Sociophonetic studies of speech perception have demonstrated that the social identity which listeners attribute to a speaker can lead to predictable biases in the way speech sounds produced by that speaker are linguistically categorized (e.g., Strand & Johnson 1996; Niedzielski 1999; Hay, Warren & Drager 2006). This has been observed where listeners use available social information about a speaker to resolve lexical ambiguity. However, less is known about the role of sociophonetic knowledge in speech perception when listeners are not faced with global linguistic ambiguity. Drawing on Strand's (2000) study of the processing effects of gender typicality, this dissertation investigates whether sociophonetic knowledge can facilitate or inhibit unambiguous spoken word recognition. Based on a survey of sociophonetic variation in the Houston metropolitan area, predictions are formulated for the processing of words containing four vowels: /ei/ and /[varepsilon]/ in the speech of older and younger Anglos, and /α/ and /Λ/ in the speech of young Anglos and young African-Americans. Houston listeners identified words containing variants of these vowels in a congruent condition and in an incongruent condition. In the congruent condition the combination of speaker identity and vowel variant was designed to match the listener's knowledge of local language variation. In the incongruent condition, it was designed to contradict it. A congruency effect was found for some but not all vowels. The results indicate that social information about a speaker can also affect speech perception in the absence of lexical ambiguity, but only where words are at least temporarily ambiguous. Where there is no linguistic ambiguity at all, perception can be unaffected by sociophonetic knowledge. These results are discussed in the context of Luce, McLennan & Charles-Luce's (2003) time course hypothesis and in the context of exemplar-based models of sociophonetic knowledge (Johnson 1997, Pierrehumbert 2001).