Disentangling task-selection failures from task-execution failures in task switching: an assessment of different paradigms

Local and global control adjustments to stimulus-based task conflict in task switching

A prominent feature of cognitive control is that its deployment is regulated depending on the environmental circumstances. Control upregulation has been widely documented in response-conflict paradigms where congruency effects are reduced both following incongruent trials, and in blocks where incongruent trials are the majority. In two pre-registered task-switching experiments, we investigated whether similar flexible mechanisms are also available when dealing with stimulus-based task conflict. Building up on previous Stroop studies, task conflict was measured as the difference in performance between bivalent congruent and univalent trials, which we name the "valency effect." If cognitive control is upregulated analogously to what observed with response conflict, valency effects should be reduced following bivalent trials and in majority-bivalent blocks. Furthermore, as cognitive control upregulation has been proposed to be task specific, we assessed whether switching to a new task eliminates the expected modulations of task. The results broadly matched our predictions. First, we observed a reduction of the valency effect following bivalent trials similar to the well-known congruency sequence effect, demonstrating similar patterns of flexible control adjustment to task and response conflict. This valency sequence effect was limited to task repetitions, indicating that local control adjustments are task specific. Furthermore, task conflict was reduced in majority-bivalent blocks, similar to the proportion-congruency effect. This finding extends previous Stroop studies suggesting that control is recruited proactively when dealing with stimulus-based task. The proportion valency effect was limited to task-switch trials, leaving open the question on the precise mechanisms behind sustained control adjustments.

Opportune moments for task interruptions: examining the cognitive mechanisms underlying interruption-timing effects

Introduction

Several studies showed that task interruptions at high mental workload moments are more harmful than task interruptions at low mental workload moments. In the present study, we used a theory-driven approach to define the mental workload during primary-task execution and to examine the effects of the interruption timing on primary-task performance.

Methods

Participants performed a primary task comprising a pre-defined sequence of six subtasks, with task interruptions occasionally occurring before the second, third, or fourth subtasks. Critically, the subtasks were organized either in two lag-2 repetition triplets or in two lag-2 switch triplets (e.g., ABA-CBC vs. CBA-CAB). This set-up allowed us to test two predictions about the effects of interruption timing on the resumption costs (i.e., the performance in subtasks following an interruption compared to the performance in the same subtask in non-interrupted primary tasks). First, we expected task interruptions before the fourth subtask being the less detrimental due to the presumed chunking of the six subtasks into two triplets. Second, in lag-2 switch triplets, task interruptions before the second and third subtasks were predicted to result in comparable resumption costs. In contrast, in lag-2 repetition triplets, task interruptions before the third subtask were hypothesized to be more disruptive than those before the second subtask. This is because the mental workload should be higher due to the need to overcome subtask inhibition.

Results

We found an interruption-timing effect with higher resumption costs for task interruptions occurring before the third subtask compared to interruptions before the second and the fourth subtasks. However, this effect did not differ across lag-2 repetition sequences and lag-2 switch sequences.

Discussion

These findings are discussed from a memory perspective and a context reconstruction perspective.

Evidence of task-triggered retrieval of the previous response: a binding perspective on response-repetition benefits in task switching

In task switching, response repetitions (RRs) usually yield performance benefits as compared to response switches, but only when the task also repeats. When the task switches, RR benefits vanish or even turn into costs, yielding an interaction between repeating versus switching the task and the response (the RR effect). Different theoretical accounts for this RR effect exist, but, in the present study, we specifically tested a prediction derived from binding and retrieval accounts. These maintain that repeating the task retrieves the previous-trial response, thus causing RR benefits. Retrieval is possible due to the task-response binding formed in the previous trial. We employed a task-switching paradigm with three response options that allowed us to differentiate error types. Across two experiments (N = 46 and N = 107) we showed that response-repetition errors in response-switch trials were more likely in task repetitions than in task switches, supporting the notion that the previous response is retrieved by the repeating task, despite being wrong. Such a finding is in line with binding and retrieval accounts but cannot be easily accommodated by the competing theoretical accounts. Thus, the present study indicates task-response binding as an important mechanism underlying RR benefits in task repetitions.

Persisting inhibition biases efficient rule inference under uncertainty

Introduction

Task set inhibition supports optimal switching among tasks by actively suppressing the interference from recently executed competing task sets. It is typically studied in cued task-switching paradigms where there is no uncertainty about the task set or rule to prepare for on each trial. While inhibition has been shown to influence the speed and the accuracy of task execution, affecting task set retrieval, preparation, or implementation in conditions of task set switching, it remains uninvestigated whether it also affects rule selection under uncertainty.

Methods

We implemented an ad-hoc four-rule card sorting task and categorized the rules selected by participants after a rule shift according to the recency of their last usage. We included a measure of working memory capacity (WMC) to control for its involvement in the rule selection process.

Results

Participants exhibited a reduced preference for recently abandoned rules than less recently abandoned ones. Furthermore, we found that such a preference was not associated with WMC.

Discussion

The results suggest that decision-making processes underlying rule inference and selection may be influenced by task-set inhibition, configuring as a conflict adjustment mechanism to the sequential history of rules application.

Examining the cognitive processes underlying resumption costs in task-interruption contexts: Decay or inhibition of suspended task goals?

To examine whether an ongoing primary task is inhibited when switching to an interruption task, we implemented the n - 2 backward inhibition paradigm into a task-interruption setting. In two experiments, subjects performed two primary tasks (block-wise manipulation) consisting of a predefined sequence of three subtasks. The primary tasks differed regarding whether the last subtask switched or repeated relative to the penultimate subtask, resulting in n - 1 switch subtasks (e.g., ABC) and n - 1 repetition subtasks (e.g., ACC) as the last subtask of the primary task. Occasionally, an interruption task was introduced before the last subtask of a primary task, changing the last subtask of the primary task from a n - 1 switch subtask to a n - 2 switch subtask (e.g., AB → secondary task → C) and from a n - 1 repetition subtask to a n - 2 repetition subtask (e.g., AC → secondary task → C). In two experiments with different degrees of response-set overlap between the interruption task and the subtasks of the primary task, we observed that switching back from the interruption task to the primary task resulted in n - 2 switch costs in the first subtask after the interruption (i.e., worse performance in n - 2 switch subtasks than in n - 2 repetition subtasks). This n - 2 switch cost was replicated in a third experiment in which we used a predefined sequence of four subtasks instead of three subtasks. Our finding of n - 2 switch costs suggest that the last subtask performed before the interruption remains activated when switching to the interruption task.