Abstract

A judgement analysis of people's social inferences of attitudes and ability was conducted. University students were asked to infer the liberalness ( N = 60; Study 1) or intelligence ( N = 40; Study 2) of targets seen in pictures. Multiple regression analyses revealed that attractiveness was the most important cue for predicting inferences of liberalness, while an ethnic cue (i.e., being Asian) was the most important cue for judgements about intelligence. Results also showed that a single-cue model was less susceptible to overfitting, but significantly less accurate than a multiple-cue model in predicting participant's intelligence judgements. Although the multiple regression models suffered a degree of overfitting, cross validation showed that they continued to have significant predictive value when applied to new data. Furthermore, a “random partner” method (comparing each participant's own regression equation with that of another, randomly selected, participant) provided evidence of significant idiosyncratic variation in the way intelligence judgements were made