Abstract
Individual variability of cognitive factors in second language (L2) grammar learning has long been the focus in the field of L2 acquisition. Most explored the issue by focusing on one factor like cognitive control, working memory, statistical learning (SL), or attention. Few investigated the topic by taking all these factors into consideration. However, different factors might interact and collaboratively contribute to the learning process. Examining the issue by considering all the factors might yield different results and facilitate our understanding of the mechanism subserving L2 grammar learning. Therefore, this study explored whether and how these factors predicted L2 grammar learning. A total of 34 college students completed a set of cognitive measurements on these cognitive factors, after which they were trained with artificial grammar over 5 consecutive days. Using multiple regression analysis and machine learning algorithms, we found that in the initial phase, SL was the more significant predictor, whereas in the intermediate and the last phases, cognitive control served as the more significant predictor. In other words, in the initial phase of L2 grammar learning, SL might play an important role, whereas in the intermediate and proficient phase, the updating component of cognitive control might play a more significant role. The findings provided empirical evidence to the neurocognitive account of grammar learning, shedding light on the mechanism of L2 grammar learning.