Search

Humans seem to arbitrate between automatic and controlled processing by optimizing a trade-off between cognitive effort and performance. Previous research has described ways of how these costs and benefits can be quantified and how the trade-off between them can be performed. However, it remains unclear how the costs should be weighed relative to the benefits and how the cost of the arbitration mechanism itself factors in. Here, we derive measures for these separate factors from a single objective: the variational free energy. We demonstrate that by minimizing this objective, the trade-off between automatic and controlled processing as well as meta-control is optimized implicitly. As a proof of concept, we show that the congruency and proportion congruency effects in the Stroop task directly result from this optimization, given an environment with specific statistical regularities.

Cognitive control theories describe the active maintenance of goal representations over temporal delays as central for adaptive behavior. Dynamic adaptations of goal representations are often measured as the congruency sequence effect (CSE), which describes a reduced congruency effect in trials following incongruent trials compared to congruent trials. Previous studies questioned active maintenance of CSEs and instead found that CSEs decrease rapidly over time (Duthoo et al., 2014; Egner et al., 2010). However, in these studies, CSEs can be attributed to both, control adaptations following conflict and binding effects due to repetition of stimulus (S) and response (R) features. In four experiments, we demonstrate that CSEs originating solely from control adaptions were not affected by temporal delays. Additional within-subject conditions partially replicated previous research showing a decrease in CSEs in task designs allowing for S-R binding effects and controlled for task complexity as a potential moderator. Together, results support theories which predict an active maintenance or retrieval of cognitive control.

Conflict processing is assumed to serve two crucial, yet distinct functions: Regarding task performance, control is adjusted to overcome the conflict. Regarding task choice, control is harnessed to bias decision making away from the source of conflict. Despite recent theoretical progress, until now two lines of research addressed these conflict-management strategies independently of each other. In this research, we used a voluntary task-switching paradigm in combination with response interference tasks to study both strategies in concert. In Experiment 1, participants chose between two univalent tasks on each trial. Switch rates increased following conflict trials, indicating avoidance of conflict. Furthermore, congruency effects in reaction times and error rates were reduced following conflict trials, demonstrating conflict adjustment. In Experiment 2, we used bivalent instead of univalent stimuli. Conflict adjustment in task performance was unaffected by this manipulation, but conflict avoidance was not observed. Instead, task switches were reduced after conflict trials. In Experiment 3, we used tasks comprising univalent or bivalent stimuli. Only tasks with univalent revealed conflict avoidance, whereas conflict adjustment was found for all tasks. On the basis of established theories of cognitive control, an integrative process model is described that can account for flexible conflict management.

We examined whether a temporal interval between an action and its sensory effect is integrated in the cognitive action structure in a bidirectional fashion. In 3 experiments, participants first experienced that actions produced specific acoustic effects (high and low tones) that occurred temporally delayed after their actions. In a following test phase, the tones that were presented as action effects in the previous phase were now presented as primes for the responses that had caused them previously and, critically, also as primes for the interval that previously separated action and effects. The tones were presented as go-signals in a free-choice test and as response-imperative stimuli in a forced-choice test. In the free choice test, participants were more likely to choose responses consistent with the previous pairing, but these responses were initiated slower than responses that were inconsistent with previous action-effect learning (Experiment 1). Effect-consistent responses were also initiated slower in the speeded forced-choice test (Experiment 2). These observations suggest that retrieval of a long action-effect interval slows down response initiation. In Experiment 3, response-contingent effects were presented with a long or short delay after a response. Reaction times in both, a forced-choice and free-choice setup, were faster in the short- than in the long-interval condition. We conclude that temporal information about the interval between actions and effects is integrated into a cognitive action structure and is automatically retrieved during response selection.