Thirty subjects rated the likableness of sixteen hypothetical persons on the basis of their personality traits. The trait descriptions were the products of a 2 x 2 x 2 x 2 factorial design based upon set heterogeneity (homogeneous vs. heterogeneous), set-size (4 vs. 8 traits), trait valence (positive vs. negative), and 2 successive judgments. For such a stimulus design, the averaging model predicted a similar set-size effect for both the homogeneous and heterogeneous sets. Analyses of the data, however, indicated that the set-size effect of the form predicted by the averaging model was limited to the homogeneous sets only (p < .0004). There was not any evidence for such an effect with the heterogeneous sets (F = 0.88). It was also found that the homogeneous sets as compared to the heterogeneous ones produced more polar judgments at both levels of trait valence (p < .0003).