Background: Nasality is contrastive for vowels in french-language varieties: speakers differentiate vowel pairs like /E/ (as in tête', eng. _head_) vs. /E~/ (as in teinte', eng. hue, note that we use the SAMPA phonetic transcription system). However, it is usually advocated that such contrasts may involve two distinct relationships between the oral and the nasal member of the pair: a phonological one and a phonetic one. On a phonological account, the 3 nasal vowels {E~, A~, O~} contrast with the 3 oral vowels {E, A, O} respectively. However, considering phonetic grounds, the oro-pharyngeal configuration of the vocal tract in these 3 nasal vowels is much closer to the 3 oral vowels {a, O, o} respectively. Also, acoustic properties of nasal vowels are known to be complex as they involve a coupling between the oral and nasal cavities that gives rise to anti-resonances. These phenomena lead to various acoustic consequences that make the distinction between nasal vowels and their oral counterparts difficult to characterize in natural speech. The aim of the present study is to investigate how various parameters of speech vocoding may influence the perceptual identification of oral and nasal vowels by typical-hearing children and to provide perspectives concerning their recognition in the context of cochear implants.

Method: Vowel perception was investigated in 2-Alternative Forced-Choice tasks involving either an identification or a discrimination procedure among 8 children (all in their 8th year). Nasal vowels were associated with both their phonological oral counterpart or their phonetic oral counterpart. Speech sounds were artificially manipulated in order to systematically remove any duration or fundamental frequency differences between the stimuli while spectral content was preserved. Speech vocoding stimuli were generated using a Pulse-Spreading Harmonic Complex carrier with respectively 4, 12 or 22 channels and either a 0 or 3 mm simulated insertion depth. The original natural speech sequences were used as a control condition.

Results: With 4-channels, performance was similarly low for both insertion depths in the phonetic and the phonological contrast conditions. Though perceptual scores increased with the number of vocoded channels, they were significantly lower for phonetic pairs than for phonological pairs, and this difference was systematically more impactful when a 3 mm insertion depth was added.

Discussion: These results confirm that acoustic information associated with strict nasal resonance are strongly affected by various vocoding parameters and are particularly impacted by insertion depth.