Traditionally used as a “heuristic device” (Labov, 1978), the sociolinguistic variable has taken on a new role as a primitive of speaker/hearer mental models in third-wave variation work (Eckert, 2005, 2008). Results from a sociolinguistic perception study suggest that at least in some cases, variants of the same variable function independently as loci of indexically linked social meaning. Listener responses were collected to three matched guises of the English variable (ING): -in, -ing, and a neutral guise with no audible (ING) tokens. The results counter the study hypothesis that listener expectation, triggered by speaker regional accent, would shape (ING)'s impact. Instead, the two variants showed distinct social associations: the -ing guises were rated as more intelligent/educated, more articulate, and less likely to be a student than either the -in or neutral guises, which did not differ significantly. In contrast, -in guises made speakers sound less formal and less likely to be gay than the -ing and neutral guises, which did not differ. These results suggest that third-wave work needs to more closely examine the role of the variable in theorizing the relationship between linguistic and social structures.