本文以累積優勢論、老化槓桿論和世代重要論等三種視角，來剖析教育對憂鬱作用的跨生命歷程和跨世代模式。藉由年齡世代迴歸模型分析「台灣社會變遷基本調查：1990、1995、2000、2005」和潛在成長模型探究「台灣中老年身心社會生活狀況長期追蹤調查：1989、1993、1996、1999、2003」，本文發現：（1）「教育」和「教育與年齡交互項」皆與橫斷面的憂鬱以及縱貫面的憂鬱斜率有顯著負向關係。當教育對憂鬱的主效果和年齡交互效果皆為負向時，因教育而別的憂鬱差異隨生命歷程而擴大。此外，較低教育者的憂鬱在中年後攀升，但較高教育者的憂鬱隨年齡增加維持平緩，故累積優勢論獲得支持；（2）「教育」和「教育與世代交互項」與憂鬱有顯著負向關係。對年輕世代而言，教育對憂鬱的影響出現較早且程度更高，故符合世代重要論；（3）橫斷面分析契合老化槓桿論，但縱貫分析僅提供有限的支持，故老化槓桿論仍需商榷。

A growing body of literature has demonstrated that the impacts of education on physical health vary across the life course and across birth cohorts in Western societies. However, since aging effects and cohort effects are difficult to disentangle empirically, most previous studies have estimated one effect while ignoring the other. In addition, whether the temporal impacts of education can be extended to depression and whether they exist in non-Western societies are much less known. Thus, this study is conducted to elucidate the patterns between education and depression across the life course and across cohorts in Taiwan, by using age-cohort regression models on the sequential cross-sectional data pooled from the Taiwan Social Change Survey and using latent growth curve models on the longitudinal data obtained from the 1989– 2003 Survey of Health and Living Status of the Elderly in Taiwan. The main findings are: (1) both approaches consistently indicate a favorable direction of life-course relationship in depression for the well educated within the same cohort. Thus, the age-specific rates of change in depression differ across levels of education in a manner that progressively enlarges the gap in late life, supporting the cumulative advantage hypothesis; (2) the education-based difference in depression is larger and appears earlier for younger cohorts, which is consistent with the rising importance hypothesis; and (3) evidence for the age-as-lever hypothesis appears in the TSCS analyses, but is limited in the latent growth analyses obtained from the elderly panels. The cross-sectional analyses are affected by selective mortality. Thus, the convergence of depression by levels of education in late life may be illusory.