Final Devoicing

This study tests the P-Map’s (Steriade, 2001) hypothesis that attested phonological patterns vary depending on phonological context due to differences in the perceptual similarity of two phonological contrasts in different contexts, and that the knowledge of these relative perceptibilities are used to explain cross-linguistic patterns of phonological typology. Specifically, the phenomena in focus are spirantization and devoicing. The study investigates if preference for spirantization of voiced stops in intervocalic position and devoicing of voiced stops in word-final position across languages is correlated with perceptual distance. Using perceptual similarity tests with native English speakers on pairs of contrasts in nonsense words, the results show that spirantization contrasts are significantly more perceptible in word-final position than in intervocalic position, explaining the preference for spirantization intervocalically while devoicing contrasts are significantly more perceptible intervocalically, explaining the preference for devoicing word-finally. The results thus support the P-Map’s claim and support a phonetics-based approach to phonology. In addition, since neither phenomenon is a process that happens in English, the lack of bias in these similarity judgements lends weight to the idea that these results reflect the universal perception of phonological processes.

Abstract

As Selkirk (1986) states, final devoicing is a pattern of phonological distribution in which both voiced and voiceless obstruents occur in a language, but at the end of a particular prosodic domain only voiceless obstruents occur. The Phonological process of final devoicing has been a well-studied topic dating back to Trubetzkoy in 1933 (translated published in English in 1969) and it has been cited by most phonologists as an example of neutralisation (Brockhaus 1991 inter alia). Many languages disfavour coda voiced phonemes; however, devoicing is subject to parametric variation (Myers 2012). Most languages resolve voiced coda in a similar fashion; by devoicing rather than nasalization, deletion, or epenthesis.
This paper argues that Central Kurdish (hence forth CK) is one of the languages that undergoes final devoicing; it also claims that Kurdish resolves coda voicing by devoicing the voiced coda rather than other means mentioned above. For example, underlying /bəɾd/ ‘stone’ can become [baɾt] but not *[bəɾ], *[bəɾm], *[bəɾdə]. Further, my analysis makes two theoretical claims: First, the prosodic domain within which coda devoicing occurs in CK includes both syllable and Prosodic word. Second, coda devoicing in CK, like many other languages (Steriade 2001/2008), can pose a hitherto unsolved problem to Optimality Theory. With the total absence of literature on this topic, I will recourse to the intuition of native speakers’ judgement to support the claim that final devoicing is occurring in CK.