Sociolinguistic research has traditionally examined stylistic variation as a way of understanding how speakers may use language indexically. Quantitatively, research has sought to correlate observed patterns of variation across such external parameters as context or topic with the ways in which speakers linguistically orient themselves to their immediate surroundings or to some other socially-salient reference group. Recently, this approach has been criticized for being too mechanistic. In this paper, I present a new method for examining stylistic variation that addresses this critique, and demonstrate how an attention to speakers' motivations and interactional goals can be reconciled with a quantitative analysis of variation. I illustrate the proposed method with a quantitative examination of systematic patterns of prosodic variation in the speech of a group of Israeli men who are all members of various lesbian and gay political-activist groups.