A test and extension of Anderson and Hubert’s model of memory for personality impressions is reported. The model suggests that memory for impression is independent of memory for the information that produced the impression. S s were told either to remember sets of trait adjectives describing persons or to form impressions. Measures of liking for the stimulus persons, recall, and recognition were administered either immediately or after 48 hours. The results supported the basic model in that the two-day delay was accompanied by considerable forgetting of the traits but had no effect on the accuracy of impressions. The results extend the empirical base and generality of the model. Small but significant correlations between retention and impression accuracy suggested that the separate memory systems are not completely independent. Accuracy was correlated with retention to a greater extent for the retention-set S s than for the impression-set S s. This partially supports an extension of the model to include a consideration of the purpose for which information about persons is attended to.