What does neuroscience have to say about free will?
Above: Mouse cortical neurons after 15 days in culture. [https://commons.wikimedia.org/wiki/File:Neuronal_web.tif]
[Editor’s Note: Aaron Schurger is principal investigator (chargé de recherche 1) with the French National Institute for Health and Medical Research (INSERM), based at the NeuroSpin Research Center near Paris, and also holds a faculty appointment in the Department of Psychology and Institute for Interdisciplinary Brain and Behavioral Sciences at Chapman University. His research focuses on the neural signatures of subjective experience and the neural antecedents of self-initiated movement. He received PhD in psychology and neuroscience from Princeton University. In 2013, Schurger was awarded the William James Prize from the Association for the Scientific Study of Consciousness (ASSC) and in 2015 was awarded the BMI-Kaloy prize from the Kaloy Foundation for his 2012 work on the role of spontaneous fluctuations in brain activity in self-initiated movement. Dr. Schurger graciously accepted an invitation from the Society of Catholic Scientists to write this article for the SCS website on his groundbreaking research and its potential implications for the age-old question of human freedom.]
Hardly any question could be more important than whether human beings have free will. Of course it is of specific importance for some religions, but also for many non-religious people. Indeed, historically, many of those who rejected religion did so in the name of intellectual freedom and called themselves “freethinkers”. The question is whether science allows us to believe in freedom of any sort, whether intellectual or moral. In the last half century, there have been claims that neuroscience was yielding evidence against the reality of human freedom. What I will discuss here is whether that is the case. My goal is not to argue either for or against the reality of freedom, but only to examine what neuroscience is at present able to say on the subject — and what it is not able to say.
The topic of free will is complicated by the fact that different people mean different things when they talk about “free will”. The everyday commonsense notion of free will is something like freedom from external influence or control: If a given action was, at least in part, determined by an external influence, whether immediately present or not, whether real or imagined, then that act was not purely an act of free will. For example, if someone compels you to do something at gunpoint, then your act is not purely an act of free will. As it turns out, this simplistic view is hardly simple at all: Even at gunpoint, you could still refuse to comply. And even in the absence of any external influence, what if you were mentally unstable when you performed the action? Does it still count as an act of free will? What if you had a brain tumor that altered your personality and behavior? Can it still be considered an act of free will? There are countless ways, according to common sense and according to the law, that one’s “free will” (and hence one’s moral responsibility) might be diminished or compromised. But this everyday notion of free will doesn’t quite answer the question of what free will is. It simply says that, barring any external influences, one can exercise it.
In neuroscientific and philosophical circles, many consider the bottom line of free will to be the ability to do otherwise. That is to say, for a given act, if it was not possible for you to have done otherwise, then that act could not be considered an act of free will, because, well, you could not have done otherwise. The debate about whether the universe is deterministic or non-deterministic is relevant here. If the universe marches along from one state to the next in a mechanistic fashion, and whatever happens next is fully determined by what happened just prior, then, for any given act, it could not be possible to have done otherwise. Each and every event, including your and my actions, would be fully pre-determined — this pre-determination would include all matter and energy in the universe, including the neural activity in your brain, which, from a scientific standpoint, determines your behavior. Thus, if the universe were deterministic, then free will in the strictest sense of “the ability to do otherwise” would not be possible. However, experts in quantum physics tell us that the universe is not deterministic because events at the quantum level are governed by probabilities and randomness. This would seem to provide us with the possibility of freedom, but it is difficult to see how it could grant us a “will”, since the will is something that should determine our actions in a meaningful way, not just randomly. This apparent impasse has led to the elaboration of other conceptions of free will such as compatibilism — the idea that we can have a kind of free will even in a fully deterministic universe. A discussion of the many different philosophical notions of free will is beyond the scope of this article,1 but it is at least worth knowing that thinkers are not all in agreement as to how to define free will. Here I will focus on the role of consciousness in decision making, because this is one area where neuroscience research is thought to undermine the possibility of free will, as most people conceive of it.
