1992, 3, 82-84.
In The Person and the Situation, Ross and Nisbett seek to answer the question "What have we really learned from social psychology?" They offer their book as a "throwback to a golden age, a tribute to our intellectual forebears and as a 'stand tall and be proud' pep talk for our colleagues (p. xv)." They succeed splendidly on all these counts.
Ross and Nisbett also offer their book as an "olive branch and invitation to more fruitful intellectual dialogue with our friends in personality research (p. xv)." Their review of the debate over the cross-situational consistency of personality (chap. 4), their analysis of the layperson's metatheory of personality (chap. 5), and their insights into how the real world conspires to confirm the layperson's metatheory (chap. 6) all provide a fresh starting point for such a dialogue. Accordingly, I accept.
Six years after Walter Mischel (1968) resurrected the debate over the consistency of personality, I suggested that the stubborn dilemma that sustains this debate and accounts for its historical durability is "the sharp discrepancy between our intuitions, which tell us that individuals do in fact display pervasive cross-situational consistencies in their behavior, and the vast empirical literature, which tells us that they do not. Intuitions or research? One of them must be wrong" (Bem & Allen, 1974, pp. 507-508). After acknowledging the "well documented biases and illusions that plague our intuitions," I nevertheless argued that "in terms of the underlying logic and fidelity to reality our intuitions are right; the research, wrong" (Bem & Allen, 1974, p. 510). At first glance, Ross and Nisbett appear to agree:
Earlier in [our] careers, [we] seriously entertained the hypothesis that most of this seeming order was a kind of cognitive illusion. We believed that human beings are adept at seeing things as they believe them to be, at explaining away contradictions and, in particular, at perceiving people as more consistent than they really are. While we continue to believe that such biased processing of evidence plays an important role in perceptions of consistency, we now believe that the predictability of everyday life is, for the most part, real. Despite all the evidence we have seen from objective studies of behavioral consistency (described in Chapter 4), and despite all we know about cognitive illusions and lay shortcomings in behavioral prediction (described in Chapter 5), we continue to believe that our own social world is inhabited by people who behave quite differently from each other in ways that are, for the most part, quite consistent. (pp. 7, 145)
But Ross and Nisbett go on to argue (in a delightful chapter entitled "The Coherence of Everyday Social Experience") that even though our intuitions about behavioral predictability are right, they are right for the wrong reasons. For example, people who are predictable because of situational role constraints are often judged to be driven instead by their enduring personality dispositions. University bursars who, among other things, will not permit seniors to graduate until they have paid their library fines are universally judged by students to be dispositionally "tight assed."
Ross and Nisbett believe that the major reason we are right for the wrong reasons is that everyday experience provides us with a badly confounded experiment. In contrast to the laboratory, where individuals are randomly assigned to conditions and subjected to treatments that do not vary as a function of their own behavior, the real world permits individuals both to choose and to modify their own conditions. This produces a consistency and a predictability of behavior that lead us to conclude erroneously that our "naïve dispositionism," as Ross and Nisbett call it, has been confirmed.
For example, I am confident that both peer ratings and objective samplings of my behavior over time and circumstance would reveal that I am consistently and predictably uncombative. But as my family will readily confirm, I also assiduously avoid conflict and actively absent myself from potentially confrontational situations. Accordingly, the hypothesis that I am really or dispositionally uncombative (or combative) never gets a rigorously clean experimental test. Nobody is going to randomly assign me to the high conflict condition! And if I never behave combatively in real life, then any hypotheses about my "real" disposition are arguably metaphysical (or psychoanalytic). The trait of combativeness is here analogous to a potbelly: Because I hold my stomach in, I do not have a potbelly. My wife, however, erroneously believes that I do have a potbelly--which, incidentally, I happen to hold in.
As Ross and Nisbett observe, the confounding of personality and situations in everyday experience also accounts for the discrepancy between controlled laboratory studies, which fail to find nontrivial consistencies in behavior, and longitudinal studies, which manage to find quite respectable continuities across the life span. In one of our own studies, for example, we found that middle-class boys who had frequent temper tantrums when they were 8 to 10 years of age became adult men who, relative to their more even tempered peers, were significantly more irritable, moody, and undercontrolled; attained lower socioeconomic status; had more erratic job histories; and were more likely to be divorced (Caspi, Bem, & Elder, 1989; also discussed by Ross and Nisbett, p. 159).
These continuities did not spring directly from enduring dispositions but from the interaction of those dispositions with situations over time. For example, we found that ill tempered boys were more likely to quit school earlier than their even tempered peers--possibly by choice, possibly by provoking their own expulsion--and path analysis showed that this was the major factor responsible for their downward socioeconomic mobility. In other words, their lower socioeconomic status as adults was not a direct effect of their dispositional ill temperedness but a "cumulative consequence" of that disposition acting at an earlier time to channel them down a particular path.
