Libet 1983 showed brain activity preceded conscious intention by 350 milliseconds. The popular reading: “your brain decides before you do.” Four decades of follow-up neuroscience, including Schurger 2012 and 2021, no longer supports that interpretation. The empirical foundation is weaker than the pop story.
For roughly four decades, one diagram has done more rhetorical work in the free-will debate than any other piece of empirical evidence. The diagram shows a slowly rising scalp voltage --- the so-called readiness potential or Bereitschaftspotential --- that begins about 550 milliseconds before a subject flicks their wrist. The diagram also shows a second event, the subject’s reported moment of conscious intention to move, occurring only about 200 milliseconds before the wrist actually flicks. The temporal gap between brain-onset and awareness-onset is about 350 milliseconds, and that gap has been narrated, in book after book and TED talk after TED talk, as conclusive empirical evidence that conscious choice is an illusion --- the brain has already initiated the action before the conscious mind notices it has “decided.”
The paper that introduced this diagram was Benjamin Libet’s 1983 publication in Brain. The interpretation that “the brain decides before you do” was not exactly Libet’s own framing --- his actual position, which we will get to, was more cautious and considerably stranger than the popular summary --- but the popular summary has hardened into a cultural fact. Sam Harris’s 2012 book Free Will leans on it. Yuval Harari’s Homo Deus cites it. Countless pop-neuroscience explainers reproduce the diagram unchanged. Critics of moral responsibility, materialists arguing against dualism, and casual readers looking for a clean scientific reason to update toward determinism all reach for the same paper.
The problem is that the neuroscience field that produced the paper has substantially moved on, and the popular interpretation has not.
The most consequential update came from Aaron Schurger’s 2012 PNAS paper, which proposed an alternative model under which the readiness potential is not a “decision signal” at all but rather an artifact of how spontaneous neural fluctuations get aligned in time when researchers select trials based on action onset. A 2021 review by Schurger and colleagues in Trends in Cognitive Sciences walked through the accumulated evidence and concluded that the canonical Libet interpretation, on which the readiness potential is a measurable neural decision preceding conscious awareness, is no longer the field consensus. The Libet diagram remains influential in popular books. It is no longer the interpretive frame that the neuroscientists who study the readiness potential rely on.
This article is about that gap. It is not an argument that we do or do not have free will. It is an argument that one of the most cited empirical pillars of the anti-free-will position is much shakier, methodologically and theoretically, than the popular literature acknowledges. For anyone evaluating “neuroscience proves we are not in control” claims --- whether the context is consumer behavior, persuasion, marketing, ethics, criminal law, or self-help --- the evidentiary foundation is worth examining carefully rather than accepted on second-hand authority.
Libet 1983 --- The Original Methodology and the 350 Millisecond Finding
Benjamin Libet was a neurophysiologist at UC San Francisco who spent much of his career on the neural correlates of conscious experience. His 1983 paper, written with Curtis Gleason, Elwood Wright, and Dennis Pearl, attempted something genuinely novel: to put a stopwatch on the relative timing of an unconscious neural precursor and the subjective experience of “deciding” to move.
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).” Brain, 106(3), 623—642. DOI: 10.1093/brain/106.3.623.
The basic paradigm worked like this. A subject sat in front of a custom-built clock with a dot rotating around its face once every 2.56 seconds, fast enough that any given clock position could be specified to within tens of milliseconds. Scalp electrodes recorded EEG over the supplementary motor area to capture the readiness potential, a slow negative-going voltage shift that had been known since Kornhuber and Deecke’s 1965 work to precede voluntary movement by several hundred milliseconds. An electromyograph on the wrist recorded the exact moment the muscle activated.
The subject was instructed to spontaneously flex their wrist whenever they felt like it, with no fixed schedule. After the movement, they reported the position the clock dot had been in at the moment they “first became aware of the wish or urge to act.” Subjects were instructed to report the earliest such moment, not the moment they actually moved.
The headline numerical finding: across forty trials per subject and multiple subjects, the readiness potential began on average about 550 milliseconds before the wrist flexed, and the reported moment of conscious intention occurred about 200 milliseconds before the wrist flexed. The difference, about 350 milliseconds, was Libet’s striking result. The brain, on this reading, had begun preparing the movement roughly a third of a second before the subject reported any conscious awareness of intending to perform it.
