The waiter asks whether you’d prefer potatoes or rice with your entrecote. Are you free to make that decision based on, well, free will? Or should you respond with “Look, I am a determinist. I will just wait and see what I order because I know that my order is determined”?

At first glance, the idea that we are not free to choose or make up our own minds seems preposterous: We all feel that we are free to make decisions and may spontaneously decide to buy ice cream or a new car or begin using Bayesian statistics rather than their frequentist counterpart.

Nonetheless, a number of influential neuroscientists have argued that free will is an illusion. On this account, the conscious articulation of a wilful decision (“I think I could do with a mint chocolate chip cone right now”) is not causal, but a consequence of neural activity that has made the decision for us. This view is shown in the figure below, where GES stands for genes, environment, and stochasticism—in other words, the variables that are truly causing our brains to make decisions before letting them become conscious as some sort of polite courtesy (and to keep philosophers in business for centuries). In other words, our brain makes us do it, but it does so in a way that creates a perfect illusion of free will—sometimes known as “willusion.”

Support for this view arises from a famous series of studies in which participants were asked to perform an action (move a finger) “at will,” at a time of their choosing. Detectable changes in activity in the brain preceded movement by half a second, and when participants were asked to announce their conscious decision to move the finger, this reported conscious awareness followed the onset of neural activity by about 300 ms (a third of a second). These observations are very much in line with the model in the preceding figure, although the studies are not without their incisive critics.

A recent article in the Psychonomic Society’s journal Attention, Perception, & Psychophysics contributed to this debate about whether intentionality can be detected in the brain by measuring brain activation during an attentional allocation task.

Research scientist Leon Gmeindl and colleagues reasoned that existing studies on this topic either provided an external cue (as in the waiter asking you about your food preferences) or required an overt response (as in the finger-movement task). Although either one or both of those elements may appear necessary at first glance, Gmeindl and colleagues found a clever way around this seeming constraint.

Their experiment involved two conditions, one of which is shown in the figure below:

In this cued condition, participants watched a sequence of visual stimuli and were instructed to fixate their eyes on the central dot. In addition, they had to shift their attention—but not their eyes!—into the direction indicated by the red letters “R” (for right) or “L” (for left). The stimuli flashed up at a rapid rate and people were instructed to keep their attention on a cued side until the cue to shift to the other side appeared. (To verify the location of attention, participants had to press a button to identify an occasional digit in the attended part of the display: in the above figure, they would press “5”, when the digits appeared after the first shift cue.)

Participants’ brain activation was measured via fMRI throughout, and during the analysis phase a so-called pattern classifier was trained to associate the location of attention—whose timing was known because it shifted in response to the cue—with the pattern of activation present at the time. In essence, a pattern classifier is a statistical machine that learns to detect the brain’s signature in response to an externally-cued event. In the present case, the classifier learned to distinguish between the brain’s activity when attention was oriented to the left versus the right stream of stimuli.

Now on to the second, uncued condition. That condition was virtually identical to the cued condition, except that the “L” and “R” cues were absent and participants were instructed to shift attention voluntarily from one side to the other, at a time of their choosing but at a pace of roughly three or four times each minute. Participants were thus left to their own devices, although they still had to identify the digits to verify the location of their attention.

The question of interest was whether the pattern-classifier that had been trained during the cued condition could identify where people’s attention was focused and, by implication, when they chose to shift it on their own account. Gmeindl and colleagues obtained a measure of the extent to which, at any point during their experiment, the pattern of activity identified by the classifier corresponded to the patterns known to be associated with attention being focused on the right or left, respectively.

The pattern classifier enabled Gmeindl and colleagues to detect clear shifts in attention for each participant, although they were of course dispersed over quite different points in time. The next question of interest was whether the time course of brain activation associated with these uncued attention shifts would differ from the time course observed in response to explicitly cued shifts.

The data that resolve this question are shown in the figure below for 4 different brain regions (which prior research had already identified as being of particular relevance):

In each panel in the figure, the graphs that plot extent of signal change (i.e., change in brain activation) are centered on a point that corresponds to the time when the attention shift was detected in the brain.

It can be seen that for at least two of the brain structures, the dorsal Anterior Cingulate Cortex (dACC; panel C) and the right Middle Frontal Gyrus (rMFG; panel B), activation preceding an uncued attention shift was detectable earlier than when attention was cued. Uncued activation was also stronger than cued activation: indeed, in one region, no change in activation was detectable when attention was cued whereas it skyrocketed for uncued attention shifts (panel C). Gmeindl and colleagues therefore conclude that this area of the brain, the dorsal Anterior Cingulate Cortex (dACC) is one of the “core components of the brain network underlying the will to act” (the other one being the right Middle Frontal Gyrus, or rMFG, in panel B).

The results of Gmeindl and colleagues mesh well with several other recent investigations of volitional acts. It appears as though we may home in on the location of “free will” in our brains.

So is our will free or is that a willusion? The results of Gmeindl and colleagues cannot resolve this basic question. However, even if it should turn out that neuroscience ultimately confirms without a doubt that our conscious recognition of a “will” always follows brain activity that we are unaware of, and if we can predict our “will” from that activity ahead of time, even then this need not do away with the concept of free will: A recent study showed that people understand our will to be free even if it can be perfectly predicted on the basis of prior neural activation. Just because “my brain made me do it,” does not mean that I didn’t do it of my own free will.

Article focused on in this post:

Gmeindl, L., Chiu, Y. C., Esterman, M. S., Greenberg, A. S., Courtney, S. M., & Yantis, S. (2016). Tracking the will to attend: Cortical activity indexes self-generated, voluntary shifts of attention. Attention, Perception, & Psychophysics. DOI: 10.3758/s13414-016-1159-7.

This is an article to be published in a special issue of Attention, Perception, & Psychophysics that is honoring Steve Yantis. This article is one of four that will be free-access for the next three months.