PERSONALITY PROCESSES AND INDIVIDUAL DIFFERENCES Intraindividual Stability in the Organization and Patterning of Behavior: Incorporating Psychological Situations Into the Idiographic Analysis of Personality Yuichi Shoda, Walter Mischel, and Jack C. Wright In nomothetic analyses, the cross-situational consistency of individual differences in social behavior, assessed in vivo in a camp setting, depended on the similarity in the psychological features of situations. As predicted by the social-cognitive theory of personality, idiographic analyses revealed that individuals were characterized by stable profiles of;/. . . then . . . , situation-behavior relationships that formed "behavioral signatures" of personality (e.g., he aggresses when warned by adults but complies when threatened by peers. Thus, the intraindividual organization of behavior variation across situations was enduring but discriminatively patterned, visible as distinctive profiles of situation-behavior relationships. Implications were examined for an idiographic reconceptualization of personality coherence and its behavioral expressions in relation to the psychological ingredients of situations.

Allport (1937) introduced the concept of idiographic analyses half a century ago, urging personologists to understand each individual deeply in terms of how that person functions, instead of just studying "the operations of a hypothetical 'average' mind" (p. 61). Nonetheless, the idiographic focus has been bypassed by mainstream personality psychology. Probably this neglect reflects not a lack of interest but an absence of appropriate methods and theory for studying individual functioning in ways that are objective and scientific rather than intuitive and clinical. In our view, understanding individual functioning requires identifying first the psychological situations that engage a particular person's characteristic personality processes and the dis-

Yuichi Shoda and Walter Mischel, Department of Psychology, Columbia University; Jack C. Wright, Department of Psychology, Brown University, and Wediko Children's Services, Boston, Massachusetts. Portions of the present results were presented in Yuichi Shoda's 1991 Society of Experimental Social Psychology Dissertation Award address, Columbus, Ohio, October 1991. Preparation of this article and the research was supported in part by Grants MH39349 and MH45994 to Walter Mischel from the National Institute of Mental Health. We thank the administration, staff, and children of Wediko Children's Services, whose cooperation made this research possible. We are especially grateful to Hugh Leichtman and Harry Parad, Wediko's directors, for their support and Mary Powers, Philip Fisher, and Cynthia Scott for their assistance in data collection. We thank Niall Bolger, Nancy Cantor, Daniel Cervone, Chi Yue Chiu, Ying Yi Hong, Kristi Lemm, and Monica Larrea Rodriguez for their valuable comments on drafts of this article. Correspondence concerning this article should be addressed to Yuichi Shoda or Walter Mischel, Department of Psychology, Columbia University, New York, New York 10027.

tinctive cognitions and affects that are experienced in them. Then, an individual's functioning should become visible in the distinctive or unique ways the person's behavior changes across situations, not just in its overall level or mean frequency. For example, a person may often behave in a warm and empathic way with her colleagues at work but almost always in a very critical manner with her family. Another person may show the opposite pattern, so that he is warm and empathic with his family but critical with his professional colleagues. If two people are similar in their behaviors averaged across situations but differ in the situations in which they display those behaviors, are these differences merely a reflection of momentary situational influences? Or do such differences reflect differences in enduring and meaningful aspects of their personality? These are the main questions addressed in this article. In social-cognitive theory,1 individual differences in patterns of behavior across situations reflect such underlying person variables as the individuals' encoding or construals of their experiences, and their expectations, values, goals and self-regulatory strategies (Mischel, 1973, 1990). These relatively enduring person variables within the individual interact with situational characteristics to generate stable but discriminative patterns of behavior. It is these "unique bundles or sets of temporally stable prototypic behaviors" (Mischel & Peake, 1982, p. 754), contextualized in relevant psychological situations, that constitute a locus in which personality coherence may be revealed (Mischel 1990, 1991; Shoda, 1990; Shoda, Mischel, & Wright, 1993a, 1993b; Wright & Mischel, 1987). A major goal of the present 1

In current usage, the terms social cognitive and cognitive social appear increasingly as essentially interchangeable descriptions of this general approach to personality and social behavior (e.g., Mischel, 1993).

