Age-standardised mortality rates are often used in epidemiologic studies to describe the dimension of social inequalities in mortality. This, however, conceals any age-dependence of social inequality. In an ecologic study, all causes and cause-specific mortality of all citizens of Bochum, FRG, who died 1988-1990, were evaluated using 13.171 death certificates. Data was aggregated on census tract level. The social status of a census tract was determined using 6 variables from the census 1987 describing the socio-economic situation in each census tract. Census tracts were grouped into quintiles according to their social status. Age and sex-specific mortality rates as well as rate ratios, using the quintile with the highest social status as reference, were calculated. Results for men (n = 6.288) indicate that social inequality is age-dependent for total mortality. Social differentials are especially marked for the age groups 35-64 years. For age group < 35 years and > 84 years no social differentiation in mortality is visible. Similar patterns are found with mortality from cardiovascular diseases (ICD-9: 390-459) and cancer (ICD-9: 140-208). Mortality from diseases related to health behaviour such as lung cancer or diseases associated with high alcohol intake are characterised by social inequalities above average in the middle age groups. For total mortality in women (n = 6.883) large social differentials are found for age groups 25-34 years and 45-54 years. Efforts to reduce social inequality on community level should especially be aimed at adolescents and young adults living in underprivileged areas.