Discrimination of word‐final voicing in German

TLDR: This paper used duplex oscillogram and intensity curves on visicorder tracings to detect the intended voicing of the final consonant of a word in a given word and played it to native German speakers who identified the word they heard.

Abstract: Traditionally, German is described as having complete neutralization of voiced versus voiceless consonants word finally. In this study native German speakers were recorded reading real words differing in the voicing of the final consonant. The recorded tokens were then randomized and played to native Germans who identified the word they heard. The listeners performed very poorly, but significantly better than chance. In addition, temporal acoustic measurements of the recorded tokens were made using duplex oscillogram and intensity curves on visicorder tracings. Statistical analysis of the acoustic measurements—nucleus duration, consonant duration, duration of voicing during closure, and final aspiration—showed significant, though small differences between intended voiced and voiceless pairs. Discriminant analysis applied to the four variables produced an acoustic criterion by which the intended voicing of tokens could be predicted with 63% accuracy, about the same as the best listener was able to do. In addition, data will be presented showing asymmetry between production and perception, in relative importance of acoustic cues and in individual subjects' discrimination performance. [Work supported by NIH.]