Attention Schema Theory

← Encyclopedia

A brain controls its body with a model of its body — schematic and physically wrong, carrying no synapses or bone density or muscle chemistry, only a useful caricature of the limbs and their reach, and it works for that reason, because a control system needs a fast summary, not a textbook. The Princeton neuroscientist Michael Graziano asks what happens when the same trick is turned on attention. The brain selects and amplifies some signals over others; that selection is a physical process with real machinery. A system controlling attention would do well to model it, and the model, like the body schema, would be a caricature that leaves the machinery out. Attention Schema Theory is the claim that this caricature has a name in ordinary speech. We call it awareness.

The theory is explicitly materialist, and its central object is a representation. On Graziano’s account the brain builds an internal model of its own attention — the attention schema — and subjective awareness just is the content of that model. The schema is mechanistically inaccurate by design. It “would not depict synapses, neurons, lateral inhibition, or electrochemical signals,” Graziano and Taylor Webb wrote in 2015; it would depict instead “something physically incoherent, a process without a physical manifestation.” A brain that consults this model, asked what attention is, finds something with all the salient properties of attention and none of the means by which it happens. So it reports a power of awareness that grasps things yet has no physical substance — a non-physical inner feel. The reported strangeness of consciousness, on this telling, is not a discovery about the world but the signature of an efficient self-model that omits its own wiring.

The obvious paraphrase of that move gets it wrong. Attention Schema Theory is not a theory of how neural activity comes to be felt; it is a theory of why brains claim and believe they are conscious — of, in Graziano’s framing, how an information-processing machine arrives at the extraordinary claim that it possesses a non-physical, subjective awareness and assigns that claim a high degree of certainty. The 2015 thesis is exactly this: the content of the model “leads a brain to conclude that it has a subjective experience.” The explanandum has shifted, deliberately, from the experience to the conviction, and that shift is what lets Graziano say the theory dissolves the hard problem rather than solving it. In the 2013 book Consciousness and the Social Brain he put it at its most direct: “The answer may be that there is no hard problem,” and the felt vividness and ethereal nature people attribute to experience “may be explainable as components of a descriptive model.” The demand for a further explanation — why does the processing feel like something? — is itself a product of the model. The brain represents its attention as a non-physical glow, then asks how a physical brain could host one, and the question inherits its mystery from the representation that generated it. Close the loop and the hard problem is not answered but disarmed — a claim the theory makes about its own reach, and everything contested in it turns on whether the loop really closes.

The theory has a social origin that gives it an evolutionary spine. It was first proposed not from introspection but from social cognition, in the 2011 paper with Sabine Kastner. The brain attributes awareness to other people constantly — a quick model of where another’s attention is pointed, useful for predicting what they will do. Graziano’s wager is that this is the same machinery, turned inward: “whatever circuitry the brain uses to attribute consciousness to others is also used to attribute consciousness to ourselves.” Both attributions implicate the temporoparietal junction, where the relevant cortical networks appear to converge, though how tightly self- and other-attribution share a mechanism remains an open empirical question. The payoff is parsimony: a model of attention would be adaptive twice over, for steering one’s own attention and for reading everyone else’s, the double duty evolution favors.

Across two decades the position was built out in stages. The 2011 paper introduced the social hypothesis; the 2013 book gave it length and the “no hard problem” passage; the 2015 paper supplied the mechanistic account — the body-schema analogy, the omission-of-physics argument, testable predictions about attention and awareness coming apart. A 2017 paper extended it toward artificial consciousness — a suitably equipped machine could be built to claim awareness on the same grounds a brain does — and a second book, Rethinking Consciousness, restated the whole in 2019. The predictions are real and still under study: attention operating without awareness should show characteristic control deficits, and the two should dissociate in conditions like blindsight and inattentional blindness. How robustly the evidence bears this out is unsettled, and the strongest early demonstrations come from the proponents’ own laboratory.