One commonly agreed upon aspect of free will, among neuroscientists, philosophers of mind, and laypeople alike, is that acts of free will should be acts that originate in a conscious decision. Unconscious motives and unconscious decisions (if such things even exist) could not, in many people’s eyes, be the basis for acts of free will. A proper act of free will should follow a chain of events where a conscious decision precedes and somehow directly causes a subsequent action. In fact, this condition is so thoroughly engrained in commonsense notions of free will that we might as well refer to it as conscious free will. The distinction is meaningful because the brain activity that triggers a given act, and the brain activity that underlies the conscious feeling of having decided to perform that act may not be the same.
In one well-known experiment by Michel Desmurget and Angela Sirigu of the CNRS in France, patients undergoing brain surgery for medical reasons had electrical stimulation applied directly to different regions of their cerebral cortex.2 Stimulation of the posterior parietal cortex caused the patient to report feeling the intention or urge to move a specific body part. At stronger stimulation intensities patients additionally reported having performed those specific movements, even though they had not moved at all. By contrast, stimulation in the premotor region of the frontal lobe provoked actual bodily movements, even though the patients “firmly denied that they had moved.” Thus, the feeling of an intention or urge to move and the initiation of the movement itself are governed by different regions of the brain, although under normal circumstances the two are tightly coupled.
From a neuroscientific standpoint, the connection with consciousness is arguably just one among a handful of necessary conditions for a given act to be an act of free will. A proposed list of necessary conditions was recently offered by cognitive neuroscientist Patrick Haggard of the Institute for Cognitive Neuroscience at University College London. According to Haggard,3 the key features of the brain activity underlying freely willed actions are the following: (1) the brain activity should lead to movement (or at least to some consequence in the outside world), (2) there should be no immediate external trigger for that brain activity, (3) the ensuing action should be reasons responsive, meaning that it should arise in relation to a reason (rather than just randomly or for no particular reason), (4) the ensuing action should be outcome- or goal-directed, (5) the ensuing action should be spontaneous or innovative, i.e. not automatic or habitual, and (6) the neural activity (and ensuing action) should involve a conscious decision. These being necessary conditions, if any one of them is cast into doubt, then the action’s status as an act of free will is also cast into doubt.
From among the necessary conditions listed above, by far the one that has been most severely challenged, by neuroscience and experimental psychology research, has been the role of consciousness. If conscious decisions do not play a direct causal role in bringing about our actions, then, the reasoning goes, we do not have conscious free will, which, let’s face it, is the kind of free will that most people would want to have. The most prominent challenges to the role of consciousness in decision making and the initiation of action have come from the electroencephalography (EEG) experiments of Benjamin Libet in the early 1980s,4 and from Daniel Wegner’s research on apparent mental causation in the 1990s,5 Subsequent research in the 2000s and 2010s helped to cement the already prevailing view that the brain makes decisions and initiates actions at a pre-conscious stage and the conscious feeling of having willed the action and being the author of the action are added on after the fact.6,7 This view had become so firmly established in the field that in 2016 the Atlantic Magazine boldly declared that “There’s No Such Thing As Free Will.” 8 In what follows, I will discuss the research that helped to establish and cement this view, as well as research of my own that ultimately upended it and reignited the dialog on the role of consciousness in decision making.
The modern era of neuroscience research on volition (the scientific term for the will) began in the 1960s with the discovery of the Bereitschaftspotential or “readiness potential” (RP). In 1965, Hans Kornhuber and Luder Deecke reported their finding, using electroencephalography (EEG), of a slow buildup of electrical potential in motor areas of the brain that preceded the onset of a freely-willed movement (e.g. pressing a button) by up to a full second 9,10 — which is a long time on neural timescales. Part of the novelty of this experiment was that there was no stimulus, no sensory cue to dictate to the subject when he or she should perform the movement. Subjects decided on their own when they wanted to perform the movement and did so, more or less, at will. Kornhuber and Deecke described their newly-discovered brain potential as “the electrophysiological sign of planning, preparation, and initiation of volitional acts.”
