Traits and social stereotypes: Levels of categorization in person perception.

Abstract

A theoretical distinction is made between trait categorization in person perception and categoriza-tion by means of well-articulated, concrete social stereotypes. Three studies test the prediction thatsocial stereotypes are both more associatively rich and more distinctive than are trait-defined catego-ries. In Study I, subjects sorted adjectives related to extroversion and introversion. A cluster analysisusing similarity measures derived from the sorting indicated that distinct social stereotypes wereassociated with each trait. This supports and extends earlier findings (Cantor & Mischel, 1979). InStudy 2, subjects generated attributes of the trait categories and stereotypes that emerged in Study1. More nonredundant attributes, especially visible features, were listed for the stereotypes than forthe trait categories. Study 3 elicited the explicit associative structure of traits and related stereotypesby having subjects rate the association between a series of attributes (derived from the responses inStudy 2) and each category label. Results showed that social stereotypes have distinctive featuresthat are not shared with the related trait category, whereas trait categories share virtually all oftheir features with related stereotypes. The implications of the trait/stereotype distinction for socialinformation processing are discussed.Substantial research has shown that the layperson assignsother individuals to social categories by means of implicit theo-ries of personality and uses these categories to predict potentialbehaviors, emotional reactions, personality attributes, atti-tudes, and values (Cantor & Mischel, 1979; Rosenberg & Sed-lack, 1972; Schneider, 1973; Wegner& Vallacher, 1977). In thisresearch, we contrast two types of representations used in socialcategorization, both of which can be observed in everyday life.First, people frequently use trait concepts to describe the essen-tial qualities of others. They may characterize people as tough,emotional, bright, or boring, for example, and in so doing as-sume that these traits have predictive power. In a second formof social categorization, people may explicitly categorize othersby thinking of them as members of a larger group. Such groupsoften bear a specific, socially shared label, such as redneck ordo-gooder, and they may be based on a number of features, in-cluding occupations and belief systems, as well as personalitytraits. These concepts are essentially well-articulated social ste-reotypes that are associated with a variety of highly visible char-acteristics, such as physical features, typical overt behaviors,and demographic identifications, in addition to relatively invisi-ble characteristics such as beliefs and internal experiences.Trait terms, in principle, designate single attributes that vari-ous kinds of people may possess. We propose on this basis that,considered in isolation, a given trait concept should not be veryWe would like to thank Deanna Bernstein, *Valerie Cathcart, and Su-sanna Perkins for their help in conducting this research.Correspondence concerning this article should be addressed to SusanM. Andersen, who is now at the Department of Psychology, New TforkUniversity, 6 Washington Place, 4th Floor, New York, New York 10003,or to Roberta L. Klatzky, Department of Psychology, University of Cali-fornia, Santa Barbara, California 93106.predictive because the individuals with this attribute may varyin a number of predictively important ways. By contrast, a per-son category representing a group of people who share manyattributes, behaviors, and attitudes should be predictive of thosevery features. In other words, trait categories should be sparsean d nonpredictive, wherea s social stereotype shoul be infor-mative and rich.Our view of trait-defined categories and social stereotypescan be undersood within the context of network approaches toknowledge representation (Anderson & Bower; 1972; Collins Q Higgins, Rholes, & Jones, 1977; Srull & Wyer,1980; 'Wyer & Carlston, 1979). In a network model, conceptualrepresentations, or nodes, are connected to one another bymeans of associative links, including connections between attri-butes and objects (e.g., soft-bed). In this terminology, we sug-gest that traits are linked by attribute-objective links to a num-ber of distinct stereotypes (and to a number of distinct individu-als). By contrast, stereotypes have object-attribute links notonly to traits but to many other types of features, includingovert behaviors and physical characteristics. Retrieval of a ste-reotype provides access to these features for purposes of socialprediction. Retrieval of a trait, on the other hand, is less predic-tive, both because its connections to other attributes are indi-rect and because the stereotypes with which it is linked are di-verse.Support for the notion that social stereotypes carry the bur-den of social prediction comes from a number of studies outsideof the trait domain. For example, it has been suggested that ste-reotypes such as housewife, mother, or temptress are embeddedwithin the category defined by woman (Ashmore & DelBoca,1979; Deaux & Lewis, 1984) and that a category defined onlyby sex may be too inclusive and abstract to be truly effective insocial prediction (e.g., Deaux & Lewis, 1984; Hamilton, 1981;235