Can Neuroscientists Measure Free Will?

EDITOR’S NOTE: This is the second of a three-part series on the Human Mind Conference, an “interdisciplinary event” held in Cambridge, England, in June 2017, “bringing together a wide range of experts from across the humanities and the cognitive sciences to discuss key aspects of mental life and experience.”

If you’ve ever read an article proclaiming that neuroscience disproves free will, you’ve probably heard of the Libet experiment.

In 1983, Benjamin Libet, a vision and neurology researcher at UC San Francisco, described an experiment in which subjects were asked to move one of their hands whenever they felt like it, and to report the instant at which they first felt the urge to move. The subjects’ brains were also being measured using an electroencephalogram, or EEG. Libet found that, just before the subjects reported feeling the conscious urge to move, there was a distinct pattern in the EEG readout: a slight increase in negative electrical potential on the alpha-wave, a specific brain wave. Libet, building on the work of earlier researchers, called this pattern the “readiness potential.”

The remarkable finding of Libet’s experiment was not the readiness potential itself, but that it apparently preceded the conscious urge to move, in some cases by up to a second. The claim advanced by Libet and many others was that the readiness potential is an empirical indicator of the underlying neurological mechanism that we experience as volition. Our conscious experience of free will is therefore merely something we report after the fact, once our brains have initiated the action.

Free will can thus be measured, the argument went — and it is the product of neurochemistry, not the conscious mind.

The Libet experiment has come under extensive fire over the years, both for methodological flaws and for its seemingly trivial definition of free will. In a talk at the Human Mind Conference, held last June in Cambridge, England, University College London neuroscientist Patrick Haggard reported on efforts to improve the Libet experiment. Haggard — who collaborated with Libet when he was still alive — argues that he and others have since created neurological tests of free will that are more clearly falsifiable and philosophically robust.

Haggard says that our understanding of the readiness potential has been “blown open” by the work of Aaron Schurger, currently a researcher at the French Institute of Health and Medical Research. Schurger noted a common criticism of the Libet experiment: that the action of the experimental subject seems not so much free as random. As Haggard explains:

If we imagine somebody in the Libet experiment who just has to press the button at some time that they feel like it, perhaps the context of the experiment — the fact that you’ve got a button on the table in front of you — moves some neural activity closer to a motor threshold.

In other words, how do we know that it isn’t just random neural noise prompting the person to feel like pushing the button at an arbitrary time? “The problem is that…when we study the readiness potential, we’re doing biased sampling,” says Haggard. “We wait until the person presses the button and then we look and see what’s happening before.” Schurger used simulations to show that this biased sampling of measurements can produce a pattern like the readiness potential from a process that was, in fact, random. This suggests that, at the very least, the implicit conception of “free action” at work in the original Libet experiment needs to be refined.

In response to these criticisms, Haggard and his team have “recently been trying to embed the idea of volition in a slightly more rich, plausible framework.” To that end, they’ve adapted an experiment already well established in monkeys. The experiment goes like this: A subject is asked to watch a pattern of still dots on a screen, then press a button when the dots begin to move. If the subject correctly indicates the direction of movement, he or she receives a reward (cash, in the case of humans). The wait time is highly variable — between 2 and 200 seconds. Subjects are thus given the option to skip a screen when they get bored, at the cost of a significantly reduced reward. The result is that test subjects face a decision with genuine stakes. And when the dots begin to move, experimenters begin checking in at random intervals to observe this internal deliberation. According to Haggard, this random sampling eliminates the bias of the original Libet experiment.

Haggard calls the impulse or decision to skip a screen when bored the “skip response.” And this response is “actually remarkably like, quote, ‘free will.’ Because the key thing about the skip response is you don’t make the skip response in response to any external stimulus.” Thus this experiment responds to the objection raised by some philosophers that the Libet experiment does not measure a truly free human action — one that has “freedom from immediacy.” Haggard contends that the “skip response” better measures this view of freedom because it differs in kind both from reflexes in response to external stimuli and from random impulses.

In additional to randomizing the sampling of skip volition, Haggard created a “critical control condition.” This control consists of a “completely separate block of trials where you cannot choose for yourself to skip but you’re told to skip.” The measured difference between these externally generated skip responses and the internally generated ones “gives us our operational definition of volition.”

What Haggard’s experiments have found, he claims, is a messier, yet still clear neurological pattern associated with volition.

