Source Count: 18 | Weighted Score: 42 | Source Confidence: [5/5] | Primary Tier: 2 | Last Updated: March 11, 2026
Keywords: integrated information theory, IIT, phi, Tononi, Koch, consciousness, information, integration, qualia, axiom, postulate, exclusion, intrinsic, composition, panpsychism, cerebellum, cerebral cortex, complexity
Category Tags: consciousness, information-theory, IIT, Tononi, neuroscience, philosophy, measurement
Cross-References: K_1_01 — Consciousness Overview · ZD_1_02 — Integrated Information Theory · ZD_1_02 — Information Theory · ZD_2_08 — Computation

QUICK SUMMARY

Integrated Information Theory (IIT), developed primarily by neuroscientist Giulio Tononi (b. 1960) at the University of Wisconsin-Madison, with significant contributions from Christof Koch (Allen Institute for Brain Science), is the most mathematically formalized theory of consciousness currently available. IIT begins not with the brain but with the phenomenology of consciousness itself — identifying five essential properties (axioms) of every conscious experience: Intrinsicality (experience exists from the intrinsic perspective of the system), Composition (experience is structured — composed of multiple distinctions and relations), Information (experience is specific — this particular experience, not another), Integration (experience is unified — it cannot be reduced to independent components), and Exclusion (experience is definite — it has a specific content and spatial/temporal grain). From these axioms, IIT derives corresponding postulates about the physical substrate that must support consciousness — culminating in the measure Φ (phi), a quantity that represents the irreducible integrated information generated by a system above and beyond its parts. A system is conscious to the degree that it has high Φ — and the specific structure of its integrated information determines what it is conscious of. IIT makes bold predictions: the cerebral cortex (with its massive recurrent connectivity) should have high Φ and be conscious; the cerebellum (with its feedforward, modular architecture), despite having more neurons, should have low Φ and contribute little to consciousness — a prediction consistent with clinical evidence. IIT also implies a form of panpsychism: any system with non-zero Φ has some degree of consciousness, including simple physical systems. The theory has attracted both ardent support and vigorous criticism — opponents argue that Φ is computationally intractable for real brains, that the axiom-to-postulate derivation is not logically tight, and that the panpsychist implications are absurd.

1. VERIFIED CLAIMS (Tier 1 — Peer-Reviewed / Established Science)

1.1 The Axioms and Postulates

• IIT's formal structure begins with five axioms (properties of consciousness taken as self-evident):
• Intrinsicality: consciousness exists from the system's own perspective, not from an external observer's
• Composition: experience is structured — composed of phenomenal distinctions and their relations
• Information: each experience is specific — it specifies "this and not that"
• Integration: experience is irreducibly unified — it cannot be factored into independent sub-experiences
• Exclusion: experience is definite — with a specific content and spatiotemporal grain
• From each axiom, IIT derives a corresponding postulate about the physical substrate:
• A substrate of consciousness must have cause-effect power (from Intrinsicality), composed of mechanisms (Composition), specifying particular cause-effect structures (Information), that are irreducible (Integration → measured by Φ), and that exist at a definite grain (Exclusion)

1.2 Phi (Φ) — The Measure

• Φ quantifies the amount of integrated information:
• Computed (in principle) by: partitioning a system in every possible way and finding the minimum information partition (MIP) — the partition that least reduces the system's cause-effect information. Φ is the information lost across this minimum partition
• High Φ = the system loses a lot of information even in its "best" partition → consciousness is high
• Low Φ (or Φ = 0 for systems that can be decomposed without information loss) → system is not or minimally conscious
• Computational intractability: computing Φ exactly requires evaluating all possible partitions of the system — the number of partitions grows super-exponentially with system size. For a system of ~300 elements, the computation is already beyond the capacity of any computer. This is a major practical limitation

1.3 Cerebellum vs. Cortex Prediction

• IIT predicts that the cerebellum — which contains ~80% of the brain's neurons (~69 billion) but has a largely feedforward, modular architecture — should contribute little to consciousness:
• The cerebral cortex — with extensive recurrent, re-entrant connectivity — should have high Φ
• Clinical evidence supports this: extensive cerebellar damage does not abolish consciousness (only motor coordination and some cognitive functions are affected), while even small cortical lesions can produce specific losses of conscious content (hemineglect, anosognosia, cortical blindness)
• This is one of IIT's strongest empirical predictions

1.4 Historical Development of IIT

2. CREDIBLE CLAIMS (Tier 2 — Academic / Debated but Supported)

2.1 IIT and the Hard Problem

• Tononi and Koch argue that IIT provides a principled solution to the hard problem:
• IIT identifies consciousness with integrated information (Φ structure) — it is not an epiphenomenon, a byproduct, or a functional property, but identical with the cause-effect structure specified by the system
• This is an identity claim (like saying temperature is mean molecular kinetic energy) — not a causal or explanatory claim
• Critics argue that the identity is asserted rather than derived — why should integrated information be consciousness rather than just correlate with it?

