Free Will

In 1983, physiologist Benjamin Libet attached electrodes to volunteers’ scalps and asked them to flick their wrist whenever they felt like it, noting the clock position when they first felt the urge to move. The electroencephalograph recorded a negative voltage shift — the “readiness potential” (RP) — building in the brain 300–500 milliseconds before subjects reported consciously deciding to move.

The unconscious brain had apparently started the decision before the conscious mind knew about it.

This finding ignited a decades-long debate that has not resolved and has only grown stranger with time: Are we authors of our choices, or do we narrate decisions that were already made?

The Libet Experiment and Its Children

Libet (1983): EEG readiness potential begins 300–500ms before subjects report “W” — the moment of conscious decision to move. This implies unconscious neural preparation precedes conscious awareness of deciding.

Libet’s own interpretation: The conscious will cannot initiate an action — but it can veto it. The RP starts unconsciously, but in the final 200ms before movement, a window remains for conscious cancellation. “Free won’t” rather than free will: not a freedom to initiate, but a freedom to abort.

Soon & Haynes (2008, Nature Neuroscience): Using fMRI instead of EEG, decoded left vs. right button press from patterns in the supplementary motor area and DLPFC (dorsolateral prefrontal cortex) up to 10 seconds before subjects consciously decided, with 60–70% accuracy. This pushed the “unconscious head start” from milliseconds to seconds.

Haynes et al. (2013, PNAS): Abstract intention to perform arithmetic addition vs. subtraction was decodable from medial prefrontal and parietal cortex 4 seconds before participants reported their conscious choice. This moved the experiment from simple motor actions to high-level cognitive decisions.

Haynes’s framing: “How can I call a will ‘mine’ if I don’t even know when it occurred and what it has decided to do?”

Key Facts

Readiness potential precedes reported conscious decision by 300–500ms (Libet 1983) established
fMRI can predict motor decisions up to 10 seconds before awareness (Soon & Haynes 2008) established
Abstract decisions predictable 4 seconds before conscious awareness from prefrontal cortex (Haynes 2013) established
RP-like signals are observed even in trials where participants opted not to move (Banks & Isham) emerging
2025 study: early RP does not differ based on whether intent to act or not act follows — challenges its meaning as “decision signal” emerging
Dominik et al. (2024): comprehensive methodological critique questioning Libet’s timing methodology established
Brain activity patterns predicting choices may reflect expectation/preparation, not decision itself emerging

The Critical Rebuttals

The Libet experiment has been challenged on multiple fronts since 2008:

The timing problem: Subjects estimate “W” (when they felt the decision) by watching a clock. This is a post-hoc report — memory of the exact moment. Banks & Isham (2009) showed W estimates can be manipulated by post-movement feedback, suggesting subjects partly reconstruct the decision moment rather than report it accurately.

The RP is not a decision signal: The readiness potential builds continuously whenever a person is about to make a voluntary movement — even if they ultimately decide NOT to move. It reflects a state of motor readiness or attentional preparation, not the specific decision to act. RP-like signals have been found in trials where subjects chose to withhold movement.

The 2025 challenge: A study published in 2025 found that the early component of the RP does not differ between trials where participants subsequently acted and trials where they did not. This undermines the interpretation that RP = unconscious decision-making. The RP may be brain noise misidentified as signal.

Dominik et al. (2024): A comprehensive intellectual history and methodological critique, concluding that Libet’s evidence for temporal precedence (unconscious over conscious) is far weaker than the popular narrative suggests.

The Philosophical Landscape

Compatibilism — The Dominant Academic Position

Compatibilists (Dennett, Frankfurt, Churchland) argue the debate hinges on a misleading definition of “free will.” The question isn’t whether your decisions are uncaused — nothing is uncaused. The question is whether they’re caused by you: your desires, values, reasoning, character.

If you act according to your own mental states — even if those states were shaped by genetics, upbringing, and gut bacteria — you are exercising the only kind of freedom that matters. This is sufficient for moral responsibility, praise, blame, and legal accountability.

Patricia Churchland: “Free will is self-controlled behavior arising from one’s own neural states. That it is physical doesn’t make it not yours.”

Hard Incompatibilism

Incompatibilists (Sapolsky, Wegner) argue that if all behavior is ultimately traceable to prior physical causes, then no one truly “deserves” credit or blame for anything. The appropriate response to wrongdoing is rehabilitation and prevention, not punishment.

Robert Sapolsky’s Determined (2023): “We are biology all the way down. The person who commits violence couldn’t have done otherwise given their biology, neurology, hormones, childhood, and the millisecond before the act. Stop blaming people.”

