Do older people tend to exaggerate their health problems? Or do they downplay them? Do such tendencies change as people age? Are they a function of cohort membership? Or are differences in health perceptions among elderly people due to differences in survivorship within the elderly cohorts? These questions are examined with longitudinal data from the Yale Health and Aging Project, a stratified probability sample of elderly persons in New Haven, Connecticut. Complete health assessments are available for 2,583 community residents 65 and older in 1982, and complete follow-up date are available for 1,319 respondents in 1988. Multivariate regressions of self-assessed health on concurrent 1982 and 1988 objective health status and longitudinal analysis of residual values from these equations show that: (a) older survey respondents gave disproportionately positive health assessments, and (b) processes of aging, selective survivorship, and cohort differences all appear to play a role in creating this pattern.