Libet’s own interpretation was more nuanced than the popular summary suggests. He proposed that conscious will did not initiate movements but could potentially veto them in the interval between intention-awareness and action --- the so-called “free won’t” account. He was explicitly hesitant to draw deterministic conclusions and remained engaged with theological and philosophical interpretations of his work for the rest of his life. The paper itself is careful to note that the findings concern the timing of awareness relative to motor preparation, not the metaphysical question of whether conscious experience has causal efficacy.
That careful original framing did not survive the second-hand citation chain. By the time the result reached popular audiences, it had been compressed into a slogan: brain first, mind second, free will is an illusion.
How the Popular Interpretation Became a Cultural Fact
The transmission of Libet 1983 into popular consciousness happened in three waves.
The first wave was inside academic philosophy of mind in the late 1980s and 1990s. Daniel Dennett discussed Libet at length in Consciousness Explained (1991), generally skeptically. Patricia Churchland used the work to argue for neurophilosophical eliminativism. Owen Flanagan and others integrated it into broader naturalistic accounts of consciousness. At this stage the interpretation was still contested and the methodological caveats were still mostly in view.
The second wave was the popular-neuroscience publishing boom of the mid-2000s. Brain-imaging technology had become a cultural fascination, and a class of trade books emerged that explained neuroscience findings to lay readers. Libet’s diagram became one of the canonical pieces of evidence in books arguing that the “self” is a kind of post-hoc narrative spun by an unconscious brain. Sam Harris’s Free Will (2012) is the cleanest example: a short book in which Libet’s data is presented as conclusive, with little discussion of methodological alternatives, and the conclusion drawn is the strong deterministic one. Yuval Noah Harari’s Homo Deus (2016) used Libet to argue that the liberal-humanist conception of the autonomous individual is empirically obsolete. By this point the original framing was effectively gone.
The third wave is ongoing. Libet’s experiment now appears in undergraduate psychology textbooks, in introductory philosophy courses, in TED talks, in self-help books that argue you should be kinder to yourself because “your brain decided before you did,” and in the standard catalog of cocktail-party neuroscience facts. The 350-millisecond number has the same cultural status as “we use only ten percent of our brain” --- something most educated adults vaguely know is supposed to mean something profound, even if they cannot quite say what.
The result is a citation chain in which most people who confidently invoke Libet have not read the 1983 paper, have not read the subsequent methodological critiques, and have certainly not read the post-2010 reinterpretations. They are working from a textbook summary of a popular-book summary of a philosophical reframing of an experiment whose author’s own interpretation was distinct from all three.
Schurger 2012 --- The Stochastic Accumulator Alternative
The most important challenge to the standard Libet interpretation came in 2012. Aaron Schurger, then at the Brain and Spine Institute in Paris and now at Chapman University, and his colleagues Jacobo Sitt and Stanislas Dehaene published a paper in PNAS that proposed an entirely different generative model for the readiness potential.
Schurger, A., Sitt, J. D., & Dehaene, S. (2012). “An accumulator model for spontaneous neural activity prior to self-initiated movement.” Proceedings of the National Academy of Sciences, 109(42), E2904—E2913. DOI: 10.1073/pnas.1210467109.
The core insight is statistical and somewhat counterintuitive. Suppose neural activity in motor planning areas is constantly fluctuating around baseline due to spontaneous noise. Suppose further that when this fluctuation crosses some upper threshold, the subject experiences an urge to move and acts. If you then run the experiment many times and average the EEG signal time-locked to action onset, what shape will the average look like?
Schurger and colleagues showed mathematically and in simulation that the average will look exactly like the classical readiness potential --- a slow ramp beginning hundreds of milliseconds before action onset. Critically, this ramp is not present on individual trials in any form that resembles “the brain deciding.” It is an artifact of the time-locking and averaging procedure combined with the threshold-crossing dynamics of a stochastic accumulator. The pre-movement EEG looks like a smooth ramp in the average because, when you select trials by action onset, you are implicitly selecting fluctuations that happened to cross threshold around then, and the average path to threshold from below has the characteristic ramp shape.
