Classifj.ing a product as creative in terms of both its originality and its practicality or sensibility has the advantage that judges need not rate creativity per se, which avoids possible problems associated

Mental Imagery and Creative Discovery

259

with the different senses in which something might be regarded as being "creative" (see Anderson & Helstrup, this volume). For example, what one person means in calling an engineering solution "creative" may be quite different from what another means in referring to a literary idea as 'kreative." A further consideration is that the judges, as well as the experimenters, should be naive with regard to the expected results of the study (Intons-Peterson, 1983). The judges should also be unaware of the particular conditions from which the products were taken, because ratings from different conditions are often compared when analyzing the results. An additional issue concerns the construct validity of these classifications. Most of the ideas and inventions that were classified as "creative" in these studies would seem to have construct validity, in that they tended to be perceived as creative by others. In fact, many people openly commented on how wonderfully creative they seemed to be, and wished they had thought of the ideas themselves.

Experimental Studies on Creative Invention All of the experiments reported in this chapter are taken from Finke (1990),and are based on ideas developed in an earlier study by Finke and Slayton (1988). They illustrate many of the methods just described. The first set of experiments address how mental imagery can be used to generate new ideas for creative inventions.

A Paradigm for Invention The experiments began by showing subjects a set of fifteen object parts, which are presented in Figure 1. Three parts were selected for each trial; in most of the experimental conditions, this was done randomly, as in Finke and Slayton (1988). This random selection was subject to the constraint that the simplest parts (the sphere, half sphere, cube, cone, and cylinder) had a 50% chance of being selected, the more specialized parts (the rectangular block, wire, tube, bracket, and flat square) had a 33% chance of being selected, and the most specialized parts (the hook, cross, wheels, ring, and handle) had a 16% chance of being selected. This reduced the

Ronald A. Finke

260

0 D

A

6 9

_------

3

x

Figure 1 Set of object parts in experiments on creative invention. The following names were used to designate each of the parts: sphere, half sphere, cube, cone, cylinder, rectangular block, wire, tube, bracket, flat square, hook, cross, wheels, ring, and handle. (From Finke, 1990.)

likelihood that the stimulus sets would consist entirely of the most specialized parts.

Mental Imagery and Creative Discovery

261

At the beginning of each trial, the experimenter named the three parts and the subjects were instructed to close their eyes and to imagine combining the parts to make a practical object or device. They were never told to try to be creative in doing the task; simply to think of an object that might have some practical value. All three of the parts had to be used, even if the same part was named more than once. The subjects could vary the size, position, or orientation of any part, but they could not bend or deform the parts, with the exception of the wire and the tube, which had been defined as bendable. The parts could be put inside one another; they could be hollow or solid, and could be made out of any material, including wood, metal, plastic, rubber, and glass, in any combination. The objects had to belong to one of the eight general object categories shown in Table 1, which allowed for a broad range of possible inventions. Depending on the experiment, these categories were specified randomly or could be chosen by the subjects. The subjects were given two minutes to discover an object using the designated parts; they were then to name the object, to draw it, and to describe what it did or how its parts functioned. These reports were then used in scoring the inventions. This scoring was accomplished by having judges rate the practicality and originality of each invention using a 5-point scale, in which ratings of "4" and "5" corresponded to ''practical'' and ''very practical," or "original" and "very original." These ratings were to be made independently; for example, an object could be practical but not original, original but not practical, or both. In addition, the judges were instructed to base their ratings on the concept represented by the drawing, not on how well the object was drawn, and on the overall design of the object, not on whether it necessarily contained all of the working parts it would actually need. All of the inventions were rated together, in random order and without knowledge of the experimental conditions. The practicality and originality ratings were then used to classify the objects into invention categories. If an object received an average practicality rating of at least 4.5, it was classified as a "practical invention"; if a practical invention received an average originality rating of at least 4.0, it was further classified as a

Ronald A. Finke

262

TABLE 1 Allowable object categories in experiments on creative invention. Category

Jkamdes

1.

Furniture

Chairs, tables, lamps, etc.

2.

Personal Items

Jewelry, glasses, etc.

3.

Transportation

Cars, boats, etc.

4.

Scientific Instruments

Measuring devices, etc.

5.

Appliances

Washing machines, toasters, etc.

6.

Tools & Utensils

Screwdrivers, spoons, etc.

7.

Weapons

Guns, missiles, etc.

8.

Toys & Games

Baseball bats, dolls, etc

Note: From Finke (1990).

"creative invention." These conventions ensured that the classifications would reflect consensus among the judges. The cutoff level for the practicality ratings was set higher than that for the originality ratings because the average practicality ratings were higher; also, this made the classificationsmore conservative. Exactly where these cutoffs are made is not crucial, however, because the major findings of these studies are based on the relative numbers of creative inventions across the different experimental conditions.

Restricting Object Parts and Categories The initial experiments tested the idea that creativity might be enhanced whenever one is forced to use unusual sets of parts or to interpret the resulting objects in unconventional ways. This was accomplished by varying whether the parts and the object categories could be freely chosen or were randomly specified at the beginning of each trial. There were three conditions of interest, in which (a) the category was random and the parts were chosen, (b) the category was chosen and the parts were random, and (c) both the category and

Mental Imagery and Creative Discovery

263

the parts were random. The other logical condition, in which both the category and the parts could be freely chosen, was not included, because it would have been trivially unconstrained. Each subject participated in 6 trials, and there were 60 subjects per condition, As shown in Table 2, the relative number of creative inventions increased significantly as the task became more highly constrained, with the greatest number of inventions occurring when both the object parts and the interpretive categories were randomly specified. This suggests that one is more likely t o generate a creative invention whenever one is forced to be resourceful in thinking about possible objects and their uses.

TABLE 2 Number of practical and creative inventions according to whether object parts and categories were freely chosen. 5 P e of Invention

Condition

categnryParts Chosen

categnrychoaen Parts Random

cwe?PYRMdam Parts Random

Practical

193

191

175

Creative

17

31

49

Note: The above categorizations are based on a total of 1080 trials, 360 for each condition. Treative" inventions were practical inventions that were rated as original. From Finke (1990).

In this last condition, subjects were able to discover a practical invention on 48.6% of the trials, and a creative invention on 13.6% of the trials. This is quite remarkable, taking into account that the subjects were never told to be creative, that they were not preselected with regard to creative ability, and that they had only two minutes to perform the task. An example of an object that was classified as a creative invention is presented in Figure 2. This was a "hip exerciser" that one subject mentally constructed using the half sphere, the wire, and the rectangular block as the three parts, and which belonged to the

264

Ronald A. Finke

category "furniture."To use the hip exerciser, you attach the wire to the walls on opposite sides of a room, stand on the base, and then rotate your hips by shifting your weight from side to side while holding onto the post.

Figure 2 The "hip exerciser," constructed using the half sphere, wire, and rectangular block, an example of an object that was classified as a "creative invention." By shifiing one's weight from side to side while standing on the half sphere, one can exercise one's hips. (From Finke, 1990.)

Figure 3 presents another example of a creative invention, the

Mental Imagery and Creative Discovery

265

"shoestring unlacer," generated within the category "personal items." To untie a knotted shoelace, one inserts the hook onto the knot and pulls out one of the loops. The opposite end of the unlacer can be used as a key chain.

Figure 3 The "shoestringunlacer,"constructed using the rectangular block, hook, and handle. This device can be used to untie a knotted shoe lace. (From Finke, 1990.1

Virtually all the creative inventions that were obtained in these and later experiments were unique. It was extremely rare for the same invention to be discovered coincidentally by different subjects, even when they had used the same sets of parts, and were given the same object category. The subjects seemed genuinely interested in pursuing their

266

Ronald A. Finlze

discoveries, even after the experiment was concluded. Many felt that, as a result of their participation, they had learned something important about how they could be more creative. Some even asked whether they were permitted to patent their ideas. Preinventive Forms When interviewed, most of the subjects reported a preference for using an exploratory strategy, in which they began by imagining interesting, suggestive forms and then trying to interpret them afterwards - as opposed to starting out with a particular concept in mind, and then trying to imagine an object that conformed to that concept. The subjects often mentioned that they would try to look for useful properties of the forms they created, and to consider novel ways in which the forms might function as workable devices. These reports suggested that it might be better t o construct an image without knowing in advance what the image was supposed to represent.

Generating and Interpreting Preinventive Forms The general method in the next experiment was to have the subjects deliberately start out by imagining forms that were not supposed to represent anything specific. Rather, the forms were simply to be interesting and potentially useful in a very general sense. These imagined forms will be referred to as preinuentiue forms. The subjects were given one minute to mentally synthesize their preinventive forms using the three randomly specified parts. This turned out to be relatively easy, as the subjects were able to generate a preinventive form of some sort on virtually every trial. Examples of their preinventive forms are shown in Figure 4. After drawing their forms, which they did at this time so that they could not change them afterwards, they were then given an object category, selected at random, and were told to interpret the forms as a practical object or device within that category. They were given one minute to provide their interpretations. In this case, note that the subjects were committed to using

Mental Imagery and Creative Discovery

267

Figure 4 Examples of preinventive forms. (From Finke, 1990.)

their preinventive forms before they knew anything about how they would have to interpret them. The results showed that this was the most successful condition of all, resulting in the greatest number of creative inventions, 65, out of the 360 trials. In addition, two-thirds of the subjects had been able to generate at least one creative invention. The next two figures present examples of creative inventions that were based on subjects' interpretations of their preinventive forms. The form shown in Figure 5 was interpreted as a 'Icontact lens remover,'' within the category "personal items." To use the remover, one places the rubber cone against the contact lens, covers the back hole with a finger, sealing off the air, and then lifts the contact off by moving the remover away from the eye. Air pressure

268

Ronald A. Finke

Figure 5 The "contactlens remover,"constructed using the half sphere, cone, and tube. One places the rubber wne against the contact lens, covers the back of the tube with a finger, lifts the wntact off the eye, and then removes the contact from the cone by releasing the finger from the tube. (From Finke, 1990.)

keeps the lens on the remover until the finger is lifted. In Figure 6, the invention shown is the "universal reacher," belonging to the category "personal items." This device can be used t o retrieve keys and other possessions that fall into hard-to-reach places. The wire is drawn out of the sphere and can be bent and

Mental Imagery and Creative Discovery

269

shaped to reach the lost item. The hook allows the device to be secured when both hands are needed to guide the wire. These findings indicate that the likelihood of producing a creative invention is increased when one initially generates a preinventive form without considering the specific ways it might be

Figure 6 The "universalreacher," constructed using the hook, sphere, and wire. The wire is drawn out of the sphere and can be shaped and bent to retrieve things that fall into hard-to-reach places, while the hook allows the device to be secured so that both hands can be used to guide the wire. (FromFinke, 1990.)

interpreted. This suggests an alternative strategy to the usual "form follows function" approach to invention and design, a topic that will be discussed further at the end of the chapter.

Using Preinventive Forms Generated by Others The next experiment explored the issue of whether it is better to use one's own preinventive forms. There are probably good reasons why a person chose to construct one form rather than another, having to do with the person's background, intuitions, and perhaps aesthetic sense. These may not be shared with anyone else; hence a preinventive form that can be readily interpreted by one person may not be so easily interpreted by another. This experiment used a procedure similar to that of the previous experiment, except that the subjects were now given preinventive forms that others had generated, instead of constructing the forms themselves. The subjects were first given an opportunity to become familiar with the forms, rating them on several measures, such as

Ronald A. Finke

270

how well the forms were drawn and how useful they were likely to be, and were then given the object categories at random and were instructed to interpret the forms in the same manner as the previous subjects. Table 3 compares the results of this experiment to those in which the subjects used their own forms. As this table shows, the present subjects came up with far fewer creative inventions, implying that successful performance on the task involves more than simply providing creative interpretations of novel forms. Apparently, it is also important that the forms be structured in a personally meaningful way.

TABLE 3 Number of practical and creative inventions according to whether preinventive object forms were generated or provided. 5 P e of Invention

Condition SuGect Category Chosen

Subject Category

Practical

120

69

Creative

65

20

Random

Note: The above categorizations are based on a total of 720 trials, 360 for each condition. "Creative" inventions were practical inventions that were rated as original. From Finke (1990).

A possible criticism is that subjects may have done better in the previous experiment simply because they had been able to practice generating their preinventive forms. Although this was not directly tested, it is an unlikely explanation given the absence of practice effects in any of the previous creative invention experiments. The advantage of using preinventive forms that you have generated yourself is related to the concept of "preparation" in creative thinking (Intons-Peterson, this volume; Wallas, 1926). Perhaps generating one's own preinventive forms plays an important

Mental Imagery and Creative Discovery

2 71

role in this preparation stage, where one begins to acquire a general sense or "feel"for the kinds of potential applications the forms might have, or the particular ways in which they might be related to one's own past experiences. This may help to explain why a preinventive form that one person constructs may not be so useful or inspiring to others. Removing Time Restrictions

The findings of these experiments are particularly striking given that subjects had only one minute to interpret their preinventive forms. With extended time to explore their preinventive forms, subjects nearly always discover a potentially useful invention or idea. In Finke (1990),I reported numerous examples of inventions that were generated in this way, where the exploration times ranged from less than a minute t o as long as fifteen minutes. Although imposing time restrictions helps to make a task empirically manageable, it also underestimates the actual potential of these creativity techniques. Spanning the Object Categories Using a Single Preinventive Form

Is it possible to reinterpret the same preinventive form as different types of inventions? In a pilot study conducted in collaboration with Ratliff and McKeown, subjects first generated a preinventive form and were then given the eight object categories, in random order. Their task was to interpret the form as a practical object or device within each of the categories, without time restrictions. Examples of the different types of inventions that were discovered using a single form are shown in Figure 7. Although exploratory, this study suggests that it is indeed possible to reinterpret the same preinventive form in a variety of different ways. In conducting experiments on preinventive forms, my colleagues and I have often encountered something we call the illusion of intentionality (Finke, 1990): The preinventive forms appear to have been conceived with the particular inventions in mind, whereas in actuality the inventions were inferred only after the preinventive forms were constructed. In looking again at the previous examples of inventions that were derived from preinventive forms, one might think that the forms were designed t o satisfy the functional requirements of those inventions. Yet the preinventive forms were

2 72

Ronald A. Finke

.

Figure 7 Multiple interpretationsof a single preinventive form, spanning the eight object categories as follows: "lawnlounger"(furniture),"global earrings" (personal items), "water weigher" (scientific instruments), "portable agitator" (appliances), '"watersled" (transportation), ''rotating masher" (tools and utensils), "ringspinner"(toys and games), and "slasherbasher" (weapons). (From Finke, 1990.)

Mental Imagery and Creative Discovery

2 73

constructed without foreknowledge of what the particular category would be. That the forms appear to have been designed for a specific purpose is an illusion.

Creative Concepts The previous experiments explored how the generation and interpretation of preinventive forms could give rise to creative insights for new inventions and designs. The next section considers extensions of the studies to cases where the forms are to be interpreted as representing an abstract concept.

Extensions of the Methods to Conceptual Domains Instead of using object categories, the subjects were now to use subject categories, which are listed, along with examples, in Table 4. After generating their preinventive forms, the subjects were told that they should interpret their forms as representing an abstract idea or concept that pertained to the specified category. In this case, the preinventive forms were to be considered as symbolic or metaphorical representations of an idea, rather than as literal depictions of a concrete object. For example, if the subject category was ''physics and astronomy," the preinventive form might be interpreted as a new concept for how atoms are combined, or a new model for how the universe was formed. The subjects were told, specifically, that their preinventive forms need not actually look like any particular object, but could simply represent the concept or idea in a general or indirect way. The rating procedure was modified slightly, with the dimension of practicality being replaced by that of sensibility. As discussed earlier in the chapter, the reason for doing so is that an idea could be sensible without necessarily being practical. The procedure was otherwise identical to that for rating and classifjring the creative inventions. If a concept received an average rating of 4.5 on the 5-point scale, it was classified as a "sensible concept," and if a sensible concept received an average rating of 4.0 on the originality scale, it was further classified as a kreative concept."

Ronald A. Finke

274

TABLE 4 Allowable subject categories in experiments on creative invention.

