This paper presents analysis of data aggregated from three experiments, investigating whether there are social differences in the processing of sociolinguistic variation. Three self-paced reading experiments tested the processing of variable subject-verb agreement in English. Sentences appeared in one of three agreement conditions: standard [singular+doesn’t] or [plural+don’t], nonstandard [singular+don’t], or “uncommon” [plural+doesn’t]. While the nonstandard form is common in vernacular English dialects, the uncommon form is not known to be a dialect variant. For participants as a whole, relative to standard sentences, both nonstandard and uncommon sentences took longer to read. Uncommon sentences took the longest, reflecting participants’ greater familiarity with the nonstandard compared to uncommon form. To explore social differences, I compared responses by grouping the data on three social dimensions: class (higher- and lower-class participants), race (white and African American participants), and sex (female and male participants). The hypothesis was that if these social groups map onto differences in linguistic experience/knowledge, such differences would be reflected in different responses to the nonstandard variant. The main findings are a) groups show differences in reading speed, independent of agreement, and b) agreement was not a significant predictor of reading times for the African American participant group. I discuss the implications for experimental methodology and future research on sociolinguistic perception.