Discusses problems with the use of stochastic (probabilistic) models for the analysis of juvenile offense data. Analyzes longitudinal data for White and non-White delinquents which show significant evidence of offense specialization and a random distribution of offenses if no specialization occurs. (Author/GC)