Category

1. 2.

3. 4. 5.

6. 7.

8.

Architecture Physics & Astronomy Biology Medicine Psychology Literature Music Political Science

Examples Concepts in building design, etc. Models of the atom, universe, etc. Methods of animal survival, etc. Mechanisms of infection, etc. Theories of personality, etc. Writing styles, techniques, etc. Composition, instrumentation, etc. Forms of government, etc.

Note: From Finke (1990).

In the first experiment, the subject categories were chosen at random, and were given to the subjects only after they had generated their preinventive forms. This turned out to be more difficult than the previous tasks. The subjects were able to discover sensible concepts on 16.9% of the trials and creative concepts on 7.8% of the trials, showing that it is possible to interpret the preinventive forms as depicting abstract concepts, though with somewhat less success. Overall, a third of the subjects were able to come up with at least one creative concept. Figure 8 presents an example of a creative concept that was discovered in this experiment. This was the concept of "viral cancellation," belonging to the category "medicine," and represented using the tube, cross, and cube. The idea is that two viruses attempting to invade the same cell may cancel one another, curing or preventing the disease. Removing Restrictions on the Subject Categories In the next experiment, subjects were allowed to choose any of the eight subject categories after having generated their preinventive

Mental Imagery and Creative Discovery

275

Figure 8 The concept of "viral cancellation,"represented using the tube, cross, and cube. The idea is that two viruses attempting to invade a cell may cancel one another, curing or preventing the disease. (From Finke, 1990.1

forms. In view of the initial findings discussed in this chapter, this

Ronald A. Finke

276

should result in fewer creative ideas because restrictions on the categories, which tend to promote creative exploration, have now been removed. Alternatively, one might expect that even more creative ideas could be generated in this case because the subjects could now select those subject categories in which they would be most knowledgeable. For example, a physics major would be more likely to interpret a preinventive form as representing a concept in physics, compared to someone who knew very little about the subject and was given that category at random. Table 5 presents the results in comparison to those of the previous experiment, where the subject categories had been randomly specified. Over twice as many creative concepts were obtained when the subject categories were restricted, in agreement with the previous findings for restricting the object categories. These techniques thus appear to be most successful when people are forced to interpret their preinventive forms in ways that are less conventional, requiring deeper explorations and considerations.

TABLE 6 Number of sensible and creative concepts according to whether subject categories were freely chosen or specified randomly. Type of Concept

Condition Su&ect Category Chosen

Subject Category

Sensible

47

61

Creative

13

28

Random

_______~

~

Note: The above categorizations are based ona total of 720 trials, 360 for each condition. "Creative"concepts were sensible concepts rated as original. From Finke (1990).

Suspension of Expertise in Forming Creative Concepts Naturally, untrained subjects are unlikely to make major

Mental Imagery and Creative Discovery

277

conceptual contributions to a highly specialized field of study, such as physics or medicine. How, then, can one account for the finding that subjects do better when they cannot choose the subject category? It is important to distinguish between experimental situations in which subjects are free to choose from among any number of categories after generating their forms, and real-life situations where experts are trying to generate new ideas in a particular, chosen field. If one can arbitrarily select any category whatsoever when interpreting one's forms, one's interpretations are likely to be superficial and less creative, as suggested by the results of the last experiment. On the other hand, if one has the necessary expertise for exploring deeper conceptual implications, and is already committed to a particular subject, one would be less likely to shift the categories simply t o fit the forms. Instead, one would persevere to discover those deeper implications.

Conclusions and Implications Summary of Experimental Findings The studies presented in this chapter show that it is possible to conduct controlled experiments in which people can generate new ideas for inventions and creative concepts. The major results can be summarized as follows: First, when using mental synthesis to explore creative possibilities, the likelihood of discovering a creative invention is greater when the component parts are restricted. Second, restricting the interpretive category also increases the likelihood of discovering a creative invention. Third, it is helpful to generate preinventive forms before one knows what the interpretive category will be. Fourth, it is helpful to use preinventive forms that one actually generated. Finally, preinventive forms can function as catalysts for conceptual exploration and discovery in abstract domains as well. As in the case of creative invention, the likelihood of discovering a creative concept is greater when the interpretive category is specified after the preinventive form is generated.

Function Follows Form The results suggest an alternative to the standard "form follows

2 78

Ronald A. Finke

function" approach to invention and design. The opposite notion, ''function follows form," seems to characterize most of these findings. Subjects typically discovered that it was better to generate preinventive forms first and then consider their possible functions and interpretations, rather than trying to make the forms conform to their initial preconceptions. Accordingly, creativity was enhanced when the task was formally structured so that the forms were generated before the interpretive constraints were imposed. These results are consistent with earlier studies on the value of "problem finding" in creative thinking, in which one searches for a new problem or idea rather than focusing on how to solve an existing problem (e.g., Bransford & Stein, 1984; Perkins, 1981). Problem finding also appears to be an important process in artistic creation; Getzels and Csikszentmihalyi (1976) found that artists produced works that were rated as higher in quality when they did not start out with a definite plan in mind, but instead, were concerned with exploring and discovering emergent structures and forms. It should be acknowledged, however, that the preinventive forms that were generated in the present studies represent only the initial phase of the creative process. Ordinarily, a considerable amount of refinement and restructuring is needed to transform an inventive concept into a working prototype. As the concept matures, the preinventive forms become more specialized, with possible directions for further exploration and refinement becoming more restricted.

Preinventive Forms as Catalysts for Creative Insight How are preinventive forms explored, once generated? These experiments offer some insight into the kinds of exploratory strategies that are most useful. The subjects reported that they often searched for interesting or novel features in the forms, imagined themselves actually using the forms, and mentally transformed the context in which the forms might occur. For example, if a preinventive form were to be interpreted as a tool, a subject might look for suggestive features of the tool, imagine holding it in his or her hand, and seeing how it might be used in different contexts or situations. A useful analogy is to consider how one might go about trying to discover the function of an unusual object or artifact. Although

Mental Imagery and Creative Discovery

2 79

initially one might have no idea what the object is, one can often gain considerable insight into its possible functions by imagining various ways in which the object might have been used. One could imagine, for example, using the object as a construction tool, a measuring device, or a utensil. This is similar t o how one might go about trying to discover the underlying meaning of an ambiguous preinventive form. (See also the discussion of recognizing emergent properties in Roskos-Ewoldsen, this volume.)

Generation and Exploration: The Geneplore Model

A model that can account for many of these findings is the Geneplore model, proposed by Finke, Ward, and Smith (in press), which considers both generative and exploratory cognitive processes. The overall structure of the model is shown in Figure 9. In an initial, generative phase, one constructs mental representations called preinventive structures having various properties that promote creative exploration. A preinventive form would be one example of a preinventive structure; others include mental blends, conceptual combinations, and mental models. Preinventive structures have various properties that stimulate creative insight and discovery; among these are novelty, ambiguity, and the emergence of unexpected features. These properties are then exploited during an exploratory phase, in which one seeks t o interpret the preinventive structures in meaningful ways. Preinventive structures can be thought of as internal precursors to the final, externalized creative products, and would be generated, regenerated, and modified throughout the course of creative exploration. If one's initial explorations are successful, the initial preinventive structure may lead directly to a creative product. However, if these explorations are unsuccessful, one of two procedures would come into play, either of which would involve a return to the generative phase. One is to abandon the initial preinventive structure and generate another that may be more promising. Another is to modify the initial structure and then repeat the exploratory phase with this modified structure. Using either procedure, one would eventually "focus" the preinventive structure on particular themes or problems or "expand'' the structure to explore more general conceptual possibilities.

Ronald A. Finke

280

GENERATION

PREINVENTIVE EXPLORATION

PREINVENTIVE STRUCTURES

INTERPRETATION

FOCUS OR EXPAND

CONSTRAINTS

Figure 9 The basic structure of the Geneplore model. In the generative phase, one constructs mental representations called preinuentiue structures. These structures have various emergent properties that are then exploited for creative purposes in the exploratory phase. The resulting creative concepts can then be focused or expanded according to task requirements or individual needs, by modifying the preinventive structures and repeating the cycle. Constraints on the final product can be imposed a t any time during either the generative or exploratory phase. (From Finke, Ward, & Smith, in press.)

This alternation between generation and exploration typically occurs when people engage in creative thinking. For example, a

Mental Imagery and Creative Discovery

281

person might construct a mental image in the generation phase that produces an interesting form, and then interpret the form as representing a new product in the exploratory phase. However, further examination of the form may lead to the conclusion that the form is incomplete in some respects. A modified form would then be generated by constructing another image or altering the existing one. This may result in a form that represents an improved or more complete design, or may lead to completely new and unanticipated interpretations of the form. The Geneplore model also considers constraints on the creativeproducts, and how they might affect the underlying cognitive processes. As shown in Figure 9, these constraints can be imposed either during the generative phase or the exploratory phase, depending on the nature of the task. This allows for both "form follows function" and "function follows form" strategies in creative thinking and design. There- are a number of implications of making a general distinction between generative and exploratory processes, as in the Geneplore model. First, it allows for the possibility that people can be creative in different ways. Some may be more skilled at generating preinventive structures, whereas others may be more skilled at interpreting them. This may help to explain why there are often dramatic individual differences in creative style. Consider, for example, the often-noted contrast between Mozart's and Beethoven's styles of composition (see Ghiselin, 1952; Perkins, 1981). Whereas Mozart generated compositions that seemed complete and fully formed in their initial creation, Beethoven revised his compositions extensively, constantly seeking new creative possibilities. Considering generation and exploration as separate aspects of the creative process also has implications for the issue of when t o suspend expertise. As the findings in this chapter suggest, it may be better to suspend expertise when generating preinventive structures, and apply it later during the exploratory phase. This is not to say that preinventive structures must be independent of expert knowledge; merely that, as a general rule, one might want to avoid deliberately imposing such knowledge onto the structures when they are initially being formed. Another implication of distinguishing generative and exploratory processes is that it allows the model to be applied at many stages of

Ronald A. Finke creative thinking. For instance, it can apply not only to early stages, where one might want to be creative in a global sense, and where the preinventive structures would tend to be relatively unconstrained, but also to later stages, where one might be close to solving a particular problem or completing a new design, and where the preinventive structures would tend to be highly constrained. As preinventive structures evolve, they can be focused more specifically on particular issues or problems, or be extended in new directions, depending on the person’s interests and intentions.

Practical Implications The methods developed in this chapter have a number of practical uses, especially for inventors and engineers. They suggest, for example, new ways in which one might attempt to come up with novel concepts for an invention or device. Because preinventive forms are assembled in a natural, instinctive way, there is beauty and coherence to their structure which can often translate into inventions that are both aesthetically appealing and innovative. These methods could thus have important implications for the development of new, innovative products. Inventions that result from interpretations of preinventive forms may have greater intrinsic fascination and appeal; these qualities would very likely enhance the marketability of the resulting creations. In fact, when people see examples of objects that were rated as highly creative inventions in these experiments, they often comment that they would be very interested in owning such a product. In addition, these studies suggest new ways for generating creative ideas using computers. For example, one might program a computer to first construct a preinventive form, and to then systematically explore the possible interpretations of the form. Conceivably,this could lead to the development of new, more creative forms of artificial intelligence (e.g., see Boden, 1991; Johnson-Laird,

1988). Professional scientists could also learn to use preinventive forms to enhance conceptual discovery. A physicist, for example, might generate a variety of preinventive forms and then try t o interpret them as representing new concepts in atomic theory or relativity. A medical researcher could explore preinventive forms to come up with

Mental Imagery and Creative Discovery

283

new ideas for how to treat a disease. At present, creativity is seldom emphasized in training people to become scientists. For example, many creative insights in physics have resulted from so-called Gedanken, or "thought" experiments (in which one envisions a hypothetical, often paradoxical situation and considers its consequences), yet physics students are rarely taught how to do this. These methods could therefore have important implications for improving scientific creativity.

The Possibility of Universal Preinventive Forms Could it be that certain types of preinventive forms have a "universal" quality, in that they give rise to creative insights across a variety of domains and applications? For example, could the same preinventive form inspire new types of mechanical designs, new concepts in physics, and new artistic styles? If so, this would help support the claim that there are often connections between the structure of artistic creations and fundamental concepts in the physical sciences (e.g., Gardner, 1982). Certain types of structures may inspire creative insights that are both aesthetically appealing and theoretically significant. In exploring this possibility, it would be useful to contact notable artists, scientists, and designers to see if these creativity methods can be applied in actual professional settings. Up to now, these methods have been used only in the context of controlled laboratory experiments, with untrained undergraduate students as subjects. By having experts generate and interpret preinventive forms, there is the very real possibility that significant artistic and scientific discoveries could result. References Amabile, T. M. (1983). The social psychology ofcreativity. New York: Springer-Verlag. Anderson, J. R. (1990). Cognitive psychology and its implications. New York Freeman. Boden, M. (1991).The creative mind: Myths and mechanisms. New York: Basic Books.

284

Ronald A. Finke

Bransford, J. D., & Stein, B. S. (1984). The ideal problem solver. New York: Freeman. Chambers, D., & Reisberg, D. (1985). Can mental images be ambiguous? Journal of Experimental Psychology: Human Perception and Performance, 11, 317-328. Cooper, L. A. (1990). Mental representation of three-dimensional objects in visual problem solving and recognition. Journal of Experimental Psychology: Learning, Memory, and Cognition, 16, 1097-1106. Finke, R. A. (1986). Mental imagery and the visual system. Scientific American, 254, 88-95. Finke, R.A. (1989). Principles of mental imagery. Cambridge, MA: MIT Press. Finke, R. A. (1990).Creative imagery: Discoveries and inventions in visualization. Hillsdale, N J Erlbaum. Finke, R. A,, Pinker, S., & Farah, M. J. (1989).Reinterpretingvisual patterns in mental imagery. Cognitive Science, 13, 51-78. Finke, R. A,, & Shepard, R. N. (1986). Visual functions of mental imagery. In K. R. Boff, L. Kaufman, & J. Thomas (Eds.), Handbook of perception and human performance (Vol. 2, Chapter 37,pp. 1-55). New York: Wiley-Interscience. Finke, R. A., & Slayton, K. (1988). Explorations of creative visual synthesis in mental imagery. Memory & Cognition, 16, 252-257. Finke, R. A., Ward, T. B., & Smith, S. M. (in press). Creative cognition: Theory, research, and applications. Cambridge, MA: MIT Press. Gardner, H. (1982).Art, mind, and brain: A cognitive approach to creativity. New York: Basic Books. Getzels, J. W., & Csikszentmihalyi, M. (1976). The creative vision: A longitudinal study of problem finding in art. New York: Wiley. Ghiselin, B. (1952). The creative process. Berkeley, C A University of California Press. Intons-Peterson, M. J. (1983).Imagery paradigms: How vulnerable Journal of are they to experimenters’ expectations? Experimental Psychology: Human Perception and Performance, 9,394-412. Johnson-Laird, P. N. (1988). The computer and the mind: An

Mental Imagery and Creative Discovery

285

introduction to cognitive science. Cambridge, MA: Harvard University Press. Koestler, A. (1964). The act of creation. New York: Macmillan. Kosslyn, S. M. (1975).Information representation in visual images. Cognitive Psychology, 7, 341-370. Kosslyn, S. M. (1980). Image and mind. Cambridge, U Harvard University Press. Miller, A. I. (1984).Imagery in scientific thought. Cambridge, MA: MIT Press. Perkins, D. N. (1981). The mind$ best work. Cambridge, MA: Harvard University Press. Pinker, S. (1984).Visual cognition: An introduction. Cognition, 18, 1-63. Reed, S. K. (1974). Structural descriptions and the limitations of visual images. Memory & Cognition, 2,329-336. Shepard, R. N. (1978).Externalization of mental images and the act of creation. In B. S. Randhawa & W. E. Coffman (Eds.), Visual learning, thinking, and communication, pp. 133-189. New York: Academic Press. Shepard, R. N. (1988).The imagination of the scientist. In K. Egan & D. Nadaner (Eds.), Imagination and education for intercultural understanding, 153-185. New York Teachers College Press. Shepard, R. N., & Cooper, L. A. (1982). Mental images and their transformations. Cambridge, MA: MIT Press. Shepard, R. N., & Metzler, J. (1971). Mental rotation of threedimensional objects. Science, 171,701-703. Wallas, G.(1926). The art of thought. New York: Harcourt, Brace, and World.

