UC Davis

A prominent idea in the study of language processing is known as Predictive Processing. This theory posits that we actively generate predictions for upcoming language input and that these predictions facilitate language processing. While this theory has been well studied with native speakers, much less is known about prediction in bilinguals despite the global ubiquity of bilingualism. Moreover, the literature on L2 Predictive Processing is mixed, with some studies suggesting that bilingual language users do not predict in their L2 and others suggesting that prediction in L2 is delayed. This dissertation aims to investigate whether native and non-native speakers of English differ in 1) the magnitude, 2) timing, and 3) content of predictively pre-activated information.Chapter 1 provides an overview of the literature on prediction during lexical-semantic processing and sets the stage for the empirical work by outlining open questions about prediction in a second language.Chapter 2 uses EEG/ERPs to examine the effects of accurate predictions and semantic relatedness on processing in the L2 of Spanish-English bilinguals and in Native English speakers. Data were collected as participants engaged in a two-word priming paradigm with a prediction task. Both groups showed clear N400 facilitation for accurately predicted words and semantically related words, and PNP effects to unpredicted related trials. In both groups, the effects of prediction accuracy preceded those of semantic relatedness, demonstrating that effects of prediction are separable from semantic matching effects. However, bilinguals showed a slight delay in the onset of both effects, suggesting that while they engage similar predictive mechanisms as monolinguals, they may maintain competing alternatives for longer, providing greater flexibility to adapt when predictions are not met. Overall, these findings indicate that predictive processing operates similarly across groups but with subtle timing differences that may reflect adaptive strategies in bilingual comprehension.Chapter 3 extends these findings by testing how specific predictive processing is with respect to different levels of linguistic representation. Namely, we investigate whether bilinguals and monolinguals pre-activate semantic features (concreteness), lexical wordform features (orthographic neighborhood), and lower-level visual features (word length) of the target word during processing. The findings showed that both groups predicted semantic and orthographic features prior to the target word, whereas evidence of prediction of visual features was less consistent. These results demonstrate that predictive processing operates at multiples levels of representation as suggested by hierarchical models not only for monolinguals but also for bilinguals in their L2. To our knowledge, this is the first study to investigate anticipation of different levels of representation during L2 processing using EEG.Chapter 4 provides a general conclusion about bilingualism and prediction given the main findings from prior chapters and discusses avenues for future research on these topics. Together, these findings indicate that bilinguals can engage predictive mechanisms in a similar way to monolinguals, while subtle temporal differences highlight the need to further investigate how language experience shapes the dynamics and flexibility of prediction. These results support hierarchical models of prediction and highlight conditions under which bilinguals predict in a similar manner to monolinguals. Future research should examine the role of factors such as proficiency, language dominance, and cognitive control in shaping these processes, as well as explore the consequences of prediction success and failure for learning and memory in bilingual contexts.