Studies of raters' cognition in performance judgment tasks might reveal the causes of rating problems in performance evaluation (Landy & Farr, 1980). Several information-processing models of performance judgment have been derived from theories and research in cognitive and social cognitive psychology (Cooper, 1981; DeNisi, Cafferty, & Meglino, 1984; Ilgen & Feldman, 1983). These theoretical perspectives have emphasized the importance of a rater's cognitive complexity in performance judgment. Cognitive complexity is a construct that emerged from Kelly's (1955) theory of personal constructs. The theory of personal constructs assumes that the number of independent dimensions or constructs employed by an individual when perceiving an object, event, or people influence discrimination and integration of information, and thus affect an individual's judgment and behavior. Of particular interest has been the total number of dimensions or constructs a person uses in organizing perceptions. The greater the number of dimensions used, the more cognitively complex the person is. Cognitive complexity is commonly defined as a person's "disposition to view the person-objects in one special environment in a complex or differentiated manner" (Vannoy, 1965, p. 385), and is measured by a role construct repertory grid. The higher a person's level of cognitive complexity, the more differentiated system of dimensions the person would have for mentally processing the behavior of others (Bieri et al., 1966). This suggests that people's level of cognitive complexity may have a bearing on how they mentally process multidimensional information in a rating judgment task. The effect of cognitive complexity on performance appraisal was first examined by Schneier (1977). Schneier defined cognitive complexity as "the degree to which a person possesses the ability to perceive behavior in a multidimensional manner" (p. 541), and measured the variable using the role construct repertory (REP) grid. In his study, cognitively complex raters demonstrated less halo, were less lenient, and used a wider range on behaviorally anchored rating scales (BARS) than did the