Cognitive Science Perspectives on Personality and Emotion [1 ed.] 978-0-444-82450-9

Table of contents :
Content:
List of contributors
Pages v-vi

Preface
Pages vii-ix
Gerald Matthews

Chapter 1 An introduction to the cognitive science of personality and emotion Original Research Article
Pages 3-30
Gerald Matthews

Chapter 2 Conation, affect, and cognition in personality Original Research Article
Pages 31-63
John D. Mayer, Heather Frasier Chabot, Kevin M. Carlsmith

Chapter 3 Introduction to the bidirectional associative memory model: Implications for psychopathology, treatment, and research Original Research Article
Pages 65-122
Warren W. Tryon

Chapter 4 Space-time, order, and hierarchy in fronto-hippocampal system: A neural basis of personality Original Research Article
Pages 123-189
Jean P. Banquet, Philippe Gaussier, Jean Claude Dreher, Cédric Joulain, Arnaud Revel, Wilfried Günther

Chapter 5 Affective influence in perception: Some implications of the amplification model Original Research Article
Pages 193-258
Shinobu Kitayama

Chapter 6 Levels of processing in emotion-antecedent appraisal Original Research Article
Pages 259-300
Carien M. van Reekum, Klaus R. Scherer

Chapter 7 Modeling individual differences in negative information processing biases Original Research Article
Pages 301-353
Greg J. Siegle, Rick E. Ingram

Chapter 8 Emotion and reason: The proximate effects and ultimate functions of emotions Original Research Article
Pages 355-396
Timothy Ketelaar, Gerald L. Clore

Chapter 9 Extraversion, emotion and performance: A cognitive-adaptive model Original Research Article
Pages 399-442
Gerald Matthews

Chapter 10 Motivational and attentional components of personality Original Research Article
Pages 443-473
Douglas Derryberry, Marjorie A. Reed

Chapter 11 Investigating cognitive processes in schizotypal personality and schizophrenia Original Research Article
Pages 475-502
Anthony Beech, Leanne Williams

Chapter 12 Attention, working memory and arousal: Concepts apt to account for the “process of intelligence” Original Research Article
Pages 503-554
Edward Necka

Subject index
Pages 555-558

Citation preview

COGNITIVE SCIENCE PERSPECTIVES ON PERSONALITY AND EMOTION

ADVANCES IN PSYCHOLOGY 124 Editors:

G. E. STELMACH E A. VROON

ELSEVIER A m s t e r d a m - Lausanne - New Y o r k - O x f o r d - Shannon - S i n g a p o r e - Tokyo

COGNITIVE SCIENCE PERSPECTIVES ONPERSONALITY AND EMOTION

editedby Gerald MATTHEWS University of Dundee Dundee, Scotland

1997

ELSEVIER Amsterdam - Lausanne- New Y o r k - O x f o r d - Shannon- Singapore- Tokyo

NORTH-HOLLAND ELSEVIER SCIENCE B.V. Sara Burgerhartstraat 25 p.o. Box 211, 1000 AE Amsterdam, The Netherlands

ISBN: 0 444 82450 2 9 1997 Elsevier Science 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 B.V., Copyright & Permissions Department, EO. Box 52 l, 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), 222 Rosewood Drive, Danvers, MA 01923. 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 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. This book is printed on acid-free paper. Transferred to digital printing 2005

List of Contributors Jean P. Banquet*. Neuroscience et Modrlisation, lnstitut des Neurosciences,

UPMC, 9 quai St Bernard, 75252 Paris cedex, France. Anthony Beech*. Department of Forensic Psychology, Fair Mile Hospital,

Wallingford, Oxfordshire OX 10 9H, England. Jean Claude Dreher. Equipe de Traitement des Images et du Signal (ETIS),

ENSEA/UCP, Umversit6 de Cergy-Pontoise, 6 Avenue du Ponceau, 95014 Cergy-Pontoise cedex, France. Kevin M. Carlsmith. Department of Psychology, Princeton University,

Princeton, NJ 08544, U.S.A. Gerald L. Clore. Deparment of Psychology, University of Illinois at Urbana-

Champaign, 603 East Daniel Street, Urbana-Champaign, IL 61820, U.S.A. Douglas Derryberry*. Department of Psychology, Oregon State University,

Corvallis, OR 97331, U.S.A. Heather Frasier Chabot. Department of Psychology, University of New

Hampshire, Durham, NH 03824, U.S.A. Philippe Gaussier. Equipe de Traitement des Images et du Signal (ETIS),

ENSEMUCP, Universit6 de Cergy-Pontoise, 6 Avenue du Ponceau, 95014 Cergy-Pontoise cedex, France. Wilfried Gtinther. Neuroklinik Bamberg, St Getreu Strasse 14-18, 8600

Bamberg, Germany. Rick E. Ingram. Department of Psychology, San Diego State University, San

Diego, CA 92182-0551, U.S.A. C~dric Joulain. Equipe de Traitement des Images et du Signal (ETIS),

ENSEA/UCP, Universit6 de Cergy-Pontoise, 6 Avenue du Ponceau, 95014 Cergy-Pontoise eedex, France. Timothy Ketelaar*. Center for Adaptive Behavior and Cognition, Max

Planck Institute for Psychological Research, Leopoldstrasse 24, 80802 Munich, Germany.

Contributors

vi

Shinobu Kitayama*. Faculty of Integrated Human Studies, Kyoto University,

Kyoto 606-01, Japan. GeraM Matthews*. Department of Psychology, University of Dundee,

Dundee DD 1 4HN, Scotland. John D. Mayer*. Department of Psychology, University of New Hampshire,

Durham, NH 03824, U.S.A. Edward Necka*. Instytut Psychologii, Uniwersytet Jagiellonski, ul. Golebia

13, 31-007 Krak6w, Poland. Marjorie A. Reed. Department of Psychology, Oregon State University,

Corvallis, OR 97331, U.S.A. Carien M. van Reekum. Department of Psychology, Universit6de Gen6ve, 9,

route de Drize, CH- 1227 Carouge-Geneva, Switzerland. Equipe de Traitement des Images et du Signal (ETIS), ENSEA/UCP, Universit6 de Cergy-Pontoise, 6 Avenue du Ponceau, 95014 Cergy-Pontoise cedex, France.

Arnaud Revel

Klaus R. Scherer*. F.P.S.E. Section Psychologie, Universit6 de Gen6ve, 9,

route de Drize, CH- 1227 Carouge-Geneva, Switzerland. Greg Siegle*. Doctoral Training Facility, San Diego State University, 6363

Alvarado Court, San Diego, CA 92120, U.S.A. W.W. Tryon*. Department of Psychology, Fordham University, Rose Hill

Campus, 441 East Fordham Road, Bronx, New York, NY 10458-5198, U.S.A. Leanne Williams. Psychology Department, University of New England,

Armidale NSW 2351, Australia.

* Corresponding author

Preface We are all cognitive scientists now. Researchers routinely use the language of cognition in developing models of personality and emotion. Constructs such as automatic processing, schemas, working memory, attentional resources and the like are now part of the essential fabric of theory. The popularity of information-processing models offers both a promise and a threat. The promise is that of a true understanding of how the different psychological faculties of perception, attention, memory and so forth are inter-woven to create the whole person, and to create the mtegrat~ adaptive reactions we call emotions. Contemporary cognitive science is at ease with multiple levels of description and explanation, and so is especially well-suited to explaining the origins and expressions of emotion and personality. But do we really speak a common language, or are we heading for a new Babel? Constructs such as schemas and strategies sometimes seem plastic enough to fit almost any theoretical conception, so that the verbal labels become private rather than shared. As subjects of inquiry, emotion and personality are particularly vulnerable to the use of language as artifice rather than as scientific discourse. The decline of psychoanalysis as a scientific enterprise illustrates the nature of the threat. In contemporary research, there is an evident risk of "cognitivism", dressing up untestable ideas in cognitive jargon. The differing perspectives provided by different strands of cognitive research are a strength, not a weakness, but communication between different perspectives requires us to work from common scientific bases. This book aims to highlight the vigour, diversity and insight of the various cognitive science perspectives on personality and emotion. It aims also to emphasise the rigorous scientific basis for research to be found in the integration of experimental psychology with neuroscience, connectionism and the new evolutionary psychology. Collectively, the contributors to this book provide a wide-ranging survey of leading-edge research topics. It is, a little arbitrarily, divided into three parts, on general frameworks for cognitive science, on perspectives from emotion research, and on perspectives from studies of personality traits. In the first, introductory chapter, I begin Part I with a personal view of the impact of the cognitive revolution, and apply the "classical theory" of cognitive science to issues in personality and emotion. As the book took shape, I came to appreciate how much a cognitive science of personality and emotion is necessarily a science of motivation too. In

Preface

viii

Chapter 2, Mayer, Frasicr Chabot and Carlsmith inter-relate these three constructs in the context of the traditional "trilogy of mind": conation, affect and cognition. They procr~ to outline a new "quatcrnity of mind", encompassing consciousness also. One of the most radical and exciting innovations of cognitive science is the use of connectionist models, and the remaining two contributors to Part I provide two different perspectives on their application. Tryon's Bidirectional Associative Memory (BAM) uses the conncctionist metaphor of memory as wells in an energy surface as a source of insight into normal emotion and pathological conditions (Chapter 3). He also outlines how psychotherapy may be directed towards re-landscaping the energy surface, by shrinking memory wells whose diameter gives them too much power over the person's experiences, for example. In Chapter 4, Banquet, Gaussier, Drehcr, Joulain, Revel and G0nthcr describe a more ncurologically-orientod conncctionist perspective on personality. They discuss how the person's sense of identity in space and time derives from circuits in hippocampus and prefrontal cortex, supporting spatio-tcmporal processing, working memory, planning and goal propagation. Part II reviews perspectives derived primarily from emotion research, which explore the interplay between emotion as a common human characteristic and individual difference factors. One of the flaws in an overly cognitivistic conception of emotion is neglect of unconscious, prcattcntivc processes which guide later, attentive processing. Kitayama (Chapter 4) presents the amplification model of affect-cognition interaction in early perceptual processing. The model describes how the emotional content of stimuli may either enhance or impede subsequent conscious rccognition, explaining phenomena such as "perceptual dcfencc". Van Rcckum and Schcrcr (Chapter 5) also address distinctions between different levels of processing, in the context of appraisal, which may be supported by sensorymotor, schematic or conceptual processing routincs. They review ncuroscicncr bases for appraisal, and link personality to different appraisal characteristics. In Chapter 6, Sicgle and Ingram explore conncctionist modelling of the negative biases in cognition characteristic of depression and other emotional disorders, expressed in appraisal, attention and memory. They focus especially on lcxical decision and valence identification as tasks which bring to thc surface the abnormalities of processing underlying pathology. The pcrspcctivc from evolutionary psychology is presented in Chapter 7 (Kctelaar and Clorc), which discusses the long-term adaptive significance of emotions, as informative and motivational signals. The authors review evidence suggesting that analysis of the evolved functions of

Preface

ix

emotions helps us to understand their more immediate effects on cognition in experimental studies. Part III is oriented towards research on personality traits, within a loosely Eysenckian framework, with contributions relating to the three superfactors of extraversion-introversion, neuroticism (anxiety) and psychotir (schizotypy). Perhaps a future volume of this kind will be able also to cover additional dimensions from the five factor model; conscientiousness, agreeableness and openness to experience. In Chapter 9, I present a cognitiveadaptive model of extraversion, which reviews information-processing correlates of the trait in the context of adaptive specialisation. Extraverts may be superior in verbal facilities such as short-term recall, retrieval and multitasking because these cognitive characteristics contribute to coping with their preferred environments. Derryberry and Read (Chapter 10) discuss the relationship between motivational and attentional aspects of anxiety, from the standpoint of cognitive neuroscience. Experimental data illustrate anxietyrelated biasing of specific attentional functions which may contribute to shaping the higher-level cognitions and motivations of anxious individuals. Beech and Williams (Chapter 11) assess the cognitive bases for schizophrema and schizotypal personality. They develop a model of activation and inhibition processes which explains priming data obtained experimentally, and the positive symptomatology of schizophrenia such as delusions and hallucinations. Finally, contemporary trait researchers are increasingly engaged with exploring the relationships between personality and ability traits. In Chapter 12, Necka links intelligence, extraversion and neurotir to an attentional resource model. Both personality and ability traits are related to arousal processes, whose impact on cognition is shown in experimental studies of dual-task performance and memory scanning. I am grateful to the Medical Research Council for their support for my research while this book was in preparation. I would also like to thank the contributing authors. I have enjoyed reading and re-reading the chapters, and my schemas and networks are greatly enriched. This is the book I would have liked to have read when I first began researching personality and emotion as a doctoral student in the early 1980s. I hope it will serve as an inspiration and a guide to all those with an interest in this exciting new research area.

Gerald Matthcws

This Page Intentionally Left Blank

Contents

P A R T I. F R A M E W O R K S F O R C O G N I T I V E S C I E N C E

Chapter 1. An Introduction to the Cognitive Science of Personality and Emotion ....................................................................... 3 Gerald Matthews Landmarks of the Cognitive Revolution .............................................. 3 A Cognitive Science Framework ...................................................... 7 Towards a Cognitive Neuroscicnce of Personality and Emotion? ......... 13 Developing Adaptive Explanations .................................................. 15 An Example: Explaining Anxiety and Cognition .............................. 20 Conclusions .................................................................................... 24

Chapter 2. Conation, Affect, and Cognition in Personality ................... 31 John D. Mayer, Heather Frasier Chabot and Kevin M. Carlsmith The Relational Model of Personality .................................................. 32 Understanding Conation, Affect, and Cognition ................................. 39 The Quaternity of Mind and Personality Dynamics ............................ 52 Conclusions and Other Considerations .............................................. 60

Chapter 3. Introduction to the Bidirectional Associative Memory Model: Implications for Psychopathology, Treatment, and Research .......................................................................65 Warren W. Tryon Bidirectional Associative Memory (BAM) ......................................... 67 Encoding Emotion ............................................................................. 70 Implications for DSM-IV Disorders .................................................. 75 Treatment ......................................................................................... 92 Research Strategies ........................................................................... 99 Conclusions .................................................................................... 101 Appendix: Description of the Bidirectional Associative Memory ....... 109

Contents

xii

Chapter 4. Space-Time, Order, and Hierarchy in FrontoHippocampal System: A Neural Basis of Personality .......................... 123

dean P. Banquet, Philippe Gaussier, Jean Claude Dreher, Cddric Joulam, Arnaud Revel and Wilfried G~tnther Hippocampal Function: An Extended View ..................................... Working Memory as Both a Cortical and a Hippocampal System ..................................................................................... Neuropsychology, Brain Imaging and Working Memory ................... Neurophysiology: Human Versus Animal Working Memory ............ Spatio-Temporal Processing in Hippocampus and Prefrontal Cortex ...................................................................................... Functional Model ............................................................................ Fronto-Hippocampal Function and Personality ................................ Conclusion .....................................................................................

126

129 135 148

151 159 176

179

PART II. PERSPECTIVES FROM EMOTION RESEARCH Chapter 5. Affective Influence in Perception: Some Implications of the Amplification Model ..................................... 193

Shinobu Kitayama The Amplification Model of Affect-Cognition Interaction ................. 196 Evaluation Criteria of the Amplification Model ................................ 202 Experiment 1 .................................................................................. 212 Experiment 2 .................................................................................. 221 The Amplification Model Evaluated ................................................ 230 Relations with Extant Theories of Attention ..................................... 232 Amplification of Attention in Other Domains ................................... 235 Perceptual Defense and Vigilance? .................................................. 238 Future Research Directions ............................................................. 240 Concluding Remarks ....................................................................... 242

Contents

xiii

Chapter 6. Levels of Processing in Emotion-Antecedent Appraisal .............................................................................................. 259

Carien M. van Reekum and Klaus R. Scherer Critique of Appraisal Notions ......................................................... Levels of Processing in Appraisal ........................... : ........................ Hierarchical Process Notions in Related Traditions .......................... Issues in Rewriting Appraisal Theory .............................................. Individual Differences in Appraisal Processes .................................. Conclusions ....................................................................................

260 263 266 277 280 289

Chapter 7. Modeling Individual Differences in Negative Information Processing Biases .............................................................. 301

Greg.1. Siegle and Rick E. Ingram Personality Research and Vulnerability to Depression: A History ...... 302 Simulating Aspects of Depression and Personality on a Computer ....................................................................................... 304 Simulating Personality Factors ........................................................ 320 A Brief Conclusion ......................................................................... 348

Chapter 8. Emotion and Reason" The Proximate Effects and Ultimate Functions of Emotions ........................................................... 355

Timothy Ketelaar and GeraM L. Clore Why Does Emotion Affect Cognition? ............................................. 356 Specific Aims of this Chapter .......................................................... 358 Consequences of Mood ................................................................... 360 Consequences of Emotions .............................................................. 365 Emotion-as-motivation and Frank's (1988) Commitment Model ....... 371 Affect-as-lnformation and Behavior ................................................ 378 The Future of Affect and Information Processing ............................. 387 Conclusion: Deficits, Biases, and Functions ..................................... 388

Contents

xiv

P A R T IIl. P E R S P E C T I V E S F R O M P E R S O N A L I T Y TRAIT RESEARCH

Chapter 9. Extraversion, Emotion and Performance: A Cognitive-Adaptive Model ................................................................ 399

Gerald Matthews ..

Extravorsion and Affect .................................................................. Extraversion and Performance ......................................................... Extraversion, Arousal and Attontion: Empirical Studies ................... An Adaptivr Framowork for Cognitive Correlates of Extraversion-lntroversion ................................................................ Conclusions ....................................................................................

400 405 409 426 434

Chapter 10. Motivational and Attentional Components of Personality ............................................................................................ 443

Douglas Derryberry and Marjorie A. Reed Biological Approachos to Personality .............................................. Assessing Attcntional Processes in Anxiety ..................................... Extensions to Complex Cognitive Processing ................................... Conclusions ....................................................................................

444 450 462 466

Chapter 11. Investigating Cognitive Processes in Schizotypal Personality and Schizophrenia .......................................... 475

Anthony Beech and Leanne Williams Mechanisms of Selective Attention .................................................. Experimental Investigations of Inhibitory Processes ......................... Inhibitory Processes in Schizophrenia .............................................. Towards a "Roducexl Cognitive Inhibition" Model of Schizophrenic Symptomatology ...................................................... Revising the Model ......................................................................... Conclusion .....................................................................................

477 478 485 490 494 497

Contents

xv

Chapter 12. Attention, Working Memory and Arousal: Concepts Apt to Account for the "Process of Intelligence" ................. 503 Edward Necka Theoretical Notions ......................................................................... Assumptions ................................................................................... "The Process o f Intelligence" ........................................................... Preliminary Empirical Data ............................................................. Cognitive Science Perspectives ........................................................

504 512 519 525 542

Subject index ......................................................................................... 555

This Page Intentionally Left Blank

PART I FRAMEWORKS FOR COGNITIVE SCIENCE

This Page Intentionally Left Blank

Cognitive Science Perspectives on Personality and Emotion-G. Matthews (Editor) 1997 Elsevier Science B.V. CHAPTER 1

An Introduction to the Cognitive Science of Personality and Emotion GeraM Matthews

The cognitive revolution has transformed the face of research on personality and emotion. Information-processing theories spring up like poppies in a cornfield, and often wither just as quickly, crowded out by more recent growth. With cognitive approaches so firmly established, it is timely to stand back a little from the intellectual ferment, and take stock of the achievements and limitations of the research area. This book represents the leading edge of research on the cognitive science of personality and emotion, and so the contributions concern a variety of specific topics. But for personality and emotion research to mean anything at all, it must above all be integrative. Both constructs hang on a multi-layered web of data and hypothesis, spanning the gamut of psychological phenomena from neuronal firing to social interaction. The aim of this introductory chapter is to outline the overall framework provided by cognitive science, and its place in personality and emotion research. In this chapter, I will sketch the progress so far of the cognitive revolution in personality and emotion (PE) research. I will then describe the classical model of cognitive science, and its three levels of explanation: the biological, the symbol-processing and the knowledge levels. Cognitive science emphasises that information-processing models are necessary but not sufficient for understanding, l will show that cognitive science explanations provide new perspectives on some old problems, and demonstrate its integrative potential by outlining its application to anxiety.

Landmarks of the Cognitive Revolution Emotion and cognition

The cognitive science of emotion has disparate roots, which demonstrate the diversity of "cognitive" approaches. The information-processing approach is based on empirical, performance-based studies of emotion, addressing problems such as the deleterious effects of anxiety on attention. It

4

Chapter 1

accommodates the various conccptualisations of emotion: as a universal but situationally-contingent human response, as an individual difference factor, and as a property of stimuli (valence). Emotion may be conccptualiscd as a dependent variable influcnce~ by processes such as appraisal, or as an independent variable which itself influences information-processing. The more sophisticated applications of the approach (e.g. Ingrain, 1984) build in feedback from appraisals of performance back into emotion. In linking emotion to bchaviour, the basic research tactic is to demonstrate moderation of effects of emotion on performance by task factors, the standard technique of experimental cognitive psychology. Emotion • task parameter interactions inspire processing models which may then be subjected to further test. The approach scores highly on scientific rigour, but like much cognitive psychology, risks degenerating into an account of the minutiae of a specific experimental paradigm with little wider relevance (cf. Neisscr, 1976). An alternative approach is design-oriented: what might be the purpose of emotion within the cognitive system as a whole? Simon (1967) linked emotion to an interrupt function, contributing to people's capacity to adapt to unpredictable environments by switching back and forth between different goals. This analysis begs the question of why the interrupt function requires all the various concomitants of emotion such as physiological arousal, biases in thinking, action tendencies and the like. Research in the Artificial Intelligence (AI) tradition simulates complex, goal-directed systems to discover basic design principles, indicating, for example, what other features arc required for interrupts to work properly. This approach generates rich and thought-provoking data, but its scientific rigour is open to question. Argument tends to proce~ by analogy and comparison of features of artificial and human systems, and it is unclear that the parallels drawn arc open to falsification or to formal test against alternative explanations. A third tradition derives from stress and clinical research, and the observation that negative emotions derive from the way people interpret and manage events, rather than from fixed properties of the events themselves. It is exemplified by the work of Lazarus (1991; Lazarus & Folkman, 1984) on the transactional model of stress, and the roles of appraisal and coping within specific, potentially stressful encounters. As theory, it has some of the characteristics of both the design and information-processing approaches. Like the design approach, it is explicitly systems-based, with emotion conceptualiseA as a "core relational theme" charactcrising the personenvironment system as a whole. However, like the information-processing tradition, transactional theory attempts to establish local cause-and-effect

G. Matthews

5

relationships open to direct empirical test, such as the effect of appraisals on emotion. Similarly, clinical accounts of anxiety and depression which emphasise the role of the person's self-knowledge and reasoning processes in generating negative emotion as an overall indicator of system functioning (e.g. Beck, 1967; Ellis, 1962). The advantages of such approaches are depth of insight obtained into the experiences of people in real environments, and practical applications to stress management and cognitive behaviour therapy (Matthews & Wells, 1996). Their shortcomings relate, first, to emphasis on self-report data, which may present a partial and distorted view of underlying processing, and, second, as with the design tradition, to difficulties in rigorous theory testing. Finally, the neuroscience of emotion has become increasing cognitive in orientation, as traditional arousal theory has fallen from favour (e.g. Robbins, 1986). Increasingly, it has becomes possible to align specific neural circuits with information-processing and behavioural function (e.g., Gray, 1982). Emotion is notoriously diffmult to localise, but advances in brain scanning technology, and in simulation of neural function are a source of optimism for the future. Pessimists focus on the extent to which feelings are intertwined with thinking, and consequent difficulties in discriminating neural and cognitive influences. There remain fundamental disagreements over the extent to which psychological phenomena are reducible to neural processes (see Lazarus, 1984, 1991, and Gazzaniga, 1992, for the end-points of the continuum of views). At the least, though, computational theories permit testable predictions concerning neural influences on behaviour, contributing to the development of cognitive neuroscience models of emotion. Personality and cognition

Much of personality research is structure- rather than process-oriented, and so unaccommodating to cognitive perspectives. The current popularity of the Five Factor Model owes much to the prodigious empirical programmes of researchers such as Costa and McCrae (1992) in deriving the Big Five as a structural description of various data sets. Personality trait theories have often been based on somewhat naive biological or conditioning models, inspired by Pavlov and J.B. Watson rather than by contemporary research. Arousal theory, in particular, has proved to be a mixed blessing. The concept undoubtedly has integrative value (K.A. Anderson, 1990), and the basic principle that personality reflects biology is becoming increasingly securely supported by behaviour and molecular genetic studies (Loehlm, 1992: Lesch

6

Chapter 1

ct al., 1996). Eyscnck and Eyscnck's (1985) application of arousal theory to personality has scored some notable empirical successes in predicting cxtraversion-introvcrsion effects on sensory thresholds and simple conditioning tasks. Unfortunately, psychophysiological data on personality arc confusing and inconclusive, and arousal theory has proved to be a poor basis for predicting personality effects on cognitive tasks (e.g. Matthews, 1985; Matthews & Deary, in press). Despite the conservatism of much personality research, there arc increasing signs that the cognitive revolution is taking root in this area also. As in the case of emotion, its expressions arc diverse. Information-processing analyses of personality effects on performance arc becoming increasingly common. The trail has been blazexl by research on anxiety traits, driven by the observation that cognitive worry is more predictive of performance than emotional and physiological tension. Detrimental effects of anxiety arc now routinely explained in terms of constructs such as attentional capacity (Sarason, Sarason, & Pierce, 1995) and working memory (Eyscnck, 1992). Humphmys and Rcvr162 (1984) have proposed an ambitious integration of individual differences research which links achievement motivation, anxiety and impulsivity to arousal and effort, which in turn influence availability of multiple resources for performing attcntional and working memory tasks. There is also a rather separate tradition with a basis in social-cognitive psychology, concerned with the knowledge structures which support personality, such as the self-schema (Cantor & Zirkcl, 1990). This approach supports some information-processing work, such as studies of self-referent processing (Klein & Loftus, 1988) and priming (Bargh, Chaikcn, Govcndcr, & Pratto, 1992), but also leans heavily on qualitative and self-report data. Hence, it resembles the transactional approach to emotion: its allegiance is to cognition but not necessarily to cognitive science. On the other hand, it is sufficiently flexible to br applied to both nomothctic and idiographic aspects of personality, and engages with individuals' actual life experiences.

Integration of personality and emotion research The distinction made between personality and emotion is artificial to the extent that much personality research has an explicit trait-state orientation, within which personality effects arc mediated by emotional states (e.g. Spiclbcrgcr's, 1966, anxiety theory). We cannot do personality research without consideration of emotion, but the converse also applies. Some studies of mood make a strong equation between positive and negative affect on the

G. Matthews

7

one hand, and extraversion and neuroticism on the other. Individual differences in mood may substantially reflect individual differences in reward and punishment systems said to be the basis for extraversion and neuroticism (Watson & Clark, 1992). Unfortunately, taken to the extreme, this approach leads to a dreary tautology, such that some unfortunates have negative genes, negative brains, negative emotions and negative personalities, and little more can be said. More promising are interactionist approaches which emphasise that individual differences in emotional response are not mechanically linked to personality, but depend on a more complex interplay between person and environment. Within the transactional model, personality is seen as biasing the appraisal and coping processes which are perhaps more direct influences on emotion (Matthews & Deary, in press). Interactionism can easily degenerate into an unfalsifiable everything-affects-everything position, but computational models can potentially supply much needed precision to theory in this area. Information-processing analyses of performance frequently attempt to discriminate trait and state effects on different processing components. Eysenck's (1992) review of the area suggests that simple traitstate models, within which trait effects are entirely mediated by gross state constructs, are not viable: trait anxiety may sometimes influence cognition and behaviour even with state anxiety controlled. Integration of trait and state research requires a more sophisticated view, such that traits affect stable parameters of processing systems which moderate their reactions to stimuli. We might link traits to knowledge structures in long-term memory (LTM) which feed into appraisal and coping (Wells & Matthews, 1994), or, from a connectionist perspective, to parameters of networks which govern the spread of activation (Matthews & Harley, 1993). In either case, moderating effects of traits are apt to be subtle, and require careful modelling.

