the homunculus in a lab coat
Here is a sentence that sounds like it explains something: consciousness exists because we observe ourselves.
It is a good sentence. It has the shape of an answer. Something looks inward, and in the looking, experience happens — the self is the act of self-regard. Say it with wave functions attached and it sounds better still. Observation collapses superposition. A system that observes itself collapses itself. The mystery dissolves into physics.
It does not. The sentence has a passenger, and the passenger has been riding along for four hundred years.
The claim
Thesis. Explaining consciousness by self-observation smuggles in an observer that is already conscious, and the quantum vocabulary does not remove it — it only makes the smuggling harder to see.
The argument runs in three steps.
- The regress. Observation in the relevant sense requires an observer, and an observer is the thing being explained. Dress it in physics and it is still the little man in the head.
- The escape closes the account. There is a good standard reply — the observing state need not itself be conscious. Grant it, and observation collapses back into ordinary physical interaction, which leaves the original hard problem untouched and the quantum machinery doing no work. Refuse it, and the homunculus is back. I cannot find a third door.
- The ground is already occupied. A theory with no quantum content — self-representationalism — delivers what the intuition wants, and the most developed quantum model in the literature takes consciousness as an input rather than an output. The physics is not paying for its seat.
Scope. This is an argument about one explanatory move, not a verdict on physics. I am not arguing that quantum effects in the brain are impossible — that is a separate question, and section 5 handles it separately on purpose. I am not claiming this connection is original; I have not checked, and I say so rather than implying otherwise.
1. Who is doing the observing
Descartes put the mind in the pineal gland and the seam has been leaking ever since. The problem is not the gland. The problem is the picture underneath: somewhere inside the head there is a place where it all comes together and gets shown — a screen, an audience, a viewer.
Daniel Dennett named it the Cartesian theater, and named the position that survives dropping the dualism: Cartesian materialism, “the view one arrives at when one discards Descartes’ dualism but fails to discard the associated imagery of a central (but material) Theater where ‘it all comes together’.” That is from the 1992 paper with Marcel Kinsbourne, the one that spends forty pages arguing the brain has no privileged finish line — no place where things become conscious at a time, so no fact of the matter about the order in which you experienced them.
That timing argument is Dennett’s real weapon, and it is not the one I need. His secondary objection is, and it comes with a qualifier worth keeping: positing a central theater, he writes, “would apparently be the first step in an infinite regress of too-powerful homunculi.” Too-powerful is doing work there. The trouble is not that something in the head does a job. It is that this particular something has to be as clever as the whole system — so its own activity is either mysterious or decomposes into subsystems that start duplicating the brain you were supposed to be explaining.
That is the homunculus: the little man in the head, watching the show. Nobody defends him. He gets in anyway, because explanations of consciousness have a habit of smuggling a small conscious thing into the machinery and hoping nobody checks its papers.
Put the sentence back up and check its papers. Consciousness exists because we observe ourselves. Who observes? Not a neuron, not a photon — an observer, in the sense that makes observation different from two rocks colliding. That observer is already doing the thing being explained.
It is the same homunculus. He is wearing a lab coat.
The coat is not the objection, and I should say so before anyone else does. Dennett, in 1995, on the diagram itself: “It is not that the homunculi watching the screen have white coats on, or that there are two of them there, or that film, not video, is the medium of transportation. The diagram is more deeply and abstractly wrong.” He is right. The picture is broken however you dress it. My claim is the smaller one he does not make: the coat is what stops people checking. Put the little man in physics and the picture reads as a mechanism instead of a placeholder.
2. The standard reply, and what granting it costs
Here is where most versions of this argument quit early, and they should not, because there is a standard reply and it is decent.
The Stanford Encyclopedia’s survey of higher-order theories of consciousness — the family that says a mental state is conscious when a higher-order state is about it — states the regress objection plainly and then defuses it: the higher-order state does not itself have to be conscious. An unconscious thought about a perception can make the perception conscious without needing a further thought about itself. The tower is two storeys, not infinite.
That reply works. Anyone deploying the homunculus objection has to answer it, and a piece that skips it is arguing against nobody.
It is not Dennett’s reply, and I should not borrow his name for it. His replacement for the theater is distribution — the homunculus’s work spread out in space and time across the brain, with no second storey anywhere. The two-storey answer belongs to the higher-order theorists, who are in some places the people he is arguing against.
So answer it. Grant the move. Let the observer be unconscious.
