This article is a study of how Roman grammarians treated a specific grammatical problem, viz., the description and classification of verbal gender in Latin. The results show that various theories were put forward and various possibilities for systematizing and explaining the data were proposed in the works of the grammarians of the 3rd and 4th centuries. Out of the discussion of these grammarians there emerged in the works of the grammarians of the 5th and 6th centuries (notably Phocas, Pompeius, and Priscian) a rather adequate description. While morphological and semantic criteria were intermingled in the earlier period, later grammarians kept them apart. We also find a tendency towards a more theoretical and less data-oriented linguistics. The picture emerging from this in scope rather limited study deviates to some extent from the theories proposed by Bar-wick (1922) concerning the history of Roman linguistics.