Generating iso-accented stimuli for second language research: methodology and a dataset for Spanish-accented English

A non-native accent can be conveyed at both the segmental and suprasegmental level. Previous studies have developed techniques to isolate the effect of segmental foreign accent by splicing accented segments from a bilingual speaker into non-accented words produced by the same speaker. The current work addresses the issue of between-segment variability by developing a technique to convert from acoustically-equal accent gradations to perceptually-equal steps. The procedure is used to derive the first corpus of Spanish-accented English composed of lexical tokens each generated with one of five degrees of nonnative accent. As an example application, corpus tokens are used to elicit accentedness judgements from four listener cohorts with first languages which differ as to whether they share the native language, the non-native (accented) language of the corpus or have a closer phonological inventory to one or the other. Findings highlight the importance of the relationship between listeners' phonological systems and those of the native and non-native languages of the corpus, especially for vowels, with respect to sensitivity to foreign accent.