Several longitudinal studies have been published showing high retest coefficients in self-report personality inventories. While the data suggest stability in personality over long periods of adulthood, critics have argued that individuals may be motivated to present themselves in consistent fashion to the investigator. It has been suggested that respondents may remember how they answered the items previously. Thus, memory and the desire for consistency may be responsible for the high stability coefficients typically observed. In the present study 85 individuals who had completed the California Test of Personality as university students in 1944 were asked to complete it again 25 years later. Respondents were instructed to remember how they had answered the inventory as students and to fill it out in that manner (Retrospective condition) in addition to responding as they were in the present (Self condition). Results indicated high retest correlations between the 1944 and 1969 Self conditions, whereas the correlations between the 1944 Self and 1969 Retrospective conditions were much lower. Subjects describing their present selves obtained scores more highly correlated with their 1944 scores than scores created by attempting to describe themselves as they had been 25 years previously. These data indicate that memory for responses to a personality inventory taken years before is inaccurate, and that it cannot account for the observed consistency in personality test scores.