A Cognitive Science Framework The brief overview above demonstrates the vigour of the cognitive approach to P E. It also shows that progress has been uneven, and the diversity of differing "cognitive" approaches. We require a general framework for examining where cognitive research has been most successful, and where its impact has so far been limited. Fortunately, the "classical theory" of cognitive science provides a ready made framework, discriminating different levels of explanation. Next, I will outline these levels, and discuss their application to PE research.

8

Chapter I

Pylyshyn (1984) presents a detailed analysis of knowledge, symbolprocessing and biological levels of explanation, from which the following account is derived (see Figure 1). The central point is that psychological events are open to qualitatively different explanations. Suppose we observe an extraverted man at a party, engaging in cheerful social interaction. How do we explain this bohaviour? One approach is to refer to his motives and goals. Perhaps he is a newcomer, and wishes to make new friendships from which he will benefit. This level of explanation is the knowledge-based or semantic level. It is concerned especially with the way the cognitive system is designed for adaptive interaction with the external environment, in pursuit of its goals. It has been developed in PE research through AI approaches to understanding emotion, through work on the adaptive functions of PE, and through social knowledge approaches to personality. Alternatively, we might present an account based on the formal cognitive architecture: a computational description of the processing structures and operations linking inputs to social behaviours. We may then identify individual differences in specific computations, such as spee~ of accessing items of social knowledge, which explain the individual's social behaviour at the processing level. Explanations of this kind are concerned with the formal characteristics of processing, rather than with the adaptive significance of processing routines. They provide the basis for much of the extensive research on information-processing models of emotion and personality previously described. Classical theory requires the architecture to be based on discrete symbols, expressing propositions. Pylyshyn distinguishes sub-levels of algorithm and functional architecture, which differentiate the logical operations performed on symbols form the cognitive structures implementing symbol processing. The centrality of symbols is a controversial area. Some authors place symbol-based accounts of processing centre stage, due to identifiability problems of modelling functional architecture (J.R. Anderson, 1990). Conversely, connectionist models see network implementations as a more powerful method for modelling behavioural data than symbolic accounts, and may even reject symbolic representations as irrelevant to theory (Smolensky, 1988). I will take the view that, in the light of the successes of connectionism, an a priori commitment to symbolic accounts may be too constraining for PE research. I will use the term "architectural explanation" to refer to explanation in terms of the formal properties of the processing machinery, irrespective of whether or not it is symbolic in nature.

G. Matthews

Knowledge

=

Goals, intentions and personal meaning, supporting adaptation to external environments

Algorithm Symbol

9

=

Formal specification of program for symbol manipulation

Functional _ Architecture

Real-time processing operations supporting symbol manipulation

/

processing

Biology

=

Physical, neuronal representationof processing

Figure 1. Levels of explanation in cognitive science.

Finally, we may look to the functioning of the neural hardware for explanation. We might use brain scanning techniques to investigate which neural structures and circuits are active during social interaction, and develop a theory linking the individual's social behaviour to the activity of the circuits concerned. We must then tackle transducaon problems; the conversion of analogue physical events into symbolic codes (Pylyshyn, 1984), or other abstract codes. In the next section of this chapter, I develop the position that information-processing models of PE are necessary but not sufficient for understanding. Processing models possess the rigour provided by computational specification, and, if adequately formulated, are readily testable against empirical data. However, a processing description of PE phenomena requires supplementation with explanations which look both downwards, to architecture and cognitive neuroscience, and upwards to knowledge-level explanations. Information-processing models." Strengths and limitations

Information-processing models of personality and emotion have an impressive track record in characterising empirical phenomena in terms of constructs such as resources, processing stages and activation of network units. The application of such models is demonstrated throughout this

10

Chapter I

volume. Multi-level models, distinguishing qualitatively different types of processing, such as stimuhs-drivcn and strategic processing, have been particularly successful in explaining empirical data (see van Rcekum & Scherer, this volume). Processing models are essential for predicting and understanding the correlates of P E, and they arc increasingly finding applications in the clinical domain (see Beech & Williams, Sicgle & Ingrain, Tryon, this vohmc). However, it is important to be clear about what such models provide and do not provide. Most models provide a snapshot description of processing at a single time epoch, although there is growing interest in learning models (e.g. Kanfcr & Ackerman, 1989). Such a description leaves open alternative types of explanation. The first question is whether effects of PE factors on processing reflect genuine differences in cognitive architecture, or differences in strategy, i.e. how the same architecture is used to support different processing sequences within a given context. It is unlikely that PE has dramatic effects on architecture; we would not expect syntactic deep structure to vary across individuals, for example (cf. Pinker, 1994). Perhaps more likely are quantitative cross-individual or cross-occasion differences in system components such as resource availabilities, short-term memory (STM) slots and speed of execution of key processes (e.g. Nccka, this vohmc). The architecture may also handle emotional stimuli differently to neutral stimuli (Kitayama, this volume). Architecture as a source o f variation

Care is needezl in showing that variance in processing reflects variation in architecture, as opposed to variation in strategy and intention (Pylyshyn, 1984). A strategy may be defined as a goal-dircctexi, voluntarily-imtiatcd processing routine. Typically, a strategy is implcmcnteA and regulated through executive processes which bias involuntary processing (see Norman & Shallice, 1985). There are rather few instances of attempts to establish systematically whether PE phenomena are strategy-dependent, although effects of emotion on strategy-insensitive processes such as early stimulus analysis (Kitayama, this volume) and procedural learning (Corr, Picketing, & Gray, 1995) are suggestive of architectural differences. More generally, processes of interest depend on both the fixed architecture and strategy, and it is difficult to disentangle the two types of influence. For example, extraverts tend to show greater STM capacity than introverts (Matthews, 1992), but this effect might reflect either individual differences in cognitive architecture, perhaps derived from physiological processes (Eysenck & Eysenck, 1985), or

G. Matthews

11

from extraverts' choice of coding strategies which tend to enhance short-term recall at the expense of long-term recall (Schwartz, 1975). If a PE effect on architecture is established, explanatory questions remain. One possibility is that PE variance in architecture reflects relatively straightforward properties of the brain. The neural substrate for emotional states may influence the formal properties of processing over short timescales. Given the heritability of personality traits, including traits related to emotionality, it is plausible that genes code for individual differences in architecture. Alternatively, the architectural difference may be more readily conceptualised as a learning effect, such as changes in control structure associated with "proceduralization" of knowledge (Anderson, 1982). We may also ask if individual differences result from biological bases for learning, or from socially-influenced exposure to learning opportunities: each level of explanation poses further questions. Strategy choice and adaptation PE effects on processing may derive not from architecture but from strategy choice. Architectural accounts of strategy implementation which describe specific executive functions (e.g. Shallice, 1988) are important but incomplete. We need also to address knowledge level questions concerning the person's goals, and choice of strategy to meet those goals. Again, answers generate new questions. How has the person acquired the goals concerned? How does the person's knowledge of strategies, such as strategy efficacy in the current context, feed into strategy choice? At one level we can answer such questions through addressing the cognitive and social factors which influence motivations and associated learning (e.g. Bandura, 1977). Understanding strategy choice may requires understanding of the Shaping of cognition within the wider social matrix, through the person's attempts to meet social norms, negotiate shared identities with others, and generally adapt to social demands (Hampson, 1988). Mayer, Frasier Chabot and Carlsmith (this volume) provide a detailed discussion of the inter-relationship between motivation, emotion and cognition. A radically different perspective is provided by evolutionary psychology (Tooby & Cosmides, 1992). The person's most important life goals are influenced by the set of genetically programmed mechanisms for solving specific evolutionary problems. Some proximate goals such as "stay warm" may be directly coded. More generally, the individual's goals are indirectly influenced by the structuring of experience imposed by the set of adaptive

12

Chapter 1

mechanisms, which, at the least, is likely to signal that certain types of stimuli and encounters are of special significance. In particular, the motivations which tend to accompany emotional states (e.g. avoidance as a correlate of anxiety) are likely to reflect adaptive pressures. There is an argument too that specific strategies, such as the decision rules used in "Prisoner's Dilemma" social encounters (Ketelaar & Clore, this volume), may be directly encoded (Cosmides & Tooby, 1992). However, the evolutionary psychologist's description of a "strategy" carries no commitment to a particular informationprocessing mechanism. The strategy might be implemented through architecture, or, alternatively, through cxxling for motivational factors. Evolutionary psychologists have perhaps shown insufficient interest in whether strategies in the evolutionary sense are contingent upon implementation of strategies in the information-processing sense previously defined. Strategies for processing reflect voluntary control and potentially complex, contingent decisions which may not be related to geneticallyprogrammed adaptations in any simple way. Evolution is an essential part of the backdrop to understanding the inter-relationship of PE and cognition, but it is simplistic to imagine that every such relationship may be traced back to the operation of an adaptive mechanism (of. Lazarus, 1991). Two qualifications are required here. First, definitions of "adaptation" differ confusingly. To evolutionary psychologists the term refers to genetically-programmed mechanisms. I prefer Lazarus' (1991) broader usage of the term to refer to any attempt to manage the demands and opportunities of an environment, which leaves open the utility of an evolutionary analysis. I will use "adaptation" subsequently in this broad sense, unless otherwise indicated. Second, in emotion research especially, it is important to distinguish explanations for emotion as a human characteristic from individual differences in emotion and associated behaviour. Adaptive explanations at the species level do not necessarily generalise to explanations for individual differences. In summary, processing models are only the beginning of the cognitive science enterprise. For further explanation, we may look either towards a reductionist approach of seeking PE effects on the cognitive architecture, which may be supported by neural mechanisms. Alternatively, we may adopt a more systems-orientexl holistie approach of establishing strategy effects, and their role in the person's adaptation to the physical and social environment. We may also nee~ to consider how neural systems, processing and motivations have been shaped by evolution. The new evolutionary psychology provides a different kind of adaptive, knowledge-level explanation to that

G. Matthews

13

afforded by motives for personal strategy choice. Next, the prospects for developing these complementary levels of explanation are discussed further. Towards a Cognitive Neuroscience of Personality and Emotion?

Investigations of the neuroscience of PE have been dogged by two fundamental problems: the use of over-generalised constructs, exemplified by general arousal theory, and nagging doubts about the causal status of physiological constructs. Criticisms of arousal theory are familiar. In brief, there are four sources of difficulty (Matthews & Amelang, 1993). Empirical criticisms focus on the failure of arousal theory predictions: the supposed inverted-U relationship between arousal and performance is simply not robust (Matthews, 1985; Neiss, 1988). Methodological criticisms relate to weaknesses in inference from empirical data, such as the difficulty in falsifying arousal theory within typical stressor-interaction designs (Hockey, 1984). Psychometric criticisms point to the failure of alternative arousal measures to intercorrelate, implying that the construct cannot be operationalised (Lacey, 1967). Conceptual criticisms concern the construct validity of "arousal" and "performance", both of which are multi-faceted (Hockey, 1984; Robbins, 1986). Hockey's cognitive critique of arousal theory is especially important: "arousal" effects vary across stressors and processing functions, and are often associated with subtle strategic effects rather than changes in parameters of the architecture. None of these considerations rule out the possibility of a better arousal theory. Such a theory would require the discrimination of different circuits whose overall activity might influence processing, a description of the specific information-processing functions sensitive to each circuit, and satisfactory methods for manipulating and measuring these multiple arousal dimensions independently. Various multi-dimensional arousal theories have been proposed (e.g. Sanders, 1990), but none have succeeded in explaining more than a small part of the empirical data. A severe barrier to theory development is the sheer complexity and interactivity of neural systems. In the personality context, Zuckerman (1991) points out that there is no one-toone mapping between neural systems and personality traits. He sees each trait as supported by several systems, and, conversely, each system feeds into several traits. Hence, even if neurological reductionism is correct in principle, it may be difficult to establish in research practice. The other basic criticism of the psychobiological enterprise may be traced back to peripheralist views of emotion and the Jamesian view that

14

Chapter I

emotion derives from perceptions of physiological reactions, perhaps through the appraisal and evaluation of autonomic nervous system activity (Schachter & Singer, 1962). The logic of this approach may be extended by denying physiological reactions any special status. Emotions may be constructed from appraisal of a variety of cues, from the external physical and social environment, as well as from physiological reactions. In contemporary research, this position has been expressed most forcefully by Lazarus (1984, 1991) who argues that the influence of physiology is always shaped by appraisal and cognition. Lazarus (1991) does suggest that there may be qualitatively different types of appraisal, trading off speed of processing against depth and complexity, which might be loosely associated with different brain structures. However, explaining how different modes of appraisal influence emotion is a cognitive- rather than a brain-level question: the distinction is between two different cognitive modules. In terms of the current framework, the explanatory questions are how the architecture supports different types of appraisal, and how appraisal and emotion are driven by adaptation to the environment. Van Reehan and Scherer (this volume) argue that multiple levels of processing must be distinguished in relating appraisal to emotion and brain mechanisms. Despite the difficulties outlined, there are several promising lines of research which elucidate mappings between brain and cognitive processes. One approach is the fine-grained analysis of neural pathways (e.g. Banquet et al., this volume; Gray, 1982; LeDoux, 1995). Gray's (1982) account of the septo-hippocampal system (SHS) as the basis for anxiety and behavioural inhibition demonstrates the potential of this approach. He explicitly describes the SHS as performing processing functions, such as calculation of the mismatch between current sensory events and expectancy. Processing is mapped onto brain circuitry to an impressive degree. However, as with other animal models, fundamental questions concerning the coding of information are left open. Some system components are clearly non-propositional, such as the "enabling signal" which gates output from the SHS, and biasing effects of ascending afferents associated with arousal. The system must also make and verify predictions about the world, a process which, in humans, we might imagine to be prepositionally coded. Rats and people may process information differently, of course, but, in any case, it is difficult to develop the theory as a cognitive account of human emotion when the computational basis for comparator function is uncertain. Predictions from Gray's theory have met with mixed success, in part, because of difficulties in operationalising its constructs in human subjects (Pickering, Diaz & Gray,

G. Matthews

15

1995). It has been most powerful when modified through integration with human cognitive neuroscience (see Derryberry & Reed, this volume). Perhaps the most promising solution to the coding problem is the use of connectionist models (see Banquet et al., this volume). It is emphasised that connectionist models do not necessarily correspond to actual neural net processes: Smolensky (1988) describes a variety of important differences between neural functioning and the connectionist architectures typically applied to psychological problems. However, connectionist models do possess some of the key formal properties of nerve cell assemblies. They comprise linked elementary processing units representing analogue information only ("activation"), which is transmitted through associative pathways. There is no direct representation of symbols, which is assumed to be distributed across units, and learning is a direct consequence of the formal properties of the net. Modelling allows testable predictions to be derived concerning behavioural consequence of neural function, as illustrated by Cohen and ServanSchreiber's (1992) work on the consequences of abnormality in dopamine function for attention in schizophrenics. However, connectionism is scientifically valuable irrespective of whether activation corresponds directly to neural functions such as rate of firing; typically, activation is best treated as a formal attribute of the cognitive architecture (of. J.R. Anderson, 1990).

Developing Adaptive Explanations Knowledge level explanations in P E research address questions of adaptation. I will take an "adaptive explanation" as a demonstration that expressions of emotion or personality arc functionally useful in achieving personal goals or dealing with environmental demands. Traditionally, personality theory has been much concerned with the challenges posed by the interplay between basic drives such as sex and power-seeking in an often threatening and unaccommodating world, as expressed in various psychodynamic theories. Although "adaptive", such explanations are unsatisfactory because of their failure to specify mechanisms in testable form (e.g. Popper, 1957). In contemporary emotion research, Lazarus (1991) places adaptation at the heart of emotion processes" emotions map onto "core relational themes" describing the adaptational relationship between person and environment. Somewhat similarly, social-cognitive approaches to personality are concerned with the person's strivings to implement "personal projects" through interaction with the social environment (Cantor & Zirkel, 1990). Such theories are testable, by and large, but rarely computational.

16

Chapter 1

Contemporary research on adaptive models of PE is open to the criticism that it is poorly integrated with information-processing models. How can such an integration be effected in future research? The processing construct bridging architectural and knowledge levels of explanation is strategy. We can describe strategies in terms of processing constructs such as selection of processing codes, criterion-setting and so forth. Understanding of strategy use also requires understanding of why the person chooses one strategy over another; the motivational guidance of strategy choice. Thus, adaptation understood cognitively refers most straightforwardly to the acquisition, selection and implementation of computationally-specified strategies which aim to facilitate the person's goals within a given environment. The primary source of data is then experimental and simulation studies which allow computional models of strategy use to be developed.

Transient adaptation and strategy selection Explanations for experimental data require an understanding of how and why PE factors are related to strategy selection. For example, much recent research on distressing environmental stressors such as loud noise suggests that their effects on strategy are often more pronounced than effects on basic structural parameters of the processing system (Hockey, 1986). Noise appears to enhance use of the dominant strategy for performance, whereas fatigue is associated with a switch to low-effort strategies. Ecological theories of stress (Hancock & Warm, 1989) see behaviour in performance contexts as driven both by strivings to perform well and strivings to maintain a comfortable task load. Negative emotion and performance degradation are influence~ by the success or failure of the strategies which implement such motivations (see Kluger & DoNisi, 1996). Strategy choice under environmental stress reflects the subject's immediate motivations, beliefs about the personal significance of the stressor, and beliefs about the efficacy of strategy use in meeting salient goals. Beliefs vary dynamically, and perhaps even on a trial-to-trial basis, as the person modifies strategy in response to error feedback (cf. Rabbitt, 1979). These "state" variables are influenced by "trait" representations in LTM of the person's goals and general beliefs relevant to the particular situation (Matthows & Wells, 1996). For example, detrimental alter-effects of noise on performance may derive from reduced use of active coping strategies, resulting from appraisals of the stressor/task environment as uncontrollable and the limited relevance of the laboratory situation to personal goals (see Cohen, 1980). Individual

G. Matthews

17

differences in susceptibility to noise may reflect the individual's beliefs about the threat and controllability of noise stimuli (Jones, 1984).

Stabilities of adaptation There are different timescales for adaptation (Revelle, 1993). In addition to "single-occasion" instances of strategy-driven behaviour, there are stabilities of adaptation associated with PE evident over periods up to a single life time (see Mayer et al., this volume). The key question here is the nature of the representation which maintains stability, and there are several options. Emotion effects on performance may often be somewhat context-specific, and contingent upon context-bound appraisals and motivations (Matthews, Sparkes, & Bygrave, 1996). At the same time, data from widely diverse contexts suggests that emotions such as anxiety and depression may have some cognitive correlates which are intrinsic to the emotional state (Martin & Jones, 1995), or at least prototypical of the emotion. Oatley and JohnsonLaird's (1987) hypothesis that emotions signal the status of current action plans implies a degree of context-independence. Sadness indicates failure of a major plan (a description of adaptive status), which in turn constrains cognitions and action. As Lazarus (1991) states, sadness is associated with appraisals of irrevocable loss, and an action tendency for withdrawal from the environment, so that a given emotion entails a given representation of adaptive status. The basis of emotions in adaptive status forces at least some consistency in emotion-cognition relationships across individuals and occasions, despite the influence of contextual factors. Similarly, personality traits may reflect stabilities of adaptation. Matthews and Dora (1995) present an adaptive account of the diversity of independent information-processing functions associated with traits such as extraversion and neuroticism. They argue that personality traits represent fitnesses for adapting to certain kinds of environment, defined in terms of their informational properties. Cognitive correlates of traits provide the building blocks for acquisition of the skills necessary for success in the environments concerned. For example, extraverts are adapted to environments characterised by high information flows, including social environments (see Matthews, this volume). Correlates of extraversion such as high STM capacity, low response criterion and efficient dual-task performance facilitate the development of skills and strategies for handling rapidly-changing inputs. Viewed in terms of information-processing alone, extraversion is associated with an arbitrary collection of cognitive correlates. The link between

18

Chapter I

processing and the central characteristics of extraversion, such as impulsivity and sociability, requires the adaptive perspective. Representation of the adaptive potentials associated with traits is distributed over a number of distract processing characteristics.

Genetic basesfor adaptation The final element of adaptive explanations is the evolutionary perspective, operating over a time scale of many lifetimes. At the species level Darwin recognised that emotional responses may be understood at the species level in terms of their functional properties in aiding survival and reproduction. Evolutionary psychologists argue that emotions solve the regulatory problems posed by the need to coordinate multiple processing modules to handle imperative situations (Tooby & Cosmides, 1992). Instructions for building modules during development are represented within the genes. Modules may then be characterised both eomputationally and in terms of their design for solving adaptive problems. Perhaps a more contentious question is how individual differences in genotype are expressed as individual differences in module functioning. Despite the controversial nature of the research, there is now convincing evidence from structural modelling of behaviour genetic data to suggest that major personality traits such as negative emotionality are partially inherited (Loehlin, 1992), and the beginnings of a molecular genetics of personality are emerging (e.g. Leseh et al., 1996). The thinking of psychobiological researchers often seems unduly linear: the implicit model seems to be that the random outcomes of the genetic dice feed forward powerfully into personality, with perhaps a little modification by gene-environment interaction. This model leads naturally to the naive good genes/bad genes perspective previously criticised. It is hardly possible to estimate the selection pressures on the various traits. However, even traits which are socially devalued, such as neuroticism and psychoticism presumably have adaptive value in some circumstances, or the genes coding for them would have been selected out. Matthews and Dora (1995) argue that neuroticism is adaptive when the environment is characterised by disguised or subtle threats, especially social threats. Similarly, psychoticism may facilitate creativity (Eysenck, 1995), perhaps through attentional mechanisms such as those described by Beech and Williams (this volume). Thus, while genes may feed forward into the cognitive correlates of traits at the level of the individual, the cognitive

G. Matthews

19

components of traits represent feedback from the environment over many generations. If a person is to function as an extravert, by relying on social interaction to promote survival, for example, then those cognitive characteristics supporting social interaction skills will be selected for. This process m turn entails selection for the neural net parameters associated with the cognitive characteristics. The patterning of cognitive/neural functions associated with traits represents, in part, the toolkit of functions required for adapting to the environments associated with the trait (Matthews, 1997). Thus, natural selection links the adaptive and biological levels of explanation: individual differences in brain functioning support individual differences in choice of environment. We can reconceptualise the ladder of explanation as a loop, as shown m Figure 2, with connectionist networks, strategies and natural selection as the key constructs bridging the levels of explanation. Adaptation, in the broad sense, is not solely driven by natural selection, of course. Learned adaptations may be equally or more important, although it is uncertain how much learning influences basic parameters of neural net functioning. The present account emphasises the importance of skills rather than processing components in determining adaptation. Good STM for words does not necessarily assist a person to function as an extravert, but being able to remember ongoing conversations most likely does. Skills must be learnt, a process which reflects the interaction between the person's choice of strategies for acquiring knowledge (knowledge level) and the processing routines which implement learning (architectural level).

Adaptation (environmental fitness)

I

Strategies (performance and/earning)

Information processing

Knowledge

Natural selection

Architecture

Connectionism

~--

Biology

t Neuroscience

Figure 2. Levels of explanation reconccptualised as a loop.

20

Chapter 1

An Example: Explaining Anxiety and Cognition In discussing levels of explanation for PE phenomena, quite a lot of ground has been covered, and the scope for confusion and over-complexity in explanation will be evident. What the researcher must do, of course, is to select levels of explanation appropriate for the research problem at hand. In this section, I illustrate the application of the cognitive science approach to explaining relationships between anxiety and attention. The phenomena to be explained are well-known: impairment of attention, bias of selective attention to threat stimuli, and the relationship between abnormality in attentional function and clinical anxiety disorder (see Wells & Matthews, 1994, for a review). There are a variety of well-regarded information-processing models in this area (e.g. Bower, 1981; Ingram, 1984; Williams, Watts, MacLeod, & Mathews, 1988). Matthews and Wells (in press a) address the question of how we can go beyond the information-processing description of phenomena to explain associations between emotion and attentional functioning, and their implications for clinical disorder.

Anxiety and information-processing The first step is to decide what kind of explanation is sought. There is a psychobiology of anxiety-related bias but it has proved difficult to integrate with studies of selective attention in humans (Wells & Matthews, 1994, pp. 325-332). For example, Gray's (1982) SHS influences attention to threat stimuli (punishment cues), but anxiolytic drugs which act on the SHS fail to influence attcntional bias on the emotional Stroop test (Golombok, Stavrou, & Bonn, 1991). Thus, while acknowledging that biological (and evolutionary) factors may be important, the most straightforward approach is to focus on the architectural and knowledge levels. The next step is to characterise the performance correlates of anxiety in processing terms. The central architectural issue here is the extent to which anxiety influences strategic and/or automatic processing. The distinction between plan-driven strategic control and stimulus-driven "automatic" control of processing has been developed in considerable detail (Norman & Shallice; 1985). Anxiety might influence both the processing routines implementing strategic or executive control, and parameters of involuntary processing. Matthews and Wells (m press b) review the evidence on the automaticity of attentional bias, and conclude that bias is predominantly strategic. There is considerable evidence for context-sensitivity of bias (e.g. Calvo & Castillo,

G. Matthews

21

1997), even with subliminal stimuli (Fox, 1996). Similarly, deficits evident on tasks with neutral stimuli, demonstrated in test anxiety research (Samson et al., 1995), appear to be associated with loss of attentional resources or working memory (Eysenck, 1992), constructs associated with strategic rather than automatic processing. The clinical literature too tends to emphasise the strategies that anxiety patients develop for interpreting and coping with a world appraised as threatening (Beck, Emery, & Greenberg, 1985; Wells, 1995). Put differently, people with anxiety traits have developed "skills" for handling threat, which are sometimes maladaptive. One effect of state anxiety may be to bias retrieval of the processing routines controlling these skills. Matthews and Harley (1996) investigated the computational basis for attentional bias using a connectionist simulation of the emotional Stroop. The network was trained to discriminate colour and semantic inputs using the backpropagation algorithm. Bias towards negative emotion semantic content was introduced through various mechanisms, and the performance of the network compared with real data. The most satisfactory mechanism was a strategic one: low-level activation of a "threat-monitoring" task demand unit during colour-naming and word reading. In other words, strategic processes modulate the spread of activation from input to output units. "Automatic" mechanisms, such as sensitivity of input units to negative stimuli, and overlearning of response to negative stimuli, generated patterns of performance incompatible with real data. Siegle and Ingram (this volume) and Tryon (this volume) discuss alternative connectionist architectures for modelling phenomena relating to negative emotion. Investigation of the underlying architecture through experiment and simulation suggests that bias is more than just an "accidental" over-sensitivity of automatic threat-processing mechanisms. However, various explanatory questions are left open. It is conceivable that the primary consequences of anxiety are architectural, such as loss of resources, and anxiety effects on strategy are an attempt to "work around" these limitations. Alternatively, anxiety may not affect the architecture at all, but, instead, it influences personal goals and motivations which directly impinge on strategy choice and acquisition of threat-management skills. Questions also remain about the inter-relationship of the various performance correlates of anxiety, which are sufficiently diverse that multiple processing mechanisms are likely to be involved (Eysenck, 1992). Diversity in component processes may be associated with unity at the knowledge level (Matthews & Dom, 1995). For example, the various processing characteristics of anxiety may all subserve an overall orientation towards hypervigilance (Eysenck, 1992).

