Variation in the production of French vowels: physiological constraints and communicatives demands
Résumé
Speech production is known to vary according to speaking style. One possible explanation for segmental reduction in spontaneous speech is that speakers adapt their pronunciation to the amount of contextual information available and/or to the communicative need of the production situation (Lindblom 1990, Meunier et al. 2005). The objective of our study is to further investigate the ability shown by speakers to adapt their speech to specific production situations, by comparing the acoustic realization of French oral vowels in a reading task vs. in spontaneous productions. Two populations are compared. The first population includes 10 healthy French speakers, for which we expect an adaptation to the different communicative demands in the two speech conditions. The second population includes 28 speakers with dysarthria due to Parkinson's disease (8), Amyotrophic Lateral Sclerosis (12), and cerebellar ataxia (8). In this population, vowel production is affected by strong physiological constraints but it is unknown whether these constraints impede the speaker ability to adapt to communicative demand. It is also unknown how the patterns of variation in healthy speech, which are commonly described as vowel reduction, can be compared to the reductions that are primarily conditioned by motoric restrictions in dysarthric speech. To answer these two questions, the acoustic realization of the French oral vowels /i, e, a, o, u/ produced by both populations in the two speech conditions are compared using the same set of acoustic metrics. Those metrics (Fougeron & Audibert 2011) are based on simple F1 and F2 formant measurements, and are meant to capture variation in many possible dimensions, such as changes in articulator mobility/displacements (size of the vowel space, displacement in F1 or in F2 dimension), changes in target stability (within category variability), and changes affecting linguistic contrast in the system (overlap between categories, centralization of vowel targets). Preliminary results on the ALS speakers show that despite large inter-speaker variability, the dysarthric population does not pattern as the healthy population: in spontaneous speech, they do not show a systematic reduction of their acoustic space, nor a centralization of acoustic target in F1 dimension, nor vowel shortening. Results on the other two other dysarthic populations will be presented at the conference.