It argues that neighborhoods are more appropriate units of analysis for studying inequality and homicide than are larger political and statistical units because neighborhoods are more likely to constitute meaningful frames of reference for social comparisons. The principle hypothesis is that a high degree of economic inequality in a neighborhood will give rise to high levels of relative deprivation and high rates of homicide. The results of a series of multiple regression analyses fail to support this hypothesis. The measure of economic inequality is weakly associated with the observed homicide rates. Similarly, the racial composition of Manhattan neighborhoods exhibited no significant association with levels of homicide, given statistical controls for other sociodemographic variables. Two neighborhood characteristics do emerge as significant predictors of homicide rates: the relative size of the poverty population and the percent divorced or separated. Homicide rates tend to be highest in those neighborhoods characterized by extreme poverty and pervasive marital dissolution. (Publisher abstract)