Timeline of integrated information theory

This is a timeline of integrated information theory, a theory of consciousness first formulated by neuroscientist Giulio Tononi in 2004, which holds that consciousness corresponds to integrated information, quantified as Φ (phi). The timeline traces the theory's roots in philosophy and neuroscience, its successive versions, clinical measures it inspired such as the perturbational complexity index, adversarial collaborations testing it against rival theories, and the critiques and controversies surrounding it, including the 2023 "pseudoscience" letter.

Sample questions

The following are some interesting questions that can be answered by reading this timeline:

• What are some important ideas and developments that preceded integrated information theory?
  • Sort the full timeline by "Event type" and look for the group of rows with value "Prelude".
  • You will see earlier work in philosophy, information theory and neuroscience that the theory would later build on.
• How has integrated information theory been formulated and developed over time?
  • Sort the full timeline by "Event type" and look for the group of rows with value "Theory".
  • You will see the founding paper, the successive versions of the theory, and its extensions to new aspects of experience and to related questions.
• How have researchers tried to measure integrated information and consciousness in practice?
  • Sort the full timeline by "Event type" and look for the group of rows with value "Measure".
  • You will see clinical measures inspired by the theory, alternative ways of calculating integrated information, and tools for computing it.
• What experiments have been conducted in connection with the theory?
  • Sort the full timeline by "Event type" and look for the group of rows with value "Experiment".
  • You will see studies in humans and other systems that test the theory's ideas or apply its measures in different states of consciousness.
• What are the main criticisms of integrated information theory?
  • Sort the full timeline by "Event type" and look for the group of rows with value "Critique".
  • You will see objections from philosophers, neuroscientists, physicists and computer scientists, concerning the theory's foundations, mathematics, testability and scientific status.
• How have proponents of the theory replied to its critics?
  • Sort the full timeline by "Event type" and look for the group of rows with value "Response".
  • You will see replies to major objections, defenses of the theory by philosophers and scientists, and reactions to public controversies about it.
• How has the theory been tested against rival theories of consciousness?
  • Sort the full timeline by "Event type" and look for the group of rows with value "Adversarial collaboration".
  • You will see collaborations in which proponents of competing theories agree on experiments in advance, from their planning to their protocols and results.
• What are some notable books, reviews and articles about the theory?
  • Sort the full timeline by "Event type" and look for the group of rows with value "Publication".
  • You will see popular books, press coverage, review articles and editorials that shaped how the theory was presented and received.
• Other events are described under the type "Event".

Big picture

Time period	Development summary	More details
Before 2004	Intellectual roots	Ideas that IIT would later combine develop separately. Gottfried Wilhelm Leibniz's mill argument (1714) frames the puzzle of how mechanism could yield experience. Claude Shannon's information theory (1948) supplies a mathematics of information. In the 1990s, Francis Crick and Christof Koch make the search for neural correlates of consciousness a respectable experimental program, and David Chalmers names the "hard problem". Working with Gerald Edelman, Giulio Tononi develops measures of neural complexity that balance integration and differentiation, and the "dynamic core" hypothesis. In 2003, Tononi and Olaf Sporns define the measure Φ.
2004–2012	Formulation and first evidence	Tononi's 2004 paper founds the theory, holding that consciousness is integrated information. The 2008 reformulation with David Balduzzi and Tononi's "provisional manifesto" give it the name IIT and a geometric account of qualia. In parallel, Marcello Massimini and colleagues use transcranial magnetic stimulation with EEG to show that the brain's capacity for integrated yet differentiated responses breaks down in sleep and anesthesia and recovers in minimally conscious patients. The Sussex group proposes practical Φ measures for real data. A New York Times profile, Koch's embrace of the theory, and Tononi's book Phi bring IIT to a wide audience, while Michael Cerullo, Eric Schwitzgebel and John Searle raise early objections.
2013–2018	IIT 3.0, clinical measures and growing scrutiny	The perturbational complexity index (2013) turns the TMS-EEG findings into a single number that separates conscious from unconscious states, later validated in unresponsive patients. IIT 3.0 (2014) recasts the theory around axioms, postulates and cause–effect structures, and its proponents locate consciousness in a posterior cortical "hot zone". Critics multiply: Scott Aaronson's "unconscious expander", Adam Barrett's call for a field-based reformulation, and philosophical attacks on the axioms and on what the theory explains. Philosophers also begin to defend it, answering Searle and arguing that IIT's information is not Shannon's. Rival Φ measures prove to behave very differently, the PyPhi toolbox appears, and the Templeton World Charity Foundation launches its adversarial collaborations.
2019–2022	Falsifiability and foundations debates	The "unfolding argument" (2019) contends that causal structure theories like IIT are either false or unfalsifiable, prompting a prolonged exchange of replies. Further critiques argue that Φ is not well defined for real physical systems, question whether consciousness can be intrinsic or whether IIT states a clear theory at all, and propose separating a testable "weak" IIT from the "strong" theory. Physicists extend IIT's measures to very large networks, at the cost of revising its assumptions. A Behavioral and Brain Sciences target article calls IIT "a case of mistaken identity", and the IIT group answers. Meanwhile proponents extend the theory to the feel of space and to causation, and adversarial collaborations against rival theories get under way.
2023–present	Adversarial tests and open controversy	The first adversarial collaboration reports in June 2023 that neither IIT nor the global neuronal workspace theory is fully confirmed. In September, 124 researchers call IIT "pseudoscience", setting off a public dispute that continues in Nature Neuroscience in 2025; Nature's editors rebuke the label when the full results appear. IIT 4.0 reformulates the theory with an explicitly idealist ontology. IIT's proponents argue that conventional computers lack the causal structure for consciousness. New adversarial collaborations publish protocols testing IIT against predictive processing theories, including its prediction that even inactive neurons shape experience, and against the workspace theory in monkeys and mice. By 2026, critics and proponents alike treat IIT as influential but unresolved.

