Acoustic measures of glottalization in native and nonnative English connected speech

Abstract

Using perceptual coding methods, previous research on English connected speech by late-onset second language speakers (L2ers) found that L2ers produced unstressed vowel-to-vowel word junctures (e.g., go héad) as “delinked” (i.e., glottalized) at significantly higher rates than native speakers (L1ers), which could not be explained by cross-linguistic influence (CLI) from the L2ers' native language (L1). In the present study, we examined whether L2ers differed from L1ers also in degree of glottalization. Comparing high-proficiency L2ers from an L1 Mandarin background (N = 5) with L1ers (N = 5), we collected four measures of spectral tilt (H1–H2, H1–A1, H1–A2, H1–A3) at vowel-to-vowel word junctures spontaneously produced in interview speech (N = 615 for L1ers, N = 841 for L2ers). Results showed that although junctures coded as glottalized tended to show more negative H1–H2 values (indicative of creaky phonation) than did junctures coded as not glottalized, there were no significant differences between L1ers and L2ers in any of the spectral tilt measures examined, for either glottalized or non-glottalized junctures. These findings thus suggest that although high-proficiency L2ers may glottalize vowel-to-vowel junctures at higher rates than L1ers, when they do glottalize they do so to a similar degree as L1ers, with no apparent CLI from the L1.