Mortality at early ages has long been known to reflect the general health of communities, and derived epidemiological parameters have been used to indicate those communities to which policies of environmental improvement and health promotion should be directed. The geographical distributions of stillbirths, and perinatal, neonatal, post-neonatal and infant mortalities in Scottish communities were analysed for the years 1959-63, 1969-73 and 1979-83. For all categories of mortality, urban-rural gradients were present in the earlier periods, but had disappeared by 1979-83. Several communities were identified which had high rates for at least two categories of mortality, and which may benefit, therefore, from intervention programmes. For 1979-83 only, bivariate correlation coefficients were calculated between the five categories of mortality and four socio-economic variables: social class, overcrowding (over 1.5 persons per room and persons per hectare), single parent households and housing tenure. For stillbirths and perinatal mortality, significant correlations were found with persons per room, housing tenure and social class. Neonatal and postneonatal mortalities were significantly correlated with council house tenure and persons per hectare respectively, whereas infant mortality was not significantly correlated with any variable. Multiple regression, using the five categories of mortality as dependent variables and social class, persons per room, persons per hectare, housing tenure and single parent families as the independent variables, showed that these socioeconomic characteristics explained between 10-32% of the variation in mortality.