Information redundancy, inconsistency, and novelty and their role in impression formation

Abstract

An experiment was conducted to investigate the effects on evaluations of persons of three mathematically related variables: the redundancy of the information presented about these persons, the inconsistency of this information, and its novelty. Subjects were asked to estimate their liking for persons described by adjective pairs. The redundancy of the two adjectives in each pair (A and B) was estimated from the extent to which the conjunctive probability of their occurrence (Pab) exceeded the product of their unconditional probabilities of occurrence (Pa and Pb), that is, by the extent to which Pab > PaPb. The inconsistency of the adjectives in a pair was defined by the degree to which Pab < PaPb. The novelty of each adjective was assumed to be inversely related to its unconditional probability of occurrence. Evaluations based upon pairs of adjectives became less polarized as the redundancy of the adjectives in those pairs increased, but became more extreme as their novelty increased. The effect of inconsistency was curvilinear, suggesting that the less polarized adjectives were discounted at moderate levels of inconsistency, but that the entire description was treated with incredulity at extreme levels of inconsistency, offsetting the discounting effect. Results were interpreted in light of current research and theory and also in terms of the mathematical and conceptual similarities between the variables investigated.