Classification is a fundamentally cognitive operation, and the significance of taxonomic issues has been widely acknowledged in the clinical literature on sexual aggression. Agreement about the level at which taxonomic differentiation among sexual offenders should occur, however, is not universal. The classification research program at the Massachusetts Treatment Center focuses on several stages, including theory formulation, implementation, validation, and integration. The application of this program to child molesters offers a good example of the importance of combining both deductive and inductive strategies in typology construction. In contrast with the child molester typology, discrepancy analyses of the rapist typology revealed certain problems without yielding hints about structural solutions. The revised rapist typology identifies opportunistic, pervasively angry, sexual, and vindictive behavioral types. Research to develop taxonomic structures for sexual offenders offers an example of a general move toward creating more particularized systems within relatively circumscribed behavioral domains. In using both deductive and inductive strategies simultaneously, the researcher generates and tests multiple structural variations and continually examines empirical feedback for emergent, consistent patterns. 60 references, 4 figures.