Full timeline

Inclusion criteria

We include:

• Precursor ideas and research that integrated information theory directly builds on, such as Leibniz's mill argument, Shannon's information theory, Edelman's neuronal group selection, the search for neural correlates of consciousness, and Tononi, Sporns and Edelman's measures of neural complexity.
• Papers that formulate, revise or extend the theory, including each numbered version (IIT 2.0, 3.0 and 4.0) and extensions to spatial and temporal experience, causation, free will and artificial intelligence.
• Measures and tools derived from or inspired by the theory, such as Φ and its rival formulations, the perturbational complexity index and its clinical validation, and the PyPhi toolbox.
• Experiments that test the theory's predictions or apply its measures, in humans, animals or artificial systems.
• Critiques published by philosophers, neuroscientists, physicists, mathematicians and computer scientists, together with the replies of the theory's proponents, so that each major objection is shown alongside its answer.
• Adversarial collaborations that test IIT against rival theories, from funding and workshops to preregistered protocols, results and joint reviews.
• Books, review articles and press coverage that significantly shaped how the theory was presented to scientists or the public.
• Notable events in the field when they bear directly on IIT, such as dedicated workshops and public controversies.
• Preprints when they mark the first public appearance of a work, dated to the preprint, with later journal publication noted in the row.

We do not include:

• Work on consciousness that does not engage with integrated information theory, even when it involves the same researchers (for example, sleep or anesthesia research with no bearing on the theory).
• Routine follow-up papers, minor technical refinements, conference talks and reprints that add nothing substantial to rows already in the timeline.
• Secondary sources when a primary source is available; news articles and reviews are used for reception and context, or when no primary source can be accessed.
• Wikipedia, other living wikis, search-engine-optimized summary sites, AI-generated pages and predatory publishers.
• Non-peer-reviewed commentary with little influence, such as personal blog reviews and low-profile preprints, unless it became a reference point in the debate (as with Scott Aaronson's 2014 blog posts).
• Duplicate coverage of the same event; a single row covers each paper, with its reception and later developments described inside that row.

Year (month and date)	Event type	Details
1720	Prelude	Gottfried Wilhelm Leibniz's Monadology appears in print, containing his famous analogy of the mill. Leibniz asks the reader to imagine walking through an enlarged brain as one would walk through a mill, observing all its mechanical operations; nowhere, he contends, would such an observer see a conscious thought, so consciousness cannot arise from mere matter. In his earlier Discourse on Metaphysics (1686), Leibniz had already allowed for infinitely many degrees of consciousness, including unconscious "petites perceptions", and he was the first to distinguish explicitly between perception and apperception (roughly, awareness and self-awareness). Both ideas would resurface in integrated information theory nearly three centuries later. IIT treats consciousness as graded, and in 2015 Giulio Tononi and Christof Koch would cite the mill, alongside John Searle's Chinese Room argument, as anticipating their claim that a computer simulation of a brain would not be conscious.
1948 (July–October)	Prelude	Claude Shannon, a mathematician at Bell Telephone Laboratories, publishes "A Mathematical Theory of Communication" in two installments of the Bell System Technical Journal. Building on earlier work by Harry Nyquist and Ralph Hartley, Shannon defines the fundamental problem of communication as reproducing at one point a message selected at another, and sets aside the meaning of messages as irrelevant to the engineering problem. He adopts a logarithmic measure of information, with units called bits (a word suggested by John Tukey), and introduces the entropy H as a measure of information, choice, and uncertainty. He shows that the entropy of a joint event is at most the sum of the individual entropies, with equality only when the events are independent, and proves fundamental limits on encoding and on channel capacity in the presence of noise. Shannon's statistical framework would supply the mathematical vocabulary for later measures of neural integration, while his exclusion of meaning would become a foil for integrated information theory, which defines information from the perspective of the system itself rather than an external observer.
1987	Prelude	Neuroscientist Gerald Edelman, a 1972 Nobel Prize in Physiology or Medicine laureate, publishes Neural Darwinism: The Theory of Neuronal Group Selection through Basic Books. Its central idea is that each individual's nervous system works as a selective system, resembling natural selection in evolution but operating through different mechanisms. By giving the categorization of things in the world a neural basis, the theory aims to unify perception, action, and learning. It also recasts memory as a dynamic process of recategorization rather than a store that replicates attributes, with implications for psychological states ranging from attention to dreaming. The book draws on developmental and evolutionary biology, anatomy, physiology, ethology, and psychology. Its publisher expects it to renew debate on philosophical questions such as the mind–body problem, the origins of knowledge, and the perceptual basis of language. The theory of neural Darwinism would become the framework within which Edelman and his later collaborator Giulio Tononi study neural integration and consciousness.
1990	Prelude	Francis Crick and Christof Koch publish "Towards a neurobiological theory of consciousness" in Seminars in the Neurosciences, helping to make consciousness a topic for experimental neuroscience. They argue that visual awareness is "a favorable form of consciousness to study neurobiologically" and propose that it takes two forms: a very fast one linked to iconic memory, and a slower one involving attention and short-term memory. In the slower form, an attentional mechanism transiently binds together the neurons that respond to the features of a single object, which they suggest happens through "coherent semi-synchronous oscillations, probably in the 40-70 Hz range" that then activate working memory. The program would shape the search for the neural correlates of consciousness through the 1990s. Giulio Tononi would later reject synchrony as the basis of consciousness, noting that epileptic seizures make brain activity more synchronized yet cause unconsciousness. Koch himself would embrace integrated information theory in 2012.
