The authors first used variables measured at program intake to predict female and male underreporting of male violence at program intake. The authors then used variables measured at program intake, as well as measures of program participation, to predict male-female disagreement (male underreporting) at 12-month follow-up. Several variables predicted underreporting, both at intake and at follow-up, but overall prediction was only marginally better than chance. Findings suggest men and women underreport violence based on situational factors such as relationship characteristics and on rational reasons rather than based on personality traits or social desirability. The ability to predict underreporting, however, is too weak for adjustment of self-reports by clinicians and program evaluators. 35 references and 2 tables