Presentation

Agnes Gao^1,2, Tamara Swaab^1,2, Matthew Traxler^1,2; ¹University of California Davis, ²Center for Mind and Brain

Monolinguals engage in context-based predictive processing, which pre-activates the semantic and perceptual features of the upcoming word (Brothers et al., 2016). Getz & Toscano (2019) showed that specific top-down lexical predictions can pre-activate phonological features prior to word onset (e.g., voice onset time), and can override the bottom-up ambiguous input. In particular, the N1’s sensitivity to an ambiguous /b-p/ sound disappeared when listeners were expecting to hear /p/ as in “park” following the highly associated prime word “amusement”, and moreover, the N1 to an ambiguous /b-p/ sound resembled the N1 to an unambiguous voiceless /p/. In comparison, bilinguals have limited predictive abilities, and past research has shown that pre-activation in bilinguals only occurs at the semantic level. This could be because bilinguals need a global language context to select the contextually appropriate language to pre-activate more specific word features other than meaning (Hoversten & Traxler, 2016). To investigate this hypothesis, we will test bilingual pre-activation of phonological features in single word context versus in sentence-level context. In Experiment 1, I will replicate the cross-modal priming paradigm of Getz et al. (2019) in proficient Spanish-English bilinguals. The prime will be presented visually, and the target will be presented acoustically. All targets will have stop consonants as onsets, and the VOTs will be manipulated to be short, intermediate (ambiguous) or long. There will three conditions: 1) association: the prime will be highly associated with the target (e.g., amusement PARK); 2) neutral: the prime will not be associated with the target (e.g., finger PARK); 3) masked: the prime will be a non-word letter string (e.g., xxxx PARK). I will measure the N1 to index VOT perception and the N400 to index the comprehension of semantics. I expect bilinguals to show sensitivity to VOT changes in the N1 in all three conditions, since VOT will not be pre-activated at the word level; however, bilinguals will show slightly reduced N400 effect to the ambiguous target word in the association condition, since word meaning will be pre-activated. In Experiment 2, each associated word pair will be embedded in a critical spoken sentence, preceded by a high or low constraint written discourse context. When provided with a high constraint context and a global English context, proficient bilinguals will show native-like patterns. Specifically, they will no longer show the linear N1 sensitivity to VOT changes, as the phonological feature will be pre-activated, which will override the bottom-up ambiguous input. Also, the N400 to the ambiguous target will completely disappear, as bilinguals will have predictively comprehended the word meaning. When provided with a low constraint context, bilinguals will not benefit from semantic context, and therefore, they won't pre-activate VOT, but they will pre-activate meaning. Overall, when only given a single priming word, bilinguals would not be able to pre-activate features other than semantics; however, when given a constraining sentence context and an appropriate global language context, bilinguals will be able to select the relevant language to engage in predictive processing to the same extent as monolinguals do.

Topic Areas: Multilingualism, Speech Perception