20.06.03 · philosophy / consciousness

Theories of consciousness: Global Workspace, Integrated Information Theory (IIT), higher-order theories

stub3 tiersLean: nonepending prereqs

Anchor (Master): Tononi, G. — Phi: A Voyage from the Brain to the Soul (2012)

Intuition Beginner

Scientists are racing to explain consciousness — the hard problem from the previous unit. Three leading theories compete. Global Workspace Theory, proposed by Bernard Baars, compares the brain to a theater. Most mental processing happens in the dark backstage: unconscious modules for vision, memory, and language. When information becomes conscious, a spotlight shines on it, and the information becomes globally available to all brain systems.

Stanislas Dehaene refined this into the Global Neuronal Workspace. When signals in the brain grow strong enough, they ignite a network of neurons across the cortex. That ignition is consciousness. Integrated Information Theory, by Giulio Tononi, takes a different route. It defines consciousness mathematically. A system is conscious to the degree that it integrates information.

Tononi calls this measure phi — a single number that says how much a system is, as a whole, more than the combination of its parts. Even simple systems can register a phi value, so the theory suggests consciousness might exist in unexpected places.

Higher-order theories take a third path. They hold that consciousness requires thoughts about thoughts. A mental state is conscious when something in your mind represents that the state is occurring — when you have a higher-order representation of it. These three theories answer the hard problem in competing ways, and the rest of this unit lays them out.

Visual Beginner

Picture three columns side by side, each a theory of consciousness. The first column is a stage lit by a single spotlight: whatever stands on stage is broadcast to the whole theater, and everything backstage stays dark. The second column is a web of glowing nodes, tightly woven, with the letter phi glowing at its centre to measure how integrated the web is. The third column is two stacked boxes — a lower state with an arrow pointing up to a higher state that monitors it.

Each diagram fixes what its theory takes consciousness to be. For Global Workspace, consciousness is the broadcast. For Integrated Information Theory, it is the integration measured by phi. For higher-order theories, it is the monitoring arrow. The diagrams look different because the theories disagree about what kind of fact consciousness is.

Worked example Beginner

Consider a classic experiment from Dehaene's laboratory. A word flashes on a screen for a fraction of a second. Just before and after it, a meaningless pattern masks the word. The subject reports seeing only a blur. Yet a brain scan shows the visual cortex registered the word. The signal travelled inward but stopped there, never reaching the rest of the brain.

Now the experimenter slows the word down slightly. The same visual signal strengthens, crosses a threshold, and ignites a vast network across the prefrontal and parietal cortex. Suddenly the subject sees the word and can report it, remember it, and act on it. On Global Workspace Theory, that ignition, the moment the information goes global, is what makes it conscious.

The contrast is sharp. Below the threshold the brain processes the word unconsciously and locally. Above it the same content is broadcast everywhere and enters experience. This is the theory's central claim: consciousness is global availability. The other two theories give different accounts of the same transition, and the sections below compare them.

Check your understanding Beginner

Formal definition Intermediate+

Building on the hard-problem vocabulary of §20.06.02, this unit fixes the leading scientific theories of consciousness in their technical form. Each theory offers a different answer to the question "what makes a mental state conscious?", and the differences are not merely terminological — they generate rival empirical predictions.

Global Workspace Theory (GWT). Bernard Baars' A Cognitive Theory of Consciousness (1988) ^{[source pending]} models the mind as a theatre. A host of specialised, largely unconscious processors — vision, audition, memory retrieval, language, motor planning — compete for access to a shared, capacity-limited global workspace. Whichever content wins the competition is globally broadcast: it becomes simultaneously available to all the processors, and that broadcast is conscious experience. On GWT, consciousness is a functional relation (global availability), not a special substance or a private feeling.

