Inhibition is arguably the function most colloquially associated with cognitive control. When we think of self-control or willpower, it is often our ability to stop ourselves from doing unwanted, distracting, or maladaptive tasks that we are thinking about. And it follows that the construct of inhibition has played a significant explanatory role in scientific ideas about the nature of cognitive control and its failures. When we perform a distracting task or have an unwanted thought or act in an inappropriate way, this could be because we failed to inhibit the impulsive action. Likewise, when we are slower to act because we are performing an uncommon action instead of a habitual one or because we are multitasking, this might be explained as a demand to inhibit the habit or the interfering task. We likewise hear accounts of the control deficits of patients or the development of control in childhood framed in terms of changes in the capacity for inhibitory control. Thus, inhibition is commonly viewed as the source of self-discipline and control, as that which withholds our hands from the temptation of life’s various cookie jars.
Yet, inhibition can be a pretender in cases like these. It is a description of an outcome of our behavior, such as when we do not produce a particular action or when the pace or vigor of a response is slowed. However, these outcomes in our behavior need not arise from an underlying inhibitory process. For example, we can avoid producing one behavior by choosing another one. Indeed, we can simply choose not to perform an action in the first place. One behavior can likewise be slowed by interference or blocking, if we choose another. These examples seem distinct from our intuition of what an inhibitory process should do. An inhibitory process should countermand an initiated action or thought, thereby suppressing or preventing its realization. Thus, merely observing that an unwanted action did not occur does not tell us that this reflects the actions of an inhibitory control mechanism in the above sense. Indeed, the scope of an inhibitory process that satisfies the above definition appears more limited.
Motor Stopping
One case where the evidence for an inhibitory process is quite convincing is that of rapidly stopping a movement. And so, it is worth introducing this example, which is treated in more detail in my book, On Task. In laboratory studies, motor stopping can be studied using tasks like the stop-signal task in which a movement is initiated to a “go” cue and then is occasionally countermanded in response to a second “stop” cue that occurs at some offset after the go cue.
This task is appealing as a test of an inhibitory process. It requires initiation of a response on every trial, and so if it is successfully countermanded, this suggests that it has been suppressed (as opposed to never initiating the movement in the first place). Further, as there is not an alternative response to make, it is harder to argue that suppression comes indirectly from actively choosing another action. And, under certain assumptions that I don’t have the space to detail here, it provides a means of measuring the strength of the inhibitory process rather than only assessing its outcome in the slowing or omission of a response.
The stop-signal paradigm has led to a wealth of information about this fast inhibitory process for movement. We have learned that it is not only fast but also global. It suppresses motor responding even for effectors that were not part of the countermanded response. We have learned that cortical-basal ganglia networks are important for this kind of fast stopping. But, they bypass the slower selective gating deliberations in the striatum and instead act through a structure called the subthalamic nucleus that causes broad inhibition of thalamo-cortical drive (Figure 1).
These observations have all concerned stopping movement, however. What about stopping a thought? There are surely instances where fast inhibition of a thought would be useful. Unwanted and unpleasant thoughts do occasionally avail themselves, and indeed, such intrusions are a feature of several psychiatric disorders ranging from post-traumatic stress disorder to ADHD to obsessive compulsive disorder.
Where thought is concerned, the evidence is more limited for suppressive inhibition versus the alternatives. Nonetheless, one place where there is growing evidence for stopping of cognition is in the fast stopping of memory retrieval, such as is tested using the think/no-think procedure. This procedure cues when to stop retrieving a memory associated with a cue. People can stop retrieval in this way, and when they do, similar cortical networks and neural dynamics are observed in the brain as seen in fast motor stopping. So, it seems likely that we do have some ability to broadly stop thought, perhaps with similar mechanisms as motor stopping.
Thus, fast, global inhibitory mechanisms for movements, and possibly thought, complement the slower selective gating mechanisms we discussed earlier. The cortico-striatal mechanisms can select or prohibit specific actions, goals, and contexts by gating working memory. They are content addressable, but also fairly slow. By contrast, the inhibitory stopping circuit is fast and global, sacrificing content addressability in favor of efficiently stopping unwanted actions. Thus, in the complexity of our everyday behavior, these systems likely interact to balance the need for speed versus specificity in control.