Ross and Nisbett's analysis of the coherence of everyday social experience is both sensible and persuasive. Accordingly, I hereby concede their basic point that our intuitive faith in a dispositional conception of personality is based in part on data that are flawed by person-situation confounds.
But the very lucidity of Ross and Nisbett's analysis has enabled me to revisit some of my earlier thoughts on this issue with a clearer understanding (e.g., Bem, 1983). As a result, I am now prepared to reaffirm my conviction that "in terms of the underlying logic and fidelity to reality," our lay metatheory of personality is, in fact, right; that it is right for the right reasons; and that our naïve dispositionism is less naïve than it appears.
The Idiographic Approach to Personality
Early in the contemporary debate over the cross situational consistency of behavior, I reactivated Gordon Allport's (1937) objection to the nomothetic assumption underlying most personality research that a particular trait dimension or set of trait dimensions is universally applicable to all persons and that individual differences are to be identified with different locations on those dimensions (Bem & Allen, 1974).
For example, the famous Hartshorne and May (1928) study, which failed to find consistency in children's "honesty" across situations, assumed that an honesty-dishonesty dimension could be used to characterize all children in the sample and that the differences among the children could be specified in terms of their degree of honesty. But as Allport observed in commenting on the finding that lying and stealing were essentially uncorrelated in the study, one child may lie because he or she wishes to avoid hurting the feelings of the teacher, whereas another may steal pennies to buy social acceptance from his or her peers. For neither of these two children do the behaviors of lying and stealing constitute instances on an honesty-dishonesty dimension, a dimension that exists in the head of the investigator, not in the behavior of the children. Accordingly, the low correlations "prove only that children are not consistent in the same way, not that they are inconsistent with themselves" (Allport, 1937, p. 250).
The implication here is that if Hartshorne and May had assessed empathy and sensitivity rather than honesty across situations, the first child would have been judged to be cross-situationally consistent; if they had assessed acceptance-seeking, the second child would have been so judged. As I phrased it in 1974 (Bem & Allen, 1974), a study will yield the conclusion that a sample of individuals is inconsistent to the degree that their behaviors do not sort into the equivalence class that the investigator necessarily imposes when he or she selects the behaviors and situations to sample. The traditional inference of inconsistency is not an inference about individuals but a statement about a disagreement between a group of individuals and an investigator over which situations and behaviors ought to be classified into a common equivalence class. The more general epistemological point here is that consistency and inconsistency are not intrinsic properties of behaviors but judgments by an observer about the match between the behaviors and his or her conceptual categorization of those behaviors.
In 1974, I essentially joined Allport in arguing that an idiographic approach to personality would prove empirically superior to a nomothetic approach. A similar idiographic approach--albeit shorn of its allegiance to trait-like variables--is also advocated by Mischel himself (Cantor & Mischel, 1979, p. 43) and by Ross and Nisbett (p. 163). I believe that the layperson belongs to this distinguished club, that our lay metatheory of personality also embraces the basic wisdom of an idiographic approach.
On Being Right: The Layperson's Metatheory of Personality
Consider how we proceed when we seek to characterize a friend. We do not typically invoke some a priori set of fixed dimensions that we apply to everyone. Instead, we analyze the data first. We review our friend's behavior and then select descriptors that strike us as pertinent precisely because they seem to conform to the patterning of his or her behavior. If Dick always does his schoolwork early, always returns his library books on time, and is meticulous about his personal hygiene, we are likely to describe him as conscientious. On the other hand, if Jane always does her schoolwork early but never returns her library books or changes her underwear, we might well describe her as a totally dedicated student who has time for little else. The point is that we would not characterize Jane as inconsistently conscientious. The trait term conscientious would not occur to us in the first place because it fails to identify any salient feature of her overall behavior. In this way we artfully finesse the inconsistency problem at the outset. Jane will not embarrass our description of her with some unexpected act of nonacademic negligence or personal slovenliness.
What I am proposing, then, is that our lay dispositional conception of personality is not a nomothetic trait theory but a context-sensitive, idiographic type theory. To understand an individual's personality, we search among our internal schemas until we find (or construct) a recognizable prototype that corresponds to the overall cross-situational configuration of the person's behaviors (cf. Cantor & Mischel, 1979).