Global Neuronal Workspace (GNW). Stanislas Dehaene and Jean-Pierre Changeux gave GWT a neuronal implementation. The workspace is realised by long-range cortical neurons, concentrated in frontoparietal cortex, that can broadcast signals across distant brain regions. When a stimulus is weak or masked, processing stays local and transient (subliminal). When the signal crosses a threshold, it triggers ignition: a sudden, synchronised, all-or-none activation of the workspace network, marked in EEG by the late P3b wave. Dehaene sorts processing into three regimes — subliminal (too weak to ignite), preconscious (strong but attention is elsewhere), and conscious (ignited and globally broadcast) ^{[source pending]}.

Integrated Information Theory (IIT). Giulio Tononi begins not from function but from phenomenology, positing axioms every experience satisfies — intrinsic existence, composition, information, integration, and exclusion — and derives a mathematical measure. The central quantity is $Φ$ (phi): the amount of information a system generates as a whole, above and beyond the information generated by its parts considered separately. A system with $Φ > 0$ is conscious to that degree; its specific experiential quality is identified with the geometry of its cause-effect structure. IIT is an ontological theory: it tells you what consciousness is (integrated information), not merely what it does ^{[source pending]}.

IIT yields striking predictions. A feedforward network, however sophisticated, has $Φ = 0$ and is not conscious. The cerebellum has more neurons than the cortex but is wired as near-independent modules, so it contributes little to consciousness — matching the clinical finding that cerebellar damage need not abolish awareness. In split-brain patients, IIT predicts two centres of consciousness; under anesthesia and dreamless sleep $Φ$ collapses, while IIT-inspired readings associate certain psychedelic states with altered integration. Because even simple systems can have nonzero $Φ$ , IIT has panpsychist leanings.

Higher-Order Theories. David Rosenthal's Higher-Order Thought (HOT) theory holds that a mental state is conscious if and only if one has a higher-order thought about it — roughly, a thought of the form "I am in mental state $M$ " ^{[source pending]}. Unconscious states lack this monitoring. HOT honours the transitivity principle: a state is conscious only if one is conscious of it. Variants differ on the nature of the higher-order state. HOROR (Lau and Rosenthal) requires only a higher-order representation of a representation, not a full thought. HOP and dispositional accounts (Carruthers) allow the higher-order state to be merely dispositional rather than occurrent. The deep shared claim is that consciousness is a representational achievement of a higher level monitoring a lower one.

Attention Schema Theory (AST). Michael Graziano's Consciousness and the Social Brain ^{[source pending]} holds that the brain constructs a simplified schema of its own attention — a model of how it selects and processes information — and that this schema is the basis of subjective experience. We attribute consciousness to ourselves because we model our own attention, and we extend the same model to others, which AST offers as an account of social cognition and of why the intuition of soul or spirit arises.

Recurrent Processing Theory (RPT). Victor Lamme argues that recurrent (feedback and horizontal) processing among cortical neurons is sufficient for consciousness, while purely feedforward propagation is unconscious ^{[source pending]}. On RPT, recurrence as early as primary visual cortex (V1) can support awareness, and prefrontal ignition is unnecessary — a direct clash with GNW over whether consciousness lives in the back of the brain or requires frontal involvement.

Key concepts — how the theories relate and where they clash Intermediate+

The theories sort along two axes: whether they are functional (characterising what consciousness does) or ontological (defining what it is), and whether they locate consciousness in frontal or posterior cortex. GWT, HOT, and AST are functional and implicate prefrontal cortex; IIT is ontological and locates the substrate of consciousness in the posterior "hot zone"; RPT agrees with IIT on the posterior site but for different reasons.

GWT versus IIT: functional versus ontological. GWT explains how information becomes globally available but, critics charge, does not explain why global availability is accompanied by any felt quality — it risks describing access consciousness and leaving phenomenal consciousness untouched (the Block-style objection carried over from §20.06.01). IIT attacks the hard problem directly by identifying consciousness with integrated information, so there is no further "why does it feel like anything" once $Φ > 0$ . The price is that the identification is controversial: Aaronson's expander-graph counterexample shows that certain mathematically contrived networks attain enormous $Φ$ yet seem paradigmatically unconscious, and computing $Φ$ for any realistic brain is exponentially intractable ^{[source pending]}. Tononi replies that the relevant measure depends on the system's intrinsic cause-effect structure, not on its combinatorial description, but the exchange remains open.

