One theory postulates that prior criminal participation has a genuine behavioral impact which increases the probability of future participation. The other explanation holds that there are stable unmeasured differences in criminal potential across the population. The model used here included measures of four personal and familial characteristics: IQ, parental criminality, poor parental child rearing behavior, and risk taking. Two findings in particular supported the persistent heterogeneity-based theories. The results showed that associations of criminal behavior with three of the four individual constructs were stable in magnitude over an extended period of time. There was also a very substantial level of persistent unobserved heterogeneity. After controlling for this factor, the association between past and future criminal participation was greatly diminished. The study does not address the question of whether social bonds can alter developmental trajectories established in childhood. 4 tables and 45 references