Several studies attempt to cross-validate the classification system by applying the sorting rules to different populations. Only one other study has tried to replicate the derivation of the types. The results of the present study, using a diverse halfway house population, are largely supportive of the Megargee types, but types Baker, Easy, and Jupiter failed to replicate. Suggestions are made for improving the test-retest reliability of the classification system, and implications for future research are proposed. (Publisher abstract)