In three experiments, we investigated whether task-choice behavior is biased by temporal cues in a temporally structured voluntary task environment. Specifically, in a task-switching setting forced-choice and free-choice trials were randomly intermixed (hybrid design) and the task in forced-choice trials was predicted by the foreperiod (Exp. 1A and 1B: 100% predictability; Exp. 2: 80% predictability). Our main interest was whether the randomly interspersed short and long foreperiods in free-choice trials biased task choice. Across all experiments, we found that compatible foreperiod-task combinations were chosen more often than incompatible ones. In the first two experiments, this effect was stronger on task-switch trials (in the last experiment, only for the short foreperiod, this dissociation was found). Furthermore, over all experiments, a compatible choice bias was only significant for the short foreperiod. Finally, temporal predictability also affected general task preferences: overall, participants were biased toward the task which was associated with the short foreperiod on forced-choice trials.
Task-choice behavior and temporal preparation: a preparation-switch account
The present experiments provide the first evidence that time-based task-expectancy effects can not only be found for task performance, but also for task choice. As reviewed in the introduction, temporal predictability may influence task performance in forced-choice settings due to time-dependent changes in task preparation. In a similar vein, task-choice behavior might also be influenced by these changes in task preparation: Specifically, we suggest that task selection in our experiments utilized bottom-up biases as introduced by time-based task expectancy. Throughout the time course of the foreperiod, the preparatory activation level for either task changed as a function of temporal predictability which further translated to task selection (for a similar argument, see Arrington, 2008
On a more functional level, this suggests that temporal predictability does not only influence the course of (active) task preparation, but also of task-selection processes. In studies on voluntary task switching, these two processes have been found to be distinguishable (Arrington et al., 2014
), but in many cases, task preparation informs task selection (e.g., Arrington & Logan, 2004
; Mittelstädt et al., 2018
). On the one hand, while task preparation is mainly reflected in performance indices like the switch costs (in RTs and ERRs), task-selection processes are marked by task choice indices like the voluntary switch rate and task bias—and correlations between these markers have been found to be rather small (Arrington & Logan, 2005
). Also, substantial interindividual differences emerge in terms of how much task selection is driven by exogenous factors (Arrington & Yates, 2009
; Orr & Weissman, 2011
), supporting a view of (partially) diverting processing streams. However, our results indicate that task-selection behavior incorporates the temporal predictabilities of tasks in forced-choice trials. This would also fit with the study by Mittelstädt et al. (2018
) in which participants used the predictable waiting time for a repetition stimulus when deciding to switch or repeat tasks. A number of studies further substantiate the claim that bottom-up influences (that is, effects triggered by the task context and/or stimulus features) assert a huge effect on voluntary task choice (e.g., preparation time, stimulus repetitions, and task difficulty; Arrington & Logan, 2004
; Mayr & Bell, 2006
; Yeung, 2010
). We suggest that the same may be true for the temporal predictability effects which we found here. In terms of the selection process, Herbort and Rosenbaum’s (2014
) as well as Volberg and Thomaschkes (2017
) studies suggest that action selection in this case, preparing the response rule and/or response hand associated with one task (Demanet & Liefooghe, 2014
) is biased by the foreperiod–task contingencies, favoring compatible task choices.
This would also fit with our finding of a significant “short” task bias, that is, a bias toward the task associated with the short foreperiod. Even though this effect was considerably reduced when excluding participants with very strong task biases (that is with less variability in their task-choice behavior), the compatible choice bias still was larger after the short foreperiod. In our view, the emergence of time-based task-expectancy effects requires that participants first prepare for one task, and then switch to preparing for the other; it is highly likely that they fail more often to do so after the long foreperiod (see also De Jong, 2000
). This idea is in accordance with the other findings of the literature. First, Pfeuffer et al. (2020
) provide evidence that the frequency of anticipatory eye movements to the temporally predicted location (and task) is larger for a short compared to a long foreperiod. Second, Volberg and Thomaschke (2017
) show that preparatory activity related to a certain temporally expected effector switches roughly at the expected end point of the short duration. Furthermore, Aufschnaiter, Kiesel, Dreisbach et al. (2018
) found an RT benefit for the short compared to the long foreperiod in a temporally structured task environment. Relatedly, it is a well-known fact that participants tend to avoid switching tasks (Kessler et al., 2009
), at least partially because of the effortful cognitive operations (e.g., reconfiguration of task sets) needed to implement those switches (Kool et al., 2010
). The same may be true for temporally predictable tasks, where a switch in preparation after the short foreperiod has passed may equally be avoided.
