What cognitive processes make up consciousness? One way of answering this question is to identify conscious processes as those involved in controlled but not in automatic behaviors. For example, if you see a bright dot appear in your field of vision, your eyes will automatically orient to that location in space. In contrast, if I have told you to look away from any bright dots that appear in your field of view, you will be able to do this - but only because you possess consciousness in the form of "cognitive control." So, what computations support "cognitive control"?
Cognitive control is sometimes argued to involve "inhibition." Although it's intuitive to think that you suppress the urge to look towards a bright dot by literally deactivating the neural representations for that eye movement, there are reasons to think this may not actually occur. For one thing, it is metabolically inefficient for the brain to spend energy deactivatingrepresentations; wouldn't it be better to simply not engage them in the first place?
Another problem with these "directed inhibition" or "selective suppression" accounts is that biological neural networks seem to have very few long-range inhibitory connections. Thus, at a neural level, it seems unlikely that one brain region might send a command like "don't move!" to another brain region, since it would actually have to send an excitatory signal directly to local inhibitory neurons. Again, this would be a metabolically inefficient architecture.
There are also reasons to suspect that many advocates of directed inhibition are actually committing levels-of-analysis fallacies, as MacLeod et al. describe in their excellent chapter "In Opposition to Inhibition."
Finally, if I tell you "don't think of a pink elephant," it is notoriously difficult to do so. Again, this suggests that you cannot directly inhibit or selectively suppress a specific representation, whether it is "pink elephant" or "look at that dot!" However, you might just suceed by thinking to yourself "blue giraffe." That is, you can avoid thinking "pink elephant" by strongly activating the mutually exclusive representation for "blue giraffe."
And this is essentially the theoretical alternative to "directed inhibition" and "selective suppression" accounts of cognitive control. In more scientific terms, representations compete with other representations for dominance in a specific network. This is compatible with known neuroanatomical evidence, and is demonstrably effective, as currently implemented in a variety of successful artificial neural network models.
However, every now and then I come across some data that is difficult to explain without recourse to "directed inhibition" or "active suppression" mechanisms. This kind of data can be found in an in-press article by Wijnen & Ridderinkhof.
In their article, Wijnen & Ridderinkhof ask subjects to perform a relatively simple task. The subject's job is to watch a display of six grey circles until five of them turn red. Subjects in the "eye movement" condition must then look towards the side of the screen containing the single grey circle (which is considered the "target"). Subjects in the "motor" condition must press either a left or right key, corresponds to the side of the screen with the target.
Sometime in between the presentation of the display and when the circles change color a new red circle appears; this distractor is to be ignored. The time at which the distractor appears is parametrically varied in the experiment. Finally, the distractor can appear on the same side of the screen or a different side of the screen as the target.
The results of this experiment showed several interesting trends. First, if the distractor was presented on the same side of the screen as the target, subjects were faster to respond than if the distractor was presented on the other side of the screen. However, this effect was much stronger if the delay between distractor and target was relatively small; if the delay was larger, the benefit of having a distractor on the same side of the screen as the target was significantly lower.
Why might this be the case?
The authors suggest that inhibition is actively directed at the distractor's location, and that this inhibition grows over time. Therefore, if the target is presented soon after the distractor, there is little inhibition, and so the correct response is much facilitated. On the other hand, if the target is not presented until well after the distractor, this location in space has become more inhibited; in this case the benefit is attentuated because that spatial location is more strongly inhibited.
Unfortunately, there are a couple of subtle problems with this account. First of all, the measurements of reaction time are thrown off by a speed-accuracy tradeoff: sure, people are faster the larger the delay between distractor and target, but they're also making significantly more errors! The "directed inhibition" account would seem to predict the opposite: responses should actually be more accurate when subjects have more time to deploy their inhibitory cognitive control.
Secondly, although the authors conclude that their similar results for the oculomotor and motor tasks reflect that "directed inhibition" is a unitary process that operates similarly across response modalities, they also note significant differences between them. For example, faster eye movements were associated with a relatively lower congruence effect than slower eye movements, and this pattern was more true the larger the delay between distractor and target. In contrast, the speed of manual responses had little influence on the congruence effect at any delay between distractor and target. Therefore, these results could just as easily contradict the conclusion that inhibition is "unitary": at least according to these measures, inhibition does not operate similarly across modalities.
This explanation would be fine if it weren't for the above problems I outlined with "directed inhibition" accounts. So, is there an alternative explanation based on a neuroanatomically-sound theory?
In fact, there is an extremely intuitive explanation for these results. Activity related to the distractor slowly disappates over time, because the location of distractors is not amplified by cognitive control. Therefore, distractors are inherently more distracting when subjects respond extremely quickly - as they tend to do with eye movements. However, for slower eye movements, and for key presses (which are slower than any eye movement by a hundred miliseconds or so), distractor-related activity has dissipated almost completely. Therefore, in these cases, there is less benefit to congruency.
According to this perspective, there is no central "directed inhibition" or "active suppression" control process, merely the slowly dissipating activity of distractors.