The Half-Second Before the Will: Libet's Experiment and the Neuroscience of Free Will

The Hook: Who Is Raising the Hand Here?

Sit down in front of a clock whose dot sweeps around the dial once every two and a half seconds or so. Your task is disarmingly simple: at some point, whenever the mood strikes you, spontaneously flex your wrist. No plan, no prior intention, no mental countdown—just a sudden urge that you give in to. And when you perform the movement, please note carefully the position of the dot at the instant you first feel the conscious wish, the urge, the conscious intention to move.

You are convinced that in this moment you are in control. The wish appears, you decide, the hand moves—in that order. That is how it feels. That is how it feels to every human being, every day, in every free action. Consciousness sits at the wheel and gives the command.

But now a mesh of electrodes rests on your head, and an electroencephalograph records the electrical activity of your brain. When the physiologist Benjamin Libet ran this experiment in San Francisco in 1983 and analyzed the data, he found something that has haunted philosophers, neuroscientists, and legal scholars ever since: your brain begins to prepare the movement before you feel the conscious will to make it. Not milliseconds later, but a good while earlier. The measurable preparation gets under way, and only afterward does consciousness announce itself with its supposed starting signal. The wheel, it seemed, turns on its own, and consciousness only reaches for it after the fact.

For someone like Sven—who works professionally with deterministic machines, with state machines, with systems whose output is fully fixed by their inputs and their internal state—this question has a special pull. Is the human brain, at bottom, just such a machine? Is the feeling of deciding freely a user interface behind which causal processes run that no one chose? This article takes you the whole way: from the ancient philosophical question through Libet's famous experiment and its even more famous misreading, to the quiet, technical rebuttal that a neuroscientist named Aaron Schurger delivered two decades later—and finally to the question of what remains, from all of this, for moral responsibility.

Part 1: The Oldest Question—and Its Three Answers

Before we attach any electrodes, we must survey the battlefield. The question of free will is not an invention of brain science; it is at least as old as written philosophy. Its modern core can be framed as a tension between two claims.

The first claim is determinism: every event, including every thought and every action, is the necessary consequence of prior events according to the laws of nature. If you knew the complete state of the universe at one moment and knew the laws, the future would be fixed. The brain is part of this universe; its neurons fire according to physical and chemical laws. So where, the determinist asks, in this seamless causal chain is there room for a "free" will that could bring about something other than what it in fact brings about?

The second claim is our lived self-understanding: we experience ourselves as the authors of our actions. We deliberate, we weigh, we choose, and we feel that we could have done otherwise. On this experience rests our entire practice of praise and blame, guilt and merit, law and punishment.

From the relationship between these two claims arise three classical positions that structure the debate to this day.

Hard determinism takes determinism seriously and draws the uncomfortable conclusion: if everything is causally fixed, then there is no free will in the strong sense. The feeling of freedom is an illusion. This position is incompatibilist—it holds free will and determinism to be irreconcilable—and it sides with determinism.

Libertarianism (in the metaphysical, not the political sense) is likewise incompatibilist, but sides with free will. It claims that we make genuine, non-predetermined decisions, and that determinism must therefore be false—at least in the realm of human action. Some libertarians locate the gap in quantum randomness; others posit a distinct causality of the acting self (agent causation).

Compatibilism, finally, often called soft determinism, denies that the question is even properly posed. Freedom and determinism, it says, are not opponents at all. To act freely does not mean to be freed from all causation—that would be mere chance and would have nothing to do with freedom. To act freely means to act in accordance with one's own desires, beliefs, and deliberations, without external coercion. Whether those desires in turn have causes is irrelevant. Compatibilism is, as the Stanford Encyclopedia of Philosophy soberly notes, by far the most widespread position among professional philosophers.

A prominent compatibilist, the philosopher Daniel Dennett, distilled this stance into a memorable phrase: it is about the "varieties of free will worth wanting." The longing for a freedom that detaches us from all causation, from our own history and our own desires, is a phantasm. The freedom we really have and need—the ability to align our actions with reasons, to weigh alternatives, to steer our behavior through reflection—is perfectly compatible with determinism.

This map matters, because the quarrel over Libet's experiment almost always suffers from the fact that the participants are talking about different free wills. Whoever means the libertarian, uncaused will sees in Libet its gravestone. Whoever means the compatibilist will shrugs. Let us hold on to this distinction—we will need it at the end.

