In 3 studies, 40, 408, and 256 college undergraduates made evaluative ratings of hypothetical persons described by personality adjectives A, B, or A and B. 12 adjective pairs were rated in the 1st study, 6 in the 2nd, and 2 in the 3rd. The extent to which each adjective in an A-B pair implied the other (i.e., redundancy) was varied by selection of adjective pairs in the 1st 2 studies and by pretraining Ss in the 3rd study. Results from all 3 studies confirmed the predictions that the rating of an AB person would tend to be more extreme than the mean of the ratings of an A person and a B person, and that this tendency would be greater the less redundant A and B. (PsycINFO Database Record (c) 2016 APA, all rights reserved)