Multiple regression analysis of data collected from a random sample of 300 Southern counties shows that structural and demographic variables explain less than 10 percent of the variance in county homicide rates in 1960 and 1970. Analysis was discontinued because interaction effects were detected. It is proposed that the association of certain structural variables and such unspecified demographic variables as age, sex, and race may account for the failure of structural variables to predict county homicide rates. Structural changes are thought to be producing a relocation of individuals who are demographically prone to commit homicide. It is argued that earlier findings regarding the association of structural variables and State homicide rates may be an artifact of the use of the State as the unit of analysis. The failure of the cultural explanation to specify a learning theory is discussed, and such a theory is tentatively sketched. It is proposed that ecological and behavior approaches may be complementary. Twenty-seven tables, 6 figures, and about 100 references are given. Five appendixes contain study-related materials.