How The Brain Manages Conflict: Global and Local Conflict Adaptation Effects

If you encounter a difficult situation, you may be extra careful afterwards, even in a different or unrelated situation. This intuitive statement has recently been confirmed in a laboratory task, and extended to show that such carry-over "conflict adaptation" effects may affect the speed with which you approach subsequent tasks very differently from how it affects the probability of making a mistake.

A task often used to look at conflict is the flanker task: when subjects must respond to an arrow symbol that is surrounded by other arrow symbols, responses will be faster when the surrounding arrows point in the same direction as the central, to-be-attended arrow. Conversely, there's a reaction time cost when the direction of the flanking arrows is incongruent with that of the central arrow.

The brain recruits special mechanisms to mitigate the conflict generated in such tasks. Neuroimaging shows that the anterior cingulate (ACC) is involved, and computational models have shown that it might subserve a "conflict-monitoring" function. The mechanisms operating downstream of the ACC are a matter of debate: maybe it increases strength of goal representations in prefrontal cortex (an established function for that part of the brain) in order to "drown out" the irrelevant stimuli; maybe it inhibits activity in subcortical motor areas to allow more time for decision-processes to operate before a response is made; maybe it does a little of both (or something entirely different).

This rather elegant story about conflict management is also supported by the behavioral phenomenon known as conflict adaptation: people are less slowed on the trials that follow incongruent flanker trials than those which follow congruent flanker trials. So it appears that the consequences of conflict detection carry-over to the subsequent trial.

If it's true that conflict detection operates at a global level, rather than being tied to the particular arrows on the preceding trial, the conflict adaptation effect should still be visible when there is no overlap between the arrows used on the current and preceding trials. This is the topic covered in a new paper by Freitas, Bahar, Yang & Banai.

In a series of three studies, Freitas et al showed that these effects are independent of stimulus repeats between trials. The first experiment involved two different sets of arrows in the flanker task - one set of arrows pointed only up or down, while another set pointed only left or right. The authors showed an interaction between the congruency of the preceding trial and the congruency of the current trial, indicating that these conflict adaptation effects are visible even when the stimuli do not overlap between trials.

Two additional experiments with 32 and 52 subjects involved trials of the Stroop task being randomly intermixed with the flanker task. (In this Stroop task, subjects had to name the ink color of incongruent color words - e.g., "GREEN" written in red ink). These results showed that the same conflict adaptation effect was visible even across these completely different tasks - in other words, an incongruent preceding trial seemed to increase conflict-mediation processes which carried over to the following trial by increasing response time on congruent trials, and decreasing response time on incongruent trials.

Although these effects could be due to influences between the current trial and the trial before the last one (a known phenomenon in task-switching) and between spontaneous moment-to-moment fluctuations in reaction time, the authors also conducted hierarchical regression analyses to control for these influences and showed that the conflict-adaptation effects remained significant in each experiment.

So, the results strongly support a domain general view of conflict-monitoring, where conflict detection operates at a relatively global level and even has carry-over effects that transcend the particular task-sets involved.

In an interesting footnote, the authors mention one exception: when the alternation of tasks or stimulus sets occurs regularly rather than randomly, similar conflict-adaptation effects can be observed in terms of error rates rather than reaction times, and these effects may be limited to particular task-sets.

There are a lot of interesting features to the data, and it's a shame that the article isn't long enough to fully explore them. Here are a few:

- If conflict-detection triggers strengthened goal representations, one would expect that to speed response times on subsequent congruent and incongruent trials alike. Instead, the actual effect is to speed incongruent trials but slow congruent trials. So it doesn't seem that ACC is acting solely on goal representations in prefrontal cortex.

- One might see a dissociation in task-specificity between response latency and accuracy conflict adaptation effects (as discussed in the author's footnote) if conflict-detection mechanisms trigger a global slowing in response time, but also trigger an strengthened task-set (i.e., task or stimulus-response specific) representation. The latter effect should manifest in terms of increased accuracy on trials when the last trial of the same task was incongruent relative to when it was congruent. This should be more apparent for current trials that are incongruent because incongruent trials are less likely to suffer from a ceiling effect, but the trend may be apparent for both.

- The authors conducted the hierarchical regression analysis to see whether the conflict-adaptation effects persisted after controlling for lagged stimulus repetition. In other words, it's possible that having the same task on trial n-2 as on trial n generally slows the response time on trial n, after termed "lag-2 repetition cost." However, the effects of lag-2 repetition here are facilitative, which is an interesting discrepancy that goes undiscussed.