22

Chapter 1

A multi-level explanatory model

The Wells and Matthews (1994, 1996) Self-Referent Executive Function (S-REF) model of attention and negative emotion integrates architectural and knowledge levels of explanation within a multi-level framework. Three main components of the architecture are distinguished: stable self-knowledge encoded in LTM in procedural form, stimulus-driven automatic processing networks, and a supervisory executive. In response to various internal and external threat stimuli, the executive retrieves generic procedures for coping with threat from LTM, and tailors them on-line to the specific demands of the situation. As in most models of this kind (e.g. Norman & Shallice, 1985), routines under executive control influence behaviour indirectly, though biasing automatic processing. In the S-REF configuration, operation of the executive is characterised by self-focus of attention, cognitive interference generated by worry, and the pursuit of self-regulative goals, such as maintaining self-esteem. The S-REF model also emphasises the dynamic interplay of components: self-knowledge drives processing of threat stimuli, but is itself often modified by self-appraisal. Clinical disorder is generally associated with dynamic disturbances, such as perseverative cycles of rumination which fail to modify self-beliefs adaptively (of. Siegle & Ingram, this volume; Tryon, this volume). Within the model, architectural and knowledge levels of understanding are linked through coping strategies (Matthews & Wells, 1996). The knowledge level specifies the personal goals and beliefs about goal attainment which influence strategy choice. For example, generalised anxiety patients are motivated to protect themselves against various (often unrealistic) threats, and they hold the metaeognitive belief that worry is a successful strategy for so doing (Wells, 1995). The architectural level delineates the specific processing routines which implement coping. The S-REF model makes two general statements about processing in distress states, consistent with empirical evidence reviewed by Wells and Matthews (1994). First, processing activities associated with worry tend to interfere with both the internal operations of the executive system, such as formulating coping strategies, and with implementing and regulating the strategies themselves, if they are attentionally demanding. Second, although there is considerable variability in coping, distressed individuals often choose the task-focused strategy of monitoring for threats congruent with personal concerns. Threat monitoring (which is voluntarily initiated but not necessarily fully conscious) is responsible for emotional Stroop effects. It remains for future research to

G. Matthews

23

determine the specific processing routines involved: the Matthews and Harley (1996) simulations illustrate how this might be done computationally. Figure 3 summarises levels of explanation for inter-relationships between cognition and anxiety (and other negative emotions). The three classical levels of explanation provide alternative ways of describing anxiety phenomena. At the knowledge level, anxiety relates to self-knowledge and goals, as in Beck et al.'s (1985) schema theory. There may also be anxiety effects on processing specified at the architectural level (de-emphasised within the S-REF model). A full account must accommodate the neuroscience of anxiety, which is becoming increasingly integrated with architectural descriptions (see Derryberry & Reed, this volume; Kitayama, this volume). In this section, we have argued that deeper understanding is obtained through use of constructs which bridge the levels, especially strategies which control the use of the architecture to serve personal goals, and neural nets which describe processing phenomena using constructs broadly compatible with neurophysiology. To the extent that anxiety is genetically-influenced, we need also to consider how the neural basis of anxiety has developed through natural selection. Anxious individuals are sensitive to threat stimuli, but often they are conspicuously poor at handling the demands of threatening

Threat-driven self-regulation

Coping strategies - threat monitoring rumination

I ~-

Knowledge

Genetic adaptation to environments characterised by subtle threats

Attentional processes - r e s o u r c e loss

-

Architecture

- bias, etc.

Neural net parameters - e.g. activation of threat monitoring units

Biological

I Cortical and s u b c o r t i c a l circuits a c t i v a t e d b y t h r e a t stimuli

Figure 3. Lcvds of explanation for associations between anxiety and cognition.

24

Chapter 1

environments. Matthews and Dom (1995) argue that the processing correlates of anxiety serve the adaptive goal of maintaining vigilance for subtle or disguised threats (especially social threats), and neural correlates of anxiety may have evolved for this purpose. Conclusions

I have suggested that the multi-faceted emerging cognitive science of personality and emotion may be clarified by distinguishing informationprocessing models from explanations of the phenomena those models describe. Sperry (1993) has claimed that cognitive science introduces a new model of causal determinism, combining traditional microdeterminism with the top-down influences of emergent, macro mental state variables. Consistent with this view, reductionistic and holistic explanatory strategies may be distinguished in PE research. Reductionism requires a focus on the transient (state) or fixed (trait) differences in cognitive architecture which may be associated with emotion and personality factors. Architectural differences may in turn be traced to properties of neural circuits. For reductionism to be scientifically valid, the mappings between these different levels must be sufficiently simple that novel, testable predictions of behaviour may be derived from theory. Predictions include those derived from connectionist models, which may provide an important bridge between neural and architectural levels of explanation. The range of phenomena open to cognitive neuroscience explanation remains to be determined. The alternative, holistic approach to explanation seeks to explore the adaptive basis of emotion and personality, in the broad sense proposed by Lazarus. We require an understanding of how state and trait characteristics subserve the goals associated with emotions and personality. That is, the functional design of the processing system for implementing and acquiring contextualised skills may vary across individuals and across occasions. Cognitive science requires that adaptive explanations are linked to computational accounts of phenomena. Over short time-scales, the link may be achieved through specifying the strategies which allow goals to be met through implementing specific processing routines. Over longer time scales, there are several approaches to explaining stabilities of adaptation. First, representations of genetic strategies in LTM may drive consistency in computation. Second, representations of adaptive status may be intrinsic to emotional states. Third, personality traits may be associated with bundles of

G. Matthews

25

relatively stable, functionally independent computational characteristics which support successful adaptation to specified environments. Finally, both reduetionist and holistic explanations may feed into evolutionary explanations. To the extent that reductionism results in neural accounts of personality and emotion, evolutionary psychology may explain how the brain systems concerned have been shaped by the pressures of natural selection. In addition, the person's goals, and/or the strategies available for satisfying those goals, may be directly or indirectly related to genetically-programmed adaptive mechanisms, it is likely that the adaptive characteristics of personality and emotion reflect some complex interplay between social learning and genetics. However, as the example of anxiety research shows, it is wise to be selective in choosing levels of explanation. Different levels within the overall cognitive science framework are appropriate to different problems in personality and emotion research. References

Anderson, J. R. (1982). Acquisition of cognitive skill. Psychological Review, 89, 369-406. Anderson, J. R. (1990). The adaptive character of thought. Hillsdale, NJ: Erlbaum. Anderson, K. J. (1990). Arousal and the inverted-U hypothesis: A critique of Neiss's "Reconceptualizing Arousal". Psychological Bulletin, 107, 96100. Bandura, A. (1977). Social learning theory. Englewood Cliffs, NJ: PrenticeHall. Bargh, J. A., Chaiken, S., Govender, R., & Pratto, F. (1992). The generality of the automatic attitude activation effect. Journal of Personality and Social Psychology, 62, 893-912. Beck, A. T. (1967). Depression: Causes and treatment. Philadelphia: University of Pennsylvania Press. Beck, A. T., Emery, G., & Greenberg, R. L. (1985). Anxiety disorders and phobias: A cognitive perspective. New York: Basic Books. Bower, G. H. (1981). Mood and memory. American Psychologist, 36, 129148. Calvo, M. G., & Castillo, M. D. (1997). Mood-congruent bias in interpretation of ambiguity: Strategic processes and temporary activation. Quarterly Journal of Experimental Psychology, 50A, 163182.

26

Chapter I

Cantor, N., & Zirkel, S. (1990). Personality, cognition, and purposive behavior. In L. A. Porvin (Ed.), Handbook of personality: Theory and research. New York: Guilford. Cohen, J. D., & Sorvan-Sehreibor, D. (1992). Context, cortex and dopamine: A conne~ionist approach to behavior and biology in schizophrenia. Psychological Review, 99, 45-77. Cohen, S. (1980). After effects of stress on human performance and social behavior: A review of research and theory. ,Psychological Bulletin, 88, 82-108. Cosmides, L., & Tooby, J. (1992). Cognitive adaptations for social exchange. In J. H. Barkow, L. Cosmides & J. Tooby (Eds.), The adapted mind: Evolutionary psychology and the generation of culture. Oxford: Oxford University Press. Costa, P. T., Jr., & MeCrae, R. R. (1992). Four ways five factors are basic. Personality and IndivMual Differences, 13, 653-665. Ellis, A. (1962). Reason and emotion in psychotherapy. New York: Lyle Smart. Eysenck, H. J. (I 995). Creativity as a product of intelligence and personality. In D. H. Saklofske & M. Zeidner (Eds.), International handbook of personality and intelligence. New York: Plenum. Eysenck, H. J., & Eysenck, M. W. (1985). Personality and individual differences: A natural science approach. New York: Plenum. Eysenck, M. W. (1992). Anxiety: The cognitive perspective. Hillsdale, NJ: Erlbaum. Fox, E. (1996). Selective processing of threatening words in anxiety: The role of awareness. Cognition and Emotion, 10, 449-480. Gazzaniga, M. S. (1994). Nature's mind: The biological roots of thinlang, emotions, sexuality, language and intelligence. Harmondsworth: Penguin. Golombok, S., Stavrou, A., & Bonn, J. (1991). The effects of diazepam on anxiety-related cognition. Cognitive Therapy and Research, 15, 459467. Gray, J. A. (1982). The neuropsychology of anxiety: An enquiry into the functions of the septo-hippocampal system. Oxford: Oxford University Press. Hampson, S. E. (1988). The construction of personality (2nd ed.). London: Routledge. Hancock, P. A., & Warm, J. S. (1989), A dynamic model of stress and sustained attention. Human Factors, 31, 519-537.

G. Mattnews

zI

Hockey, G. R. J. (1984). Varieties of attentional state: The effects of the environment. In R. Parasuraman & D. R. Davies (Eds.), Varieties of attention. New York: Academic. Hockey, G. R. J. (1986). A state control theory of adaptation to stress and individual differences in stress management. In G. R. J. Hockey, A. W. K. Gaillard, & M. G. H. Coles (Eds.), Energetics and human information processing. Dordrecht: Martmus Nijhoff. Humphreys, M. S., & Revelle, W. (1984). Personality, motivation and performance: A theory of the relationship between individual differences and information processing. Psychological Review, 91, 153-184. Ingram, R. E. (1984). Toward an information-processing analysis of depression. Cognitive Therapy and Research, 8, 443-478. Jones, D. M. (1984). Individual and group differences in the response to noise. In D. M. Jones & A. J. Chapman (Eds.), Noise and society. New York: Wiley. Kanfer, R., & Ackerman, P. L. (1989). Motivation and cognitive abilities: An integrative/aptitude-treatment interaction approach to skill acquisition. Journal of Applied Psychology, 74, 657-690. Klein, S. B., & Loflus, J. (1988). The nature of self-referent encoding: The contributions of elaborative and organizational processes. Journal of Personality and Social Psychology, 55, 5-11. Kluger, A. N., & DeNisi, A. (1996). The effects of feedback interventions on performance: A historical review, a meta-analysis, and a preliminary feedback intervention theory. Psychological Bulletin, 119, 254-284. Lacey, J. I. (1967). Somatic response patterning and stress: Some revisions of activation theory. In M. H. Appleby & R. Tumbull (Eds.), Psychological stress. New York: Appleton-Century-Crofts. Lazarus, R. S. (1984). On the primacy of cognition. American Psychologist, 37, 1019-1024. Lazarus, R. S. (1991). Emotion and adaptation. Oxford: Oxford University Press. Lazarus, R. S., & Folkman, S. (1984). Stress, appraisal and coping. New York: Springer. LeDoux, J. E. (1995). Emotion: Clues from the brain. Annual Review of Psychology, 46, 209-235. Lesch, K. -P., Bengel, D., Heils, A., Sabol, S. Z., Greenberg, B. D., Petri, S., Benjamin, J., Miiller, C. R., Hamer, D. H. & Murphy, D. L. (1996). Association of anxiety-related traits with a polymorphism in the serotonin transporter gene regulatory region. Science, 274, 1527-1531.

28

Chapter 1

Loehlin, J. C. (1992). Genes and environment in personality development. Newbury Park, CA: Sage. Martin, M., & Jones, G. V. (1995). Integral bias in the cognitive processing of emotionally linked pictures. British Journal of Psychology, 86, 419436. Matthews, G. (1985). The effects of extraversion and arousal on intelligence test performance. British Journal of Psychology, 76, 479-493. Matthews, G. (1992). Extraversion. In A. P. Smith & D. M. Jones (Eds.), Handbook of human performance. Vol. 3: State and trait. London: Academic. Matthews, G. (1997). Intelligence, personality and information-processing: An adaptive perspective. In W. Tomic & J. Kingsma (Eds.), Advances in cognition and educational practice (Vol. 4), pp. 475-492. Greenwich, CT: JAI Press. Matthews, G., & Amelang, M. (1993). Extraversion, arousal theory and performance: A study of individual differences in the EEG. Personality and Indi~dual Differences, 14, 347-364. Matthews, G., & Deary, I. J. (in press). Personality traits. Cambridge: Cambridge University Press. Matthews, G., & Dorn, L. (1995). Cognitive and attentional processes in personality and intelligence. In D. H. Saklofske & M. Zeidner (Eds.), International handbook of personality and intelligence. New York: Plenum. Matthews, G., & Harley, T. A. (1993). Effects of extraversion and self-report arousal on semantic priming: A connectionist approach. Journal of Personality and Social Psychology, 65, 735-756. Matthews, G., & Harley, T. A. (1996). Connectionist models of emotional distress and attentional bias. Cognition and Emotion, 10, 561-600. Matthews, G., Sparkes, T. J., & Bygrave, H. M. (1996). Stress, attentional overload and simulated driving performance. Human Performance, 9, 77-101. Matthews, G., & Wells, A. (1996). Attentional processes, coping strategies and clinical intervention. In M. Zeidner & N. S. Endler (Eds.), Handbook of coping: Theory, research, applications. New York: Wiley. Matthews, G., & Wells, A. (in press a). The cognitive science of attention and emotion. In T. Dalgleish & M. Power (Eds.), Handbook of cognition and emotion. New York: Wiley.

G. Matthews

29

Matthews, G., & Wells, A. (in press b). Attention, automaticity and affective disorder. Behavior Modification. Neiss, R. (1988). Reconceptualizing arousal: Psychobiological states in motor performance. Psychological Bullean, 103, 345-366. Neisser, U. (1976). Cognition and reality. San Francisco: Freeman. Norman, D. A., & Shallice, T. (1985). Attention to action: Willed and automatic control of behaviour. In R. J. Davidson, G. E. Schwartz & D. Shapiro (Eds.), Consciousness and self-regulation: Advances in research (Vol. 4). New York: Plenum. Oatley, K., & Johnson-Laird, P. (1987). Towards a cognitive theory of emotions. Cognition and Emotion, 1, 29-50. Popper, K. (1957). The poverty of historicism. London: Routledge & Kegan Paul. Pickering, A. D., Diaz, A., & Gray, J. A. (1995). Personal@ and reinforcement: An exploration using a maze-learning task. Personality and Individual Differences, 18, 541-558. Pinker, S. (1994). The language instinct. Harmondsworth: Penguin. Pylyshyn, Z. W. (1984). Computation and cognition: Toward a foundation for cognitive science. Cambridge, MA: MIT Press. Rabbitt, P. M. A. (1979). Current paradigms and models in human information processing. In V. Hamilton & D. M. Warburton (Eds.), Human stress and cognition: An information processing approach. London: Wiley. Revelle, W. (1993). Individual differences in personality and motivation: "Non-cognitive" determinants of cognitive performance. In A. Baddeley & L. Weiskrantz (Eds.), Attention: Selection, awareness and control Oxford: Oxford University Press. Robbins, T. W. (1986). Psychopharmacological and neurobiological aspects of the energetics of information processing. In G. R. J. Hockey, A. W. K. Gaillard, & M. G. H. Coles (Eds.), Energetics and human information processing. Dordrecht: Martinus Nijhoff. Sanders, A. F. (1990). Issues and trends in the debate on discrete versus continuous processing of information. Acta Psychologica, 74, 123-167. Sarason, I. G., Sarason, B. R., & Pierce, G. R. (1995). Cognitive interference: At the inteUigenee-personality crossroads. In D. H. Saklofske, D. H., & M. Zeidner, M. (Eds.), International handbook of personality and intelligence. New York: Plenum. Schachter, S., & Singer, J. E. (1962). Cognitive, social, and physiological determinants of emotional state. Psychological Review, 69, 379-399.

30

Chapter I

Schwartz, S. (1975). Individual differences in cognition. Journal of Research in Personality, 9, 217-225. Shallice, T. (1988). From neuropsychology to mental structure. Cambridge: Cambridge University Press. Simon, H. A. (1967). Motivational and emotional controls of cognition. Psychological Review, 74, 29-39. Smolensky, P. (1988). On the proper treatment of eonnectionism. Behavioral and Brain Sciences, 11, 1-74. Sperry, R. W. (1993). The impact and promise of the cognitive revolution. American Psychologist, 48, 878-885. Spielberger, C. D. (1966). The effects of anxiety on complex learning and academic achievement. In C. D. Spielberger (Ed.), Anxiety and behavior. London: Academic Press. Tooby, J., & Cosmides, L. (1992). The psychological foundations of culture. In J. H. Barkow, L. Cosmides & J. Tooby (Eds.), The adapted mind: Evolutionary psychology and the generation of culture. Oxford: Oxford University Press. Watson, D., & Clark, L. A. (1992). On traits and temperament: General and specific factors of emotional experience and their relation to the fivefactor model. Journal of Personality, 60, 441-476. Wells, A. (1995). Meta-eognition and worry: A cognitive model of generalised anxiety disorder. Behavioural and Cognitive Psychotherapy, 23, 301-320. Wells, A., & Matthews, G. (1994). Attention and emotion: A clinical perspective. Hove: Edbaum. Wells, A., & Matthews, G. (1996). Modelling cognition in emotional disorder: The S-REF model. Behaviour Research and Therapy, 34, 881888. Williams, J. M. G., Watts, F. N., MacLeod, C., & Mathews, A. (1988). Cogniave psychology and emotional disorders. Chiehester: Wiley. Zuckerman, M. (1991). Psychobiology of personality. Cambridge: Cambridge University Press.

Cognitive Science Perspectives on Personality and Emotion -G. Matthews (Editor) 9 1997 Elsevier Science B.V. All fights reserved. CHAPTER 2

Conation, Affect, and Cognition in Personality John D. Mayer, Heather Frasier Chabot and Kevin M. Carlsmith

During much of the 20th Century, personality psychology has been a field divided into competing schools of psychodynamic, trait, humanistic, and other perspectives, with little communication among perspectives, and no common language. Recently, however, a consensus view of the field has been developing which considers personality from a systems perspective and attends to (a) the location of personality, (b) its parts, (c) its organization, and (d) its development (Mayer, 1993; 1995a, b, Pcrvin, 1980; Sears, 1960). For instance, pcrsonality's location is defined in relation to such neighboring systems as biology and sociology. Personality's parts include components that arc relatively basic such as hunger, happiness, and working memory, and more complex components as well, including extraversion, the self and the ego. Thousands of parts of personality have been proposed (Allport, 1958), and of these thousands, at least 400 parts are regularly discussed (Mayer, 1995b). Keeping 400 parts of personality in mind is a near impossibility, so one alternative strategy is to consider them in groups or classes (e.g., Barratt, 1985, Buss & Finn, 1987; Mayer, 1995a,b). Most classification systems for these components employ one or more of three categories of mind that have a centuries-old tradition: the conaave, affective, and cogmtive - what Hilgard (1980) has referred to as the trilogy of mind. According to this division, c~nation (or motivation) includes components that propel or move the organism such as the hunger drive, and the need for achievement. The affect group, principally containing emotion, includes such basic feelings as anger and happiness, along with related parts such as the mental programs for emotional facial expressmns. The cognition group, containing thought-related processes and mechanisms, includes such elements as worlang memory, judgment, and reasoning. The division of the mind into r affect, and cognition is so embedded in our discipline that many of our journals are named a~er those parts: Cogmtion, Motivation and Emotion, Cogmtion and Emotion, and so on. Despite this, many of us would be hard-pressed to recall the origin of this classification system, or to describe the differences among the three

32

Chapter 2

categories. Along these lines, Henle (cited in Hilgard, 1980, p. 115) remarked: as we become absorbed in our own specialties we often become cryptosystematists, that is, our beliefs are embedded in larger systems of thought that are not explicit but may serve to perpetuate errors. Indeed, the differences among motivation, affect, and cognition can become paper thin. A person's associations to the word "success" may reveal her need for achievement (conative), while also being influenced by her mood (affect), and memory (cognition). To accommodate such blended areas of performance, there exist blended areas of study such as "cognition and affect", and "motivation and emotion." Still, in what sense is one such class of mental process to be distinguished from the others? In this chapter we clarify the meaning of this tripartite division. We will begin by examining a general systems model of personality (already introduced at the outset). This model's further development relies in part on the distinction among classes of conative, affective, and cognitive components. The systems model illustrates how the three spheres of conation, affect, and cognition, can be used to classify aspects of personality psychology. The usefulness of the three spheres, however, relies on a clear understanding of each one's meaning. Following description of the systems model, we focus on conation, affect, and cognition, including (a) their historical origins, (b) their changing description across time, (c) their conceptualization, and (d) a recommended update of their meaning. Finally, we return to questions of conation, affect, and cognition in personality and in contemporary research, and discuss how the trilogy may be integrated into a picture of the person as a whole.

The Relational Model of Personality Several contemporary models of personality employ one or more classes

of conation, affect, and cognition in their construction (e.g., Barratt, 1985; Buss & Finn, 1987; Mayer, 1995a,b). Examination of one such model demonstrates one way the trilogy of mind is used today, and highlights some of the issues surrounding its use. The specific model employed here is the relational model of personality, so-called because personality and its parts are all described in relation to one another and their neighboring

e /

I

-

'

GROUPS INCLUDING OR INTERACTING WITH

PERSONALITY

INTERNAL PERSONALITY

NERVOUS SYSTEM

EXTERNAL SITUATION

SITUATIONAL ELEMENTS

[

Figure 1 . An view of the personality system amidst its neighboring systems, includmg biology, sociology, and situations. A molecular-molar dunension is represented vertically, an internal-external dimension horizontally, and an organismic dependent-constructed dunension depthwise.

,I.D. Mayer, H. Frasier Chabot and K.M. Carlsmith

L

I

33

34

Chapter 2

systems (Mayer, 1995a,b). This relational model is typically developed according to four systems-oriented topics: that is, according to personality's location, components, organization, and development. One aspect of the relational model that makes it particularly worth discussing is its highly integrative aspects; it contains or subsumes several models developed by others (e.g., Buss & Finn, 1987). Certain conceptual dimensions can be employed to distinguish personality from its neighboring fields of scientific study. The most important of these include a molecular-molar dimension, that distinguishes more molecular brain sciences which underlie personality from personality itself, and also distinguishes personality from more molar social structures that "contain" it such as the family and society. A second, internal-external dimension, distinguishes inside mental processes from outside observable behavior. To this, a third, organismic-constructed dimension can be added, which distinguishes between those parts of personality that are most constrained by the biological organism (i.e. basic motivations) from those that are most independent (i.e., formal reasoning). The use of three dimensions makes possible a three dimensional pictorial representation of personality and its component parts (see Figure 1). The purpose of this initial picture is to orient personality amidst its neighboring system in the three-dimensional space. Internal personality is contained in a box labelled "personality" on the left-hand side of the figure, mid-way between nervous system substrates beneath it, and family and social systems above it. In the picture, this vertical dimension represents the molecular-mOlar continuum in the sense that the lower brain sciences are more molecular than personality whereas the family and other social groups above personality are more molar. The second, horizontal dimension, represents the internal-external continuum with internal personality to the left, and personality's external manifestation (i.e., its interaction with the environment) to the right. Finally, the third, depth dimension, distinguishes more organismic parts of personality (to be added momentarily) in the foreground from more constructed parts (also to be added) in the background. The empty personality box can now be filled with classes of personality components in a manner that is consistent with each of the three dimensions. For example, in Figure 2, conation, affect, and cognition are placed along the floor of the cube, near the biological level, with a slight rise toward the back indicating the greater molarity of cognition relative to conation. This particular placement implies that conation, affect, and cognition refer to

J.D. Mayer, H. Frasier Chabot and K.M. Carlsmith

GROUPSlNCLUDlNG OR INTERACTING WITH PERSONALITY

Figure 2. A second view of the personality system including the enablers: Conation, affect, cognition, and consciousness (modified from Mayer, 1995a, Figure 2).

35

36

Chapter 2

internal, more molecular components of mind - that is, close to the biological level, or, only minimally influenced by learning. Notice that toward the innermost part of personality a fourth category has been added, consciousness. The placement of consciousness near conation, affect, and cognition suggests that consciousness, like them, is a more molecular, biological phenomenon, which may interact with the other three. Too little is known about consciousness to place it definitively anywhere, of course. One very respectable and influential tradition views consciousness as analogous to an imago in a hologram, in that it emerges from layered information within the cerebral cortex (Pribram, 1971, p. 171). This view would place consciousness at the ceiling of the personality box. But the rdational model puts it to the left bottom for reasons to be developed later. Within this relational model, conation, affect, cognition, and consciousness arc subgroups of a class of personality components containing them, termed enablers. Enablcrs are mechanisms that carry out, or enable, the basic functions of personality. The r arc one of four broad classifications that collectively contain all the parts of personality. The other three classes arc establishments, themes, and agencies. Establishments arc so-called because they are established (or leamod, or constructed) models of the self, the world, and the self in the world. Examples of establishments include the self-conc~t, self-esteem, attachment patterns, and expert knowledge. Establishments develop from experience and learning, and utilize the cnablcrs' functions to operate. For example, the self concept's self-love or self-hatred will be generated and intcrprcteA by emotional enablers, its self assessment will require cognitive cnablers. The connection between cnablcrs and establishments is often limited, however, to the fact that cnablers support establishments. At the establishment level, for example, expert knowledge can be fairly independent of a good or bad memory at the enabler level. That is, children may construct expert knowledge about dinosaurs independent of whether they possess an impoverished or superior memory. Thus, the establishment can be dcfine~ primarily according to its specific content. Establishment models arc illustrated in Figure 3, as the three floating cubes of internal personality. They are more molar than the enablcrs, and arc more independent of the organism as they proceed back toward models of the world. Note that all parts of personality arc vicwod as connected to all others; no arrows or connections are drawn in, however, as such a thicket of connections would obscure the rest of the depiction.

,I.D. Mayer, H. Frasier Chabot and K.M. Carlsmith

N E R V O U S S Y S T E M

37

Figure 3. A more complete view of the personality system now including all four major classes of personality components. The enablers (wnation, affect, cogmtion, and consciousness) are on the floor of the personality box. The establishments (models of the self, world, and self-in-world) are represented as boxes floating in the inside of the cube. The themes combine features of enablers and establishments; one theme, extroversion, is illustrated toward the back center of the Figure. Finally, agencies are larger supercomposites of individual components that collectively act as sub-personalities; one such agency, James' self-as-knower, is represented, as a cloud that intersects with the "Models of the Self' box (modified from Mayer, 1995a, Figure 2).

