On your marks, get set, stop (by K. Dunovan)

A blog post on the article Competing basal-ganglia pathways determine the difference between stopping and deciding not to go (pdf).

As kids, we are constantly reminded of the need for self control – by our teachers, (“sit still and be quiet!”); by our parents (“stop bothering your sisters!”); by our coaches (“wait for it… swing!”). Over time we get better at controlling our behavior, relying more on our past experiences and future goals to gauge when to exercise restraint and when to swing for the fences. Unfortunately, as in the case of baseball, you can’t hit a homerun off of sheer will power. You still need external cues about the speed and trajectory of the ball in order to make a decision, giving you only a few hundred milliseconds once it leaves the pitcher’s hand to decide if it’s worth swinging. Pointing your bat at the fence is easy, the hard part is suppressing the urge to swing when it’s high and inside.

(Unless you’re this kid. In which case, you still blast it out of the park )

There are two general mechanisms for inhibiting an action – 1) we can proactively choose not to move because it conflicts with our current goals or 2) we can reactively cancel a planned action after encountering an environmental cue to stop (i.e., putting the brakes on an action). While it is easy to see that these are distinct forms of inhibitory control, they might in fact interact with each other. For instance, if you’re less certain about whether or not to execute a particular action, are you better at reactively cancelling when needed? Intuitively, it would seem so; however, experimental evidence for such an interaction has been mixed. In fact, the separability of different control mechanisms has become a central topic of debate in cognitive neuroscience, largely driven by ambiguity in the neural networks underlying reactive and proactive forms of control.

Proactive and reactive inhibitory control are classically associated with a network of brain areas known as the basal ganglia. This common neural circuitry has led some to argue, or at least speculate, that proactive and reactive inhibitory control originate from the same brain computation. However, recent evidence has shown that they may rely on distinct neural pathways that both pass through the basal ganglia. As it turns out that there are many pathways that run through the basal ganglia. Activation of the direct pathway serves to facilitate action execution, giving it the nickname of the “Go” pathway. Whereas activation of the indirect pathway provides the selective action suppressing signals necessary for proactive control (e.g., decide not to swing at first pitch before the ball is thrown), giving it the nickname of the “No-Go” pathway. Finally, activation of the hyper-direct pathway performs rapid, global action cancellation, acting as a late “Brake” for reactively stopping a planned action (e.g., stop swinging if pitch is high and inside).

While all three pathways are traditionally considered to operate as parallel, non-interacting processes, there are several areas within the basal-ganglia where they overlap with each other. For instance, all three pathways converge before exiting the basal-ganglia, allowing for the possibility that reactive and proactive interact as they pass through this network. If so, what are the underlying dynamics that distinguish proactive action suppression from reactive action cancellation? How might these different control mechanisms interact?

In a recent study published in the journal ELife, we propose a novel model in which contextual factors, such as uncertainty and reward expectations, are encoded by the dynamic competition between action facilitation (e.g., direct, “Go” pathway) and suppression (e.g., indirect, “No-Go” pathway) (Dunovan et al., 2015). In this framework, a single decision process tracks the direct-indirect pathway difference, accumulating toward the decision threshold as activation of the direct pathway overpowers that of the indirect pathway. A distinguishing feature of this model is that proactive modulation of the direct-indirect competition determines the initial state of a separate “braking” process (i.e., hyper-direct), making it more difficult to reactively cancel a planned action that is closer to the execution threshold.

It kind of works like this (http://goo.gl/RXpszN.): Over time, as you prepare to make a movement, there is a buildup of “Go” activity (green squiggly line). If a cue to stop comes later in time, there is more “Go” activity to override than if the stop cue had been presented early on, which makes it more difficult for the “Braking” process to cancel the movement (red squiggly line).

We put our model to the test, assessing its ability to account for behavioral data in two different inhibitory control tasks, one in which the decision to inhibit a response was based on a stop cue (e.g., reactive stopping task) and another in which subjects used prior knowledge to decide whether or not to act (e.g., proactive no-go task). In addition to making behavioral predictions, we wanted to know if our model could account for subjects’ brain activity during the task. To this end, we used functional magnetic resonance imaging, or fMRI, to record neural activation of the basal ganglia and other inhibitory control regions and compared these recordings with simulated “brain activity” generated by the model. Compared to previously proposed models, our so-called Dependent Process Model (DPM) provided a much better description of reactive stopping behavior, suggesting that cancelling a planned response depends on how prepared you are to either execute or suppress your movement. We also found that, when we increased the penalty for failing to inhibit a response, subjects exercised more caution in their decisions by proactively slowing down the buildup of the execution process, making it easier for the braking process to terminate the action.

The way the penalty slowed down people’s decisions hints at a new way of thinking about proactive control. Perhaps what happens when you make a “no-go” decision is that you slow down the buildup of your decision so much that you effectively run out the clock. If so, this would provide insight into how proactively deciding not to go is fundamentally different from reactively stopping an action. We tested this by fitting our model to behavior from a modified proactive control task, in which the decision to execute or withhold a response was based solely on prior knowledge about which of the two outcomes would be correct (i.e., rewarded). Indeed, the model that best explained the data was the model where the rate of accumulation (or “drift” as they say in the decision making literature) was affected by context. In other words, it seems that subjects accelerated or decelerated their decision process based solely on their expectations of being rewarded or penalized for making a response. Not only could we explain the behavioral data, but the model also accurately captured the fMRI activation in the basal ganglia as participants completed this proactive decision task. So we could capture both behavior and neural responses in the same model. Two birds, one computational stone.

The true benefit of this framework is that, using the known architecture of neural pathways we can build a decision model that provides a parsimonious and unifying description of how context affects control decisions, while at the same time distinguishing between reactive (i.e., braking) and proactive (i.e., deciding not to go) mechanisms for stopping our actions.

Source Article: Dunovan, K., Lynch, B., Molesworth, T., & Verstynen, T. (2015). Competing basal-ganglia pathways determine the difference between stopping and deciding not to go. eLife, e08723.


Leave a comment