This study aims at contributing to a more elaborate understanding of the historical phonology of Mawé (Tupí family, Brazil). The paper discusses two patterns of diachronic correspondence involving j and w, arguing, first, that the diachronic correspondence involving j is more insightfully viewed as stemming from the blocking of a regular process of change and, second, that the process of change underlying the correspondence of the glide w is better conceptualized as a dissimilatory process. A unified, conspirational account of these changes is proposed by invoking a constraint banning homorganic glide-vowel sequences that acted in two seemingly contradictory ways: blocking a change that targeted t in the proto-language and triggering a change that dissolved wu sequences. Finally, I discuss the singular character of the correspondence w > h and propose a hypothesis accounting for this reflex. Wider implications of these accounts are also considered, and the hypotheses and analyses advanced here are evaluated and compared to alternative accounts.