| Abstract: | This paper examines theories of impression formation which represent stimulus information as distributions on subjective dimensions rather than points along a continuum. These theories explain why unfavorable information often overrides favorable information. Subjects were asked to imagine hypothetical persons described by single adjectives or adjective combinations and to estimate the probability that each person would have various degrees of likeableness. Three models were considered to describe how the likeableness distribution for the adjective combination depends on the distributions of the single adjectives. All three models assume that the mean of the distribution for the adjective combination is described by the equal probability criterion which, for symmetric distributions, implies a weighted average of the single adjective means with weights inversely proportional to their standard deviations. The models can be distinguished on the basis of the standard deviations of the adjective combinations. Estimated standard deviations of the single unfavorable adjectives were smaller than those for the single favorable adjectives. Furthermore, the standard deviation associated with an adjective combination tended to fall between the standard deviations of the single adjectives, consistent with the horizontal averaging model. |
