This study examines some cases of heteroclisis in the history of Romanian dialects, and concludes that the data call for a reconsideration of Stump's distinction (Language 82:279-322, 2006) between 'cloven' heteroclisis, where the intraparadigmatic 'split' is aligned with some morphosyntactic feature distinction, and 'fractured' heteroclisis, where this is not the case and the pattern of heteroclisis is purely morphological. Stump's account creates the impression that the 'cloven' variety is universally predominant, and that the 'fractured' variety tends to follow very closely the available 'cloven' patterns of the language. I shall suggest, instead, that the 'fractured-only' situation may in fact underlie heteroclisis cross-linguistically, the phenomenon being in general sensitive not directly to morphosyntactic content, but rather to characteristic, and often purely 'morphomic', patterns of stemallomorphy.