Neutralization of obstruent voicing in german is incomplete

Abstract

A familiar phonological neutralization rule is the merger of voiced and voiceless obstruents in syllable-final position in German. Thus, the /d/ in Leid ‘sorrow’ is pronounced [-voice] like the /t/ in leit ‘lead.’ A spectrographic test of minimal pairs by nine Germans, however, revealed (1) vowels before underlying voiced consonants were 10% longer than before underlying voiceless ones, (2) there was a slightly longer interval of glottal pulsing into the consonant closure for the underlying voiced consonants, yet (3) consonant closure durations were the same. Since these consonants all sounded voiceless to us, a listening test was performed. Native speakers of German listened to these utterances and indicated which word of the minimal pair was heard. Subjects were able to guess the intended word about 70% correct—significantly better than chance. This “semicontrast” must be nearly useless in conversation. These results show that voicing neutralization is only partially achieved in German and disprove any analysis of the form [+voice] → [−voice]. Instead we propose that Germans employ a “motor implementation rule” of final devoicing that simulates the voiceless feature but is distinct from it. [Research supported by NIH grant HD12511.]