Journal of Personality and Social Psychology, 1994, Vol. 67, No. 4, 674-687 Copyright 1994 by the American Psychological Association, Inc. OO22-3514/94/S3.0O

674

STABILITY OF INTRAINDIVIDUAL BEHAVIOR PATTERNING

research was to obtain empirical evidence relevant to the validity of this conception of intraindividual personality coherence. Our analysis of the organization of the individual's behavior is conditional or contextual in the sense that the fundamental unit of observation is not the unconditional probability of traitrelevant behavior (e.g., the tendency to be extraverted), but rather the conditional probability of a given type of behavior in given types of psychological conditions or situations (e.g., Patterson, 1982; Patterson & Reid, 1984; Shoda, Mischel, & Wright, 1989; Wright & Mischel, 1987, 1988). The traditional use of the unconditional probability of trait-relevant behavior is appropriate and useful for many purposes, such as selecting persons on the basis of their predicted future behaviors on average in unspecified situations. However, beyond identifying individual differences in average levels of behavior, we view the intraindividual variability of behavior itself as the behavioral phenomenon of interest. These patterns of if . . . then . . . relations link psychological situations to the person's relevant behaviors (e.g., Mischel, 1973). Responsive to the many calls for a shift from a variable-centered approach in personality research to one that is more person-centered (e.g., Carlson, 1971; John, 1990), our analysis is indeed person-centered and focuses on the within-person (intraindividual) organization of behavior and on personality coherence. However, in the search for intraindividual stability, we do not pursue the traditional configuration of global dispositions or behavioral tendencies that characterize a person (e.g., high in extraversion and low in conscientiousness). Instead, our approach is unique in focusing on how a given type of behavior (e.g., aggression) by an individual varies distinctively but predictably across different types of psychological situations (e.g., ifAheX,butifBheY). Stability in these patterns of if. . . then . . ., situation-behavior relations is predicted to the degree that there is stability in the underlying person variables, such as the individual's ways of construing the situation and his or her relevant goals, values, expectancies, and the like, as they are activated in the particular situation (Mischel, 1973, 1990). Suppose for example that one person encodes events like being teased or provoked by his peer as an offense that requires a response in kind, whereas he sees being warned by his supervisor as a situation in which he has to comply to avoid negative consequences. For another person, however, being teased or provoked by a peer is encoded as normal and acceptable bids for interaction, whereas warning by a supervisor is construed as a personal violation by an unaccepted authority. Such differences in the subjective meaning of different situations, reflecting stable differences in encoding (e.g., Dodge, 1986), may result in activation of different expectancies, goals, values, and other person variables. Behaviorally, the effects should be visible as distinctively patterned if. . . then . . . , situation-behavior configurations, expressed as stable profiles of behavior variability across situations that differ in their psychological meaning for the individual. In the present article, we examine empirical data on the stability of such intraindividual situation-behavior configurations to test this hypothesis. Direct evidence that such patterns of variability in behavior are stable and distinctive within individuals would allow one