The paper went on to test predictions of the accumulator model against experimental data and found that the model fit the data better than the classical “decision signal” interpretation. The implication is severe: the readiness potential might not be the neural correlate of a specific decision at all. It might be a statistical signature of how trial selection interacts with ongoing spontaneous activity.
This was not a fringe critique. The paper appeared in PNAS, was widely covered in the neuroscience press, and prompted a substantial follow-up literature attempting to test the accumulator account against alternative models. The original Libet interpretation, on which the readiness potential is a measurable preparatory signal that crosses some kind of decision threshold to produce action, became one model among several rather than the default reading of the data.
The accumulator account also dissolves the most rhetorically powerful version of the Libet-implies-no-free-will argument. If the readiness potential is not a “decision signal” but a noise-aligned artifact, then the 350-millisecond gap between RP-onset and reported intention does not show that the brain decided before the subject was aware of deciding. It shows that, when subjects spontaneously move at moments their internal noise reaches threshold, the EEG averaged across many such moments has a characteristic ramp. The deterministic punch line does not follow.
Schurger 2021 --- The Comprehensive Reinterpretation
Nine years after the PNAS paper, Schurger returned to the topic in a substantial review for Trends in Cognitive Sciences, co-authored with Pengbo Hu, Joanna Pak, and Adina Roskies (a philosopher who has worked extensively on the free-will implications of neuroscience).
Schurger, A., Hu, P. (Ben), Pak, J., & Roskies, A. L. (2021). “What is the readiness potential?” Trends in Cognitive Sciences, 25(7), 558—570. DOI: 10.1016/j.tics.2021.04.001.
The review is striking in its framing. The opening question --- “what is the readiness potential?” --- is not asked rhetorically. It is asked because, after more than fifty years of research, the field genuinely does not have a settled answer that commands consensus. The authors walk through the history of the Bereitschaftspotential from Kornhuber and Deecke’s original 1965 description, through the standard “motor preparation” interpretation, through the Libet-paradigm spontaneous-movement variant, and into the post-2012 alternatives.
Their central conclusions are several and worth listing.
First, the readiness potential as measured in the Libet paradigm (spontaneous, self-initiated movements with no external cue) is dynamically and possibly mechanistically distinct from the readiness potential measured for cued or planned movements. Lumping them together as a single phenomenon is a category error that has propagated through the textbook literature for decades.
Second, the standard interpretation on which the RP reflects a specific “decision” or “intention” to move is not well supported by current evidence. The accumulator account, several refinement variants, and a number of competing models all fit the existing data approximately as well, and the question is empirically open.
Third, the inference from Libet 1983 to claims about free will requires a specific interpretation of the RP that the field has not validated. Whether or not free will exists, Libet’s data is not the empirical bedrock the popular literature has treated it as. The review explicitly notes that much of the philosophical work that has used the RP as evidence about the nature of voluntary action has relied on an interpretation of the RP that is no longer well supported.
Fourth, methodological refinements in the post-2012 literature have repeatedly weakened rather than strengthened the original interpretation. Studies using better temporal resolution, alternative measures of motor preparation, single-trial analyses rather than averages, and explicit comparison of accumulator-style and classical models have generally favored accounts in which the RP is not the discrete decision signal Libet’s framing implied.
The 2021 review is, in effect, a senior researcher in the field telling the broader scientific and popular community that the textbook story is out of date. It has not had the cultural impact of the original 1983 paper. The relevant audiences --- philosophers writing about free will, popular-science authors writing about consciousness, undergraduates encountering the topic for the first time --- are mostly still working from the older summary.
Methodological Limits the Original Paper Carried From the Start
Even setting aside the post-2012 reinterpretation, the original Libet paradigm had methodological limitations that have been discussed in the literature since the 1980s. Several deserve attention.
