Hearing vocalizer race from voiced laughter sounds

Abstract

Previous experiments testing whether an individual's race can be identified from speech sounds have produced inconsistent results. Current work examined this question using spontaneous laughter sounds to help isolate voice quality from dialect effects. In Experiment 1, six laughs from each of six black and six white vocalizers were edited to create 72 bouts of three to four contiguous bursts. These bouts were then further edited to create 72 single bursts. Fourteen undergraduate participants heard both sets of laughs in separate blocks. Sounds were presented twice on each trial and categorized as “Black” or “White.” Mean percentage-correct was 61.1 (SD = 10.1) with multi-burst versions, and 55.8 (SD = 12.7) with single bursts. Both outcomes were statistically higher than chance performance in t-test comparisons. Experiment 2 tested 12 new listeners in a balanced, same-different design with paired, bout-length stimuli from same- or different-race vocalizers. There were 120 stimuli in all, combining 5 laughs from each of 5 laughers of each race. Mean d' was 0.92 (SD = 0.44), which was statistically different from chance. Taken together, these experiments suggest modest, but reliable sensitivity to vocalizer race from voice-quality alone.