675

to conceptualize them as behavioral signatures of personality, rather than as measurement error or as data contradictory to personality coherence. This calls for an idiographic analysis of behavioral coherence, using an extensive observation system to assess people's behaviors in natural social interactions over many occasions. Unlike many earlier studies of person-situation links that relied exclusively on self-reports, we required behavioral data that would allow us to examine individual differences in people's distinctive patterns of relating to these naturally occurring social situations. In our long-term research program, these patterns were obtained by systematic behavior observations in vivo over the course of a summer in a children's residential camp setting, Wediko Children's Services' summer program in Hillsboro, New Hampshire. The results yielded an extensive archival database that allows systematic analyses to identify personality coherence in behavior as it unfolds across naturalistic situations and over many occasions (e.g., Shoda, 1990; Shoda etal., 1989,1993a, 1993b). Within a given ecological setting, such as the Wediko camp, situations may be conceptualized at different levels and with alternative units (Mischel, 1991; Shoda, 1990; Shoda et al., 1993b). At one level are the nominal situations that have been operationalized in studies of behavioral consistency traditionally (e.g., Hartshorne& May, 1928;Newcomb, 1929; Mischel & Peake, 1982). Typically they were dictated by the structure of the particular ecology (the setting), rather than by their potential psychological impact on, and meaning for, the person or by the generalizability of the observations obtained within them. Usually these nominal situations are highly complex and contain a wide array of different psychological features (Shoda et al., 1993a). In a summer camp, such as Newcomb's (1929) Camp Wawokiye or the present site, Wediko, woodworking, for example, may be a nominal situation that contains such diverse interpersonal psychological events as being praised, frustrated, teased, and punished. Nominal situations such as woodworking in the camp also tend to limit generalizability to other life settings. Thus, individual differences in relation to a specific nominal situation, even if highly stable, necessarily would be of modest psychological interest beyond the setting. On the other hand, at a deeper level, situations may be defined to capture basic psychological features or ingredients that occur in many different nominal situations and settings. In that case, information about an individual's behavior tendencies in relation to them is potentially generalizable to other situations that also contain these features. The utility of analyzing behaviors in terms of their stable relations to particular psychological features, hinges mostly on how widely the features occur in diverse nominal situations and different ecological settings. Table 1 summarizes and illustrates the distinctions among ecological settings, nominal situations, psychological features, and the interpersonal situations used as units of situations in the present study. As the table indicates, such events as being teased, provoked, or threatened are embedded within different nominal situations and contain the salient psychological features that are encoded by individuals and that affect their behaviors dynamically in the stream of social interactions. Just as individuals' responses to particular medications can be understood more fundamentally by considering the specific active ingredi-

676

Y. SHODA, W. MISCHEL, AND J. WRIGHT

Table 1 Examples ofEcological Settings, Nominal Situations, Interpersonal Situations, and Psychological Features Setting

Nominal situations

Camp

School

Home

Interpersonal situations

Psychological features

Woodworking

When peer initiated positive contact When peer teased, provoked, or threatened When praised by an adult When warned by an adult When punished by an adult

peer, positive peer, negative adult, positive adult, negative adult, negative

Cabin meeting

When peer initiated positive contact When peer teased, provoked, or threatened When praised by an adult When warned by an adult When punished by an adult

peer, positive peer, negative adult, positive adult, negative adult, negative

Playground

When peer initiated positive contact When peer teased, provoked, or threatened When praised by an adult When warned by an adult When punished by an adult

peer, positive peer, negative adult, positive adult, negative adult, negative

Classroom

When peer initiated positive contact When peer teased, provoked, or threatened When praised by an adult When warned by an adult When punished by an adult

peer, positive peer, negative adult, positive adult, negative adult, negative

Mealtime

When peer initiated positive contact When peer teased, provoked, or threatened When praised by an adult When warned by an adult When punished by an adult

peer, positive peer, negative adult, positive adult, negative adult, negative

Watching TV

When peer initiated positive contact When peer teased, provoked, or threatened When praised by an adult When warned by an adult When punished by an adult

peer, positive peer, negative adult, positive adult, negative adult, negative

ents rather than the brand names, the social-cognitive analysis of situations focuses on the psychologically active features of situations. Whereas nominal situations (such as woodworking) tend to contain heterogeneous sets of psychological features, in the present study we focused on interpersonal situations, each of which contains a relatively more homogeneous, distinct set of psychological features (Shoda et al., 1993b), as the units of analysis. The challenge in this type of analysis is to capture those features that are encoded distinctively by perceivers and that activate other relevant cognitive social person variables (e.g., expectancies and values) in the mediating process. Individual differences in response to nominal situations, such as the daily activities within a camp, then, may be analyzed in terms of the person's stable cognitive, affective, and behavioral responses to the encoded "active," psychological features within the nominal situations (e.g., Mischel, 1973). These psychological features, in turn, may consist of combinations of even more specific features and may be analyzed in terms of their overlap and similarity. Focusing on interpersonal situations as the situation units of analysis embedded in their nominal situations within the eco-