The reported moment of conscious intention required subjects to introspectively note the position of a fast-rotating clock dot at the exact moment they first became aware of an urge to move. This is a difficult and arguably ill-defined introspective task. The reported time depends on attention allocation, the inherent latency of perceiving the clock position, the salience of the urge experience, and the subject’s interpretation of the instructions. Several follow-up studies have shown that the reported W (the “moment of will”) time can be shifted by experimental manipulations that, on the strong deterministic reading, ought not to affect it --- including feedback about the subject’s own reported times, instructions to attend more or less to the urge, and the inclusion of cued probes.
The action itself was trivial. Flexing a wrist on no particular schedule, with no consequence, is about as far from a meaningful decision as a voluntary action can get. Whether the timing relationship Libet documented for this paradigm generalizes to decisions that have consequences, that involve deliberation between alternatives, that are temporally extended, or that involve any of the things people usually mean by “free will” is an open empirical question. The popular interpretation has consistently treated the wrist-flex as a microcosm for human agency. That treatment is not obviously warranted.
The selection of trials was self-paced and post-hoc. Subjects flexed when they felt like it, and trials were identified by the action. As Schurger’s accumulator account makes vivid, this selection procedure interacts with whatever generates the EEG signal in ways that are not obvious from the averaged result. The original analysis did not consider whether the apparent pre-movement ramp could be an artifact of this interaction rather than a representation of decisional content.
The sample sizes were small by current standards. The original 1983 paper reported data from six subjects across forty trials each. The replication base for the basic finding is much larger now, but the original interpretation rested on data that, by 2026 methodological standards, would barely pass review for a single small study.
The introspective reliability of the moment-of-will report has never been validated against any independent measure. We have no way to verify that the reported W time corresponds to any specific neural or psychological event. The whole interpretive edifice rests on subjects accurately reporting the timing of an introspective experience whose precise definition was left implicit.
Banks and Pockett, in their chapter for the Blackwell Companion to Consciousness (2007), reviewed many of these limitations in detail and concluded that the philosophical conclusions drawn from Libet’s work had outrun what the methodology could support.
Banks, W. P., & Pockett, S. (2007). “Benjamin Libet’s work on the neuroscience of free will.” In M. Velmans & S. Schneider (Eds.), The Blackwell Companion to Consciousness (pp. 657—670). Blackwell Publishing. DOI: 10.1002/9780470751466.ch52.
Their judgment, written years before the Schurger reinterpretation, was that the Libet experiments showed something interesting about the timing of self-reports relative to motor preparation but did not licence the dramatic free-will conclusions popular accounts had drawn. The subsequent decade has, if anything, strengthened that assessment.
Why Pop Neuroscience Still Cites Libet
Given the state of the empirical literature, an obvious question is why the popular Libet narrative has proven so durable. Several factors are worth naming.
The diagram is rhetorically powerful. A simple two-line plot, with brain activity preceding awareness by 350 milliseconds, is exactly the kind of clean visual evidence that makes for compelling exposition. Schurger’s accumulator model, by contrast, requires a paragraph of statistical reasoning to explain and does not produce a single iconic image. The case against the standard interpretation is structurally harder to summarize, which is a real disadvantage in the attention economy.
The conclusion it appears to support is congenial to several active intellectual movements. New atheists, eliminative materialists, behavioral economists arguing for “nudge” paternalism, and various reductionist projects in the cognitive sciences all have reasons to want a clean empirical demonstration that conscious choice is not what it seems. Libet 1983 looked like exactly that demonstration. Letting go of it is costly for the broader argumentative structures that have come to depend on it.
The original paper is genuinely well-executed for its time and place. Libet was a careful researcher, the methodology was novel and clever, and the data is real. The problem is interpretive, not methodological in the “fraud” or “p-hacking” sense that has characterized many of the takedowns in this hub. There is no original paper to retract, no co-author recantation, no failed preregistered replication. The reinterpretation is more subtle: the data exists and means something, but probably not what the popular literature claims it means.
Citation chains have inertia. Once a finding is established in textbooks and trade books, it tends to propagate in those forms even as the primary literature moves on. The half-life of an outdated interpretation in popular science is on the order of decades, not years. Sam Harris’s Free Will is still in print and still cited, with no edition update reflecting the Schurger reinterpretation. The Homo Deus passage on Libet is unchanged in the 2020 paperback.