2.2 The Perturbational Complexity Index (PCI)

• Casali et al. (2013) developed a practical proxy measure — PCI — that does not require computing Φ directly:
• PCI measures the complexity of the brain's response to a transcranial magnetic stimulation (TMS) pulse — high PCI = complex, integrated, differentiated response
• PCI reliably distinguishes conscious (wakefulness, dreaming, locked-in syndrome) from unconscious (deep sleep, general anesthesia, vegetative state) states in clinical settings — accuracy >95%
• PCI is one of the most practically significant applications of IIT-inspired research
• Sleep and anesthesia support: Deep NREM sleep reduces cortical integration (measured by PCI/TMS-EEG) compared to wakefulness and REM sleep — the "bistability" of cortical responses during NREM (only local, stereotyped responses to TMS) contrasts with the complex, widespread, differentiated responses during consciousness; ketamine at sub-anesthetic doses preserves sensory processing but disrupts integration patterns
• Other Φ proxies: Beyond PCI, approximate measures include Φ*, Φ_R, and Φ_E — their exact relationship to true Φ remains debated, but they offer computationally tractable alternatives for systems with 10–15+ elements

PCI Clinical Results:

• PCI is IIT's strongest empirical achievement — a theory-derived clinical tool that works; the detection of consciousness in apparently vegetative patients has direct ethical implications for clinical decision-making

2.3 Adversarial Collaboration — COGITATE and Templeton Results

• The Templeton World Charity Foundation funded a ~$20M adversarial collaboration ("Accelerating Research on Consciousness") to pit IIT against GNWT (Stanislas Dehaene, Jean-Pierre Changeux, Bernard Baars); both camps pre-registered specific predictions before experiments were run
• A preregistered adversarial test of IIT vs. GNW using EEG, fMRI, and intracranial recordings (Melloni et al., 2023): IIT predicted sustained posterior cortical activity for conscious visual perception; GNW predicted transient frontoparietal ignition

• Neither theory was fully validated — consciousness science remains immature; IIT outperformed GNWT in this specific test, but "winning" ≠ "proven"
• Nature Neuroscience commentaries (March 2025) showed deep disagreement even among collaborators about what the results mean
• The adversarial collaboration model was widely praised as methodologically important for consciousness science regardless of outcome

2.4 Unfolding Argument

• Scott Aaronson (2014) raised a computational challenge: IIT attributes consciousness to simple expander graphs (networks with high Φ by construction) — including systems that seem intuitively unconscious
• Tononi and colleagues responded with the unfolding argument (Tononi et al., 2016): a feedforward system can simulate any input-output mapping of a recurrent system, but it lacks the intrinsic cause-effect structure → the intrinsic perspective matters

2.5 The 124-Scholar "Pseudoscience" Letter (2023)

• In September 2023, 124 scholars signed a PsyArXiv letter claiming IIT's panpsychism implications make it pseudoscientific
• However, a survey of signatories revealed that only ~8% fully agreed with the "pseudoscience" label they had signed
• Many signatories objected specifically to panpsychism, not to IIT's mathematical formalism or its empirical predictions
• The letter was published before the Templeton adversarial collaboration results — IIT's subsequent 2-of-3 confirmed predictions from a $20M pre-registered study sit uncomfortably alongside the "pseudoscience" characterization
• The controversy highlights ongoing tensions between IIT's mathematical rigor and clinical utility on one hand, and its counterintuitive metaphysical implications on the other

3. SPECULATIVE CLAIMS (Tier 3 — Possible but Unverified)

3.1 Panpsychism

• IIT implies that any system with Φ > 0 has some measure of consciousness:
• This includes thermostats, photodiodes, and simple physical systems — though their Φ is vanishingly small and their "experience" correspondingly simple
• IIT's panpsychism is a graded, quantitative panpsychism — not the claim that rocks have human-like consciousness, but that there is a continuous scale from virtually zero to high consciousness

