Disambiguation of Segmental Dependencies by Extended Phonetic Context

Abstract

This research investigated listeners' ability to recognize adjacent vowels and consonants that are conveyed in part by a common temporal cue – vowel duration. The stimuli consisted of a large sample of natural speech containing nonsense syllables made by combining four vowels that differed in inherent duration (/i/, /ε/, /i/, /e/) with two syllable-final consonants that differed in phonological voicing (/z/ and /s/), a distinction that is partially cued by vowel duration. Experiment 1 examined the effect of vowel identity, fricative identity, and contextual factors on vowel duration, voiced-frication duration, and voiceless-frication duration in speech production, and the potential effectiveness of these durational measures as classification cues. In Experiment 2, listeners identified the syllables after they had been excised from sentential context. Accuracy in recognizing /s/ varied with the inherent duration of the preceding vowel, with the number of errors increasing with inherent vowel duration. This suggests that listeners did not sufficiently account for the effect of vowel identity when interpreting vowel duration as a cue to syllable-final voicing. In Experiment 3, listeners heard the syllables in sentential context. Accuracy improved, and no relation was found between inherent vowel duration and accuracy in recognizing /s/. This indicated that the sentential context of a syllable conveys information that is useful to listeners in correctly determining the segmental basis of the internal temporal structure of a syllable.