Purpose. This study explores the social, biological, and genetic determinants of depression in later life. It adds complexity to the idea that later life depression is a natural outgrowth of vascular impairment, antagonistic pleiotropy, or compromised neuroanatomical structures. Approach. The study uses the NAS-NRC Twin Registry of World War II Veterans. The use of twins allows for the exploration of gene-environment interplay. A recent survey instrument associated with the registry contains numerous indicators of health, as well as an established measure of geriatric depression. Findings. The results show that education has a strong negative relationship with depression among those in their 70s and early 80s. Although this relationship is partly explained by lower rates of cardiovascular disease and diabetes among the well-educated, the relationship between education and many common physical illnesses is quite small. Most people of this age experience at least one chronic illness. The relationship between education and depression is explained, instead, by reduced impairments in activities of daily living. These impairments are not an inevitable outgrowth of declining health. The well-educated are better able to moderate the impact of poor health on daily functioning. Moreover, the well educated are able to avoid the otherwise strong genetic risks for depression in later life. Gene × environment models show a high heritability for later life depression on average, but also reveal that this heritability declines with increasing education. Among those with a four-year college degree, the heritability of depression is very small. Value. These patterns are interpreted in light of models for understanding compensatory gene × environment interactions. These models emphasize the importance of especially enriched environments for overcoming genetic risk.