This Page Intentionally Left Blank

Imagcry, Creativity, and Discovery: A Cognitive Pcrspcctive 13. Roskos-Ewoldson, M.J. Intons-Peterson and R.E. Anderson (Editors) 0 1993 Elscvier Science Publishcrs B.V. All rights rcscrvcd.

287

Chanter 10

IMAGERY AND

DISCOVERY

Stephen K.Reed Department of Psychology San Diego State University San Diego, CA 92182 USA Robert McKim (1980) begins his book, Experiences in Visual Thinking,with the observation that visual thinking can be found in all forms of human activity, from the abstract and theoretical t o the down-to-earth and everyday. A motorist driving along the highway, an interior decorator designing a room, a football coach outlining a new play, and a surgeon planning an operation are all thinking visually. Visual thinking, according to McKim, consists of three kinds of interrelated activities: seeing, imagining, and drawing. Discoveries from direct perception are common in science, such as Alexander Fleming's observation that bacteria around one of his cultures had died. Although many scientists would have failed to act on this observation, Fleming realized the importance of his observation and used it to discover penicillin. Imagery has also contributed to many creative discoveries. Nikola Tesla, the inventor of the fluorescent light and the AC-generator, could project before his eyes a detailed picture of every part of a machine, enabling him to make complex designs without drawings. Because most of us lack imagery powers of this magnitude, drawings and sketches help clarify our visual thoughts. James Watson (1968) remarks in his book, The Double Helix, that an important idea about the structure of DNA came while he was drawing the rings on paper. McKim represents the interaction among seeing, imagining, and drawing as three overlapping circles to symbolize that visual thinking is most effective when people can easily move from one activity to another. These reports of scientists about their discoveries indicate that

288

Stephen K.Reed

visual imagery has contributed to important insights. Shepard (1988)has also studied the self-reports of famous scientists and found that imagery helped them achieve some of their major discoveries. Scientists such as Einstein, Maxwell, Tesla, Watson, and Hawking are among those whose major discoveries were aided by the ability to create novel spatial representations of a problem. For instance, Shepard reports that Einstein's discoveries in theoretical physics depended on soaring leaps of spatial and physical intuition. Einstein, himself, claimed to have achieved his insights about space and time by means of thought experiments in which he mentally pictured the relative motion of light waves and idealized physical bodies. The reported use of imagery in everyday and in scientific thought has motivated relatively few laboratory investigations of how imagery promotes creative thinking, current contributions notwithstanding. My goal in this chapter, therefore, is to explore when our use of imagery is likely to help us discover new information. The first part of the chapter reviews previous findings that illustrate how images may fail us, either because they are ineffective, or because they are effective but insufficient for discovering new information. I then want to summarize several experiments that show the successful use of imagery to synthesize new information. The second part of the chapter contains four recommendations that should help us increase our understanding of the creative use of imagery. First, we need to better understand how illustrations and diagrams facilitate problem solving. Second, we need to extend research on imagery to include interesting problem-solving tasks. Third, we need to determine how imagery can lead to successfid problem representations that differ from the more standard, symbolic representations. Fourth, we need to extend the study of self-reports to include a wider range of creative insights, such as those obtained by architects. I will return to these recommendations after summarizing some of our current knowledge about the role of imagery in discovery.

Imagery and Discovery

289

When is Imagery Useful Let’s begin by reviewing past research, searching for clues for when imagery can lead t o new insights. I will first review several studies showing the limitations of imagery, because it is either ineffective or insufficient, and will then review several studies showing the successful use of imagery. Imagery is Ineffective My initial research on visual imagery was motivated by an interest in determining whether people could discover new parts in their images (Reed & Johnsen, 1975). We selected parts of patterns that we thought would be relatively easy to find if the pattern were physically present, as in the standard embedded-figures test. Figure 1 shows some of the patterns and parts used in our research. But we also wanted to test how well subjects could find the same parts when they had to examine an image of the pattern. To compare these two conditions we randomly assigned subjects to either a perception condition or an imagery condition. In the perception condition we showed a possible part, followed by the complete pattern. Subjects searched the pattern and indicated whether it contained the part at the end of a 10-second interval. In the imagery condition the pattern occurred before the part, so the search for the part would require using a visual image of the pattern. When subjects responded positively in the imagery condition, we asked them whether they had seen the part when shown the pattern, or whether they had discovered the part in their image. Our results showed that subjects failed to detect the part on 14% of the trials in the perception condition and on 48% of the trials in the imagery condition. When we eliminated those image trials in which subjects recognized the part before examining their image, the percentage of undetected parts increased to 72% in the imagery condition. These findings suggest that it is difficult to use an image to reanalyze a description of a pattern and discover new information. Further evidence of this difficulty was provided by Chambers and Reisberg (1985)in their study of whether people could reinterpret an ambiguous figure representing either a duck or a rabbit. Subjects were unable to reinterpret the figure from a visual image, but could

Stephen K. Reed

290

readily find the alternative interpretation after drawing the figure. Chambers and Reisberg reported that failure to reverse images occurs despite hints, coaching, and a moderate amount of training.

XA0 A QD> 18

1

2

1A

2A

10

1c

1E

2c

28

20

2E

Figure 1 Patterns and parts used by Reed and Johnson (1975).

Results showing the limitations of imagery should challenge us to find what constrains our ability to reinterpret images. Recent work reported in this volume (Peterson, this volume; RoskosEwoldsen, this volume) shows how the perceptual characteristics of patterns constrains the discovery of new information. Peterson’s research studied how the processes of shape recognition influence people’s ability to reinterpret ambiguous patterns. She proposed that the ducWrabbit figure might be particularly difficult to reinterpret because it requires a reference frame reversal - the front of the duck is the back of the rabbit. Her subjects were more successful in using an image to reinterpret an ambiguous figure (snail-elephant) that did not require a reference frame reversal. Other investigators have studied how perceptual variables influence the discovery of parts in patterns or images. Roskos-

Imagery and Discovery

291

Ewoldsen (this volume) proposed that people should be more accurate in detecting parts when the parts have high figural goodness and the patterns have low figural goodness. She reasoned that figurally good patterns are more coherent and are therefore harder to reorganize to discover new information. The results provided partial support of her hypotheses and depended on whether subjects were examining images or perceptual patterns. Good parts enhanced accuracy only for the image group and poor patterns enhanced accuracy only for the perception group. These findings reminded me of a correlational study that Angaran and I did to determine which perceptual variables most strongly correlate with the difficulty of detecting an embedded figure (Reed & Angaran, 1972). One of the best predicting variables was what we called analysis complexity, defined as F + G - P,where F is the complexity of the (embedded) figure, G is the complexity of the ground, and P is the complexity of the (complete) pattern. Our reasoning was that detecting an embedded figure should be difficult to the extent that the complexity of the parts exceeds the complexity of the whole. The correlation between analysis complexity and the time needed to find an embedded figure was .71 for mentally retarded children, .62for children in lower elementary school, .61for children in middle elementary school, and .57 for children in upper elementary school. Discovering new information, in general, may be much more difficult when one has to reorganize information that is already well organized. For instance, Mayer (1983) discussed this topic in the context of Tresselt and Mayzner’s (1966) study on reorganizing anagrams to make words, The letter transition probability (LTP) of anagrams is determined by how frequently pairs of letters occur in the English language. Low LTP anagrams, such as rhtue, consist of successive letter pairs that don’t frequently occur in language. High LTP anagrams, such as ahter, consist of successive latter pairs that do frequently occur. Tresselt and Mayzner (1966)found that poorly organized (low LTP) anagrams were solved much more quickly than well organized (high LTP) anagrams. Notice that organization, in this case, is measured quite differently than the figural goodness of patterns.

292

Stephen K.Reed

Imagery is Effective but Insufficient A second limitation on using imagery to make new discoveries is that people may be able to effectively use their images to mentally scan patterns or simulate events, but more than imagev is required to discover new information. I shall argue that discovering functional relations between concepts is an example. A typical finctional relation is the relation between time, rate, and distance, Kosslyn, Ball, and Reiser (1978)studied this relation by measuring the time to scan between two objects in a visual image of an island. They found that scanning time was a linear function of the distance between the two objects when subjects imagined a dot moving at a constant rate from one object to another. Kosslyn and his colleagues interpreted this result as evidence that people can accurately scan visual images, but critics have argued that the result could be based on estimated time intervals if subjects have tacit knowledge of how rate and distance influence scanning time (Pylyshyn, 1981). Mitchell and Richman (1980)showed that people do know this relation because their estimated #can times were also linearly related to the distance between two objects. Hock, Lockhead, and I attempted to rule out the tacit knowledge explanation of mental scanning by designing an experiment in which people could not predict the results (Reed, Hock,& Lockhead, 1983). We measured scanning time of line configurations that varied in both length and shape. The shape was either a straight line, a spiral, or a maze consisting of lines meeting at right angles. Subjects in the perception condition were told to scan a pattern projected on a screen, and subjects in the imagery condition were told to scan an image of the pattern, following a 0.5 sec projection of the pattern on the screen. Scanning times for the perception (Figure 2a) and imagery (Figure 2b) conditions were virtually identical. In both conditions the rate of scanning differed significantly among the three shapes. In a second experiment we investigated whether people could predict their scanning rates. Following a 0.5 second presentation of each pattern, subjects estimated how long it would take to scan an image of that pattern. Figure 2c shows that the estimated scan times differed from the actual scan times, We concluded from these h d i n g s that people actually do mental scanning in a mental

Imagery and Discovery

"f 40

2 35$j

30-

58 2.5 -

E 205

15-

6-l 0

1.0

Q5

-

'25 2

Figure 2 Effeet of pattern configuration on (a) perceptual scan time, (b) imaginal scan time, and (c) estimated scan time. From Reed, Hock, & Lockhead (1983).

293

294

Stephen K.Reed

scanning task and that they do not have good tacit knowledge of how the shape of patterns influences their scanning rates. But does scanning images improve one’s knowledge of how different shapes influence scanning rates? If people make predictions, scan images, and then again make predictions, would their predictions become more accurate? Unfortunately, we did not do this experiment, but I would guess that the answer is no because learning how the shape of a pattern influences scanning time requires more than doing the task. One must also be able to (1) judge the scanning time for each pattern and (2) integrate the scanning times over different lengths to determine the correct functional relationship for each shape. Another variation of Kosslyn’s mental scanning procedure was included in a study by Intons-Peterson and Roskos-Ewoldsen (1989). They asked subjects to imagine transporting a weight (either a 3ounce balloon, a 3-pound ball, or a 30-pound cannonball) as they moved between pairs of buildings on a campus map. Imaginal-group subjects imaged a map that they had previously learned and visualmap subjects saw the map. Notice that if the carried weights influence the rate of scanning, the balloon, ball and cannonball should cause different slope differences, similar to those shown in Figures 2a and 2b. However, the obtained scanning times were more similar to the parallel curves shown in Figure 2c. For the imaginalmap group, the slopes of the reaction-time functions did not differ across weights, but the y-intercept of the cannonballwas significantly higher than the y-intercept of the other two objects. Subjects’ estimates of their scan times for a short and a long distance were also highly parallel across weights. The mean estimates for the short distance were 3.01seconds t o transport the balloon and 4.10 seconds to transport the cannonball - a differences of 1.09 seconds. The mean estimates for the long distance were 7.03 seconds for the balloon and 8.14 seconds for the cannonball - a difference of 1.11 seconds. The differences between objects stayed the same, rather than diverged, as the distance increased. These aspects of the results could be explained by subjects’ tacit knowledge because the scan times more closely matched the parallel functions of subjects’ estimates than the diverging functions that would result from different rates of scanning. But other aspects of the results could not be explained by tacit knowledge or by demand

Imagery and Discovery

295

characteristics of the experiment (Intons-Peterson, 1983). Subjects’ actual scan times were much slower than their estimated scan times. They also scanned the imagined map more slowly than the visual map, but the estimated times were the same for both maps. Although the concept of tacit knowledge was originally introduced by Pylyshyn (1981) to question whether imagery is a necessary theoretical concept, it is also useful to determine when imagery leads to the discovery of new information. People may effectively use imagery without discovering new knowledge because either the imagery produces results that correspond t o tacit knowledge or the difference between imagery and tacit knowledge does not lead to the revision of tacit knowledge. One constraint on attempting to improve our judgments about events by mentally simulating the events is that observation may not be sufficient to correct misconceptions, even when observation is based on perception. McCloskey and Kohl (1983) asked subjects to select the trajectory that a ball would follow if a string broke while someone was twirling the ball at a high speed in a circle. One group selected a trajectory from six trajectories drawn on paper. Another group selected a trajectory after watching computer simulations of the six trajectories. Contrary to expectations, the number of correct selections did not differ significantly for the two groups. Computer simulation of trajectories also failed to improve predictions of the path taken by a ball after exiting a curved tube. These results suggest that accurate mental simulation of the different trajectories would not be sufficient for selecting the correct trajectory. Mentally simulating the different paths would seem to be a reasonable imagery task. But, as suggested by the lack of effect of perceptual observation, accurate mental simulation of the alternative paths would likely be insufficient for improving predictions. If students did not improve their judgments after viewing the alternative trajectories, it is unlikely that they would improve their judgments after imagining alternative trajectories.

Imagery is Effective and Sufficient The previous analyses illustrate two ways in which imagery could disappoint us as a means to creative insights. First, our images will not always be sufficiently transformable to allow new

296

Stephen K.Reed

interpretations of their structure. Second, processing images successfully will not always provide sufficient information for creating new knowledge. 1 would now like to examine a different paradigm that offers greater promise for making new discoveries. The paradigm requires mental synthesis - combining parts to make a whole. An early demonstration of mental synthesis is Shepard and Feng's (1972)study of mental paper folding. The task required judging whether arrows on two squares would meet if the squares were folded into a cube, Response times increased linearly with the total number of squares carried along during the folds, suggesting that subjects were using visual images to mentally construct the cube. Another example of a perceptual synthesis task is Palmer's (1977)research on combining two spatially separated 3-segment parts to form a 6-segment line pattern. Palmer was interested in the structure of patterns rather than in visual images, but good visual imagery skills should be very helpful for this kind of task. He measured how long it took subjects to combine the two parts and then measured their error rate by having them judge whether their synthesis matched a test pattern. The error rate was fairly low, and response times varied from 1.5 seconds for high-goodness parts to 4.5 seconds for medium- and low-goodness parts. Cooper (1990)has recently found that students constructed a three-dimensional representation of complex objects to make judgments about the relation among the top, front, and side views. She gave engineering students two of the three views and asked them to judge whether a third view was compatible with the other two views. She later showed them pairs of three-dimensional objects and asked them to identify which object could be formed from the two-dimensional projections they had seen earlier. T w o aspects of her findings suggested that subjects had previously synthesized a three-dimensional object to make judgments about the compatibility of the two-dimensional views. First, they were fairly accurate in identifjing the correct object. Second, they were significantly more accurate in identifylng the correct object when they had been correct on the compatibility task (90% correct identifications) than when they had been incorrect on the compatibility test (72% correct identifications).