Graziano groups the theory with the illusionist family of Daniel Dennett and Keith Frankish, but with a qualification easy to flatten and important to keep. In the 2019 review, written with Arvid Guterstam and others, the authors split consciousness in two. There is i-consciousness — “i for information” — the selecting, enhancing, and broadcasting of signals, which “literally and mechanistically exists.” And there is m-consciousness — “m for mysterious” — the experiential essence people believe accompanies it. The thesis is a sentence: “I-consciousness is what the brain actually has; m-consciousness is what the brain thinks that it has.” The brain genuinely processes information; it is merely “slightly misled” about its nature, and there is, they insist, someone home. This is weak illusionism in Chalmers’s taxonomy: consciousness exists but lacks a property it seems to have, as against strong illusionism, on which it does not exist at all. The label sits uneasily on Graziano even so. He declined the bare “illusionist” description in 2016, then in 2019 called the theory “a type of illusionist theory” — held at arm’s length, not quite set down.

That placement reaches in two directions at once. Toward Global Workspace Theory, Graziano takes the workspace as an account of i-consciousness — access and broadcast — and folds it in rather than fighting it. Its weakness, he argues, is that broadcasting information does not explain why a system would believe it has a subjective experience; the attention schema is the added layer that supplies the missing piece, a higher-order model of the workspace itself. The two are stacked, not opposed, and the schema becomes, in his phrase, “the simplest possible way to unify” higher-order thought and the global workspace. Toward Integrated Information Theory and panpsychism, the relation is flat opposition. Those views make consciousness fundamental — identical to a system’s integrated causal structure, present by degree wherever that structure is, basic as mass or charge. Attention Schema Theory runs the ledger the other way: the mysterious extra is not fundamental, not a real property, but a model fully reducible to information processing. Graziano’s hostility to the integrated-information camp is open, and is polemic rather than verdict; he has called it a magicalist theory with no chance of scientific success, a line marking the temperature of the dispute, not its resolution.

The criticisms converge on a single seam, the one the theory was built to cross. Does explaining the belief in experience explain the experience? David Chalmers gives the objection its sharpest handle with the meta-problem of consciousness — the problem of explaining why we think there is a problem of consciousness, a question about verbal reports and so, in principle, easy. He places Attention Schema Theory among meta-problem solutions and credits its ambition, then withholds the conclusion: positing a self-model “does not itself constitute anything close to a solution,” because we model our own beliefs all the time without any sense of mystery, so a self-model alone does not explain why this one should generate problem intuitions, or represent awareness as ethereal at all. There is a double edge that cuts toward Graziano as much as away: a meta-problem solution that truly explained our consciousness reports without invoking consciousness would underwrite the illusionist’s debunking argument, so a working theory would help the deflationist, not the realist. Frankish presses from the other flank, that only strong illusionism can dissolve the hard problem, and that the theory’s weak stance may be too weak for the work Graziano asks of it.

The deepest objection is older and simpler. To misrepresent attention as phenomenal, the misrepresentation must appear to something — an illusion needs a subject to be fooled, and a pattern of activation, the philosopher Raymond Tallis among others has urged, is not confused about anything. The theory helps itself to the language of seeming while denying the phenomenality that makes seeming coherent. The dilemma tightens: if representing attention as phenomenal is enough to make it phenomenal, then phenomenality is real after all; if it is not enough, the theory has not explained why anything even seems to be experienced. François Kammerer adds a resistance problem — why the conviction of inner experience is so stubborn that no degree of belief in the theory makes the glow recede. Each charge meets a rejoinder: the model, the theory replies, is itself the someone to whom things are presented, no further audience required; seeming is just what the model’s content amounts to; and the recalcitrance of the intuition is what an efficient, cognitively sealed self-model would predict. No party has landed a blow the other concedes.

The traditions documented in these pages put mind near the ground of things — consciousness primary, not a late accident of matter. Attention Schema Theory occupies the far pole, and the interest of holding the two in one view is the clean opposition: a serious, materialist, testable account on which the felt interior is not the foundation of anything but a fiction the brain tells itself for control — the strongest naturalistic case that the inner glow the old doctrines built upon is a representation reporting on machinery it cannot see. Whether it succeeds is open, and an open question leaves both standing. What the theory does, beyond dispute, is locate the disagreement precisely. Either the brain has an experience, or it has a model that says so, and from the inside — by the theory’s own admission — there is no way to tell which.

→ Related: Hard Problem Of Consciousness · Global Workspace Theory · Integrated Information Theory · Panpsychism