The prefrontal dispute. GWT and HOT both treat prefrontal cortex as load-bearing for consciousness — GNW because ignition is frontoparietal, HOT because higher-order monitoring is prefrontal. Tsuchiya and Koch press the opposite case: activity in prefrontal cortex may reflect report and metacognition rather than the conscious experience itself, and "no-report" paradigms suggest consciousness can persist without frontal engagement. IIT and RPT converge on a posterior substrate. This is not a quibble: if prefrontal cortex is required, certain lesion patients should be unconscious in ways they are not; if it is not, the GWT and HOT mechanistic stories need revision.

Higher-order theories and the targetless-HOT objection. HOT faces a structural puzzle: what if a higher-order thought represents a first-order state that does not exist — a "targetless HOT"? If consciousness is conferred by the higher-order representation, a targetless HOT should produce a conscious experience of nothing, which is incoherent. Weisberg ("Abandoning the Factives") and Neander ("The Second-Order View") sharpen versions of this problem; defenders revise the theory so that higher-order states must be appropriately grounded to generate genuine phenomenology.

Empirical signatures. Despite their disagreements, the theories converge on a shared empirical project: finding signatures that distinguish conscious from unconscious processing. GWT predicts global ignition and the P3b; IIT predicts that posterior, integrated activity tracks the content of experience more faithfully than frontal activity does; HOT predicts that prefrontal integrity tracks the presence of consciousness. The Templeton World Charity Foundation's adversarial collaboration (2019–2021) tested IIT against GWT on shared data and produced a mixed verdict, with each theory fitting some predictions and failing others — a productive but unresolved standoff.

Theory	Style	Key quantity / mechanism	Locus
GWT / GNW	functional	global broadcast; ignition; P3b	frontoparietal
IIT	ontological	$Φ$ (integrated information); cause-effect structure	posterior hot zone
HOT / HOROR	functional-representational	higher-order thought / representation	prefrontal
AST	functional	attention schema	parietal / TPJ
RPT	structural	recurrent processing	posterior (from V1)

Exercises Intermediate+

Exercise 2 (medium, critique).

Reconstruct Aaronson's objection to IIT and Tononi's reply, then say why the exchange is not settled.

Hint

The objection turns on a class of networks that score very high on the measure yet strike us as plainly non-conscious. What does Tononi say the measure actually tracks?

Answer

Aaronson exhibits families of networks (including expander graphs and certain grids of logic gates) that attain arbitrarily large $Φ$ under IIT's measure, which would make them more conscious than a human — an absurd conclusion. Tononi replies that high $Φ$ requires a rich, irreducible cause-effect structure evaluated from the system's own intrinsic perspective, and that mere combinatorial connectivity does not guarantee it. The exchange stays open because the cause-effect structure is so costly to compute that the counterexamples cannot be decisively refuted on real systems, and because IIT's defenders and critics disagree about which abstract objects the axioms should be applied to.

Exercise 3 (medium, argument reconstruction).

Explain the targetless-HOT problem for Rosenthal's theory and indicate how a defender might respond.

Hint

If the higher-order thought is what confers consciousness, what happens when its first-order target is missing?

Answer

If a higher-order thought makes a first-order state conscious by representing it, then a HOT with no actual first-order target should still generate a conscious experience — an experience of a state that does not exist, which seems incoherent. A defender can respond by requiring that HOTs be suitably grounded in their targets (so a genuinely targetless HOT produces no phenomenology), or by moving to HOROR-style accounts in which the higher-order representation is constitutive but constrained by lower-level activity. Each response weakens the clean biconditional ("conscious iff there is a HOT about it") and shifts work onto the theory of how higher-order states acquire their content.

Exercise 4 (hard, application).

