Résumé :
This thesis explores the relationship between neural and linguistic alignment for word meaning, examining whether prediction drives alignment. Linguistic alignment refers to the tendency of conversational partners to synchronize mental states, words, and concepts (Clark, 1996; Pickering & Garrod, 2004). Neural alignment occurs when brain activity in speakers and listeners becomes correlated (Stephens et al., 2010; Hasson et al., 2012). Despite the evidence for both forms of alignment, whether they represent the same phenomenon remains unclear. The first objective of this thesis is to investigate whether neural alignment reflects linguistic alignment by examining whether it is sensitive to specific linguistic knowledge, such as word meanings. The second objective is to explore whether prediction enhances both linguistic and neural alignment, since it has been proposed as a key mechanism of alignment (Pickering & Garrod, 2013; Friston & Frith, 2015; Pickering & Gambi, 2018). Although predictive processing has been well-studied in individuals, its role in dyadic interactions is less understood. To address these goals, a novel experimental paradigm using EEG hyperscanning and interactive language games was developed. Participants engaged in simple semantic association tasks with or without predictability manipulation. This setup allowed investigation of (1) neural alignment for word meaning and (2) the effect of prediction on meaning-specific alignment. Results demonstrated neural alignment sensitive to word meaning and showed that prediction facilitates earlier emergence of meaning-specific neural alignment. These findings suggest neural alignment could be the brain’s signature of linguistic alignment, and offer a potential neural implementation of the interactive alignment model, highlighting the role of prediction in conversational behavior.