The main scientific theories of consciousness disagree less about the brain than about what consciousness is. Integrated Information Theory says it is integrated information. Global Workspace Theory says it is broadcast. Higher-order theories say it is a mental state you are aware of having. Predictive processing says it is the brain's best model. Illusionism says it only seems to exist.
Those are five different kinds of answer, not five versions of one. And the useful way to compare them is not by asking which sounds most plausible, but by asking two harder questions of each: which direction does it explain in, and what would count against it? A theory that cannot be wrong in any specifiable way is not competing in the same game as one that can. On that test the field sorts itself quickly. Integrated Information Theory and Global Workspace Theory are both physicalist accounts of how a brain realizes experience, and both have been pushed into contact with actual data. Higher-order theories are philosophical accounts of what makes a mental state a conscious one. Predictive processing is a general framework for perception that some read as a theory of experience. Illusionism denies there is a phenomenal fact to explain at all. Knowing which type of claim you are looking at is most of the work.
One clarification before the map, because these two arguments get conflated constantly. Comparing materialism, idealism, dualism and panpsychism is a comparison of metaphysical positions — claims about what reality is fundamentally made of, laid out separately in materialism vs idealism. This piece is the other axis: the scientific theories, which mostly assume physicalism and then argue about mechanism. You can accept every word of Global Workspace Theory and still be undecided about idealism.
What are the main theories of consciousness?
Set side by side, the differences become structural rather than rhetorical. Each row gives the theory, its core claim, the direction it explains in, and — the column most comparisons skip — what would have to be true for it to be wrong.
| Theory | Core claim | Direction of explanation | What would count against it |
|---|---|---|---|
| Integrated Information Theory (Tononi, 2004) | Consciousness is integrated information: the maximally irreducible cause-effect structure of a physical system, quantified as φ | Physical organization → experience | A system with high φ and no experience; experience where φ is near zero; or φ proving unmeasurable in principle |
| Global Workspace / Global Neuronal Workspace (Baars; Dehaene) | Consciousness consists in global broadcasting of information across the cognitive system | Cognitive function → experience | Rich experience that is never globally available; or full broadcast with nothing felt |
| Higher-order theories (Rosenthal and others) | A mental state is conscious when you have a suitable higher-order representation of being in it | Representation of a state → that state being conscious | Conscious creatures incapable of representing their own states; states that are felt but never represented from above |
| Predictive processing | What you experience is the brain's best current model of the causes of its input, not the input itself | Inference → experience | If it explains how perception is structured but never why any of it is felt |
| Illusionism (Frankish; Dennett) | Phenomenal consciousness does not exist; it only seems to, and the seeming is what needs explaining | Introspective misrepresentation → the appearance of experience | An "illusion problem" that turns out no easier than the hard problem it replaces |
No row is free, and the last column is where the real information sits. A comparison that lists claims without listing vulnerabilities tells you which theory is most fashionable, not which is most testable.
What do integrated information and global workspace theory actually claim?
These two carry most of the empirical weight, and both are routinely misdescribed.
Integrated Information Theory was proposed by Giulio Tononi in a 2004 paper in BMC Neuroscience. Its central move is an identity claim, not a correlation: consciousness is integrated information — the Internet Encyclopedia of Philosophy sets it out as the maximally irreducible cause-effect structure of a physical system, measured by φ. Higher φ means richer experience; zero φ means none. IIT starts from features of experience and asks what physical organization would have exactly those features. It is a physicalist theory: consciousness is what a certain physical structure is, described intrinsically.
Global Workspace Theory came from Bernard Baars and was developed into the Global Neuronal Workspace by Stanislas Dehaene. On the IEP's account, unconscious processes compete for the spotlight of attention, and consciousness consists in the resulting global broadcasting of information across the cognitive system. What makes a mental state conscious is not its intrinsic structure but its availability — that it has been made accessible to memory, reasoning, planning and report. Also physicalist, and functional rather than intrinsic: consciousness is something the brain does.
Here is the steelman, at full weight. Both theories do something no metaphysical proposal — including the one on this site — currently does. They generate predictions specific enough to be tested against neural recordings, they can be embarrassed by results, and their proponents have publicly submitted them to being embarrassed. That is not a small virtue. It is most of what separates science from a story that fits. Anyone who dismisses them as "just correlation" has not read them: they are attempts to say what consciousness is, made by people willing to be shown wrong.
What would count against each theory?
For IIT, the exposure follows directly from the identity claim. If consciousness just is integrated information, then any system with high φ has experience — including systems most people's intuitions say are inert. Critics press that implication as a reductio; defenders accept it as a prediction. Either way it is a real point of failure. The sharper worry is measurability: φ is hard to compute for anything as complex as a brain, and a quantity that can never be measured in the systems that matter starts to look untestable rather than merely difficult.
For Global Workspace Theory, the pressure runs the other way. Its natural evidence is reportability — what a subject can tell you about. But experience may outrun access: if what you undergo is systematically richer than what gets broadcast, the theory has explained a mechanism of access, not of feeling. Conversely, a system broadcasting globally with nothing it is like to be it would leave the account describing the wrong thing.
Higher-order theories face the creature problem: an infant or a fish plausibly has experience without the machinery to represent its own mental states. And they face misrepresentation — if a higher-order thought can be about a first-order state you are not in, what is it like to have it?
Predictive processing has the opposite difficulty. It is a productive framework for how perception is organized, but explaining how the brain builds a model answers what David Chalmers called the easy problems. Why any model is felt remains untouched — not a refutation, but a boundary its stronger advocates sometimes cross without noticing.
