Acoustic variability shapes adaptation to single versus multiple sound sources

Many studies have demonstrated superior recognition of words spoken by a single talker compared to multiple talkers. Recently, Shorey et al. (2023 AP&P) reported parallel results when labeling words from talkers and tones from musical instruments. Here, we further examined the effects of stimulus variability in both speech and music perception. Experiment 1 adopted the paradigm of Stilp and Theodore (2020 AP&P) with multiple mixed-talker blocks where voices had Low Variability (similar mean f0s) or High Variability (dissimilar mean f0s); instruments with Low Variability (similar attacks and spectra) or High Variability (dissimilar attacks and spectra) were also presented. Music perception was slower and less accurate at each successive increase in stimulus variability. However, the effects of increased stimulus variability were less pronounced in speech blocks than in music blocks. Experiment 2 used this paradigm to measure vowel identification spoken by multiple talkers with dissimilar mean f0s saying the same two target words (heed and hoed, Low Variability) or eight different target words that all shared the /i/ or /o/ target vowel (High Variability). Speech responses became slower at each successive increase in stimulus variability, paralleling music blocks. Thus, adaptation to sound sources (formerly “talker normalization”) displays domain-general responses to stimulus acoustic variability.