Moffitt's developmental taxonomy anticipated that cognitive abilities will relate to criminal activity differently throughout the population of offenders, mattering more for "life-course-persistent" offending than "adolescence-limited" offenders. The current study tested this theory in a longitudinal study. The subjects in the Philadelphia portion of the NCPP were infants of mothers who received inexpensive maternity care provided by a public clinic at Pennsylvania Hospital. The sample for the current analysis consisted of 987 subjects selected from the first 4 cohorts (1959-62) of 2,958 African-American mothers who participated in the NCPP. In the early 1980's, a research team from the University of Pennsylvania collected information related to school functioning and criminal history, including all police contacts through age 17. In the summer of 1998, a criminal history follow-up was completed for all 987 subjects. Data on adult offending were obtained from police records. Independent variables were several operationalizations of life-course-persistent offending. The primary independent variable was cognitive abilities/neuropsychological risk, which was measured at age 7 and 8, prior to the onset of criminal activity, by the digit span component of the WISC verbal scale. The second measure was the total battery score of the California Achievement Test, which was administered when the subjects were in grades seven and eight (ages 12 to 14). Several other variables were used as controls in the model, including sex, socioeconomic status, mother's age at childbirth, whether the mother was single at childbirth, and number of school disciplinary charges. The analyses of both the full and the offender-only samples found that higher scores on tests of cognitive abilities protected against life-course-persistent patterns of offending. This finding held after controlling for several other variables presumed to be associated with criminal activity. The main findings of this study also replicated some recent research conducted in Europe. Although these results provide support for Moffitt's hypothesis that cognitive abilities constitute one important discriminator between adolescence-limited and life-course-persistent offending, the data were limited in their ability to produce generalizations. Suggestions are offered for future research. 3 tables, 5 notes, and 60 references