The convergent validity of the two most frequently used methods for assessing violent offending in juveniles (i.e., self-reports and arrests) was evaluated. Participants were 87 serious juvenile offenders and their maternal figures, primarily from disadvantaged families. Validation measures tapped established behavioral, family, and peer correlates of delinquency. Results failed to support the ability of either arrests for violent crimes or self-reported violent offenses to index violent criminal behavior accurately. Several methodological features of the study support our hypothesis that the findings were not spurious. Procedural and conceptual implications of the findings are discussed.