A Unified Model of Mono- and Bilingual Intelligibility

The construct of intelligibility in L2 speech has primarily been operationalized functionally in terms of speech being classified as intelligible if the listeners successfully recovered the intended message (Munro & Derwing, 1995). In this paper, I will operationalize intelligibility psycholinguistically in terms of spoken word recognition. We do not need to invoke any special machinery for intelligibility in bilinguals; monolinguals and bilinguals process speech in the same way (Libben, 2000; Libben & Goral, 2015). Listeners have to segment the speech stream and the parser maps the phonetic elements onto higher-level linguistic representations such as phonemes, syllable nodes and metrical feet. The role of experience in the listener is modelled analogously to high-variability phonetic training (HVPT) via broadening the prior likelihood (in a Bayesian sense) of the mapping of an L2 phone onto an extant phonological category. I conclude by discussing pedagogic implications, and suggesting that pedagogic models that advocate a single non-native variety of English, which will be intelligible to all ears (i.e. parsable by all grammars), are problematic psycholinguistically.