ExLing 2019: Proceedings of 10 th International Conference of Experimental Linguistics, 25-27 September 2019, Lisbon, Portugal Lexical selection and gender agreement processing Masayuki Yamada, Maria do Carmo Lourenço-Gomes Institute of Arts and Humanities, University of Minho, Portugal https://doi.org/10.36505/ExLing-2019/10/0052/000414 Abstract This study investigates the influence of lexical selection of adjectives on gender agreement in writing by late Chinese learners of European Portuguese (EP), with an online paradigm. Participants were asked to translate sentences (from Chinese to Portuguese) containing adjectives that ranged from easy to difficult, on a platform that recorded the writing process. A greater pause was observed in the difficult condition than in the easy condition at the critical position, i.e., just before the adjective. In addition, there were more errors in the difficult condition. These results suggest that when lexical selection is difficult it demands more resources from working memory (WM), and, hence, less resources are left for the agreement process, which results in more errors. Keywords: gender agreement, lexical selection, Portuguese, Chinese Background Acquisition of gender agreement by late L2 learners has generated much research in the literature. There are, however, few studies on gender agreement in EP. Most have examined written corpora data and discussed the relationship between errors and noun features such as morphology and animacy (e.g., Brito, 2015; Pinto, 2014). Despite their relevant findings, their results should be interpreted with caution, because they did not distinguish between attribution errors and agreement errors in the analysis, which renders difficult to discuss the nature of the errors (whether lexical or syntactic). Furthermore, as these analyses were conducted on the final version of the texts, the online process of writing was not examined. For these reasons, Yamada et al. (2018) analysed gender agreement errors in writing by late Asian learners of EP, by using a platform developed by the author (demo: http://bit.ly/2M4FCgR), which enabled the recording and later review of how the writing process unfolded in real time due to a keystroke logging technique (for a review, see Van Waes, Leijten, Wengelin, & Lindgren, 2011). As for noun-adjective agreement, the results suggested an independent relationship between noun’s features (transparency and animacy) and occurrence of agreement errors. Moreover, an analysis of the writing process revealed that a relatively great pause was found just before the occurrence of some errors. A post-interview to the participants unveiled that they were engaged in the lexical selection of adjectives during the