Interaction of cues to vowel identity and consonant voicing: Cross-language perception

Abstract

Canadian English has two high front vowels differing in spectral and duration properties, Spanish has one high front vowel, and Japanese has two high front vowels differing in duration only. Vowel duration is a major cue to post-vocalic consonant voicing in English, but not in Japanese or Spanish. Canadian English, Japanese, and Mexican Spanish listeners identified members of a multidimensional edited speech continuum covering the English words bit, beat, bid, bead. The continuum was created by systematically varying the spectral properties of the vowel, and the durations of the vowel, the consonant closure, and the carrier sentence. English listeners had a categorical cutoff between /i/ and /I/ based primarily on the spectral properties of the vowel. Half the English listeners identified consonant voicing using vowel duration. Japanese listeners had a categorical cutoff between the English vowels based primarily on the duration of the vowel. The location of the cutoff was the same as the categorical cutoff between Japanese long /i:/ and short /i/. Japanese listeners identified consonant voicing at random. Spanish listeners identified the English vowels using vowel duration but did not have a categorical cutoff. Half the Spanish listeners identified consonant voicing using the spectral properties of the vowel.