Publication Date



Psyche Loui; Gloster Aaron; John Kirn


Neuroscience and Behavior


The acoustic cues that influence one’s impression of a speaker’s race were investigated in a series of three experiments. Two identical sentences read by 17 Southern Black and 17 Southern White speakers were selected from the TIMIT corpus and analyzed using the speech synthesizer Tandem-STRAIGHT (Kawahara et al., 2008). Sentences were kept the same (“original” speech; Experiment 1), raised and lowered in fundamental frequency (F0) by 1, 2, and 3 semitones (Experiment 2), and increased and decreased in vocal tract length (VTL) by 5, 10, and 15% (Experiment 3). Amazon Mechanical Turk listeners (N = 211) rated original and manipulated speech on a 6-point scale: 1-most likely Black, 2-probably Black, 3-possibly Black, 4-possibly White, 5-probably White, and 6-most likely White. Ratings for original Black speech were significantly lower than ratings for White speech, demonstrating that listeners could distinguish speakers by race (Experiment 1). Analyses suggest that listeners relied on speakers’ vowel quality, height, and use of African American Vernacular English phonology to inform their ratings. F0 manipulation did not affect listener ratings, confirming that F0 information is not used in auditory race perception (Experiment 2). VTL manipulation did significantly affect listener ratings. Sentences with longer VTL elicited lower ratings (closer to 1-most likely Black) and sentences with shorter VTL elicited higher ratings for Black and White speech alike (Experiment 3). This indicates that race perception is influenced by a speaker’s VTL. The study’s findings contribute to understanding how race discrimination based on the voice may occur.



© Copyright is owned by author of this document