1994 (May)	Prelude	Neuroscientists Giulio Tononi, Olaf Sporns, and Gerald Edelman (the latter a Nobel Prize in Physiology or Medicine laureate) publish "A measure for brain complexity: relating functional segregation and integration in the nervous system" in the Proceedings of the National Academy of Sciences. The paper starts from a tension in the brains of higher vertebrates: local areas are anatomically and physiologically specialized, yet they work as an integrated whole during perception and behavior. To capture this interplay, the authors introduce a measure called neural complexity (CN), built from how far subsets of a neural system deviate from statistical independence. CN is high when segregation and integration coexist, and low when components are either fully independent or fully dependent. Computer simulations of cortical areas show that typical cortical wiring (dense connections, strong local connectivity forming neuronal groups, patchy links between groups, and prevalent reciprocal connections) yields high CN. The authors suggest the approach may also apply to gene regulation and embryogenesis. The idea that the brain balances differentiation with integration would become a direct precursor of Tononi's later integrated information theory.
1995	Prelude	Philosopher David Chalmers publishes "Facing Up to the Problem of Consciousness" in the Journal of Consciousness Studies. He argues that progress requires confronting consciousness directly, and isolates its truly hard problem (explaining subjective experience itself) from more tractable problems about cognitive functions and mechanisms. Chalmers critiques recent work that addresses consciousness with reductive methods, arguing that such methods inevitably fail to come to grips with the hardest part of the problem. In the second half of the paper, he proposes that a naturalistic account of consciousness is still possible through a new kind of nonreductive explanation. His candidate is a theory built on principles of structural coherence and organizational invariance, together with a double-aspect theory of information, in which information has both a physical and a phenomenal side. The paper would be reprinted in the 1996 Tucson volume Toward a Science of Consciousness and in Chalmers' 2010 collection The Character of Consciousness. The hard problem framing would become the backdrop against which integrated information theory presents itself, starting from the properties of experience rather than from brain mechanisms.
1996 (April)	Prelude	At the second Tucson conference on consciousness, philosopher David Chalmers gives a talk titled "On the Search for the Neural Correlate of Consciousness", published in 1998 in the MIT Press volume Toward a Science of Consciousness II. Chalmers observes that the field has produced a "neural correlate zoo" of competing proposals for the neural correlate of consciousness (NCC). Among them are Francis Crick and Christof Koch's 40-hertz cortical oscillations, Gerald Edelman's reentrant thalamocortical loops, and Bernard Baars's global workspace. He argues that the search faces a foundational obstacle: consciousness cannot be measured directly, and for principled reasons no consciousness meter seems possible. Researchers therefore rely on pre-experimental bridging principles, chiefly verbal report, to judge whether a system is conscious and what it is conscious of. Since these principles are not themselves tested experimentally, Chalmers contends that everyone searching for the NCC is engaged in philosophical reasoning. The absence of a principled measure of consciousness is the gap that integrated information theory would later try to fill, by proposing a quantity to be computed from a system's physical structure.
1996 (April 16)	Prelude	Giulio Tononi, Olaf Sporns and Gerald Edelman publish "A complexity measure for selective matching of signals by the brain" in the Proceedings of the National Academy of Sciences. It builds on their 1994 measure of neural complexity (CN), which quantifies the amount and heterogeneity of statistical correlations within a neural system in terms of the mutual information between subsets of its units. Because CN had been applied only to systems isolated from the environment, the authors introduce matching complexity (CM), the change in CN after a system receives sensory signals. CM measures "how well the ensemble of intrinsic correlations within a neural system fits the statistical structure of the sensory input". In simulations, CM is low when a cortical area's intrinsic connectivity is random and high when that connectivity amplifies the correlations sensory input enhances. A small amount of incoming information can thus trigger "a large increase in the mutual information between many functionally specialized subsets within the brain". The emphasis on a system's intrinsic correlations, rather than on its input, would carry into the dynamic core hypothesis of 1998.
1998 (December)	Prelude	Giulio Tononi and Gerald Edelman publish "Consciousness and complexity", a review article in Science. They contrast their approach with conventional research on consciousness, which typically looks for the contribution of specific brain areas or groups of neurons. Instead, they ask what kinds of neural processes could account for key properties of conscious experience. They apply measures of neural integration and complexity (extending the neural complexity measure the two had introduced with Olaf Sporns in 1994) and analyze extensive neurological data. This leads them to a testable proposal about the neural substrate of consciousness, which they call the dynamic core hypothesis. The paper shifts the question from where consciousness sits in the brain to what kind of neural process supports it, a framing Tononi would later develop into integrated information theory.
2000	Prelude	Neuroscientists Gerald Edelman, director of the Neurosciences Institute and chairman of the Department of Neurobiology at the Scripps Research Institute, and Giulio Tononi, a senior fellow in theoretical and experimental neurobiology at the same institute, publish the book A Universe of Consciousness: How Matter Becomes Imagination through Basic Books. Aimed at a general readership, the book asks why physical events inside a small mass of brain tissue give rise to the world of conscious experience. Its publisher presents it as the first full-scale theory of consciousness grounded in direct observation of the working human brain, one that challenges much of the conventional wisdom on the subject and bears on language, thought, emotion, and mental illness. The book brings the pair's research program on neural integration and complexity to a broad audience, a few years before Tononi would publish the first formal statement of integrated information theory.
2003 (December 2)	Theory	Giulio Tononi and Olaf Sporns publish "Measuring information integration" in BMC Neuroscience, setting out the measure Φ in detail. They note that communication theory measures how information is transmitted from a sender to a receiver, but not how a distributed network like the brain integrates it. To fill that gap, they introduce effective information: the causal effect one part of a system has on another when the first part's outputs are replaced by noise. Φ is the effective information across a subset's weakest bipartition. A complex is a subset with Φ above zero that is not contained in a subset of higher Φ. Simulations of small networks show that Φ is highest when elements are both specialized and densely interconnected, and low when networks are homogeneous or modular. The authors argue that this explains why the thalamocortical system, unlike the modular cerebellum, supports consciousness, and they suggest that basal ganglia loops may underlie unconscious processes. They acknowledge that an exhaustive search is feasible only for about two dozen elements, and they propose probing real brains with transcranial magnetic stimulation. IIT 4.0 would later count this paper, with Tononi's 2004 paper, as IIT 1.0.