Imagery and Discovery

297

All of these mental synthesis tasks have a single correct answer. A single answer makes it easy to evaluate the accuracy of performance but makes it difficult to study creativity. The experiments described by Finke (1990) in his recent book on creative imagery allowed him to study creative synthesis by measuring how successfully people could combine parts to make recognizable objects or inventions. One interesting finding was that subjects created as many recognizable objects by using imagery as by using physical synthesis. Furthermore, mental synthesis of imagined parts resulted in significantly more recognizable objects than mental synthesis of perceived parts. Continuation of this approach should be particularly beneficial for increasing our understanding of the relation between imagery and creativity (Anderson & Helstrup, this volume; Finke, this volume). The results of mental synthesis tasks are impressive in demonstrating the effectiveness of imagery. Perhaps these findings are impressive because they require the synthesis of new knowledge rather than modifying previous knowledge. Finding a new part in a pattern, reinterpreting an ambiguous figure, and learning correct trajectories all require modifjrlng old beliefs. Changing one’s mind may be difficult, in general, regardless of the specific role of visual imagery. Training Studies

The results of the previously cited studies present a mixed picture of the effectiveness of imagery for discovering new information. Some results are encouraging, but others challenge us to find techniques that will enhance people’s ability to discover new information in their images. The experiments by Hyman and by Peterson (this volume) show that helpful instructions and appropriate training patterns can increase the number of successful reversals of ambiguous figures. For the ducwrabbit figure, Hyman told subjects to consider the back of the head in their current interpretation to be the front of the head for a different animal. For the chefldog figure, he told subjects to rotate their image. Peterson found providing training patterns that reversed in the same way as the test figure increased the number of reversals. Subjects who practiced on the goosehawk figure were

298

Stephen K. Reed

more successful than subjects who practiced on the chefldog figure in reversing the dudrabbit figure. Both the goosehawk and duck/ rabbit figures require reference frame reversals. These findings provide a good start, but we need to provide more extensive training of a variety of methods to promote successful transfer across a wide range of problems. Practice on a specific technique, such as reference frame reversal, will only help us make discoveries when reversing the reference frame creates interesting information. People need to learn a variety of different methods to manipulate images so they have alternative approaches for making discoveries. We still know very little about how well people can transfer a promising method from one problem to another. Most work on analogy has focused on the transfer of specific solutions rather than on the transfer of general methods. In contrast, Novick's (1990) recent work focuses on the transfer of methods, such as using a matrix to organize information in a problem. Her research may provide helpful clues about how to train methods that will facilitate the discovery of new information across a variety of problems. But we will still need to learn much more about imagery-based techniques to design effective training procedures. The remainder of this chapter discusses four recommendations.

Recommendations The previously discussed findings provide a starting point for further work on how imagery can aid the discovery of new information. We clearly have more to learn, so I would like to make several recommendations to guide this work. First, we need to better understand how illustrations and diagrams facilitate problem solving. Second, we need to extend research on imagery to include realistic problem-solvingtasks. Third, we need to determine how imagery can lead to successful problem representations that differ from the more standard, symbolic representations. Fourth, we need to extend the study of self-reports to include a wider range of creative insights, such as those obtained by architects.

Imagery and Discovery

299

Construct Theories of How Pictures Aid Problem Solving We would have a better understanding of how images aid problem solving if we had a better understanding of how pictures aid problem solving. This argument does not imply that pictures and images are identical. As discussed previously, pictures are better than images for reinterpreting patterns (Chambers & Reisberg, 1985; Chambers, Reisberg & Logie, this volume; Reed & Johnsen, 1975) and imagined parts are more useful than perceived parts for synthesizing patterns (Finke, 1990). But the substantial similarity between the functional equivalence of pictures and images (Finke, 1985) should help us predict when images are likely to facilitate performance based on the beneficial effects of pictures. For instance, Mayer (1989) found that adding labeled illustrations to scientific text helped students recall more explanative information in the text but did not improve recall of nonexplanative information. The passage explained how mechanic, hydraulic, and air braking systems operated, and the illustrations showed how the key parts of each system changed when applying the brakes. Of particular interest is Mayer’s finding that students who received the illustrations gave better answers to problem-solving questions, such as what should be done to make brakes more reliable, or more effective. A detailed model of how diagrams can facilitate problem solving (particularly physics and geometry problems) was developed by Larkin and Simon (1987).They proposed that diagrams often display information that is only implicit in a text, and therefore has to be computed to make it explicit. Even when both diagrammatic and text representations contain equivalent information, computational demands can differ in accessing the information. Related information is often located at adjacent locations in diagrams making it easier to recognize patterns, search for information, and make inferences. Diagrams are typically better representations not because they contain more information, but because they support more efficient computations. Larkin and Simon concluded their paper with a comment on visual imagery: In this paper, we have represented external diagrams symbolically as list structures, and the

300

Stephen K. Reed inference processes as list processes in a production system language. These representations and processes could equally well be interpreted as denoting mental images and imagery processes in the brain. But much difficult psychological research, the exact character of which we can only dimly perceive, will be required to test this hypothesis. (Larkin & Simon, 1987,p. 98).

Although Larkin and Simon believe that external diagrams and images have similar properties of localization of information, they also believe that substantially less detail can be stored in images. In the next section, I propose some promising areas in which imagery may facilitate problem solutions because the diagrams do not require excessive detail that would prevent them from becoming useful imagined diagrams.

Determine How Imagery can Supplement Problem Solutions My second suggestion is that we need to expand our efforts to understand how imagery can facilitate solving problems. The relation between imagery and successful problem solving has not been ignored by psychologists, as is evident from the extensive review of this topic by Kaufmann (1990). However, Professor Kaufmann began his review with the statements that (1) our knowledge about the role of imagery in problem solving lags behind our knowledge of imagery in learning and memory, and (2) the field suffers from a lack of explicit theoretical formulations that could integrate previous findings and direct future research. My own research efforts in recent years have focused on the study of algebra word problems. Although many of these problems require that students make inferences about spatial relations, I have not studied the influence of diagrams on their inferences. It seems to me, however, that constructing diagrams might be beneficial and that some of the diagrams are simple enough to construct mentally. One of the reasons that I think diagrams would be beneficial is that we have recently discovered that inferences about spatial relations are systematically biased by irrelevant, verbal information (Reed & Zelmer, 1991). We gave psychology students 24 motion problems and asked them to classify the problems according to

Imagery and Discovery

301

whether the two travelled distances mentioned in the problem should be added, subtracted, or equated. Consider the following problem: An athlete trains by running for 1.5 hours and biking for 1 hour, covering a total distance of 25 miles. If his running speed is 10 mph slower than his biking speed, how fast does he run? This is an example of a motion problem, consisting of two travelled distances, two speeds, and two times. Should the distances be equated, added, or subtracted? We wanted to determine whether an irrelevant relation, the relation between the two speeds, influences subjects’ decisions about distances. We had reason to expect such an influence. Previous studies have shown that novices often solve problems by using a meandend analysis search that focuses on the unknown variahle (Gick, 1986; Larkin, McDermott, Simon, & Simon, 1980; Sweller, Mawer, & Ward, 1983). Although the relation between the two speeds is irrelevant to deciding the relation between two distances, it may nonetheless bias students’ decisions because of its importance as an unknown variable. We tested these hypotheses by creating four questions for each problem by orthogonally varying whether the goal was to find the slower or faster of the two speeds, and whether the word slower or faster described the relation between the two speeds (the different questions was a between-subjects variable). Table 1shows the four questions for three of the problems used in our study. The correct answer is that the two distances should be added in the first example, subtracted in the second example, and equated in the third example. Our findings supported our suspicions. Subjects responded significantly more often that the two distances should be added when the goal was t o find the faster speed, and subtracted when the goal was to find the slower speed. Similar results were found for the biasing effect of the words slower and faster, although the bias was significant only for the addition responses. Students correctly classified only 44% of the problems, not much higher than the 33% correct classifications expected by chance. The large number of misclassifications and the significant biasing effect of irrelevant, nonspatial relations (such as speed) indicate that we need to find effective methods for improving spatial inferences.

302

Stephen K.Reed

TABLE 1 Example of problems and questions used by Reed & Zelmer (1991) to study how irrelevant verbal information influences spatial inferences. Unknown Description Problem and Questions An athkte trains by running for 1.5 hours and biking fir 1 hour, covering a total distance of 25 miks. If his running speed is 10 mph slower than his biking slower slower speed, how fast does he run? faster slower If his biking speed is 10 mph faster than his running speed, how fast does he run? slower faster If his runningspeed is 10 mph slower than his biking speed, how fast does he bike? faster If his biking speed is 10 mph faster than his running faster speed, how fast does he bike? Karen’s boat can travel 160 miles further in 8 hours than Jane’s boat can travel in 10 hours. slower slower How fast is Jane’s boat is Jane’s boat is 25 mph slower? faster slower How fast is Jane’s boat if Karen’s boat is 25 mph faster? slower faster How fast is Karen’s boat if Jane’s boat is 25 mph slower? faster faster How fast is Karen’s boat if Karen’s boat is 25 mph faster? n m drove to his vacation home in 7 hours and returned by the same route in 5 hours. slower slower How fast did he drive to his vacation home if his initial speed was 18 mph slower? faster slower How fast did he drive to his vacation home if his return speed was 18 mph faster? slower faster How fast did he return from his vacation home if his initial speed was 18 mph slower? faster faster How fast did he return from his vacation home if his return speed WBB 18 mph faster? Note: Unknown refers to whether the question asks for the value of the slower or faster speed, and Description refers to whether the word slower or faster describes the relation between the two speeds.

Imagery and Discovery

303

One approach would be to ask students to physically construct simple vector diagrams to represent the two distances in each problem. Following practice on constructing vector diagrams, with feedback, they would be instructed to mentally construct vector diagrams of the problems. Vector diagrams are sufficiently simple that their mental construction should be feasible if a high level of performance can be achieved during physical construction. Determine How Imagery can Replace Standard Solutions My third suggestion is that we need to discover how imagery can lead to novel representations of problems that can replace standard representations. Although the use of either physicallyconstructed or imagined diagrams might improve spatial inferences in solving standard word problems, few of us would classify these solutions as creative. The role of imagery in these situations is to facilitate the construction of standard solutions based on algebraic equations. Let’s now look at a more challenging problem that can be solved by a creative approach based on spatial inferences (Trismen, 1988). A man is standing on a bridge, 300 feet from the near side and 500 feet from the far side. A train is approaching the near side. If the man runs at a speed of 10 mph toward the train, he will reach the near end of the bridge just as the train does. If he runs at a speed of 10 mph away from the train, he will reach the far end of the bridge just as the train overtakes him. What is the speed of the train? You can solve this problem without using paper and pencil and without constructing equations if you can make the appropriate spatial inferences. My experience has been that people find this a challenging problem, and that even good problem solvers often prefer to search for an algebraic solution, rather than look for the simpler approach, See if you can find the solution by discovering helpful spatial relations. If you have not yet solved the problem, see if you can solve it after answering the following question: If the man runs away from

304

Stephen K.Reed

the train, where will he be when the train reaches the near end of the bridge? The answer is that if the man runs 10 mph away from the train he should travel 300 feet by the time the train reaches the near end of the bridge. If he continues running at 10 mph, he and the train will reach the far end at the same time. The train therefore has to travel 800 (500 + 300)feet in the same time that the man travels 200 (500 - 300)feet. The train must therefore be travelling four times as fast as the man, or 40 mph. This kind of problem should be particularly useful for studying creativity and imagery, First, the solution based on spatial inferences is challenging and requires that we ignore the more direct approach of searching for algebraic equations. Second, making spatial inferences should be facilitated by imagery. Even if people physically construct a diagram to represent the initial state of the problem, they must appropriately move the train and man to discover helpful information. My own experience in solving this problem is that I initially thought about constructing an algebraic equation to represent all the relevant information. However, I quickly abandoned this approach for two reasons. First, I wasn’t certain that I could construct a correct equation. Second, because the distances are measured in feet and the speeds are measured in miles per hour, I might have to make some conversions to a common unit of measurement. I then decided to use a more imagery-based approach that did not require solving an equation. I mention my experience because it fits Kaufmann’s (1990) formulation of how imagery is used in problem solving. He proposes that imagery is a back-up system that is used when computational processes break down, because of either a lack of rule-based knowledge or because of the strain on working memory created by a high information load. Images allow the problem solver to construct perceptual-like mental models based on perceptual operations that are simpler than the computational operations. As I was writing this chapter I decided to try to solve the bridge problem by constructing algebraic equations so I could compare the algebraic solution with the imagery solution, Constructing equations initially requires that the problem solver determine what t o equate. The bridge problem specifies that the man and train reach the near

Imagery and Discovery

305

side of the bridge at the same time when the man runs toward the near side, and both reach the far side at the same time when the man runs toward the far side. I therefore decided to equate the amount of time taken by the man and train to reach one end of the bridge. Time can be expressed as distance divided by speed. The time taken by the man to reach the near side is therefore proportional to 300 feet divided by 10 miles per hour. I decided to ignore the difference in units, hoping that conversion to a common unit would "cancel out" in the equation and therefore not be necessary. The time taken by the train to reach the near side is proportional to d divided by s, where d is the distance (in feet) of the train from the near side and s is its speed (in rnph), both of which are unknown. This leads to Equation 1.

300 feet -- -d 10 mph

s

Because there are two unknowns, I needed another equation t o solve for s. The other equation is the time to reach the far side of the bridge. The time required by the man is proportional to 500 feet divided by 10 mph. The time required by the train is proportional to d + 800 divided by s, where d + 800 is the distance of the train from the far side of the bridge. This leads to Equation 2.

10 mph

S

Solving Equation 1for d yields d = 30s. Substituting 30s for d in Equation 2 and solving for s yields s = 40 mph, which fortunately is the same answer obtained from the imagery solution. Shepard (1978) has suggested several reasons why mental imagery and spatial visualization can facilitate creative problem solving. These reasons are discussed in some detail by IntonsPeterson (this volume) so I will only briefly mention their relevance t o this example. One reason that imagery may lead to creative solutions is that it provides an alternative approach that differs from

306

Stephen K. Reed

more traditional approaches. I believe that constructing equations is the more traditional approach for solving word problems because this method is typically emphasized in algebra classes. There is nothing wrong with this method if students can apply it correctly, but the imagery approach provides an alternative method for finding the solution. Shepard argued that images precede language development and can result in more intuitive solutions. The algebraic solution of the bridge problem depends on knowledge of algebra that may be lacking in many students. Those students who are able to solve the problem by making spatial inferences therefore have an advantage over students who rely solely on algebra. A third reason for using imagery is that the richness of imagery suggests new relations that are not immediately apparent in the verbal statement of the problem. Imagining the man running 300 feet toward the far end of the bridge has the advantage that we now know the location of both the man and the train, which greatly simplifies the problem. The algebraic solution required solving two equations because both the speed and location of the train were unknowns in the initial statement of the problem. A fourth reason is that imagery may help reveal structural symmetries and invariances. An example in the bridge problem is that it doesn’t matter whether the man runs 300 feet at 10 mph toward the near end or far end - in both cases the train will arrive at the near end. Recognizing this invariance lead to the relatively simple solution based on imagery. Collect Self-Reports from Nonscientists

My final suggestion is that we need to collect self-reports from nonscientists t o determine how imagery helped them make new discoveries. Self-reports of how people made important discoveries can further our understanding of the role of imagery in those discoveries. Shepard’s (1988) chapter, cited at the beginning of this article, contains many examples of how imagery aided scientific discoveries. It is important t o note, however, that Shepard did not seek out these examples until after he was confident that he could objectively study imagery in the laboratory. Relating laboratory findings to self-reports should be a promising approach t o relating imagery, which can be more effectively studied in the laboratory, to

Imagery and Discovery

307

creativity, which in its highest form typically occurs when psychologists are not present to study it. The collection of self-reports has focused on scientific discoveries, which is a good place to begin, but we should extend this collection to other domains. Architecture is another occupation in which visual creativity is important. Figure 3 shows a dramatic new building in San Diego, the Emerald-Shapery Center. Developer Sandor Shapery worked out the basic ideas for the design by following Frank Lloyd Wright’s philosophy that good architecture often results from studying nature. Shapery began to study crystals and noticed that many crystals have a hexagonal structure, The final design, by architect C. W. Kim, consisted of eight six-sided spires combined in clusters to form the towers. The hexagonal shapes, besides their aesthetic appeal, have advantages for both construction and the quality of the interior spaces. All horizontal building faces are the same size, allowing for cost savings from the standardization of parts. In addition, there is more window surface area per square foot of floor area than in conventional, rectangular buildings. Every room opens toward the windows, creating a more open, spacious environment than suggested by the physical dimensions. The formation of the eight hexagonal towers into clusters reminds me of the laboratory synthesis tasks that I described earlier. These tasks require that subjects combine several parts to make a whole object. A hexagonal tower was the starting point for the design shown in Figure 3, but the planners needed to decide how to combine towers t o make an aesthetically pleasing and functional design. Mr. Shapery began using three-dimensional models early in his exploration of how to combine these shapes. Physical models are obviously helpful, and I am not arguing that complex architectural designs are created only with visual imagery. But psychologists should not be overly restrictive in emphasizing solely the role of imagery in visual thinking. As I mentioned in my introduction, McKim (1980)has argued that visual thinking is most effective when one can easily move between perception, imagery, and drawing. Many creative discoveries likely involve an integration of all three activities.

