A perceptual study of the voiced-voiceless contrast in syllable-final stops

Abstract

Computer editing techniques were used to remove voicing in 10-ms steps from syllable-final closure intervals of natural tokens of [pɛb,pɛd,pɛg, pag, pig, pug]. Release bursts were retained and a pitch-synchronous editing routine was used to equalize vowel durations at about 110 ms. Identification results from 23 listeners showed that, in general, relatively large amounts of voicing had to be removed from the closure intervals before perception changed from voiced to voiceless. For some of the continua (e.g., /pɛb-pɛp/), voiceless responses began to predominate when more than 95% of the closure interval was rendered voiceless. For other continua (e.g., /pag-pak/), the change from voiced to voiceless responses did not occur until voicing during the closure interval and a portion of the VC transition were removed. We found no place effect for the syllable-final stop; a significant, but relatively small effect for the vowel context was observed. In a second experiment, 11 listeners identified stimuli from which the release bursts had been excised. No significant differences were found between identification boundaries with or without bursts. Findings are discussed in relation to the nature of voicing cues in syllable-final stops. [Work supported by NIH: T32 NS 07100-04 and NIH Biomedical Sciences Support Grant: 5 S05 RR07028.]