2004 (November 2)	Theory	Neuroscientist Giulio Tononi, of the Department of Psychiatry at the University of Wisconsin, publishes "An information integration theory of consciousness" in BMC Neuroscience, the founding paper of what would become integrated information theory (IIT). Starting from two phenomenological properties of experience, differentiation (a vast repertoire of possible experiences) and integration (the unity of each one), Tononi proposes that consciousness corresponds to a system's capacity to integrate information. Thought experiments contrast a person with a photodiode, which gains a single bit when it tells light from dark, and with a megapixel camera, whose sensors distinguish countless states without interacting. He introduces Φ, the effective information across a subset's informational "weakest link", and defines a complex as a subset with Φ above zero that is not part of a larger subset of higher Φ. The quality of experience is set by the informational relationships within the complex. Tononi uses the theory to explain why the thalamocortical system, unlike the cerebellum, supports consciousness, and why consciousness fades in dreamless sleep. He concludes that consciousness is a fundamental, graded quantity, present in infants and animals, and that conscious artifacts should be possible to build.
2005 (September 30)	Experiment	Marcello Massimini, Giulio Tononi, and four colleagues publish "Breakdown of cortical effective connectivity during sleep" in Science, asking why consciousness fades in sleep even though the brain stays active. Combining transcranial magnetic stimulation (TMS) with high-density electroencephalography (EEG), they stimulate the premotor cortex and track how the activation spreads. During quiet wakefulness, an initial response is followed by waves that travel to connected cortical areas several centimeters away. During non-rapid eye movement sleep, the initial response is stronger but quickly dies out without spreading. The authors conclude that the fading of consciousness in some sleep stages may reflect a breakdown in cortical effective connectivity. Tononi's 2004 paper had announced that such TMS–EEG experiments were underway. In a note, the authors argue that gamma activity and synchrony, proposed as correlates of consciousness by Francis Crick and Christof Koch among others, cannot account for the change: they are equally low in REM sleep, when experience is usually vivid, and can be high during anesthesia. IIT's proponents would later cite the study among evidence that complexity measures inspired by the theory can distinguish conscious from unconscious states.
2007 (May 15)	Experiment	Marcello Massimini, Fabio Ferrarelli, Steve Esser, Giulio Tononi and colleagues report in the Proceedings of the National Academy of Sciences of the United States of America that transcranial magnetic stimulation can trigger sleep slow waves noninvasively and reliably. In sleeping subjects, each pulse over sensorimotor cortex evokes a high-amplitude slow wave that "originates under the coil and spreads over the cortex". The stimulation deepens sleep from stage 2 to stage 4 and boosts slow-wave activity up to eightfold near the stimulation site. The same pulses fail to evoke slow waves during wakefulness, producing instead a "differentiated pattern of long-range cortical activation". Building on their 2005 finding that effective connectivity breaks down in sleep, the authors conclude that the sleeping brain "loses its ability of entering states that are both integrated and differentiated". Instead, it either "breaks down in independent modules" or "bursts into an explosive and aspecific response". Citing Tononi's integrated information theory, they suggest both features may account for the fading of consciousness early in sleep. They also propose practical uses, such as enhancing short naps or treating disturbed sleep.
2008 (June 1)	Publication	Christof Koch and Giulio Tononi publish "Can Machines Be Conscious?" in IEEE Spectrum, as part of a special report on the singularity. Their answer is "Yes", though "perhaps not in the way that the most popular scenarios have envisioned it". Drawing on clinical cases, they argue that consciousness requires neither emotion, memory, attention, self-reflection, language, nor interaction with the environment. What matters instead is integrated information. A one-megapixel camera can enter an enormous number of states, but its photodiodes are independent, whereas a conscious brain is "a single integrated entity with a large repertoire of states". They describe Φ, measured in bits, as "an intrinsic property of the system", different from Shannon information. They propose a Turing test for consciousness in which a machine must grasp the gist of any scene, such as recognizing a still of a man pointing a gun in a liquor store as a robbery. They are skeptical that simulating the brain will produce a conscious machine. In 2015 the pair would argue that a computer simulation of a brain would not be conscious.
2008 (June 13)	Theory	Giulio Tononi and David Balduzzi publish "Integrated Information in Discrete Dynamical Systems: Motivation and Theoretical Framework" in PLOS Computational Biology. Earlier versions of the theory measured integrated information only for stationary systems modeled as Gaussian processes, leaving out time and changing dynamics. The paper introduces a time- and state-dependent measure, φ, of how much information a system generates as a whole when it enters a particular state through causal interactions among its elements, above and beyond what its parts generate independently. In small simulated networks, φ peaks in balanced states and collapses when a network is inactive or hyperactive, which the authors liken to coma and generalized seizures. A system that merely replays an activity sequence without interacting generates no integrated information, while modular, homogeneous, feedforward, and lattice architectures prove poor or inefficient integrators. The authors describe φ as an intrinsic, observer-independent property, contrasting it with the traditional notion of information sent from a source to a receiver. They also acknowledge that the combinatorial explosion of possible partitions makes computing φ for the human brain unfeasible, a limitation later critics would press.
2008 (December)	Theory	Giulio Tononi publishes "Consciousness as Integrated Information: A Provisional Manifesto" in The Biological Bulletin, setting out the framework explicitly as integrated information theory (IIT). As in his 2004 paper, the theory starts from phenomenology and thought experiments and claims that consciousness is integrated information. The quantity of consciousness corresponds to the integrated information (Φ) generated by a complex of elements, defined as the information the complex generates above and beyond its parts. The quality of experience is specified by the set of informational relationships within the complex. To capture this, Tononi introduces qualia space (Q), in which each axis is a possible state of the complex and the informational relationships generated by its causal mechanisms form a shape that completely specifies a particular experience. The manifesto argues that the framework explains why consciousness depends on some neural systems and not others, why it fades in dreamless sleep and generalized seizures, and how different cortical architectures shape the quality of experience. It closes by noting that equating consciousness with integrated information has broader implications for our view of nature.
