Discipline: Social, Behavioral, and Economic Sciences
Subcategory: Social Sciences/Psychology/Economics
Room: Park Tower 8216
Deema Farraj - Brooklyn College CUNY
Listeners have little difficulty identifying the sex of a speaker’s voice, but how this is achieved remains poorly understood. Adult male and adult female voices differ in two prominent ways: fundamental frequency/pitch (f0) and range of formant frequencies (due to the length of vocal tract). Because these two acoustic properties ordinarily occur together in the same individual, it is unknown whether one of them plays a more important role than the other in listeners’ ability to identify the sex of a speaker. It is also unknown how f0 and formant frequencies interact, and if either one of them would have an effect independently of the other. If one of these two features were neutralized, would the influence of the other on sex identification be equal between f0 and formants? This talk reports two experiments designed to neutralize f0 and formants (respectively) to determine whether listeners make use of either feature in isolation in their perception of speaker sex. Listeners were presented with synthesized speech where formant ranges and f0 varied independently from one another, and asked to judge whether the token sounded “more male” or “more female” using a Likert scale. In one experiment, listeners were presented with a set of stimuli with “ambiguous” formants in between the typical adult male and female ranges. The f0 contours of the stimuli, however, varied in nine equal steps from the typical male range (85-120 Hz) to the typical female range (170 to 240 Hz.) In the second experiment, the f0 was kept in a constant, ambiguous region, midway between typical male and female ranges, but formants were varied in 9 equal steps, ranging from typical male to female values. Formants were estimated using recordings of a male speaker, and then raised by approximately 3% for each step. The stimuli were synthesized using a modified protocol for Sinewave Speech that adds a perceptual cue for pitch. The results of experiment 1 indicated that when formants were kept constant in an ambiguous range, lower (perceived) f0 values resulted in clear judgements that the voice was “male” while higher perceived f0 values resulted in judgements of “female.” When perceived f0 was held constant in an ambiguous range (experiment 2), judgments of speaker sex partially correlated with formant ranges. Lower formant frequencies were never judged to be “female” (the average judgment for the tokens with the lowest formant values was halfway between male and female). Listeners tended to judge stimuli with higher formants as “female.” These results suggest that pitch has a stronger influence than formant frequencies on judgments of speaker sex, but that formants also play a role. A direction for possible future research is how listeners would perceive a voice that exhibits “mixed” properties, i.e. f0 in the typical male range but formants in the typical female range, and vice versa?
Funder Acknowledgement(s): Funding was provided by NSF REU SITE Intersection of Linguistics, Language, and Culture to Brooklyn College and Long Island University (LIU) REU Site Grant SMA# 1659607
Faculty Advisor: Jonathan Nissenbaum, JNissenbaum@brooklyn.cuny.edu
Role: I designed and carried out the study with input and feedback from my faculty mentor and other members of his lab.