Pending intentions: Effects of prospective task encoding on the performance of another task

In the present article, we use the term “task preparation” exclusively for preparation that can be shown to be specific for a particular task. One indication of task specificity is a reduction of switch costs as a function of the opportunity to engage in advance task preparation. In the literature on task switching, it is not unusual to speak of task preparation on the basis of observing improved performance as a function of the time interval that separates two trials (when the task sequence is known in advance) or the interval that separates the presentation of a precue and the presentation of an imperative stimulus even without observing significantly reduced switch costs. One possibility to account for this kind of “preparation effect” is to assume that participants engaged in a process of goal shifting that does not entail processes like the retrieval of task rules that would be necessary for a reduction of switch costs (e.g., Gotler, Meiran, & Tzelgov, 2003; Koch, 2003). More generally, what seems to be behind these preparation effects that are not accompanied by reductions of switch costs is a reduction of uncertainty about which task to perform and when to perform it. Reductions of uncertainty are well known to speed up performance (cf. Welford, 1980) but are not necessarily linked to task preparation. Perhaps the following illustration will make the distinction clearer: Imagine you are told that you will go on a trip tomorrow. As long as you do not know your destination (e.g., whether it will be hot or cold there), there is not much you can do to prepare for the trip. Once you are told where you will go to, your uncertainty is reduced and you can prepare for the trip, but you don’t have to. Nevertheless, once you decide to prepare for the trip (for example, by packing your bag or buying tickets), the reduction of uncertainty may speed up the initiation of this preparation. Now imagine that you are told that your destination has changed, a situation that is comparable to a situation involving incorrect precues, or that the trip is cancelled altogether, a typical go-nogo situation. As long as you did not engage in preparation for the trip, you have to shift your goal, but it should not matter whether your original and your revised destination are nearby or in different hemispheres. However, once you have started with “trip-specific” preparation, this difference will matter a lot, so the relation between your original and your final destination will determine the costs associated with revising your plan.

As a further test of whether there was a structural limitation in the ability to uphold preparation for Task 2 while performing Task 1, we correlated for each pair of trials the individual reaction times of Task 1 with the reaction times of Task 2. This was done separately for each experiment and each combination of stimulus congruence, the relation between the two tasks (inter-task relation), and, for Experiment 2, the three RSIs. Generally, the mean individual correlations were quite low (range: - .01 - .23), and, more important, with one exception (out of 32), they were always positive. This speaks against the assumption that preparation for one task occurred at the expense of preparation for the other task of the trial pair.

The “underadditive switch-cost profile” that can be reliably observed with this task combination consists of the observation of an ordered gradation of switch costs as a function of the relation between two successive tasks (i.e., the inter-task relation). Switch cost are largest with a change of only the type of judgement, smallest with a change of only the mapping, and intermediate with a change of both task features. This switch-cost profile shows up when participants switch among the four tasks without foreknowledge (e.g., Kleinsorge & Heuer, 1999; Kleinsorge et al., 2001), and it is not restricted to this particular task combination but seems to be present with any combination of binary task dimensions that can be represented in a hierarchically structured manner (cf. Kleinsorge, 2004). When participants have reliable foreknowledge about the next task, the switch-cost profile is largely alleviated (cf. Kleinsorge, Heuer, & Schmidtke, 2002), as it was in the present experiments (cf. Fig. 2).