logical setting of the research site, in this article we examine the consistency and stability of situation-behavior relations that characterize individuals. Guided by the social-cognitive approach to personality (e.g., Mischel, 1973, 1990; Shoda & Mischel, 1993), we pursued an idiographic strategy. Specifically, we focused on the intraindividual organization of behavior in terms of the specific patterns in which that behavior varied across interpersonal situations, examining the stability of this pattern over time within each individual. We hypothesized that there would be significant intraindividual stability in the distinctive pattern by which the person's behavior varied predictably across particular types of these situations, visible as intraindividually stable "profiles" of if. . . then . . . , situationbehavior relations. Second, we examined the implications of these hypothesized intraindividually stable profiles of situation-behavior relations for the nomothetic analysis of cross-situational consistency. Namely, we hypothesized that the same underlying processes that generate stable and distinctive intraindividual profiles of behavior variation across these interpersonal situations also should generate cross-situational consistency in behavior to the

STABILITY OF INTRAINDIVIDUAL BEHAVIOR PATTERNING

extent that the situations are similar in their psychologically active features. Then the degree of consistency in individual differences in behavior across different situations should be a function of the similarity in the psychological features that they shared. To test this hypothesis, we also examined a more traditional nomothetic aspect of behavior organization, focusing on the cross-situational consistency of individual differences examined separately for each type of behavior observed in each type of psychological situation.

Method

677

Subjects The subjects were the 84 children (60 boys and 24 girls) from the research program described above, who resided for 6 weeks in a summer camp residential setting (Wediko Children's Services) in New Hampshire. They ranged in age from 6 years 5 months to 13 years 2 months, with a mean of 10 years 2 months, and they resided in cabin groups of 6-10 same-sex peers similar in age. This population is characterized by significant social adjustment problems, particularly with inadequate prosocial behavior and aggressive behavior in the home or school environment, and most of the youngsters are from low-income families in the Boston area. Data dealing with other questions and results from the same children assessed in the same setting and summer in thefieldstudy are reported in Shoda et al. (1989, 1993a, 1993b) and Rodriguez et al. (1989).

Research Program and Design The present article reports aspects of a large-scalefieldresearch program conducted at Wediko Children's Services, a summer camp residential program in New Hampshire (Shoda, 1990). The general population and setting have been described previously (e.g., Rodriguez, Mischel, & Shoda, 1989; Wright & Mischel, 1987, 1988). In the research program on which this article draws, a total of 84 children (60 boys and 24 girls, mean age 10 years and 2 months) were observed for the entire duration of one summer session (6 weeks) in the camp. The exceptionally rich database that we collected in this research program yielded a unique data archive for the systematic analysis of social behavior as it unfolds over time and across settings. These data range from highly molecular (e.g., coding of videotaped social interactions in 10-s units), to more molar (e.g., hourly ratings of behavior), to relatively global (e.g., dispositional judgments by counselors at the end of the summer). The data were collected by a team of 77 adult observers (a total of 14 to 28 observers per child), with an average total of 167 hr of behavior observations for each child over the 6-week summer. Our design enabled studies of diverse facets of the organization and nature of the individual's distinctive and stable behavior patterns at various levels of specificity and depth. Specifically, at the level of nominal situations, we analyzed individual differences in the organization of social behavior, focusing on the role of competencies as the characteristic of the individuals and as a demand characteristic of the situations (Shoda et al., 1993a). At the level of interpersonal situations, we addressed the relationships between judgments of global dispositions and the behaviors of individuals who were good exemplars of dispositional categories (Shoda etal., 1993b). We designed the research program to include assessments of interpersonal situations with distinct sets of salient psychological features that are relevant, important, and consensually encoded by people in the population sampled. Five interpersonal situations were selected representing two of the most salient and observable psychological features identified in earlier research (Wright & Mischel, 1988), namely valence (positive vs. negative) of the interaction and type of person (adult counselor vs. child peer) involved in the interaction. Thefivesituations selected to represent each combination of the two psychological features are summarized in the third column of Table 1. Specifically, throughout the 6-week summer, within each hour of camp activity, observers recorded the frequency with which each of thesefivetypes of situations occurred and whether the child responded with any of thefivebehavior categories of interest, namely, verbal aggression (teased, provoked, or threatened); physical aggression (hit, pushed, physically harmed); whined or displayed babyish behavior; complied or gave in; and talked prosocially (Shoda, 1990; Shoda et al., 1989, 1993b). In the present research we tested the hypotheses concerning the nature and stability of the intraindividual organization of social behavior in relationship to these situations from both the idiographic and nomothetic perspectives.

