Sine-wave and noise-vocoded sine-wave speech in a tone language: Acoustic details matter

Sine-wave speech (SWS) is a highly simplified version of speech consisting only of frequency- and amplitude-modulated sinusoids representing the formants. That listeners can successfully understand SWS has led to claims that speech perception must be based on abstract properties of the stimuli far removed from their specific acoustic form. Here it is shown, in bilingual Cantonese/English listeners, that performance with Cantonese SWS is improved by noise vocoding, with no effect on English SWS utterances. This manipulation preserves the abstract informational structure in the signals but changes its surface form. The differential effects of noise vocoding likely arise from the fact that Cantonese is a tonal language and hence more reliant on fundamental frequency (F0) contours for its intelligibility. SWS does not preserve tonal information from the original speech but does have false tonal information signalled by the lowest frequency sinusoid. Noise vocoding SWS appears to minimise the tonal percept, which thus interferes less in the perception of Cantonese. It has no effect in English, which is minimally reliant on F0 variations for intelligibility. Therefore it is not only the informational structure of a sound that is important but also how its acoustic detail interacts with the phonological structure of a given language.