Part 2: Libet's Experiment (1983)

The Setup

Benjamin Libet, a physiologist at the University of California in San Francisco, wanted to do something that had until then seemed barely possible: measure the subjective time of a conscious act of will and compare it with objective brain activity. His tool was an already known signal, the readiness potential (originally described in German as Bereitschaftspotential by Kornhuber and Deecke in 1965). It is a slow, gradually rising negative electrical voltage over the brain's motor areas that precedes a self-initiated movement. You find it by recording many movements and aligning the EEG traces backward from the moment of movement, then averaging them.

Libet's stroke of genius was the second measurement: that of the conscious will. The subjects sat before a special cathode-ray clock on which a spot of light circled. They were to flex the wrist spontaneously and, in doing so, note the position of the spot at the moment they first felt the conscious wish, urge, or intention to move. Libet called this time "W" (for will). At the same time, an electromyogram over the muscle measured the actual onset of movement ("M" for movement), and the EEG supplied the onset of the readiness potential.

Libet thus had three time markers on the same axis: the onset of neural preparation, the moment of conscious will, and the moment of muscle action.

The Result

The order was not the one our sense of self expects. For spontaneous, unplanned movements, the readiness potential began roughly 550 milliseconds before the actual muscle action. The conscious will "W," by contrast, was reported by the subjects only about 200 milliseconds before the movement. Converting this, the onset of measurable brain activity lay about 350 milliseconds before the appearance of the conscious intention.

In Libet's own words, published in 1983 in the journal Brain: the cerebral process of a freely voluntary act begins unconsciously, well before the conscious intention even arises. The "unconscious initiation of a freely voluntary act"—as the title put it—seemed to demote consciousness to a spectator who believes himself the author of a text that another agency has, in truth, already begun to write.

It is hard to overstate the force of these numbers. If the brain sets the movement in motion a full third of a second before the conscious "Now!", then the conscious will is apparently not the cause of the action but arrives too late to the party. It reports on a decision that has already been made. That is exactly how the experiment was read for decades: as empirical proof that free will is an illusion.

Libet's Own Rescue: "Free Won't"

Remarkably, Libet himself did not draw this radical conclusion. He pointed to a small but decisive window of time. Between the appearance of the conscious will (about 200 ms before movement) and the actual muscle command (the final roughly 50 ms in which the motor pathways fire irrevocably) there remains a gap of about 100 to 150 milliseconds. In this window, Libet argued, consciousness can still stop the already-initiated movement. He called this the veto function—later often mocked as "free won't" rather than "free will."

On Libet's picture, the brain initiates actions unconsciously, but consciousness retains a right of control: the ability to suppress the impulse at the last second. Free will is thus less an initiator than a gatekeeper. This reading is elegant, but it has a problem: the veto too is a neural process, and one can ask whether it is not itself preceded by an unconscious "veto readiness potential." The rescue merely shifts the question; it does not answer it.

Part 3: The Escalation—Soon et al. (2008)

A quarter of a century after Libet, neuroscience seemed to tighten the screws even further. A team led by John-Dylan Haynes at the Bernstein Center in Berlin published a study in Nature Neuroscience in 2008 with the programmatic title "Unconscious determinants of free decisions in the human brain."

Instead of an EEG, Chun Siong Soon, Marcel Brass, Hans-Jochen Heinze, and Haynes used functional magnetic resonance imaging (fMRI), which spatially resolves blood flow and thereby, indirectly, activity throughout the brain. And instead of a mere movement, the task involved a genuine, if trivial, choice: the subjects were to decide freely whether to press a button with the left or the right hand, and to note the moment of the decision.

The result was even more astonishing than Libet's. Using pattern-analysis methods (multivariate pattern classification), the outcome of the decision—left or right—could be predicted from the spatial activity pattern of certain brain regions, and indeed up to seven seconds before the person reported having consciously decided. Taking into account the sluggishness of the measured blood-flow signal (the so-called hemodynamic response), the predictive neural information preceded the conscious decision by up to ten seconds.

Two regions provided this early warning: the frontopolar cortex (Brodmann area 10) far forward in the frontal lobe, and an area in the parietal lobe stretching from the precuneus into the posterior cingulate cortex. An often-overlooked detail matters here: the prediction lay not in a general increase in activity but in the fine spatial pattern—as if a quiet bias were hidden in the distribution of activity that statistically influenced the later choice.

