Integrated Information Theory (IIT): Tononi's Mathematical Theory of Consciousness

In 2004, Giulio Tononi — an Italian-trained psychiatrist and neuroscientist at the University of Wisconsin-Madison — published An information integration theory of consciousness in BMC Neuroscience. The paper proposed something almost no contemporary consciousness theory dared: that consciousness is not a property that emerges from something else, but that consciousness IS a specific mathematical quantity, and that quantity can be computed.

Twenty years later, Integrated Information Theory (IIT) is the most mathematically formal theory of consciousness in mainstream cognitive science. It's also one of the most contested — culminating in September 2023 when 124 researchers signed an open letter calling it "pseudoscience."

The debate has not closed. What follows is what IIT actually claims, what's at stake, and where reasonable people disagree.

What does IIT actually claim?

IIT begins with phenomenological axioms — properties Tononi argues that every conscious experience necessarily has, just from being a conscious experience:

1. Intrinsic existence — experience exists for itself
2. Composition — experience is structured, not undifferentiated
3. Information — each experience is THIS specific experience, ruling out others
4. Integration — experience is unified, not divisible into independent parts
5. Exclusion — each experience has definite borders (it includes some content, excludes others)

From these axioms, IIT derives "postulates" — properties that a physical substrate must have if it's going to support consciousness:

• A substrate of consciousness must exist intrinsically (it must have cause-effect power on itself)
• It must be composed of parts
• It must specify particular cause-effect relationships
• Those relationships must be integrated (irreducible to independent parts)
• It must have a definite border

The mathematical formalism quantifies these properties. The headline measure is Φ ("phi"): the amount of integrated information a system generates above and beyond the information generated by its parts considered independently.

Φ = 0 means no consciousness. Φ > 0 means some consciousness. Higher Φ means more consciousness.

[Tononi G. (2004). An information integration theory of consciousness. BMC Neuroscience, 5, 42.]

What does IIT predict?

Several specific predictions distinguish IIT from competing theories:

1. Cerebellum has little Φ despite many neurons. The cerebellum has ~80% of the brain's neurons but they're arranged in highly modular parallel circuits. IIT predicts low Φ here. Empirically: cerebellar damage produces motor problems but not loss of consciousness. ✓ Consistent with prediction.

2. Thalamocortical loops have high Φ. IIT predicts the cortex + thalamus + their reciprocal connections generate most of human Φ. Empirically: damage to these regions reliably produces unconsciousness; cortex generates the rich content of experience. ✓ Consistent.

3. Anesthesia reduces Φ. Under propofol or similar agents, Φ should decrease as integrated information capacity drops. Empirically: this is measurable using techniques like the Perturbational Complexity Index (PCI), which is derived from IIT principles. ✓ The PCI has clinical utility in distinguishing vegetative state vs minimally conscious state vs locked-in patients.

4. Feed-forward systems have zero Φ even if highly complex. IIT predicts that purely feed-forward networks (no recurrent connections) cannot be conscious, regardless of computational power. This is the controversial implication: it means current large language models, if they're purely feed-forward, have Φ = 0 even if they pass behavioral consciousness tests.

5. Simple systems can have tiny but non-zero Φ. Photodiodes, single neurons, simple circuits — these have very small Φ values but non-zero. This is the panpsychism-adjacent implication that has drawn the most heat.

[Tononi G, Boly M, Massimini M, Koch C. (2016). Integrated information theory: from consciousness to its physical substrate. Nature Reviews Neuroscience, 17(7), 450-461.]

The 2023 "pseudoscience" letter

On September 15, 2023, 124 researchers including some prominent consciousness scientists published an open letter in PsyArXiv titled The Integrated Information Theory of Consciousness as Pseudoscience. The signatories argued:

• IIT's central claims are unfalsifiable as currently formulated
• The theory's panpsychism-like implications (rocks could have tiny Φ; computer simulations could have higher Φ than humans) are unscientific
• Recent experimental tests (the "adversarial collaboration" between IIT and Global Workspace Theory) were promoted as more conclusive than they actually were
• Calling IIT "the leading theory of consciousness" misrepresents the field's actual state

The letter was deeply controversial. Many consciousness researchers (including some IIT critics) thought "pseudoscience" was too strong a label and that the letter blurred the line between scientific disagreement and accusation.

IIT proponents (Tononi, Christof Koch, Marcello Massimini) responded forcefully, arguing the letter mischaracterized the theory and that the panpsychism-adjacent implications are derivations from rigorous axioms, not bugs.

[Fleming SM, Frith CD, Goodale MA, et al. (2023). The Integrated Information Theory of Consciousness as Pseudoscience. PsyArXiv pre-print.]

The adversarial collaboration

In 2018-2023, the Templeton World Charity Foundation funded an "adversarial collaboration" between IIT and Global Workspace Theory (GWT, Bernard Baars/Stanislas Dehaene). Both camps pre-registered predictions; experiments were run by neutral labs.

