I Historical Perspective
1 Personality and Cognition: Something Borrowed, Something New?
Walter Mischel
Stanford University
I hope that this volume is a step toward the increasing integration of orientations, insights, and methods from the subdisciplines of personality, cognitive, and social psychology. Like most (or all?) scientific developments, the progress reflected in the contributions that follow combines the new and old, a mix of innovation and recreation of elements whose historical roots may be discerned in earlier movements. In viewing these contributions, a proper evaluation of their originality lies somewhere between "déjà vu" and "revolutionary." To help arrive at a reasonable perspective, I begin by commenting on what I see as some of the highlights of the field of personality, especially as they may relate to some of the major themes in the present volume.
Searching for Traits
Historically, the field of personality has been influenced by a few main positions, of which one of the oldest and most enduringly influential is trait psychology. Beginning about the turn of the century, and inspired mostly by the success of the intelligence testing movement and especially by the work of Alfred Binet, psychologists became interested in seeing whether the success achieved with mental (i.e., "cognitive") measurement might be repeated if one tried to quantify social characteristics.
The trait theories of the 1920s and 1930s were primarily a psychology of common sense, in which the layperson's "natural," everyday theory of what people are like became the scientific theory of what people are like. People were held to differ on dimensions such as friendliness, honesty, conscientiousness, and aggressiveness. The scientific innovation—and it was a most important one—was the development of statistical methods to give dimension to these qualities systematically and to quantify them rigorously and carefully. The responses of the subject were primarily self-reports, sometimes ratings by others, represented by check marks on multiple choice or true-false inventories and questionnaires. Recall, for example, the Bell Adjustment Inventory, or the Woodworth Personal Data Sheet in World War I ("I am a good mixer"; "I am at ease with people"). These efforts began in a very reasonable way, intended as shortcuts for the actual sampling of behavior (for instance, the psychiatric interview).
I think a giant conceptual leap was taken at that time, perhaps without full realization of just how big a leap it was. This was the shift from sampling people's relevant specific performance by assessing what they could do (which was the heart of intelligence testing) to asking people to report about what they were like, in general, on such broad situation-free dimensions as friendliness, conscientiousness, or introversion. These responses were used not as samples of the respondents' relevant behavior but as signs or indicators of their generalized dispositions. This leap was not carefully thought through, and we have seen its implications over the last 10 or 20 years as we began to realize the conceptual and methodological problems that one gets into if one does not distinguish very carefully between people's subjective judgments about themselves and an objective sampling of what they actually do (their performance under specific circumstances). We see what people do, for example, when we ask them to count the beads or arrange the blocks or say the digits backward, or when we directly observe their behavior as it unfolds. Such specific behaviors are very different from how people characterize themselves globally as introverted, aggressive, or friendly, and the differences have serious measurement implications (Mischel, 1968).
Early trait psychology was a psychology guided mostly by words in the dictionary, as in Allport's (1937) search, yielding the hundreds of terms finally culled from an even larger number of adjectives. Other trait psychologists attempted to find ways to hone down this list more finely, using a variety of often ingenious techniques, most notably factor analysis. The struggle became an attempt to find a finite and, hopefully, relatively small taxonomy of the basic dimensions of personality and social behavior. This remains one of the main objectives of the field, with a distinctive methodology and, I think, with distinctive uses, contributions, and limitations.
The lone voice speaking out against this dimensionalization of personality with numbers was Gordon Allport, who, beginning in the 1930s, insisted on each person's uniqueness and individuality. Most who heard Gordon Allport were excited by the idiographic approach he espoused. But, curiously, his own work (Allport, Vernon, & Lindzey, 1960) inspired still more attempts to categorize people on slots or dimensions that were the preferred yardsticks of the psychologist rather than on the equivalence classes of the person being assessed. More forcefully than others before him, Allport (1937, p. 3) articulated the essence of the trait position, claiming that the individual is an "amazingly stable and selfcontained system" that contains enduring generalized dispositions. Our task, in his view, was to search for those stable broad dispositions as manifested behaviorally, and that has been the major mission of trait psychology for many decades.
How fruitful has this search been? In trying to answer this question it is important to discriminate clearly between demonstrations of temporal stability on the one hand and cross-situational generality or consistency on the other. More than a decade ago the available research provided evidence for significant temporal stability but also—and far more surprising at the time—for discriminativeness or "specificity" in how behavior varies across situations. Summarizing those findings, I concluded: "Although behavior patterns often may be stable, they usually are not highly generalized across situations [Mischel, 1968, p. 282]."
In the long debate on this topic no one seriously questions that lives have continuity and that we perceive ourselves and others as individuals who maintain a stable identity over time, even when our specific actions change across situations (Mischel, 1968, 1973, 1977). Although temporal stability in the patterning of individual lives, in self-perceptions, and in how others view us seems evident, there is room for serious disagreement about the nature, degree, and meaning of the "erratic and uneven" relationships typically found when cross-situational consistency is studied with objective measures of behavior. To be sure, this result may reflect measurement problems, as Block (1977) and Epstein (1979) suggest. Better measures will surely provide better support for the existence of meaningfully organized behavior patterns. But, in my view, better measures and more fine-grained analyses should also make it even more apparent that individuals discriminate among situations according to subjective equivalences. Individual behavior is organized in terms of these personal meanings, not those of the trait psychologist. Although sometimes the subject's equivalences will coincide with the nomothetic trait categories of the assessor, often they will not (Mischel & Peake, work in progress). Whether or not the degree of correspondence is judged adequate and useful depends on a host of considerations, and especially on one's purpose (Mischel, 1979).