The quantum version cannot survive its own rescue. If the observing does not require a conscious observer, then “observation” means what it means in the rest of physics — a physical interaction between systems. And now the claim has quietly become: certain physical interactions produce subjective experience. Which is the hard problem, in its original packaging, with nothing added. The quantum machinery bought nothing. It just moved the mystery from the word consciousness to the word measurement, and measurement is a word physics has been arguing about since the 1920s without any of the arguing being about what it feels like to be something.
Take the reply and the physics is idle. Refuse the reply and the homunculus is back. There is no third door I can find.
3. A theory with no quantum content already owns this ground
The strongest thing against the quantum framing is not an objection. It is a competitor with no quantum content at all.
Self-representationalism holds that a conscious state represents both the world and itself. One state, doing both jobs. There is no second observer, no inner audience, no regress to stop — consciousness by self-presentation rather than by inspection. That is very nearly a paraphrase of the sentence we started with, and it arrives without a single wave function.
Which sets a bill. If a quantum account of self-observation is going to be worth having, it has to say what the physics buys that self-representationalism does not already deliver for free. I have not found anyone paying that bill, and it is not a small one.
4. The best-developed model runs the arrows the other way
The most developed treatment of consciousness and quantum collapse in the literature is Chalmers and McQueen’s, circulated as a preprint dated 3 July 2020 and forthcoming in Gao’s Consciousness and Quantum Mechanics from Oxford. It is careful work and worth reading.
It also does the opposite of what the sentence needs.
The authors say it themselves, in the text: “In this article we are not trying to solve the hard problem of how physical processes give rise to consciousness. We are giving an account of the causal role of consciousness that can be combined with many different approaches to the hard problem.” And the model requires a theory of consciousness as an input: they use integrated information, and it is the physical correlate that fixes when collapse happens.
Read the direction of the arrows. Consciousness goes in. Collapse comes out. Nothing in the machinery manufactures experience, and the authors are not pretending it does.
The sentence needs collapse to produce consciousness. The literature’s best-built model needs consciousness to produce collapse. The one serious version of the theory is running backwards relative to the intuition it seems to vindicate — that is my read of the two papers side by side, not a claim either author makes.
5. A separate question: is the brain quantum at all?
Nothing above depends on this. I am answering it anyway, because conflating it with the argument is the signature error of writing on this topic.
Independence. The brain could be full of quantum effects with consciousness playing no role in collapse. Consciousness could cause collapse in a brain that is entirely classical at the relevant scale. Both. Neither. The two questions do not constrain each other.
Not evidence. There is a research programme that uses quantum formalism to model decision-making and judgment, with no claim about brain physics at all. Those results get cited constantly as support for quantum brains. They are math that fits behaviour.
For. 2024 and 2025 gave the microtubule camp its first new material in years. Rats given a microtubule-stabilising drug took about 69 seconds longer to lose consciousness under isoflurane (eNeuro, August 2024) — eight rats, and the authors themselves note the result is potentially consistent with classical models. Add an in-vitro result on ultraviolet superradiance in tryptophan networks, and one in-vivo human MRI signal that drew a published comment and a reply.
Against. Tegmark’s decoherence estimate from 2000: coherence times in neural tissue on the order of 10⁻¹³ to 10⁻²⁰ seconds, against dynamical timescales of 10⁻³ to 10⁻¹ seconds. Ten orders of magnitude, at minimum. The estimate is disputed, and the dispute is live.
Verdict. The pile moved from nothing to small-and-contested. It did not move to supported.
Calibration. There is no tidy mainstream to measure that against. Integrated information theory and global neuronal workspace — two of the most-tested non-quantum theories — were run head-to-head in a preregistered adversarial collaboration, proponents agreeing in advance on the experiments and on what each outcome would mean, 256 subjects across fMRI, MEG and intracranial recording, published in Nature at the end of April 2025. It substantially challenged key tenets of both. Nobody won. And the publicity around its early findings was itself enough that in September 2023, 124 researchers signed a letter calling integrated information theory pseudoscience; the word they objected to hardest was leading. That is not a licence to believe anything. It is a reason to distrust confidence in every direction, including the debunking direction.
What survives
The sentence does not survive as physics. Quantum mechanics is not where consciousness explains itself, and the strongest work in that corner explicitly says it is not trying to.
It survives as something better than physics, which is a structural observation about a class of explanations. Any theory that explains experience by invoking an observer owes you an account of the observer. Say neuron and you have to explain how neurons see. Say higher-order thought and you have to explain why the thought does not need its own thought — the higher-order theorists have an answer, and it is the best answer going. Say quantum measurement and you are betting that a word from physics will do work that the same word has never done in physics.
The little man in the head does not go away when you dress him better. He goes away when the theory stops needing anyone to be watching.