The influence of voice parameters on perceptual grouping in dynamic cocktail-party listening

Abstract: Cocktail parties pose a difficult, but solvable, problem for the auditory system (reviewed by Shinn-Cunningham, 2008). The cocktail-party problem is made considerably easier when all sounds within the target stream are spoken by the same talker (Bressler et al., 2014). The so-called “voice-continuity benefit” occurs because speech sounds from a single talker are all similar in terms of certain acoustic features. This makes it easier to group together these sounds as a perceptual object than speech sounds from different talkers. Two of the most salient voice features are glottal-pulse rate (GPR) and vocal-tract length (VTL) (e.g., Baumann and Belin, 2010). GPR determines the talker’s perceived pitch, whereas VTL contributes to the perception of talker’s voice timbre. It is known that these cues are important for determining a specific talker’s identity (Lavner et al., 2000), but it is not known whether they are just as important for perceptual grouping in the context of cocktail-party listening. The proposed project aims to determine the relative contributions of GPR and VTL to the voice-continuity benefit in cocktail-party listening. To simulate a natural cocktail-party situation, we will use a free-field environment with loudspeakers at different positions. Previous research has also shown that individuals differ largely in their ability to solve cocktail parties (Ruggles et al., 2011). Using electroencephalography (EEG), we will be able to investigate whether these differences can be explained by the fidelity of neural temporal coding.