Speaker-dependent differences in voicing and devoicing of German obstruents

To investigate the different phonetic parameters of final devoicing an acoustic analysis of German obstruent realizations was carried out focusing on the adequacy of the word material employed and the consistency of the phonetic realizations produced by speakers from different regions of Germany. Three types of sample words were used: (1) words containing (a) voiced obstruents and (b) their devoiced counterparts (i.e., words in which these obstruents appear in syllable- or word-final position), (2) voiceless control items and (3) corresponding words ending in nasals and liquids (dummy items). The test items-embedded into a sentence frame-were read by six native speakers taken from the three main dialectal areas of Germany. They were manually segmented with respect to the durational parameters: vowel duration (VOC), voicing into closure or friction (VC), plosive or fricative duration (CODA). In a MANOVA analysis the durations of VC, VOC and CODA in plosives revealed significant effects of voiced vs. voiceless obstruents in medial (nonneutralizing) position. These differences are neutralized in syllable-final and word-final positions: In utterance-final position VOC and VC are completely neutralized by all speakers, while South German speakers preserve the durational differences in CODA between underlying voiced and voiceless plosives. In fricatives, there is complete neutralization in utterance-final position for all speakers. (C) 2003 Elsevier Ltd. All rights reserved.