A patient in a vegetative state shows intact local sensory responses but no global ignition on EEG, and posterior $Φ$ estimates are near zero. What would GWT and IIT each predict about the patient's consciousness, and what clinical consequence follows?

Hint

Tie each theory's marker (broadcast for GWT, $Φ$ for IIT) to a verdict, then consider what that verdict means for detecting covert awareness.

Answer

GWT predicts the patient is unconscious in the access sense: without ignition, no content is globally broadcast, so there is no conscious access to stimuli. IIT predicts little or no phenomenal consciousness, since near-zero $Φ$ indicates that the system's cause-effect structure has disintegrated. The clinical consequence is caution: both theories agree the markers are bleak, yet cases of covert awareness (Owen's imagined-tennis fMRI paradigm) show that behavioural unresponsiveness and even absent ignition do not guarantee an absence of experience. The theories therefore sharpen the diagnostic question but do not settle it, and they motivate multi-marker assessment combining ignition, $Φ$ -style integration measures, and active command-following paradigms.

Theoretical frameworks in depth Master

Predictive processing and consciousness

Predictive processing recasts the brain as a hierarchical prediction machine that generates top-down models of its sensory input, propagates prediction errors upward, and updates its models accordingly (Friston, "The Free-Energy Principle"; Clark, Surfing Uncertainty; Hohwy, The Predictive Mind ^{[source pending]}). Precision weighting modulates which error signals are trusted; active inference extends the framework to action, so that organisms change the world to fit their models. Within this architecture, several authors locate consciousness in specific aspects of the predictive hierarchy.

Anil Seth (Being You ^{[source pending]}) frames conscious experience as a controlled hallucination: the brain's best guess about the causes of its sensory input, including interoceptive input from the body. On this view perception is a controlled inference, and the self is the brain's model of itself. Seth and Suzuki extend the account to emotion as interoceptive inference, aligning with Barrett's theory of constructed emotion (How Emotions Are Made; see §29.11.02). Hobson and Friston's "A General Theory of Consciousness" treats waking and dreaming (REM/NREM) as different regimes of the same predictive engine. The framework re-poses the hard problem (§20.06.02): if experience is inference, the explanatory gap shifts from "why does processing feel like anything" to "why does a system performing this particular kind of inference have an inside", a question predictive processing addresses but does not close.

IIT: mathematical details and critiques

IIT has been refined through successive versions (IIT 2.0, 3.0, 4.0). The formal procedure starts from a system in a state, considers every possible partition into two parts, measures how much the system's cause-effect information is reduced by the minimum-information partition, and defines $Φ$ as the irreducible residue. The full cause-effect structure — the constellation of concepts the system specifies — is identified with the quality of experience, and the system's $Φ$ with its quantity of consciousness ("qualia space", the unfolded cause-effect structure) ^{[source pending]}. The exclusion axiom states that only the maximum- $Φ$ structure exists from the intrinsic perspective, a strong constraint meant to rule out overlapping conscious entities within one substrate.

The formal apparatus has drawn heavy criticism. Scott Aaronson's "Why I Am Not An Integrated Information Theorist" showed that expander graphs score very high on $Φ$ . Cerullo's "The Integrated Information Theory of Consciousness: A Case of Mistaken Identity" argues that IIT conflates an identity claim with an empirical measure, so the theory risks being unfalsifiable. The computation of $Φ$ is exponentially intractable for realistic brains, forcing proponents to use proxies. Tegmark's "Consciousness as a State of Matter" attempts to extend IIT into physics, and the theory's panpsychist implications connect it to Russellian monism (§20.06.02) and to mathematically formulated panpsychism. Whether IIT explains consciousness or redescribes it in intractable formal terms is the standing dispute.

