This study tested whether partners can be relied on to provide congruent reports about abuse in their relationship. The authors examined whether interpartner agreement (IA) varies as a function of whether the perpetrator is the man or the woman, and by whether the abusive behavior being reported is physical or psychological. Guided by psychometric test theory, the authors examined whether weak 14 about specific behaviors can be improved by aggregating behavior items into scales and by controlling for random measurement error. A representative sample of 360 young couples was studied. IA did not vary with the perpetrator's gender or with the nature of the abusive behaviors, but victims (both men and women) reported somewhat more abuse than did their perpetrators. IA about specific abusive behaviors was only poor to fair, but it became very good when items were aggregated into scales and even better when measurement errors were removed from the reports. The findings suggest that reports of abuse can be aggregated to form internally consistent scales that show strong IA, thereby fulfilling criteria for reliability. Moreover under research conditions that guarantee confidentiality, either abuser reports or victim reports are suitable methods for use in research on partner abuse.