308

Stephen K.Reed

Figure 3 The Emerald-Shapery Center, San Diego.

Imagery and Discovery

309

Conclusions In conclusion, I have attempted to examine current research findings for clues regarding when imagery is likely to lead to new discoveries. New discoveries may be limited by either the ineffectiveness or the insufficiency of mental images. In the first case, images are too vague to support new discoveries. For instance, people usually find it difficult to reinterpret images to find a new part or an alternative interpretation. In the second case, images are effectively processed but don’t provide enough knowledge for new discoveries. People may be able to mentally simulate events, without changing their current beliefs. A third case is the successful use of images to discover new information, such as often occurs during mental synthesis tasks. These successful cases should motivate us to find effective training procedures to improve people’s ability to make difficult discoveries. These procedures should train people on a variety of methods for manipulating images and investigate the transfer of these methods to new problems. Improving people’s ability to effectively use images will depend on how much we learn about the role of imagery in creative thought. We would have a better understanding of how imagery aided visual thinking if we had better theories of how pictures aided visual thinking. We also need to extend imagery paradigms into problem solving domains that emphasize spatial knowledge. Some algebra word problems and most physics problems are good candidates. We need to explore how imagery can help us apply standard solutions and lead to changes from symbolic representations to novel, pictorial representations. Finally we need to link laboratory findings to selfreports of creative thinkers to identify the kind of spatial operations that are used across a variety of real-world tasks.

References Chambers, D., & Reisberg, D. (1985). Can mental images be ambiguous? Journal of Experimental Psychology: Human Perception and Performance, 11, 317-320. Cooper, L. A. (1990). Mental representation of three-dimensional objects in visual problem solving and recognition. Journal of

310

Stephen K. Reed

Experimental Psychology: Learning, Memory, and Cognition, 16, 1097-1106. Finke, R. A. (1985).Theories relating mental imagery to perception. Psychological Bulletin, 98, 236-259. Finke, R. A.(1990).Creative imagery: Discoveries and inventions in visualization. Hillsdale, N J : Erlbaum. Educational Gick, M. (1986). Problem solving strategies. Psychologist, 21, 99-120. Intons-Peterson, M. J. (1983).Imagery paradigms: How vulnerable Journal of are they to experimenters’ expectations? Experimental Psychology: Human Perception and Performance, 9,394-412. Intons-Peterson, M. J., & Roskos-Ewoldsen, B. B. (1989). Sensoryperceptual qualities of images. Journal of Experimental Psychology: Learning, Memory, & Cognition, 15,188-199. Kaufmann, G. (1990). Imagery effects on problem solving. In P. J. Hampson, D. F. Marks, & J. T. E. Richardson (Eds.), Imagery: Current developments (pp. 169-196). London: Routledge. 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 Perception and Performance, 4, 47-60. Larkin, J. H., McDermott, J., Simon, D. P., & Simon, H. A. (1980). Expert and novice performance in solving physics problems. Science, 208, 1335-1342. Larkin, J. H., & Simon, H. A. (1987).Why a diagram is (sometimes) worth ten thousand words. Cognitive Science, 11, 65-99. Mayer, R. E. (1983). Thinking, problem solving, cognition. New York: W. H. Freeman. Mayer, R. E. (1989).Systematic thinking fostered by illustrations in scientific text. Journal of Educational Psychology, 81,240-246. McCloskey, M.,& Kohl, D. (1983). Naive physics: The curvilinear impetus principle and its role in interactions with moving objects. Journal of Experimental Psychology: Learning, Memory, and Cognition, 9,146-156. McKim, R.H.(1980).Experiences in visual thinking. Belmont, CA: Wadsworth.

Imagery and Discovery

31 1

Mitchell, D. B., & Richman, C. L. (1980). Confirmed reservations: Mental travel. Journal of Experimental Psychology: Human Perception and Performance, 6, 58-66. Novick, L. R. (1990). Representational transfer in problem solving. Psychological Science, 1 , 1.28-132. Palmer, S. E. (1977). Hierarchical structure in perceptual representation. Cognitive Psychology, 9,441-474. Pylyshyn, Z. W. (1981). The imagery debate: Analogue media versus tacit knowledge. Psychological Review, 88, 16-45. Reed, S. K., & Angaran, A. J. (1972). Structural models and embedded-figure difficulty for normal and retarded children. Perceptual and Motor Skills, 35, 155-164. Reed, S. K., Hock, H. S., & Lockhead, G. R. (1983). Tacit knowledge and the effect of pattern configuration on mental scanning. Memory & Cognition, 11, 137-143. Reed, S. K., & Johnsen, J. A. (1975). Detection of parts in patterns and images. Memory & Cognition, 3, 569-575. Reed, S. K., & Zelmer, R. (1991, November). Acquired modularity for assigning relations in problems. Paper presented at the 32nd annual meeting of the Psychonomic Society, San Francisco. Shepard, R. N. (1978). Externalization of mental images and the act of creation. In B. S. Randawa & W. E. Coffman (Eds.), Visual learning, thinking, and communication (pp. 133-189). New York: Academic Press. Shepard, R. (1988). The imagination of the scientist. In K. Egan & D. Nadaner (Eds.), Imagination and education (pp. 153 -185). New York: Teachers College Press. Shepard, R. N., & Feng, C. (1972). A chronometric study of mental paper folding. Cognitive Psychology, 3, 228 -243. Sweller, J., Mawer, R. F., & Ward, M. R. (1983). Development of expertise in mathematical problem solving. Journal of Experimental Psychology: General, 112, 639-661. Tresselt, M. E., & Mayzner, M. S. (1966). Normative solution times for a sample of 134 solution words and 378 associated anagrams. Psychonomic Monograph Supplements No. 15, 1, 293-298. Trismen, D.A. (1988). Hints: An aid to diagnosis in mathematical problem solving. Journal for Research in Mathematics Education, 19, 358-361.

312

Stephen K. Reed

Watson, J. D. (1968). The double helix. New York: New American Library.

Imagery, Creativity, and Discovcry: A Cognitive Perspcctive B. Roskos-Ewoldson. M.J. Intons-Peterson and R.E. Anderson (Editors) 0 1993 Elsevier Science Publishers B.V. All rights reserved.

313

Chapter 11 ~~

~

IMAGERY, CREATIVITY, AND DISCOVERY= CONCLUSIONS AND IMPLICATIONS Beverly Roskos-Ewoldson Margaret Jean Intons-Peterson Department of Psychology Department of Psychology Indiana Uni vers i ty University of Alabama Tuscaloosa, AL 35486 Bloomington, IN 47405 USA USA Rita E. Anderson Department of Psychology Memorial University St. John's, Newfoundland Canada A l C 5S7 This book represents our conviction that a cognitive perspective can inform research and yield insights into the relations among imagery, creativity, and discovery. Our main goal for this volume has been to explore the role of imagery in creativity and discovery, and to determine whether the introspective reports of imagery's use in creative problem solving are epiphenomena1 or cognitively valid. The chapters in this volume assessed, through discussion of current theory and research, our progress with respect to the following central issues: (a) identifylng the processes involved in creativity and discovery, and determining how imagery influences these processes; (b) determining which properties of visual and auditory imaginal processes facilitate or limit creativity and discovery; (c) determining whether the purported limitations of images reflect limitations of the imaginal system or are part of a more general limitation of cognitive

B. Roskos-Ewoldsen et al.

314

processing capacity; and (d) determining other aspects of the imaginal process that may be used in the creativity/discovery process. The following summary recapitulates the main themes of the chapters and includes comments from discussion among the conference participants during the concluding session. We first describe the problems encountered while attempting to define creativity and discovery. Next, we discuss imagery’s role in the creative enterprise. Finally, we examine the insights delivered by applying the cognitive perspective to the relations among imagery, creativity, and discovery, and identify areas in need of further research.

What is Meant by “Creativity”and “Discovery”? We sought definitions of terms such as “creativity,” “creative,” and “discovery” so that we would share a common understanding. This goal was partially thwarted by some instructive and constructive disagreements.

Multiple Definitions One view (e.g., Anderson & Helstrup; Finke) emphasized the discovery process within the larger realm of the creative process. Discovery, in this view, is embedded in creativity. Proponents consider the cognitive processes involved in a visual discovery task; namely, construction and interpretation (Anderson & Helstrup) or generation and exploration (Finke). Another view (e.g., RoskosEwoldsen) described “creativity” as a process that results in a form or figure that may or may not be judged to be creative. Combinational play, similar to the others’ generation and construction, comprises the creativity (rather than the creative) process. Interpretation and exploration constitute the discovery process. This latter view is somewhat more sequential, with the products of the creativity process feeding the discovery process, although the processes can act independently. All views share the commonality of hypothesizing that cognitive processes are fundamental to both discovery and creativity. They differ in the relations and ordering of the processes of creativity and

Conclusions and Implications

315

discovery to each other, and in the cognitive processes hypothesized to contribute to each. The views are compared and contrasted in more detail below. Three Views of the Creative Process Anderson and Helstrup describe the processes involved in the general visual discovery task, defined as the set of tasks in which subjects find parts in whole patterns, or synthesize or construct a whole from the parts. In this view of figure construction, the parts to be used are entered into a memory store and then are combined and constructed in a visual buffer to form a figure. The constructed figure is interpreted or searched for signs of a recognizable figure. Both controlled, goal-directed, and more gestalt-like automatic figure formation and segmentation processes are assumed to be involved in the construction and interpretation phases. Once a decision is made that the figure is recognizable, the product is reported by labelling and drawing it. Finke proposed the “Geneplore” model, a general view of creative discovery. In his model, Finke identifies two separate processes - generation of preinventive structures and preinventive exploration and interpretation. The inventor begins by generating a structure that may or may not correspond to any particular, useful product. The generated structure is explored for useful purposes or interpretations. The exploration itself may suggest the generation of a new structure, either by focusing or expanding a concept or part of the initial structure. In both of these views, the construction and interpretation (discovery)phases are of equal importance and creativity is assumed to be possible during either phase. In principle, constructions or interpretations can be identified as creative; in practice, however, the interpreted structure is judged for creativity by external judges. In contrast, Roskos-Ewoldsen redefines the concepts of creativity and discovery in terms of the types of cognitive processes that constitute each one. In her view, the phase of generation, construction, or combinational play, discussed by Anderson and Helstrup and by Finke, is defined as creativity. That is, creativity is manipulational, combinational, and generational play. Discovery, then, becomes the interpretation or exploration of the structures

31 6

B. Roskos-Ewoldsen et al.

produced during combinational play, or simply the interpretation of existing structures. Clearly, the term creativity is used in quite different ways between the first two views and the last. Specifically, creativity, as used by Anderson and Helstrup and by Finke, refers to the larger realm of the creative enterprise, of which creative discovery is a part. In this sense, creativity involves an ability to see purposes for newly constructed forms that are unusual or original, Contrast this with Roskos-Ewoldsen's version of creativity, where the term is used in a more specific way to differentiate it from discovery. On her view, the process of creativity involves combinational play, whether or not the product is subsequently judged to be creative. The term discovery is also used differently in the three views. Again, Roskos-Ewoldsen takes a more narrow view of discovery than either Finke or Anderson and Helstrup. The latter researchers describe the discovery task as including both generation-construction and exploration-interpretation processes, whereas Roskos-Ewoldsen defines discovery as the process of exploration-interpretation only. The differences and similarities of these views are both constructive and instructive, for they lead to explorations of the views, thereby advancing our understanding of the creative enterprise.

Other Considerations The difficulty of defining creativity prompted the question: Can we or should we attempt a universal definition of creativity? The resulting discussion suggested that there may be different species of creativity. First, creating may be a form of problem solving. If so, creativity might occur only in relation to a goal or set of goals. Second, creativity may differ depending on the level at which one focuses. For example, what is considered creative for society may not be the same as that considered creative for an individual. Third, if we assume that creation is a process, is the final product a part of creativity? If the final form is a part of creativity, what criteria should be used to judge whether products or their interpretations are creative? Novelty or originality are ofien considered to be indicators of creativity. But if novelty is used as a criterion, problems remain because a product can be novel on different dimensions, and,

Conclusions and Implications

31 7

conversely, products with similar values on one dimension may not be perceived as equally creative, overall. Fourth, how do creative scientists (artists, etc.) fit with our notions of creativity? Are these creative people qualitatively different than other, less creative scientists and artists? Or are they at one extreme on a continuum of creative abilities? These discussions reflect the difficulty of deciding on a definition of creativity, and led one conference participant to suggest that the attempt to define creativity may define it out of existence! We turn now to a discussion of imagery and its role in creativity and discovery.

Imagery's Role in Creativity and Discovery Properties of Imagery

What began as an investigation of imagery's role in creativity and discovery led to an opportunity to revisit what we know about imagery. Some general properties of images emerged as themes in many of the chapters. Imagery paradigms most often test a hybrid of language and perception, Subjects typically receive verbal instructions to image something, and the "something" may have to be retrieved from longterm memory. Hence, images are not exclusively visual replications of previously perceived stimuli, nor are they propositions. Rather, they are a true hybrid of both, having both depictive and descriptive qualities (see Chambers; Kaufmann & Helstrup). From most accounts, images are not pictures in the head, complete with all details, waiting to be searched for any piece of information. Instead, images are often produced with intention for some purpose. As such, images tend to be inherently meaningful (Chambers). That is, a description of the image to be formed, based on the intent of the imager, may guide the actual depiction or formation of the image. Second, images occur in many forms, Images range from conceptual images (e.g., intentional thought) to spatial images to mental pictures (e.g., experiential sensation), with spatial images occurring more frequently than either conceptual images or mental pictures (Kaufmann & Helstrup). Furthermore, there is a distinction

318

B. Roskos-Ewoldsen et al.

between images that are generated through combinational play and those that are retrieved as visual “pictures” from memory or those that are replayed in the mind as if listening to a tape recording. Similarly, there is a difference between imagining with the use of efferent aspects of sound production (i.e., the musculature of the mouth; enacted imagery) and imagining without their use (“pure” imagery; Reisberg & Logie). By viewing imagery in terms of afferent and efferent partnerships, we begin to understand how, for example, visual imagery differs from auditory imagery. A third property of imagery is that it is part of the cognitive system and, consequently, is dependent on our limited-capacity working memory (Anderson & Helstrup; Chambers; Hyman; Kaufmann & Helstrup; Reisberg & Logie; Roskos-Ewoldsen). Images must be constructed from information stored in long-term memory or from the results of perceptual analysis. Once an image is constructed, it must be maintained over time. Not all information from memory or perceptual analysis will be represented in an image, nor will all information be maintained to the same extent. Because an image is generated and maintained in a limited-capacity system, the image may not be clear everywhere; some aspects of the image may appear to the imagery t o fade in and out, whereas other aspects may never have been there to begin with (i.e., not all details are filled in). Hyman suggested that imagery is like reconstructive memory. Images, like reconstructive memory, may embody the gist of the information originally encountered. Furthermore, aspects of the image that are congruent with our intent for the image may receive more attention and therefore be maintained more so than aspects that are less relevant to the intent (Chambers; Hyman). How are these properties of imagery related to its role in the creative enterprise? More specifically, how do these properties influence imaginal creativity and discovery? Imagery and Discovery Consider first the role of imagery in discovery. Though there is some disagreement about what is considered to be discovery (i.e., whether discovery involves both combinational play and interpretation or just interpretation), there is agreement that interpretation is a major part of the discovery process. With this

