A resource-sensitive multifactorial account of working memory effects on multilingual structural priming

Recent studies showed that structural priming may rely on cognitive resources such as working memory (WM). However, it’s unclear whether WM effects differ across languages of varying similarity. We studied Mandarin-Cantonese-English multilinguals in three structural priming experiments by using a digit-recall task between the prime and target to induce high and low WM load. In within-language processing (Experiment 1), both abstract structural priming and lexical boost effect were stronger under low WM load than high WM load. In between-language processing (Experiments 2 and 3), when the prime and target involved different actions, structural priming was similar between different languages (Cantonese-to-Mandarin vs. English-to-Mandarin) and between different WM load conditions (high vs. low). When prime and target involved same action, structural priming and translation-equivalent boost were stronger in the high-similarity condition (Cantonese-to-Mandarin) and under low WM load. These findings support a multifactorial implicit learning framework in which structural priming arises from the joint influence of shared syntactic representations and domain-general cognitive resources, positioning multilingual syntactic processing as a dynamic, resource-sensitive system.