Abstract
To perform a task that differs from a previously performed one, it is necessary to prepare for the new task as well as to disable the task set of the preceding task. In a series of three experiments we examined whether preparation for a task shift implies a direct update of the task set that is carried over from the preceding trial. To this end, in Experiments 1 and 2 we factorially varied the relation of the task in trial n, first, to the task in trial n − 1 (intertask relation) and, second, to the task that was precued for trial n (precue-to-task relation). Invalid precues resulted in substantial costs, which increased with longer precueing intervals. However, this increasing effect of the precue-to-task relation was not accompanied by a decreasing effect of the intertask relation. Furthermore, both effects had different qualitative properties. These findings suggest that two tasks can be represented concurrently but on two different levels of representation. In Experiment 3 we observed that the persistence of the effect of the intertask relation depends on the possibility that task repetitions can occur in trials in which a task shift is precued. This suggests that the persistence of the effect of the intertask relation is controlled in a context-sensitive way.
Acknowledgments
The research reported in this article was supported by Grant KI 1205/2–1 of the Deutsche Forschungsgemeinschaft. We thank Gero Szepannek and Dana Lembke for assistance in running the experiments and analysing the data, and Stefan Lapp for providing the software. In addition, we thank Iring Koch and an anonymous reviewer for helpful comments on an earlier version of this article.
Notes
An analysis of the even-numbered trials yielded no observations that are relevant to the purposes of the present study. Generally, participants were faster in the even-numbered trials than in the odd-numbered correctly precued task-repetition trials. Because the even-numbered trials were both always correctly precued and always task repetitions, it is impossible to determine the relative contribution of these two factors to the lower RT level in the even-numbered trials. As is seen in Experiment 2, the main findings of Experiment 1 cannot be attributed to the inclusion of these filler trials, so they are not mentioned any further.
An analysis that included the individual tasks and the response sequence as factors revealed that neither the effect of the intertask relation nor the effect of the precue-to-task relation entered into a significant interaction with the task factor. With respect to the response sequence, the usual response-repetition costs in task-shift trials were observed (cf. CitationKleinsorge & Heuer, 1999).
This detailed feedback was mainly intended to inform the experimenter. Participants' attention was directed on the percentage score.
It might be suspected that in the task-repetition-possible condition the shift-cost profile was preserved because participants also stayed in a “repeat modus” in the odd-numbered trials although a task shift was more probable than a task repetition in these trials. However, a comparison of mean RTs in even-numbered task-repetition trials with mean RTs in odd-numbered repetition trials revealed that participants were much faster in the even-numbered trials (493 ms) than in the odd-numbered trials (685 ms), t(11) = 5.94, p < .001. This was so despite the fact that task repetitions in odd-numbered trials were actually third repetitions of the same task, which should have benefited from additional positive transfer if participants in fact were set for another task repetition in the odd-numbered trials. This effect is also likely to mask the number of shift costs in the task-repetition-possible condition, which was unusually low. The number of “true” shift costs is probably estimated more appropriately by a comparison of RTs from even-numbered trials with RTs in odd-numbered trials with correct left precues than by a comparison of RTs from odd-numbered trials with correct right versus correct left precues. The former comparison involves task repetitions too, but is affected by the effects of an incorrect precue and a high probability of a task shift.