Limitations of the Classical Approach
The impact of trait psychology has been limited seriously because human consistencies and psychological equivalences are more complex and cognitively constructed than nomothetic trait theory suggests. Traditionally, the search for consistent personality traits and types has assumed well-defined, distinct, and nonoverlapping categories in which each member of a personality category has all the defining features of that category. But although such neatness and equality may be built into artificial, logical taxonomies, it does not seem to characterize the categories of objects, people, and situations encountered in the natural world. As Wittgenstein (1953) first noted, the members of common everyday categories do not share all of a set of singly necessary and jointly sufficient features critical for category membership. When one examines a set of natural objects all labeled by one general term, one will not find a single set of features shared by all members of the category; instead, one finds a pattern of overlapping similarities, a family resemblance structure.
Many linguists, philosophers, and psychologists following Wittgenstein (Labov, 1973; Lakoff, 1972; Lehrer, 1970; Rosch, Mervis, Gray, Johnson, & Boyes-Braem, 1976; Smith, 1976; Tversky, 1977) now suggest that natural semantic categories are ' 'fuzzy sets" that violate the expectations of the classical all-or-none position: Natural categories are organized around prototypical examples of focal stimuli (the best examples of a concept), with less prototypical members forming a continuum away from the central prototypic exemplars (Rosch, 1975). Guided by this prototypicality view, and recognizing that person categories are "fuzzy sets," Nancy Cantor and I have been exploring the rules people use in making judgments about the clearest or most prototypical exemplars of such person categories as extraverts (Cantor, this volume; Cantor & Mischel, 1979). It is to be hoped that such studies of prototypicality rules and ' 'family resemblance'' principles will ultimately allow a better understanding of the nature of consistency and coherence extracted from variations in behavior. Someone who has seen a thousand Picassos, each of which is unique, can determine (usually quite easily) whether a previously unseen work is a true Picasso, an imitation, or the work of someone else. The same processes that allow this kind of pattern recognition—extraction of a central distinctive gist or unity from great diversity—must surely be basic for our recognition of identity and coherence in the face of behavioral variability.
Trait psychology has also been limited because its focus is not addressed at the flow of behavior nor in any way linked to the dynamic interactions that go on where people actually are living their lives. It has to do instead with the effort to taxonomize the world, to cut up human qualities and the stream of actions into neat dimensions, and to provide some kind of taxonomy which cuts people in their places—but on the psychologist's theoretically preferred favorite dimensions rather than on the person's. The result is a focus on between-person differences, between-person variance, rather than on within-person variance as it relates to environmental changes. These efforts have consumed much energy and yielded many interesting useful classification schemes (Guilford, 1975). Unfortunately these efforts do not address the processes underlying personality, tending rather to classify subgroups or people than to capture and explain the ongoing flow of their behavior.
Stimulated especially by the thoughtful writings of such personologists as Henry Murray (1938) and David McClelland (1951), some psychologists have long recognized these limitations and tried to move away from an exclusive person-focus to study the interactions of persons with their worlds in a more process-oriented fashion. Perhaps most influential in the discovery of the role of social contexts was Kurt Lewin's (1951) articulation of the field of forces operating in any given situation at the moment. Disappointment with the bundles of correlation coefficients yielded by traditional trait psychology added to the appeal of the experimental approach to the analysis of ' 'basic processes'' influencing attitudes and behavior espoused by the mainstream social psychology of Lewin's disciples (Festinger, 1957). Unfortunately, the personological concerns of this new breed of experimental social psychologists were generally limited to throwing in a few "personality measures" (often homemade and usually of unknown reliability) while searching for experimentally manipulated main effects in laboratory studies of interesting pieces of social behavior. For many social psychologists, individual differences at best became ancillary, minor data; at worst, they were error variance to be obviated by cleaner methodology and more potent independent variables. In the 1950s and 1960s, those who studied social processes experimentally thus tended to have little interest in personality, whereas those who were committed to studying persons found it most convenient to study them in relative isolation with the nonexperimental (i.e., correlational) methods. The rift between personality and social psychology widened, to the detriment of both endeavors.
The Psychodynamic Search for Underlying Dispositions
It is, I think, because of the static, descriptive, and nonanalytic quality of trait psychology that many of us were so excited when we first came across the psychodynamic approach. We found the writings of Freud and his followers to be an enormously stimulating, invigorating way of getting beyond labeling or groupings and into each person's unique psyche. For me, the appeal of the psychodynamic approach, when I first encountered it as a graduate student in the 1950s, was that at last there seemed to be a way of finding consistencies as they exist for the individual in a way that was not obvious at the level of surface behavior. These equivalencies were not documented by the kind of counting of specific behavioral similarities that characterized the work of Hartshorne and May (1928), who hoped to show that the child who is honest about not cheating when it comes to an arithmetic test is also honest about not cheating when -it comes to a spelling test. Instead, psychodynamic theorists asserted that the relationship between indicators, intrapsychic dispositions, and overt behavior may not be cumulative and additive, as the trait approach had assumed. Rather, the relationship between sign and disposition may be subtle, indirect, and contradictory and may involve all kinds of transformations.
Indeed, much of the excitement of the psychoanalytic approach was that it provided the notion of transformation—the notion that the same motive could manifest itself in all sorts of ways, displaying a wide range of vicissitudes. Thus Joe, who repeatedly says he is very aggressive on questionnaires, may not be more aggressive than Charles, who claims to be unaggressive and endorses mostly timid adjectives; indeed, the more Joe insists on his aggressiveness, the less aggressive and the more passive-dependent he might really prove to be, whereas Charles, the ostensibly passive one, may indirectly reveal himself to be a cauldron of stifled aggressive impulses. The hope was that recognition of underly...