Anyone reading the headlines might have gotten the impression that the matter was settled: the brain makes the choice, and consciousness learns of it seconds later like a press spokesperson announcing a decision long since made. But here a closer look pays off. The prediction accuracy was not 100 percent but only a few percentage points above the chance level of 50 percent—just enough to be statistically reliable. The brain thus revealed not an already-made decision but, at best, a faint tendency, a thumb on the scale. A bias is not a determination. We will see in a moment why this difference changes everything.

Part 4: The Quiet Rebuttal—Schurger's Accumulator

A Signal Misunderstood

The decisive turn came not from a philosopher but from a neuroscientist who asked a seemingly technical question: what is the readiness potential, really? Aaron Schurger, then at Stanislas Dehaene's research center in Paris, published a model in PNAS in 2012 together with Jacobo Sitt and Dehaene that turned the interpretation of the entire Libet paradigm on its head.

The starting point is an observation about statistics. Recall how the readiness potential becomes visible at all: you record many movements, align each EEG trace to the moment of movement, and average them. This procedure—backward alignment and averaging—has an insidious property: it produces an apparently gradually rising curve even when the underlying signal is in truth only fluctuating randomly.

The Model

Schurger transferred a tool well known from decision research to spontaneous movement: the leaky stochastic accumulator. Imagine an inner level upon which random neural noise constantly acts—sometimes a bit upward, sometimes downward (technically: a drift–diffusion process with temporally correlated noise). The level is called "leaky" because, without inflow, it slowly sinks back down. As soon as this fluctuating level crosses a certain threshold, the movement is triggered.

Now comes the crux. In Libet's task there is no external trigger, no reason to move now rather than in two seconds. The only instruction is: move at some point, spontaneously. In this situation, Schurger argued, the brain does not wait for an inner decision—it simply lets the random noise run and takes a spontaneous upward fluctuation as the occasion to cross the threshold and get going. The movement occurs when the noise happens to reach a peak.

And what happens when you now average backward from the movement? You inevitably select precisely those moments in which the level has just shot upward—after all, that is why the movement occurred. Average many such random upward excursions, aligned to their peaks, and on average a smoothly rising curve emerges that looks like purposeful preparation—but that is nothing other than the statistical echo of cherry-picked random spikes. The readiness potential would thus be no signal that announces an already-made decision, but an artifact of the analysis method. Schurger's model even predicted the exact shape of the curve correctly.

The Consequence

With that, the dramatic interpretation collapses. If the rise of the readiness potential is only random noise, then its "onset" marks no point in time at which the brain secretly resolved to move. The onset does not really exist as a sharp point in time—it is an averaging effect. And then the temporal comparison between "onset of the readiness potential" and "moment of conscious will," which carries Libet's entire argument, is robbed of its foundation.

In a widely noted review of 2021 in Trends in Cognitive Sciences with the plain title "What Is the Readiness Potential?", Schurger summarized this reassessment. His message is precise and, for an engineer, pleasingly sober: the new models force a reassessment of the ontological standing of the readiness potential—whether it is even a real, causally efficacious signal—of the meaningfulness of the temporal comparisons, and they strongly relativize the relevance of the readiness potential for debates about free will. At the same time Schurger fairly emphasizes: the Libet experiment does not disprove free will, but it does not prove it either. It is simply far less telling than was claimed for decades.

Fittingly, later meta-analyses of Libet-style experiments have found that the reported moments of conscious will are methodologically fragile and heavily dependent on measurement conditions—another reason not to draw metaphysical conclusions from a single number like "350 milliseconds."

Part 5: The Philosophers Strike Back

Even if one set Schurger's reinterpretation aside and took the Libet numbers at face value, philosophers have good reasons why the experiment could not decide the question of free will anyway. Three objections are especially strong.

First: the wrist-flexing question is no model of real decisions. The philosopher Alfred Mele, one of the sharpest critics of the hasty Libet reading, points to a simple fact: Libet's task explicitly demanded an arbitrary, groundless, spontaneous movement without any deliberation. But precisely such fidget-decisions are the least interesting case. When I decide whether to accept a job, take out a loan, or enter a relationship, I do not sit before a clock waiting for a twitch impulse. I weigh reasons, over hours or weeks. That a groundless finger movement is co-steered by unconscious fluctuations says almost nothing about deliberate, reason-guided decisions. One studied the most boring conceivable sort of "decision" and generalized the results to all others.

