Body mass is a key determinant of fitness components in many organisms, and adult mass varies considerably among individuals within populations. These variations have several causes, involve temporal and spatial factors, and are not yet well understood. We use long-term data from 20 roe deer cohorts (1977-96) in a 2600 ha study area (Chizé, western France) with two habitats contrasting in quality (rich oak forest in the North versus poor beech forest in the South) to analyse the effects of both cohort and habitat quality on adult mass (i.e. median body mass between 4 and 10 years of age) of roe deer (Capreolus capreolus). Cohort strongly influenced the adult body mass of roe deer in both sexes: males born in 1994 were 5.2 kg heavier when aged between 4 and 10 years old than males born in 1986, while females born in 1995 were 4.7 kg heavier between 4 and 10 years old than females born in 1982. For a given cohort, adult males were, on average, 0.9 kg heavier in the rich oak forest than in the poor beech forest. A similar trend occurred for adult females (0.5 kg heavier in the oak forest). The effects of cohort and habitat were additive and accounted for ca. 40% of the variation observed in the adult mass of roe deer at Chizé (males: 41.2%; females: 40.2%). Population density during the spring of the birth accounted for about 35% of cohort variation, whereas rainfall in May-June had no effect. Such delayed effects of density at birth on adult body mass probably affect population dynamics, and might constitute a mechanism by which delayed density-dependence occurs in ungulate populations.