Kornhuber and Deecke’s experiment opened the door to the neuroscientific study of “freely-willed action”, but did not undermine everyday beliefs about free will. The defining experiment would come nearly 20 years later, in 1983, when neuroscientist Benjamin Libet published his now classic study of the chronometry of conscious volition.4 Libet recorded EEG while his subjects performed spontaneous voluntary movements, similar to what Kornhuber and Deecke had done, but with a new twist. Libet also asked his subjects to monitor a rapidly rotating clock dial, with a small dot that made one cycle around the clock face every approximately 3 seconds (about 50 milliseconds per tick on the clock). Subjects were instructed that, after each movement, they should report the location of the dot on the clock when they first were aware of their will (Libet used the word “urge”) to perform the movement (a flick of the wrist). The results are about as well-known as any scientific experiment could ever be: On average, the reported time of awareness of intent to move was about 200 milliseconds (2 tenths of a second) before the onset of the movement, whereas the EEG readiness potential began its buildup at least 500 milliseconds or more before the onset of the movement. The brain, it seemed, had already “decided” to initiate the movement well before subjects were aware of their own decision to initiate the movement. According to Libet, “…the brain evidently ‘decides’ to initiate or, at the least, prepares to initiate the act at a time before there is any reportable subjective awareness that such a decision has taken place. It is concluded that cerebral initiation even of a spontaneous voluntary act, of the kind studied here, can and usually does begin unconsciously …These considerations would appear to introduce certain constraints on the potential of the individual for exerting conscious initiation and control over his voluntary acts.” (Libet, 1983, p. 640-1).
Libet’s experiment has been debated, discussed, written about, and challenged in numerous ways since it was first published, but at least one thing is clear: the experiment replicates well. When others have later performed the same or a similar experiment, they have, by and large, obtained the same results. In 2011, neurosurgeon and neuroscientist Itzhak Fried repeated Libet’s experiment on patients who had electrodes implanted inside their brains and obtained essentially the same result.11 So it is perhaps not surprising that the Atlantic Magazine took aim at free will in 2016 and essentially declared that science had put the idea to rest.8 Through all of this, and through all of the years since its discovery in 1965, no one stopped to ask, what is the readiness potential in the first place? What does this apparent buildup mean? How should we interpret it? We started with an untested assumption — that the RP reflected the brain’s decision to initiate a spontaneous voluntary action — and almost immediately considered it a standard point of reference indicating the time at which the brain had decided to initiate a movement. Neuroscientists and experimental psychologists were quite ready to accept the idea that consciousness of exerting our will comes far too late in time to play any kind of role in movement initiation — the feeling of intending must itself originate in the same brain that is responsible for triggering movements, and our conscious percepts are sometimes misleading.
This interpretation of Libet’s classic experiment has held strong for four decades, but not without being challenged. Some of these challenges came from empirical data and others from theoretical or philosophical counter-arguments. For example, Lau, Rogers, and Passignham in 2006 argued for something like Heisenberg’s uncertainty principle in spontaneous voluntary action: measuring the process of movement preparation may actually affect the perceived onsets of the decision to move.12 Trevena and Miller in 2009 reported that the RP was just as evident before decisions not to move as it was before decisions to move.13 And Miller, Shepherdson, and Travena in 2010 reported14 that the act of clock monitoring actually changed the EEG data and could be partly responsible for the slowly-building brain activity preceding spontaneous voluntary movements.
