Study participants were involved in the Dunedin Multidisciplinary Health and Development Study, a longitudinal investigation of the health, development, and behavior of a cohort of consecutive births between April 1, 1972, and March 31, 1973, in Dunedin, New Zealand. Data were obtained at birth and at age 3 for 1,037 children (52 percent male and 48 percent female). In socioeconomic status and ethnicity, the children's families were representative of the general population of New Zealand's South Island. The sample was reassessed with a diverse battery of psychological, sociological, and medical measures at ages 18 (n=993), 21 (n=992), and 26 (n=980). Measures were official reports of offending and self-reported delinquency. A measure of neuropsychological health was taken when participants were 13 years old, along with a measure of personality (modified version of the Multidimensional Personality Questionnaire). The analyses determined how well the number of violent offenses could be modeled by using a single binomial probability. Findings differed as a function of data sources, such that official reports provided no evidence of offense specialization; whereas, specialization was evident in self-reports of offending. The findings suggest that violence in an individual's criminal history is an important factor that marks a subgroup of offenders who begin offending early and are characterized by a particular set of personality traits. The individuals who committed some violent acts apparently differed from those who committed none. These findings support Moffitt's (1993) taxonomy of antisocial behavior, which posits two types of offenders: life-course-persistent offenders, who manifest antisocial behavior early in the life course and continue to offend into adulthood, and adolescence-limited offenders, who offend only during adolescence. 3 tables and 21 references