GWT empirical signatures

Dehaene and Naccache ("Towards a Cognitive Neuroscience of Consciousness") catalogued the signatures of conscious access: a nonlinear, all-or-none ignition across frontoparietal cortex; the late, distributed P3b wave in EEG; sustained activity in a global neuronal workspace; and the amplification of high-frequency gamma. Non-conscious processing, by contrast, is local and transient — masked primes activate sensory cortex briefly without igniting the workspace ^{[source pending]}. These signatures are probed through three canonical paradigms: masking (a target rendered invisible by flanking patterns), the attentional blink (a second target missed when it follows a first within a few hundred milliseconds), and inattentional blindness (failure to perceive unattended stimuli). The signatures are robust but contested: critics argue the P3b reflects post-perceptual report rather than consciousness itself, motivating "no-report" variants that decouple ignition from behavioural response.

HOT empirical signatures and the prefrontal question

Lau and Rosenthal's "Empirical Support for Higher-Order Theories of Consciousness" ^{[source pending]} argues that prefrontal cortex carries the higher-order representations HOT requires: lesions and stimulations of prefrontal areas alter metacognitive access and the presence of awareness, and binocular rivalry and multistable perception dissociate the stimulus from the conscious percept in ways higher-order monitoring can explain. The theory connects to the developmental literature on theory of mind and metacognition in children (Flavell; see §29.06), since the machinery for representing one's own states overlaps with that for representing others'. The recurring obstacle is the same prefrontal dispute that divides the field: if prefrontal activity reflects report and control rather than experience, HOT's mechanistic claim weakens, and the theory must specify what distinguishes consciousness-conferring monitoring from mere metacognition.

Neural correlates of consciousness

Crick and Koch ("Towards a Neurobiological Theory of Consciousness") defined the neural correlates of consciousness (NCC) as the minimal set of neural events jointly sufficient for a specific conscious experience. The search has converged on a frontoparietal network but split on whether the posterior hot zone (temporal, parietal, and occipital cortex) is sufficient, with prefrontal involvement reflecting report and control rather than experience (Koch, The Feeling of Life Itself ^{[source pending]}). Modern tools — CLARITY tissue clearing, connectomics, and dense multi-electrode recording (Allen Institute; Brain Initiative) — promise finer-grained NCC identification. The NCC programme is theory-light: it hunts for correlations without deciding between GWT, IIT, and HOT, so its results can adjudicate among theories only when a correlate is shown to be necessary and sufficient under the theories' competing definitions.

The comparative debate and type-A physicalism

The theories are often read as type-A physicalist programmes (in Chalmers' terminology, §20.06.02): each assumes that fixing the relevant functional or informational facts fixes the phenomenal facts, and that the hard problem will dissolve once the correct function or quantity is identified. This makes them targets for the hard-problem objection that they explain access or integration, not why such processing is accompanied by feel. Roundtables and edited volumes (Cognition, Computation, and Consciousness; the ASSC community) stage these disagreements directly. The adversarial collaboration between IIT and GWT is the most ambitious attempt to let shared data decide: each theory specified predictions in advance, and the results fit each theory on some measures, leaving the competition productive but unresolved. The meta-question is whether any functional or informational theory can close the explanatory gap, or whether they are all, in the end, theories of access rather than of phenomenology.

Animal consciousness

The Cambridge Declaration on Consciousness (2012) and Birch's work on animal sentience (the "Edinburgh Declaration on Animal Consciousness") ^{[source pending]} consolidated the case that many non-human animals are conscious, with direct ethical consequences (Singer, Regan; §20.02.07, §20.05.*). The theories offer different criteria. IIT invites estimates of $Φ$ in octopus, bees, and C. elegans, predicting graded consciousness by integration; GWT asks whether ignition-like global broadcasts occur in bird and corvid brains; HOT-style accounts demand metacognitive capacity that may be absent in many species. The criteria diverge enough that they licence different verdicts on the same animal, which sharpens the ethical stakes: whether a given creature is conscious, and how much, depends on which theory one accepts, and the welfare consequences of that verdict are large (§31.04 biological anthropology; §19.06 speciation).

