Network researchers have been concerned with evaluating the accuracy of individuals' descriptions of their personal networks for many years. This paper examines the problem of "forgetfulness" and the extent to which it influences the measurement of ego-centric or personal social networks over time. The data come from the Indianapolis Network Mental Health Study (INMHS), a longitudinal study of the social networks of a cohort of individuals newly diagnosed with mental health problems. We focus on 114 people who completed two or more waves of the INMHS and explore the frequency that names mentioned in prior waves were forgotten. The results indicate that the membership of the respondents' networks changed significantly over the three waves; however, less than 5 percent of the observed change was due to respondents forgetting ties mentioned in prior waves.  The vast majority of changes observed in their networks reflected "true" change in the composition of the respondents' social networks.  Overall, the results suggest that people with mental health problems can provide descriptions of the changes in their social networks that are reliable and relatively free of recall bias. The implications of these findings for measuring personal networks over time are discussed.