This process becomes even more apparent when the strategy stalls, when we actually do detect behaviors that we construe as inconsistent. Typically we take this as a sign that we do not yet understand the person; we have not yet solved the concept attainment problem before us. For example, most students are familiar with the instructor who seems stiff and unfriendly in the lecture hall but relaxed and friendly in less formal settings; this is a familiar and recognizable type, and students perceive no inconsistency here. In contrast, students are less prepared for my own peculiar behavioral pattern, expecting on the basis of my friendly lecturing style that their visit to my office will be an intimate encounter-group rap. They are, alas, invariably disappointed. But they are puzzled by my apparent inconsistency only until they realize that the pattern of my behavior corresponds to the prototype of the stage performer who is introverted or aloof in private life. At that point, the concept-attainment problem has been solved and their initial, provisional verdict of inconsistency evaporates.
The context-sensitivity of our intuitive approach to personality not only finesses the inconsistency problem but also automatically solves the problem of situational specificity that can so easily embarrass a nomothetic trait theory. By defining Jane as a dedicated-student-who-has-time-for-little-else type, we virtually specify the equivalence class of situations across which she will be consistently conscientious and the equivalence class of situations across which she will be consistently nonconscientious. And if our initial descriptive attempt encounters empirical difficulties (we learn that Jane always does her physics assignments early but her history assignments late), we can further modify or qualify the type description: Jane is a dedicated-science-student-who-has time-for-little-else type.
Admittedly, this promiscuous freedom to generate open-ended personality typologies "on the fly" does not make for great science. Our intuitive metatheory of personality will never win any prizes for elegance or parsimony. But it does have the virtue of being right.
On Being Right for the Right Reasons
I conceded earlier that everyday experience can illegitimately appear to confirm our dispositional conceptions of personality. But I also believe that everyday experience provides legitimate confirmation for the kind of idiographic metatheory just outlined.
In one of their book's most illuminating sections, "Making Sense of 'Consistency' Correlations," Ross and Nisbett analyze the degree of behavioral predictability that can be achieved when behaviors display the low cross situation correlations typically found in most controlled studies. Their analysis shows that even when the cross-situation correlation of a particular behavior is only .16,
knowing that a particular individual has displayed an extremely "high" response on even a single occasion makes it safe to conclude that the person's response on some other occasion is much more likely to be extremely high than extremely low. And knowing that an individual's mean response over a great many observations is extreme makes these differences in the relative likelihood of particularly high or particularly low future scores reach dramatic levels indeed. (p. 111)
This statistical conclusion is not of much practical utility if we were to pursue a nomothetic strategy of assessment because, by definition, most individuals will be nondistinctive on any given dimension. But our intuitive idiographic strategy both permits and prompts us to select precisely those dimensions on which the target individual is characteristically and cross situationally distinctive. Accordingly, we will be frequently rewarded with palpable evidence that our predictive utility has reached "dramatic levels indeed." Our lay metatheory is not only right, it is right for the right reasons.
On Being Wrong for the Wrong Reasons:
Luring the Idiographic Expert into the Nomothetic Trap
If laypersons are such natural personological experts, why do they do so badly on the hundreds of predictive tasks reviewed in this book? Because social psychologists like Ross, Nisbett, and company force them to perform unnatural nomothetic acts, that's why. The layperson is equipped to play the idiographic game, "given a person, specify the situations in which he or she will behave distinctively and what those distinctive behaviors will be." But the studies reviewed invariably require the layperson to play either the nomothetic game, "given a person, specify how he or she will behave in this particular situation," or the nomothetic personnel-selection game, "specify the person who will behave in a particular way in this particular situation." And what do these studies find? That laypersons are no better at these nomothetic tasks than professional personality psychologists and that laypersons think they are better at these tasks than they actually are--just like professional personality psychologists. From this, Ross and Nisbett conclude that the layperson's metatheory is wrong.
I don't think that's right.
Allport, G. W. (1937). Personality: A psychological interpretation. New York: Henry Holt.
Bem, D. J. (1983). Toward a response style theory of persons in situations. In R. A. Dienstbier & M. M. Page (Eds.), Nebraska Symposium on Motivation 1982: Personality--Current theory and research (Vol. 30, pp. 201-231). Lincoln: University of Nebraska Press.
Bem, D. J., & Allen, A. (1974). On predicting some of the people some of the time: The search for cross-situational consistencies in behavior. Psychological Review, 81, 506-520.
Cantor, N., & Mischel, W. (1979). Prototypes in person perception. In L. Berkowitz (Ed.), Advances in experimental social psychology (Vol. 12, pp. 3-52). New York: Academic.
Caspi, A., Bem, D. J., & Elder, G. H., Jr. (1989). Continuities and consequences of interactional styles across the life course. Journal of Personality, 57, 375-406.
Hartshorne, H., & May, M. A. (1928). Studies in the nature of character: I. Studies in deceit. New York: Macmillan.
Mischel, W. (1968). Personality and assessment. New York: Wiley.
Beginning of article
Other abstracts & articles available online
List of publications
Bem Home Page