Differential effects of foreperiod length can also be found in the variable foreperiod paradigm mentioned in the introduction (e.g., Steinborn et al., 2008
), where sequential modulations can only be found with a current short foreperiod. Steinborn et al., (2009
) used varying warning stimulus modalities (or features, Steinborn et al., 2010
) to show that this sequential modulation was largely reduced when modality (or sufficiently distinct features within one modality) shifted, further providing evidence that short-foreperiod trials are influenced by preceding trials in a way that long-foreperiod trials are not: after-effects of the previous trial, e.g., its reinforced time point of peak readiness, thus seem to be limited to comparably short foreperiods. Other processes, such as conceptually driven, more intentional preparation processes, may prevail for longer foreperiods (Langner et al., 2018
Similar effects can be seen in the task-switching domain, where interference from the previous task set is largest with no or very short preparation intervals (Meiran, 1996
). Consequently, task-switching performance as well as voluntary switch rates increase with more time between trials (Arrington & Logan, 2004
; Monsell & Mizon, 2006
). In our paradigm, stronger compatibility effects were obtained for the short foreperiod and also the task associated with it. Thus, one could assume that while task selection of short-foreperiod trials depends on availability biases induced by the foreperiod-task contingency manipulation, any such biases are reduced the more time passes. Hence, the “short” task bias is a temporal bias in effect propagation of foreperiod compatibility: the compatibility bias may simply fade with time or become more noisy (as a sort of passive decay, cf. Meiran, 1996
) or may sometimes be overruled by other factors impacting task choice (cf. Arrington & Logan, 2005
; Langner et al., 2018
The present results could also be interpreted in terms of an episodic-retrieval account (Hommel, 2004
; Los et al., 2014
; Mayr & Bell, 2006
; Thomaschke & Dreisbach, 2015
). According to this account, on each trial, a binding between the current foreperiod and the task is established that carries over to the next trial(s): a repetition of the current foreperiod automatically retrieves the task that was associated with it in the previous trial. Given that foreperiods and tasks in forced-choice trials were highly correlated, task-choice behavior in a following free-choice trial could simply reflect such an automatic retrieval of previous foreperiod-task bindings (for a similar argument, see Los et al., 2014
) thus most likely result in a compatible choice. However, this account would predict more compatible choice repetitions than switches (cf., Mayr & Bell, 2006
; Los et al., 2014
), which we did not find in the current results. Nevertheless, episodic retrieval is suggested to be an aiding factor during learning of foreperiod–task associations (Thomaschke & Dreisbach, 2015
) and could influence task choice in the current paradigm in addition to learned foreperiod–task associations. Future studies should directly test this account by contrasting trials where both foreperiod and task repeat or switch with trials were only one repeats.
Temporal predictability effects on (voluntary) task performance
Importantly, we also found time-based task-expectancy effects on task performance. RTs and, somewhat attenuated, ERRs on free-choice trials mirrored the advantage of compatible task choice. Repetition and switch trials seemed to profit in a similar way by compatibility; or at least, with the current results, a clear dissociation is not possible (Exp. 1A: no effect for switch RTs; first two experiments: no effect for switch ERRs; Exp. 2: no effect for repetition ERRs). The similarity between task choice and task-performance results fits well with the preparation-switch account which we introduced earlier: time-based task expectancy acts on task preparation processes (as reflected in task-performance indices) that inform task selection (reflected in task-choice indices). However, the face validity of the performance-choice similarities has to be treated with caution—causal attributions cannot be made so far and further investigation is needed to corroborate this claim.
In a last 100% forced-choice test block in Exp. 1A and 1B, we checked whether effects of compatibility as established in the previous test phase would transfer to a block where foreperiod–task combinations were completely random. Here, we could show that indeed previously established foreperiod–task combinations were responded to faster and (only in the case of repetitions in Exp. 1A) more accurately. This reflects a replication and extension of the findings by Aufschnaiter, Kiesel, and Thomaschke 2018
) who showed for forced-choice task switching that time-based task expectancy survives a change in absolute time environment. The current results extend these findings to a free-choice context.
Implications for future research
Using different or a larger variety of foreperiods may be interesting with respect to the “short” task bias. Given that preparation in voluntary task switching as well as in foreperiod and interval timing studies is known to be successful only after some hundred milliseconds have passed, the question arises whether the “short” task bias would still be found if the short foreperiod was considerably shortened (e.g., to 100 ms). In this case, the default may rather be to start preparation only after this short foreperiod has passed given that only then task preparation (Kiesel et al., 2010
) and accurate timing (Lewis & Miall, 2009
) are possible. Furthermore, previous research on the variable foreperiod paradigm (Langner et al., 2018
; Steinborn et al., 2009
) has shown that the short-foreperiod “bias” in terms of susceptibility to sequential modulations can be largely increased when using more than two foreperiods as well as a greater range. In an experimental setup that was optimized for revealing differential effects of foreperiod length, Langner et al.’s (2018
) Experiment 2 employed foreperiods of 800 ms, 1600 ms, and 2400 ms. Note that using more than two foreperiods also means that participants have to not only learn two foreperiod–task associations, but three. While increasing the overall task demand, this would also allow to investigate other task-switching phenomena in the context of temporal predictability, such as backward inhibition (Koch, Gade, Schuch, & Philipp, 2010
This methodological approach may be informative in more than one respect: it may also allow to investigate whether the effect of temporal predictability on task-choice behavior is based on relative or absolute timing. Previous studies on time-based expectancy effects on performance seem to prompt the relative timing idea ((Aufschnaiter, Kiesel, & Thomaschke, 2018
; Thomaschke, Kunchulia, & Dreisbach, 2015
). That is, participants learn that one event appears after the interval that is relatively shorter/longer than another interval, rather than learning the correlation between the exact time period and the event. If we think of the example in the beginning of waiting for an unpunctual friend, relative timing information seems sufficient for changing the anticipated course of action: the more time passes, the likelier it is that the person will not show up and I will have to act accordingly. (Aufschnaiter, Kiesel, & Thomaschke, 2018
make the legitimate claim that finding time-based task expectancy to only involve relative timing information may be due to the experimental design involving only two foreperiods. Given that many real-life scenarios require absolute timing (e.g., a pilot operating in a cockpit), it may be a fruitful endeavor to see whether time-based task-expectancy effects, particularly on task-choice behavior, can be shifted toward absolute timing.