A combination of information integration methodology and measures of retention was used to investigate how subjects differentially attend to and weight information in judgmental tasks. Subjects were shown sets of test scores for hypothetical students and were asked to rate the performance of each student or predict each student's performance on a comprehensive final exam. Subjects given the prediction task showed a recency effect--their predictions were weighted more heavily by the second half of the scores than by the first--and tended to discount a deviant score. Subjects given the rating task showed a recency effect only when explicitly told to take into account trends in scores. The prediction group that showed the most discounting of deviant scores was also the group that had the highest rate of recall of deviant scores on tests of retention. Discounting thus appears to be an active process, rather than a lowered level of attention. (Author)