38

Chapter 2

The third class of components, the themes, represent thematic connections between establishments and enablers. Themes combine features from enablers and from establishments so as to form conceptually related mixtures that reveal themselves to observers in a coherent fashion. Whereas establishments are focussed on contents, themes are focussed on common or integrated features across enablers, across establishments, or across the two combined. Thus, a need for stimulation alone is an enabler; a model of "joining friends for a party," is an establishment. But the two can be viewed as thematically related. Thus, extroversion, according to Eysenck (1982), involves both a need for stimulation, (the eonative enabler), and establishment models of things such as how to throw a party. Extroversion is illustrated as elliptical features found in both eonation and in models of the world; these features are labelled ("extroversion features") to the right of the internal personality cube. The fourth class of components, the agencies, refer to large subdivisions of personality that carry out much of a personality's activities, but in partial independence of the whole; these include the id, ego, or superego. Another example of an agency is James' concept of the self-as-knower, which comes close to a self-conscious free spirit or free will. The self-as-knower is represented as a cloudlike column that runs through the Models of the Self. A more comprehensive discussion of the classification of personality components into enablers, establishments, themes, and agencies, and their twenty-one subcategories can be found elsewhere (Mayer, 1995a,b). Here, we are particularly interested in conation, affect, cognition, and consciousness, the subgroups of enablers. Enablers, as already noted, are viewed as close to the biological level in the relational model. For that reason, there must be plausible biological bases for the operation of these parts, and their division. Moreover, these parts form a larger class that describe mechanisms that carry out the functions of personality. Hence, the enablers must be divided and understood foremost according to what they enable, that is, what functions they perform. Because enablers are so basic, and perform basic functions of personality, almost all other parts of personality rely on them and are influenced by them. Better defining conation, affect, and cognition, and understanding the rationale underlying these concepts, can clarify understanding of personality as a whole.

J.D. Mayer, H. Frasier Chabot and K.M. Carlsmith

39

Understanding Conation, Affect, and Cognition Conation, affect, and cognition through recent history Hilgard's (1980) classic article The Trilogy of Mind... recounts the rise and fall of these three concepts from the early 1700's to early 1900's, and offers a rationale and recommendation for their resurrection. Surprisingly, Hilgard's work omits virtually any discussion of the meanings of conation, affect, or cognition, aside from their special status as a three-fold classification for the overall mind. Nonetheless, his article provides a basis for such an exploration by tracing the major figures who developed the trilogy over its history.

Faculty psychology and the trilogy of mind Hilgard (1980, p. 108) starts with the German faculty psychologists of the 18th century. He credits, in particular, Moses Mendelssohn's Letters on Sensation for bringing together the three concepts for the first time. Mendelssohn distinguished conation, affect, and cognition according to the fact that they operated differently from one another and that they might even interfere with one another. For example, when reason (cognition) "laboriously investigates the origin of pleasure," he wrote, "pleasure may be destroyed" (Mendelssohn, 1755/1971, p. 66) 1. There is both a phenomenological quality to this statement, indicating a sensitivity to the inner conscious experience of cognition and affect, and also a functional notion, identifying that cognition "investigates" pleasure. Mendelssohn also noted the independent behavior of the three components, writing that "convictions...belong in the realm of man's cognitive psychology," and that "by their very nature, [convictions] cannot be influenced by coercion or bribe" (Mendelssohn, 1983/1969, p. 44). On the other hand, will or motivation could be encouraged or discouraged by "reward and punishment" (Mendelssohn, 1983/1969, p. 44). Mendelssohn's approach is a partly functional one in the sense that he is specifying the conditions under which operations of the three spheres can be teased apart. The faculty psychology of late 18th century Germany gradually spread 1 Mendelssohn'swork is not yet translated in English. Hans G. Hirsch was kind enoughto translate fragments of the work which at least suggest some flavor of the original writings (see also Mayer, 1995b).

40

Chapter 2

to England and Scotland in the early 19th century. A number of psychologists contributed to classifying aspects of the mind during this period. For example, Thomas Reid, the great Scottish faculty psychologist, divided mental faculties into the intellectual (cognitive) and the active (motivational), dropping out emotion. By the late 19th century a summary of these British works was published in Alexander Bain's two-volume English textbook on psychology. Bain was fairly critical of attempts such as Reid's to reduce the trilogy to only two categories. He wrote that Reid's "submerged department of Emotion," could not be made to disappear but rather that its parts, such as emotions, feelings, and so on, "will be found partly taken in among the Intellectual Powers...and partly treated among the Active Powers," (Bain, 1855/1977, pp. 6-7), where they did not plainly fit. "Mind," wrote Bain (1855/1977, p. 1) at the outset of The Senses and the Intellect, ...possesses three attributes or capacities. I. It has Feeling, in which term I include what is commonly called Sensation and Emotion. II. It can Act according to Feeling. IIl. It can Think. Bain's trilogy, however, differs from the contemporary. For Bain, Feeling included sensation, whereas today's mental divisions typically group sensation with perception, outside the trilogy. Additional information concerning Bain's views on each member of the trilogy appear in the top portion of Table 1, which has three columns. Table 1 indicates the views of several central theorists, beginning with Bain. The three columns are divided so as to represent that theorist's view of conation, affect, and cognition. For example, in Table l's affect column, Bain says feeling and consciousness are "one and the same;" a statement which appears untenable today given contemporary research on unfelt, unexpressed, or unconscious emotions (e.g., Taylor, 1984). At the same time, Bain successfully develops a contemporary understanding of conation as he distinguishes between mental actions, which are part of the mental sphere, and those external actions that are not (Table 1, under "conation").

Chapter 2

41

Table 1. Historical and Contemporary Views of Conation, Affect, and Cognition: Direct Quotations and Brief Summaries from Key Figures. .

.

Conation .

.

.

.

.

.

.

.

.

.

.

.

.

Emotion .

.

.

.

.

.

.

Cognition

BAIN (1855/1977) "Action is...The putting forth of power to execute some work or perform some operation...in speaking of Action, however, as a characteristic of mind, we must render explicit the distinction between mental actions and such as are not mental...mental actions [are]... under the prompting and guidance of Feeling." (pp 2-3) "...There are in the human system movements and tendencies to movement 9..The eyes may open of themselves, the voice may break forth into utterance ...Yet those movements belong to the sphere of mind. The term Volition applies...to the entire range of mental or feelingprompted actions ." (p. 5)

"The three terms, Feeling, Emotion, and Consciousness, will, I think be found in reality to express one and the same fact or attribute of mind..." (p. 1) "...for a notion of what feeling is, I must refer each person to their own experience. The warmth felt in sunshine, the fragrance of flowers, the sweetness of honey..." (p. 2)

"...discriminating with preference, and the performance of intermediate actions to attain an end, are the most universal aspects of intelligence, inasmuch as they pervade the whole of the animal kingdom." (p. 6) "...the intellect [is]...a distinct endowment following laws of its own, being sometimes well developed and sometimes feeble without regard to the force or degree of the other two attributes." (p. 6) Intellect is distinct from emotion and volition because it allows for sensations and ideas to be relived without the stimulus (pp. 315-316) "Reason without affect would be impotent, affect without reason would be blind." (p. 112)

"In the evolutionary transition from reptiles to mammals, three cardinal behavioral developments were (1) nursing in

"The neocortex [can be described as]...ballooning out progressively in evolution and reaching its greatest proportions in the

MACLEAN (1990) "The protoreptilian formation is represented by a particular group of ganglionic structures located at the base of the

42

Chapter 2

Table 1 continued. forebrain in reptiles, birds, and mammals...these ganglia must be of 'enormous significance'for otherwise they would not be found as a constant feature in the vertebrate forebrain...[It is involved in] such basic behavior as the struggle for power, adherence to routine, 'imitation,' obeisance to precedent, and deception." (pp. 15-16)

conjunction with maternal care, (2)audiovocal communication for maintaining maternal-offspring contact, and (3) play...The limbic system plays a basic role in thymogenic functions reflected as emotional behavior...Two evolutionarily older subdivisions...have proved to be involved, respectively, in oral and genital functions...The third subdivision, for which there appears to be no

human brain...[it] has afforded a progressive capacity for problem solving, learning, and memory of details... linguistic translation and communication of subjective states..." (p. 17)

rudimentary counterpartin reptiles...[involves] parental care,audiovocal communication, and play behavior" (pp. 16-17) TOMKINS (1962) "In the human being the drive system plays a central role in... self-maintenance and reproduction." (p. 29) The system's primary function is to provide "motivating information" "information that drives and a drive that informs"specific to survival. (pp. 3031) It communicates "...where and when to do what- when the body does not know otherwise how to help itself." (p. 31 )

"The affective system [possesses]...numerous invariant instigators of any particular affect... [and] numerous invariant reducers of the same affect...It is this differentiated coupling and uncoupling characteristic which permits the affect system to assume a central position in the motivation of man." (p. 23) "Affects are sets of muscle and glandular responses located in the face and also widely distributed through the

[Not compared]

J.D. Mayer, H. Frasier Chabot and K.M. Carlsmith Table 1 continued. "The drive system with its relatively primitive signal and feedback mechanisms will work well enough [signalling internal changes] because of this predictable and small variability of the internal environment." (p. 124) "...a variety of materials must be regularly transported in and out of the body and thus drive signals wax and wane." (p. 125)

body, which generate sensory feedback which is either inherently 'acceptable' or kmacceptable'." (p. 243) Affects (associated with the reticular activating system, p. 90) such as interest, enjoyment, surprise, fear, shame, arise in response to learned or unlearned triggers (p. 22, p. 337). There is a partly invariant trigger-affect relation (p. 23). Affect is partially independent of the motivational system; it can mask motivation, or amplify the drive system so as to motivate the individual (p. 22). "This [affect] system is the primary provider of blueprints for cognition..." (p. 22) "There is here no essential rhythm as there is with respect to the drive system." ([- 125)

PLUTCHIK (1984) Aroused by changing internal states of the organism" (p. 214) "Aroused by the absence of homeostatically significant stimuli" (p. 214) "There are specific 'natural' objects toward which motives direct the organism (e.g., food, water)" (p. 214)

"Aroused by external stimuli" (p. 214) "Aroused by the presence of a survival-related event" (p. 214) "There are few 'natural' objects in the environment toward which emotions are automatically directed" (p. 214)

[Not compared]

43

44

Chapter 2

Table 1 continued. "Induced before the process of search is begun" (p. 214) "Tend to have a rhytlunic character" (p. 214)

"Induced aRer an object is seen or evaluated" (p. 214) "Depend on events in environment which may occur on a random basis" (p. 214)

IZARD (1993) "Drives such as hunger, thirst, sex...are cyclical in nature." (p. 72) "[Drives are] dependent upon peripheral physiological processes" (e.g., stomach growling; p. 73) "Drives provide specific information regarding the time and place that something needs to be done..." (p. 73) Drives, "cue a relatively specific set of responses..." (p. 73)

An emotion has no temporal cycle (p. 73) "...an emotion...is not dependent on peripheral physiological processes" (e.g., stomach growling) (p. 73) "...can be associated with a virtually limitless variety of phenomena" (p. 73) Emotions "can motivate an equally wide range of cognitions and actions" (p. 73) "the emotions system preceded the cognitive system in evolution and outpaces it in ontogeny" (p. 73)

"Clearly, information processing consists of several types or levels... ranging from that which leads to the color of an eye to that which produces a Mona Lisa or a theory of relativity" (p. 73) "I propose four differentiable sorts of information processing: cellular, organismic, biopsychological, and cognitive...the first three of the forgoing categories involve types of noncognitive information processing" (p. 70) Cognition is about knowledge- learning, memory, symbol manipulation, thinking, and language (p. 73) Emotion-cognition interactions occur in all the many coping activities that require stimulus appraisal and judgment before action (p. 73)

J.D. Mayer, H. Frasier Chabot and K.M. Carlsmith

45

Hilgard (1980, pp. 113-114) concludes his survey of the trilogy of mind shortly after his discussion of Bain, with the psychologists of the 1920's and 1930's. He comments: Those in America who were proposing a new experimental or laboratory psychology rejected faculty psychology and along with it the classification of mental activity into three categories.., with [the American psychologist] McDougall the history of the trilogy of mind appears to have ended, nearly two centuries after it began in Germany and Scotland. In part, the fading of such a "generally accepted" view may have coincided with the decline of a felt need for such a comprehensive classification of mental processes. To be sure, Hilgard (1980, p. 113) wrote, "the trilogy of mind was still familiar in the vocabulary of psychology," but psychologists of the time were more interested in experimental advances than in the classification systems of the past. We believe that Hilgard's own interest in the trilogy suggests that its history was - and is - not over, although it may no longer occupy so central a place in the field. For that reason we proceed to more recent developments.

MacLean and the influence of psychiatry on the trilogy of mind By the mid-20th century enough had been learned about the brain structure and function that some initial statements could be made regarding its relation to mental faculties. Of course, this had been attempted earlier. Phrenologists had attempted to connect mental faculties such as learning or feeling to specific brain areas, for the purpose of charting personality according to a shape of an individuars cranium. Thus, someone with a cranial indentation alongside the presumed brain-site for imagination would be regarded as having a stodgy, uncreative personality. But phrenology was based on pure speculation, and as a consequence, was discredited. Brain localization became a reality, however, with the identification of some language abilities in Broca's area. And it was shortly thereafter, with the writings of Paul MacLean (e.g., 1949, 1973, 1990), that the trilogy of mind found a possible home in brain science. MacLean inferred from the structure of the human brain the existence of three partially independent subbrains, or brain divisions, which reflected three distract epochs in the human brain's evolutionary development. The first such brain, which was structurally innermost, was shared in all its essentials with the complete brain

46

Chapter 2

of reptiles. The second brain, which corresponded to the limbic system, was shared in common with most mammals. The third brain, which corresponded to the cerebral cortex, was most highly developed in humans. MacLean (1990, p. 9) wrote: In popular terms the three evolutionary formations might be imagined as three interconnected biological computers, with each having its own special intelligence, its own subjectivity, its own sense of time and space, and its own memory, motor, and other functions. Although MacLean never emphasized the point, parallels exist between conation and the reptilian brain, affect and the old-mammalian brain, and cognition and the neo-mammalian brain. For example, the reptilian brain had associated with it, "such genetically constituted forms of behaviour as selecting homesites, establishing territory, engaging in various types of display, hunting, homing, mating, bree~ing, imprinting, forming social hierarchies, and selecting leaders." (MacLean, 1973, pp. 9-10; 1990; see also Table 1). The old mammalian brain, "plays an important role in elaborating emotional feelings that guide behaviour with respect to the two basic life principles of self-preservation and the preservation of the species..." (MacLean, 1973, pp. 12-13). The third, neomammalian brain, is concerned with higher cognitive processes. MacLean suggests a number of innovative comparisons among the three brains. He notes that "the limbic system might be imagined as particularly designed to amplify or lower the intensity of feelings involved in guiding behavior required for self-preservation and preservation of the species." (1991, p. 17). He further notes that the different brains vary as to their external orientation, with the neomammalian (cognitive) brain most external in that it receives its information through signals conducted from the eyes, ears, and somatic receptors (MacLean, 1991, p. 19). MacLean's writings were influential in the 1950's and it is not surprising that they turned up, shortly thereafter, in psychological writings more explicitly identified with the mental trilogy.

Modern psychologists and the trilogy of mind Silvan Tomkins, an evolutionary emotions psychologist, focussed on the function of psychological processes and may have been influenced by

J.D. Mayer, H. Frasier Chabot and K.M. Carlsmith

47

MaeLcan's writings. Recall that MacLean saw the limbic system, which was largely emotional, as amplifying survival-relatod feelings; Tomkins raised this idea again, arguing that the emotion system's role was to amplify motivation. Recall also that MacLcan described the nco-mammalian brain as more closely connected to the outside world than were the palco-mammalian or reptilian brains. Tomkins was perhaps influenced by this comparison when he notexi that the emotion system was directed toward the outside world whereas the conativc system was directed to the internal world. Finally, Tomkins shared with MacLcan and others of the time the use of an informationprocessing metaphor, describing r for example, as providing "readouts" of the organism's internal states. For Tomldns, conation has evolutionary significance in that it "plays a central role in...self maintenance and reproduction" (Tomkins, 1962, p. 29) as well as an information-processing aspect in which "primitive signal and feedback mechanisms" provide a readout of the internal homeostatic rhythms of the organism (Tomkins, 1962, p. 124). Tomkins went on to earcfuUy detail some of the characteristics that distinguished the conativc system from the affectivc. For example, Tomkins noted that "internal states" trigger conation, and that conation is typically rhythmic. In contrast, "external stimuli" trigger emotion, and emotion follows no particular set timclinc. These ideas have become generally accepted. For example, Robert Phtchik's (1980) side-byside comparisons of conation and affect included those and other distinctions that had been outlined by Tomkins. Plutchik's comparisons can also be found in Table 1. Tomkins and Plutchik both distinguish conation from emotion, with less attention paid to cognition (the cognition columns of Table l arc essentially empty for these theorists). The conation-affcct distinction was likely viewed as requiring more theoretical attention because motivation and emotion are so inextricably intertwined in behavior. There is something so different between conation and affect, on the one hand, and cognition, on the other, that the difference was often unattended to (Bain, 1855, p. 6, made this same point). Nonetheless, there arc some difficulties involved in distinguishing conation and affect from cognition. A central problem is caused by the frequent use of an information-processing metaphor to describe both the functions of conation and affect. If both conation and affect arc processing information, what is unique about cognition? Tomkins' former student, Cal Izard, recently addressed this problem by distinguishing between non-cognitive and cognitive information processing. Non-cognitive information processing inchdes that accomplished by genetic

48

Chapter 2

codes, chemical reactions, and "reflective instinctive, and biologically prepared or genetically disposed behavior" (Izard, 1993, p. 70). Cognitive processing, in contrast, "involves more general and flexible processes that operate on experience based learning and memory. Cognitive activities involves judgment, planning, problem-solving and understanding." Trends in thinking on the trilogy across time Considerable shifts in meaning of the tfilogy's categories have taken place, even from Alexander Bain's writings in the late 19th century to the present. This progression reflects (to us) a cumulative understanding of the utility of the trilogy, and of the differences among the tripartite areas. Several trends appear to best describe this progression: a trend toward identifying the trilogy as taking place exclusively internal to personality, a trend toward localizing each member of the trilogy in one or more brain areas, a trend toward an information-processing metaphor to describe them, and a reformulation of each class so as to create a more meaningful trilogy. The trend toward distinguishing the internal from the external. There has been a more or less constant recognition that conation, affect, and cognition are internal mental events, i.e., associated with brain function rather than with external events. Mendelssohn's comments that pleasure and pain change a person's will but not their cognition suggests that cognition is something intrinsically private, hidden and autonomous (Mendelssohn, 1755/1971, p. 66). A century later, Alexander Bain struggled to define will's internal location. Bain (1855, p. 2) referred to will as conative action that required the "putting forth of power to execute some work." Bain (1855, pp. 2-3) noted that, "In speaking of Action...as a characteristic of mind, we must render explicit the distinction between mental actions and such as are not mental." Bain's clarification that action was "a characteristic of mind," and therefore internal, was probably necessitated by his description of mental action as "putting forth power," which could readily be mis-understood as taking place externally. This metaphorical difficulty evaporated with MacLean's switch to the use of information processing metaphors for brain function, which suggested an internal computer. The trend toward brain localization and informaaon processing. Consistent with the internalization of these three processes was the attempt to find serious associations between the three classes and brain function. Although a non-scientific beginning to this pursuit originated with the phrenologists, serious connections awaited the works of MacLean, in

J D. Mayer, H. Frasier Chabot and K.M. Carlsmith

49

biological psychiatry, and Tomkins, in psychology. Although MacLean's work focussed on brain localization, Tomkins' work provided an interesting supplemental conception by extending localization to the larger nervous system. For example, "affects" were "sets of muscle and glandular responses closely associated with the brain's reticular activating system" (Tomkins, 1962, p. 243). Along with the increased focus on the brain and nervous system was the aforementioned shift in metaphor from industrial machines to an information processing paradigm. Bain's view of conative action as the "putting forth of power to execute some work" seems embedded in his own era of mechanical engines, whereas Tomkins' (1962, p. 124) view that conation provides "signal and feedback mechanisms" of internal organismic information, seems embedded in an era of computers. Although the information processing metaphor is today dominant it is still possible that multiple metaphors can best describe the phenomenon, just as in physics, light is both described as a wave and a particle (Bohr, 1963). For example, conation seems best described by combining Bain's and Tomkins' descriptions, so that conation is said to provide "a primitive readout of the internal, more or less homeostatic rhythm of the organism", and generates "power to execute some work." The trend toward finding more homogeneous categories at a common level offunction. There has also been an important narrowing of the trilogy's members such that each category is individually more circumscribed, and so that they operate collectively at a common level of function. For example, Bain's category of affect originally included the three concepts of feeling, consciousness and sensation, whereas contemporary views have essentially restricted the category to emotions and closely related feeling states such as calmness and arousal. This narrowing of focus represented a growing recognition that consciousness, sensation, and affect are incommensurate processes that perform different functions, are localized separately, and therefore are best treated separately. In today's Introductory Psychology books, sensation has been paired off with perception, and consciousness is treated, if at all, in its own chapters. The remaining affect category retains only emotion and closely related feelings. This narrowed version of the affect category seems more parallel to the similarly narrowed categories of conation and cognition. A similar and no less important transition occurred for conation, which originally referred to will, but with the transition from Mendelssohn to Tomkins has come to refer to more-or-less basic, unlearned motivations. The conation category now includes only basic motivations, which are, once

50

Chapter 2

again, both more homogeneous, and easier to compare to the similarly revised category of affect. The concepts of "will" and "consciousness", although excluded from the trilogy, were not plainly grouped with other parts of the mind. "Will" is perhaps covered in personality in discussing self-control and selfmanagement. Consciousness, however, could perhaps form a fourth category added to the trilogy of conation, affect, and cognition - a possibility we examine shortly. The trend toward emphasizing unlearned or innate qualiaes. As the categories of eonation, affect, and cognition have been more narrowly focussed, the focus has been directed toward their unlearned or innate qualities. The effort to distinguish these three mental categories has almost always best suece,exted when descriptions of them focus on their developmentally early, unlearned states. Thus, to say that motivations are "rhythmic"~ whereas emotions are not, is to emphasize such motivations as hunger, thirst, and sex, rather than more learned, less rhythmic motivations such as a desire for education or achievement. Similarly, to focus on the fact that emotions are triggered by external events is to emphasize their basic nature rather than more complex, learned emotions that might be triggered by reminiscence. This lower level, more mechanical conception was yet another reason to homogenize the categories and dispense with those parts, such as consciousness and will, that did not fit well. What remains in each category is a set of mechanisms, or basic functions of personality. Recall that it was their basic mechanical qualities that led to the label of enablers for conation, affect, cognition, because they help personality get the job done. The reason this emphasis on innate, or minimally learned qualities of the enablers is so important, is that as learning increases, more complex structures are created that are less plainly divisible into the three categories. For, as the enablers engage together in more complex functions it is clear that they become inexorably combined and intertwined. There exist a relatively few pure psychological enablers: pure conative urges for food and water, or pure affective joy or sadness, and pure memory networks. Soon atter these enablers begin work, they construct a much larger set of established thoughts that combine them. For example, a person develops models of the self, or a self concept, that includes conation (what I want), affect (what I feel about myself), and cognition (what I know about myself). But the general selfconcept, which includes all three, by necessity integrates the enablers. It was sensitivity to this point that led McDougall (1923, p. 266) to say that the trilogy work cooperatively rather than individually:

J.D. Mayer, H. Frasier Chabot and K.M. Carlsmith

51

We often speak of an intellectual or cognitive activity; or of an act of willing or of resolving, choosing, striving, purposing; or again of a state of feeling. But it is generally admitted that all mental activity has these three aspects, cognitive, conative, and affective; and when we apply one of these three adjectives to any phase of mental process, we mean merely that the aspect named is the most prominent of the three at that moment. Each cycle of activity has this triple aspect; though each tends to pass through these phases in which cognition, conation, and affection are in turn most prominent; as when the naturalist, catching sight of a specimen, recognizes it, captures it, and gloats over its capture. The trend toward more limited inclusiveness. Through the time of Bain, some claim was made that the trilogy encompassed all mental function. With the increasingly focussed meaning of the three classes of mentation, it became easier to eject some concepts outside the trilogy. As has already been noted, sensation and perception were paired outside the trilogy. Similarly, will and consciousness were moved outside. The trilogy is no longer a trilogy of the entire mind, perhaps, but remains a critical trilogy operative within the more molecular, basic aspects of personality - and remains of considerable research importance. Caveat emptor This particular reading of the history of the trilogy of mind is, of course, our own, and alternatives are possible. The relational model of personality was constructed in part according to this reading of the evolution of the categories and employs those categories according to their outline here; alternative models are possible. Still, the relational model has very evident strengths in relation to classification models that have been developed before (see Mayer, 1995b), and it is worth, therefore, further considering how the trilogy of mind can be developed within it. Clarifying the trilogy m an expanded quaternity of mind

Although conation, affect, and motivation have been narrowed and clarified across time, many of the original distinctions among them still apply, even more clearly. The above discussion, atter all, has distinguished the three realms in several important ways. Phenomenological distinctions focus on different conscious experiences of the trilogy - that conation, affect, and motivation all "feel" differently from one another. Structural brain distinctions focus on differences in brain localization of the trilogy.

52

Chapter 2

Functional distinctions focus on the different actions of the three systems, and so on. These distinctions, as well as a number of others, can be summarized across theorists in a new, enlarged format. To create this summary, we chose the clearest statements from the Table 1, edited them, and supplemented them where necessary, in Table 2. Although Table 2 was constructed on the basis of the above discussion of the trilogy of mind, the table denotes a quatcmity - consciousness has been added. Some comment is necessary on this. As noted, Bain joined consciousness to feeling, but consciousness nowadays is just as likely to be joined to cognition (e.g., Bower, 1981), or denoted as a blackboard to represent all three (e.g., Bower & Cohen, 1982). In fact, consciousness is implicated whenever any of the three systems reach a high enough level of activation. For these reasons, it seems useful to separate consciousness from any single one of the other three and provide it with a place of its own. Because one interpretation of consciousness is that it is basic and elemental, a place among the enablers seems one possibility. Such a classification is useful from a systemic viewpoint because, just as the conativc-cnabler class includes urges, instincts, and mental energy, so a conscious-enabler class could include such components as the stream of consciousness, the phenomenal field, and so on. This provides a strong classificatory rationale, if nothing else, for provisionally converting the trilogy into a quatcmity, with the addition of consciousness.

The Quaternity of Mind and Personality Dynamics If the discussion until now seems removed from contemporary concerns that is one of the problems frequently encountered with discussions of classification. Contemporary research is concerned with dynamics - causal or mutual influences among different parts of personality. Another difference between the classification thus far and contemporary research is the sheer generality of the discussion. So far, we have talked of all affect as if it were a single entity, when in fact, it is divisible into many parts. The contemporary researcher, in contrast, typically is interested in more specific personality parts and their dynamics. So, whereas up-to-now we have discussed the interaction between affect and cognition, the researcher might be more interested in the influence of happiness on memory. Discussion at the global level has indisputable value, however, because it can make clear the conceptual background within which more specific research is conducted.