There is no easy replacement narrative. If you tell a popular audience “Libet’s experiment does not prove free will is an illusion, and also we do not have a clean alternative story about the neural basis of intention,” you have not really given them anything to take home. The popular Libet story is so durable in part because the more accurate scientific position is “this is genuinely an open question and the evidence is messier than the popular framing suggests” --- which is not a satisfying ending.
Strategist Implications for “Neuroscience Proves” Claims
This article is for strategists, not philosophers. The practical takeaway is not “you have free will after all” or “Libet was wrong about the timing.” Those are matters for further specialized inquiry. The takeaway is about how to evaluate the broader class of “neuroscience proves” claims that show up in commercial, persuasive, and applied contexts.
The Libet pattern recurs constantly. A specific empirical finding gets translated into a strong general claim about human nature, the strong general claim becomes the citation-anchor for a commercial or political argument, and by the time the original finding’s interpretation has been substantially revised by the field, the strong general claim has acquired sufficient cultural momentum that it continues to circulate independently of the empirical update. Marketing-adjacent examples include the misuse of mirror neuron research to justify storytelling claims, the popular invocation of left-brain versus right-brain personality theory, the persistence of the ten-percent-of-the-brain myth, and the recent uptake of polyvagal theory in trauma-informed marketing despite the original neuroanatomy being substantially contested. The pattern is sufficiently common that “neuroscience proves” is now most usefully read as a warning sign about an argument rather than as a sign of empirical strength.
The general protocol for evaluating such claims is the one this hub keeps repeating in different forms. Find the primary paper. Check whether the field’s current interpretation matches the popular interpretation. Check whether attempted replications and alternative models have updated the picture. Check whether the original methodology actually licences the general claim being made --- specifically, whether the experimental task is meaningfully connected to the real-world phenomenon being discussed, or whether the connection requires a substantial inferential leap.
In Libet’s specific case, the methodological connection between flexing a wrist on no particular schedule and “free will” in any sense that matters for everyday human agency is far less obvious than the popular interpretation pretends. The relationship between the readiness potential and the experience of intending to move is also far less obvious than the popular interpretation pretends. Both of these gaps existed in 1983 and have only become more visible over time.
For a strategist evaluating a vendor pitch, a research summary, an academic-flavored consulting deck, or a book chapter that cites Libet, the right response is something like: I have read this paper, I have read the Schurger 2012 alternative model, I have read the Schurger 2021 review, and the current state of the literature does not support the claim you are making. If the vendor or author cannot engage with that reply, they are working from a textbook version of a finding whose primary literature they have not actually consulted, and the rest of their evidence base should be evaluated with the same skepticism.
This is a calibration exercise. The Libet example is useful precisely because it is so culturally entrenched and because, on close inspection, the empirical foundation is much weaker than the cultural status suggests. Apply the same scrutiny to the next “neuroscience proves” claim that crosses your desk. Most of them will not survive it.
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).” Brain, 106(3), 623—642. DOI: 10.1093/brain/106.3.623.
- Schurger, A., Sitt, J. D., & Dehaene, S. (2012). “An accumulator model for spontaneous neural activity prior to self-initiated movement.” Proceedings of the National Academy of Sciences, 109(42), E2904—E2913. DOI: 10.1073/pnas.1210467109.
- Schurger, A., Hu, P. (Ben), Pak, J., & Roskies, A. L. (2021). “What is the readiness potential?” Trends in Cognitive Sciences, 25(7), 558—570. DOI: 10.1016/j.tics.2021.04.001.
- Banks, W. P., & Pockett, S. (2007). “Benjamin Libet’s work on the neuroscience of free will.” In M. Velmans & S. Schneider (Eds.), The Blackwell Companion to Consciousness (pp. 657—670). Blackwell Publishing. DOI: 10.1002/9780470751466.ch52.
- Kornhuber, H. H., & Deecke, L. (1965). “Hirnpotentialänderungen bei Willkürbewegungen und passiven Bewegungen des Menschen: Bereitschaftspotential und reafferente Potentiale.” Pflügers Archiv, 284, 1—17. DOI: 10.1007/BF00412364.
- Harris, S. (2012). Free Will. Free Press.
