The Relationship Between Concurrent Speech Segregation, Pitch-Based Streaming of Vowel Sequences, and Frequency Selectivity

Simultaneous and sequential segregation form the basis of auditory scene analysis and are likely involved in concurrent speech segregation. However, previous work showed that speech-in-noise perception was uncorrelated with simultaneous segregation, whereas it appeared to be related to the pure-tone fusion threshold of sequential streaming. The current study aimed to clarify the relationships between pitch-based speech-in-speech segregation, pitch-based streaming, and frequency selectivity. Twenty-three listeners with close to normal hearing were involved. Speech-in-speech perception was measured using words presented in a time-reversed single talker background, with various pitch differences between target and masker. Streaming performance was measured using an objective order-naming task on vowel sequences. Auditory filter widths were derived using a notch-noise method. Results showed a correlation between the effect of pitch on speech-in-speech perception and the effect of pitch on streaming performance. However, frequency selectivity was found to correlate with average speech-in-speech perception but not with streaming, and only in the region of the second formant. These latter results are consistent with the hypothesis that pitch-based streaming probably relies on pitch discrimination, which is only poorly correlated to frequency selectivity. Further, these results suggest that mild impairments in frequency selectivity do not systematically impair pitch-based streaming.