Why do children attend to global structure in speech signals?

Abstract

Adults can use global structure such as that found in sine wave and vocoded signals for speech perception. Children appear to weight global spectral structure (represented at the syllabic level by formant transitions) even more than adults, prompting the suggestion that children first attend to the relatively slow vocal-tract movements that create global spectral structure rather than to the rapid and discrete gestures more closely associated with individual phonetic segments. But this interpretation hinges on an articulatory account of speech perception, which lacks general support. An alternative explanation is that children attend to formant transitions because they are found in voiced signal portions and so adhere to principles of auditory scene analysis, such as having a common fundamental. This work tested that hypothesis by using stimuli with no fundamental. Adults and children (3, 5, and 7 years) were asked to label fricative-vowel syllables with three kinds of vocalic portions: natural, sine wave, and whispered. Results showed that all listeners attended to formant transitions as much in the sine wave and whispered stimuli as in the natural, leading to rejection of the hypothesis that children prefer formant transitions because they are signal components that adhere to principles of auditory scene analysis. [Work supported by NIDCD Grant No. DC-00633.]