2009 (June 17)	Measure	Mario Rosanova, Marcello Massimini and colleagues in Milan publish "Natural frequencies of human corticothalamic circuits" in the Journal of Neuroscience. Using transcranial magnetic stimulation with high-density EEG, the same perturbational approach applied to sleep in 2005, they find that stimulated occipital, parietal and frontal areas each oscillate at a characteristic "natural frequency" (alpha, beta and fast beta/gamma respectively), and that each area tends to keep its own rate even when engaged indirectly through brain connections. The authors argue that the method can reach any cortical area, including in deafferented or locked-in patients, and could "probe the state of human thalamocortical circuits at the patient's bedside". Casali, a co-author, would lead the 2013 study introducing the perturbational complexity index.
2009 (August 14)	Theory	Giulio Tononi and David Balduzzi publish "Qualia: the geometry of integrated information" in PLOS Computational Biology, developing the qualia space (Q) from Tononi's 2008 manifesto into a formal framework for the quality of experience. Q has an axis for each possible activity state of a complex. Each submechanism of the complex specifies a point in Q (a repertoire of system states), and arrows between these points define informational relationships. Together the arrows form a quale, a shape that completely characterizes the quality of a conscious experience, while Φ corresponds to the height of that shape. A new quantity, entanglement, measures how irreducible informational relationships are to their components and is used to specify concepts and modes. Among the corollaries: two systems with identical activity patterns can generate different qualia; inactive elements contribute to experience but inactivated ones do not; and modalities and submodalities correspond to subshapes in Q. Specific qualities such as the "redness" of red are generated by a local mechanism but cannot be reduced to it. The authors suggest the framework may offer a principled way to translate qualitative properties of experience into mathematics.
2010 (January 25)	Experiment	Fabio Ferrarelli, Marcello Massimini, Giulio Tononi and colleagues at the University of Wisconsin publish "Breakdown in cortical effective connectivity during midazolam-induced loss of consciousness" in the Proceedings of the National Academy of Sciences. Starting from the theoretical claim that consciousness depends on the brain's ability to integrate information, they test whether the breakdown of cortical effective connectivity seen in early sleep in 2005 also occurs under anesthesia. In six subjects rendered unresponsive with midazolam, transcranial magnetic stimulation of the premotor cortex evokes a large but brief local response instead of the activity that, in wakefulness, lasts over 300 milliseconds and spreads to distant areas. An index of how far evoked currents propagate, significant current scattering, reliably separates consciousness from unconsciousness. The authors contrast their "theoretically motivated" approach with commercial anesthesia monitors such as the bispectral index, which they say "lack theoretical validity", and propose extending it to coma, vegetative and minimally conscious states. The study would be featured in the New York Times later that year, and the group would introduce the perturbational complexity index in 2013.
2010 (September 20)	Publication	Science writer Carl Zimmer profiles Giulio Tononi and integrated information theory in The New York Times under the headline "Sizing Up Consciousness by Its Bits". The article explains phi through Tononi's photodiode and camera examples. A camera's million photodiodes each carry information independently, while consciousness needs "specialists who talk to each other, so they can behave as a whole". It also reports Tononi's argument that IIT explains why epileptic seizures cause unconsciousness: synchrony reduces the brain's possible states and lowers phi. Christof Koch calls IIT "the only really promising fundamental theory of consciousness" but notes "immense practical problems", since phi cannot be computed exactly even for the 302-neuron worm C. elegans. David Chalmers calls it "a useful starting point" that "doesn't solve the problem of consciousness". Zimmer describes experiments in which sedation shortens and simplifies the brain's response to magnetic pulses, and a collaboration with Steven Laureys aimed at a "consciousness index" for unresponsive patients. That line of work would produce the perturbational complexity index in 2013, and Koch would embrace IIT in his 2012 book.
2011 (January 20)	Measure	Adam Barrett and Anil Seth of the University of Sussex publish "Practical Measures of Integrated Information for Time-Series Data" in PLOS Computational Biology. They note that the 2008 version of Φ (which they call ΦDM) is defined only for discrete Markovian systems, "which are unusual in biology", and so can rarely be measured in practice. They introduce two alternatives that can be applied to continuous time-series data without perturbing the system. ΦE measures information against the system's actual past states rather than a hypothetical maximum-entropy repertoire. ΦAR, for non-Gaussian data, is based on autoregressive prediction error. Because the new measures track what a system does rather than what it could do, the authors argue that they imply treating consciousness as a process, a "Jamesian modification" of IIT. They also show that the normalization used to find a system's minimum information bipartition can make Φ unstable and dependent on the choice of coordinates, which "undercuts ascription of physical meaning" to it. Comparisons of competing Φ measures would continue, notably in a 2019 study by Pedro Mediano and colleagues.
2011 (October 13)	Measure	Neuroscientist Anil Seth and colleagues Adam Barrett and Lionel Barnett publish "Causal density and integrated information as measures of conscious level" in Philosophical Transactions of the Royal Society A. They describe an outstanding challenge for neuroscience: developing measures of conscious level that are both theoretically grounded and practically applicable, high in alert wakefulness and low in dreamless sleep, coma, and general anesthesia. The paper reviews measures of dynamical complexity, focusing on causal density and integrated information, which capture in different ways how far a system's dynamics are simultaneously differentiated and integrated. Because conscious scenes share these features, the authors argue, such measures are good candidates for tracking conscious level. New simulations show where the measures agree and where they diverge, and the authors discuss remaining obstacles to applying them to empirical data. The paper treats integrated information as one candidate measure among several rather than as a definition of consciousness. Both the 2013 paper introducing the perturbational complexity index and the 2016 Nature Reviews Neuroscience overview of IIT would cite it.