For each of the two sets of trial subjects — those who were externally prompted and those internally prompted — Haggard measured the standard deviation (the degree of variation) of the EEG signal:

If the brain is doing everything and anything prior to a voluntary action, you might expect the standard deviation to be high. If the brain is doing one specific, particular precursor activity of getting ready to make a voluntary action, you might expect the standard deviation to be quite low.

And, indeed, Haggard found that the signal’s variance was low prior to the voluntary actions — or at least, significantly lower than prior to the externally prompted actions. Haggard interprets this difference as evidence that “there’s some extra thing going on, if you like, in the brain, which causes a low variability” in the voluntary trials. That is, “the brain is going through a relatively consistent process prior to the voluntary action.”

Other experiments seem to corroborate these findings. In one variant on the experiment described above, subjects who have already gone through the internally prompted trial are asked to repeat the test. But this time, they are allowed only a limited number of skips, increasing the effective cost and meaningfulness of the choice. The result is an even stronger convergence — that is, a lower variance — in the EEG signal.

“Philosophers always insist to me that I can’t talk about actions unless there’s something reasons-responsive,” said Haggard. Here he is alluding especially to the work of moral philosopher G. E. M. Anscombe, who argued that free will requires that an agent’s actions be at least partially the result of rational deliberation. The arbitrary choice of the Libet experiment doesn’t seem to meet this standard. But Haggard argues that his versions of the experiment are not only reasons-responsive, but proportionately so: Their operational measure of volition (a decrease in alpha-wave variance) increases as actions become less habitual and more deliberate. Accordingly, voluntary action may be explainable in part as “some kind of executive process which can self-regulate your own brain noise.”

Only at the end of his talk does Haggard make clear that he does indeed view his own work as offering a more robust confirmation of Libet’s original conclusion. In other words, Haggard remains a neurodeterminist about free will.

Haggard’s revisions of the Libet study are rather ingenious, laudable as a model of how psychological experiments can be responsive to philosophical criticism, so that the disciplines need not remain entirely at odds. His case that he has resolved the methodological flaws of the Libet experiment — that he has devised a measurement of some aspect of volition — is persuasive. But his case for neural determinism is not.

On Haggard’s interpretation, his work complicates the conventional deterministic view while still ultimately confirming it:

Crucially, conscious volition is not only retrospectively confabulated post-hoc. This is what most [scientists] seem to believe, particularly in North American psychology. In my view, it may also involve some real-time read-out of neural precursor activity. The will — just to ontologize it — the will may not be free, but I think there is a bona fide neural-cognitive process of volition, and I think it has some associated phenomenology [or felt experience].

But even if we accept that Haggard has discovered a genuine neural precursor to volitional action, this finding hardly commits us to a deterministic view of free will, or to the view that our conscious minds are merely spectators or troubadours of the brain’s sovereign play.

The problem is that Haggard has made volition the beginning and end of freedom. Haggard’s experiment, for example, would also seem to confirm views of action like those offered by Aristotle or modern-day virtue ethicists. These philosophers not only accept but emphasize pre-conscious cognitions as central components in rational action. For many modern thinkers, the “will” is like a deliberative force, fighting against something, such as our own impulses — a view that owes much to Kant. For virtue ethics, by contrast, freedom need not manifest as struggle, but is instead understood as a means to achieving some rational end. (In fact, the notion of a will does not seem to appear at all in Aristotle’s account of rational choice.) If virtue ethicists are right, then perhaps freedom can be understood to work in concert with decisional impulses, rather than being threatened by them.

For example, let’s say I decide not to commit murder. My decision is rational not only because I have deliberated about the reasons not to do it, but also because my decision flows from a character that has been formed in a rational way. When faced with the choice to murder, my dispositions have already been shaped, e.g., by membership in a society that professes to value human life, by individual reflection, or by both. And if this is the case, then when confronted with the choice to murder, my decision is not simply the result of a logical process of deliberation but also a response to a well-formed impulse — something that might well appear from a neurological perspective to be a “real-time read-out of neural precursor activity.”

In other words, Haggard offers a convincing case for the measurability of a reasons-responsive impulse to act. But he has not shown why the existence of such impulses must doom human freedom. Why, when faced with a choice — whether to kill, for instance — am I confronted with this impulse, this sense of will? Might it not be central to the work of a free, rational agent to interrogate and reshape these responses?

Neuroscientists continue to claim that they have annexed free will to their domain. But, at least for now, freedom belongs still to the philosophers.