Second: what is "W" anyway? Libet asked his subjects to report the moment of the first conscious urge. But an urge is not the same as a decision. It is entirely plausible that the feeling of an urge enters consciousness only late and approximately, while the actual resolve itself carries no sharp timestamp at all. The assumption that there is a precise instant "W" at which free will is pressed like a start button is itself already a questionable theory of the will—precisely that naive, pointlike conception that the compatibilist rejects anyway.

Third: the unconscious brain is not the enemy of the self. Here lies perhaps the deepest error of reasoning. The dramatic interpretation tacitly presupposes that "I" am identical with my consciousness, and that everything unconscious is a foreign manipulator who overrules me. But why should that hold? The unconscious neural processes that prepare my movement are part of me. It is not that "the brain" decides against "me"—the brain is the organ with which I decide. If my preferences, my history, my values are embodied in neural processes, then an action that arises from these processes is precisely my action, whether it becomes conscious a second earlier or not. Dennett's compatibilism lives on exactly this insight: freedom does not require consciousness to monitor every causal precursor, but that the action stem from the right inner process.

Herein also lies the bridge to a related idea: the brain as a prediction machine. If perception, as the model of predictive processing suggests, is largely a brain-generated, continually corrected model of the world, then the after-the-fact feeling of having willed "now" may itself be such a construction—a narrative the brain lays over its own process in retrospect, in order to experience itself as a coherent author. This does not devalue the narrative; it only shows that the sharp "moment of will" is not raw data but a product of self-modeling.

Part 6: What Is Really at Stake

If neuroscience thus neither disproves nor proves free will—why all the fuss? Because something practical hangs on the question that concerns us all: moral responsibility.

The philosopher Derk Pereboom holds a position he calls hard incompatibilism and which is systematically developed in his book Living Without Free Will (2001). His argument is more radical than classical hard determinism. Pereboom claims that we do not possess the kind of free will required for moral responsibility—and this holds independently of whether determinism is true. For even the randomness that quantum mechanics brings into play does not help: an action caused by genuine chance is no more "my" responsible deed than one fixed by strict causation. In both cases the deep, ultimate authorship that we actually presuppose in the concept of desert is missing. For Pereboom, this leads not to nihilism but to a milder stance: we should scale back the reactions grounded in "desert"—resentment, moral indignation, retributive punishment—in favor of prevention, restitution, and treating causes.

Against this stands the compatibilist tradition, which holds on to responsibility without needing libertarian metaphysics. The philosopher Harry Frankfurt, with his famous thought experiments, the Frankfurt cases, even shook the long-taken-for-granted principle of alternative possibilities—the assumption that someone is responsible only if he could also have acted otherwise. Frankfurt's case goes like this: Jones wants, of his own accord, to do X. In the background lurks a manipulator who would intervene and force Jones to do X should Jones begin to waver. But Jones does not waver; he does X entirely on his own. The manipulator never has to intervene. Jones could not have done otherwise (the manipulator would have prevented it)—and yet Jones is obviously fully responsible, because the action really did come from him. The lesson: responsibility does not hang on whether one could have acted otherwise, but on whether the action is one's own in the right way. And this condition can be met even in a determined world.

This quarrel is not salon philosophy. It has tangible consequences. Studies in social psychology (Vohs and Schooler and follow-up work) have suggested that people whose belief in free will is experimentally weakened subsequently behave somewhat less honestly or less helpfully—though the reproducibility of these effects is contested and should be treated with caution. I am of the opinion that the cautious reading is the right one here: the public discourse about free will is itself a causal factor in the behavior of societies, and precisely for that reason one should not derive bold headlines about the "death of free will" from uncertain neuroscientific findings. The question also feeds into law: criminal law presupposes capacity for guilt, and defense attorneys increasingly bring neuroscientific findings to bear—mostly with limited success, because a readiness potential says nothing about whether a defendant could, in the relevant sense, have acted otherwise.

A Framework: Who Believes What?

For orientation, an overview of the positions and of what Libet means for each.

The table makes something important visible: only for one of the four positions—metaphysical libertarianism—is Libet's experiment even a serious threat. For the most widespread position, compatibilism, it is simply beside the point. A large part of the decades-long excitement arose because popular accounts tacitly assumed the libertarian notion of will—that uncaused ghost in the control room—which hardly any serious philosopher of the twenty-first century still defends.