J.D. Mayer, H. Frasier Chabot and K.M. Carlsmith

53

T a b l e 2. Conation, Affect, Cognition, and Consciousness Compared. Characteristic

Conation

Affect

Cognition

Consciousness

FUNCTION

To direct the organism to carry out basic acts so as to satisfy survival and reproductive needs

To organize a limited number of basic responses quickly, adaptively, and in an organized fashion; to link those responses to complex situational environments

To learn from the environment and to problem solve so as to assist with motives and emotions

To assign mental activity where needed; to intervene flexibly in conation, affect, or cognition, where new responses are called for

CONSCIOUS MANIFESTATIONS

If conscious, specific urges, e.g., toe.at, to drink

If conscious, the pleasure and pain of objects and stimuli; also, specific emotions such as happiness, fear, anger, etc.

Conscious and unconscious parts; conscious examination of problem

Direct consciousness itself; also reflective awareness of existence

AGENCY

Involuntary

Partly involuntary; partly voluntary

Mostly voluntary Partly voluntary; partly involuntary

DEVELOPMENTAL ONSET

Basic urges present immediately, including hunger, thirst, comfort..

Two or more basic emotions (e.g., pleasure, pain) present immediately; later development includes more complex emotions

Concrete reasoning early on, later the ability to reason with abstract information

Unknown; selfawareness from 18 months; continuous conscious identity from around age 3 with the end of infantile amnesia

Chapter 2

54 Table 2 continued.

i,

Characteristic

Conation

Affect

Cognition

Consciousness

INrITATION OF Predominantly RESPONSE responsive to internal bodily states

Predominantly responsive to external environment

Responsive either to internal or external environment

Responsive to non-habituated, i.e., novel, or unusually intense, internal or external events

TEMPORAL CHARACTERISTICS

Motivations precede action; rise and fall rhythmically or cyclically

Emotions often respond to events; they possess no set timeline

Occurs any time; Alternates no set timeline according to the sleep-wake cycle.

INFORMATIONAL SPECIFICITY

Specific as to what is lacking and what must be done

Identifies a class of possible events that must be addressed, without necessarily being specific

Either specific or general depending upon problem requirements, work accomplished, and mental capacity

Can incorporate and become aware of a wide variety of information; is very plastic in how it interprets information and proceeds

BRAIN LOCALIZ-

The limbic system is a subcortical structure, near the center of the cerebral hemispheres. It encircles the top of the brainstem. It is commonly divided into three tracts, or circuits, composed of different

Emotion is commonly associated with the limbic system, particularly with the amygdala, and secondarily with the hypothalamus. There is also recent evidence that the frontal cortex of the left hemisphere may

Information processing can be distinguished from higher level cognition. Although the entire brain processes information, we reserve cognition to encompass flexible processing based on learning and memory; this

May be located in the reticular activating system, or may be an emergent property of the mind as a whole

ATIONS

J.D. Mayer, H. Frasier Chabot and K.M. Carlsmith

55

Table 2 continued. Characteristic

Conation

Affect

Cognition

structures. One mechanism of importance involves the hypothalamus which controls hormones that target various parts of the body and may regulate drives, e.g., of hunger and sex (Reeve, 1992).

specialize in processing positive emotion, the right hemisphere in negative,

includes judgment, planning, problem solving, and understanding. These are commonly viewed as dependent upon the association cortex and the cerebral cortex.

DESCRIPTION Unmotivated OF QUANTITY Motivated

Unemotional Emotional

Unthinking Thinking

Unconscious Conscious

SOCIALLY Constructive vs. DESIRED AIMS Destructive Motivations

Pleasant vs. Unpleasant Emotions

Intelligent vs. Unintelligent Thinking

Spiritually conscious vs. self-conscious

OPEN VERSUS Accepting vs. CLOSED/INAC- Repressed CESSIBLE

In Contact vs. Out of Contact with Feelings.

Flexible vs. Rigid

Receptive versus Unreceptive

JOINT MOLECULARMOLAR DEVELOPMENTAL CONTINUUM

*Basic emotions; e.g. happiness, anger, fear **Complex emotions, e.g., shame, guilt, mixed emotions ***Sentiments (emotions attached to objects) e.g., loving one's country,

*Basic cognition: sensory motor operations, learning **Middle cognition: concrete operations, symbol learning ***Complex cognitions: formal operations, abstract thought.

*Basic consciousness **Reflective consciousness ***Higher consciousness (e.g., reflective, spiritual, etc.).

*Basic urges, e.g., hunger, thirst, physical contact; **Learned motivations: e.g., pleasing others, achievement ***Functionally autonomous motives, e.g., doing a good job, helping others,

Consciousness

56

Chapter 2

For example, such a general discussion can provide hints as to where the more important enabler-to-enabler interactions will take place. Treating conation, affect, cognition, and consciousness as equals would suggest there exist 6, i.e., (4 • 3)/2, equivalently important sets of interactions to cover. An interesting alternative view, however, suggests that the central interactive areas among the classical trilogy will be more limited. Recall MacLean's triune brain that emerges in stages from conation to affect to cognition. If we assume adjoining areas (in terms of brain localization) have more interactions, greater interactions should occur between the adjoining areas of conation and affect, and affect and cognition, than between conation and cognition. This seems borne out by (our admittedly subjective impression of) today's research literature, which focusses on the former two interactions. Limitations of time and space have encouraged us to focus on the central conative-affective, and affective-cognitive interactions. The interactions between consciousness and the trilogy will be considered briefly at the end.

Conation and affect To recap, conative phenomena concern include hunger, thirst, and reproduction. Conative functions chart homeostasis in the body and alert the organism about needs for survival and reproduction. Thus, hunger tells us we should eat; thirst tells us we should drink, and so forth. In contrast, affect is concerned with such feeling states as happiness, joy, and alertness. Its primary concern is to provide us with signals about our relations with external individuals and objects. Thus, happiness tells us we are in harmony with others, and anger that we are treated unjustly. It is plain that conation and affect must serve the same master to some extent (e.g., overall personality). Thus, basic-level motivations provide constraints on emotions that ensure survival. Say you agree to eat your bagged lunch with someone late in the day. Then, during a walk in the woods you become hungry and think of the bagged lunch you brought along. You are likely to feel frustrateA, but you won't eat immediately because you know it will make you feel guilty later. Should the motivation to cat become stronger, however, most people will cat, so as to promote their energy and clear-headedness - their likelihood for survival. In the above instance, motivation (conation) and emotion work together, assessing different necessities, and balancing one against another. In that example, whether motivation or emotion "wins" is a matter of which signal (i.e., hunger or guilt) is the strongest. Often, however, more sophisticated

J.D. Mayer, H. Frasier Chabot and K.M. Carlsmith

57

interactions may take place. For example, the emotion system (which is the more flexible) may "filter" motivations by allowing expression of those that are adaptive in a given situation, and by (at least temporarily) disallowing or suppressing those needs that are inappropriate. For example, if one is hungry, and there are people around who are eating, but none offer food, the original sense of hunger may be replaced by a feeling of injustice. An angry injustice might be a motivator for requesting food even though the act could be viewed as impolite or even improper (making a request might be suppressed by guilt). Say that, in this instance, the anger does replace the original hunger motivation and redirects the individual to ask for food. This is in part what Tomkms (1962, p. 22) meant when he wrote that "Affect...can mask motivation, or amplify the drive system so as to motivate the individual." Similarly, Oatley and Johnson-Laird (1987) view emotions as coordinating motivational urges and plans. Finally, motivation and emotion may contribute to one another more directly. Say you become happy because you have accomplished an important goal. You may need companionship as a consequence, and the motivational system may provide urges - phenomenological bursts of energy - to assist you to pursue social companionship. As another example, you may suddenly become sad; motivationally you may need to return to your own territory, or as the present idiom has it, you "need space." Helpful or harmful though this motivational accompaniment may be that moment, it is hard to change its directional quality. Research on the interaction between motivation and affect often reflects explorations in physiological, non-verbal communication, and evolutionary psychology. A review of such literature can be found in the chapter, "Motivation and Emotion," in Mook's (1996) textbook, Motivation. Because this area has been reviewed so recently, and because a large portion of it lies outside our own areas of expertise, we will move ahead to the relation between affect and cognition.

Affect and cogmtion We have already recapped the affect system, focussing on its depiction of relationships between oneself and the external world. The cognitive system, on the other hand, is useful for more flexible understandings of the world and the events in it. One of affeet's most important contributions to cognition is to prioritize it (Mandler, 1984). Thus, when working on a project, a fear of something going on at home, although distracting at first,

58

Chapter 2

may turn one's attention to what is, ultimately, a higher priority to one's survival. Not only do affects interrupt cognitions, but they can also change them in ways that may promote better judgment and creativity. One of the major influences of affect on cognition is through that of the mood-congruent cognition effect. Modified slightly from Mayer, Gaschke, Braverman, & Evans (1992, p. 129), the mood-congruent cognition effect: ...states that people's cognitions are sensitive to the correspondence between the pleasant-unpleasant quality of their mood and the pleasant-unpleasant connotations of their ideas. An affective match between a person's moods and ideas increases both the memorability and the judged merit, broadly defined, of those ideas. For example, mood-congruent concepts will be more readily learned and recalled. In addition, mood-congruent ideas will be judged richer in their associations, mood-congruent attributes will be judged as more applicable, mood-congruent examples of categories will be judged as more typical, and mood-congruent causes and outcomes will be judged more plausible. It is possible to read into this effect another way mood facilitates cognition: As a person's moods shift, the shift will force changes in a person's perspective on the surrounding world. Changing perspectives, in turn, allows for creative thinking about a problem, and the construction of a greater number of alternative courses of action. Such mood shifts drag the cognitive system along with them, forcing alterations in thinking and motivating changes in perception, and potentially enhancing planning and creativity (see Mayer, 1986, or discussion in Mayer, McCormick, & Strong, 1995). At a still broader level, cognitions seem to keep affects tolerable. That is, much thinking involves doing something for the emotion system, and consequently, for the motives those feelings relate to. This is what Tomkins (1962, p. 22) meant when he wrote that, "...this [affect] system is the primary provider of blueprints for cognition..." It is also at least loosely related to Freud's notion that the ego derives its energy from the id. The more one's emotions are satisfied, the less directive they are and the more chance the cognitive system has to operate well according to its own rules of logic, propositions, and formalism. Although cognition follows the blueprint of affect, it can also turn around and change affect where affect (or motivation) seems

d.D. Mayer, H. Frasier Chabot and K.M. Carlsmith

59

counterproductive. For example, cognitions can help manage affects when they get out of hand, and separate good or useful affects, from misleading ones. So-called meta-or reflective experiences of mood (e.g., "This mood is clear to me," "This feeling is unacceptable," etc.) involve cognitive attempts to evaluate and regulate moods so as to improve their responsiveness beyond a simple reflexive attempt at survival (e.g., Mayer & Gaschke, 1988; Mayer & Stevens, 1994; Salovey et al., 1995). The recently developed concept of emotional intelligence (e.g., Mayer & Geher, 1996; Mayer & Salovey, in press; 1993; Salovey & Mayer, 1990) is basically a compendium of the areas in which emotion facilitates thought, and thought improves emotion. One recent definition of emotional intelligence (Mayer & Salovey, in press) describes it as including four broad classes of abilities: ...the ability to perceive accurately, appraise, and express emotion: the ability to access and/or generate feelings when they facilitate thought; the ability to understand emotion and emotional knowledge; and the ability to regulate emotions to promote emotional and intellectual growth. The role of consciousness

It is hard to resist a mention of consciousness' function at this point. To us, consciousness plays a role similar to that of a family bulletin board upon which messages are placed (cf., Bower & Cohen, 1982, pp. 309-310). The consciousness "bulletin board," more specifically, receives messages from conation, affect, and motivation: urges, such as "need water," emotions, such as "anxiety", and thoughts, such as "l should talk more at my upcoming meeting to appear more assertive." Just as in a family, each member has different handwriting, so too, conation, affect, and cognition, have their own individually recognizable modalities, their signature phenomenology. An integrated personality recognizes messages from each source because it experiences each differently, and evaluates each system on its own terms, much as one evaluates messages from family members on the basis of their recognizable styles. That is, an adult personality uses consciousness to recognize that an urge is an urge, and as such, has a different status than a logical proposition. Ideally, it weighs the urge ("l am increasingly hungry") with the thought ("This project would best be finished before l eat") and wisely chooses which to follow depending on circumstances.

60

Chapter 2 Conclusions and Other Considerations

The four cnablers of conation, affect, cognition, and consciousness represent only the lower level portions of personality. Emerging from them arc establishments, including models of the self, the world, and the self-inthe-world, and themes, coherent collections of features drawn from cnablcrs and establishments that arc expressed as behavioral traits. Conation, affect, and cognition work closely together to support these more complex structures. For example, research on cognition and affect as they extend into a person's models of the self and world (i.e., establishments) are being conducted by Fiskc and her colleagues on affect-triggered schemata (Fiskc, 1982); by Higgins and his colleagues on self-schema and affect (Higgins, 1987), and by Petty and his colleagues on attitudes (e.g., Pricstcr & Petty, 1996). Summary. Researchers in the area of cognition and affect are, by virtue of their interdisciplinary interest, unusually broad in the problems they pursue. Successful research across affect and cognition may be facilitated by better understanding the scope of affect and cognition, the distinctions between them, and their relationship to personality. To better understand cognition and affect, their original grouping: conaaon, affect, and cognition the so-called trilogy of mind - was examined in considerable detail. We provided a historical review of the trilogy of min~! and attempted to discover some trends in their evolving meaning. The dofufitions of conation, affect, cognition, were refined and updated. An alteration of the trilogy to a quaternity was recommended so as to include consciousness. This quatemity/trilogy was located within one possible contemporary model of personality, the relational model. Finally, the relevance of the quatemity and the interactions among its members were briefly applied to a discussion of some contemporary research in cognition and affect. References

Allport, G. W. (1958). What units shall we employ7 In G. Lindzey (Ed.), Assessment of Human Motives (pp. 239-260). New York: Rinehart & Company, Inc. Bain, A. (1855/1977). The senses and the intellect. London: John W. Parker & Son. [Roprintexi in D. N. Robinson (Ed.), Significant contributions to the history of psychology: 1750-1920 [Series A: Orientations; Vol. 4]. Washington, DC: University Publications of America.

J.D. Mayer, H. Frasier Chabot and K.M. Carlsmith

61

Barratt, E. S. (1985). Impulsiveness defined within a systems model of personality. In C. D. Spiclbcrgcr & J. N. Butcher (Eds.), Advances in personality assessment (Vol. 5, pp. 113-132). HiUsdalc, NJ: Lawrence Erlbaum. Bohr, N. (1963). Essays, 1958-1962, on atomic physics and human knowledge. New York: Wiley. Bower, G. H. (1981). Mood and memory. American Psychologist, 36, 129148. Bower, G. H., & Cohen, P. R. (1982). Emotional influences in memory and thinking: Data and theory. In M. S. Clark & S. T. Fiskc (Eds.), Affect and cognition. Hillsdalc, NJ: Lawrence Erlbaum. Buss, A. H., & Finn, S. E. (1987). Classification of personality traits. Journal of Personality and Social Psychology, 52, 432-444. Clark, M. S., & Fiskc, S. T. (1982). Affect and cognition: The seventeenth annual Carnegie Symposium on cognition. Hillsdalc, NJ: Lawrence Erlbaum. Eyscnck, H. J. (1982). Personality, genetics, and behavior. New York: Pracgcr. Fiskc, S. T. (1982). Schema-triggered affect: Applications to social perception. In M. S. Clark & S. T. Fiskc (Eds.), Affect and cognition. Hillsdalc, NJ: Lawrence Erlbaum. Higgins, E. T. (1987). Self-discrepancy: A theory relating self and affect. Psychological Review, 94, 319-340. Hilgard, E. R. (1980). The trilogy of mind: Cognition, affection, and conation. Journal of the History of the Behavioral Sciences, 16, 107117. Izard, C. E. (1993). Four systems for emotion activation: Cognitive and noncognitivc processes. Psychological Review, 100, 68-90. MacLcan, P. D. (1949). Psychosomatic disease and the 'visceral brain'. Recent developments bearing on the Papcz theory of emotion. Psychosomatic Medicine, 11, 338-353. MacLean, P. D. (1973). A triune concept of the brain and behaviour. Toronto: University of Toronto Press. MacLcan, P. D. (1990). The triune brain m evolution: Role in paleocerebralfunctions. New York: Plenum Press. Mandlcr, G. (1984). Mind and body: Psychology of emotion and stress. New York: W. W. Norton & Co. Mayer, J. D. (1993). A system-topics framework for the study of personality. Imagination, Cognition, and Personality, 13, 99-123.

62

Chapter 2

Mayer, J. D. (1995a). The system-topics framework and the structural arrangement of systems within and around personality. Journal of Personality, 63, 459-493. Mayer, J. D. (1995b). A framework for the classification of personality components. Journal of Personality, 63, 819-877. Mayer, J. D., & Gaschke, Y. N. (1988). The experience and recta-experience of mood. Journal of Personality and Social Psychology, .55, 102-111. Mayer, J. D., & Geher, G. (1996). Emotional intelligence and the identification of emotion. Intelligence, 22, 89-113. Mayer, J. D., McCormick, L. J., & Strong, S. E. (1995). Mood-congruent recall and natural mood: New evidence. Personality and Social Psychology Bulletin, 21,736-746. Mayer, J. D., & Salovey, P. (1993). The intelligence of emotional intelligence. Intelligence, 17, 433-442. Mayer, J. D., & Salovey, P. (in press). What is emotional intelligence? In P. Salovey & D. Sluyter (Eds.), Emotional development and emotional intelligence: Implications for educators. New York: Basic Books. Mayer, J. D., & Stevens, A. (1994). An emerging understanding of the reflective (meta-) experience of mood. Journal of Research in Personality, 28, 351-373. Mendelssohn, M (1971). Moses Mendelssohn: Gesammelte Schrifien Jubilaumsausgabe (Band 1: Schriflen zur Philosophie und Astheak). Stuttgart: Friedrieh Frommann Verlag (Gunther Holzboog). (Original work published 1755). Mendelssohn, M. (1969). Jerusalem (A. Jospe, Trans. & Ed.). New York: Schocken. (Original work published 1783). Mook, D. G. (1996). Motivation: The organization of action (2nd ed.). New York: W. W. Norton. Oatley, K., & Johnson-Laird, P. N. (1987). Towards a cognitive theory of emotion. Cogniaon and Emoaon, 1, 29-50. Pervin, L. A. (1990). A brief history of modem personality theory. In L. A. Pervin (Ed.), Handbook of personality theory and research (pp. 3-8). New York: Guilford. Plutchik, R. (1984). Emotions: A general psychoevolutionary theory. In K. R. Scherer & P. Ekman (Eds.), Approaches to emotion. Hillsdale, NJ: Lawrence Erlbaum. Priester, J. R., & Petty, R. E. (1996). Gradual threshold model of ambivalence: Relating the positive and negative bases of attitudes to subjective ambivalence. Journal of Personality and Social Psychology,

J.D Mayer, H. Frasier Chabot and K.M Carlsmith

63

71, 431-449. Pribram, K. H. (1971). Languages of the brain: Experimental paradoxes and principles in neuropsychology. Englewood Cliffs, NJ: Prentice Hall. Reeve, J. (1992). Understanding motivation and emotion. Fort Worth, TX: Harcourt, Brace, Jovanovich. Salovey, P., Mayer, J. D., Goldman, S., Turvey, C, & Palfai, T. (1995). Emotional attention, clarity, and repair: Exploring emotional intelligence using the Trait Meta-Mood Scale. In J. W. Pennebaker (Ed.), Emotion, disclosure, and health (pp. 125-154). Washington, DC: American Psychological Association. Salovey, P. & Mayer, J. D. (1990). Emotional intelligence. Imagination, Cognition, and Personality, 9, 185-211. Sears, R. R. (1950). Personality. Annual Review of Psychology, 1, 105-118. Taylor, G. J. (1984). Alexithymia: Concept, measurement, and implications for treatment. American Journal of Psychiatry, 141,725-732. Tomkins, S. S. (1962). Affect, imagery, consciousness. Vol. 1: The positive affects. New York: Springer. Author Notes

Paul Presson was instrumental in developing the graphics for the relational model of personality; his patience during design sessions enabled us to develop a far clearer picture than we would have otherwise, and we are grateful for his assistance.

This Page Intentionally Left Blank

Cognitive Science Perspectives on Personality and Emotion - G. Matthews (Editor) 9 1997 Elsevier Science B.V. All rights reserved. CHAPTER 3

Introduction to the Bidirectional Associative Memory Model: Implications for Psychopathology, Treatment, and Research Warren W. Tryon

Learning and memory are arguably the two most fundamental psychological processes. Without learning, infants would not acquire the skills that make them children and adults. Without memory, cumulative learning could not occur; we would continuously relearn everything. All connectionistic neural networks (CNNs) both learn and remember; they entail a learning and memory mechanism. It is therefore impossible to discuss learning in the absence of memory or memory in the absence of learning. An important advantage of CNNs is that they are also compatible with biological and genetic explanations. The possibility that the synaptic network comes preset at birth with sensitivities to, and biases for, processing information in certain ways was addressed by Seligman (1970) and Seligrnan and Hager (1972) in terms of biological preparedness. It is also possible that not all aspects of the CNN are equally modifiable by experience. It may be that certain networks function essentially unchanged throughout the subject's lifetime. These possibilities do not detract from the fact that many organisms, especially humans, learn a great deal during their lifetime and that some of what is learned plays an important role in developmental changes. Personality is heavily dependent upon memory. Persons with Alzheimer's Disease provide empirical support for this assertion. Their personalities gradually dissolve as they forget their life experiences including where they have been, what they have done, and who their children and parents are or were. Psychopathology and psychotherapy are also highly dependent upon memory. A phobic person is afraid only because they have anxious memories about certain stimuli. If the anxious memories of a car phobic can be replaced with memories of positive experiences, then the person will no longer fear automobiles. Other feelings not generated by immediate environmental stimuli are also memories. This includes feelings of depression, insecurity, and low self-image. Lotius (1980, p. xiv) described a hypothetical future memory doctor as being able to cure psychological disorders by modifying the memories giving rise to the associated feelings. Schafer's (1978) hermeneutic psychotherapy seeks relief in just such a way; by recalling and altering

66

Chapter 3

memory for past events through reinterpreting them so that they are experienced more positively and in an integrated fashion. Wachtel's (1977) description of psychoanalysis includes recapturing disassociated memories and diffusing their emotional impact through catharsis. Psychoanalysis can be accurately summarized as a theory of conscious and unconscious memory formation and recall especially under stressful conditions. The Bidirectional Associative Memory (BAM), one variety of CNN, enables psychologists to address many of the same topics as psychoanalysts but with much more testable models since CNNs, including the BAM, can be implemented on a computer and are therefore fully open to analysis and experimentation. An added benefit of CNNs is their fundamental compatibility with neuroscience and biological psychiatry. Contemporary behavior therapy is dominated by cognitive and cognitivebehavioral models. Reference to emotion or affect is conspicuously absent; Ellis excepted (1962, 1980). Hollon and Beck's (1994) description of cognitive and cognitive-behavioral therapies discusses thinking, beliefs, and interpretations as important elements but does not include emotionalmotivational variables. Blatt and Bers (1993, p. 165) observe that "The role of affect is not only ignored in most cognitive behavioral considerations of self-schemas, but it is often considered an impediment to the assessment of them. Rather than viewing the self-schema as a cognitive-affective structure, research from a cognitive-behavioral orientation often attempts to eliminate or control current mood as possibly confounding the assessment of schemas". The authors subsequently noted that cognitive-behavioral theorists are generally reluctant to explore motivational, affective, and developmental issues. Cognitive and information processing models of normal and abnormal behavior stress intellectual control. Contemporary behavior therapies for children and adults emphasize corrective thinking for emotional as well as behavioral disorders. Put otherwise, psychologists have over intellectualized emotional disorders. Any comprehensive explanation of normal and abnormal behavior must address emotion as well as cognition and behavior. The main purpose of the purpose of this chapter is to augment interest in modeling mechanisms underlying normal and pathological phenomena using connectionistic neural networks by applying one particular CNN, the BAM, to several areas of interest. The fact that CNNs in general and the BAM in particular are new to many psychologists means that little empirical work has been conducted to date. Hence, this chapter cannot review and evaluate the BAM in terms of quantitative empirical data. The scope of this chapter is

W. W. Tryon

67

therefore restricted to discussing the heuristic value of the BAM for understanding a wide range of phenomena related to cognition, emotion, and psychopathology. This chapter extends the BAM so that it learns emotions of varying intensities in specific contexts thereby forming affective memories. It is important to note that the same learning mechanism used to form intellectual memories is also capable of forming affective memories. In order to fully understand this approach to "hot cognition", we must review how the BAM stores and recalls memories.

Bidirectional Associative Memory (BAM) The BAM model was selected for the following masons. First, it is designed to form and recall memories. Second, because the BAM is equally able to associate among stimuli, emotions, and behaviors, it is applicable to the full spectrum of psychological and behavior disorder. Third, the concepts of memory well and basin of attraction associated with the BAM provide new ways to conceptualize psychopathology and treatment; both psychological and biological. Fourth, the BAM is a relatively simple system and consequently is a good point of departure. The BAM entails symmetric interconnections that the brain does not have and is therefore less biologically plausible than some other neural networks. However, the BAM is not intended to be an exact brain copy of an actual brain structure but rather to simulate memory formation and recall using selected brain functions such as parallel distributed processing and local processing at each node. The present discussion derives mainly from Kosko (1987a, 1987b, 1988) and Wasserman (1989). The Appendix provides details regarding how the BAM works. Because it is not entirely necessary to understand every detail of how the BAM functions to appreciate its heuristic value in understanding psychopathology, a succinct overview of the most important elements is given next. The stimuli and responses that the BAM learns to associate are represented as vectors, a sequence of numbers, of l's and O's defining the presence or absence of a set of characteristics. The attributes coded for can be cognitive, affective, and/or behavioral which makes the BAM a highly general model of memory formation. Any level of detail can be modeled. At a very low level of abstraction, vector entries can represent the state of individual sensory neurons and motor fibers. At a high level of abstraction, vector entries can represent the results of other neural networks dedicated to

Chapter 3

68

recognizing perceptual features (red hair), affeetive states (see below), and/or behavioral dispositions (response vs. no response or flight vs. fight). If the elements of one vector (A) index rows and the dements of the second vector (B) index columns, the resulting square matrix (outer product) constitutes a memory matrix (M) for the AB association. For example, if vector A = 1, -3, 5, 7 and vector B = 2, 4, -6, 8, then memory matrix M is defined as follows:

2 Vector A

1

2

-3 5 7

-6 10 14

.

.

Vector B 4 -6 4 -6 -12 18 20 -3O 28 -42 .,

8 ....... 8 -24 40 56 ,,,,

The numerical values of the matrix dements simulate functional synaptic properties of excitation (positive values) and inhibition (negative values). Multiple memories, up to a computable limit, can be accurately encoded into a single memory matrix by summing corresponding cells over all individual memory matrices. Memory recall is accomplished by multiplying a stimulus vector by the composite memory matrix. If the result of multiplying vector A times memory matrix M is not exactly vector B (correct recall), then the obtained result is fed back through the memory matrix by multiplying the obtained result by the transpose of M. The result of this calculation is used as a modified stimulus and therefore multiplied by M, as was vector A. The result will either be vector B or something closer to it This active reverberating and reconstructive process of memory recall, continues until vector B is fully recalled or no further improvement can be obtained in which case the memory recalled is, as with people, the best approximation that can be generated. This process enables pattern completion where a whole memory can often be reconstructed from a partial stimulus. Neural networks are good at Gestalt psychology. This pattern completion property will be emphasized in our discussions of psychopathology. Because of parallels with physics, an "energy" value can be calculated for each memory. This calculation provides the two dimensional memory matrix with a third dimension; height in this case, that enables one to visualize memory formation as the creation of memory wells in an otherwise

W. W. Tryon

69

fiat surface. This is because the state of minimum energy is the state of best fit between the AB vector pair. Memory recall occurs whenever this energy state recurs. Imagine a fiat rubber sheet upon which a ball bearing has been placed causing a vertical indentation. Since the ball bearing comes to rest at a point below the surface, it is associated with a negative, and therefore minimum, energy state (see Figure 1). That memory formation is associated with a minimum energy state can be understood as similar to how "best fit" occurs when the deviation of data points about a regression line is minimized. Both are measures of fit.