3.2 The Exclusion Postulate Controversy

• The axiom that experience is definite (occurs at a single spatiotemporal grain and has definite borders) leads to the exclusion postulate — consciousness exists only at the spatial and temporal scale that maximizes Φ; this prevents "double counting" of consciousness at multiple scales simultaneously
• Critics argue the exclusion postulate is ad hoc and lacks clear justification independent of the desire for theoretical tidiness; it is difficult to test empirically and remains one of IIT's more controversial structural choices

3.3 Digital Computers Are Not Conscious

• IIT predicts that digital computers — which can simulate any input-output function but lack intrinsic causal structure (they operate through sequences of simple yes/no gates) — have very low Φ regardless of what they compute:
• A perfect simulation of a brain on a digital computer would not be conscious according to IIT — because the simulation lacks the intrinsic integrated information structure
• This is the most controversial practical prediction of IIT — directly opposed to functionalist views (which hold that consciousness depends on computation, not substrate)

4. DUBIOUS CLAIMS (Tier 4 — No Credible Source / Contradicted by Evidence)

4.1 Φ Has Been Measured in the Human Brain

• [MISLEADING] No study has actually computed Φ (per the mathematical definition) for the human brain — the computation is intractable. PCI and other proxy measures are inspired by IIT but do not measure Φ

4.2 IIT Is Experimentally Proven

• [OVERSTATED] While IIT-inspired measures (PCI) have clinical utility, the core theory — the identity between Φ structure and consciousness — remains a theoretical framework, not an experimentally proven theory

Counter-Arguments & Criticisms

Giulio Tononi’s Integrated Information Theory (IIT) and its measure phi (Φ) have faced significant criticism. Scott Aaronson (2014) demonstrated that IIT assigns high Φ values to simple grid-like systems that intuitively should not be conscious, challenging the theory’s explanatory adequacy. Critics argue that IIT’s mathematical framework, while precise, makes counterintuitive predictions (e.g., high consciousness in certain simple computational structures). The phenomenological axioms from which IIT derives its postulates have been questioned as insufficiently justified starting points. Practical measurement of Φ in biological neural networks remains computationally intractable, limiting empirical testability.

IMAGES

No images assigned yet.

