The origin of switch costs: Task preparation or task application?

Abstract

Recent studies have shown that switch costs (i.e., slower responding on task-alternation trials than on task-repetition trials) are not observed when on the preceding trial a no-go signal instructed the participant to withhold the response to the target stimulus. This finding suggests that neither task set is inhibited on no-go trials, and that the origin of switch costs is located in the application of the task set to the target stimulus. However, these studies also showed that responding after a no-go trial is substantially slower than after a go trial. This suggests that both task sets are inhibited on no-go trials and that switch costs originate from the preparation of a task set. In two experiments we created conditions that revealed an absence of switch costs in conjunction with relatively fast responding after no-go trials. Together these findings confirm that switch costs originate from the application of the task set.

Keywords:

• Task switching
• Cognitive control
• Preparation

Acknowledgments

We thank Nachshon Meiran, Marco Steinhauser, and an anonymous reviewer for helpful comments on an earlier draft of this article.

Notes

¹ Taking this point of view, a go trial after a no-go trial could be legitimately referred to as an alternation trial, even when the same task set is cued on those trials. However, to maintain consistency with previous reports, we categorized a task set as repetition or alternation on the basis of the cue, regardless of whether it was applied or not.

² Schuch and Koch (2003) used symbolic cues (squares and rectangles) instead of the verbal cues we used here. Symbolic cues typically lead to greater switch costs than verbal cues (Mayr & Kliegl, 2003), which may reflect that the elaborate processing they require is particularly susceptible to intertrial priming (e.g., Logan & Bundesen, 2003; Schneider & Logan, 2005). Since these priming effects were not of present interest, we tried to minimize them by using verbal cues.

³ Two 3-way interactions approached significance: Task Transition × TTI _n  × TTI _n−1, F(1, 15) = 4.10, MSE = 1,534.38, p = .06, η² = .22; Task Transition × Response _n−1 × TTI _n , F(1, 15) = 4.23, MSE = 1,316.46, p = .058, η² = .22. However, since both these interactions were far from significant in Experiment 1 (F values < 1), we ignore them here.

⁴ For instance, one reviewer suggested that participants obligatorily prepare as soon as the cue is presented. However, on the assumption that on no-go trials the preparatory phase must be concluded by task-set suppression to prevent the task set from being executed, this view boils down to the maximal preparation strategy. And, as argued in the Discussion section of Experiment 1, this strategy cannot explain relatively fast responding after a no-go trial.

⁵ To ensure that this finding is not an artefact of imperfect task randomization, we checked whether trial n + 2 repetitions and alternations were equally frequent. This turned out to be the case: In Experiment 1, 49.76% of the trials were trial n + 2 repetitions; in Experiment 2 this percentage was 50.00%.