EUROSLA Yearbook, 5 (2005), c5–222. issn 1568–1491 / e-issn 1569–9749 © John Benjamins Publishing Company Cognitive task complexity and second language writing performance Folkert Kuiken, Maria Mos and Ineke Vedder University of Amsterdam / Tilburg University / University of Amsterdam Tis paper reports on a study in which two models proposed to explain the influence of cognitive task complexity on linguistic performance in L2 are tested and compared. Te two models are Robinson’s Cognition Hypoth- esis (Robinson 2001a, 2001b) and Skehan and Foster’s Limited Attentional Capacity Model (Skehan 1998, Skehan and Foster 2001). Sixty-two Dutch university students of Italian performed two writing tasks with prompts of differing cognitive complexity. Linguistic performance was operationalized in terms of syntactic complexity, lexical variation and accuracy. Te study provides partial support for the Cognition Hypothesis, in so far as the writ- ten products of the cognitively more demanding task turned out to be more accurate, with significantly lower error ratios per T-unit than those of the cognitively less demanding task. In addition stronger effects of cognitive task complexity were found for high-proficiency learners than for low-proficiency learners. No effects could be observed on measures of syntactic complexity or lexical variation. Introduction Since the introduction of task-based language pedagogy in the 1980’s, the po- tential benefits of using tasks in the second language classroom have interested both teachers and researchers. Language teachers agree that unless learners engage in language activities such as those found in real-world tasks, they may not succeed in developing the kind of L2 proficiency needed to commu- nicate fluently and effectively. Second language acquisition researchers have recognised the importance of tasks for documenting how learners structure and restructure their interlanguages, and how they perform when they are not attending to language form (Long 1985, Crookes and Gass 1993, Skehan 1996, 1998, Willis 1996, Lee 2000).