BIBLIOGRAPHY

1. Tononi, Giulio | 2004 | "An Information Integration Theory of Consciousness" | BMC Neuroscience | ∅ | 5.42::1–22 | ∅ | ∅ | doi:10.1186/1471-2202-5-42 | ∅ | ∅ | ∅
2. Tononi, Giulio | 2008 | "Consciousness as Integrated Information: A Provisional Manifesto" | Biological Bulletin | ∅ | 215.3::216–242 | ∅ | ∅ | doi:10.2307/25470707 | ∅ | ∅ | ∅
3. Tononi, Giulio, Melanie Boly, Marcello Massimini; Christof Koch | 2016 | "Integrated Information Theory: From Consciousness to Its Physical Substrate" | Nature Reviews Neuroscience | ∅ | 17::450–461 | ∅ | ∅ | doi:10.1038/nrn.2016.44 | ∅ | ∅ | ∅
4. Oizumi, Masafumi, Larissa Albantakis; Giulio Tononi. e1003588 | 2014 | "From the Phenomenology to the Mechanisms of Consciousness: Integrated Information Theory 3.0" | PLoS Computational Biology | ∅ | 10.5:: | ∅ | ∅ | doi:10.1371/journal.pcbi.1003588 | ∅ | ∅ | ∅
5. Casali, Adenauer G., et al. ra105 | 2013 | "A Theoretically Based Index of Consciousness Independent of Sensory Processing and Behavior" | Science Translational Medicine | ∅ | 5.198::198 | ∅ | ∅ | doi:10.1126/scitranslmed.3006294 | ∅ | ∅ | ∅
6. Koch, Christof, Marcello Massimini, Melanie Boly; Giulio Tononi | 2016 | "Neural Correlates of Consciousness: Progress and Problems" | Nature Reviews Neuroscience | ∅ | 17.5::307–321 | ∅ | ∅ | ∅ | ∅ | ∅ | ∅
7. Aaronson, Scott | 2014 | "Why I Am Not an Integrated Information Theorist (or, The Unconscious Expander)" | ∅ | ∅ | ∅ | Blog post, . [Discussed in Tononi et al | ∅ | ∅ | ∅ | ∅ | 2016 response.]
8. Cerullo, Michael A. e1004286 | 2015 | "The Problem with Phi: A Critique of Integrated Information Theory" | PLoS Computational Biology | ∅ | 11.9:: | ∅ | ∅ | ∅ | ∅ | ∅ | ∅
9. Barrett, Adam B.; Anil K | 2011 | "Practical Measures of Integrated Information for Time-Series Data" | PLoS Computational Biology | ∅ | 7.1:: | Seth. e1001052 | ∅ | ∅ | ∅ | ∅ | ∅
10. Massimini, Marcello, et al | 2005 | "Breakdown of Cortical Effective Connectivity During Sleep" | Science | ∅ | 309.5744::2228–2232 | ∅ | ∅ | ∅ | ∅ | ∅ | ∅
11. Tononi, Giulio; Christof Koch | 2015 | "Consciousness: Here, There and Everywhere?" | Philosophical Transactions of the Royal Society B | ∅ | 370.1668::20140167 | ∅ | ∅ | ∅ | ∅ | ∅ | ∅
12. Albantakis, Larissa, et al. e1011465 | 2023 | "Integrated Information Theory (IIT) 4.0: Formulating the Properties of Phenomenal Existence in Physical Terms" | PLoS Computational Biology | ∅ | 19.10:: | ∅ | ∅ | ∅ | ∅ | ∅ | ∅
13. Merker, Bjorn | 2007 | "Consciousness Without a Cerebral Cortex: A Challenge for Neuroscience and Medicine" | Behavioral and Brain Sciences | ∅ | 30.1::63–81 | ∅ | ∅ | ∅ | ∅ | ∅ | ∅
14. Tegmark, Max | 2015 | "Consciousness as a State of Matter" | Chaos, Solitons & Fractals | ∅ | 76::238–270 | ∅ | ∅ | ∅ | ∅ | ∅ | ∅
15. Melloni, L. et al. , 18, , e0268577 | 2023 | "An Adversarial Collaboration Protocol for Testing Contrasting Predictions of Global Neuronal Workspace and Integrated Information Theory" | PLOS ONE | ∅ | ∅ | ∅ | ∅ | ∅ | ∅ | ∅ | ∅
16. Doerig, A. et al | 2019 | "Hard Criteria for Empirical Theories of Consciousness" | Cognitive Neuroscience | ∅ | 10.4::195–213 | ∅ | ∅ | ∅ | ∅ | ∅ | ∅
17. Melloni, L. et al | 2025 | "An Adversarial Collaboration to Critically Evaluate Theories of Consciousness" | Nature | ∅ | ∅ | ∅ | ∅ | ∅ | ∅ | ∅ | ∅
18. Koch, C. | 2019 | ∅ | The Feeling of Life Itself: Why Consciousness Is Widespread but Can't Be Computed | ∅ | ∅ | MIT Press | ∅ | ∅ | ∅ | ∅ | ∅

CROSS-REFERENCE INDEX

Generated from V4 expansion plan. Last Updated: March 11, 2026

<table border="1" cellpadding="12" cellspacing="0" style="border-collapse: collapse; border: 2px solid #888; margin-top: 2em; background: #fafafa;">

<tr><td>

⚠️ AI-Assisted Research Disclaimer

This document was generated and structured with the assistance of AI tools.

While every effort is made to ensure accuracy, AI-assisted content may

contain errors, misattributions, or unintended inaccuracies. **Always

verify claims, dates, and sources independently** before citing or relying

on any information presented here.

• Sources may contain errors. Bibliography entries and cross-references

are checked by automated systems, but mistakes can occur. If something

looks wrong, it may be.

• Speculative and unverified claims are clearly labeled. This project

uses a four-tier evidence system:

• Tier 1 — Verified: Peer-reviewed, established scientific consensus.
• Tier 2 — Credible: Academically supported, debated but grounded.
• Tier 3 — Speculative: Plausible but unverified by mainstream science.
• Tier 4 — Dubious: No credible support or contradicted by evidence.
• This project maps multiple perspectives — not a single truth. Mainstream,

alternative, and skeptical viewpoints are presented side by side for

critical comparison, not endorsement. Inclusion does not imply agreement.

• We are actively improving. Source verification, factuality scoring,

and bibliography enrichment are ongoing. Each revision adds stronger

citations, corrects identified errors, and expands coverage.

📖 For full details on our verification methodology, scoring systems, and

quality metrics, see: Fact-Checking & Verification Systems

Think Openly. Check the sources. Draw your own conclusions.

</td></tr>

</table>