Abstract
In large scale low stakes assessments, students usually choose their own speed at which to work on tasks. At the same time, previous research has shown that in hard tasks, the time students invest is a positive predictor of task performance. From this perspective, a relevant question is whether student dispositions other than the targeted skill might affect students' time on task behavior, thus potentially affecting their task performance and in turn their estimated skill in the target domain. Using PISA 2009 computer based assessment data, the present research investigated for the domain of reading digital text whether three variables that can be assumed to predict performance in digital reading tasks, comprehension skill, enjoyment of reading, and knowledge of reading strategies would also predict how much time students would devote to digital reading tasks, and in particular, whether they would adapt time on task to task difficulty. To address this question, two linear mixed models were estimated that predicted the time students spent on a task, and the average time students spent on relevant pages within each task, by the interaction of task difficulty with comprehension skill, enjoyment of reading, and knowledge of reading strategies. To account for time on task being nested in students and tasks, random effects for persons and tasks were included. The interaction of task difficulty with gender and Socio-Economic Status (SES) was included for control purposes. Models were estimated individually for 19 countries, and results integrated meta-analytically. In line with predictions, for both time on task indicators, significant positive interactions were found with comprehension skill, enjoyment of reading, and knowledge of reading strategies. These interactions indicated that in students with high comprehension skill, enjoyment of reading, and knowledge of reading strategies there was a stronger association of task difficulty with time on task than in students low in either of these variables. Thus, skilled comprehenders, students enjoying reading, and students in command of reading strategies behaved more adaptively than lower skilled, motivated, or knowledgeable students. Implications of these findings for the validity of self-paced computer-based assessments are discussed.