The SARP was designed to replicate the Minneapolis domestic violence experiments (Sherman and Berk, 1984), which found that arresting domestic-violence suspects contributed to a lower risk for repeat domestic violence. In order to examine the external validity of this result, in 1986 the National Institute of Justice funded the SARP to replicate the Minneapolis study in six other cities in geographically diverse regions of the country. The SARP data were selected for the current study because they facilitated an analysis of whether domestic-violence offenders exhibited specialization in violence. Also, the data provided information from victim interviews on the nature of the violence in both the presenting incident and in subsequent victimization incidents. This permitted an analysis of the extent to which the severity of offenders' attacks against the same victim increased, decreased, or stayed about he same. The study found that the majority of domestic-violence offenders with prior official criminal records had been involved in nonviolent criminal behavior in addition to domestic violence. Regarding variations in the seriousness of domestic violence over time, three SARP sites manifested a heterogeneous mix of offenders who escalated and de-escalated the severity of their attacks over the relatively short follow-up periods; however, one other site showed pronounced tendencies for offenders to escalate the severity of their attacks when the presenting case involved minor injury. There was no tendency at this site for offenders to de-escalate the severity of their attacks when the presenting incident involved serious injuries. 3 tables, 14 figures, and 50 references