While subjects are generally faster to respond to a stimulus if it is presented at a previously “cued” location, they are paradoxically slower to respond if the spatial “cue” occurred more than 300 ms beforehand. This strange phenomenon is known as inhibition of return (IOR), and is thought to result from the reflexive “suppression” of previously-attended locations in space. IOR can be considered an adaptive characteristic of neural information processing, in that IOR ensures that the environment is being constantly searched for new information.
It turns out that IOR may have other information processing benefits as well. If subjects are asked to name the ink color of a color word, they are slower to do so for incongruent than congruent and neutral stimuli (e.g., GREEN is slower than GREEN and HOUSE). This is known as the Stroop interference effect. However, this difference disappears if these stimuli are presented at locations subject to IOR (i.e., they were cued more than 300 ms ago).
Why should this be the case? In a new article from the Journal of Cognitive Neuroscience, authors Chen, Wei and Zhou examine this effect using fMRI and a clever experimental design. They used a simplified version of the Stroop task, in which subjects could only respond with “red” or “green.” Only two types of stimuli are important for our purposes: those that are incongruent & response eligible (e.g., GREEN) and those that are incongruent & response ineligible (e.g., GREEN).
If the Stroop interference effect is eliminated at regions with IOR because IOR mitigates response level conflict, then “response ineligible” stimuli should still show the interference effect (since there is no response level conflict for a word that refers to an ineligible response), but the “response eligible” stimuli should no longer show the interference effect. In contrast, both stimulus types contain “pre-response” conflict, and so if IOR affects Stroop interference at that level, both stimulus types should show similar patterns of interference at each type of location.
The behavioral results replicated previous work by showing that IOR completely eliminates the Stroop interference effect, such that both types of incongruent stimuli were as fast as neutral words when presented at the previously cued location. There were no behavioral differences between response-ineligible and -eligible trials.
In contrast to the behavioral results, the imaging data demonstrated that left rostral anterior cingulate (ACC) was more active during pre-response interference at the uncued location than at the cued location. The authors interpret this as reflecting the fact that information at the uncued location receives more attention (since that location is not suppressed) and thus generates more interference, which is then detected by ACC. The authors suggest this indicates that ACC is sensitive to pre-response conflict, but this rests on the assumption that reponse-ineligible stimuli actually have no response conflict. This is a tenuous assumption, particularly since some previous work has shown ACC to be sensitive only to response-level conflict.
Left dorsolateral prefrontal cortex (dlPFC) was more active for interference (possibly at the pre-response level, at the response level, or both) at the cued location than at the uncued location. This seems to indicate that dlPFC is sensitive to response-level conflict, but again this rests on the assumption that response-ineligible trials do not have response-level conflict.
It is tempting to accept Chen et al’s conclusions, in part because they tell such a nice story: ACC activity is related to pre-response conflict, whereas dlPFC activity is related to response-level conflict. Unfortunately, as I have alluded to above, these are problematic conclusions.
So, what can safely be concluded from this study?
One interesting thing is that many of the effects here are actually driven by changes in the time to name the color of neutral words. That is, neutral trials are much slower at the cued location, and much faster at the uncued location. So rather than saying that Stroop interference disappears at suppressed locations, it’s more fair to say that even neutral trials are subject to interference when the stimuli appear at suppressed locations!
Likewise, Chen et al’s assumption that response-ineligible incongruent trials contain less conflict than response-eligible incongruent trials is belied by the fact that ineligible trials take numerically longer on average.