Bilinguals automatically access both languages based on "irrelevant" acoustic detail: evidence from lexical tone

Bilinguals simultaneously activate words from both languages that are consistent with the input, even in contexts/tasks which emphasize only one language. However, it is unknown whether/how the speech perception system integrates or ignores "irrelevant" acoustic cues which might constrain bilingual parallel activation. We leverage the opportunity of Mandarin-English bilinguals in which Mandarin uses an acoustic cue to contrast words (lexical tone) that is not relevant in English. For example, Mandarin distinguishes /bei/ with a high tone (cup) from the same word with a falling tone (quilt). Thus, when a bilingual hears bay with a falling prosody do they use the prosodic information embedded in 'bay' to preferentially activate /bei4/ or they simply ignore this cue because it is irrelevant in identifying 'bay'? Experiment 1 tested this hypothesis with English words of naturally falling prosody and Mandarin competitors that either matched the prosodic form (Tone 4) or did not (Tones 1-3). Participants fixated competitors more frequently than control distractors only when the target words contained the prosodic forms matched with the Mandarin competitors. Experiment 2 replicated this finding with English words superimposed with Tones 1-3 but only found the effect when English stimuli were strongly Mandarin-like. Experiment 3 replicated Experiment 1 by showing no difference between natural tokens and tone-synthesized ones. Together, these results demonstrate that bilingual parallel activation is automatic and that lexical tone, irrelevant to the present language and task, is obligatorily interpreted in real-time processing, such that both tones and segments are required in the acoustic input.