Neural correlates of adaptive cognitive control in working memory.

Conflicts in working memory (WM) can occur when retrieval cues activate competing items, which impairs the efficiency of retrieval. It has recently been shown that WM retrieval adapts similarly to these conflicts as predicted by conflict monitoring theory for selective attention tasks. Here, we utilized event-related potentials (ERPs) to investigate whether conflict and adaptive control in WM are reflected by the same neural markers that have previously been described for selective attention tasks. In our task, participants encoded two differently colored memory lists that contained four digits each (i.e., 2 5 7 1 and 4 5 9 1), and had to recognize whether a probe item from a specific list and position was correct or incorrect. Conflict during retrieval emerged when digits at corresponding positions (e.g., 2 and 5 at the first position) were different (incongruent), but not when these digits were the same (congruent). In behavioral data, we found a congruency sequence effect, that is, responses to incongruent probe items were slower, and this effect was reduced following trials with incongruent probe items. In ERPs, this behavioral marker of adaptive control was accompanied by two effects. First, congruency affected the amplitude of an N450, and this conflict effect was reduced after incongruent trials. Second, the posterior P3 amplitude varied with the congruency of the current and the previous trial. Both results resemble those found for the Stroop task and thus highlight the similarity between conflict and adaptive control in WM and selective attention tasks.

Adaptive control of working memory.

The present study investigated mechanisms of adaptive cognitive control in working memory (WM). WM is conceived as a system for short-term maintenance, updating and manipulation of representations required for goal-directed action. Adaptive control refers to the finding of flexible adjustments of control processes based on conflict. For instance, a higher frequency of incongruent stimuli, that is, stimuli evoking conflicting response tendencies, leads to a higher level of cognitive control as reflected by smaller congruency effects (i.e., the difference between congruent and incongruent items). Likewise, conflict on the previous trial leads to a higher level of cognitive control on the current trial. To investigate adaptive control in WM, we used a modified Sternberg paradigm. Participants memorized two differently colored lists of four digits (i.e., 2 5 7 1), in which corresponding positions in both lists contained the same digits (congruent items) or different digits (incongruent items). Participants were required to make a match/mismatch judgement (Experiment 1 and 2) or to recollect the correct digit at a probed position in one of the two lists (Experiment 3). In all experiments, we could replicate both hallmark effects of adaptive control, the proportion congruency effect, and the congruency sequence effect. Our results strongly support the assumption that WM representations can be dynamically adapted based on the amount of conflict, and that adaptive control of WM follows the same principles that have previously been shown for selective attention.

Same same but different? Modeling N-1 switch cost and N-2 repetition cost with the diffusion model and the linear ballistic accumulator model.

In three task-switching experiments, we investigated the relationship of n-1 switch cost and n-2 repetition cost. N-1 switch cost is observed when participants are asked to switch from one classification task to another, e.g., from judging a digit as odd or even to judging a digit as smaller or larger than five. N-2 repetition cost is observed when participants are asked to switch among three tasks (thereafter called A, B, and C). This cost is observed when the task on trial n-2 is repeated in trial n (i.e., in task sequences like ABA) compared to when it is switched (i.e., in task sequences like CBA). So far, the n-1 switch cost is assumed to be caused either by reconfiguration processes or by episodic-memory inertia from the previously activated task-set. N-2 repetition cost is thought to reflect lingering inhibitory processes for resolving conflict among tasks. Whereas both views are integrated in some models, it is up to date unclear whether n-2 repetition cost is related to n-1 switch cost. To examine this relationship, we decomposed the processes underlying n-1 switch cost and n-2 repetition cost using a diffusion model analysis as well as a linear ballistic accumulator model. The results showed that n-1 switch cost reflects interference caused by the residual activation of the previous task set as indicated by slower evidence accumulation processes. In contrast, there were no consistent parameter modulations underlying n-2 repetition cost. These findings emphasize that different cognitive processes are involved in n-1 switch cost and n-2 repetition cost.