Consciousness across substrates and AI

Functionalism and IIT give opposite answers to substrate independence. Functionalism (the default of cognitive science, §20.06.02) holds that consciousness is multiply realisable: any substrate implementing the right functional organisation is conscious. IIT holds that consciousness is substrate-dependent, because $Φ$ depends on the physical cause-effect structure, and standard feedforward artificial neural networks have $Φ = 0$ regardless of their behavioural sophistication. The question of AI consciousness (Chalmers, "The Singularity: A Philosophical Analysis"; Schneider, Artificial You; §20.02.06) therefore splits: a sufficiently complex functional architecture might be conscious on GWT or HOT, while a feedforward deep network would not be conscious on IIT. Searle's Chinese Room (§20.06.01) presses that passing behavioural tests is not enough, and recurrent or neuromorphic architectures are the live candidates for artificial consciousness. Hybrid systems — neural prostheses interfaced with biological brains (§35.08) — raise the further puzzle of how consciousness extends across mixed substrates.

Disorders of consciousness

Clinical states — vegetative state, minimally conscious state, coma, and brain death — supply a severe test for any theory of consciousness (§29.10, §35.05). Adrian Owen's "Detecting Awareness in the Vegetative State" used fMRI to show that some behaviourally unresponsive patients can imagine playing tennis on command, demonstrating covert awareness undetectable at the bedside ^{[source pending]}. The theories reframe this finding: GWT predicts that covertly aware patients retain a partial workspace capable of supporting the imagined-tennis task; IIT predicts that their posterior integration is preserved even when overt behaviour and global report are lost. Both theories aspire to turn $Φ$ -style measures and ignition markers into diagnostics that distinguish vegetative from minimally conscious patients — a clinical application where getting the theory right carries direct consequences for end-of-life decisions.

Psychedelic states

Carhart-Harris and Friston's REBUS model ("REBUS and the anarchic brain") ^{[source pending]} treats psychedelics as relaxing the brain's high-level priors, increasing entropy and widening the range of predictions the system entertains. Read through IIT, psychedelic states involve altered integration (with some measures suggesting increased $Φ$ ); read through GWT, they involve reduced workspace efficiency and weakened top-down control. The free-energy framework unifies these descriptions by treating psychedelics as agents that reduce precision on high-level priors, opening the system to bottom-up prediction error. These models connect to therapeutic applications (psilocybin for depression, MDMA-assisted therapy; §29.10.03) and to the broader study of altered states — flow, peak experiences (Maslow), and contemplative or meditative states (§20.10 Eastern philosophy, §29.06 adult development) — where the standing question is whether any theory can predict what an altered state feels like from its neural or informational structure.

Connections Master

Consciousness, the hard problem, and qualia 20.06.01 and the hard problem 20.06.02 pending are the direct prerequisites. They supply the vocabulary (qualia, the explanatory gap, type-A/type-B physicalism) against which these scientific theories are measured; every theory here is, among other things, a response to the hard problem.
The measurement problem in quantum mechanics 20.03.01 connects through the Wigner–von Neumann strand in which consciousness causes wave-function collapse; if a scientific theory fixes what consciousness is, it constrains whether such a collapse role is even coherent.
AI ethics and alignment 20.02.06 and consequentialism and animal ethics 20.02.07 pending connect through moral status: whether a machine or an animal is conscious depends on which theory one accepts, and that verdict underwrites the ethics of treatment, of welfare, and of switching systems off.
Philosophy of biology [20.05.*] connects through the evolution and distribution of consciousness across species, and through whether consciousness has an adaptive function or is, as the zombie argument suggests, functionally idle.
Neuroscience [29.02], sensation and perception [29.03], cognitive development [29.06], and emotion and motivation [29.11] supply the empirical substrates — frontoparietal networks, masking and rivalry paradigms, the development of metacognition, and interoceptive inference — that the theories attempt to unify.
Therapy and biological treatments [29.10], especially psychedelic-assisted therapy 29.10.03 pending, connect through REBUS and the clinical application of integration and ignition markers.
Eastern philosophy [20.10 Confucianism, 20.11 Buddhism, 20.12 Advaita Vedanta, 20.13 Daoism] offers first-person phenomenologies of awareness and meditative states that resist the third-person framing the scientific theories presuppose, and that supply data on altered states the theories must eventually accommodate.