Conditional Probabilities ofBehaviors Within Each Interpersonal Situation The conditional probabilities of behaviors were based on thefivetypes of behaviors described above, observed within each of thefivetypes of interpersonal situations shown in Table 1. We use the term situationbehavior relation to refer to each of the 25 i/[or when]. . . then . . . conditional probabilities of the five behaviors in response to the five psychological situations recorded (e.g., prosocial talk when teased; prosocial talk when faced with a positive peer contact). The conditional probability of physical aggression when praised by an adult, however, was 0 for all except one child and therefore had little variation across individuals. Thus, this condition-behavior relation was excluded from further analyses. To compute conditional probability of subjects' behavioral responses in each psychological situation reliably, it was necessary that the subjects encounter each situation repeatedly. To ensure adequate reliability and resolution in the conditional probabilities in the analyses reported below, we therefore required that the subject encounter each type of situation at least a total of six times. Individuals differed in the number of times they encountered each of thefivesituations we observed. Specifically, the mean frequencies of encountering each type of situation were as follows (SD shown in parentheses): Peer teased, provoked, or threatened, 10.3 (6.5); Adult warned the child, 42.9 (19.5); Adult gave the child time out, 22.8 (16.0); Peer initiated positive social contact, 39.8 (10.5); Adult praised the child verbally, 66.5 (14.6). We also used this subset to compute cross-situational consistency of conditional probabilities of behavioral responses so that all correlations reported in Table 3 are based on the same set of subjects regardless of the situation pairs involved.

Intraindividual Situation-Behavior Profiles The if. . . then . . . pattern with which a given type of behavior displayed by an individual varied across the situations constituted an intraindividual situation-behavior profile. For example, Figure 1 shows a situation-behavior profile for verbal aggression for one individual and depicts how the person's verbal aggressiveness varied over thefiveinterpersonal events shown along the horizontal axis. One may plot the absolute conditional probabilities of behavior across situations, but the variability of such a profile would reflect the differences among the situations in how people behave in them in general, constituting the normative level for each situation. For example, most people are more likely to display aggression when teased by peers than when praised by an adult. To plot the aspect of behavior variation that is distinctive for each person, therefore, we subtracted the normative (mean) profile observed in this sample of subjects from the individual's "raw" profiles, and the results were rescaled using as units the standard deviations of the behav-

678

Y. SHODA, W. MISCHEL, AND J. WRIGHT Child # 17 profile stability: r = 0.96

Child # 9 profile stability: r = 0.89

PEER TEASE ADULT WARN PEER APPROACH ADULT PRAISE ADULT PUNISH

PEER TEASE ADULT WARN PEER APPROACH ADULT PRAISE ADULT PUNISH

Child # 28 profile stability: r = 0.49

Child #48 profile stability: r = 0.11

PEER TEASE ADULT WARN PEER APPROACH ADULT PRAISE ADULT PUNISH

PEER TEASE ADULT WARN PEER APPROACH ADULT PRAISE ADULT PUNISH

Figure 1. Illustrative intraindividual profiles of verbal aggression acrossfivetypes of psychological situations. The two lines indicate the profiles based on two different, nonoverlapping samples of occasions in which the child encountered each type of psychological situation, shown as Time 1 (solid) and Time 2 (broken).

ior within each situation. In short, the distinctive profile of behavior variability across situations was identified for each individual by the pattern of standardized deviations from the normative pattern in terms of standard scores computed in each situation. Figure 1 illustrates such a profile. In the present analyses, we assessed the stability of these situation-behavior profiles within individuals to test the hypothesis that the distinctive ways in which an individual's behavior varies across situations constitutes an enduring aspect of personality, rather than merely reflectingfluctuationsdue to uncontrolled, random factors.

