At what point in the causal chain does someone become "responsible" for their action?
The person committed murder. They are responsible; they should be punished.
But why did they commit murder? Because they had a rage disorder, elevated testosterone, and poor impulse control. Did they choose to have a rage disorder? No. Did they choose their testosterone levels? No. Did they choose their developmental environment that wired their impulse control circuits? No.
So we go back one more level. They had these neurobiological substrates because of genetics and early experience. Did they choose their genetics? No. Did they choose the family they were born into? No. Did they choose the stress exposure in infancy that dysregulated their stress systems? No.
At what point in this chain does "choice" enter? At what point do we stop excusing and start blaming?
This is Sapolsky's core problematic: neurobiology traces every human action to causes outside the person's conscious control. Yet we retain a categorical need to assign responsibility, to distinguish the guilty from the innocent, to punish the criminal and exonerate the person who "didn't choose" their circumstances.1
The neurobiology shows no clear boundary between "responsible" and "not responsible" action. It shows a continuum. A person with a prefrontal tumor that destroys their moral reasoning is less responsible than a person without the tumor. A person with childhood trauma that dysregulated their threat-detection is less responsible than a person with secure attachment. A person with psychopathic traits (reduced amygdala-insula activation) is less responsible than a person with intact moral affect systems.
Yet all of these people—tumored, traumatized, psychopathic—retain the neurobiological capacity to process information, form intentions, and execute motor sequences. They are not automatons. They are neurobiology-determined agents acting from neural substrates, exactly like everyone else.
The question becomes: if everyone's action is neurobiologically determined, what does "responsibility" actually mean?
Sapolsky's position is agnostic. He documents the neurobiology in clear detail, shows how it determines action, and then refuses to resolve the philosophical problem of moral agency.2 He treats it as a genuine paradox: neurobiology explains action without determining responsibility. We are free to hold people responsible in a practical sense (we need to manage behavior, protect society, incentivize prosocial conduct). But we cannot claim that responsibility is metaphysically real—that there is some fact about the universe that determines who "really" deserves punishment.
The criminal justice system requires categorical judgments: guilty or not guilty, responsible or insane, deserving punishment or requiring treatment. But neurobiology offers only gradients. There is no moment at which responsibility suddenly switches on. There is no neural marker that divides the person who should be punished from the person who should be treated.
A person with a clear prefrontal tumor causing behavior change might be judged insane (not responsible). A person with the same tumor but who has managed to partially compensate through behavioral strategies might be judged sane (responsible). A person with elevated testosterone and poor impulse control (neurobiologically very similar to many criminal perpetrators) is held fully responsible if they commit violence.
The categories (responsible/not responsible, guilty/not guilty, deserving punishment/requiring treatment) do not map onto the neurobiology, which offers only gradients of diminished capacity, distributed causation, and partial agency.
This creates a foundational tension: the system (criminal justice) requires categorical judgments. The neurobiology (science) describes gradients. The two are fundamentally incompatible.
Sapolsky's greatest intellectual virtue is his refusal to resolve this tension prematurely. He does not try to argue that neurobiology proves people are not responsible (they might still be, in some sense; the neurobiology just doesn't determine the answer). He does not try to argue that we should abandon responsibility concepts (they serve pragmatic and social functions, even if not metaphysically grounded).
Instead, he leaves the tension active: neurobiology shows that action is caused by factors outside conscious control, yet we must assign responsibility and punishment for practical and social reasons. The tension reveals something important about human judgment. We need to attribute responsibility, even if we cannot philosophically justify it.
This generates a productive question: if responsibility cannot be grounded in neurobiology, on what grounds are we assigning it? The answer may be: pragmatic, institutional, and narrative grounds. We assign responsibility where it is most likely to deter harmful action, where it aligns with social intuitions about fairness, and where it fits the narrative of how we understand agency.
But these grounds are fragile. They acknowledge that responsibility is not discovered in nature; it is constructed by our social systems for our social purposes. That construction is legitimate if it serves good purposes (reducing harm, protecting vulnerable people). But it should be undertaken with awareness that it is construction, not fact.
The history of criminal law shows that categorical judgments about responsibility are not timeless truths but historical constructions reflecting the values and understanding of particular moments.
In some historical periods, children as young as seven were held criminally responsible for capital crimes. In other periods, minors are completely exempted from criminal responsibility below a certain age. In some legal traditions, insanity exempts you from responsibility. In others, moral insanity is no excuse.
None of these legal framings is "correct" in a neurobiology sense. They all represent institutional choices about when and how to assign responsibility. As neurobiology becomes more sophisticated, these legal framings become harder to maintain (if neurobiology shows that criminal behavior is heavily influenced by factors outside conscious control, the category of "responsibility" becomes thinner and thinner).
What this cross-domain connection reveals: Responsibility is not a natural kind discovered by neuroscience; it is an institutional and historical category that serves social functions. As our understanding of neurobiology deepens, we may need to fundamentally reconstruct how we assign responsibility—moving toward preventive and treatment-focused systems rather than purely punitive ones.
Eastern philosophical traditions, particularly Buddhism, propose a framework that bypasses the responsibility paradox entirely. Instead of asking "who is responsible?" Buddhism asks "what are the conditions that produce suffering, and how can those conditions be changed?"
