Accent bias and racial accents in academic comprehension

Abstract

The moment a person speaks, listeners are given access to a wealth of information: class, race, gender, and other environmental and cultural details. When we listen to people speak, we are checking to see if we can understand what is being said. Accents, however, can cause someone to deem another person unintelligible not because of the content of their production, but rather because of how they said it—the phonological or acoustic properties of the production. This study seeks to determine if racial accents are perceptible in quasi-academic (academic level, but not in an academic setting) contexts, and how many visual cues to a speaker’s race influence how or whether they are heard and the comprehension of the information conveyed in the various accents. To test these hypotheses, three online experiments consisting of audio recordings and photos were administered to a total of 240 participants split into three groups. In all three groups, participants were presented with photos of three women of different races (white, Latina, Black) who were depicted as having recorded the texts played auditorily that the participants read. In each experiment, two of the voices and photos were “mismatched”: the speaker’s voice and a photo depicting the race of the speakers were swapped. In each experiment, then, one photo-voice pairing was the “matched” baseline. This design, over the course of all three experiments, attempted to dissociate audio and visual input in determining whether and how both factors might contribute to both comprehension of texts and affective responses to the speakers. The results show clear evidence of accent bias, though the unique contributions of pictures and voices remain somewhat unclear.