'Hard' theories successfully describe the particular sets of events for which they have been designed. Their weaknesses are that theories appear to be confirmed by instances for which the theory has no obvious application, beliefs can be adjusted to allow for any evidence, and the way in which events are described determine which theories will be supported by the evidence. One of the more enduring 'fresh' theories about crime causes implicates broken homes. Too little attention, however, has been given alternative explanations to warrant concluding that broken homes cause crimes. 'Soft' theories spread into criminology after attaining a degree of acceptance as explanations for noncriminal behavior. These theories are so nebulous they continue to influence criminologists despite evidence that developments from their premises have failed to produce predicted results. 'Semihard' theories account for criminality in terms of rewards, punishments, and behavioral models. These theories assume universal psychological hedonism, although psychological hedonism has not received the critical examination that its role in these theories requires. New metatheories, which integrate 'simple' theories, are likely to improve criminality predictions, if for no other reason than they involve more variables than simple theories. As a trade-off, however, underlying conceptual confusions and false assumptions are likely to be more difficult to detect. Having a reasonable theory should require establishing that events occur in the sequences implied by the theory. 115 references.