Abstract :
[en] In this study, we present data about the amount of nasalization in contextually nasalized French oral vowels and we discuss results in connection with Kingston and Diehl's hypothesis about phonetic control (Kingston and Diehl, 1994).
The selected dependent variable, i.e. the mean proportion of nasal to total airflow and volume, varies significantly according to the oral vowel considered. More specifically, proportional nasal airflow (PNA) of oral vowels that have a phonological nasal counterpart is nearly twice as small as PNA of other oral vowels. We examine some constraints that could account for the observed variation. The lack of a satisfactory explanation leads us to consider whether this phonetic variation in the implementation of the feature [oral] could be due to an active control of velum activity.
Eight native French speakers (from Belgium) took part in the experiment. Their task was to read lists of words containing the stimuli. Nasal and oral airflow were recorded with the Physiologia Workstation. Corpus contained 198 items with a NVN, NV or VN structure where N is one of the three French nasal consonants and V one of the ten French oral vowels. The PNA variable has the advantage of neutralizing the differences due to gender as well as the effects of changes in overall intensity.
The three main results are the following : PNA varies significantly according to (1) nasal consonant context type, (2) existence (or not) of a phonemic nasal counterpart to the oral vowel, and (3) vowel tongue height.
As regards context, progressive nasalization was found on average to be five times as large as regressive nasalization. Oral vowels preceded and followed by a nasal consonant are even more nasalized. However, an identical pattern occurs within each context : oral vowels can be divided into two groups : [a,E,9,O], the vowels that have a phonemic nasal counterpart [a~,e~,9~,o~], have a much smaller and steadier PNA than the remaining [e,2,o,i,y,u] (see Table 1).
By reason of the well-known diachronic evolution of French nasality, high nasal vowels do not exist in this language. Thus, the 'nasal counterpart' variable interacts with the tongue height variable. Nevertheless, the relationship between PNA and tongue height is not linear. In the first group, the lowest vowel /a/ has consistently more PNA than the three low-mid ones whereas within the second group (vowels without phonemic nasal counterpart), high-mid vowels are less nasalized than high ones (see Table 2).
Actually, high-mid vowels show PNA values much closer to low-mid than high vowels. This should not be surprising since the following facts are taken into account: (1) in most languages, mid vowels (either high-mid or low-mid) have a common and typical behavior with respect to nasalization, (2) in standard French, the phonological contrast between high-mid and low-mid vowels is neutralized in final position; most of the time, this is not true for Belgian speakers but there happens to be more variability in the realization of these minimal pairs in this specific context. In our data, the context variable interacts significantly with the relationship between high-mid and low-mid vowels. Finally (3), at least for the O/o pair, the vocal tract configuration of the nasal vowel (o~) is closer to the high-mid than the low-mid oral vowel in modern French. Both phonemes present rather similar values for PNA in our data while the E-e and 9-2 differences are much greater.
The results presented above do not provide an explanatory variable (i.e. a phonetic constraint) that would predict the observed variation in the amount of nasalization for contextually nasalized oral vowels with as much accuracy as does the phonemic nasal counterpart variable. Neither the coarticulatory source of nasalization type nor the features of the vocalic phoneme account for such a variation.
Thus, we propose to consider Kingston and Diehl's hypothesis about phonetic control. Control of the [oral] feature phonetic implementation in nasal context implies an active reduction of the amount of nasalization when a contrast with a nasal vowel has to be maintained, in order to maximize the acoustic and perceptual distinctiveness of [nasal] vs [oral] phonetic realization. Otherwise, control deliberately allows a greater amount of nasalization, which minimizes production effort.
