Search Abstracts | Symposia | Slide Sessions | Poster Sessions
Do Heritage Bilinguals Predict Speech Sounds? ERP Evidence from Ambiguous VOT Perception
Poster Session F, Friday, October 2, 2:45 - 4:45 pm, Wangari Maathai
Sarah Wang1, Tamara Swaab1,2; 1University of California, Davis, 2University of Birmingham, United Kingdom
Existing research suggests that the brain predicts upcoming linguistic content during comprehension. These predictions can operate at multiple levels, from meaning-based expectations about upcoming words to more specific expectations about word form (Kuperberg & Yaeger, 2016). In spoken language, this raises the possibility that language context may shape not only which word listeners expect, but also how they perceive the speech sounds that unfold over time. Getz and Toscano (2019) used the auditory N1 component, which varies as a function of voice onset time (VOT), to show that predictive processing can influence speech perception at the phoneme level. However, prediction may operate differently in bilinguals, because the mapping between acoustic cues and phonological categories can vary across languages, and bilingual listeners may need to manage competing phonological representations during speech perception. The present study asked whether highly proficient Spanish-English heritage bilinguals (N=25) used lexical-semantic expectations to shape perception of ambiguous English VOT contrasts, and whether this bias is reflected in early auditory processing, indexed by the N1, or later semantic processing, indexed by the N400. Participants saw visual primes followed by auditory targets whose word-initial consonant was either unambiguous or ambiguous. In the unambiguous condition, targets began with clearly voiced (short VOT) or voiceless (long VOT) English plosives (b/p, d/t; g/k) and could be related or not to the prime (e.g., brown- bear, lamp-bear). In these trials, prime-target relatedness was defined by the actual spoken target word. In the ambiguous condition, targets began with a phoneme whose VOT was ambiguous between two phoneme categories (e.g., brown- #ear, and lamp #ear, where # indicates an ambiguous onset that could be perceived as bear or pear). Thus, for ambiguous targets, relatedness depended on whether the prime generated an expectation and whether participants’ phoneme-identification response was consistent with the predicted or alternative interpretation. After each target, participants selected which initial consonant they heard. Statistical analyses (lmer) focused on the mean amplitude of the N1 and N400 effects to the auditory targets. If heritage bilinguals use lexical-semantic expectations to predict phoneme-level information, ambiguous targets perceived as short VOT should elicit larger N1 amplitudes than those perceived as long VOT, paralleling the unambiguous VOT effect. Similarly, N400 effects of relatedness in the unambiguous condition and perceived relatedness in the ambiguous condition should be comparable. Results provided mixed evidence. In unambiguous targets, N1 amplitudes were larger for short- than long-VOT consonants (p <.05), but this effect was absent for ambiguous targets classified by perceived phoneme (p >.05). In contrast, N400 effects of relatedness and perceived relatedness were both significant (p’s < .05). These findings suggest that heritage bilinguals used lexical-semantic expectations to support later meaning-based interpretation of ambiguous speech, but did not show evidence of prediction-based pre-activation of phonological information as indexed by the N1.
Topic Areas: Multilingualism, Meaning: Lexical Semantics