Phonological but not lexical processing alters the perceptual weighting of mean fundamental frequency and vocal-tract length cues for voice gender categorisation

Abstract

Listeners use various voice cues to segregate different speakers, or to infer speaker-related information such as perceived gender. Two important anatomically related voices cues used for speaker identification, including perceived gender, are mean fundamental frequency (F0), related to the glottal pulse rate, and vocal-tract length (VTL), correlating with body size. Voice cue processing seems to be affected by linguistic processes, such that voice perception is more precise when listeners hear speakers in their native language compared to a non-native language. In addition, recent research shows that F0 and VTL sensitivity is lower for words compared to time-reversed words, either because time-reversed words are unintelligible or phonemes are distorted in voice-onset times and aspirations, pointing to effects of lexical or phonological processing. However, voice cue sensitivity and using these cues to infer speaker-related information may rely on different mechanisms. Here, we studied effects of lexical and phonological processing on F0 and VTL cue weighting for one aspect of speaker identification, namely voice gender categorisation, by manipulating these cues in three linguistic conditions: meaningful words; phonotactically plausible nonwords; and phonotactically implausible time-reversed nonwords. We found that F0 and VTL weighting for voice gender categorisation was affected by phonological but not by lexical processing.