- Harari, Y. N. (2016). Homo Deus: A Brief History of Tomorrow. Harper.
- Dennett, D. C. (1991). Consciousness Explained. Little, Brown and Company.
Related Reading
- Wegner’s Illusion of Conscious Will --- the companion popular-neuroscience argument that the experience of willing is post-hoc, and the methodological grounds for revisiting it.
- Mirror Neurons --- another neuroscience finding whose popular interpretation has substantially outrun what the primary literature supports.
- The Ten Percent of the Brain Myth --- the most-cited piece of pop neuroscience that has never had any empirical basis at all, and what its persistence tells us about how brain claims propagate.
- Left Brain / Right Brain Personality Theory --- the lateralization-and-personality narrative that survives in popular culture despite four decades of contrary neuroscience.
- Polyvagal Theory --- a more recent example of contested neuroscience that has nevertheless been widely adopted in adjacent fields including trauma-informed marketing.
FAQ
Did Libet himself believe his work disproved free will?
No. Libet’s own interpretation was substantially more cautious than the popular one. He argued for a “free won’t” account on which conscious will, although it did not initiate movements, could veto them in the roughly 150-millisecond window between the reported moment of intention and the actual movement. He remained engaged with theological and philosophical interpretations of consciousness throughout his career and was on record as resisting strong deterministic readings of his own data. The popular interpretation attributed to Libet is largely a second-hand simplification.
Is Schurger’s accumulator model now the consensus view?
It is the most influential alternative to the classical interpretation, but the field has not converged on a single replacement model. The 2021 Trends in Cognitive Sciences review treats the question of what the readiness potential actually represents as genuinely open, with the accumulator account, several refinement variants, and other competing models all having some empirical support. The honest summary is that the classical “the RP is a decision signal preceding awareness” interpretation is no longer the field default, but a clean replacement has not emerged.
Have the Libet experiments themselves been replicated?
The basic timing pattern --- readiness potential beginning hundreds of milliseconds before action, reported W time occurring some shorter time before action --- has been replicated many times across labs and decades. What is contested is not the existence of the timing relationship but the interpretation of what it means about the underlying neural process. A finding can be empirically robust in the sense of replicating and theoretically unstable in the sense of not having a settled mechanistic interpretation. Libet’s data is in that category.
Does any of this change the philosophical question of whether free will exists?
Not directly. The philosophical question is essentially independent of whether the Libet interpretation holds up. Free will is a metaphysical and conceptual issue that has been debated for thousands of years on grounds that do not require neuroscience at all. What does change is the role that Libet 1983 plays as an empirical input to that debate. If you were previously persuaded by the popular interpretation that Libet had provided clean empirical evidence against free will, the more recent reinterpretations should reduce your confidence in that specific empirical pillar. They do not, by themselves, resolve the broader question in either direction.
What about the 2008 Soon et al. study using fMRI that pushed the timing back to several seconds?
Soon, Brass, Heinze, and Haynes published a study in Nature Neuroscience in 2008 reporting that fMRI activity in prefrontal and parietal cortex could predict the outcome of a left-versus-right button-press decision up to 10 seconds before the subject reported being aware of deciding. The study was widely cited as a dramatic extension of Libet. The reported predictive accuracy, however, was only modestly above chance (around 60% for a 50% baseline), the predictions involved a very specific and constrained binary choice rather than meaningful agency, and subsequent work has raised some of the same accumulator-and-noise concerns that apply to Libet 1983. The Soon study is real and interesting but does not provide cleaner evidence for the “brain decides before you do” narrative than Libet did. It mostly reproduces, with fancier technology, the same set of interpretive problems.
What is the practical takeaway for someone evaluating behavioral-science or neuroscience claims in a commercial context?
Treat “neuroscience proves” or “research shows the brain” framings as warning signs, not as evidence. Ask which specific paper is being cited, whether the primary author would endorse the popular interpretation as stated, whether the field’s current consensus has moved beyond the original interpretation, and whether the experimental task actually licences generalization to the commercial or applied context being discussed. In a substantial fraction of cases the answer to one or more of those questions will be “no,” and the citation should carry correspondingly less weight than the vendor or author wishes it to.