2011 (October 20)	Experiment	Jeffrey Edlund, Nicolas Chaumont, Arend Hintze, Christof Koch, Giulio Tononi, and evolutionary biologist Christoph Adami publish "Integrated information increases with fitness in the evolution of animats" in PLOS Computational Biology. Noting that the ability to integrate disparate sources of information is a hallmark of biological organisms, yet hard to quantify, they evolve simulated agents, or animats, over thousands of generations to solve a navigation task in a simple maze. Along the evolutionary line of descent, measures of information integration rise together with fitness as the animats adapt. The authors compare these measures with more conventional information-processing measures and conclude that high fitness requires both processing and integration, with integration the better predictor when the task demands memory. They suggest that such measures can quantify an organism's functional complexity even without fitness data. A 2014 follow-up by Larissa Albantakis and colleagues would extend the approach to environments of increasing complexity. In 2015, Tononi and Koch would cite these simulations as IIT's account of why consciousness is adaptive and therefore evolved.
2011	Critique	Michael Cerullo publishes "Integrated Information Theory: A Promising but Ultimately Incomplete Theory of Consciousness" in the Journal of Consciousness Studies, an early critique of Giulio Tononi's theory. He notes that because IIT grounds consciousness in information, quantified with a modified version of Claude Shannon's definition, it allows consciousness to be realized in many kinds of systems. Its "main failings", he argues, are the absence of a link between conscious experience and awareness, and its reliance on "Shannon's limited data based definition of information", which cannot link information to the causal properties of the brain. These limitations, he contends, prevent the theory from satisfying David Chalmers's principles of structural coherence and organizational invariance. It is unclear whether IIT can be repaired "without simply reducing the theory to a general statement of computational functionalism". Cerullo would return with a sharper critique of IIT 3.0 in 2015, and later defenders would argue that IIT's information is not Shannon's.
2012 (January 5)	Experiment	Mario Rosanova, Olivia Gosseries, Mélanie Boly, Giulio Tononi, Steven Laureys, Marcello Massimini and colleagues report in Brain (journal) a bedside test of consciousness in severely brain-injured, non-communicating patients that requires neither intact sensory pathways nor the ability to follow instructions. Building on the premise, which they attribute to theoretical and experimental work, that consciousness requires specialized cortical areas to "engage in rapid causal interactions" (effective connectivity), they combine transcranial magnetic stimulation with high-density EEG. In patients in a vegetative state, who are awake but unresponsive, stimulation triggers "a simple, local response", like that seen in dreamless sleep or anesthesia. In patients in a minimally conscious state, it "invariably triggered complex activations" spreading to distant areas, as in locked-in patients, who are conscious. In patients who gradually recovered, this change could appear before reliable communication was established and before the spontaneous EEG changed. The group would condense these responses into a single number, the perturbational complexity index, the following year.
2012	Publication	Neuroscientist Christof Koch publishes Consciousness: Confessions of a Romantic Reductionist with MIT Press, a book that is part scientific overview, part memoir, and part futurist speculation. Koch recounts his collaboration with Francis Crick in the 1990s and 2000s and the emergence of consciousness, once considered a fringe subject, as a legitimate topic of scientific investigation. He places a handful of colleagues at the center of that shift, including Ned Block, David Chalmers, Stanislas Dehaene, Giulio Tononi, and Wolf Singer. Alongside chapters on attention and awareness, neurons that respond to celebrities, free will, and the loss of his belief in a personal God, Koch devotes one chapter to arguing that consciousness is a fundamental property of complex things. That echoes Tononi's 2004 claim that consciousness is a fundamental quantity. Another chapter outlines an electromagnetic gadget to measure consciousness. The book signals Koch's embrace of the framework he would later co-develop with Tononi, whom he joins as co-author of the 2016 Nature Reviews Neuroscience overview of IIT.
2012 (June–September)	Theory	Giulio Tononi publishes "Integrated information theory of consciousness: an updated account" in the Archives Italiennes de Biologie. Building on thought experiments, the article recasts IIT as a set of phenomenological axioms (information, integration, exclusion) paired with "ontological postulates". The information postulate holds that only "differences that make a difference" from the intrinsic perspective of a system matter: a mechanism generates information if its present state has specific past causes and specific future effects within the system. The integration postulate admits only information that cannot be partitioned among independent components. The exclusion postulate keeps only maxima, so each mechanism specifies a single maximally irreducible set of causes and effects, called a concept. A complex is a set of elements specifying a maximally irreducible constellation of concepts, evaluated at its optimal spatio-temporal scale, and this structure, or quale, is identical with an experience. The article also links the temporal grain of integration to metastable dynamics in the corticothalamic system. A corrected version is republished in the journal's December 2012 issue. In 2014, the formal presentation of IIT 3.0 would point to this account as its precursor.
2012 (June 25)	Critique	Philosopher Eric Schwitzgebel of the University of California, Riverside circulates a draft paper, "If Materialism Is True, the United States Is Probably Conscious". He argues that materialists have no principled reason to deny consciousness to spatially distributed but informationally integrated entities, which would make the United States a candidate for literal consciousness. Turning to Tononi, he notes that IIT avoids this conclusion through an "anti-nesting principle": when one integrated system is nested in another, consciousness occurs only at the most highly integrated level. Schwitzgebel questions why Tononi, who grants a photodiode a bit of consciousness, should resist group consciousness at all. He also argues that the principle has its own absurd consequences. If a large enough election, with results communicated back to every voter, integrated information more than any one brain, every voter would lose consciousness. And since "greater than" is all-or-none, a single mail-in ballot could tip the balance, with no detectable effect on anyone's behavior. The objection would echo through later critiques, and Tononi and Koch would answer that IIT allows "nothing-it-is-like-to-be" the United States.
