The current study explores how multilingual speakers with three typologically different languages (satellite-framed, verb-framed and equipollent-framed) encode and gauge event similarity in the domain of caused motion. Specifically, it addresses whether, and to what extent, the acquisition of an L2-English and an L3-Japanese reconstructs the lexicalization and conceptualization patterns established in the L1-Cantonese when the target language is actively involved in the decision-making process. Results show that multilingual speakers demonstrated an ongoing process of cognitive restructuring towards the target language (L3) in both linguistic encoding (event structures and semantic representations) and non-linguistic conceptualization (reaction time). And the degree of the restructuring is modulated by the amount of language contact with the L2 and L3. The study suggests that learning a language means internalizing a new way of thinking and provides positive evidence for L3-biased cognitive restructuring within the framework of thinking-for-speaking.
