Drawing on data gathered during fieldwork in an Indo-Guyanese village (1994–96), this article shows that the gender patterns for variable pronominal usage are strikingly stable over time. In both this study and one conducted by John Rickford more than twenty years ago, women, compared with men, use more basilectal variants in the category of 1sg. subject (mi vs. ai), but fewer in the category of 3sg. objects (am vs. shi/ii/it). Rather than explain this variation as a “leveling out” (Rickford 1979), it is here suggested that variants must be understood in terms of their contribution to an unfolding interactional engagement. The conclusion remarks on the continuing confusion among sociolinguists regarding the analytical relevance of gender as an external constraint on variation. A more developed understanding of these issues depends on recognizing the way in which language variation serves as an indirect and constitutive index of gender.