The Central Takeaway

The story of Libet's experiment is, at its core, not a story about free will at all. It is a story about measurement and interpretation—and about how easily a statistical artifact rises to a metaphysical sensation when it serves the right expectation.

The readiness potential looked like the brain's secret command, issued before consciousness learned of it. In truth it was, as Schurger's accumulator model shows, in all likelihood the averaged echo of random neural fluctuations, fished out by precisely the procedure with which it was made visible. That is a lesson reaching far beyond philosophy, one familiar to anyone who works with data: if you align your events to their trigger and then average backward, you almost inevitably produce a curve that looks purposeful—even from pure noise. The same trap lurks in the analysis of log data, in A/B tests, in every post-hoc analysis that searches for "precursors" of an event by looking only at the cases in which the event occurred. Survivorship bias and selective alignment are twins.

The practical prompt to action is therefore twofold. First, in thinking: be skeptical when a single experiment "proves" an ancient philosophical question—first ask whether what is measured is really what is meant, and whether the measurement method itself is not producing the result. Second, in self-understanding: the fact that unconscious processes precede your decisions does not disenfranchise you. These processes are not your opponent; they are the material of which your willing consists. You are free not when no neuron fires before you think "yes," but when your actions arise from your reasons, values, and deliberations—without external coercion. That is the only freedom worth wanting, and no experiment has ever touched it.

A Closing Question for Reflection

Suppose a future procedure could actually predict your binary decisions with 80 or 90 percent accuracy a few seconds in advance—the way a good language model predicts the next word, without your therefore writing "unfreely." Would this predictability really destroy your feeling of deciding freely? Or would you recognize that predictability and freedom have never contradicted each other—that the very fact that your decisions are predictable from you, and not from pure chance, is the best thing one can say about a free will?

Cross-References in the Vault

This inquiry touches several threads spun elsewhere in the vault:

• The Child in the Road: Moral Luck and the Limits of Responsibility – the flip side of the responsibility question: if chance (moral luck) already runs through our judgments, how much control does responsibility presuppose at all?
• The Predictive Brain: Predictive Processing and the Illusion of Perception – the brain as constructor of its own model, including the model of its own willing.
• What Mary Did Not Know: The Knowledge Argument and the Riddle of Consciousness – the question of whether consciousness is something that goes beyond the physical facts.
• The Chinese Room: Searle and the Question of Whether Machines Can Understand – behavior versus inner state, there in the machine, here in the human.
• Meant or Merely Foreseen: The Doctrine of Double Effect and the Power of Intention – the moral weight of intention, which makes Libet's "W" so hard to pin down.
• The Two Boxes: Newcomb's Paradox and the Battle Over Rational Choice – the predictability of a decision and its compatibility with rationality and choice.

Sources

• Libet, B., Gleason, C. A., Wright, E. W., Pearl, D. K. (1983): Time of conscious intention to act in relation to onset of cerebral activity (readiness-potential). The unconscious initiation of a freely voluntary act. Brain 106(3), 623–642. https://pubmed.ncbi.nlm.nih.gov/6640273/
• Soon, C. S., Brass, M., Heinze, H.-J., Haynes, J.-D. (2008): Unconscious determinants of free decisions in the human brain. Nature Neuroscience 11, 543–545. https://www.nature.com/articles/nn.2112
• Schurger, A., Sitt, J. D., Dehaene, S. (2012): An accumulator model for spontaneous neural activity prior to self-initiated movement. PNAS 109(42), E2904–E2913. https://www.pnas.org/doi/10.1073/pnas.1210467109
• Schurger, A., Hu, P., Pak, J., Roskies, A. L. (2021): What Is the Readiness Potential? Trends in Cognitive Sciences 25(7), 558–570. https://pubmed.ncbi.nlm.nih.gov/33931306/
• McKenna, M., Coates, D. J.: Compatibilism. Stanford Encyclopedia of Philosophy. https://plato.stanford.edu/entries/compatibilism/
• Pereboom, D. (2001): Living Without Free Will. Cambridge University Press. Review: https://ndpr.nd.edu/reviews/living-without-free-will/
• Meta-analysis of Libet-style experiments (2021), Neuroscience & Biobehavioral Reviews. https://www.sciencedirect.com/science/article/abs/pii/S0149763421002578

← All articles