Cross-domain to the philosophy of language [§22]: the private-language argument and the ineffability of qualia bear on whether workspace broadcast, $Φ$ , or higher-order monitoring could ever be communicated from the inside.

Historical and philosophical context Master

The scientific study of consciousness is younger than the philosophical one. Through most of the twentieth century, behaviourism and early cognitive science treated consciousness as unfit for scientific study; the word was largely avoided in respectable neuroscience. The change began in the 1980s. Bernard Baars' A Cognitive Theory of Consciousness (1988) ^{[source pending]} introduced Global Workspace Theory and made consciousness a legitimate target for cognitive modelling by redescribing it as global broadcast among otherwise unconscious specialists. Gerald Edelman and Antonio Damasio developed parallel neurobiological theories of reentry and of the self, broadening the field.

The 1990s brought two decisive moves. First, Francis Crick and Christof Koch ("Towards a Neurobiological Theory of Consciousness", 1990) proposed hunting for the neural correlates of consciousness, giving the field an empirical programme independent of any single theory. Second, the hard-problem literature (Levine 1983; Chalmers 1995, 1996; §20.06.02) sharpened what any scientific theory would have to explain, raising the bar that GWT and its rivals had to meet.

Higher-order theories descend from the seventeenth-century intuition that consciousness involves a perception of one's own mental activity, but their modern form is due to David Rosenthal ("Two Concepts of Consciousness", 1986; Consciousness and Mind, 2005 ^{[source pending]}). Hakwan Lau and Rosenthal later assembled the empirical case for higher-order representations and tied the theory to prefrontal cortex, while Peter Carruthers developed dispositional variants. Victor Lamme's recurrent processing theory (2000s) and Michael Graziano's attention schema theory (Consciousness and the Social Brain, 2013) ^{[source pending]} added structural and modelling-based alternatives, both locating consciousness differently from the frontoparietal emphasis of GWT.

Integrated Information Theory arrived in 2004 with Giulio Tononi's "An Information Integration Theory of Consciousness", refined through IIT 2.0, 3.0, and 4.0, and given a popular form in Phi: A Voyage from the Brain to the Soul (2012) ^{[source pending]}. IIT broke with the functionalist mainstream by defining consciousness ontologically, and its panpsychist implications and its intractability provoked a literature of critiques (Aaronson 2014; Cerullo 2015) that remains active. Predictive processing, growing out of Friston's free-energy principle and developed by Clark, Hohwy, and Seth, supplied a unifying computational language in which ignition, integration, and higher-order monitoring can all be partially recast.

The contemporary field is institutionally visible through the Association for the Scientific Study of Consciousness (ASSC) and empirically ambitious through the Templeton-funded adversarial collaboration (2019–2021) that pitted IIT against GWT on shared data. The collaboration produced no winner, and the deeper philosophical question — whether any theory of function or information can dissolve the hard problem, or whether they all remain theories of access — is unresolved. Animal-consciousness declarations (Cambridge 2012; the New York Declaration on Animal Consciousness 2024) and clinical applications to disorders of consciousness have pushed the theories out of the seminar room and into ethics and medicine, where their competing verdicts have real consequences.