Indexing Profile Similarities to Determine Their Stability To index and statistically test the stability of the intraindividual, situation-behavior profiles, the total available observations of each type of situation were randomly divided to form two sets of observations. Within each set, for each subject, conditional probabilities of each type of behavior in each type of situation were computed. This procedure yielded for each person two intraindividual situation-behavior profiles observed on two different sets of occasions. The similarity of the "shapes" of the two profiles was then indexed by an ipsative correlation coefficient, computed within each individual separately, using the interpersonal situations as the units of analysis. For example, if a person's behavior profile for thefivesituations from one set of nominal situations (camp activities) was [ 1.0,0.5,-1.0, —0.5,

0.0], and the one from the other was [0.5, 0.0, -1.5, -1.0, -0.5], the correlation between the two would be +1.0. Note that the correlation essentially reflects the stability of the rank order among the interpersonal situations within the same individual in how the individual's behavior in each situation deviates from the respective norm in each situation. In this example, both profiles indicate that the standardized (i.e., relative to the situation norm) behavior probability was highest in the first situation, followed by the second, fifth, fourth, and then the third. The stability coefficient of +1.0 indicates that the intraindividual rank order of situations was preserved perfectly over time. If the profile from the second set was [-1.0, -0.5, 1.0, 0.5, 0.0], it would indicate a complete reversal of the intraindividual rank ordering of thefivesituations, and the profile stability correlation would be — 1.0.

Computing Mean Consistency Coefficients Within Versus Across Interpersonal Situations As noted in the introduction, in addition to the idiographic analysis of intraindividual profile stability we also pursued a more nomothetic route to test the cross-situational consistency of if. . .then. . .relationships separately for each type of behavior in relation to each type of interpersonal situation. Therefore, to test the hypothesis that the degree of cross-situational consistency in individual differences in behavior should be a function of the similarity in the psychological features that

STABILITY OF INTRAINDIVIDUAL BEHAVIOR PATTERNING they shared, we examined traditional, nomothetic, cross-situational consistency coefficients of individual differences in each behavior, computed for each distinct pair of interpersonal situations. To ensure sufficient samples of observations for each type of interpersonal situation, observations from different camp activities were pooled. Specifically, the 14 camp activities were randomly grouped to form two sets of 7 activities, and within each set, conditional probabilities of each type of behavior in each type of interpersonal situation were computed. The consistency of individual differences in these conditional probabilities was computed across the two sets. Different counselors led different activities, and behavior observations were made at the end of each activity on a computer-scored behavior tracking sheet by the counselor who led that activity. Because observations of behaviors in different activities were made by different counselors, consistency coefficients computed across them do not reflect possible links due to overlap in observers. To minimize chance associations present in any specific random grouping of the camp activities, this procedure was repeated 100 times, and the results were averaged using Fisher's r-to-z transformation. These average correlations should be more stable with a smaller standard error than is usually expected for a single correlation, because such averages are less subject to the sampling error associated with the specific random grouping of the 14 camp activities into two sets of 7. If the 100 iterations had been conducted in independent samples, then one should be able to compute the expected standard error of the mean coefficient following the central limit theorem. However, the 100 iterations do not constitute independent samples, and therefore one cannot compute estimates of the standard error by simply applying the theorem, which assumes independence of sampling. Therefore, we used the bootstrapping procedure, whose method and theoretical rationale are described elsewhere in detail (Diaconis & Efron, 1983; Efron, 1981, 1985; Efron ATibshirani, 1986). Briefly, this computation-intensive, nonparametric method of estimating standard errors involves drawing random samples from the obtained data pool and computing the statistics of interest in each random sample. The bootstrapping procedure calls for sampling with replacement; that is, after a subject is drawn from the pool, the chosen subject is replaced back to the pool so that she or he can be chosen again. Therefore, even though it requires forming random samples of the same size as the size of the obtained data pool from which they are drawn, the exact composition of the random sample varies, due to the fact Liat in each random sample some subjects are represented multiple times while some others are not chosen. The distribution across the random samples of the statistic of interest computed in each sample (which in the case of the present analysis is the mean consistency coefficient) provides an estimate of the sampling distribution. Specifically, the standard errors reported in Table 2 are based on 500 such random samples. Within each sample, 10, rather than 100, random groupings of camp activities were made because of the limitation of computer resources; thus, we obtained conservative overestimates of standard errors and p values.

