A lot has been written about domain-general processing in prefrontal cortex, and a very old lesson often gets overlooked: there are very basic hemispheric asymmetries (particularly in PFC) that divide information processing by modality. A very nice study by Morimoto et al provide a nice reminder of this important feature of neuronal organization, and illustrates that very specific regions of PFC are the only ones to show such hemispheric specialization in tasks that require cognitive control.
Morimoto et al first trained subjects to press (for example) button “1″ in response to “RED”, “BLUE”, █ or █, and button “2″ in response to “GREEN”, “YELLOW”, █ or █, with the different stimulus types separated into different blocks. As you can tell, some of the words require the same response as the color they refer to (in this case, red and yellow) while other words require the opposite response as the color they refer to (blue and green). Next, they intermixed these trial types, and after that they introduced a flanker task using these stimuli, such that subjects had to respond only to the central item. For example, which button would you press for this?
The answer is 2 – and that counts as an easy trial, since both █ and “GREEN” are associated with button 2. This would count as a stimulus-incongruent but response-congruent trial, since the stimuli refer to different colors but both require the same response. In contrast, this one should be harder:
because the flanker – █ – is associated with a different response from the one you’re supposed to pay attention to – “RED.” This is considered a stimulus-incongruent and response incongruent trial, since the stimuli both refer to different colors and to different responses. In any case, the authors never included more than one flanker (with an x on the other side – as above), and threw in a few trials with a neutral asterisk flanker as a control condition. Notably, this control condition also has “hard” and “easy” trials, since some of the stimuli refer to colors that are mapped to two different responses (hard; blue and green in the example above), and other stimuli refer to colors that are associated only with a single response (easy; yellow and red above). Subjects practiced this task (with completely intermixed trial types) until they got 90% correct, and then completed the final task in the scanner.
The behavioral results showed what you’d expect: reaction times were slower on trials with incongruent flankers – regardless of whether they were incongruent due just to the stimuli, just to the response, or to both (and there were no differences between these conditions). But it’s neuroimaging that’s really interesting…
The easy control trials elicited greater activity in the left inferior frontal junction (IFJ) when the stimulus was a word, and greater activity in the right when it was a colored shape. This was only true for the IFJ – the more ventral IFG region simply showed greater activity overall on the shape trials. However, the flanker trials elicited the opposite effect: greater activity was seen in the right IFG when the flanking stimuli were patches, and greater activity was seen in the left IFG when the flanking stimuli were words, each relative to unflanked trials. The effects in IFG were driven by the flanker trials with incongruent mappings relative to the hard control trials (those with stimuli that referred to colors with two possible responses). Furthermore, these results differed depending on the accuracy of the subjects (although it’s not clear in which direction – I’m waiting for clarification from the authors on this point).
This study is fairly incredible because it shows that these modality effects are specific to very particular regions of the lateral prefrontal cortex. Furthermore, the regions that show these modality effects depend on the task-relevance of the material; the more ventral IFG regions show the interaction for the stimuli that are task-irrelevant, as though the IFG is attempting to suppress them in a modality specific way. Conversely, the more dorsal IFJ region is modulated depending on the task-relevance stimulus – the one that needs to be selected.
This study poses a major problem for theorists that suggest the IFG mediate cognitive control in ways other than directly suppressing task-irrelevant information. The problem is that one would expect greater activity in those regions for the to-be-selected stimulus if they are accomplishing some form of activation-based selection, but the hemispheric interaction went in the opposite direction: the IFG showed greater activity for the to-be-suppressed stimulus! I consider this a very strong challenge to activation-based accounts of selection in cognitive control tasks, and very strong support for inhibition-based accounts of selection.
OK, after talking with a lab mate about this, and after hearing from the authors about an unclear statistic, I’ve changed my attitude slightly.
The authors reported an interaction of subjects’ accuracy with these hemispheric patterns, and via personal communication I’ve learned that the interaction was stronger in the low-accuracy group. This could suggest that the hemispheric asymmetry was actually detrimental to performance – far from the prediction of the inhibition-based account of selection, in which greater recruitment of “inhibitory” PFC regions would be expected to yield higher accuracy. Instead, it’s consistent with the account offered by my labmate: the effect of flankers on functional hemispheric asymmetry in the inferior frontal gyri reflects competition from those flankers (greater in the left from verbal material, greater in the right from more visual material) that increases demands on those regions for selection of the appropriate response. Thus, this paper essentially reports a task-difficulty manipulation, such that verbal flankers make the task more difficult for the left hemisphere, visual flankers make it more difficult for the right hemisphere, and such that these patterns are exaggerated for those that have more difficulty with the task.
Hiroki M. Morimoto, Satoshi Hirose, Junichi Chikazoe, Koji Jimura, Tomoki Asari, Ken-ichiro Yamashita, Yasushi Miyashita, Seiki Konishi (2008). On Verbal/Nonverbal Modality Dependence of Left and Right Inferior Prefrontal Activation during Performance of Flanker Interference Task Journal of Cognitive Neuroscience, 20 (11), 2006-2014 DOI: 10.1162/jocn.2008.20138