2012 (August)	Publication	Giulio Tononi, professor of neuroscience at the University of Wisconsin, publishes Phi: A Voyage from the Brain to the Soul through Pantheon Books, a 384-page book that presents integrated information theory to a general readership. Rather than an exposition, the book is a literary journey in three parts: an aging Galileo Galilei is guided by three famous scientists. Francis Crick, co-discoverer of the structure of DNA, shows him how portions of the brain function separately and together. Alan Turing, a founder of artificial intelligence, helps him link these facts into "a scientific theory of consciousness". Charles Darwin explores how consciousness evolves. Notes at the end of each chapter, written in a voice separate from the narrator's, expand on the themes; the last one remarks that the book's closing symbol "must hold some great significance to the author but could not be deciphered." Publishers Weekly calls the book "both playful and philosophical" and "extravagant", praising its lavish artwork and literary references as showing how fully complementary art and science can be. The book appears the same year as Christof Koch's Consciousness: Confessions of a Romantic Reductionist.
2013 (January–March)	Critique	Philosopher John Searle reviews Christof Koch's Consciousness: Confessions of a Romantic Reductionist in The New York Review of Books (January 10 issue), arguing that information theory cannot explain consciousness: consciousness is observer-independent, but information exists only relative to an observer. In a letter published on March 7, Koch and Giulio Tononi reply that this is exactly IIT's own starting point. IIT, they write, introduces a "non-Shannonian" notion of integrated information, "differences that make a difference" to a system from its intrinsic perspective. Against Searle's objection that consciousness "cannot spread over the universe like a thin veneer of jam", they add that only local maxima of integrated information exist, so there is no superordinate consciousness of the United States. Searle answers by quoting their own words. He cites Tononi's claim that a binary photodiode "enjoys exactly 1 bit of consciousness", and Koch's book, which calls IIT "an elaborate version of panpsychism" and finds a modicum of Φ even in protons. Searle maintains that the integrated information in such systems is observer-relative too, and concludes that the theory "does not seem to be a serious scientific proposal."
2013 (August 14)	Measure	Adenauer Casali, Marcello Massimini, and colleagues, including Steven Laureys and Giulio Tononi, publish "A Theoretically Based Index of Consciousness Independent of Sensory Processing and Behavior" in Science Translational Medicine. Motivated by the lack of an objective bedside measure of consciousness for unresponsive brain-injured patients, they introduce the perturbational complexity index (PCI). Extending the team's earlier TMS–EEG work, PCI perturbs the cortex with magnetic pulses to engage distributed interactions (integration) and then compresses the resulting electrical responses to measure their algorithmic complexity (information). Applied to existing recordings, PCI ranges from 0.44 to 0.67 in 32 awake healthy subjects but drops to 0.18–0.28 in non-REM sleep, with similarly low values under midazolam, xenon, and propofol anesthesia. Among patients who emerged from coma, those in a vegetative state score 0.19–0.31, those with locked-in syndrome 0.51–0.62, and those in a minimally conscious state in between. The journal's editorial summary cautions that clinical validity still requires prospective trials. The 2016 Nature Reviews Neuroscience overview of IIT would cite PCI among the theory's contributions to assessing consciousness in non-communicative patients.
2013 (November 18)	Theory	Erik Hoel, Larissa Albantakis and Giulio Tononi publish "Quantifying causal emergence shows that macro can beat micro" in the Proceedings of the National Academy of Sciences. It is widely assumed that once the micro level of a system is fixed, macro levels are fixed too (supervenience), and that only the micro level is causally complete, leaving no room for causation at higher levels. The authors test this with effective information, a measure introduced by Tononi and Olaf Sporns in 2003, which is higher the more a system's mechanisms constrain its possible past and future states. In simple systems with fixed micro mechanisms, they show that effective information can peak at a macro level in space or time, when coarse-grained mechanisms are more deterministic or less degenerate than the underlying ones. Although the macro level supervenes on the micro, it "can supersede it causally", which they call causal emergence. The question of the right grain matters for IIT, whose exclusion postulate holds that the substrate of consciousness is specified at the spatial and temporal grain that maximizes integrated information.
2013	Critique	Philosopher Anthony Peressini of Marquette University publishes "Consciousness as Integrated Information: a Provisional Philosophical Critique" in the Journal of Consciousness Studies. It critically examines the theory Giulio Tononi had set out in his 2008 "provisional manifesto". Peressini raises some concerns internal to the theory and then broader philosophical ones, before assessing its prospects as a fundamental theory of consciousness. He finds that much of its scientific promise does carry over to philosophical theorizing about qualia and consciousness, but "not as directly as Tononi suggests", because the account focuses on the qualitative character of experience rather than on consciousness itself. He therefore proposes renaming it an "integrated information theory of qualia" (IITQ). The question of what exactly IIT is a theory of would recur in later philosophical critiques, such as Adam Pautz's 2019 catalogue of questions about the theory.
2014 (February 4)	Critique	Adam Barrett of the Sackler Centre for Consciousness Science at the University of Sussex publishes "An integration of integrated information theory with fundamental physics" in Frontiers in Psychology. He argues that existing formulations of IIT "are not applicable to standard models of fundamental physical entities". Its Φ measures apply only to networks of discrete elements, so they depend on an observer's choice of graining and cannot capture truly intrinsic information. Since modern physics treats fields as fundamental, and brains process information electrically, he proposes the "field integrated information hypothesis": consciousness arises from information intrinsic to fundamental fields, in practice the electromagnetic field. This would require a frame-invariant measure for continuous field configurations, which he admits does not yet exist. He also defends IIT's implied panpsychism against John Searle's objection, arguing that minute "germs" of consciousness everywhere are "no reason to dismiss" the theory. He concedes the hypothesis is "rather speculative" and hard to test directly. Barrett and colleagues would return to the call for a field-based reformulation in a 2026 assessment of IIT.