I Iglll IBm

60

LU

X

ILl

4}1t

2il

IOlD IBm ....J" / ....

". . . .

direction of movement

/

/.__~ ~

Motor Area

--~"

Associative motor area

Figure 10. Goal backpropagation: the model is used to learn transitions between stable states such as to go from A to B (AB) and next from B to C (BC)... on CA1 neurons. CA3 neurons are used to merge the recognition of the current state coming directly from the entorhinal cortex EC2 and delayed information about the previous state that we suppose comes from the Dentate Gyrus and more precisely from the Granular Cells (the GC developing a time spectrum expression of the information coming from EC2). The CA3 representation of the state transition is then learned on CA1 neurons and copied on prefrontal cortical neurons which learn the "anti-causal" (backpropagated) links between pairs of transition states. These prefrontal neurons also learn associative links with motivation nodes if they are activated simultaneously with them. Assuming the activation of a motivation coming from the limbic system (the will to eat for instance), the model explains the propagation of the goal to subsequent subgoals on the prefrontal cortex. When a subgoal transition can be achieved because the robot state corresponds to the first part of the transition, the associated learned action is the trigger and allows the robot to reach a new state which is the starting point for triggering the recognition of a new transition. When several transitions are possible from a current state (like BC or BD) the level of activation of the prefrontal neurons associated with the different possible transitions allows the choice of taking the shorter or the most interesting pathway from the current state to the final goal.

174

Chapter 4

1989). These sharp waves present some of the characteristics, in particular in amplitude and frequency, of the stimuli capable of inducing LTP. They could therefore participate in the reactivation of cortical patterns necessary for the LTM storage at cortical level. The linkage between successive events constituting transitional states would result in a playback not of single events but of sub-sequences of events.

From hippocampus to prefrontal cortex This memory of states and transitions forming a sequence of events is, in the model, ~rther integrated at CA1 level, with a possible partial restitution there of the "cortical" topology thanks to the direct pathways from entorhinal cortex third layer to CA1. From CA1 there are direct as well as indirect connections to prefrontal cortex. At this prefrontal site the long-term registration of a sequence can be considerd as a result of the successive activation of nodes which can be assimilated to cortical columns. By the very orderly nature of a sequence encoded by a spatial pattern of activation, one could assume that the best neurophysiological support for such a storage would be a unidirectional facilitation of the synaptic weights of a specific path in a network architecture, as it has been implemented in Bapi and Levine (1994). It is plausible that such an oriented unidirectional facilitation takes place in primary or secondary cortical areas. Nevertheless, the prefrontal cortex is the most plausible site for the linkage between sensory and motor sequences at least at a high level of controlled processes (Figure 11). This does not preclude the possibility of sensory-motor links at subcortical or even lower levels as schematized by Figure 11. At cortical level, the execution of a sensory-motor sequence is necessarily linked, at least implicitly, to the completion of a goal selected by motivation. A goal in the model corresponds in fact to a secondary goal, i.e. a situation which allows the satisfaction of a basic drive or of a sublimation of this basic drive. At the executive controlled level of the prefrontal cortex there is clear, even if only subjective, evidence that the goal is usually present and therefore activated at the very onset of the sequence. Thus it can influence the choices of subgoals and the hierarchical unfolding of specific endeavors to reach them. The most parsimonious implementation of this psychophysical reality requires the instantiation of a bidirectional facilitation of the different pathways leading from the starting point to the goal of a sensory-motor sequence. In this way, the activation of a goal induces a retropropagation of activity, similar to a priming by top-down activation from the categorial

J P. Banquet et al.

,2

Posterior association cortices

175

Premotor cortex

e~

// cL

Primary motor cortex

Primary sensory cortex

,// Perception

Action

I I I

External environment I I I

Figure 11. Flowchart of multilevel information transfers: 1) between sensory and motor areas (horizontal connections) 2) between sensory (or motor) and associative areas of increasing complexity (vertical connections) 3) between internal drives and planification capabilities (prefrontal cortex).

nodes in an ART architecture (Grossberg, 1976a, b). Yet, here the priming process concerns an entire sequence of events, and accordingly, is implemented according to a gradient. This subliminal backward priming of a sequence in conjunction with a bottom-up activation from subcortical structures such as the hippocampus helps the selection of the best sequence of actions to reach a specific goal. This process of goal retropropagation is not solely efficient for selecting the optimal way for goal attainment. It also operates in deciding the order of goal satisfaction, when several goals are

176

Chapter 4

simultaneously active or in competition. As such, it implements aspects of hierarchy setting performed by Pc. Similar disambiguation between several possible sequences takes place at Hs level during the playback of the sequence if the prediction of the future event is based on several previous events instead of just the immediate preceding one. When such a neural network is used for the control of a mobile robot, different types of behavior in order to select and reach a goal are exhibited by the robot. These choices are dependent on three parameters of the system: the relative strength of the different drives, the weights of the connections implicated in the different paths, and the required match level between perceived and memorized steps towards the goal. Any node of the cognitive map leams transitions between pairs of learned places. The level of activation of these nodes results from the addition of bottom-up (match-related) and topdown (drive and path length-related) activations. These combined activations of the nodes can lead to a variety of behaviors of the system which have counterparts in real life. If the top-down influences are too weak, the system is unable to follow a specific path for the attainment of a specific goal. It is susceptible to distraction by any new input previously associated with a different behavior. Conversely, if the saliency of the top-down input is too strong, the recognition of a situation could be biased in the direction of a situation corresponding to the satisfaction of its goal. The initiation of such erroneous recognitions can be self-reinforcing. All these situations correspond to pathologies of frontal lobe.

Fronto-Hippocampal Function and Personality Personality in humans presents three key "primitives": 1) A temporal function which seems to obey some principle of symmetry of past and future, memory and prospective, with respect to the present. This temporal memory function is mostly based on the capacity to evaluate and record the order of occurrence of event sequences (a kind of segmentation lost in frontal patients), and eomplementarily the capacity to recognize new from familiar events (a kind of fusion of events lost in hippocampal patients). We have seen the importance of the Hs in the recording of "one exposure" events. The consequence of the suppression of this memory function is illustrated by anterograde amnesia, i.e. the incapacity starting at a period of life, corresponding to some time prior to the lesion, to build up a continuing history, just as if the factual life of a person had stopped at this moment. Yet, the Pc is also involved in this historical function, as one of the favoured

J.P. Banquet et al.

177

cortical sites for the permanent recording of these event memories. Symmetrically, the prospective function supposes the capacity to project in the future an orderly sequence of planned events in order to either actually perform them or merely simulate them. This planning capacity is also an hallmark of personality. This function is as important as the previous one, and in fact intimately linked to it. Our capacity to make plans, i.e. to project our actions in the future, is narrowly dependent on a library of past behavioral schemes and of their consequences. As our personal history goes back in time as far as our early childhood, our ability to project our life in the future concerns more or less remote time. The range of this prospective capability is closely linked to the strength and integration of our personality, and supports our motivation. Mostly from neuropsyehological studies, the role of the Pc in this function is well documented. Pc is essential not only for the strict and logical ordering of events or actions. Furthermore, it operates in the determination of an hierarchy of subgoals and actions to reach a predetermined goal. The incapacity to forecast the consequences of actions could be responsible, along with the neutralisation of affective life, for the incoherent and self-destructive behavior eventually encountered in prefrontal patients. 2) Working memory can still be considered as a part of the temporal function. Nevertheless its unifying role, and its implication in practically every other function related to personality deserves a separate account. The historical and prospective function, in particular, could not exist without the support of an "extended present", i.e. a working memory. The capacity to link successive, logically related events oriented towards the performance of a task, or the accomplishment of a goal, is essential to the development of personality. WM is not present in early childhood and this absence explains the non-permanence of hidden objects in the field of consciousness as internal representation, and therefore the incapacity to perform delayed tasks. This capacity progressively develops during infancy, and probably supports the unfolding of logical reasoning. This logical function is a prototypical illustration of the characteristic of WM defined as both maintenance and manipulation of information over an extended period of time. Classically, Pc is endowed with WM capacity, but so also are temporal and plausibly other cortical or subcortical structures. Our contention is that Hs also partakes of an automatic aspect of WM, even though delay neurons have not been recorded there, as in different cortices. We have proposed at least three subsidiary mechanisms that could support this function at Hs level.

178

Chapter 4

3) Emotions and motivations arc another important facet of personality. One could hardly contend that a robot is a person even though by its previous experience it can have a semblance of history. Emotions could be viewed as resulting from sublimation, but certainly not a suppression, of basic drives including the instinct for survival and even the instinct of death. Motivation could result from an integration and a trade-off between the need for drive satisfaction, emotions, and social constraints. Motivation is usually dependent on the degree of satisfaction of these needs. Hero also limbic system and Pc act conjointly. Amygdala is as important for emotional lifo as hippocampus proper is important for correlational and WM functions. Similarly, orbitofrontal and meAial Pc are essential for the integration of drives, emotions and motivations while more cognitive information processed in dorso-lateral Pc. The suppression of any type of affr162 colour, positive or negative, in the life of severely damaged prefrontal patients, as after lobotomy, induces a disengagement from real life. This underlies the importance of the integrative function of Pc, in particular, between cognition and emotion. Those different functions are not compartmentalized. In particular emotional charge of events, as previously mentioned, modulates the process of memory consolidation in the hippocampo-cortical system. The cooperation of these different functions is perhaps best perceived in the mechanism of attainment of goals. 4) Attainment of goals can be considered as the uttermost expression of the cooperation between Pc and limbic system. This function presents sensory aspects which consist of recognition of goals and evaluation of the outcomes of action, and a motor aspect made up of the setting and execution of motor programs. In the classical learning theory, such as proposed by Skinner (1953), the necessary chaining of sequences of sensory-motor events results from associative (or operant) conditioning of a neutral stimulus by a reinforcer. Cascades of secondary, and higher order, conditioning could account for linking sequences of events together. Obviously, this process can be and has been accounted for without the extensive implication of prefrontal cortex as in our model (Gray ct al., 1991). The main structures concerned are hippocampus, amygdala and basal ganglia. They certainly correspond to a kind of automatic operation mode for the attainment of goals. Nevertheless, several problems arise if the basic components of sequences, plans, or chained actions remain limited to stimulus-response conditional associations. In particular, latent learning (Tolman, 1948) does not obey any clearly defined drive satisfaction, motivation or goal attainment. The conditioning process seems to work correctly for simple sequences of actions. But, taking

J.P. Banquet et al.

179

into account simply the present states in the recognition or reproduction of long sequences of actions leads to a combinatorial explosion of possible paths, which can only be avoided by an active representation of more than one pair of events. Furthermore, the strength of secondary or higher order reinforcers seems to sharply decrease with the distance to the unconditional stimulus. Sigmficant progress has been achieved by the identification of latent learning useful to build cognitive maps, even if these maps are used independently of any prespecified goal. We think that the definition of goals based on the satisfaction of basic (or not so basic) needs conjointly with the learning of more or less complex maps is a further step required to account for complex behaviors. The possibility for diffusion, and in particular retropropagation of goals allows the discovery of solutions that have never been experienced during learning, and thus are created from new by the system. This is an actual illustration of creativity. This functioning mode requires a supplementary degree of freedom in the system, independent from both sensory and motor processing, but still bridging the two systems. This extra degree of freedom is provided by Pc. The efficacy of an algorithmic version of the model for the parsimonious solution of several problems of robotic learning and navigation either in free space or maze constraint does not automatically deliver a certificate of biological plausibility for the system. Nevertheless, the fact that this efficacy has been obtained thanks to a stringent taking into account of essential neurobiological constraints, makes us confident that the model is oriented in a relevant direction. Conclusion

Two different forms of memory, "active" and "dormant", supported respectively by post synaptic potentials (PSPs) and synaptic potentiation are present everywhere in the brain. The interplay between the two forms of memory and in particular the transition and/or the modulation of one form by the other are at the bases of the different processing modes and memory capacities of the brain. Variations in the implementation modalities and in the ranges of these two types of memory along with variation of connectivity give functional specificities to the different systems. This is specially true for Hs and Pc. Cortical processing depends essentially on two memory registers, STM and permanent LTM. The transition from the long-term to the shortterm seems to be direct and normally encounters few problems. Plausibly, at a gross level of analysis the anatomical substrates are topographically identical (STM representing active forms of LTM). Nevertheless, at a fine

180

Chapter 4

grainexi level, the neurophysiological supporting mechanisms must be necessarily different as previously mcntionexi, involving respectively either electrical or durable structural-chemical changes. Conversely, the transition from STM to LTM store follows a more intricate path, probably for the sake of optimizing the amount of information stored, but also for securing the storage of unique events which built up the unique history of each living being. Between these two extreme ranges (STM and LTM), only minor variations from primary to associative areas can be recorded at the sole cortical level, with a tendency to an increase in the temporal range of memory with the increasing level of complexity in processing performed by these areas (Lii ~ al., 1992). Prefrontal and temporal cortices arc endowed with delay neurons that can bridge a gap between two sensory or sensori-motor events. Nevertheless, the tg~nporal range in the usual experimental tests of this property remains largely in the domain attributed to STM, i.e. less than 30 SCCS.

The specificity and vantage point of Hs concerns both topographical and temporal facets. The topographical aspect of Hs specificity as a unique compact site of input convergence and output divergence has been extensively emphasized. It has been credited with the correlational function of Hs which implies some loss of the cortical topology. This functional characteristic is corrected and complemented by a loose topological correspondence between cortical and hippoeampal system in the longitudinal direction. This loose correspondence could be transformed into a dynamic learning-dependent precise mapping between hippocampal and cortical neuronal populations in order to implement the topologically specific consolidation function. This function could be implemented thanks to the fast-transient learning capacities present both within the hippocampus itself, and also at the interfaces between cortex and hippoeampus. The emphasis placed on the spatial aspects of Hs function was detrimental to the exploration of the no less important temporal function. This function results from the capacity of Hs to interact very flexibly with a whole spectrum of registers from the short-term to the longterm, and also possibly to be detached from permanent LTM. The unique characteristic of Hs would be the conjunction of this array of registers with a wide variety of loops of various sizes providing for an easy transition between dormant-inactive and re-activated forms of memory. Beyond these range differences between cortex and Hs memory registers, some more subtle differences could exist in the implementation modalities of active memory. Extensive research has been conducted on delay cells in Pe or temporal cortex, as a support for WM. Indeed, this type of activity can bridge

J.P. Banquet et aL

181

the gap between two sensory or sensory-motor events. Up to now, delay cells have not been located at the level of the Hs. Nevertheless, the equivalent function in Hs could be performed by different mechanisms subserving slightly different functions. First, loop iterative activation could operate the punctual reenactment of recorded patterns of activation either during information processing in WM, or during the more lengthy process of LTM consolidation. Second, spectral timing as performed in our model by DG could also operate the function equivalent to that of the cortical delay neurons. This function consists of maintaining significant information in an active state, while waiting for correlation with a new significant event. This process creates the chaining of basic components of the sequence. The Hs functional specificity would be in multimodal fusion and correlations. Finally, event locked and modulated theta activity could constitute, at least for some species, a basic mechanism for the maintenance of a pattern in an active state, thus making possible a crosS correlation with forthcoming significant patterns. These types of complementary "hardware" constraints in the implementation of active memory and in the range of "dormant" registers determine the type of cooperation established between the two structures Hs and cortex. Further complementarity results from the direct contact of the cortex with environment, favoring externally triggered activation. Conversely, Hs is the only brain structure so easily prone to autoactivation. That property leads, in the pathological domain, to seizure activity. The specific import of Pc to this processing chain seems to result from its unique position at the top of the hierarchy of sensory-motor and motivational streams of information (Figure 11). Both, its independence from and its close contact with multisensory and complex motor representations or codes provides the entire system with an extra degree of freedom. This feature gives to the brain the capacity for: Recording and simulating both sensory and motor sequences independently of their actual implementation, in relation with planning and adaptation; Motivated hierarchical selection of goals and subgoals, and goal attainment; Finally, invention and creativity. These properties can be considered as the highest expression of all these capacities. -

-

-

182

Chapter 4 References

Atkinson, R. C., & Shiffrin, R. M. (1968). Human memory: A proposed system and its control processes. In K. W. Spence & J. T. Spence (Eds.), The psychology of learning and motivation" Advances in research and theory (Vol. 2, pp. 89-185). New York: Academic Press. Baddeley, A. D. (1986). Working memory. Oxford: Oxford University Press. Baddeley, A. D. (1995). Working memory. In M. S. Gazzaniga (Ed.), The cognitive neurosciences (pp. 755-764). Cambridge, MA: MIT Press. Baddeley, A. D., & Hitch, G. J. (1974). Working memory. In G. A. Bower (Ed.), The psychology of learning and motivation: Advances in research and theory (Vol. 8, pp. 47-90). New York: Academic Press. Baddeley, A. D., & Warrington, E. K. (1970). Amnesia and the distinction between long-and short-term memory. Journal of Verbal Learning and Verbal Behavior, 9, 176-189. Banquet, J. P. (1983). Inter and intra-hemispheric relationships of the EEG during sleep in m a n . Electroencephalography and Clinical Neurophysiology, .55, 51-59. Banquet, J. P., & Grossberg, S. (1987). Probing cognitive processes through the structure of event-related potentials during learning: An experimental and theoretical analysis. Applied Optics, 26, 4931-4946. Banquet, J. P., Renault, B., & Les~vre, N. (1981). Effect of task and stimulus probability on evoked potentials. Biological Psychology, 13, 203-214. Banquet, J. P., Smith, M. J., & Renault, B. (1990). Bottom-up versus topdown: An alternative to the automatic attended dilemma? Behavioral and Brain Sciences, 13, 233-234. Banquet, J. P., & Contreras-Vidal, J. L. (1992a). Temporal order, timing, and probability context effects on pattern recognition and categorization. In I. Aleksander & J. Taylor (Eds.), Artificial neural networks 2 (pp. 1885-1890). Amsterdam: Elsevier. Banquet, J. P., & Contreras-Vidal, J. L. (1992b). An integrated neural network-event related potentials model of temporal and probability context effects on categorization. In Proceedings of the International Joint Conference on Neural Networks (pp. 541-546). Hillsdale, NJ: Lawrence Erlbaum. Banquet, J. P., & Contreras-Vidal, J. L. (1993a). Learning temporal contexts and priming preparation modes for pattern recognition. In Proceedings of the International World Conference on Neural Networks (pp. 126131). Hillsdale, NJ" Lawrence Erlbaum.

JP. Banquet et al.

183

Banquet, J. P., & Contreras-Vidal, J. L. (1993b). Spectral timing and integration of multimodal systemic processes. In J. Taylor (Ed.), Artificial neural networks (pp. 350-354). Amsterdam: Elsevier. Banquet, J. P., & Contreras-Vidal, J. L. (1994). Medium and long-term memory in context processing: A network model of cortex-hippocampus relations. Proceedings of the International Conference on Neural Networks, 4, 647-654. Banquet, J. P., Contreras-Vidal, JL., Gaussier, P., Gissler, A., & Bumod, Y. (1997). The cortico-hippocampal system as a multirange temporal processor: A neural model. In R. Park & D. Levin (Eds.), Fundamentals of neural network modelling for neuropsychologists. Boston: MIT Press. Bapi, R. S., & Levine, D. S. (1990). Networks modelling the involvement of the frontal lobes in learning and performance of flexible movement sequences. In Proceedings of Twelth Annual Conference of the Cognitive Science Society (pp. 915-922). Bapi, R. S., & Levine, D. S. (1994). Modelling the role of the frontal lobes in sequential task performance. I: Basic structure and primacy effect. Neural Networks, 7, 1167-1180. Baylis, G. C., & Rolls, E. T. (1987). Responses of neurons in the inferior temporal cortex in short-term and serial recognition memory tasks. Experimental Brain Research, 6.5, 614-622. Berger, T. W., & Thompson, R. F. (1978). Neuronal plasticity in the limbic system during classical conditioning of the rabbit nictitating membrane response. I: The hippocampus. Brain Research, 14.5, 323-346. Br6bion, J. (1994). Memoire de travail, comprehension de texte et vieillissement. Unpublished doctoral dissertation, University Ren6 Descarte, Paris. Brown, J. (1958). Some tests of the decay theory of immediate memory. Quarterly Journal of Experimental Psychology, 1O, 12-21. Bullock, D., Fiala, J. C., & Grossberg, S. (1994). A neural model of timed response learning in the cerebellum. Neural Networks, 7, 1101-1114. Burgess, N., Reece, M., & O'Keefe, I. (1994). A model of hippocampal function. Neural Networks, 7, 1065-1081. Buzsaki, G. (1988). Polysynaptic long-term potentiation: A physiological role of the perforant path - CA3/CA1 pyramidal cell synapse. Brain Research, 4.55, 192-195. Buzsaki, G. (1989). Two-stage model of memory trace formation: A role for 'noisy' brain states. Neuroscience, 31, 551-570.

184

Chapter 4

Cave, C. B., & Squire, L. R. (1992). Intact and long-lasting visual object priming in amnesic patients. Journal of Experimental Psychology: Learning, Memory, and Cognition, 18, 509-520. Cressant A., Muller R., & Poucet B. (1997). Failure of centrally placed objects to control the firing fields of hippocampal place cells. Journal of Neuroscience, 17. Darke, S. (1988). Anxiety and working memory capacity. Cognition and Emotion, 2, 145-154. Dehaene, S., Changeux, J. P., & Nadal, J. P. (1987). Neural networks that learn temporal sequences by selection. Proceedings of the Natwnal Academy of Science of the USA, 84, 2727-2731. Denham, M. J., & Boitano, J. (1996)A model of the interaction between prefrontal cortex, septum and the hippocampal system in the learning and recall of goal-directed sensory-motor behaviours. Technical Report NRG-96-01, University of Plymouth School of Computing. Drachman, D. A., & Arbib, M. (1966). Memory and the hippocampal complex. Archives of Neurology, 15, 52-61. Eichenbaum, H., Otto, T., & Cohen, N. J. (1994). Two functional components of the hippocampal memory system. Behavioral and Brain Science, 17, 449-518. Eysenck, M. W. (1982) Attention and arousal: Cognition and performance. New York: Springer-Verlag. Funahashi, S., Bruce, C., & Goldman-Rakic, P. W. (1989). Mnemonic coding of visual space by neurons in the monkey's dorsolateral prefrontal cortex revealed by an oculomotor delayed-response task. Journal of Neurophysiology, 61, 331-349. Fuster, J. M., & Alexander, G. E. (1971). Neuron activity relatext to shortterm memory. Science, 173, 652-654. Fuster, J. M. (1980). The prefrontal cortex: Anatomy, physiology and neuropsychology of the frontal lobe. New York: Raven Press. Fuster, J. M. (1995). Memory in the cerebral cortex. Cambridge, MA: MIT Press. Fuster, J. M., & Jervey, J. P. (1981). Inferotemporal neurons distinguish and retain behaviorally relevant features of visual stimuli. Science, 212, 5255. Gaussier, P., & Zrehen, S. (1994a). Complex architectures for autonomous agents. In P. Gaussier & J. D. Nicoud (Eds.), PerAc (pp. 278-290). Lausanne: IEEE Press.

J.P. Banquet et al.

185

Gaussier, P., & Zrehen, S. (1994b). Navigating with an animal brain : A neural network for landmark identification and navigation. In Proceedings of Intelligent Vehicles (pp. 399-404), Paris. Gaussier, P., & Zmhen, S. (1995). PerAc: A neural architecture to control artificial animals. Robotics and Autonomous Systems, 16, 291-320. Gaussier, P., Joulain, C., Revel, A., & Banquet, J. P. (1996). Are shaping techniques the correct answer for the control of an autonomous robot ? In Proceedings of UKACC International Conference on Control '96, IEEE. Gaussier, P., Joulain, C., Revel, A. Zrehen, S, & Banquet, J. P. (1997a). Autonomous robot learning: What can we take for free? To appear in Proceedings of IEEE conference on ISIE, IEEE. Gaussier, P., Joulain, C., Zrehen, S. Banquet, J. P, & Revel, A. (1997b). Visual navigation in an open environment without map. To appear in Proceedings of IROS Conference. Glanzer, M., & Amitz, A. R. (1966). Two storage mechanisms in free recall. Journal of Verbal Learning and Verbal Behavior, 5, 351-360. Goldman-Rakic, P. W. (1988). Topography of cognition: Parallel distributed networks in primate association cortex. Annual Review of Neuroscience, 11, 137-156. Goldman-Rakic, P. W. (1994). Working memory dysfunction in schizophrenia. Journal of Neuropsychmtry and Clinical Neuroscience, 6, 348-357. Granger, R., Whitson, J., Larson, J., & Lynch, G. (1994). Non-Hebbian properties of long-term potentiation enable high-capacity encoding of temporal sequences. Proceedings of the Nanonal Academy of Science of the USA, 91, 10104-10108. Gray, C. M., KOnig, P., Engel, A. K., & Singer, W. (1989). Oscillatory responses in cat visual cortex exhibit intercolumnar synchronization which reflect global stimulus properties. Nature, 338, 334-337. Gray, J. A., Feldon, J., Rawlins, J. N. P., Hemsley, D. R., & Smith, A. D. (1991). The neuropsychology of schizophrenia. Behavioral and Brain. Sciences, 14, 1-20. Grossberg, S. (1976a). Adaptive pattern classification and universal recoding. I: Parallel development and coding of neural feature detectors. Biological Cybernetics, 23, 121-134. Grossberg, S. (1976b). Adaptive pattern classification and universal recoding. II: Feed-back, expectation, olfaction, and illusions. Biological Cybernetics, 23, 187-202.

