The premises of the study are that crime is not randomly distributed within a city, that the risks of punishment are not constant within a large and differentiated urban area, and that the primary sources of risk-related information within the neighborhood are friendship networks. The neighborhood-level data were collected from calls for service to the Oklahoma City Police Department over a 100-week period beginning in June 1986. The nature of each call; the month, day, year, and time of call; and the address to which police were dispatched were recorded; although this paper only reviews robberies. The findings suggest that, although there is some weak evidence of significant neighborhood differences in the deterrence process, on-site arrests, in general, do not have a deterrent effect on reported robberies. The authors suggest that the results reflect a short-term period of equilibrium in the neighborhood during which the levels of crime and arrest are relatively stabilized. 3 tables, 1 figure, 5 notes, and 46 references (Author abstract modified)