Russian, Japanese, and Spanish late second language learners’ perception of American vowels produced in sentences with varying consonantal contexts: Perception in the phonological mode.

Abstract

We examined native Russian‐, Japanese‐, and Spanish‐speaking L2 learners’ perception of American English (AE) vowels produced in nonsense trisyllables (gaCVCa) embedded in a sentence. Vowels in 4 contexts (bVb, bVp, dVt, dVp) in the phrase “five gaCVCa” excised from “I said five gaCVCa this time” were presented in a categorial ABX task in which the AB stimuli were both gabVba or both gadVpa, and X contained either the same or a different consonant context. Thirteen pairings of eight vowels [i:, ɪ:, ɑ:, ε, æ:, ʌ, u, u:] constituted three types of contrasts: four Control pairings (non‐adjacent height contrasts), four spectral + duration contrasts (four short‐long adjacent height contrasts), and five spectral only contrasts (three front‐back contrasts and two height contrasts between short vowels). The native AE listeners’ performance was quite accurate overall on the nine experimental contrasts (89% correct on same‐context trials, 79% on different‐context trials). Non‐native groups did more poorly, especially on experimental contrasts in different‐context trials (Russian 73%, Japanese 67%, and Spanish 54%) and were poorest at differentiating mid‐low and low vowels: [æ:‐ɑ:, ɑ:‐ʌ, ε‐ʌ]. Comparisons of results with perceptual assimilation patterns, discrimination of citation /Vpə/ disyllables, and with spoken English proficiency (Versant test) are discussed. [Work supported by NSF.]