We investigate primary school precursors to criminal involvement in early adulthood. Two large longitudinal datasets, Children of the National Longitudinal Youth Study (NLSY) and the Beginning School Study (BSS), provide us with estimates of the association between criminal involvement by around age 21 on the one hand and four primary school indicators — reading and math skills, attention problems, and antisocial behavior problems — on the other. Looking first at bivariate associations, we find significant correlations between early-adult crime and all of our early measures of skills and behaviors. Controls for either a handful of family background measures or concurrent primary-school skills and behaviors reduce all but the early antisocial behavior measure to statistical insignificance. Our detailed look at the persistence of early antisocial behavior problems show that children, particularly boys, with antisocial behavior that begins early and persists beyond age 10 or 11 are at the highest risk of later arrest or incarceration.