| The purpose of the study was to examine the relationships between writing strategy use and second language (L2) writing ability on an academic writing test. This study utilized a new 38-item cognitive and metacognitive strategy use questionnaire and two composition prompts to gather writing strategy use and writing ability data from 214 college and graduate ESL participants. Multi-faceted Rasch measurement was employed on the writing data to determine test score reliability and assessment error associated with the performance test. It was also used to examine the characteristics of and relationships among the test takers, the raters, the prompts, and the scales. Exploratory factor analysis (EFA) was used to identify the latent factors in both strategy use and writing data. Confirmatory factor analysis (CFA) was utilized to examine the relationships between observed and latent process variables. Finally, full-latent variable structural equation modeling (SEM) was used to model the relationships between participants' reported strategy use and writing performance.; The results showed that metacognitive processes had no direct effect on L2 writing ability, but had a significant, positive, direct influence on the comprehending and text generating processes. Comprehending processes, in turn, showed a significant, positive, and direct effect on L2 writing ability, while text generating processes indicated a significant, negative, and direct impact on L2 writing ability. In other words, the more test-takers reported using comprehending strategies, the better they performed; the more they invoked text generating strategies, the worse they performed in the context of a timed writing assessment. The current results also suggested that, in order to succeed on L2 writing tests, test-takers require a control of both cognitive and metacognitive strategies. |
