Mixing costs in task shifting reflect sequential processing stages in a multicomponent task

We investigated the hypothesis that mixing costs in task shifting reflect the sequential selection of task components (e.g., stimulus categories) during task execution. This proposition was supported by Hübner, Futterer and Steinhauser (2001), who showed that the amount of mixing costs depends on the number of mixed task components (e.g., stimulus level and judgment). However, their results could also be explained by a task set selection account, because task components and task sets were confounded. In Experiments 1 and 2, we compared conditions in which either the number of task sets varied and the number of mixed task components was constant or vice versa. Only the number of mixed task components was predictive for the mixing costs. In Experiment 3, we replicated the additivity of mixing costs from level and judgment mixing. Our results suggest that the mixing costs reflect a selection strategy in which interference is reduced in a stepwise manner