Developing Intelligence

When you need to stop yourself from committing some response, do you simply freeze – like a deer in the headlights – or can you selectively inhibit only the undesired action? The question is important because the ability to stop or inhibit a planned or prepotent action may be a central feature of so-called “executive functions,” those capacities which are tightly related to fluid intelligence, academic success, and genetic factors. Yet surprisingly little is known about whether stopping can be global, selective, or both.

Some answers to this question come from selective stopping paradigms. For example, key presses may be required from both index fingers for one stimulus, and for both middle fingers for another stimulus; occasionally, a stop signal will occur that is specific to one hand. These paradigms reveal a “stopping interference” effect, in which the response from one hand is slowed if the other hand’s response must be inhibited – as though a global stopping mechanism has been engaged by default.

In a new Psych Science paper, Aron & Verbruggen argue this “stopping interference effect” reflects the underlying neural circuity of response inhibition. Specifically, they suggest that global stopping engages a “hyperdirect” pathway between prefrontal cortex and the basal ganglia (think deer in the headlights), but that more selective stopping is implemented by an “indirect” pathway (between the thalamus, external & internal globus pallidus, and subthalamic nucleus).

This hypothesis motivates a few predictions: first, true selective stopping should be slower than global stopping, since the direct pathway is monosynaptic and the indirect pathway is multisynaptic. Second, subjects should be encouraged to use a more selective stopping mechanism – and thereby show less of the stopping interference effect – when they have foreknowledge about which hand might require inhibition on each trial.

Aron & Verbruggen confirmed the first of these predictions by showing that “stopping” was less efficient when subjects had foreknowledge about the hand that may need to be inhibited (as indexed by SSRT – a common but complex dependent measure in stop signal paradigms). As the authors note, however, this difference could reflect additional cognitive load induced by having informative foreknowledge. A second experiment showed that less informative foreknowledge (where the foreknowledge pertained only to where in visual space the stop signal would occur, and not to preparation of responses) was not associated with less efficient stopping (again, as indexed by SSRT).

Aron & Verbruggen also found that the stopping inteference effect – slowing of the response that was not-to-be inhibited – was smaller in the case where subjects had foreknowledge. Aron & Verbruggen suggest that this evidence in combination with the slower SSRT for “selective” stopping are strong evidence for two mechanisms of inhibition.

I want to briefly explore a few other possibilities for what this “stopping interference” effect means. One might observe less slowing of the not-to-be-stopped response (with foreknowledge, relative to no foreknowledge) if:

1) subjects additionally “fortify” that not-to-be-stopped response ahead of time (this makes a simple prediction in terms of LRPs). The authors imply this kind of explanation is unlikely, since go reaction times on non-stop trials didn’t differ between conditions, but on the other hand, a) strong conclusions cannot be drawn on a null effect, b) this type of additional preparation could easily fit within the 2s foreperiod, and c) although not significant, numerical trends go in the right direction (lower RT for noncued than cued hand, and lower correlations between these RTs in the foreknowledge condition).

2) subjects always engaged a global stopping mechanism to stop signals, and experienced some facilitation in reinitiating not-to-be-stopped action due to their foreknowledge. This would be expected to yield some stopping interference cost even in the foreknowledge condition, but smaller than that in the no-foreknowledge condition. Indeed, this was observed in experiment 1. This would not necessarily predict less efficient stopping in one condition vs. the other, unless the engagement of the global stopping was delayed by informative foreknowledge.

3) subjects undergo attentional capture due to the stop signal, as a function of its unexpectedness (akin to an oddball effect), and foreknowledge reduces this attentional capture. This yields less stopping efficiency (since, in this case, attentional capture improves stopping efficiency) and less stopping interference (since attentional capture incurs a global cost on responses). Although both this account and Aron & Verbruggen’s would predict that individual differences in stopping efficiency should correlate with those of stopping interference, as well as individual differences in the change of both as a function of foreknowledge, only this “attentional capture” account predicts these effects should also be observable in attentional capture tasks that do not involve inhibition.

In summary, Aron & Verbruggen have used a innovative behavioral task to attempt to index different neural pathways, and while there are some alternative explanations (as for any scientific study), this work is nonetheless a huge contribution to our understanding of stopping. It appears that the “deer in the headlights” response is not all we have at our disposal, and that with appropriate foreknowledge, we can be more selective in our use of inhibition – albeit with some cost in efficiency.


  1. #1 Thad
    December 17, 2008

    How does this theory work for a checked swing in baseball? Is there any kind of foreknowledge that the swing will be stopped? e.g. I might decide at the last minute to not swing, but I can’t make that `decision` until the last 1-2 milliseconds.

  2. #2 CHCH
    December 17, 2008

    what an interesting question. I wish I’d framed the study in that context!

    It’s true that baseball players know exactly what muscles must need to be stopped, and they’re probably very aware that this is a possibility for every pitch. This could actually be detrimental if you believe that all they need to do is “freeze” – a lack of foreknowledge may encourage more efficient global freezing via the direct pathway. On the other hand, it’s more likely that they need to inhibit some responses and enact others; for example, they may actually need to retract the bat so that it doesn’t cross the plate. In this case, it seems that the enacting of alternative responses (retracting the bat) may be the more important variable for performance, and would not use the inhibitory system described here.