Learning effects in information integration: Manipulation of cue validity in an impression formation task.

In Stage 1, Ss predicted whether the majority of students in a previous study liked or disliked each of a series of hypothetical persons described by pairs of personality trait adjectives. One trait within each pair was highly polarized on a social desirability dimension, and the other was highly polarized on an intellectual desirability dimension. In Stage 2, new person descriptions were presented and Ss either predicted likabihty ratings for the reference group or gave their own personal ratings. Responses on Stage 2 varied as a function of feedback mampulations on Stage 1. The effect of feedback was to increase the relative weight of the dimension associated with reinforcement on Stage 1. This effect was more pronounced for Ss predcting the ratings of the reference group on Stage 2 than for Ss making personal ratings. These results illustrate how the weighting of information in an integration task can be manipulated through reinforcement contingencies.