Retour vers liste

Détail de la contribution


Segmental and prosodic effects on voicing assimilation: a gestural blending account

Abstract/Résumé: This paper presents new data to argue for the re-analysis of a well-established topic in Spanish phonology, the voicing assimilation of /s/ before a voiced consonant ([rázgo] vs. [rásko]), as the result of gestural blending rather than of a categorical voicing feature change. We present acoustic data from two experiments to show that this assimilation, albeit gradient, is conditioned by certain prosodic and segmental factors, and that it can be best modeled as the result of gestural blending. Gradient assimilatory effects have been analyzed as the result of increased overlap among adjacent gestures. Gestural blending is the result of increased gestural overlap between two gestures specified for the same articulator–voicing assimilation results from blending of two overlapping laryngeal gestures. This study tests the effect on voicing assimilation of two prosodic factors, stress and prosodic boundaries, shown to affect gestural overlap: we expect more assimilation when stress falls on the syllable following /s/ than on the syllable containing it, and the degree of assimilation is expected to decrease moving to higher prosodic boundaries. Our first experiment tests the degree of voicing assimilation of /s/ preceding a voiced obstruent in Northern Peninsular Spanish under different stress conditions (stressed vs. pre-stressed) and prosodic boundaries: word internal vs. word boundary vs. intonational phrase boundary. Three acoustic cues to voicing were measured: preceding vowel and fricative duration, and voicing during frication. The percentage of voicing during frication was used to categorize each token as unvoiced, partially voiced or fully voiced. The results show variation in the degree of voicing. Stress does not condition assimilation, but prosodic boundary type does influence the degree of assimilation in the predicted direction. We further identified another possible conditioning, i.e., manner of articulation of the following stop. All the voiced stops following /s/ were realized as approximants. We coded the manner of the following consonant as open approximant or close approximant (Martínez-Celdrán 1991). We find a relation between the voicing category and the following consonant manner with a following open approximant favoring voiced realizations. Given these results, a second experiment was designed to further test the effect of stress (post-stress positions) and of other manners of articulations (sonorants) in the degree of assimilation. Two new sets of stimuli from 8 more Peninsular Spanish speakers were acoustically analyzed as in the previous experiment. Results indicate that post-stressed vs. other positions show an effect on the assimilation, and that sonorants trigger less assimilation than obstruents.