Speech perception at its best: Extracting linguistic information from acoustically underspecified input. The case of singing

Deme, Andrea (2015) Speech perception at its best: Extracting linguistic information from acoustically underspecified input. The case of singing. In: 18th International Congress of Phonetic Sciences, August 2015 , Glasgow.

Preview

Text Deme_ICPHS0168.pdf Download (463kB) | Preview

Abstract

High-pitched sung vowels are “underspecified” due to i) the tuning of the F1 to the f0 accompanying pitch-raising, and ii) the wide harmonic spacing of the voice source resulting in the undersampling of the vocal tract transfer function. Therefore, sung vowel intelligibility is expected to decrease as the f0 increases. On the basis of the literature of speech perception it is often suggested that sung vowels are better perceived if uttered in CVC context (than in isolation) even at high f0, but the results for singing are contradictory. In the present study we further investigate this question. We compare vowel identification in sense and nonsense CVC sequences and show that the positive effect of the context disappears if the number of legal choices is similar in both conditions, meaning that any positive effect of the CVC context may only stem from the smaller number of possible responses, i.e. higher probabilities.

Actions (login required)

Edit Item