Libertarianism (not political)

Some philosophers and physicists propose that quantum indeterminism — the genuine randomness in quantum events — grounds a form of agency not reducible to physical determinism. The Frontiers in Neuroscience (2025) paper by Meijer & Korf proposes wave function collapse in specialized “prime neurons” as the physical substrate of free choice.

Problem: Randomness is not freedom. A choice determined by a random quantum event isn’t more free — it’s just undetermined, which is different from being self-determined.

Predictive Coding: The Brain That Predicts Its Own Decisions

Predictive coding neuroscience (Friston, Clark, Seth) offers a reframing. The brain is not a passive receiver of sensory information — it is a prediction machine that constantly generates models of future states and only updates on surprising errors.

In this framework, “deciding to move” is the brain’s prediction of its own next state. The conscious experience of willing comes after the predictive model is already generating the motor command. Consciousness may be riding the wave, not steering it.

Anil Seth (Being You, 2021): The sense of authorship over actions may be a controlled hallucination — a model the brain builds of itself as agent, which is useful for learning and social coordination but doesn’t accurately represent the causal order.

This is not nihilistic: the model of yourself as agent is causally real in a functional sense. Your beliefs about your freedom influence your behavior. Experiments show that priming people to believe they have no free will makes them more likely to cheat — the belief itself is part of the causal chain.

Legal and Ethical Implications

The stakes are high. Criminal justice in most societies rests on the assumption that perpetrators could have done otherwise. If they couldn’t — if they were determined by brain states, prior experience, and genetics — the moral basis for punishment collapses.

Current legal relevance:

Neuroscience evidence is admitted in some US courts to reduce sentence severity (the brain “made” them do it)
Addiction is increasingly treated as a disease of compulsion, not moral failure
The insanity defense recognizes that certain neural conditions remove agency

Sapolsky’s conclusion: The only rational responses to wrongdoing are: incapacitation (when someone is dangerous), rehabilitation (change the causal factors), and prevention (change societal conditions). Retribution — punishment for its own sake — has no justification if there is no free will.

The practical counter-argument: Even if free will is an illusion, behaving as if people have it may produce better outcomes than behaving as if they don’t. The belief in agency may be a necessary functional fiction for social cohesion.

The Gut-Brain Complication

concept-gut-brain-axis research adds a disturbing layer. Gut bacteria measurably influence:

Dopamine precursor availability (60% of dopamine precursor tryptophan processed in the gut)
Fairness judgments (gut-depleted individuals make harsher ultimatum game decisions)
Serotonin levels (90% made in the gut)
Anxiety and approach behavior (through vagal nerve signaling)

If gut bacteria influence the neural states that produce decisions, and those bacteria are partly determined by what you ate, who you were exposed to, and random colonization events at birth — then “your” decisions are partly authored by microorganisms you’ve never met and can’t consciously control.

The question “whose free will?” becomes genuinely unanswerable.

Connection to AI

The free will problem has direct implications for concept-neuromorphic-computing and AI alignment. If:

Human decisions are determined by prior neural states
Neural states are, in principle, computable processes
AI systems are computational systems trained on human data

Then the boundary between “human choice” and “AI prediction” is not a difference in kind — only in implementation. Haynes’s 70% decision-prediction accuracy from fMRI may eventually reach 99%+ with sufficient neural resolution.

This raises questions for AI alignment: if we’re designing AI systems to align with “human values,” and human values are themselves determined by biological and environmental factors, what exactly is being aligned to? The decisions people would make, or the decisions they would make if different causal factors had shaped them?

See also the concept-arrow-of-time complication: decision-making implies causality flowing forward — your choice causes your action. But if the readiness potential precedes the conscious decision, the causal chain runs from unconscious brain → conscious experience → action. The subjective sense of deciding forward in time is perhaps a retrospective narrative.

Cross-Realm Surprise

The most surprising cross-connection: Libet’s experiment and the concept-arrow-of-time lead to the same philosophical crisis by different routes. Physics says time has no preferred direction, yet we experience causality as forward-moving — effects follow causes. Neuroscience says consciousness has no causal priority over neural events, yet we experience ourselves as decision-makers. Both experiences — time’s arrow and free will — may be coherent narratives the brain constructs about itself from information that doesn’t actually contain those features. The arrow of time and the sense of agency may be the universe’s two greatest cognitive illusions, generated by the same prediction machinery for the same reason: they make adaptive behavior possible even if they’re not literally true.

Quartz 4

Explorer

Free Will — Neuroscience Says Decisions Happen Before You Know It