Conclusions and Implications

319

agreement as a basis, we can discuss the limitations and successes of the use of imaginal processing in the discovery-interpretation process. It appears that images can be reinterpreted or reconstrued there does not appear t o be as absolute a limitation on the interpretability of images, as originally indicated by Chambers and Reisberg's (1985) results. If reconstruals that do not match experimenter-defined correct reconstruals (e.g., the ducWrabbit reconstrued as a fish) are accepted as possible answers, reconstruals are quite common (Kaufmann & Helstrup; Peterson). Although images can be reinterpreted to some degree, imaginal reinterpretations are not as easy as perceptual reinterpretations. This difficulty may occur because imagery is a process that is dependent on a limited capacity memory system and, hence, not all details will appear in an image. The details that appear tend to reflect the intentions of the imagery, and the demands of the task. Therefore, an image may not be reconstrued easily because the image has been formed with a priori meaning. If another image were to be formed from information stored in memory, the content may change, depending on the intent of the imagery (Chambers). Furthermore, an image requires internal support, unlike a drawing. That is, maintaining an image entails more processing capacity than looking at a drawing. The more working memory that is required to maintain the image, the less processing capacity there will be for imaginal discoveries (Anderson 8z Helstrup; Roskos-Ewoldsen). Factors other than processing capacity may also influence imaginal discovery. One factor involves the way an image is inspected, which is determined in part by the type of practice figure provided. Figural reinterpretations can range from reconstruals, where the parts of a figure take on new meaning as different parts of the alternative interpretation, to reference frame reversals, where tophottom and fronthack change from one interpretation to the next. If a figure requires a reference-frame reversal for reinterpretation, imaginal reinterpretation is easier when an appropriate practice figure (i.e., a figure that also requires a reference-frame reversal) is used (Peterson). The type of prompt provided when the subject does not spontaneously reverse an imagined ambiguous figure also affects the way the figure is inspected. If participants are given information

320

B. Roskos-Ewoldsen et al.

about both the category to which the alternative interpretation belonged, and the orientation of the alternative (i.e., which perspective is top), then almost half of the participants are able to discover the alternative interpretation (Chambers;Hyman; Peterson). Only 10-15% of the participants who received minimal prompts were able to reinterpret their imagined figures (Hyman; Kaufmann & Helstrup, strict criterion; Peterson). Another factor influencing imaginal discovery is skill at spatial visualization. Kaufmann and Helstrup showed that highly skilled visualizers, as measured by a version of the Minnesota Form Board test, were more likely to reverse an ambiguous figure than Chambers and Reisberg’s subjects, who were unselected for visual or spatial abilities. A final factor involves the perceptual organization of a stimulus, which appears to influence how easily imaginal discovery can occur (Roskos-Ewoldsen). If an imagined pattern has been put together from good (well-organized) parts, discovery of emergent patterns embedded within the imagined pattern occurs more easily than when the imagined pattern was put together with poorly organized parts. The organization of the pattern itself also influences discovery within imagined patterns. Discovery is more difficult (i.e., takes longer) when the pattern is organized than when the pattern is poorly organized. Imagery and the Creative Process

As with discovery, there was some disagreement with what we mean by creative process. The creative process may be the process that gives rise to creative inventions, interpretations, and discoveries (e.g., Anderson & Helstrup; Finke). Roskos-Ewoldsen prefers to discuss not the creative process but creativity per se (i.e., combinational play and manipulation). Despite differences in opinion regarding creativity and the creative process, there is agreement that generation, construction, and combination are major components of the creative process. Given this agreement, we return to the question: Can imagery play a role in the creative process? The answer appears to be yes. First, research by Anderson and Helstrup, Finke, Hyman, and Intons-Peterson has shown that participants unselected for their

Conclusions and Implications

321

imagery ability (or creative ability, for that matter) can imaginally combine simple shapes to form recognizable patterns, some of which are later judged to be creative patterns. But, is imagery critical to the creative endeavor?

Does Imagery Have a Special Status? Many people have claimed that imagery is critical to the creative process (e.g., Miller, 1984; Shepard, 1978). Shepard for example, argues that imagery has a special status primarily because it is relatively free of linguistic constraints. In particular, he claims that imagery has a richness and relation to external sources that is not fully preserved by language. That is, language may have a constraining influence on imaginal processing and, consequently, on the use of imagery in creativity and discovery. In fact, much of the research reported in this volume lends substance to Shepard's proposal that the conventions of language constrain the kinds of mental imagery operations involved in the processes of creativity and discovery. At one level, this is bound to be true because most experimental tasks involving these processes are presented via verbal instructions. In addition, Chambers and Kaufmann and Helstrup note that both depiction and description presumably are involved in imagery and, consequently, in creativity and discovery. The descriptive, linguistic shaping of imaginal processes, however, seems to be deep, tenacious, and difficult to overcome. Hence, linguistic features and demands will affect performance on tasks involving imagery, creativity, and discovery. The effects of language are shown clearly with respect to the linguistic and conceptual interpretation of images. As much of the research reported, the initial (verbal) interpretation of an image may resist reinterpretation or reconstrual. The most dramatic demonstration is Chambers and Reisberg's (1985)failure to find evidence for any reversals of ambiguous figures, of course, but even hints about using different linguistic-conceptual or spatial perspectives have not produced marked increases in the frequencies of reconstrual (e.g., Chambers, Hyman, Kaufmann & Helstrup, Peterson). Furthermore, Anderson and Helstrup found that asking subjects to use a verbal strategy reduced the number of recognizable patterns produced, compared to visual strategy instructions.

322

B. Roskos-Ewoldsen et al.

This apparent suppression of recognizable patterns by a verbal strategy seems at odds with Finke's demonstration that subjects produced more responses judged creative when the responses had to fit specific categories than when they were not constrained t o a single category. Do Finke's results signal linguistic facilitation of creativity? They might, because language can be used to overcome at least some of the resistance to reinterpreting images. For example, when subjects were told which side of a tilted image of the state of Texas was the top, they were able to identify the shape, although they were unable to do so when given less precise verbal hints. Additional evidence of linguistic facilitation comes from other kinds of precise hints (e.g., Hyman, Peterson, Reisberg & Logie). Note, however, that the latter evidence is facilitation only in the sense that precise hints may combat resistance to reinterpretation of images. It does not really address the creativity issue, as Finke's results do. But we return to the question: Do his results signal linguistic facilitation of creativity? Again, the answer must be that it is possible. An alternative explanation does exist, however: Having a category label may aid the search and retrieval process. Subjects in Intons-Peterson's work often remarked that they "hardly knew where to begin" when drawing their images. Having a target category may focus the search, as required by the task. This perspective argues that the apparent facilitation of creativity afforded by specifying linguistic category may arise from the targeting of the search among alternatives to the number of images specified by the instructions. Summarizing the interrelations between language and imagery, language clearly affects imaginal activities by defining and constraining the interpretations of the mental products of imaginal processing. Language also may aid release from these constraints. Thus, although we are not yet in a position to generate finely tuned predictions of the interrelations of language and imaginal-spatial processing, we must be sensitive to language's existence and likely influence on all experimentation in this area. Shepard also proposed that imaginal activity has a special richness that is conducive to creativity. Intons-Peterson provides an example of the facilitating effects of imagery. Though she found no differences in the number of creative figures constructed during an initial test session among conditions that varied the likelihood of

Conclusions and Implications

323

using imagery to perform the task, she found that, after practice, more creative figures were produced by subjects in the imageryinducing conditions than by subjects in the conditions less likely to induce the use of imagery. Other evidence described in the chapters is tangentially related to Shepard's point that imagery is a rich alternative compared to language. Reed suggests that diagrams and pictures (and images, by extension) are particularly useful, not because they contain more information than verbal descriptions, but because they support more efficient computations than strictly verbal descriptions. Specifically, there are some problems that can be solved by imagining the spatial relationships among the elements of the problem to be solved, and "seeing" how the relationships change over time. This spatialimaginal inferential reasoning can be easier and less time-consuming than a more direct approach such as searching for algebraic equations (see also Lindsay, 1988). Reed also suggests that although diagrams may encourage this inferential reasoning, diagrams are static. Because mental images can be dynamic, they may facilitate the dynamical transformation of the diagram information. In addition, Peterson argued that reconstruals dominate in imagery; reference frames in perception. It is not clear from her research that the reconstruals of imagery provide greater richness than the reference frames of perception, but Shepard's view is that imaginal-spatial activity may have its own special richness. Part of that richness may be related to Peterson's reconstruals. Thus, in imagery, as in reconstruals, the manipulation of internal parts may be fairly easy, and modifying the overall configuration may be less easy, a position entirely consistent with Roskos-Ewoldsen's evidence. If so, these reconstruals are exactly the kind of mental manipulation that would be expected to foster both creativity and discovery. Shepard's third suggestion was that imagery, with its aspects of intuition and manipulation, may precede language developmentally. Evidence that young children develop efficient navigational systems before extensive communication (Mandler, 1983) and that monkeys show patterns of responses that resemble mental rotation even though they do not speak in the usual sense (Georgopoulos, Lurito, Petrides, Schwartz, & Massey, 1989) can be cited to support this view, These claims may be disputed, however, by people claiming

324

B. Roskos-Ewoldsen et al.

that many nonhuman animals possess effective communication systems. Finally, Shepard notes that symmetries may have a special status within visualization. Impressive support for this view emerged from a number of laboratories. Roskos-Ewoldsen found that symmetry, roughly equated to goodness ratings, has complicated effects, which may be related to Peterson's distinction between reconstrual and reference frame reversals. Roskos-Ewoldsen found that patterns previously judged to be poor were not as cohesive as good patterns, hence it was easier to dissemble or reparse poor patterns into their component parts than to dissemble or reparse good patterns into their components. The poor patterns tended to be less symmetrical than the good patterns and, thus, may be akin to the sensitivity of images to reconstrual of internal parts. Good (symmetrical)patterns cohere, require less processing time, and, like reference frame reversals, resist fragmentation or reinterpretation. In brief, then, the effects of symmetry (goodness) depend upon whether the symmetry is manifested primarily in the outer edges or perimeter of the object or internally. As a final note on Shepard's claim that imagery has a special status in creativity, we mention that imaginal manipulation of stimuli is often more difficult than doodling with, seeing, or hearing the stimuli. This difficulty is exemplified by Kaufmann and Helstrup's research. Their findings show that although participants who were chosen for their high spatial and visualizing abilities performed better than other unselected subjects (e.g., Chambers & Reisberg, 1985; Hyman; Peterson), less than 30% of the high ability students could reconstrue their own images, according to Chambers and Reisberg's reconstrual criterion. The bottom line is that creativity and discovery can occur with the use of imagery; we have not yet shown that imagery facilitates the processes, in comparison to the use of external support &e., doodling, seeing, hearing). Instead, what appears to be most important for the creative process is to be able to transform information from one code t o another--for example, from verbal to imaginal and back again--so as to gain multiple perspectives on the task at hand, a point mentioned during the final discussion session of the conference (see also Peterson).

Conclusions and Implications

325

Future Directions So far we have considered the cognitive aspects or Greativity and discovery, and we have explored the properties of mental imagery that facilitated or limited its use in the creative process. We now venture into even less charted territory by suggesting future research directions.

Training The training of people to use mental imagery in creativity and discovery has had a colorful past (e.g., McKim, 1972). Our research suggests that training should and can be informed by theory. Specifically, we know that the type of hints provided about changing orientations, or captions of “top” and “bottom,” is crucial for discovery (Chambers; Hyman). Likewise, the type of stimuli used in practice is important for discovery (Peterson). Further, practice itself appears to facilitate discoveries which are judged to be creative, especially when imaginal processes are likely to be invoked (IntonsPeterson). What about creativity? What kinds of training leads to improved combinational play? Reed argues that to understand the circumstances under which imagery will work, we should begin by investigating the circumstances under which diagrams work, for they may serve similar purposes. In addition, Reed suggests that we should investigate the work of individuals from domains where imagery is likely to be central, such as architecture.

Problem Solving Reed also suggests that we extend imagery paradigms to problem solving. How does imagery work within this realm? What are the cognitive processes that allow the use of imagery in problem solving? How does the use of imagery compare to the use of algorithms? How does using imagery compare with using diagrams? In a sense, research by Anderson and Helstrup, Finke, and IntonsPeterson already deals with problem solving. When participants are asked to combine parts in such a way that they form a recognizable form, or are asked to interpret a structure in terms of a given

326

B. Roskos-Ewoldsen et al.

category, the task can be perceived as a problem to be solved, with the solution being open-ended Le., there is no single correct answer).

Individual Differences Another direction for future research involves the range of cognitive abilities people may have to play combinationally with parts or to reinterpret or discover embedded figures. These are the abilities that drive the creativity and discovery processes. To understand these processes is to understand creativity and discovery. Intons-Peterson raises several questions related to individual differences. Can creative discovery be taught, or is it the special domain of a few gifted individuals? Some imagers may generate and search their images more rapidly than others - are these people more likely to reverse their images, make discoveries within their images, or be more creative than others who generate and search more slowly? Conversely, are creative individuals quicker at reversing or discovering from images than others who are less creative? Perhaps speed of generatiodsearch is not as important as the strategies used to complete the task. Creative individuals may focus their search for a solution differently than less creative individuals.

Cognitive Processes in Visual and Auditory Imagery Reisberg and Logie present an intriguing analysis of the similarities and differences between visual and auditory imagery, based on differences in working memory for auditory and visual information. More research exploring the afferentlefferent qualities of audition and vision, and their auditory and visual imagery counterparts, needs to be conducted.

Imagery within the Limited-Capacity Cognitive System There should be more emphasis not only on defining what we mean by creativity and discovery, and how imagery is able to be used in these processes, but also on understanding how imagery, creativity, and discovery fit into the cognitive system. Not only do we need to relate imagery to perception, but we need to understand better how limited processing capacities limit imaginal performance.

Conclusions and Implications

32 7

We also need to examine how the interactions between language and imagery affect creativity and discovery.

Motivation and Affective Concomitants Shepard (1978) argues that images are more likely to engage affective systems than verbal language. Finke talked about the "haunting" quality of many of the forms generated by his subjects, and the interest they had in pursuing the implications of their inventions. Intons-Peterson noted that although little work has been conducted in this area, recent work by D. Roskos-Ewoldsen and Franks (personal communication) suggests that affect and imagery are interconnected in the cognitive system. At this time it is unclear whether imagery drives the affective system or vice versa. Nor is it clear how motivational factors enter into our picture of creativity and discovery.

Summary Throughout this volume we hoped to define and discuss the major theoretical issues, including (a) processes that are involved in creativity and discovery, and how imagery influences these processes, (b)properties of visual and auditory imaginal processes that facilitate or limit use of creativity and discovery, (c) purported limitations of images, considered either as a limitation of the imaginal system or as a general constraint of processing capacity, and (d) other aspects of the imaginal process that may be used in the creativity-discovery process. We close by noting that in her introduction, Intons-Peterson revealed the limited creativity of psychologists with respect to our understanding of creativity. The four stages of creativity suggested by Wallas (1926)have not changed in the last seventy-five years. Our hope is that a cognitive perspective may help us to move beyond the old to new conceptions of creativity.

