Time-on-task effects in digital reading are non-linear and moderated by persons' skills and tasks' demands

•Time on task effects in digital reading were positive in weak readers and hard tasks.•Time on task effects in digital reading were null in strong readers and negative in easy tasks.•Time on tasks effects in digital reading had a strong negatively u-shaped quadratic component.•Results fit with a dual-processing account of digital reading.•Results replicate across 19 educational systems from Europe, Asia, America, and Oceania.

Time-on-task effects on response accuracy in digital reading tasks were examined using PISA 2009 data (N = 34,062, 19 countries/economies). As a baseline, task responses were explained by time on task, tasks' easiness, and persons' digital reading skill (Model 1). Model 2 added a quadratic time-on-task effect, persons' comprehension skill and tasks' navigation demands as predictors. In each country, linear and quadratic time-on-task effects were moderated by person and task characteristics. Strongly positive linear time-on-task effects were found for persons being poor digital readers (Model 1) and poor comprehenders (Model 2), which decreased with increasing skill. Positive linear time-on-task effects were found for hard tasks (Model 1) and tasks high in navigation demands (Model 2). For easy tasks and tasks low in navigation demands, the time-on-task effects were negative, or close to zero, respectively. A negative quadratic component of the time-on-task effect was more pronounced for strong comprehenders, while the linear component was weaker. Correspondingly, for tasks high in navigation demands the negative quadratic component to the time-on-task effect was weaker, and the linear component was stronger. These results are in line with a dual-processing account of digital reading that distinguishes automatic reading components from resource-demanding regulation and navigation processes.