The present study was performed to investigate the effects of information redundancy on the integration of informations as the amount of informations increases. 60 Ss were randomly assigned to 2 experimental conditions: 30 Ss were assigned to High Redundancy condition (Hi-R), and 30 Ss to Low Redundancy condition (Lo-R). The degree of redundancy was manipulated by varying the similarity in meaning of adjectives involved in a set. Each S of both Hi-R and Lo-R rated their favorableness for persons described by sets of 2, 4, 6 and 8 adjectives along with a 21-point scale. Results obtained tend to provide support for the prediction that an increase in the degree of redundancy among adjectives would decrease the response polarity. Especially, this prediction was strongly supported when stimulus persons were presented by more neutral levels of favorableness of adjectives (M+ and M-). Furthermore, the set size effect that an increase in the amount of informations increases response polarity did not appear in responses for Hi-R, but was found in Lo-R. The remarkable results were that the effects of differences in redundancy increased with an increase in the amount of informations in responses of M+ sets. As on account of failure to interpret these results within the weighted averaging formulation, they were interpreted in light of the discounting hypothesis proposed by Anderson & Jacobson (1965).