Previous studies have found that educational differences in mortality are weaker among the elderly. In this study I examine whether either cohort or period effects may have influenced the interpretation of age effects. Six 10-year birth cohorts are followed over 30 years through decennial censuses. Differential survival is inferred from changes in the relative proportions of a cohort in each education category as the cohort ages. In cross-section, younger persons generally show stronger education effects on survival, although this pattern is clearer for women than for men. There is evidence of period effects. Within cohorts, relative survival tends to increase with age.