More recently, a radically different interpretation of the RP, called the stochastic decision model (SDM), was proposed15 and has sparked significant debate. This interpretation seriously undermines the challenge that the RP has posed for free will.16,17 According to the SDM, movement is initiated when neural activity in premotor brain circuits reaches a certain threshold. This idea of a threshold-crossing triggering movement is supported by a wealth of evidence from research on perceptual decision-making: An animal is presented with a stimulus and has to quickly recognize what it is, or what category it belongs to. When neural activity in a specific part of the animal’s cerebral cortex reaches a certain critical threshold then the animal initiates a movement (pressing a button, pulling a lever, or making a reaching movement) in order to report its decision. It is not such a stretch to consider that maybe movements initiated for no particular reason (as in Libet’s task) are also triggered by a threshold-crossing — when the intensity of neural activity in motor areas reaches a certain level. So, the new interpretation goes, when the imperative to move is weak — as in Libet’s experiment (there is no sensory cue telling you when to move) — then the precise moment of that threshold crossing, and subsequent movement initiation, is largely determined by random sub-threshold fluctuations in neural activity. The brain is rife with such ongoing, seemingly random, background neural activity. Even when you are not doing anything in particular, neural activity in your brain (including in motor and pre-motor circuits) has its own ongoing ebb and flow, like waves on the ocean. If the onset of movement is more likely to coincide with crests in these ongoing waves, then when you look at the brain data aligned to the moment of movement onset (as researchers typically do in this field) you will indeed find a slow buildup preceding the moment of movement onset.
But does that buildup reflect the outcome of a pre-conscious decision to initiate movement? Not necessarily, no. It might reflect the slow-cresting nature of the aforementioned random ebb and flow of brain activity. As long as that ongoing random neural activity plays some role in determining the precise moment of movement onset, then those waves of activity will be apparent when we average the data together time-locked to the onset of movement. This is more than just an intriguing idea: In an actual EEG experiment,15 subjects were faster to react to surprise stimuli while doing Libet’s task when the stimulus was preceded by a slow buildup of neural activity over the motor cortex. This slow buildup could not reflect preparation for movement because the surprise stimuli were delivered at random times, unknown even to the experimenter. You cannot prepare for a movement that you do not know you are going to make!
What this means is that the slow buildup might reflect the neural antecedents of a decision to move rather than a consequence of it. The decision, according to the SDM, happens very close in time to the onset of movement, consistent with the moment at which subjects report first being conscious of their decision. Perhaps Libet should have taken his subjects at their word — when they reported feeling like they had felt the urge to move at t-minus-200-milliseconds, maybe that was because that was precisely when the decision was made. Instead, Libet assumed that his subjects must have been wrong about the time of their decision to move, because the RP was visible so much earlier. Thus, the SDM is a late-decision account of the RP, rather than an early-decision account like the classic interpretation. And if the brain’s decision to initiate movement coincides in time with the conscious decision to initiate movement, then this at least leaves open the possibility of conscious will. (It does not, however, account for the brain activity that precedes both the conscious decision and the threshold crossing that triggers the movement — that is another matter.)
In another more recent account along similar lines, Schmidt and colleagues18 proposed that the motor system is subject to an ongoing (presumably random) slow-varying ebb and flow called the slow cortical potential (SCP). The motor system in the brain is, according to Schmidt, more excitable in the “up” phase of the SCP, which means that, all other things being equal, movement is more likely to be initiated in the up phase compared to the down phase of the SCP. Thus, if you average your EEG data time locked to movement onset, then the resulting waveform will be dominated by the up phase of the SCP and this will appear in the average as a slow buildup. This model (called the slow cortical potential sampling hypothesis) has the same implications for conscious will as the SDM — it at least leaves open the possibility of conscious volition. As with any new set of findings and new theoretical framework, this one will require further research to confirm and pin down. It remains open to debate.
What about free will? What does neuroscience have to say about it in the end? Well, you can argue about free will on purely conceptual and philosophical grounds, and find various ways to refute its existence or allow for it, depending on how you define it. But could neuroscientific data be used to prove that free will does not exist? That would be difficult, though perhaps not impossible. Given the societal importance that attaches to the idea of free will, the bar should be set quite high. If you could predict human actions (based on brain activity or otherwise) well in advance of the initiation of the action, so that your predictions came far before any conscious decision (e.g. many seconds in advance) with near-perfect accuracy, then we would want to consider that evidence very carefully because it would indeed be a serious threat to conscious will. But no such evidence exists, and, in my opinion, it is unlikely that it ever will. Why? Well, consider that a single brain is arguably more complex than the entirety of the Earth’s atmosphere, and even the local weather can be challenging to predict, despite millennia of trying to learn how to predict it. So, it seems unlikely that simply applying machine learning techniques to brain data is going to lead to that level of prediction. Therefore, if you believe in free will, there is as present no hard neuroscientific data (that I know of) that could be used to seriously undermine that belief.
