For the past 30 years, the study ofaccuracy in person perception has been a neglected topic in social and personality psychology. Research in this area was stopped by a critique of global accuracy scores by Cronbach and Gage. They argued that accuracy should be measured in terms of components. Currently, interest in the topic of accuracy is rekindling. This interest is motivated, in part, by a reaction to the bias literature. We argue that modern accuracy research should (a) focus on measuring when and how people are accurate and not on who is accurate, (b) use each person as both judge and target, and (c) partition accuracy into components. The social relations model (Kenny & La Voie, 1984) can be used as a paradigm to meet these requirements. According to this model, there are four types of accuracy, only two of which are generally conceptually interesting. The first, cared individual accuracy, measures the degree to which people's judgments of an individual correspond to how that individual tends to behave across interaction partners. The second, called dyadic accuracy, measures the degree to which people can uniquely judge how a specific individual will behave with them. We present an example that shows high levels of individual accuracy and lower levels ofdyadic accuracy. The topic of accuracy in interpersonal perception is a fundamental issue in social and personality psychology. In this article, we present a new and integrative approach. We begin with a historical review of the topic and the Cronbach and Gage critique of global accuracy scores. We then propose that accuracy research should be nomothetic, interpersonal, and componential. Finally, we show how the social relations model fulfills these requirements and so provides a methodology to study interpersonal accuracy. Historical Survey Accuracy in person perception is one of the oldest topics in social and personality psychology. The roots of this research lie in the success of standardized intelligence testing. Researchers reasoned that if it was possible to measure individual differences in cognitive skills, it should be possible to measure individual differences in social skills. Psychologists rushed to the task of measuring individual differences in accuracy in person perception. Whether it was called accuracy, empathy, social skills, Understanding, or sensitivity, the goal was always the same: to differentiate people in their ability to know the social world surrounding them. The individual-difference orientation fostered during World This research was supported in part by Grant R01 MH402950 ! from the National Institute of Mental Health. We thank Ronald Anderson, who generously provided us with his data. Also, Bella M. DePaulo, Larry Galpert, Judith Hall, Doug Kenrick, Brewster Smith, and William Swarm provided us with valuable comments on an earlier draft. Finally, Raymond R. Reno assisted us in the data analysis.

This research focused on the target effect on a perceiver's judgments of personality when the perceiver and the target are unacquainted. The perceiver was given no opportunity to interact with the target, a condition we refer to as zero acquaintance. We reasoned that in order to make personality judgments, perceivers would use the information available to them (physical appearance). Consensus in personality judgments would result, then, from shared stereotypes about particular physical appearance characteristics. Results from three separate studies with 259 subjects supported this hypothesis. On two of the five dimensions (extraversion and conscientiousness) on which subjects rated each other, a significant proportion of variance was due to the stimulus target. Consensus on judgments of extraversion appears to have been largely mediated by judgments of physical attractiveness. Across the three studies there was also evidence that the consensus in judgments on these two dimensions had some validity, in that they correlated with self-judgments on those two dimensions. If individuals were asked to make judgments about the personality characteristics of individuals with whom they were unacquainted, how much consensus would there be? That is, to what extent would perceivers agree as to where each target stands on a given trait? With no behavioral information on which to make judgments, intuitively we might speculate that consensus should be near zero. If, however, perceivers have access to the physical characteristics of the strangers they are judging, consensus may result from the use of shared stereotypes regarding the personality concomitants of these cues. Although there are studies that have measured consensus between acquainted individuals (

Consensus refers to the extent to which judges agree in their ratings of a common target. Consensus has been an important area of research in social and personality psychology. In this article, generalizability theory is used to develop a percentage of total variance measure of consensus. This measure is used to review the level of consensus across 32 studies by considering the role of acquaintance level and trait dimension. The review indicates that consensus correlations ranged from zero to about .3, with higher levels of consensus for ratings of Extraversion. The studies do not provide evidence that consensus increases with increasing acquaintance, a counterintuitive result that can be accounted for by a theoretical model (D.A. Kenny, 1991, in press). Problems in the interpretation of longitudinal research are reviewed.

Consensus, self-other agreement, and meta-accuracy were studied within and across nonoverlapping social groups. Thirty-one target persons were judged on the Big Five factors by 9 informants: 3 family members, 3 friends, and 3 coworkers. Although well acquainted within groups, informants were unacquainted between groups. A social relations analysis conducted within each social group showed reliable consensus on the Big Five personality factors. A model specified to estimate the consistency of a target person's effect on perceptions by others across social groups showed weaker agreement across groups. That is, targets were perceived consensually within groups, but these consensual perceptions differed between groups. The data suggest that personality and identity are context specific; however, there was some evidence of agreement in perceptions across groups.

Interpersonal perception among well-acquainted individuals in a social context was studied. High acquaintance was expected to provide perceivers with a large sample of target behaviors across situations. In turn, memory for acquaintances should be organized by social group and personality characteristics, as predicted by the social context-personality index theory. Differentiation of the target's traits in memory should produce a target effect on perception that is stronger than the perceiver effect. Furthermore, evidence for accuracy, meta-accuracy, independence of self- and other-perception, and reciprocity of affect were anticipated. A social relations analysis of data from a multiple-interaction, reciprocal design was used to study these phenomena. At the individual level, analyses indicated that perceptions of targets were determined primarily by target characteristics and secondarily by perceiver construction of the judgment. Also, perceivers judged targets as targets judged themselves, and targets knew in general how perceivers viewed them. Self- and other-perceptions were largely independent. Surprisingly, we did not observe dyadic meta-accuracy.