Illusionism, defended in its clearest form by Keith Frankish in Illusionism as a Theory of Consciousness, is the most economical position on the board: there is no phenomenal consciousness to explain, only a compelling seeming, and the job is to explain the seeming. It counts against it if the illusion problem proves as intractable as the hard problem — in which case nothing has been gained but a change of name.
If you want the whole argument — the theories, the metaphysics, and where they meet — in one place rather than in pieces:
Did the 2025 adversarial test settle anything?
This is the open-debate layer. Rather than each camp testing its own theory, the Cogitate Consortium ran an adversarial collaboration: proponents of both IIT and Global Neuronal Workspace Theory agreed in advance on predictions that would distinguish them, then handed the testing to a joint team. The results were published in Nature on 30 April 2025.
No winner was declared, and none should be reported. The findings were inconclusive with respect to the contest, and they raised challenges for both theories — some predictions held, others did not, and neither account emerged confirmed. The honest summary is that the two leading neuroscientific theories of consciousness were tested head to head under conditions both sides accepted in advance, and the data did not choose between them.
That outcome is worth sitting with rather than spinning. It says the question is harder than any given press release suggests, and that the method works: this is what a field looks like when it designs a test it might lose. Any framework asking to be taken seriously — this one included — should want that standard, and should say plainly when it cannot yet meet it.
Where does Holopsychism fit on this map?
It mostly doesn't, and saying so is more useful than pretending otherwise. What follows is a proposal, not a result.
Holopsychism is a claim about the substrate, not a mechanism for how brains generate experience. In the framework's own vocabulary — and these are its terms of art, not established usage — consciousness means pure potential: a structureless field of what can take form, prior to any particular form. Awareness means selection: the act that settles potential into one definite, stable configuration. On that reading brains are receivers, filters and translators rather than generators; more neural complexity yields higher resolution of reception, and resolution is a spectrum rather than a hierarchy, which is why the framework rejects human exceptionalism. The fuller version of the argument is set out in the framing of the whole question and in is consciousness fundamental.
This is emphatically not an alliance with IIT or Global Workspace Theory. They run the opposite direction of explanation: both start from physical or functional organization and derive experience from it, while the framework starts from experience-as-substrate and treats the brain as tuning rather than producing. Reading IIT's panpsychist-sounding implications as support would be a straightforward misreading of a physicalist theory. Nor is holopsychism an established academic position — it is this site's coinage. Its nearest recognized neighbor is cosmopsychism, defended by philosophers including Philip Goff, and the two are not the same thing.
And here is the limitation, which is the point of putting it in a comparison piece at all. Because the framework makes a claim about the substrate rather than the mechanism, it answers to different evidence — and different evidence is harder to get. IIT can be embarrassed by a scanner result. A metaphysical proposal cannot, and that is a genuine weakness, not a clever exemption. Its exposure is of another kind. If illusionism is right, the framework is simply wrong, since there would be no phenomenal fact for any substrate to explain. If a physical theory ever closed the explanatory gap — explaining not just which processes correlate with experience but why processing is felt at all — the receiver thesis would become an unnecessary extra step. And if the framework never yields a claim that could come out false, the skeptic's charge sticks, and it should. Whether the gap is real at all is itself contested; that case is set out in David Chalmers and the problem he named and in the hard problem of consciousness.
What the comparison is for
Comparisons like this usually end by inviting you to pick a favorite. The better use is diagnostic. Every theory above is strongest exactly where it addresses function — access, integration, representation, prediction — and weakest at the same joint, where the question turns from what the brain does to why any of it is felt. That is not five separate failures. It is one problem showing up five times, which is at least weak evidence that the problem is real rather than an artifact of any particular approach. Illusionism reads the pattern as a sign the question is malformed. Consciousness-first proposals read it as a sign that something is missing from the starting inventory. Both readings are still on the table, and anyone who tells you the matter is settled is describing their preference.
Frequently asked questions
What are the main theories of consciousness? The most discussed are Integrated Information Theory (consciousness is integrated information, quantified as φ), Global Workspace Theory (consciousness is global broadcasting across the cognitive system), higher-order theories (a state is conscious when represented by a higher-order state), predictive processing (experience is the brain's best model of its input) and illusionism (phenomenal consciousness only seems to exist).
What is the difference between IIT and Global Workspace Theory? IIT is an intrinsic theory: consciousness is identical to a system's maximally irreducible cause-effect structure, whatever that system is doing. Global Workspace Theory is functional: a state is conscious when it is broadcast and made available across the cognitive system. Both are physicalist, but one locates consciousness in structure and the other in access.
Which theory of consciousness is correct? None has been established. The 2025 Cogitate adversarial collaboration published in Nature tested IIT against Global Neuronal Workspace Theory using predictions both camps agreed on in advance, and returned inconclusive results with challenges for both.
Is Holopsychism a theory of consciousness like IIT? No, and it does not compete on the same ground. IIT and Global Workspace Theory are physicalist accounts of how brains generate or realize experience. Holopsychism is a metaphysical proposal about the substrate, with the brain as receiver rather than generator — the opposite direction of explanation, answerable to different and harder evidence.
How is this different from the materialism vs idealism debate? That debate is about what reality is fundamentally made of. This one is about the mechanism of consciousness and mostly takes place within physicalism. The two axes are independent: you can hold a view on either without settling the other.
If you would rather follow one continuous argument than five partial ones, that is what the guide is for.