1.. T. O’Connor and C. Franklin, “Free Will”, in The Standford Encyclopedia of Philosophy, E.N. Zalta, Editor. 2021, Stanford University: Stanford.
2.. M. Desmurget et al., “Movement intention after parietal cortex stimulation in humans”, Science, 2009. 324: p. 811-813.
3.. P. Haggard, “The Neurocognitive Bases of Human Volition”, Annual Review of Psychology, 2019. 70(1): p. 9-28.
4.. B. Libet et al., “Time of Conscious Intention to Act in Relation to Onset of Cerebral-Activity (Readiness-Potential) – the Unconscious Initiation of a Freely Voluntary Act”, Brain, 1983. 106(SEP): p. 623-642.
5.. D.M. Wegner, “The Mind’s Best Trick: How We Experience Conscious Will”, Trends Cogn Sci, 2003. 7(2): p. 65-69.
6.. C.S. Soon et al., “Unconscious determinants of free decisions in the human brain”, Nat Neurosci, 2008. 11(5): p. 543-545.
7.. C.S. Soon et al., “Predicting free choices for abstract intentions”, Proc Natl Acad Sci USA, 2013. 110(15): p. 6217-22.
8.. S. Cave, “There’s No Such Thing as Free Will”, in The Atlantic, 2016.
9.. H.H. Kornhuber and L. Deecke, “Hirnpotentialänderungen bei Willkürbewegungen und passiven Bewegungen des Menschen: Bereitschaftspotential und reafferente Potentiale” (Changes in brain potentials with willful and passive movements in humans: the readiness potential and reafferent potentials), Pflügers Archiv, 1965. 284: p. 1-17.
10.. H.H. Kornhuber and L. Deecke, “Brain potential changes in voluntary and passive movements in humans: readiness potential and reafferent potentials”, Pflügers Archiv – European Journal of Physiology, 2016: p. 1-10.
11.. I. Fried, R. Mukamel, and G. Kreiman, “Internally Generated Preactivation of Single Neurons in Human Medial Frontal Cortex Predicts Volition”, Neuron, 2011. 69: p. 548-562.
12.. H.C. Lau, R.D. Rogers, and R.E. Passingham, “On Measuring the Perceived Onsets of Spontaneous Actions”, J. Neurosci., 2006. 26(27): p. 7265-7271.
13.. J. Trevena and J. Miller, “Brain preparation before a voluntary action: Evidence against unconscious movement initiation”, Consciousness and Cognition, 2009. 19(1): p. 447-456.
14.. J. Miller, P. Shepherdson, and J. Trevena, “Effects of Clock Monitoring on Electroencephalographic Activity”, Psychological Science, 2010.
15.. A. Schurger, J. Sitt, and S. Dehaene, “An accumulator model for spontaneous neural activity prior to self-initiated movement”, PNAS, 2012. 109(42): p. E2904-E2913.
16.. A. Schurger, M. Mylopoulos, and D. Rosenthal, “Neural Antecedents of Spontaneous Self-Initiated Movement: A New Perspective”. Trends Cog Sci, 2016. 20(2): p. 77-79.
17.. A. Schurger et al., “What Is the Readiness Potential?” Trends in Cognitive Sciences, 2021. 25(7): p. 558-570.
18.. S. Schmidt et al., “’Catching the waves’ – slow cortical potentials as moderator of voluntary action”, Neuroscience & Biobehavioral Reviews, 2016. 68: p. 639-650.