186

Chapter 4

Grossberg, S. (1978). A theory of human memory: Self-organization and performance of sensory-motor codes, maps, and plans. In R. Rosen & F. Snell (Eds.), Progress in theoreacal biology (Vol. 5, pp. 233-374). New York: Academic Press. Grossberg, S., & MerriU, J. (1992). A neural network model of adaptively timed reinforcement learning and hippocampal dynamics. Cognitive Brain Research, 1, 3-38. Guigon, E., Dorizzi, B., Bumod, Y., & Schultz, W. (1995). Neural correlates of learning in the prefrontal cortex of the monkey: A predictive model. Cerebral Cortex, 2, 135-147. Halgrcn, E., Squires, N. K., Wilson, C. L. Rohrbaugh, J. W., Babb, T. L., & CrandaU, P. H. (1980). Endogenous potentials generated in the human hippocampal formation and amygdala by infrequent events. Science, 210, 803-805. Hasselmo, M. E., & SchneU, E. (1994). Laminar selectivity of the cholinergic suppression of synaptic transmission in rat hippocampal region CA l: Computational modeling and brain slice physiology. Journal of Neuroscience, 14, 3898-3914. Horel, J. A. (1994). The localization of general memory functions. Behavioral and Brain Sciences, 17, 482. Howarth, E., & Eysenck, H. J. (1968). Extraversion, arousal, and pairedassociate recall. Journal of Experimental Research in Personality, 3, ll4-116. Ivry, R. B., & Keele, S. W. (1989). Timing functions of the cerebellum. Journal of Cognitive Neurosciences, 1, 134-150. James, W. (1890). The principles of psychology. New York: Holt. Johnson, R., Jr. (1995). On the neural generators of the P300: Evidence from temporal lobeetomy patients. Electroencephalography and Clinical Neurophysiology, 44 (Supplement), 110-129. Johnson, R., Jr., & Donchin E. (1982). Sequential expectancies and decision making in a changing environment: An electrophysiological approach, Psychophysiology, 19, 183-199. Jones, R. S. G. (1993). Entorhinal-hippocampal connections: A speculative view of their function. TINS, 16, 58-64. Jonides, J., Smith, E. E., Koeppe, R. A., Awh, E., Minoshima, S., & Mintun, M. A. (1993). Spatial working memory in humans as revealed by PET. Nature, 363, 623-625.

J.P. Banquet et al.

187

Levine, D. S., & Park, R. (1992). Frontal lesion effects on verbal fluency in a network model. Proceedings of the International Joint Conference on Neural Networks, 2, 39-44. Lii, Z. L., Williamson, S. J., & Kaufman, L. (1992). Human auditory primary and association cortex have differing lifetimes for activation traces. Brain Research, 572, 236-241. Matthews, G. (1992). Extraversion. In A. P. Smith & D. M. Jones (Eds.), Handbook of human performance. Vol. 3: State and trait (pp. 95-126). London: Academic Press. McCarthy, G., Blamire, A. M., Puce, A., Nobre, A. C. Bloch, G., Hyder, F., Goldman-Rakic, P. W., & Shulman, R. (1994). Functional magnetic imaging of human prefrontal cortex activation during a spatial working memory task. Proceedings of the National Academy of Science of the USA, 91, 8690-8694. Miller, E. K., Li, L., & Desimone, R. (1993). Activity of neurons in anterior inferior temporal cortex during a short-term memory task. Journal of Neuroscience, 13, 1460-1478. Miller, E. K., & Desimone, R. (1994). Parallel neuronal mechanisms for short-term memory. Science, 263, 520-522. Milner, B. (1966). Amnesia following operation on the temporal lobes. In C. W. M. Whitty & O. L. Zangwill (Eds.), Amnesia. London: Butterworths. Milner, B., Corkin, S., & Teuber, H.-L. (1968). Further analysis of the hippocampal amnesia syndrome: 14-year follow-up study of H. M. Neuropsychologia, 6, 215-234. Milner, B. (1982). Some cognitive effects of frontal lobe lesions in man. Philosophical Transactions of the Royal Society, London, B, 298, 211226. Miyashita, Y., & Chang, H. S. (1988). Neuronal correlate of pictorial shortterm memory in the primate temporal cortex. Nature, 331, 68-70. Niki, H. (1974). Differential activity of prefrontal units during right and left delayed response trials. Brain Research, 70, 346-349. N~i/it~en, R. (1982). Processing Negativity: An evoked-potential reflection of selective attention. Psychological Bulletin, 92, 605-640. O'Keefe, J., & Nadel, L. (1978). The hippocampus as a cognitive map. Oxford: Oxford University Press. Olton, D. S., Becker, J. T., & Handelmann, G. E. (1979). Hippocampus, space and memory. Behavioral and Brain Science, 2, 313-365.

188

Chapter 4

Owen, A. M., Morns, R. G., Sahakian, B. J., Polkey, C. E., & Robbins, T. W. (1996). Double dissociations of memory and executive functions in working memory tasks following lobe excisions, temporal lobe excisions or amygdalo-hippocampectomy in man. Brain, 119, 15971615. Paulescu, E., Frith, C. D., & Frankoviak, R. S. J. (1993). The neural correlates of the verbal component of working memory. Nature, 362, 342-345. Peterson, L. R., & Peterson, M. J. (1959). Short-term retention of individual verbal items. Journal of Experimental Psychology, 58, 193-198. Petrides, M., Alivisatos, B., Evans, A. C., & Meyer, E. (1993). Dissociation of human mid-dorsolateral from posterior dorsolateral frontal cortex in memory processing. Proceedings of the National Academy of Science of the USA, 90, 873-877. Rolls, E. T., Baylis, G. C., Hasselmo, M. E., & Nalwa, V. (1989). The effect of learning on the face selective responses of neurons in the cortex in the superior temporal sulcus in the monkey. Experimental Brain Research, 76, 153-164. Rolls, E. T., & O'Mara, S. M., (1995) View-responsive neurons in the primate hippoeampal complex. Hippoeampus, 5, 409-424. Ruchkin, D. S., Sutton, S., Mahaffey, D., & Glaser, J. (1986). Terminal CNV in the absence of a motor response. Electroencephalography and Clinical Neurophysiology, 63, 445-463. Ruehkin, D. S., Johnson, R., Jr., Canoune, H., & Ritter, W. (1991). Eventrelated potentials during arithmetic and mental rotation. Electroencephalography and Clinical Neurophysiology, 79, 473-487. Ruehkin, D. S., Johnson, R., Jr., Grafman, J., Canoune, H., & Ritter, W. (1992). Distinctions and similarities among working memory processes: An event-related potentials study. Cognitive Brain Research, 1, 53-66. Sehacter, D. L., Chu, C. -Y., & Oehsner, K. N. (1993). Implicit memory: A selective review. Annual Review ofNeuroscience, 16, 159-82. Seoville, W. B., & MiMer, B. (1957). Loss of recent memory after bilateral hippoeampal lesions. Journal of Neurology, Neurosurgery, and Neuropsyehiatry, 20, 11-21. Shiffrin, R. M., & Schneider, W. (1977). Controlled and automatic information processing. II: Perceptual learning, automatic attending and a general theory. Psychological Review, 84, 127-90.

J.P. Banquet et al.

189

Sidman, M., Stoddard, L. T., & Mohr, J. P. (1968). Some additional quantitative observations of immediate memory in a patient with bilateral hippocampal lesion. Neuropsychologia, 6, 245-54. Singer, W. (1983). Neuronal activity as a shaping factor in the selforganization of neuron assemblies. In E. Basar, H. Flohr, H. Haken, & A. J. Mandell (Eds.), Synergetics of the brain. New York: SpringerVerlag. Skinner, F. (1953). Science and human behavior. New York: McMillan. Solomon, P. R. (1980). A time and a place for everything? Temporal processing views of hippocampal function with special reference to attention. Physiological Psychology, 8, 254-61. Squire, L. R., Ojemann, J. G., Miezin, F. M., Petersen, S. E., Vdeen, T. O., & Raichle, M. E. (1992). Activation of the hippocampus in normal humans: A functional anatomical study of memory. Proceedings of the National Academy of Science of the USA, 89, 1837-41. Tolman, E. C. (1948). Cognitive maps in rats and men. The Psychological Review, 55, 189-208 Zrehen., S. (1995). Elements of brain design for autonomous agents. Unpublished PhD thesis, Swiss Federal Institute of Technology, Lausanne. Zrehen, S., & Gaussier P. (1997). A neural architecture for motivated landmark-based navigation. ETIS internal report (submitted for publication). Author Note

This research was supported by INSERM, NATO and DGA/DRET Grant # 911470/A000/DRET/DS/DR.

This Page Intentionally Left Blank

P A R T II PERSPECTIVES FROM EMOTION RESEARCH

This Page Intentionally Left Blank

Cognitive Science Perspectives on Personality and Emotion - G. Matthews (Editor) 1997 Elsevier Science B.V. CHAPTER 5

Affective Influence in Perception: Some Implications of the Amplification Model Shinobu Ki tayama

Will affectively charged stimuli be perceived any differently from affectively neutral ones? Will affect inherent in a focal stimulus disrupt perceptual processing? Or will it facilitate the latter so that affective stimuli stand out in the perceptual field? After years of waxing and waning (e.g., Dixon, 1980; Erdelyi, 1974, 1985), the influence of stimulus affect on immediate perception remains as a topic of considerable significance. Mechanisms underlying immediate conscious perception are both logically (e.g., Helmholtz, 1884) and empirically (e.g., Marcel, 1983a) preconscious. Thus, if it can be shown that the immediate perception of a stimulus is indeext influenced by affect inherent in the stimulus itself, we will have identified a window through which to observe what Kihlstrom (1990) has called the psychological unconscious. Championed by Freud and his successors (e.g., Freud, 1895/i966), the functional structure of the unconscious, especially the one involving affect, has turned out to be one of the most formidable problems in psychology, often evading scientific scrutiny. However, with rigorous experimental methodologies and theoretical tools now available at hand, recent investigations on detectionless processing (e.g., Bargh, Bond, Lombardi, & Tota, 1986; Carr & Dagenbach, 1990; Greenwald, Klinger, & Lui, 1989; Marcel, 1983a,b; Niedenthal, 1990; Shevrin, 1990), automatic processing (e.g., Uleman & Bargh, 1989; Shiffrin & Schneider, 1977), and implicit memory (e.g., Schacter, 1989) have taken significant steps toward more comprehensive and accurate understanding of the unconscious. And a new theoretical framework has begun to emerge (e.g., Erdelyi, 1985; Lewicki, 1986; Kihlstrom, 1990; Marcel, 1983b; Rumelhart, 1989; Zajonc, 1980). The present paper seeks to contribute to this literature. We will examine whether and how the perceptibility of a faintly shown stimulus can vary with the affective significance of the stimulus itself. The goal is to identify distinctly affective phenomena in a perceptual identification task, and integrate them with current theories of cognition, affect, and attention, thereby laying a solid foundation for understanding other forms of

194

Chapter 5

"hot cognitions" (Abelson, 1963). Should perceptibility depend on affective significance, that would suggest that affect can be preconsciously elicited by an impinging stimulus and, moreover, that the elicited affect influences subsequent processing required to develop a conscious percept. Accordingly, the present work has the potential of revealing the nature of a preconscious interaction between affect and cognition. The current approach emphasizes a non-associative, energizing consequence of affect (of. Osgood, 1962) and analyzes how stimulus affect influences the perception of the stimulus itself. Specifically, it will be proposed that affect evoked through preattentive processing amplifies attentive processing, thereby either enhancing or impairing the emerging conscious percept. This work therefore will supplement a presently dominant, largely associative approach to affect (e.g., Bargh et al., 1986; Bower, 1981; Fazio, Sanbonmatsu, Powell, & Kardes, 1986; Greenwald et al., 1989; Isen, Shalker, Clark, & Karp, 1978; Johnson & Tversky, 1984; Lang, 1084; Niedenthal, 1990; Zajone, Murphy, & Ingelhart, 1989), which has proved powerful in analyzing how affect of one stimulus (prime) can bias the perception of another (target). It is typically assumed in this literature that the activation of affective information, caused by the prime, can spread to related information within a network of associative memory, thus biasing the perception of the target. A historicalperspective: The "New Look" and its aftermath The general issue of affect-cognition interaction in perception can be traced back to the literature of "New Look" in perception in the 1950s (Bruner, 1957). It was then proposed that perception depends not only on exogenous factors, but also on endogenous factors including perceptual set, expectation, motivation, personality, and affect (see e.g., Allport, 1955, for a review). In a pioneering experiment on affect and perception, MeGinnies (1949) examined the perception of affcctivdy charged words. He briefly flashed either a taboo word or a neutral word, and found recognition threshold to be considerably higher for the taboo word than for the neutral word. He maintained that this resulted from perceptual processes; affect ("anxiety") evoked by a taboo word recruited the process of psychological defense, which blocked further perceptual processing, thus diminishing the conscious percept (see also Blum, 1954). Other studies observed that affect sometimes enhanced perception (e.g., Postman, Bronson, & Gropper, 1953).

S. Kitayama

195

The proponents of perceptual defense then suggested the operation of perceptual vigilance. Critics quickly pointed out an apparent paradox involved in McGinnies's assertion that one can feel "anxiety" without knowing the identity of the stimulus. They maintained that the finding could be explained most parsimoniously by post-perceptual response biases such as reluctance or readiness to report a taboo word (Eriksen, 1963; Goldiamond, 1958: see Erdelyi, 1974; Dixon, 1980 for reviews). There are some methods, however, that allow one to examine perceptual accuracy independent of post-perceptual response bias. For instance, subjects may be asked to choose the item shown from a pair of equivalently valenced words. Research employing this and other similar procedures has demonstrated that the affective tone of a stimulus does influence accuracy in perception (e.g., Bootzin & Natsoulas, 1965; Dorfman, 1967). Thus, McGinnies was correct in this regard. Further, to be reviewed below, recent cognitive research has strongly suggested that conscious perception is the end product of a number of preconscious operations. Hence, McGinnies's notion that affect can be induced by an impinging stimulus before the stimulus is consciously identified is no longer considered paradoxical (Erdelyi, 1974). Nevertheless, his theory that processes of defense/vigilance mediate the effects of affect in perception has faced serious challenges. Neither he nor his successors articulated the mechanisms of defense or vigilance. Hence, no prediction about the direction of the influence of affect is possible. Further, evidence suggests that affect can influence the perceptibility of a stimulus whether its valence is positive or negative (Broadbent & Gregory, 1967; Kitayama, 1990, 1991). In retrospect, then, it would seem that McGinnies was correct in that the effect he observed was, at least in part, perceptual. However, his hypothesis of defense/vigilance as an underlying mechanism is increasingly suspect.

The present approach The current paper presents a model of affect-cognition interaction designed to account for perceptual influence of affect. We reconsider this old problem, traditionally studied under the rubric of defense and vigilance, from a new perspective afforded by a number of theoretical and methodological innovations accomplished in the interim. Informed by current theories of cognition, affect, and attention, the model hypothesizes that affect induced through preattentive processing of an impinging stimulus amplifies

196

Chapter 5

subsequent attentive processing. We will review past studies on preattentive processing, attention, and affect, which together form the empirical and theoretical basis for the proposed model. It will be shown that the model provides a coherent account for an anomalous pattern of past findings in the defense/vigilance literature. Further, the model has guided more recent empirical investigations on the topic. Evidence for the model will be reviewed, and two new experiments will be reported. Finally, broader implications of the proposed model will be discussed and directions for future research explored. The present attempt to test implications of the model strategically focuses on just one form of perception: word perception. Studying this relatively simple case should make it possible to exert precise experimental control and, thus, to test the model in a more rigorous fashion. Further, in the domain of word perception, there is considerable overlap between research on affect and research on cognition. Much of past research on affect and perception, dating back to McGinnies's original contribution, was done with words as stimuli, and word perception has been extensively studied in current cognitive psychology (see e.g., Posner, 1989). The Amplification Model of Affect-Cognition Interaction A model

A model of affect-cognition interaction in early perceptual processing (Kitayama, 1990, 1991; Kitayama & Howard, 1994) is illustrated in Figure 1. The basic tenet of the model is that affect induceA through preattentive processing of an impinging stimulus amplifies subsequent attentive processing, thereby either enhancing or impairing the conscious percept of the stimulus. Unlike the defense/vigilance hypothesis, this model assumes that an influence of affect in perception results from interaction among three component processes commonly implicateA in ordinary processes of perception, i.e., preattentive processing, attentive processing, and affect. Preattentive processing. According to current cognitive theories of perceptual processing of lexieal materials such as words (e.g., McClelland & Rumelhart, 1981; Posner, 1978) and certain graphic stimuli such as faces (Bauer, 1984; Damasio, Damasio, & Van Hoesen, 1982; Tranel & Damasio, 1985), an impinging stimulus is initially processed automatically without any involvement of attention. Through preattentive processing, the graphic and possibly the semantic perceptual codes that correspond to the stimulus can be

S. Kitayama

197

ii

Preattentive processing

,11111

i

)" i|1 ii iii

I Engagement of attention in a relevant

i

i

|l

Activation of affective circuits and subsequent amplification of attentive processing

Attentive

processing

Conscious

percept

Response (i.e., reading out relevant features, selecting responses, etc.)

Figure 1. A schematic illustration of the amplification model of affectcognition interaction in early perceptual processing. The model is composed of, as its major components, preattentive processing, attentive processing, and activation of affect and subsequent amplification of attentive processing, and response. These components are highlighted in bold squares.

activated before the conscious percept of the stimulus is developed. A number of recent studies with a semantic priming paradigm have shown that semantic information (and, by implication, graphic information as well) can be activated by a word that is pattern masked and thus made undcteetable in consciousness (e.g., Allport, 1977; Balota, 1983; Carr, McCaulcy, Spcrber, & Parmclee, 1982; Carr & Dagcnbach, 1990; Grccnwald ct al., 1989; Fowler, Wolford, Sladc, Tassinary, 1981, Marcel, 1983a, 1983b). This

198

Chapter 5

literature lends strong support to the hypothesis that attention, in the sense of a set of processing mechanisms selectively and serially applied to a confined sp~/tial or semantic region, is not necessary for the activation of meanings of an impinging lexieal stimulus, let alone the activation of its shape. The foregoing conclusion might seem at odds with cognitive research on vision, which has traditionally assumed that preattentive vision is very crude (Egeth, 1977; Neisser, 1967). For example, Sagi and colleagues (Braun & Sagi, 1990; Sagi & Julesz, 1985) have proposed that preattentive vision allows one to note that a field contains a discrepancy, but does not enable one to identify any object. Similarly, for Treisman (1988), preattentive vision is sufficient to identify separate features of a field (such as "green" or "square"), but not any object defined by a conjunction of more than two features (such as "green square"). However, the vision literature does not necessarily contradict the above evidence for the preattentive activation of shape and meaning of a perceptual object. The two lines of research typically use conspicuously different stimulus materials. On the one hand, the vision literature has focused primarily on simple and to a large extent, arbitrary graphic stimuli (e.g., colors, lines, simple geometric figures) that have no obvious, unique meanings. On the other hand, the studies attesting to the presence of preattentive activation of shape and meaning employ meaningful stimuli that are routinely encountered in everyday life, viz., mostly lexical materials such as words, but occasionally certain complex and realistic graphic materials such as faces. It goes without saying that some kind of preexisting processing structures such as the ones exemplified in connectionist networks are required for preattentive activation of shape or meaning to take place (McClelland & Rumelhart, 1981). These structures will develop gradually from everyday encounter with relevant stimuli (e.g., LaBerge & Samuels, 1973; Shiffrin & Schneider, 1977) although those for certain phylogenically significant stimuli such as faces may be hard-wired through evolution (Field, 1985). Thus, the extent of preattentive processing can vary from very crude (as in the case of arbitrary and/or meaningless stimuli for which no ready-made processing structure is available) to very thorough and sophisticated (as in the case of meaningful/lexical materials for which elaborate processing structures have been established and, thus are readily available). A series of studies by Shevrin and his colleagues provided some evidence (Shevrin & Fritzler, 1968; Shevrin, Smith, & Fritzler, 1971). They found that evoked potential to a subliminally shown picture is significantly more intense if the picture is meaningful than if it is meaningless. Further, the meaningfulness of the

S Kitayama

199

picture systematically influenced subsequent free associations as well, suggesting that semantic activation was actually caused by the subliminally shown picture if it was meaningful. All in all, then, once a meaningful stimulus commonly encountered in daily life such as a word is presented, it will be processed automatically, and corresponding perceptual codes are activated. Although this activation is preconscious and quite weak especially if the stimulus is impoverished, it has been shown to be sufficient for affect associated with the stimulus to be elicited covertly (Ohman, 1985). Perhaps, the activated codes summon affective circuits of the brain located in the limbic or subcortical regions. LeDoux (1987, 1989) has reviewed neuroanatomical evidence suggesting numerous neuronal connections between sensory processing areas of the brain and the limbic regions. Furthermore, several studies (e.g., Corteen & Wood, 1972; Lazarus & McCleary, 1951; Zajonc, 1962) have shown a reliable autonomic response to a subliminal affective stimulus. Although this literature has been criticized on methodological grounds (e.g., Merikle, 1982; Holender, 1986), more recent research with a strict criterion for awareness has also shown that affect can be elicited by undetectable stimuli (e.g., Dawson & Schell, 1982; Greenwald et al., 1989; Niedenthal, 1990; Tassinary et al., 1984; see also Kunst-Wilson & Zajonc, 1980). Attentive processing. It is reasonable, then, to postulate that affect elicited via preattentive processing in turn influences subsequent attentive processing, which is generally believed necessary for conscious perception of the stimulus (Neely, 1977; Posner & Snyder, 1975). Unlike preattentive processing, attentive processing is selective, limited solely to a perceptual code to which attention has been directed. Thus, once a relevant perceptual code has been automatically and preconsciously activated by an impinging stimulus, two operations need be performed. First, attention is shifted and directed to the relevant code and, second, once so directed, attention furthers the code's processing (cf. Posner, 1980). Through attentive processing, a more elaborate perceptual and, perhaps, semantic image of the stimulus is developed, which corresponds to the immediate conscious percept of the stimulus. Finally, the conscious percept may be scanned and its more specific features may be read out to control subsequent action (Allport, 1989). In order to grasp the proposed relationship between preattentive processing and attentive processing, a metaphor of attention as a spotlight is useful (Crick, 1984; LaBerge, 1983; Moser, 1988; Posner, 1980). According to this metaphor, preattentive processing activates the relevant one of numerous perceptual codes in long term memory. This activation itself,

200

Chapter 5

however, is not enough to produce conscious perception. For the latter to occur, an attentional spotlight must be deployed. The spotlight must first be shifted to the relevant code and then used to illuminate the code. This illuminating of the relevant information amounts to additional processing performed on the latter and, as such, is thought to enable conscious perception of the stimulus. In this formulation, attentive processing is distinct from preattentive processing in its selective nature. Whereas preattentive activation can occur simultaneously at multiple loci (e.g., letter, graphic, semantic codes that are valid as well as, especially when the stimulus is impoverished, those that are invalid; see below), attentive processing can be focused on only one of them. Another implication of the current formulation is that attention and consciousness are distinct even though there is substantial overlap between them. Generally, preattentive (nonselective) processing takes place without conscious awareness, whereas attentive (selective) processing is mostly conscious. However, attentive (selective) processing required to produce conscious awareness is necessarily preconscious. Amplification by affect. One widely postulated property of affect is arousal, or its ability to amplify a variety of psychological functions. In his pioneering work, Tomkins (1962, 1980) has proposed that various basic emotions such as joy and anger can be described in terms of differential patterns of amplification of a nervous system. Although Tomkins's analysis may no longer seem feasible, arousal or an intensity dimension of affect has been shown to be essential in defining a variety of everyday vocabularies of emotion and concepts in general, and suggested to be universal across cultures (Osgood, 1962; Russell, 1980). Another major dimension of affect identified in this literature is pleasantness. From the very beginning, it has been widely recognized that an amplifying property of affect can have a variety of consequences on psychological processes. Under the guise of drive, this assumption is central to a behavioral theory of learning proposed by Hull, Spence, and Taylor in the 1950s (e.g., Spence, 1956). It also is at the core of the Yerkes-Dodson law (Yerkes & Dodson, 1908), as well as its modem extensions by H. Eysenck (1967) to analysis of personality dimensions of extraversion/ introversion and impulsivity. It has also proved applicable to social facilitation (Zajone, 1965). More recently, Revelle, Humphreys, and their colleagues (Humphreys & Revelle, 1984; Revelle & Loftus, 1990; see also M. Eysenck, 1976) have elaborated on some specific consequences of arousal on different stages of memory processes.

S. Ki tayama

201

As has been pointed out by a number of researchers, the notion of generic, uni-dimensional arousal involving all aspects of the sympathetic nervous system and those of cortical processes seems too simplistic (e.g., Lacey & Lacey, 1968). Nevertheless, the hypothesis that affect amplifies some aspects of psychological processes remains both reasonable (Lindsley, 1951) and empirically viable (Stembach, 1968). And, as such, it has the potential of clarifying ways in which affect influences cognition. Extrapolating from this literature, it may be hypothesized that affect elicited through preattentive processing amplifies attentive processing. This simple hypothesis suggests both (i) the conditions in which affect associated with an impinging word is most likely to enhance the perception of the word itself and (ii) those in which the affect is most likely to impair the perception. Enhancement and impairment of perception by affect. First and most obvious, if attention has accurately been directed to a relevant perceptual code (i.e., the one corresponding to an impinging word), affect and ensuing amplification of attention should enhance the veridical perception of the impinging stimulus. In this case, affective stimuli will be more accurately perceived than neutral stimuli (affective enhancement). Suppose, however, that a stimulus is presented in an extremely impoverished manner, as is often the case in perceptual identification experiments. Under these conditions, the relevant perceptual code will not receive strong activation. As we have reviewed earlier, this weak activation seems sufficient to produce a degree of affect, thus amplifying subsequent attentive processing. Nevertheless, the weak activation will cause considerable difficulty in computing exactly which perceptual code corresponds to the impinging stimulus, especially because residual activations caused by past experience are likely to remain for many other irrelevant codes (e.g., Jacoby, 1983; Higgins & Bargh, 1987). Because of this difficulty in locating the relevant code, attention may be misdirected to an irrelevant code. Under these conditions, affect produced through preattentive processing will amplify attentive processing that has been directed, accidentally, to invalid perceptual information and, as a consequence, it will impair an emerging conscious percept. In this case, affective stimuli will be less accurately perceived than neutral stimuli (affective impairment). In terms of the spotlight metaphor introduced earlier, preattentive processing of an affective stimulus activates the corresponding code and, as a consequence, evokes associated affect, which in turn increases the illumination of the attentional spotlight. However, because the activation of the relevant code is weak, perhaps no stronger than residual activations

202

Chapter 5

remaining in irrelevant codes, the spotlight is likely to be locked on to one of the irrelevant codes, accidentally illuminating the latter and, thus, causing an impairment of valid perception: In sum, the present model (the amplification model hereafter) predicts that affective impairment should be most likely to occur when the presentation of a target word is extremely impoverished. It further implies that the impairment effect should disappear or even reverse itself once the difficulty in locating the relevant code is alleviated, that is, when the activation of the relevant code is increased relative to the activation of other irrelevant codes. Under these conditions, attention will be successfully directed to the relevant perceptual code and, as a consequence, affect and subsequent amplification of attentive processing should enhance the emerging percept. The general prediction tested, therefore, can be stated in terms of affective enhancement (higher accuracy for affective than for neutral stimuli within a given experimental condition) or affective impairment (lower accuracy for affective than for neutral stimuli within a given experimental condition): Any variable that increases the activation of a relevant perceptual code relative to the activation of other, irrelevant codes will increase the likelihood of affective enhancement and~or decrease the likelihood of affective impairment. Evaluation Criteria of the Amplification Model Two points must be made explicit before setting out to test implications of the amplification model. First, the model predicts that stimulus affect should influence perceptual accuracy independently of response bias either for or against reporting an affective stimulus. Earlier studies in the defense and vigilance literature were criticized largely because they used recognition threshold as a dependent variable. With this measure it is extremely difficult to separate perceptual accuracy from response bias (Eriksen, 1963; Goldiamond, 1958). As noted above, however, there are some methods, most notably forced choice between two affectively equivalent stimuli, that allow one to control for response bias (Natsoulas, 1965). In the following, we will draw primarily on those studies that have adequately controlled for response bias. Second, the engagement of attentive processing in a preattentively activated perceptual code is only one of several distinct operations that can

S. Kitayama

203

contribute to overall perceptual identification (see Figure 1). The model therefore assumes that any variable that can enhance attentional engagement may also either improve or impair other operations, and thus either increase or decrease the overall perceptual identification independently of stimulus

affect. To illustrate, consider stimulus complexity, which is likely to have opposing effects on preattentive processing and response selection. To begin with, the more complex a stimulus is, the greater the number of features it contains. Because each of these features will serve as an additional constraint in preattentive processing, as stimulus complexity increases, the corresponding perceptual code may be more unequivocally activated. According to the amplification model, under these conditions the percept will be more accurate for affective stimuli than for neutral stimuli (affective enhancement). In addition, however, once the percept has been developed, the respondent will subsequently have to make an overt response. Most of the studies to be reviewed or reported in the current paper examine accuracy in a forced choice, whereby the respondent supposedly scans and compares the percept with available alternatives. Because complex stimuli contain more features to be compared and matched in the choice, stimulus complexity should make response selection more difficult, thereby leading to poorer overall performance. In short, stimulus complexity is likely to increase the likelihood of affective enhancement, while simultaneously decreasing choice performance. Once these two effects of stimulus complexity are super-imposed on each other, one will observe performance for affectively neutral stimuli to decline with stimulus complexity. Relative to this base line defined by the neutral stimuli, performance for comparable affective stimuli should improve. Yet, this improvement due to stimulus affect may or may not compensate for the decline of overall performance due to choice difficulty. This means that performance for affective stimuli may or may not actually improve with stimulus complexity. The crucial prediction of the amplification model in this case then, is that a decline of performance as a function of stimulus complexity is less for affective stimuli than for neutral stimuli. In general, it is safe to assume that any variable that can enhance attentional engagement (e.g., stimulus complexity) may also either improve or impair other operations (e.g., response selection), and thus either increase or decrease overall performance independently of stimulus affect. Accordingly, the amplification model must be evaluated in terms of its ability to predict either affective enhancement or impairment, rather than its ability to predict

204

Chapter 5

an absolute increase or decrease of performance for affective or neutral stimuli with the manipulated variable. With these evaluation criteria at hand, we now turn to a review of extant studies pertinent to some predictions of the amplification model.

