The authors analyzed cases of police officers accused of culpable homicide or murder that were reported in the English- language South African press during 1986-92. The number of fully reported cases (reports that included the facts of the case and the outcome of a trial) was only 30; data were also drawn from homicide cases that attracted only partial press coverage, serious assault cases, and inquests (approximately 100 additional cases). The data were sufficient for a preliminary analysis, but conclusions should be considered tentative because of the limitations of the newspaper sources. Circumstances surrounding the killings were classified into four types: conventional policing, reactive unrest policing, proactive unrest policing, and off-duty incidents. Unrest cases were coded as "reactive" if officers claimed that their actions were provoked by public disturbances. Cases were coded as "proactive" if they involved unprovoked actions against persons thought to be opposed to the regime. The sample consisted of 15 unrest-related cases (7 reactive and eight proactive), 12 conventional policing cases, and 3 off-duty cases. Findings show that the courts tend to rule against the accused in certain circumstances, particularly when the killing was premeditated or clearly unjustified by the evidence, but the courts are charitable in cases where there was some provocation or where the accused was highly agitated. The net effect of judicial lenience in cases of reactive unrest policing and deaths of persons in custody is arguably to normalize the use of deadly force in these circumstances. 2 tables and 101 footnotes