The 2023 results (Cogitate consortium) showed:

• Partial support for both theories — neither IIT nor GWT was fully confirmed or fully refuted
• IIT-predicted patterns in posterior cortex during conscious perception ✓
• GWT-predicted patterns in prefrontal cortex during conscious access ✓
• The data didn't cleanly distinguish the two theories

Interpretation depends on prior commitments. IIT supporters cite the posterior cortex findings; GWT supporters cite the prefrontal findings; neutral observers note that neither was decisively confirmed.

[Cogitate Consortium. (2023). An adversarial collaboration to critically evaluate theories of consciousness. Nature.]

How is IIT different from materialism?

This is the philosophically interesting part. IIT is technically a form of identity theory: consciousness IS integrated information. It's not "produced by" the brain — it IS a mathematical property of certain physical structures.

This has unusual consequences:

• It makes consciousness mind-independent: the substrate generates Φ regardless of whether anyone observes it
• It makes consciousness substrate-independent in principle: any system with the right cause-effect structure has consciousness, even if not made of neurons
• It avoids "emergent consciousness" hand-waving: there's no mysterious step from physics to phenomenology
• BUT: it implies panpsychism-like consequences (any Φ > 0 system has SOME consciousness)

Critics dispute the panpsychism implications. IIT supporters argue these are mathematical consequences of the axioms, not flaws — and that the axioms themselves are well-grounded in phenomenological analysis.

How does this compare to the consciousness-research alternatives?

	IIT	GWT	Higher-Order	Predictive Processing	Idealism (Kastrup)
Consciousness IS	Integrated information (Φ)	Global broadcasting	Higher-order representation	Active inference	Fundamental
Mathematical formalism	Yes (rigorous)	Partial	Limited	Yes (Bayesian)	No
Panpsychism implication	Yes (anything Φ>0)	No	No	Limited	Yes (different form)
Empirical predictions	Specific	Specific	Limited	Specific	Limited
Mainstream acceptance	Contested	Mainstream	Mainstream	Growing	Marginal

IIT and idealism (covered here) end up at adjacent conclusions despite very different paths.

What's the current status?

After the 2023 controversies, IIT remains:

• One of two or three "leading theories" cited in mainstream consciousness research
• The most mathematically formal proposal in the field
• The basis of clinically useful measures (PCI in coma assessment)
• Highly contested in its strong forms (especially the panpsychism implications)
• Actively developed (Tononi and collaborators continue revising; current version is IIT 4.0)

The 2023 letter has not "settled" the question. Most researchers consider IIT a serious theory with serious problems, and they disagree about how seriously to take the problems.

What can we conclude?

Well-established:

• IIT is the most mathematically formal theory of consciousness in mainstream science
• It generates specific empirical predictions, some of which (cerebellum, anesthesia) are confirmed
• The PCI clinical measure derived from IIT has demonstrated utility
• It does not explain "why" consciousness exists in some metaphysical sense — it claims consciousness IS Φ

Not established:

• Whether Φ is actually computable for systems as complex as human brains (current computations are tractable only for small systems)
• Whether the panpsychism implications are problematic features or accurate descriptions
• Whether IIT is genuinely falsifiable in its strong form

Open:

• The 2023 adversarial collaboration was inconclusive
• The "pseudoscience" charge is itself contested
• IIT 4.0 incorporates revisions responding to critics

How does this connect to the broader pillar?

IIT sits within the scientists pillar as the mainstream-science alternative to less formalized consciousness theories. Its work intersects with:

• Donald Hoffman's Interface Theory — different formal framework
• Bernardo Kastrup's analytic idealism — adjacent conclusions, different argument
• Rupert Sheldrake's morphic resonance — empirically engaged alternative

For the broader theoretical landscape, see Consciousness Theories 2026.

Sources

• Tononi G. (2004). An information integration theory of consciousness. BMC Neuroscience, 5, 42.
• Tononi G, Boly M, Massimini M, Koch C. (2016). Integrated information theory: from consciousness to its physical substrate. Nature Reviews Neuroscience, 17(7), 450-461.
• Oizumi M, Albantakis L, Tononi G. (2014). From the phenomenology to the mechanisms of consciousness: Integrated Information Theory 3.0. PLOS Computational Biology, 10(5), e1003588.
• Cogitate Consortium. (2023). An adversarial collaboration to critically evaluate theories of consciousness. Nature (preprint).
• Fleming SM, et al. (2023). The Integrated Information Theory of Consciousness as Pseudoscience. PsyArXiv.
• Casali AG, Gosseries O, Rosanova M, et al. (2013). A theoretically based index of consciousness independent of sensory processing and behavior. Science Translational Medicine, 5(198), 198ra105. — the Perturbational Complexity Index paper