328

B. Roskos-Ewoldsen et al. References

Chambers, D., & Reisberg, D. (1985). Can mental images be ambiguous? Journal of Experimental Psychology: Human Perception and Performance, 11, 317-328. Georgopoulos,A. P., Lurito, J.,Petrides, M., Schwartz, A., & Massey, J. (1989). Mental rotation of the neuronal population vector. Science, 243, 234-236. Lindsay, R. K. (1988). Images and inference. Cognition, 29, 229-250. Mandler, J. M. (1983). Representation. In P. H. Mussen (Ed.), Handbook of child psychology, Vol. IIZ. (4thed.) (pp. 420-494). New York Wiley. McKim, R. H. (1972).Experiences in visual thinking. Monterey, C A BrookdCole. Miller, A. I. (1984). Imagery in scientific thought: Creating 20th century physics. Boston: Birkhauser. Shepard, R. N. (1978).Externalization of mental images and the act of creation. In B. S.Randawa & W.E. Coffman (Eds.), Visual learning, thinking, and communication (pp. 133-189). New York: Academic Press. Wallas, G. (1926). The art of thought. New York Harcourt Brace,

329

AUTHORINDEX A Ahsen, A., 126,146 Amabile, T.M.,258,283 Anderson, J.R., 129, 135,146,256, 283 Anderson, R.A., 62,67 Anderson, R.E.,5, 8,10,11, 13, 14, 15,20,21,27,28,30,31,32, 33,35,50,75,77,94,126,127, 145,147,189,190,191,192, 193,194,195,198,200,202, 215,219,223,224,226,231, 233,235,236,237,238,243, 246,250,251,259,297,314, 315,316,318,319,320,321, 325 Angaran, A.J., 291,311 Arieti, S., 2,33 Arnheim, R., 2,33,84,93 Attneave, F.,204,210,218

Bialystok, E., 44, 74 Biederman, I., 156,157,182,232, 250 Binford, T.O.,156,183 Bishop, D.V.M., 8,33,67,69 Block, N.,143,146,198,218 Boden, M.,282,283 Bradshaw, G.L., 30,31,35 Brandimonte, M.A.,8,33,48,67,69 Bransford, J.D., 101,120,278,283 Brentano, ,F.,79,95 Brody, B.A., 64,69 Brooks, L., 62,69 Brown, J.L., 203,204,220 Bruce, D., 224,229,250 Bruner, J., 41,69 Bruning, J.L., 165,183 Brunn, J., 229,251 Bryant, P., 54,73 Buchanan, M.,55,69 Butters, N.,64,69

B Baddeley, AD., 52,53,54,56,59, 61,62,65,69,72,73,74,76 Ball, T.M., 292,310 Barclay, C.R., 101,120 Bartlett, F.C., 100,116,118,120, 134 146 Barton, M.,64,69 Bassok, M.,31,33 Baxter, D.A., 23,25,36,50,51,57, 75,198,199,220 Begg, I.M., 31, 36 Bellugi, U.,59,69 Besner, D.,52,53,69

C Calvanio, R., 61,62,69 Caminiti, R., 64,72 Campbell, R., 59,69 Carpenter, P.A., 61,69 Cary, L.,203,220,229,251 Cave, K.R., 229,251 Chambers, D., 8,22,23,24,25, 26, 27,34,39,40,42,43,44,45, 46,47,48,50,57,66,69,75, 78,80,81,82,83,84,88,89, 90,91,92,93,94,95,97 99, 100,102,103,104, 105,108,

330

Author Index

109, 117, 118,120,121,128, 130,131, 135, 137,140, 142, 143,145,146,149, 152, 155, 156,157, 158, 159, 161,162, 165,167,168,183,185,194, 198,199,216,218,234,250, 258,283,289,299,309,317, 318,319,320,321,324,325, 328 Christie, D.F.M., 62,74 Clark, J.M., 180,183 Clayton, T.,85,87,88,96 Clement, D.E., 204,210,219 Clifton, C., 52,76 Cocude, M., 10,34 Cohen, C., 204,210,220 Conrad, R., 55,69 Content, A., 203,220,229,251 Cooper, L A , 44,64,76,78,97,187, 188, 198,221,255,256,283, 285,296,309 Cornell, E.H., 228,251 Csikszentmihalyi, M., 258,278,284

D Daniels, S., 52,69 Davies, J., 52,69 Dean, G.,62,69 Della Sala, S.,56,62,69 DeMers, S.T.,6,7,9,36 Denis, M., 10,34 Dennett, D.C., 84,94,129,130,131, 146 Donnenwerth-Nolan,S.,63,75 Duncker, K.,216,219

E Eisenberg, P., 61,69 Engelkamp, J., 63,69,132,146

Ernest, C., 7,34 Essick, G.K., 64,69

F Fagen, R., 248,250 Farah, M.J., 10,23,34,61,62,69, 81,94,96,100, 109, 117,120, 152, 173,183, 184, 198,200, 215,219,230,231,234,250, 258,284 Fanes, J., 99,107 Feng, C., 296,311 Findlay, C.S., 248,250 Finke, R.A.,7,10, 11, 12, 13, 14,20, 21,23,28,30,32,34,39,50, 61,65,66,69,72,74,75,76, 84,92,94,95,96,100, 101, 109,117,120,126, 133,134, 137, 145,146,152,183, 189, 190, 191,192,193,194,195, 197, 198,199,200,201,202, 204,215,216,219,223,224, 225,226,227,229,231,232, 233,234,236,237,239,242, 244,245,247,249,250,255, 256,257,258,259,260,262, 263,264,265,267,268,269, 270,271,272,274,275,276, 279,280,284,297,299,310, 314,315,316,320,322,325 Fischman, D., 62,72 Fisher, G.H.,153, 159,183 Flew, A.G.N., 132,146 Fliegel, S.L.,173,184 Fodor, LA., 79,96,134,146 Franks, J.J., 28, 101,120 Fusella, V., 79,97

Author Index G Gardner, H., 283,284 Garner, W.R., 204,210,219 Gathercole, S.,65,72 Georgopoulos, A.P., 28,34,63,64, 72, 323,328 Gerbino, W.,48,67,69 Getzels, J.W., 258,278,284 Ghiselin, B., 2,34,123,146,256, 281,284 Gibson, B.S., 155,183,184 Gick, M.L., 30,34,301,310 Glisky, M.L., 7,21,23,24,30,31, 36,81,83,97,154, 158, 165, 169,173,174, 175, 176,184 Goldman, A.I., 143,146 Gordon, R.,6,34 Guilford, J.P., 6,34

H Hadamard, J., 123, 146, 188,219 Hadden, S.,63,72 Halper, F.,87,88,96 Hammond, K.M., 61,62,69 Hampson, P.J.,124, 127, 148 Hannay, A., 132,133,146 Harman, G.H., 143,146 Harnish, R., 234,253 Harvey, E.R.,155,184 Hayes, J.R., 31, 34,35 Hazen, N.,228,251 Heil, J., 130,146 Helstrup, T.,5, 8, 10,11,13, 14,15, 20,21,25,27,28,30,31,32, 33,35,47,49,50, 75,81,94, 102,119, 126, 127, 132, 145, 147,189, 190,191,192,193, 194,195, 198,199,200,202, 215,217,219,223,226,228,

331

231,233,234,235,236,237, 238,243,246,249,250,251, 259,297,314,315, 316,317, 318,319,320,321,324,325 Heth, C.D., 228,251 Heuer, F.,62,72 Hilgard, E.R., 1,35 Hill, W.E.,153,183 Hinsley, D.,31, 35 Hinton, G., 40,72,151, 173,183, 203,216,219 Hitch, G.J., 8,33,67,69 Hochberg, J., 42,72,87,96,173, 204,210,219 Hock, H.S., 292,293,311 Hoffman,D.D., 155,156,183 Hoffman, R.,190, 191,219 Holton, G., 188,219 Holyoak, K.J., 31,33,34 Horowitz, L.M.J., 126,127,147 Hyman, I.E.,Jr., 8,21,24,25,27, 29,35,47,48,49, 73,75,81, 82,84,94,96,99,101,102, 120,128, 137,162, 165,184, 194, 199,234,239,297,318, 320,321,322,324,325

I Intons-Peterson, MJ., 10,28,30,35, 123, 134, 145,147, 181, 190, 194,198,200,202,204,215, 216,219,224,230,240,249, 251,257,258,259,270,284, 294,295, 305,310,320,322, 323,325,326,327

J James, W., 86,96 Jastrow, J., 80,88, 90, 93,152,159,

332

Author h d e x

161,164,184 Jenkins, J.J., 101,120 Johnsen, J.A., 22,36,151,185,203, 220,223,252,289,299,311 Johnson, N.S., 101,121 Johnson, P.,62,73 Johnson-Laird,P.N.,229,251,282, 284 Jolieoeur, P., 157,172,173,184 Jonides, J., 61,73

K Kahn, R., 61, 73 Kalaska, J.F.,64,72 Kanizsa, G., 42,73 KaufmaM, G.,5,8,21,25,28,47, 49,77,102,119,123,124,125, 126,127,130,144,147,148, 194,198,217,226,228,234, 249,300,304,310,317,318, 319,320,321,324 Keenan,J.M., 61,76 Kern, N.H., 61,73 Kihlstrom, J.F., 7,21,23,24,30,36, 81,82,97,154,158,165, 169, 173,174,175,176,184 Kimura, Y., 54,73 Kinbch, w., 101,120 Kintz, B.L.,165,183 Klatzky, R.L., 10,35,203,204,220, 221 Klima, E.S.,61,69 Koestler, A, 256,285 Kohl, D., 295,310 Kolbet, L., 84,85,88, 92,97 Kolers, P.A., 127,148 Kolinsky, R., 203,220,229,251 Kosslyn, S.M.,27,35,39,44,62,64, 69.73. 77. 78. 86. 94.96. 101.

102,118,120, 121,126,127, 133,137,148, 173,184,198, 203,220,225, 226,229,230, 234,251,252,255,285,292, 294,310 Krueger, T.H., 31,35 Kurtzman, H., 39,72

L Lachman, J.L., 227,251 Lachman, R.L., 227,251 Langley, P., 30,31, 35 Larkin, J.H., 299,300,301,310 Leak, S., 62,75 Levine, D.N., 61,62,69 Levy, B., 52,73 Lieberman, K.,61,62,69 Lewis, V.J., 53,56, 69 Lindsay, R.K., 323,328 Lockhead, G.R., 292,293,311 Logie, R.H., 25,28,31,55,59,61, 62,63,69, 73,74,102,106, 128,134,135,142,152,155, 157,198,199,232,234,299, 318,322,326 Lubart, T.I.,217,220,221 Luchins, A.S.,216,220 Lumsden, C.J., 248,250 Lurito, J., 28,34,64,72,328

M Maier, N.R.F., 182,184 Mandler, J.M., 28,35,101,121,229, 251,323,328 Marchetti, C., 61,62,63,69,74 Marks, D.F., 6,35,78,96,132,148 Marmor, G.S., 61,74 Marr, D., 153,156,184 Marschack. A.. 227. 248,251

Author Index Massey, J.T., 28,34,64,72,323,328 Mawer, R.F., 301,311 Matthews, W.A.,62,74 Mayer, R.E., 291,299,310 Mayzner, M.S., 291,311 McAlister, E., 204,210,219 McCloskey, M., 295,310 McDaniel, M.A., 134,147 McDermott, J., 301,310 McKellar, P., 126,132,148 McKim, R.H., 287,307,310,325, 328 Metzler, J., 255,285 Miller, A.I., 1,2,3, 33,35,39,74, 123,148,255,285,321,328 Mitchell, D.B., 292,311 Morais, J., 203,220,229,251 Morris, N., 62,74 Morris, P.E., 62,69,123,127,148 Murray, D., 52,74

N Needham, D.R., 31,36 Neisser, U., 8,24,35,48,72,81,82, 96,99,101,102,120,121,130, 131, 132,148,162,165,184, 229,252 Newell, A., 125,148 Nishihara, H.K.,156,184 Novick, L.R., 31,36,298,311

333

Palmer, S.E., 203,204,210,220, 296,311 Papagno, C., 65,69,74 Parsons, L.M., 173,183,227,252 Pendleton, L.R., 62,76 Perkins, D.N., 278,281,285 Peterson, L.R., 204,210,220 Peterson, M.A., 7,21,23,24, 28,30, 36,42,44,46,47,48,49,72, 77,81,82,84,87,94,97,102, 109,128,137,154, 155,158, 165,169,173, 174,175, 176, 184,194,199,234,236,289, 297,319,320,321,322,323, 324,325 Petrides, M., 28,34,64,72,323,328 Phillips, W.A., 62,74 Pinker, S.,10,23,34,39,74,81,94, 97,100,101,109,117, 118, 120,152,157,183,185,198, 200,215,219,232,234,250, 252,255,258,285 Pope, K.S., 127,149 Price, H.H., 133,149 Price, J.R., 153,184 Pylyshyn, Z.W., 77,97,292,295,311

Q Quinn, J.G., 62,74, 75

R 0 Okovita, H.W., 61,74 Olson, D.,44,74 O'Shaughnessy, M., 59, 75

P Paivio, A., 7,36,61,74,127,148, 180,183

Raaheim, K., 123,148,149 Raichle, M., 230,252 Ralston, G.E., 62,75 Rappaport, I., 59,75 Rawlings, L.,204,210,220 Reed, S.K.,5,22,23,28,30,31,36, 151,185,203,204,220,223, 252,258,285,289,291,292,

334

Author Index

293,299,300,302,311,323, 325 Reisberg, D., 8,22,23,24,25,26, 27,28,31,34,36,39,40,42, 43,44,45,46, 47,48,49, 50,52,53,54,57,59,62,66, 69,72, 76,76,78,79,80,81, 82,83,84,88,89,90,91,92, 93,94,95,97,99,100,102, 103,104,105,108,109,117, 120,121,128,130,131, 135, 137, 140,142,143,145,146, 149,152,155, 156,157,158, 159,161,162, 165,167,168, 183,185,194, 198,199,216, 218,220,232,234,250,252, 258,283,289,299,309,318, 319,321,322,324,326,328 Reiser, B.J., 173,184,292,310 Richards,W.A., 155,156,183 Richardson, J.T.E.,39,52,75,126, 149 Richman, C.L., 292,311 Rock, I., 42,44,75,82,83,86,87, 97,216,220 Roe, A., 2,36 Rose, P.M., 7,21,23,24,30,36,81, 82,97,154,158,165,169,173, 174,175,176,184 Roskos-Ewoldsen,B., 10,21,23,28, 29,30,36,134,147,199,220, 238,245,249,252,258,279, 289,294,310,314,316,318, 319,320,323,324 Roth, J.D., 234,252 Rothenberg, A., 144,145,149,181, 185 Rozin, P.,61,73 Rubin, D.C.,101,120

Rubin, E.,155,185 Ryle, G., 129, 130,132,149

s Saltz,

E.,63,75

Samuels, M.,188,221 Samuels, N.,188,221 Sanford, A.J., 125,149 Schwartz, A.,28,34,64,72,323, 328 Segal, S.J.,79,97 Shand, M.,59,76 Shaw, G.A., 6,7,9,36 Sheikh, A.A., 127,149 Shepard, R.N., 2,4,5,6,7, 8,9,27, 28,29,30,31, 32,34,36,44, 64,76,78,97,123,144,149, 181,182,185,187,188,194, 198,221,227,229,230,252, 255,256,284,285,288,296, 305,306,311,321,322,323, 324,327,328 Shorter, J.M., 130,131, 149 Shwartz, S.P.,101,121 Siegel, R.M., 64,69 Simon,D.P., 301,310 Simon, H.A., 30,31,34,35,123, 125,149,299,300,301,310 Singer, J.L., 127,149 Siple, P., 59,69 Slayton, K.,10,11, 13,14,34,50, 66,69,72,81,96,190,192, 193,198,199,203,204,215, 219,223,231,232,233,234, 239,244,250,257,259,284 Slee, J., 86,97 Slowiaczek, M.,52,76 Smith, G.E.,101,121 Smith, J.D., 23, 25,36,50,51, 52,

Author Index 53, 54,57, 75,76,198,199, 220 Smith, S.M., 194,219,279,280,284 Smyth,M.M., 62,76 Smythe, W.E., 127,148 Sobel, R.S., 144,145,149 Solso, R.L.,1, 37 Sonenshine, M.,23,25,36,50,51, 57, 75,198,199,220 Stein, B.S., 278,283 Sternberg, R.J., 123,150,194,217, 220,221 Sweller, J., 301,311

T Tarr, M., 157,185 Thompson, A.L., 10,35,203,204, 220,221 Thomson, N.,55,69 Tinbergen, N.,23,37,159,185,224, 253 Titchener, E.B., 86,97 Tolman, E.C., 229,253 Torrance, E., 6,37 Tresselt, M.E., 291,311 Trismen, D.A.,303,311 Tsal, Y.,85,86, 87,88,92,97 Tudor, L.,216,220 Tye, M., 198,221