Empirical evidence Expectation and word frequency. Research on priming has amply demonstrated that when individuals are led to expect the identity of a target, the corresponding perceptual code receives extra activation (e.g., Higgins & Bargh, 1987; Neely, 1977; Posner & Snyder, 1975). It can be hypothesized, therefore, that in a perceptual identification task, the difficulty of locating a relevant perceptual code is relieved by a correct expectation about the target word. Thus, affective enhancement will be more likely and affective impairment less likely in the presence of a valid expectation than in its absence (Kitayama, 1990). Initial support for the prediction was uncovered in a review of the literature of perceptual defense and vigilance. Because most studies in this literature examined recognition threshold and failed to control for response bias, their status as evidence for the current analysis is uncertain (Eriksen, 1963; Goldiamond, 1958; see e.g., Dixon, 1980; Erdelyi, 1974, for reviews). Nevertheless, two experiments which manipulated expectation tended to support the amplification model (Freeman, 1954; Lacey, Lewinger, & Adamson, 1953). In these studies, when there was no expectation, recognition threshold was higher for affective words than for neutral words (affective impairment); but when an expectation about the identity of a target was provided, recognition threshold for affective words was lower than that for neutral ones (affective enhancement; see also Postman et al., 1953, for a similar result). Additional evidence for the present analysis can be found in more recent, methodologically more sophisticated studies that assess perceptual accuracy independently of response bias (either readiness or reluctance to report affective rather than neutral stimuli). Kitayama (1990) located nine such studies (see Table 1). In most of these studies (Bootzin & Natsoulas, 1965; Broadbent & Gregory, 1967; Dorfman, 1967; Dorfman, Grossberg, & Kroeker, 1965), the dependent variable was correct response rate, with appropriate adjustments made for response bias. Two additional studies used different methods to minimize response bias. Chapman and Feather (1972) examined the ability to detect (rather than identify) a novel graphic stimulus using a signal detection procedure. They assigned an affective tone to the

S. Kitayama

205

stimulus by classically conditioning an electric shock to it. Sales and Haber (1968) minimized response bias by having subjects report individual letters of a flashed word rather than the word itself, and analyzed the number of letters correctly reported.

Table 1. Comparison of nine experiments in which effects of response bias were minimized (adapted from Kitayama, 1990). Experiment

Expectation

Outcome

Chapman & Feather (1972) Dorfman (1967) Dorfman et al. ( 1965) Mathews & Wertheimer(1958) Minard (1965) Van Egeren (1968) Bootzin & Natsoulas (1965) Broadbent & Gregory(1967) . Sales & Harber (1968)

Clear Clear Clear Vague Vague Vague Absent Absent Absent

Enhancement Enhancement Enhancement Inconsistent1 Inconsistent2 No effect Impairment Impairment Impairment

1 Significant impairment effect was found for "high-hysteria" subjects, but no effect was obtained for "high-psychasthenia"subjects. 2 Significant impairment effect was found for males, whereas significant enhancement effect was found for females.

Among these nine studies, three obtained affective enhancement (Chapman & Heather, 1972; Dorfman, 1967; Dorfman et al., 1965). Interestingly, all the three studies inadvertently used a procedure that assured that the subjects had a clear expectation about the target stimulus. In two experiments by Dorfman subjects were shown the target word plus a nontarget word immediately before the target was actually flashed. They were told that one of the pre-target words would be flashed on that trial. Chapman and Feather (1972) had subjects keep in mind the target stimulus while seeing a visual display. Thus, both methods provided subjects with a clear expectation. Some other experiments implanted subjects with vague expectations by familiarizing them with experimental stimuli at the beginning of the session. In these studies there was no systematic pattern. Mathews and Wertheimer (1958) and Minard (1965) found the influence of affect to depend

206

Chapter 5

on certain individual difference variables, and Van Egeren (1968) failed to find any influence of affect on perceptual accuracy. The remaining did not use any of the above procedures and reported affective impairment (Bootzin & Natsoulas, 1965; Broadbent & Gregory, 1967; Sales & Haber, 1968). Kitayama (1990) subsequently conducted an experiment in which expectation was systematically manipulated. In this experiment word frequency was also varied. It was hypothesized that as word frequency increased, the valid perceptual code would be more strongly activated and, as a consequence, affective impairment would become less likely and affective enhancement more likely. Subjects were exposed to a 25 ms flash of a target word. They then chose the target word from a word pair. In half the trials, this word pair was given before the flash to create an expectation. Further, on some trials no target was presented although subjects were led to believe that it was actually shown. Analysis of the data from these trials revealed no response bias for or against reporting affective stimuli, so choice hit rate was used as a measure of perceptual accuracy. Consistent with the amplification model, both expectation and word frequency increased the likelihood of affective enhancement and decreased the likelihood of affective impairment. As can be seen in Figure 2, when words were low in frequency (10-50 occurrences per million) and an expectation was absent, affective words were identified significantly less accurately than neutral ones (affective impairment). This pattern, however, was reversed to show a reliable enhancement effect when high-frequency words (more than 100 occurrences per million) were examined and an expectation was present. Finally, the influence of affect in the remaining two conditions (high frequency/unexpected and low frequency/expected) was no greater than that in the former two conditions. The Kitayama (1990) study thus generally confirmed the predictions of the amplification model. Nevertheless, it was not totally conclusive. First, it tested only a small number of words (12 in total). Second, it found the predicted effect of expectation only for high-frequency words. There was no such effect for low-frequency words: as can be seen in Figure 2, affeetive impairment of evidently equal strength was observed regardless of expectation. Stimulus contrast. The amplification model states that affective impairment is most likely when the presentation of a target is extremely impoverished. Another recent set of experiments with a larger number of stimulus words (126 in total, ranging from 8 to 65 per million in frequency of occurrence) has provided support (Kitayama, 1991). In Study 1, a target was

S. Kitayama

207

0.68 0.66

"

0.64

--

['7

Affeetivq

1

Neutral

0.62 0.60

--

0.58 0.56 Im qD

0.54 0.52 0.50

i

.......

I

i

..0

Word Frequency/Expectation

Figure 2. Perceptual accuracy (hit rate) as a function of target affectivity, word frequency, and expectation (adapted from Kitayama, 1990). presented in dark gray (luminance = .72 if-L) in a black background (.67 ilL), so the contrast between the target and the background was extremely low. As predicted, a significant affective impairment effect was obtained - that is, the identification of affective words was less accurate than the identification of neutral words. In Study 2, however, the contrast was increased so that the target was shown in lighter gray (.75 if-L). Under the latter condition, there was no influence of affect. Exposure time. According to the amplification model, a relevant perceptual code needs to be located quite early in the processing before attention is directed. It then follows that effects of exposure times should depend crucially on the range in which they are manipulated. When relatively long exposure times are manipulated, it will be only late in the processing that these variations begin to increase the activation of the relevant code. Thus, under the conditions of extremely impoverished stimulus contrast, attention should be misdirected to an irrelevant code regardless of the exposure times. In Study 1 of Kitayama (1991) described above, three relatively long exposure times (100, 150, and 200 ms) were tested. As predicted, affective impairment of evidently equal magnitude was observed in all the three

208

Chapter 5

exposure time conditions. By contrast, when relatively short exposure times are tested, an increase in exposure times should increase the activation of a relevant code early in processing, thus mitigating the difficulty in locating the code. Thus, an increase in relatively short exposure times should result in a lesser likelihood of affective impairment and a greater likelihood of affective enhancement. Kitayama (1989) showed that affective impairment observed with a 25 ms exposure (Kitayama, 1990) could disappear with a 40 ms exposure. Word length. Virtually every past study in this area has examinexl only relatively short words (less than 6 letters long). However, current cognitive models of word recognition (e.g., McCleUand & Rumelhart, 1981) suggest that word length may systematically change the likelihood of affectivr enhancement and impairment. According to these models, the initial, preattentivr processing of a visual input process in parallel, leading to simultaneous activation of parts of the entire input. Currently, there is no consensus about exactly what defines functional parts of a word. Drawing on some prominent models of word recognition (e.g., McCleUand & Rumelhart, 1981), we assume here that "word-parts" correspond fairly closely to individual letters, although, for the purposes of the present argument, however, it makes little difference whether the units are letters or something else. Once individual letters have been activated, they in turn impose significant constraints on the likely identity of the input, permitting only a limited number of English words as reasonable candidates for the input. All else being equal, as word length increases, a greater number of letters should be activated and the letter-level information should more strongly constrain the word-level identity. To illustrate, imagine that the processing of a fourletter word successfully activated half of the constituent letters, say, "LxxE." There are several 4-letter candidate words that meet these constraints, say, "LIVE," "LIKE," "LOVE," "LAKE," and so on. In contrast, if half of the constituent letters are activated in a word that is 10 letters long, say, "AxTRAxxlxx," there will be very few 10-letter words other than "ATTRACTIVE" that fully meet the constraints. Thus, as word length increases, the perceptual code corresponding to an impinging word will be more unequivocally and uniquely activated and attention will be more likely to be directed to the valid perceptual code. As word length increases, there should be a greater chance of affective enhancement as opposed to affective impairment.

S. Ki tayama

209

In a recent experiment, both word affect and word length were systematically manipulated (Study 1 in Kitayama, 1991). Words comparable to Kitayama's (1990) low-frequency words were presented with an extremely diminished stimulus contrast and no expectation was provided - a condition the amplification model predicts to be highly conducive to affective impairment. A reliable affective impairment effect was observed. However, this impairment effect was also observed for longer words, thus failing to confirm the model's prediction. Perhaps, with the extremely diminished stimulus contrast examined in this study, there was only marginal activation of a relevant perceptual code regardless of word length. It would seem reasonable that an increase in word length could contribute to the unique and unequivocal activation of a relevant perceptual code only when there was enough stimulus input. In the experiments to be reported below, therefore, targets were presented with greater stimulus contrast. The perceptibility of the target was then reduced by presenting a masking stimulus immediately after the disappearance of the target. Under these conditions of backward pattern masking, an increase in word length was predicted to decrease the likelihood of affective impairment and to increase the likelihood of affective enhancement. Valence of affect. One potential divergence between the defense/vigilance hypothesis and the amplification model concerns the effect of the valence (positive or negative) of affect. Unlike the amplification model, the defense/vigilance hypothesis has never been explicit enough to advance clear-cut predictions for affective enhancement and impairment. Yet, it would seem to predict that the processing is either prohibited (the defense) or enhanced (the vigilance) if and only if "anxiety" (or, equivalently, "psychodynamic conflict") is evoked. Since "anxiety" is more closely linked with negative than positive affect, the perceptual influence of affect should be obtained primarily with negative affective words. In contrast, the amplification model is non-committal in this regard. It is possible that attention is amplified once the significance or the interest value of an impinging stimulus has been detected. If this is the case, the perceptual influence of affect need not depend on the valence (positive or negative) of the affect; for the significance or interest value can be signalled by any affect either positive or negative, In virtually all the past studies that examined accuracy in perception independently of response bias for or against affective stimuli, only taboo (mostly affectively negative) words were used (see Kitayama, 1990, for a review). The exclusive use of taboo words was justified on the supposition

210

Chapter 5

that Freudian defense should mediate the perceptual influence of affect (e.g., Blum, 1954, Erdelyi, 1974; McGinnies, 1949), but ironically made it impossible to test the supposition itself. A few studies, however, examined both positive and negative affective words while controlling for response bias. It was found that the perceptual influence of affect was mostly identical whether affect was positive or negative (Broadbent & Gregory, 1968; Kitayama, 1990, 1991). This evidence is consistent with the amplification model, but raises some doubt on the defense/vigilance hypothesis.

The present experiments In an attempt to further test the implications of the amplification model, two experiments were conducted. As noted above, initial support for the model was obtained in a review of past studies that differed in the extent to which a valid expectation was available to subjects. It was hypothesized that a valid expectation should activate the relevant code prior to the presentation of a target stimulus, thus alleviating the difficulty in locating the code in perceptual identification. The expectation, therefore, should increase the likelihood of affective enhancement and decrease the likelihood of affective impairment. So far, however, only a few studies have actually manipulated expectation (Freeman, 1954; Laccy ct al., 1953; Kitayama, 1990; Postman ct al., 1953). Although these studies supported the predictions of the amplification model, they were not conclusive. Freeman (1954), Laccy ct al. (1953), and Postman et al. (1953) measured recognition threshold, so response bias may in part account for their fmdings. Although response bias was controlled in the Kitayama (1990) experiment, only a small number of words (12 in total) were tested, leaving open the generality of the fmdings. Thus, the effects of expectation was further examined in the present experiments. Another variable tested was word length. According to current models of word processing (e.g., McClelland & Rumclhart, 1981), an increase in word length should impose more constraints on the identity of the word and, thus, conduce to unequivocal activation of the relevant perceptual code. Thus, affcctive enhancement should be more likely and impairment less likely with an increase of word length. Only Kitayama (1991) studied the effect of this variable, and failed to find any evidence. To test the conjecture that this failure was due to the highly degraded input, the current series of experiments employed a pattern-masking procedure, whereby a target stimulus was presented with a relatively high stimulus contrast, but was immediately

S. Kitayama

211

followed by a pattern mask. In these experiments word frequency was held constant at the level comparable to the low-frequency condition of the Kitayama (1990) experiment (the level also examined in the Kitayama (1991) study) to maximize the comparability across different studies. Finally, both positive and negative affective words were tested to determine whether the perceptual influence of affect depends on its valence. Predictions. We predicted that both expectation and word length would increase the likelihood of affective enhancement and decrease the likelihood of affective impairment. Thus, our first two predictions were: (1) Affective impairment will be most likely when short words are used and no expectation is provided. (2) Affective enhancement will be most likely when long words are tested and an expectation is provided. It was not certain exactly how expectation and word length would jointly operate. According to the amplification model, in order for these variables to have additive impacts on the likelihood of affective enhancement or impairment, two conditions must be met. First, expectation and word length must additively increase the activation of a target perceptual code relative to the activation of other irrelevant codes. Second, the relative increase of the activation of the target code must linearly increase the likelihood of affective enhancement (or decrease the likelihood of affective impairment). Neither assumption has been explicitly tested in the literature. Thus, no a priori prediction could be made regarding whether the two variables would interact or have additive effects. Thus, our third prediction was: (3) The influence of affect in the remaining two conditions (i.e., short words/expected, long words/unexpected) will fall somewhere between the above two extremes (short words/unexpected, and long words/expected); in other words, the influence of affect in the former conditions will be no greater than that in the latter. In addition to their hypothesized role to improve attentional engagement and thereby to increase the likelihood of affective enhancement and to reduce that of affective impairment, there are some suggestions in the literature that both expectation and word length may have some extraneous effects on forced choice performance. To begin with, it has been demonstrated that explicit

212

Chapter 5

formulation of a "hypothesis" or expectation can impair perceptual identification (e.g., Bruner & Potter, 1964; Lawrence & Coles, 1954). Perhaps, a clear expectation may direct one's attention to only one type of information that is potentially available (e.g., orthographic, phonemic, semantic, etc., in perceptual identification). Information on the ignored dimensions may then be unavailable in response selection, leading to poorer performance in the presence of an expectation than in its absence. Note that this restrictive effect of expectation should occur equally regardless of stimulus affect. Similarly, word length is also likely to depreciate overall performance independently of stimulus affect. In a forced choice task tested in the present research, the respondent scans the percept and compares it with available alternatives. The longer the word, the greater the number of features (e.g., individual letters) that must be matched and, therefore, the more difficult response selection should be. Furthermore, as we shall show below (p. 219), this difficulty in response selection for longer words may be exacerbated by the fact that any given pair of long words tend to share a greater number of common letters than a pair of short words. All in all, as word length increases, response selection will be more difficult and, further, this effect of word length on response selection will occur regardless of word affect. In sum, we hypothesized that both expectation and word length would depreciate overall performance in perceptual identification, while simultaneously increasing the likelihood of affeetive enhancement and decrease that of affeetive impairment. Taken together, we predicted a general decline of perceptual identification with expectation and word length, and further expected this decline of performance to be significantly less for affective words than for neutral words. Notice that this latter prediction amounts to the three predictions stated above.

Experiment 1 Method Overview and subjects. There were 128 trials, divided into two blocks, differing in the length of the words (long versus short). The order of the two blocks was counter-balanced over subjects. On each trial subjects were exposed to a 33 ms flash of either an affectively positive, negative, or neutral target word, immediately followed by a pattern mask (a string of "&"s of the same length as the word). They were then presented with the target word and an equivalently valenced word of the same length, and asked to choose the

S. Kitayama

213

one presented. Whereas word length was a within-subject variable, expectation was a between-subject variable. Thus, for half the subjects the two choices were presented immediately before the target was flashed to create an expectation, but for the other half there was no such expectation. Seventy undergraduates at the University of Oregon (both males and females) participated in the experiment to partially fulfill their introductory psychology course requirements. All the subjects claimed to be a native English speaker. Materials. The one-hundred twenty-eight words used in the present experiment are listed in Appendix A. There were an approximately equal number of affectively positive, negative, and neutral words. Thirty three undergraduates who did not participate in the present experiment were asked to judge the affective quality of each word (1 = unpleasant, 5 = pleasant). The positive words were rated as more positive (M = 4.3, s = .23) than the neutral words (M = 3.1, s = .21), which in turn were rated to be more positive than the negative words (M = 1.8, s = .30). About half of the words of each affect type were long (more than nine letters long) and half were short (fewer than six letters long). Frequency of occurrence ranged from 8 to 65 appearances per million words, as determined by Kucera and Francis's (1967) norms. The mean frequencies of occurrence for the six word categories (3 affective types • 2 length types) were practically identical (varying from 27 to 37 occurrences per million). This frequency range roughly corresponded to the low-frequency condition of the Kitayama (1990) experiment. Sixty-four pairs were formed between words with equivalent valences and lengths, as shown in Appendix A. Each word served as target once and as nontarget once, resulting in 128 experimental trials. The order of these trials were randomized within each block for each subject. All stimulus words were shown in uppercase letters. Equipment. The experiment was controlled by an AMDEK-286 personal computer with an AMDEK-132 VGA adapter. Stimuli were presented on an AMDEK-732 color graphic monitor. Both the brightness and contrast of the screen were kept maximal. To the monitor was attached a translucent tube. The inner contour of the tube was 17 cm • 23.5 cm, and the length, 60 cm. One end of the tube was fitted to the monitor, and subjects watched the screen through the other end. Target words were presented at the center of the monitor. The height of the words was approximately 4 mm. The length of a five-letter word was 12 mm, and that of a nine-letter word was 23 mm, resulting in the visual angle of approximately 1.15 ~ and 2.06 ~ for the short and the long words, respectively. The experiment was controlled by the MEL (Micro Experimental Laboratory) system developed by Schneider (1988).

214

Chapter 5

This system synchronizes the presentation of stimuli with the refresh cycle of the monitor, enabling accurate control over the exposure durations. The experiment was conducted under normal room illumination. With the VGA facility, it is possible to create 62 shades of gray in addition to white and black. In a black background (luminance = .67 ft-L [foot-Lambert]), a target was shown in the 23rd shade"of gray (1.02 ft-L). This level of stimulus contrast was considerably higher than the level employed in the Kitayama (1991) study (15th shade of gray; .72 It-L). A fixation point, a mask, an expectation, and response choices were shown in the 32nd shade of gray (2.65 i~-L). Procedure. Subjects were tested individually. Upon arrival, they were randomly assigned to one of the four conditions representing the presence or the absence of an expectation and the order of the short-word and the longword blocks. They were instructed to look into the tube, and to place their left and fight index fingers respectively on the Z and the M keys of the computer keyboard. The subjects were told that the experiment was concerned with perception of briefly shown words. After the procedure was described (see below), they were given the following instruction: "As I mentioned to you, words are presented very briefly. We want to know how accurately people can recognize a word under such impoverished viewing conditions. So, the p r ~ u r e is set up so that you cannot perfectly see the word, yet you can still recognize some fragments or parts of the word. For instance, you may be able to recognize a letter or two, or even a small part of some letter. Or you may be able to recognize the contour of the word. Such partial information has proved very useful in performing this task. I will explain to you exactly what can be le~meA from this sort of experiments later. For the time being, even though you might occasionally feel that you are merely guessing, don't be discouraged or disturbed by this. Instead, try to pick up as many physical cues from the flash as possible. In this way your responses will be most accurate. Even when you don't think that you have enough information to make a choice, give us your very best guess. Please never use any intuition or gut-feeling in making judgment; once you do this, your performance cannot be better than chance. From past research we know that it is essential that you try to pick up physical features such as letter segments and overall contour if you are to perform this task at a better-thanchance level."

S. Kitayama

215

In the expectation-present condition, each trial started with the presentation of a pair of words with one to the fight of the center of the screen and the other to the left. After 2 see, the pair was replaced by a fixation point at the center of the screen. The subjects in this condition had been told that one of the two words would be a target in the upcoming trial and that it Was important to keep the pair in mind in order to perform well in the task. In the expectation-absent condition the trial began with the presentation of a fixation point. The two expectation conditions were otherwise identical. Thus, in both conditions, when the subject simultaneously pressed the two response keys, the fixation point disappeared and, 200 ms later, a target word was presented for 33 ms, immediately followed by a pattern mask. A sequence of "&"s of the same length as the target word served as a mask. The mask was presented for 1500 ms. Immediately after the disappearance of the mask, the corresponding word pair (the one shown at the beginning of the trial in the expectation-present condition) was shown. The subjects had been instructed to press either the Z key (with the left index finger) if the word on the left side of the fixation point had been presented, or the M key (with the right index finger) if the word on the right side had been presented. Response time was measured from the onset of the choice pair in milliseconds. The next trial automatically started after an interval whose length varied randomly between 2 and 7 see with the average of 4.5 sec. At the completion of the first block, the subjects were given a short break prior to the second block. The experimental trials were preceded by 26 practice trials. On the practice trials both affective and neutral words not used in the experimental trials were presented. On the first practice trial the target word was shown for 200 ms. On each of the subsequent practice trials, the exposure time was gradually reduced so that by the 21st trial, it was set at the exposure level used in the experimental trials, i.e., 33 ms.

Results and discussion Two dependent variables were examined. First, the percentage of correct choices (hit rate or the accuracy score) was computed for each condition. In the present procedure the subjects chose between two affectively equivalent stimuli. Hence, there was no room for response bias for or against affective stimuli to come into play, and the accuracy score can be safely taken as an unbiased index of the perceptibility of the target (Natsoulas, 1965). Second, the time required to make a choice was analyzed. Because few existing studies in this area reported response time, it was not clear whether the

216

Chapter 5

perceptibility under the present conditions could be reliably and adequately indexed by response time. It was considered reasonable, however, that as the perceptibility of a target word increased and the accuracy score increased, response time would decrease. Accuracy. Two separate analyses were pefformexl on the accuracy scores. First, mean accuracy scores were calculated separately for each subject and then submitted to. an Analysis of Variance (ANOVA) with subjects as a random variable. A significant F statistic obtained in this analysis (designated as F1) indicates the gcneralizability of the effect across subjects. Second, mean accuracy scores were computed separately for each word pair and thon submitted to an ANOVA with word pairs as a random variable. A significant F statistic from this analysis (designated as F2) indicates the generalizability of the effect across word pairs. Clark (1973) has recommended the use of mmF', which indicates the extent to which the effect can be generalizeA simultaneously over both subjects and word pairs. A mmF' can be computed from the corresponding F 1 and F 2 according to the formula, F 1,,F2/(FI+F2). Some have argued, however, that this statistic is too conservative, prone to Type II errors (Wike & Church, 1976); both F 1 and F 2 will have to be quite large before the corresponding mmF' can approach statistical significance. In the present paper, thus, F 1 and F 2 are used to gauge the reliability of effects. MinF' will be reported only when it attains statistical significance. A preliminary analysis involving two between-subject variables (expectation and the order of long- versus short-word blocks) and two withinsubject variables (word affect and word length) showed no effect of block order (Fs < 1). This variable, thus, was dropped in subsequent analysis. The relevant mean accuracy scores are given in the first half of Table 2. Note, first, that in all the conditions the accuracies for positive and negative affective words were virtually identical, lending support to the prediction that the influence of stimulus affect on immediate conscious perception should not depend on the valence of the affect. Consistent with the amplification model, yet somewhat contrary to theories based on the notion of Freudian defense (e.g., Blum, 1954; Erdelyi, 1974; McGinnies, 1949), this apparent irrelevance of the valence (positive or negative) indicates that attention was amplified once the significance or the interest value of an impinging stimulus had been detected. Because the effect of the valence was negligible, the two valence categories were subsequently combined.

S. Ki tayama

217

Table 2. Means (and standard deviations) of accuracy and response time in perceptual identification as a function of word affect, word length, and expectation (Experiment 1). Expectation Word

Present (N=34)

Length

Absent (N=36) Word Affect

Positive

Negative

Neutral

Positive

Negative

Neutral

Accuracy Long Words

.63 (.11)

.63 (.14)

.57 (.12)

.60 (.13)

.57 (.14)

.59 (.09)

Short Words

.61 (.16)

.61 (.14)

.67 (.10)

.63 (.15)

.57 (.15)

.67 (.15)

Response Time (ms) Long Words

827 (363)

799 (326)

890 (431)

1760 (571)

1792 (740)

1684 (571)

Short Words

762 (397)

862 (458)

719 (402)

1651 (481)

1645 (490)

1571 (431)

A 2 x 2 x 2 (expectation, word affect, and word length) ANOVA showed a significant main effect for word length (/71(1,68)=22.5 , p