The main sources of national information on homicides are the Supplementary Homicide Reports (SHR) that local law enforcement agencies provide the FBI through the Uniform Crime Reporting system. However, a detailed comparison of Baltimore's 1983 SHR reports and summaries of police investigations showed that the classification used by the SHR has low reliability. The sources of this low reliability are the ambiguous definitions used, the fact that a homicide can be placed in only one category, and the failure to use all the available information for categorizing some cases. Another study examined the relationship between homicide and poverty using the 1970 SHR data for the 49 largest United States cities. Cities with large poverty populations had higher rates of homicides, especially robbery homicide, than did other cities. Another study used previously collected data on Detroit's felonies during 1976-1978. Offenders who inflicted injuries during robberies were more likely to have a prior history of violent offenses. A final study used Detroit data to examine the role of weapons and robbery homicide. The use of guns was strongly associated with the incidence of robbery homicide. Figure, data tables, 9 references.