A phonetic and phonological analysis of final devoicing in Turkish

Abstract

Including the current study, the phonetic realization of devoiced and voiceless final obstruents has been investigated in 5 languages, German, Polish, Russian, Catalan, and Turkish. In terms of orthograp- hic representation, these five languages can be divided into to basic classes: languages which reflect final devoicing in the orthography (Catalan and Turkish), and languages which do not reflect final devo- icing in the orthography (German, Polish, and Russian). The two lan- guages which reflect final devoicing in the orthography show comparab- le phonetic results. In all three studies of Catalan (Dinnsen and Charles-Luce, 1984; Charles-Luce and Dinnsen, 1987; Charles-Luce, 1987), as in the Turkish data reported here, the overall acoustic differences between devoiced and voiceless obstruents were not signi- ficant (except that a 1 ms difference was found to be significant for one parameter in one of the Catalan studies)1. In the perceptual experiments, overall performance was at chance level. Thus, the Cata- lan and Turkish data suggest that final devoicing is neutralizing in these two languages. Large acoustic differences between devoiced and voiceless obstru- ents have been reported only for the languages which maintain a final orthographic viocing distinction. However, the phonetic studies of these languages can be subdivided in to two groups based on their experimental design: studies which controlled for an orthographic influence, and studies which did not. This subdivision is necessary in reevaluating the findings for these languages as the claims of the two groups are substantially different. In general, temporal differen- ces between devoiced and voiceless obstruents have been reported for studies which did not eliminate the orthographic influence. Such findings have led to the criticism that thr differences were an arti- fact of the orthography. Some support for this criticism is provided by studies which failed to find temporal differences between devoiced and voiceless obstruents when the experimental paradigm controlled for the possible orthographic influence (e.g., Fourakis and Iverson, 1984; Jassem and Richter, 1989).