Phonetic vs. phonological aspects of voicing
Eugeniusz Cyran
Catholic University of Lublin
The paper discusses the role of phonetic grounding in establishing phonological categories (features) with a specific application to voicing of consonants. It appears that such phonetically defined properties as vibration of vocal folds cannot be automatically assumed to correspond to discrete phonological categories. A "hard-core" phonologist may prefer to establish melodic primes independently of articulatory or even auditory indications, for example, by applying the following two criteria:
a) phonological primes should express only relevant contrasts
b) phonological primes should be established on the basis of phonological behaviour
While the first criterion excludes the presence of a phonological property [voice] in sonorants, as this would suggest a contrast with voiceless sonorants, the second criterion provides us with the dilemma concerning the nature of primes. That is, whether they are equipollent: [+voice] implies the existence of [−voice] , or privative: the representation of laryngeal contrasts is asymmetrical.
The choice between the two systems leads to different predictions. For example, the symmetrical (equipollent) system will neatly capture symmetrical phenomena involving voice, e.g. regressive voice assimilation in Polish, but it will struggle with the overtly asymmetrical phenomenon of devoicing (final voice neutralisation). Logically, the privative system will be best suited for such asymmetrical phenomena, and will find symmetry problematic.
Typological facts concerning laryngeal specification as well as markedness tendencies strongly support the privative systems, which, surprisingly also finds some support in the realm of phonetics, though acoustic and not articulatory. It will be demonstrated that if pushed to its logical conclusion an asymmetrical (privative) system makes interesting predictions concerning linguistic variation, but it also must further detach itself from phonetics as a consequence.