Full text loading...
Abstract
This study aimed to test if morphological complexity (MC) could predict L2 English writing quality (L2WQ) and if the predictive power of MC would differ under integrated and independent task conditions. The data sources of the study were 234 integrated and 639 independent reliably-scored argumentative essays. Sixty-six MC indices were calculated for each text in the corpora, and the strongest possible prediction models were obtained using ElasticNet regression with recursive feature elimination and cross-validation. The results showed that the best model for integrated L2WQ was obtained with 13 MC indices, explaining almost half of the variance. For independent L2WQ, the best model could explain slightly less than one tenth of the variance with 13 MC indices. The findings underscore the need to devise an L2 MC development framework across language proficiency levels and to include MC as a component in the assessment of L2WQ. Additionally, the findings emphasize the necessity to increase morphological awareness among L2 learners.
Article metrics loading...
Full text loading...
References
Data & Media loading...