Elliptical speech reveals the use of broad phonetic categories aids noise-degraded speech perception

We investigated links between perceptual gradiency in phonetic categorization and speech-in-noise (SIN) perception in listeners with varying music backgrounds. Categorization was measured for vowels and stops using phoneme labeling tasks. Speech discrimination and transcription were assessed using “elliptical speech” sentences [Miller and Nicely (1955). JASA , 27, 338–352] which use featural substitutions that renders them meaningless in clean conditions but surprisingly improves their recognition under noise. We hypothesized listeners who use broader perceptual equivalency classes at the phoneme and/or sentence level would show better SIN. Perception of elliptical speech was indeed resilient to noise but this elliptical benefit varied with music background; nonmusicians showed larger susceptibility and noise-related benefit from ellipses than musicians. Phoneme categorization and elliptical sentence perception were also associated with QuickSIN performance but in opposite ways depending on musicianship. Broader phonetic category usage was related to better SIN in nonmusicians but poorer SIN in musicians. Findings suggest listeners can use broader/narrower perceptual equivalency classes and top-down inference to cope with noise degradation but this depends on auditory demographics. Musically naïve listeners can use broader phonetic categories to aid SIN perception while expert listeners can use narrower categories in otherwise similar speech contexts. PACS numbers 43.66, 43.71