V Valentine, T., 65,74 Vallar, G.,55, 56,65,69,76 van Dijk, T.A., 101,120

W Wallach, R.W., 229,251 Wallas, G., 1, 3,37,270,285,327,

335

328 Ward, M.R., 301,311 Ward, T.B., 194,219,279,280,284 Watson, J.D., 287,312 Weber, R.J.,234,253 Weidenbacher, H.,155,184 Weisberg, R.W., 181,182,185 Wertheimer, M.,188,221 Wheeler, D.,216,220 White, W.,52,53, 76 Wilding, J., 52,53,76 Wilson, M., 25,37,50,52,53, 54,75, 76 Wittgenstein, L., 129,150 Wynn, V.,69

Z Zabeck, L.A., 61,74 Zelmer, R., 300,302,311 Zimler, J., 61,76 Zivin, G.,65,76 Zucco, G.M., 62,74 Zytkow,J.M., 30,31,35

This Page Intentionally Left Blank

337

SUBJECT INDEX A Adaptation 225,227,248 Affective systems 5,20,28-29,327 Merent channel 58,66,127,318, 326 Algorithms 30,125,325 Anecdotal accounts 2-5,9-10,187188, 194,198,255-256 Architects 5,256,288,307 Aristotle 1 Articulation 52,54 -rehearsal loop 55-59 -suppression of 8 Artists 5, 282-283 Associations 175-182,194 Attention 85-86,92 Audition 48-54

B Beethoven, L. 281 Bohr,N.3

C Cardan, G. 5 Cognitive resources 232-236,248249 Componential analyses 156, 157, 163,231 Computer simulation 30 Construals 2,23-27,40-48, 77-92, 99-120,128, 130, 137, 140-145, 152-182,319,321,323,324 Construction 190-203,314-316 Cortex -Parietal 63-64

Creativity -and combinations 9,187-218, 225,258,314-316,318, 320,325 -and concepts 257-283 -and intelligence 6-7,194 -and inventions 39, 192,200218, 258-283,287-309 -and language 5-21,31, 189191, 198-215,257-283 -and motivation 28-29,217218 -and patterns 9-21,191-215, 223-249,256-283 -and people 2-5,30-33,123, 138, 187-189,217-218, 255,317 -and thought 387-309 -and visualization 255-283, 287-309 -as process or product 316-317 -constraints(restrictions)on, 82-84,93-94,102-103, 248-249,257,281,327 -definition of 10, 123, 181, 189-191,314-317,320 -inventions of 1-6,39,193-218, 258-283,287-309 -judgments of 10-11,13-21, 233-247,258-282 -models of 124, 134-137,142145,187-218,231-232, 279-282,315

338

Subject Index

D Decomposition 223,232-249 Demand characteristics 257,294-295 Depiction 8,25,40,44-47,57,68, 77-95,131, 132,317,321 Description 8, 13,22,77-95,128132, 134,317,321 Designers 5 Developmental perspectives 225-227, 230,249,323 -on creativity 27-28,225-227, 230,249 -on imagery 5, 27-28,225-227, 230,249,323 Diagram 31,255,288,299,300,303, 323,325 Didion, J. 3 Dirac 3 Discovery 3, 12-20,39,40-47,99120, 128,202,225-227.232249 -and imagery 1, 10,42-52 -and mental synthesis task 921,203-218,232-249, 287-309 -and drawing 10, 13-21,40-47, 57,99-105,111, 131, 232-249 -as interpretation 187-218 -constraints on 80-84,102-103, 248-249 -creative 12-20,187-218,231249,255-283 -definition of 189-191,256, 314-316 -models of 124,134-137,142145, 187-195,231-232, 279-282,315 -properties of 188

Domains -conceptual 273-274 Drawing -and discovery 10, 13-21,4047,50, 57,80-91,99-105, 112, 114-117,131,152, 159,210,227,234-249, 3 19 -and reconstrual 40-47,80,99105,110,152,159,227, 319 -as externalized imagery 41, 43,48-69,80,102-110, 137-142,232-249, 323324 Dynamic imagery 100,137,188,323

E Efferent channel 58,66,318,326 Einstein, A. 2,3,4,6,144,188,192, 194,255,288 Emerald-Shapery Center 307-308 Emergent patterns 30,195,197-218, 258,279-283,320 Emotions 5 Enacted imagery 25,162-169 Equivalence, functional 126,128, 133, 143-145,299 Evolutionary perspectives 225,227, 229,230,248-249 -on imagery and creativity 2728, 225,227,229,230, 248-249 Expectations 9 -of experimenter 236-239 -of subjects 236-239 Experimenter bias 236-239,257 Expert-novice comparison 4,6-7,9, 32,324

Subject Index Eye of beholder 241-249 Eye of creator 241-249

F Familiarity 13,21-22,216 Faraday, A. 3,4,5,255 Feynman, R. 3,255 Figure-ground 41-43,48-50,83-84, 87,95,155,158,161,165, 166168,195 -organization 41-42,83-84 -reversals 40,50,82-95,158 Figures, ambiguous 8,22-28,4060, 57,79-95,99-109,135, 137138,140-143,152,289,297, 319 -chefldog 99-104,138,140-143, 159-161,166-168,297 -Jastrow ducwrabbit 8,22-26, 40,48,49,50,66,67, 80-82,85,87-93,95,99107,135, 137-138,140143, 152, 159-168,172182,289,297,319 -Fisher snaillelephant 153, 159, 170-172 -goosehawk 23, 159-160,165, 166-167,297 -Hill wifdmother-in-law 138, 153 -Mach book 23, 153, 158, 168 -Necker cube 39,51,103,107, 138,152,153,158,167 -Rubin vasdface 41, 104, 107, 138,159,161 -Schroeder staircase 103, 107, 138, 152 -Texas 43-47,49,82,83, 100102, 104

339

Figures, embedded 22, 151,214, 223,230,233,289,291 Figures, geometric 111-117 Fleming, A. 287 Form -and fundion 277-283 -preinventive 192-283,315 Franks, J. 29,327 Functional perspective 223-232,249

G Geneplore model 195,279-282,315 -generative phase 195,279-282 -representations 195,279-282 -preinventive structures 195 Goodness -of parts 23, 29-30,110-120, 202-218,230,291,296, 320,324 -of patterns 29-30,110-120, 202-218,223-224,230, 237-240,242-244,291, 296,320,324

H Hawking, S. 288 Heisenberg, W.3 Heuristics 30, 125 Homophone judgments 53-55 Hume, D. 133

I Illumination 1,3 Imagery -ability (see Marks’ Visual Inventory Questionnaire) 78 -and affectivdmotivational systems 5,20,28-29,32,

340

Subject Index

327 - a d l h i t a t i ~ n39-48,8044, ~ 86,93-94,98,134, 158, 289-295,304,319,320, 326-327 -auditory 23,25,50-58,199, 326,327 -definition of 133, 135 -dynamic 100, 137,188,323 -enacted 25, 59,62-29 -functional role of 126-137 -grammar of 129-132 -loss of 83 -motonc 59,60-64,68-69 -static 63,323 Images -amount of detail 173,175 -amount of information 39 -and discovery 4,42-47,99120, 129-145,288-309 -and motivation 5, 20,28-29, 32,327 -construal/reconstrual of 2,2327,40-48,77-95,99-120, 128, 130, 137, 140-145, 152-182,319,321,323, 324 -definition of 125-126,131, 134, 151 -depictive 8,25,40,44-47,57, 68,77-95,131, 132,317, 321 -descriptive 8, 13,22,77-95, 128-132,134,317,321 -detection of 130 -drawing of 10,13-21,40-47, 50,57,80-91,99-105, 110, 112, 114-117,131, 152, 159,210,227,319

-enacted 25-59,62-69 -externalization of 41,43,4869,80,102-110,137-142, 232-249,323-324 -fading of 85,87, 137,142,318 -kinetic 188 -orientation of 42-47,87 -prototypical 88,241 -rotation of 10,28,44-47,64, 78,93-95,97, 100, 157, 233,323 -spatial 6,9,27-32,61-63,135, 223, 228,229,287-309, 317,322 -usefulness of 289-309 -vividness of 63,86 Incubation 1,3 Individual differences 15-19,30-33, 63, 119, 138, 141-142,217-218, 229,249,326 Inner -ear 52, 55-59,65 -eye 60-66,132-134 -scribe 58,60-66 -speech 52,57,65 -voice 52-56,58-59 Intelligence 6-7,194 -,see also Creativity Intentions 79,84,95,128,129,130, 133,135,137,142, 144,317 Interpretation 25, 79,99-120,133, 140-145,152, 161-162,176182,190-203,225,231,257, 314-316,318 -of images 8, 25-27,79-95,99120, 133,140-145,152, 161-162,176-182,255, 231, 232-249,257 -of percepts 25-27,79-95,99-

Subject Index 120,133, 140-145,152, 161-162,171-182,225, 231,232-249 Interference 8,9,12,20-21,52,62, 79, 173 Introspection 78,86,188,256,313 Intuitive spreading approach 225227 Inventions 1-6,39,193-218,258-283, 287-309

K Kekul6, F.A. 2, 181,255 Kim, C.W. 307 Kinesthetic cue 52,59 Kofia, K. 86

L Learning 127 -from images 39-53,78-95 Levels of analysis 224-231 -framework 225-231 Levels of processing 126

M Marks’ Visual Inventory Questionnaire 78 Maxwell, J.C.2, 4, 6,144,288 Meaning 79-95,317 Memory -limited capacity 210,214 -shape 24, 151-182 -working 39-69,101, 125,249, 304,319,326 Mental tasks -construction/reconstmction 927,44-45,79-81,99-120, 138-143,158-180 -discovery 12-20,40-47,99-

341

120,158-180,200-211, 223-227,232-249 -paper-folding 296 -rotation 10,27,44-47,64,78, 93-95,97,100,157,233, 323 -Scanning 42,61,64,78,86, 100,292-295 -sQipts 180 --,see also Schemata -simulation 30,295,309 -synthesis 9-21,236-249,258, 277,296-297,315 -in studies of imagery 109-117 -and creativity 256-259 -traditional methods in 256259 -cognitive science 256-259 Mind‘s eye 132-134,198,200 -,see also Inner Eye Minnesota Paper Form Board 138, 217,229 Models -Anderson and Helstrup’s model of visual discovery 231-232 -Bartlett 99-120 -Biederman’s Recognition-byComponents 156-157 -computational 124,225 -computer simulation 30,295, 309 -connectionist 230 -creativity 187-218,279-282, 315 -depictive 8,25,40,44,57,68, 77-95,131,132,317,321 -descriptionalist 8, 13,22,7795,128-132,134,317,

342

Subject Index

321 -discovery 195,279-282 -dispositional 131 -empiricist 133, 134 -equivalence 133-134 -Finke’s Geneplore 195,279282, 315 - W e ’ s model of imagery 133134 - K a ~ f m a24, ~ ’ 134-137, ~ 142145 - K o s s l ~ ’ 118, s 127 -Paivio’s dual code 127 -pictorialist 132-134,137 -reconstructive remembering 99-120 -Roskos-Ewoldsen’smodel of creativity and discovery 194-197 Motivation 5,20,28-29,327 Motoric imagery 59-64,68-69 Mozart, W.A. 2, 3, 281

N Newton, I. 4 Nicklaus, J. 2 Novelty 10,11, 13, 68,93-95,123, 124, 125,181, 195,216, 218, 241,246,247,248,256,258, 278,279,282,288,303,309, 316 Neurophysiological evidence 64-65, 225-227,230 -Parietal lobe 63-64

0 Organization 291 -Ofparts 188,200-218 Orientation 42-48,82-83,93,95,

100, 157 Originality 258,261

P Pascal, B. 5 Patterns 223,233-249,315 -complexity 234,238 -emergent 199-218,247 -goodness 23,29 -pixel 26, 41,93 -properties of 199 Pauli, W. 3 Perception 24-28,40-45,77-95,99120, 124, 125,133,142, 144, 153,193,195,318,320 Percepts 40-42,77-95,151 Perceptual linkages 42 Personality 194 Philosophical issues 129-137,143145 Phonological coding (store) 52 Physicists 3 Pixel patterns 27,41,93 Plato 1 Play -combinational 9,187-218, 225,258,314-316,318, 320, 325 Poincarb, H.1,3,33 Positron emission tomography (PET) 230,249 Practicality 258,261,273,276 Practice effects 13-21,31-32,216, 323,325 Practical implications 214-217,282283,287-309,317-327 Preinventive forms 195,266-283, 315 Preparation 1, 3, 270

Subject Index Problem -definition 123 -finding 278 -programming 123,144 -solving 30-31,39, 47,123, 127, 181,182,288-309, 316, 325-326 --and use of diagrams 31,255, 288,299,300,303,323, 325 --and illustrations 30,288 --and use of imagery 30,47, 181,288-309 --four stages of 1-3,325 --simulation of 30-31 --transfer t o analogous problems 30-31,288-309 Processes -automatic 194 -controlled 194 Productivity -,see also versatility 15-21 Propositions 22,77-78,124,125, 126,151,181,317

Q Quantum mechanics 3

R Ratings -correspondence 13-21,233247, 258-282 -creativity 13-21,233-247,258282 Recognition 190 -of patterns 44-47,190,198218,290 -of shapes 44-47,153, 157, 172, 173,190,198-218,

343

290 Reconstructive memory 99-120,318 Reconstrual 2,23-27,40-48,77-95, 99-120,128,130,137,140-145, 152-182,319, 321,323,324 -structural 176-180,176-178 Reference-frame 21-27,29, 32,4252,57-58,66,67,84,94,153182 -definition of 23,42,153 reversals 23-27,42-52,58, 290,297-298,319 Rehearsal loop 55-56,58 Reinterpretation 22-27,32,40-50, 77-95,99-110,187,200-218, 227,297,319-322,324 -of ambiguous figures 22-27, 40,50,99-110,200-218 -of images 22-27,40-50,77-95, 99-110,227 Relativity 3 Representation 23-24,31,54,58,7795,145, 156,182,223,227230,238,239,241,288 -analog 75, 124 -conscious 124 -depictive 8,40,44,77-95 -descriptive 8-13,22,77-95 -imaginal 30 -mental 30,54,81,95,288 -pictorial 30,288 -propositional 77-78,124,125, 126, 182 Reversals 23-27,29,52,60,66,67, 79-95,99-110,140-145,152182 --,seealso reconstrual, reinterpretation -figure-ground 140-145

344

Subject Index

-reference-frame 23-27,42-52, 58-66,79 Rhyme judgments 53-54 Rigidity 216 Roskos-Ewoldsen, D. 29,327 Russell, W.12

S Schemata 101,117,119,180,321322 Scratch pad 56 Schrodinger, E. 3 Self-reports 2-7,61, 68,307 -,see also anecdotal accounts Spatial ability 6-7 Spatial inferences 223,228,229, 288,300,303-306,309,317, 322 Static imagery 62-63 Stimulus -geometry 42-49,299-300 -support 41,43,48-69,80, 102110, 137-142,232-249, 323-324 Story construction 8, 101-102,110121 Strategies 101, 117, 119,180,321322 -for creative discovery 197-199, 202-218,236-249,260283 -symbolic 124 Structure -description of 22 -exploratory 260 -invariances of 6 -reconstruals 162 -symmetry in 6,29-30,32, 153,324

-three-dimensional 81, 151, 296,307 Subjects -design students 138-143 -engineering students 296 -gifted 6-7 -unselected undergraduates 10,11-21,28-29,40, 204-211,234,236,276, 283,300,320,324 Subvocalization 25-26,51-62, 66-69 Surls, J. 2, 188,194, 198 Synthesis 9-21,236-249,258,277, 296-297,315

T Tacit knowledge 294-295 Tesla, N. 5,9,287,288 Thinking 127,131,144,270,283, 287 -homospatial 145 Time-sharing 56-58 Training to use imagery 29-31,295309,325 Transfer 15 Transformations 8, 10,145,233, 234,238,324

V Verification 1, 3 Versatility 15-21 -principle of 15 -,seealso productivity 15-21 Vividness 86-87 -Marks' Visual Inventory Questionnaire 6,35,78, 86,95

Subject Index

W Watson, J. 2, 6,287,288 Working memory 39-69,101,125, 249,304,319,326 -articulatory loop 55-56,58 -central executive 56 -limited capacity 210,214, 249,304,318,326 -scratch pad 56 Wright, F.L. 307

345

This Page Intentionally Left Blank