Inhibition as an account of self-control
Today’s discussion should make clear why it is perilous to take any instance of error, impulsive act, or intemperance of the moment and attribute it to a failure of inhibition. Indeed, the same disinhibited behavior could arise from multiple distinct underlying mechanisms. Thus, it is no surprise that when people have tried to correlate one measure of inhibition, such as stopping, with real world impulsive behaviors, the results have been mostly weak. This does not mean that we should not take these mechanisms seriously. Rather, it warns against theorizing that relies on one general mechanism of inhibitory control as an account of our complex and multi-faceted ability to control ourselves.
Hi David. I was going to ask about thought control so I’m glad you’ve brought it up here. You didn’t talk about thought as an action in previous posts so I wasn’t sure if you included it. I take it there’s a background idea here that talking is an action like other actions that we can build up generatively from components. Inner speech is just talking where we don’t let the words come out.
My intuition is that we can never switch off a thought, but as you suggest- we can think about something else instead. It does seem like we have some control over how the thought comes out though- such as whether we think in words or how vivid the image is.
One consideration- in your previous post you talked about the control process interfering between pre-motor and motor cortex (if I understood correctly). If inner speech is understood as the planning of outer speech in the pre-motor, then might the gating processes you described be unable to apply?
Hi Tom. Thanks. This is an interesting question about inner speech. If I understand how inner speech relates to outer speech in this case, then perhaps the multiple nested gating loops I discussed in the 3rd post are relevant to consider. I would agree with you that the specific cortico-striatal loop that regulates premotor to motor interactions wouldn’t apply to planning of inner speech, as you describe it. However, the higher-order cortico-striatal gates that regulate prefrontal cortex would. These loops regulate the input and output of the prefrontal cortex and control memory for more abstract goals, plans, and thoughts. Further, in this framework, these multiple gating loops allow contexts themselves to be conditioned on other contexts and so offer a means of implementing deeper hierarchical rules. For example, I can condition my use of driving as a context for different actions in the car based on some other contextual factor, like the local laws where I’m driving. That is helpful for control of componential thought and action, and I think, for the conception of inner speech you describe here.
Hi David,
I just “binge read” your posts on cognitive control, and I really enjoyed them.
You identified a number of core themes in cognitive control research. I wanted to get your thoughts on another one, an area of I’ve been thinking and writing about a lot lately: cognitive control and evidence accumulation. We know from decades of work in mathematical psychology that many response types can be modeled as the accumulation of evidence towards a response boundary, and such models are applicable across motor actions, value-based choice, attention, memory, etc.. Building on this, one potential unifying way to characterize what cognitive control actually does in our psychology is that it modulates spontaneous patterns of evidence accumulation, biasing evidence accumulation in a goal-appropriate direction. So-called “conflict DDMs” provide a more precise realization of this idea.
The nice thing about this approach is that provides a nice theoretical unification between two areas of computational psychology (DDMs and cognitive control research) that haven’t gotten together much until recently. So I thought I’d ask you for your take on the prospects for this approach for providing a general characterization of the “upshots” of cognitive control.
Hi Chandra,
Thanks for reading. In line with your point, evidence accumulation models are increasingly influential in cognitive control research, and they have been quite useful in some settings. As one example, motor stopping through the subthalamic nucleus that I discuss in this post has been implicated in boundary shifts in these decision making models. In simple terms, there is evidence that when there is uncertainty in a choice, the STN will drive inhibition of the motor response. This effectively puts the breaks on responding and allows more time for evidence to accumulate. This dynamic can be modeled using an evidence accumulation model as a dynamically increasing threshold or decision bound. I have cited a couple papers on this below for your reference. There are limits to these models, of course. Nonetheless, they can be a powerful theoretical tool for understanding cognitive control.
Ratcliff R, Frank MJ (2012) Reinforcement-based decision making in corticostriatal circuits: mutual constraints by neurocomputational and diffusion models. Neural Comput 24:1186 –1229.
Cavanagh JF, Wiecki TV, Cohen MX, Figueroa CM, Samanta J, Sherman SJ,
Frank MJ (2011) Subthalamic nucleus stimulation reverses mediofrontal influence over decision threshold. Nat Neurosci 14:1462–1467.
Frank, MJ, Gagne, C, Nyhus, E, Masters, S, Wiecki, TV, Cavanagh, JF, and Badre, D (2015) fMRI and EEG predictors of dynamic decision parameters during human reinforcement learning. Journal of Neuroscience 35(2): 485-94.