Results Idiographic Analyses oflntraindividual Profile Stability Figure 1 presents examples of idiographic situation-behavior profiles for individuals, illustrating varying levels of profile stability for allfivesituations. Thefirstprofile in Figure 1 indicates that the pattern by which verbal aggression of Child 17 varied across thefivesituations was stable and distinctive. Specifically, this individual was more verbally aggressive than others when punished by an adult (standard score of over 2.0), but his level of aggression was lower than the average level when teased, provoked, or threatened by a peer (standard score of about 0.5); while in the other three situations his level of verbal aggression

679

was near average in each respective situation. The two lines indicate the profiles based on two different, nonoverlapping samples of occasions in which the child encountered each type of psychological situation, shown as Time 1 and Time 2. At both Time 1 and Time 2, his profile was characterized by the fact that his level of verbal aggression, relative to the level of the peers in each situation (i.e., the "norm"), was the highest when adults punished him and lowest when another child teased. Child 9, whose profile of verbal aggression is shown in the second panel of Figure 1, was most distinctively verbally aggressive when warned by adults and his overall profile indicated substantial stability. Child 28 (the third panel of Figure 1), on the other hand, was most distinctively verbally aggressive when peers approached him. The profile stability was relatively modest, however, because his profile with regard to the remaining psychological situations changed over time. The last panel of Figure 1, for Child 48, illustrates a case of low profile stability. His profile of verbal aggression was distinctive at Time 1 in that he was most verbally aggressive, relative to others, when praised by an adult, but this was no longer the case at Time 2. As these examples illustrate, children differed widely in the stability and nature of their intraindividual profiles for each type of behavior. To test the hypothesis that on the whole these patterns of behavior variation across situations constitute intraindividually stable profiles, rather than "error variance," the mean profile stabilities for each behavior were computed by averaging each subject's profile stability using Fisher's r-to-z transformation. Because the profile stability is computed ipsatively, we considered each child's observed stability coefficient as independently sampled from a distribution of profile stabilities and tested the statistical significance of the group mean stabilities by t tests as estimates of the sampling error using the standard deviations of the stability coefficients across individuals. The first section of Table 2 shows the mean stability coefficients of intraindividual situation-behavior profiles over all five interpersonal situations listed in Table 1. To provide a reliable assessment of the conditional probabilities of behavioral responses in relation to each situation, yet to retain a maximum number of subjects, we included all subjects as long as'they experienced each psychological situation included in a profile at least six times in the course of the 6-week summer, as indicated in the Method section. Of the total of 84 individuals in the sample, 53 encountered all five situations sufficiently to meet this criterion and thus were included in this analysis. On average, they had mean stability coefficients of. 19 (p < .05) for prosocial talk, .28 (p < .001) for whining, .41 (p < .001) for compliance, and .47 (p < .001) for verbal aggression in their intraindividua! profiles of behavior variability across allfivesituations. It was possible to include more individuals in the analysis by excluding from the profiles the less frequently encountered types of situations. Specifically, excluding the least frequent situation (when teased, provoked, or threatened by a peer), 73 people encountered each of the remaining four types of situations at least six times, and we used that subset of subjects to compute the stabilities of profiles of behavior variability over the four situations. As shown in the second section of Table 2, the average stability coefficients for these profiles were .28 (p