Research data from West Germany consisted of a 5 percent sample of all criminal offenders convicted for theft, fraud, breaking and entering, assault, traffic offenses, and similar crimes during 1972 in the State of Baden-Wurttemberg. British data were obtained from a statistical study of 5,000 offenders convicted in January 1971. To explore whether criminals tend to move between specific crime types, transition matrixes were examined and tested for the null hypothesis of no relationships (independence). A Markovian model was applied to determine how much subsequent convictions depend on past crime careers. The Markovian model assumes that current transitions are not influenced by the past. Several measures were used to estimate the extent of crime type specialization, but the ultimate research goal was to observe and predict criminality. Analysis of transitions among crime types in successive convictions suggested that these transitions conform to a quasi-independent model and that deviations show no systematic patterns or trends, although there appeared to be some specialization, at least for property and traffic offenses. No escalation in serious over crime career was observed, nor were there significant groupings among crime types, except for a slight similarity between property offenses and fraud. The finding that transition matrixes were close to quasi-independence suggested that prior offense records are not very good predictors of future convictions and thus provide little guidance for developing sentencing policies that seek to account for incapacitation of special deterrence. 19 references, 19 tables.