Over the last decade, the number of homicides in Porto Alegre has increased to the point where external causes are now the main group of causes of death in the 5–34-year age group. Preventing these deaths depends fundamentally on identifying factors related to excess violence in population groups. The overall aim of this study is to analyse the spatial distribution of homicide victims by place of residence in Porto Alegre, the capital of the southernmost Brazilian State of Rio Grande do Sul, in 1996, in order to identify and understand the socio-spatial context. Demographic and socioeconomic indicators based on the 1991 census and 1996 population count were used to build a multivariate classification characterizing the 1851 census tracts. Homicides occurring in 1996 were located using the municipality’s Geographic Information System. Four socioeconomic groups were identified, mainly differentiated by housing indicators. Small areas on the urban periphery in which slums (favelas) are concentrated presented higher homicide rates. Homicide rates were lower in the two groups with higher income and educational level. The second step was to classify the census tracts according to the homicide indicator. In this case, areas were differentiated by the number of household inhabitants per room, income, schooling, and median age. We conclude that the multivariate socioeconomic classification presents a limited capacity to identify populations exposed to homicides, suggesting that socioeconomic conditions themselves do not determine violent behaviour. On the other hand, the spatial methods allowed us to identify small areas where deaths are concentrated and whose populations should receive special attention in planning measures to prevent violent deaths.