Effects of information content and evaluative extremity on positivity and negativity biases.

This article extends models that assume that the integration of mixed (positive and negative) information results in a negativity bias in the morality domain but a positivity bias in the competence domain. Using functional analysis, this study predicted a positivity bias for moderately evaluated information and a negativity bias for extremely evaluated information. Content domain (competence vs. morality) and evaluative extremity produced main effects in a 2 x 2 experiment where 108 Ss were provided with descriptions of negative and positive behaviors of fictitious targets. As expected, the negativity effect in the morality domain was strong for extremely evaluated information but weak or nonexistent for moderately evaluated information. In contrast, the positivity effect in the competence domain was stronger for moderately than for extremely evaluated information. We suggest that positive and negative evaluations serve as approach and avoidance cues in interpersonal perception. The border between good and evil is located within rather than between persons: Usually, the same person shows both positive and negative attributes. An exception may be characters in the trivial literature, but this literature is considered trivial exactly because it is populated with black and white characters. According to the well-known averaging model of information integration (N. H. Anderson, 1981), observers' impressions of a person should be neutral if this person shows the same number of positive and negative attributes, each to the same extent. Numerous studies reveal, however, that even if the relative number of positive and negative attributes is controlled and balanced, negative impressions tend to emerge (e.g., Birnbaum, 1973; Fiske, 1980; Oden & Anderson, 1971; Reeder & Coovert, 1986; Reeder & Spores, 1983; Roskey & Birnbaum, 1974; Van der Pligt & Eiser, 1980; Warr, 1974). This suggests that negative information, even if equal in extremity to positive information, is given more weight in information integration, thus yielding a negativity effect. This effect has been explained in terms of the relative novelty of negative information (Fiske, 1980), its lower ambiguity (Birnbaum, 1972; Wyer, 1973), or its greater discrepancy with the general positivity of the typical or average person (Simpson & Ostrom, 1976). According to Skowronski and Carl