Bibliography Master

Baars, B. J. — A Cognitive Theory of Consciousness (Cambridge University Press, 1988).
Dehaene, S. — Consciousness and the Brain: Deciphering How the Brain Codes Our Thoughts (Viking, 2014).
Dehaene, S. and Naccache, L. — "Towards a cognitive neuroscience of consciousness: basic evidence and a workspace framework", Cognition 79(1-2), 1-37 (2001).
Dehaene, S. and Changeux, J.-P. — "Experimental and theoretical approaches to conscious processing", Neuron 70(2), 200-227 (2011).
Tononi, G. — "An information integration theory of consciousness", BMC Neuroscience 5, 42 (2004).
Tononi, G. — "Integrated information theory of consciousness: an updated account", Archives Italiennes de Biologie 150(2-3), 56-90 (2012).
Tononi, G. — Phi: A Voyage from the Brain to the Soul (Pantheon, 2012).
Oizumi, M., Albantakis, L., and Tononi, G. — "From the phenomenology to the mechanisms of consciousness: Integrated Information Theory 3.0", PLOS Computational Biology 10(5), e1003588 (2014).
Rosenthal, D. M. — "Two concepts of consciousness", Philosophical Studies 49(3), 329-359 (1986).
Rosenthal, D. M. — Consciousness and Mind (Oxford University Press, 2005).
Lau, H. and Rosenthal, D. M. — "Empirical support for higher-order theories of conscious awareness", Trends in Cognitive Sciences 15(8), 365-373 (2011).
Carruthers, P. — Phenomenal Consciousness: A Naturalistic Theory (Cambridge University Press, 2000).
Weisberg, J. — "Abandoning the factives: aboutness and Horor", Philosophical Issues 23(1), 316-333 (2013).
Neander, K. — "The second-order view: a defence of the perceptual model of introspection", manuscript (2017).
Lamme, V. A. F. — "Why visual attention and awareness are different", Trends in Cognitive Sciences 7(1), 12-18 (2003).
Lamme, V. A. F. — "Towards a true neural stance on consciousness", Trends in Cognitive Sciences 10(11), 494-501 (2006).
Graziano, M. S. A. — Consciousness and the Social Brain (Oxford University Press, 2013).
Friston, K. — "The free-energy principle: a unified brain theory?", Nature Reviews Neuroscience 11(2), 127-138 (2010).
Clark, A. — Surfing Uncertainty: Prediction, Action, and the Embodied Mind (Oxford University Press, 2016).
Hohwy, J. — The Predictive Mind (Oxford University Press, 2013).
Seth, A. — Being You: A New Science of Consciousness (Faber, 2021).
Barrett, L. F. — How Emotions Are Made: The Secret Life of the Brain (Houghton Mifflin Harcourt, 2017).
Hobson, J. A. and Friston, K. — "A general theory of consciousness: an update", Journal of Consciousness Studies 28(9-10), 109-126 (2021).
Carhart-Harris, R. L. and Friston, K. — "REBUS and the anarchic brain: toward a unified model of the brain action of psychedelics", Pharmacological Reviews 71(3), 316-344 (2019).
Crick, F. and Koch, C. — "Towards a neurobiological theory of consciousness", Seminars in the Neurosciences 2, 263-275 (1990).
Koch, C. — The Feeling of Life Itself: Why Consciousness Is Widespread but Can't Be Computed (MIT Press, 2019).
Tsuchiya, N. and Koch, C. — "The relationship between consciousness and top-down attention", Trends in Cognitive Sciences 19(9), 505-513 (2015).
Aaronson, S. — "Why I am not an integrated information theorist (or, the unconscious expander)", Shtetl-Optimized blog, 21 May 2014.
Cerullo, M. A. — "The integrated information theory of consciousness: a case of mistaken identity", Journal of Consciousness Studies 22(7-8), 188-208 (2015).
Tegmark, M. — "Consciousness as a state of matter", Chaos, Solitons & Fractals 76, 238-270 (2015).
Owen, A. M. et al. — "Detecting awareness in the vegetative state", Science 313(5792), 1402 (2006).
Melloni, L. et al. — "An adversarial collaboration protocol for testing contrasting predictions of global neuronal workspace and integrated information theory", PLOS ONE 18(2), e0268577 (2023).
Birch, J. — "Animal sentience and the precautionary principle", Animal Sentience 2(16) (2017).
Chalmers, D. J. — "The singularity: a philosophical analysis", Journal of Consciousness Studies 17(9-10), 7-65 (2010).
Schneider, S. — Artificial You: AI and the Future of Your Mind (Princeton University Press, 2019).