This is a fundamentally different question. It does not require assigning responsibility. It requires understanding causation and intervening at causal points to reduce suffering. A person who commits harm due to neurobiological dysfunction is not "irresponsible" in the Buddhist framework; they are someone whose conditions have led to harmful action. The response is compassion and intervention directed at changing conditions, not punishment directed at assigning blame.
This represents a different way of handling the causation problem: not denying that action is caused (neurobiologically or otherwise), not trying to locate some metaphysical fact of responsibility beneath the causation, but instead accepting causation fully and responding with what contemplative traditions call karmic understanding—understanding that all action arises from conditions, and working skillfully with those conditions.
What this cross-domain connection reveals: There may be ways of organizing society that respond to harmful behavior without requiring the problematic category of responsibility. The challenge is redesigning institutions around prevention, treatment, and harm reduction rather than punishment and blame.
Three people in three rooms, each handed the same trolley scenario. One pulls the lever and refuses to push the man. One pulls and pushes both. One refuses both. They are not three personality types having different opinions. They are three nervous systems wired so that different circuits dominate the moment of decision. That is what the courtroom is asking when it asks "is this person responsible" — and the courtroom is asking a question that doesn't fit the brain it's asking about.
The Trolley Problem in the Brain catches the plurality this page argues for, in concrete numbers. Greene's data sorts populations into a 30/30/40 split: thirty percent consistent deontologists whose amygdala-insula veto is strong enough to override every utilitarian arithmetic, thirty percent consistent utilitarians whose dlPFC dominance silences the somatic brake, forty percent context-dependent whose verdict flips with framing. Same scenario, three brains, three answers. The "moral agent" the courtroom wants to indict is at least three different neural configurations producing different conclusions from identical facts. Asking which one is the real agent that should be held responsible is asking which of three real things is the real one.
Two defendants can stand before a judge with the same brain scans, the same histories, the same deliberation patterns, and have produced opposite moral conclusions because different neural systems dominated their decisions in the moments of their acts. Sentencing them as if they were the same agent applying different reasoning to different situations misses the structure underneath. They were, in those moments, different agents — sharing a body and a name, but not a circuit configuration.
This sharpens what the page already argues for: if responsibility is a tool we build for social functions rather than something discovered in the universe, the tool has to fit the substrate it actually operates on. Deterrence threats reach the dlPFC — they speak the language of rules and consequences. They do not reach the amygdala-insula, which doesn't reason about deterrence; it just fires or doesn't. Punishment systems built on the assumption that the punished agent will integrate the threat into future decision-making are right about the dlPFC and wrong about the visceral system that actually overrides the dlPFC in many of the acts they're trying to deter. Treatment that works on amygdala-insula recalibration — somatic therapy, exposure-based intervention, certain contemplative practices — reaches a layer punishment cannot. These aren't softer or harsher versions of the same intervention. They're interventions on different circuits.
The harder thing the handshake says: the 30% pure utilitarians who would push the man, harvest the organs, smother the baby are not missing their moral system. They are running the same dlPFC-dominant configuration modernity celebrates in some contexts (the surgeon making the triage call, the general accepting tolerable casualty rates, the public health official computing the curve) and condemns in others (the psychopath, the sociopath, the killer who feels nothing). The neural configuration is identical. What changes is whether the institution sanctioning the act is recognized as legitimate. Responsibility, in this reading, doesn't live in the agent. It lives in the institution that frames the act — and we have been asking the courtroom to find it inside the wrong skull.
See Free Will & Moral Responsibility for the compatibilist position the trolley data complicates — the "deliberation flowing from values" defenders of free will rest the case on isn't a unified process; which deliberation, processing which values, in which circuit? See Voice Dialogue for a clinical approach that already treats the agent as plural and built a method around it.
The Sharpest Implication
Neurobiology does not prove that you are not responsible. It proves that responsibility cannot be grounded in neurobiology. This means responsibility is not a natural fact; it is a social construction. The question becomes: what social purposes does responsibility serve, and can we serve those purposes better through different frameworks?
If the purpose of assigning responsibility is to deter harmful action, neurobiology suggests this is inefficient. Most harmful action is driven by factors (poverty, neurobiological dysregulation, social desperation) that punishment does not address. If the purpose is retribution (the guilty deserve to suffer), neurobiology suggests this is philosophically incoherent (if they didn't choose their action, why do they deserve suffering?).
What neurobiology suggests might actually work is addressing the conditions that generate harmful action: treating neurobiological dysregulation, addressing poverty and desperation, reducing social conflict, providing prosocial alternatives.
Yet our institutions remain locked in a responsibility-and-punishment framework because it serves other purposes: social catharsis (we feel satisfied when the guilty are punished), narrative coherence (our stories require heroes and villains), and institutional stability (the legal system is built around categorical judgments about responsibility).
Neurobiology is not the obstacle to this framework. Institutional inertia is.
Generative Questions
If neurobiology makes responsibility philosophically indefensible, but we continue to assign responsibility for pragmatic reasons, what are the limits of that pragmatism? At what point does continuing to assign responsibility become ethically indefensible?
A society organized around treatment rather than punishment would look radically different—no prisons, preventive intervention, addressing root causes. What institutions would we need to create? Would the vast criminal justice infrastructure be repurposed or dismantled?
If we accept that responsibility is a social construction rather than a natural fact, does this undermine the entire basis for blame and moral judgment? Or can we have a robust blame practice that knows it is a construction and defends it on pragmatic grounds?