2014 (May 8)	Theory	Masafumi Oizumi, Larissa Albantakis, and Giulio Tononi publish "From the Phenomenology to the Mechanisms of Consciousness: Integrated Information Theory 3.0" in PLOS Computational Biology. The paper calls the earlier stationary-system version IIT 1.0 and the state-dependent version IIT 2.0. It starts from axioms about experience (existence, composition, information, integration, and exclusion) and turns them into postulates about how physical mechanisms such as neurons or logic gates must be configured. Information becomes "differences that make a difference" within a system, measured with the earth mover's distance. The central identity holds that an experience is a maximally irreducible conceptual structure, a constellation of concepts in qualia space, whose irreducibility ΦMax sets its quantity. Worked examples yield several corollaries: inactive elements can contribute to experience; a two-element "photodiode" with feedback is minimally conscious; and a feedforward network can be a "zombie", unconscious yet matching a conscious system's input–output behavior. The authors build that zombie by unfolding a recurrent system's elements over time. Five years later, critics would turn this very construction against the theory in the unfolding argument.
2014 (May 21)	Critique	Theoretical computer scientist Scott Aaronson publishes "Why I Am Not An Integrated Information Theorist (or, The Unconscious Expander)" on his blog Shtetl-Optimized, answering a challenge from physicist Max Tegmark, an enthusiastic adopter of IIT. Aaronson coins the "Pretty-Hard Problem of Consciousness" (deciding which physical systems are conscious) and calls IIT a serious attempt at it that nonetheless fails. He shows that a system that merely multiplies its input by a Vandermonde matrix, or that applies error-correcting codes built on expander graphs, can have an arbitrarily large Φ, so IIT would deem it "unboundedly more conscious than humans are". In the comments, a reader notes that IIT 3.0, published two weeks earlier, had dropped the normalization his construction exploits. Aaronson nonetheless places IIT in the top 2% of mathematical theories of consciousness, precisely because it can be shown wrong. Tononi would reply nine days later, accepting the conclusion that such networks are conscious.
2014 (May 30)	Response	Scott Aaronson publishes "Giulio Tononi and Me: A Phi-nal Exchange" on Shtetl-Optimized, sharing a 14-page rebuttal Tononi had sent him, "Why Scott should stare at a blank wall and reconsider (or, the conscious grid)". Tononi notes that Aaronson's calculations did not use IIT 3.0, but sets the point aside. He then affirms that a large network of XOR gates is conscious, and adds that the expander structure is unnecessary: even a simple two-dimensional square grid of such gates would be conscious. Rejecting appeals to common sense, he argues that staring at a blank wall reveals a richly structured experience of 2D space (locations, relative positions, distances), which requires a grid-like physical substrate to create it intrinsically. Aaronson credits Tononi's "intellectual courage" but is unpersuaded. He points out that a one-dimensional line of gates would have almost no Φ, although staring at a rope is also a structured experience. He also charges that IIT appeals to "largely uncontroversial" facts, such as an unconscious cerebellum, while rejecting intuition about grids. He concludes the two are at "a philosophical impasse". Tononi's blank-wall argument would grow into IIT's 2019 account of spatial experience.
2015 (May 19)	Publication	Giulio Tononi and Christof Koch publish "Consciousness: here, there and everywhere?" in Philosophical Transactions of the Royal Society B, setting out what IIT 3.0 implies about which systems are conscious. They argue that behavioral and neural correlates cannot settle hard cases such as brain-damaged patients, preterm infants, distant animal species, and machines. The authors acknowledge that IIT shares central intuitions of panpsychism: consciousness is a fundamental property, graded, likely widespread among animals, and present in small amounts even in very simple circuits. Unlike panpsychism, however, IIT holds that not everything is conscious. Its exclusion postulate answers the combination problem: two people talking, a nation, or a sand dune form no further conscious entity. Purely feedforward networks, such as those used in deep learning, are unconscious however capable, so no Turing test can suffice. Even a digital computer faithfully simulating a human brain would likely experience next to nothing, just as a simulated star does not bend space–time, while neuromorphic hardware might not face this limit. The authors propose validating the theory on human consciousness first and extrapolating outward from there.
2015 (July)	Publication	Physicist Max Tegmark publishes "Consciousness as a state of matter" in Chaos, Solitons & Fractals, proposing that consciousness can be understood as a state of matter, which he calls "perceptronium", much as solids, liquids, and gases are distinguished by how their atoms are arranged. Crediting Tononi's 2008 manifesto with making compelling the idea that a conscious system must both store a large amount of information and integrate it into a unified whole, Tegmark asks how these principles could be grounded in fundamental physics: which physical properties of moving particles should count as bits, and which particle arrangements are conscious? He explores four candidate principles (information, integration, independence, and dynamics) and generalizes Tononi's framework from neural networks to arbitrary quantum systems. He uses them to attack what he calls the "quantum factorization problem": why conscious observers perceive the world as classical space and as a hierarchy of integrated, relatively independent objects. Along the way he finds links to error-correcting codes, condensed matter criticality, and Quantum Darwinism, and a connection between the emergence of consciousness and the emergence of time.
2015 (September 17)	Critique	Michael Cerullo publishes "The Problem with Phi: A Critique of Integrated Information Theory", a Perspective in PLOS Computational Biology, the journal that had carried several of IIT's own formulations. He argues that IIT rests on a principle of "information exclusion" (that consciousness grows with the number of alternatives a system rules out), for which Tononi offers no justification beyond intuition. To test IIT's claimed empirical support, Cerullo invents a deliberately trivial rival, "Circular Coordinated Message Theory", which ties consciousness to information circulating in feedback loops and reproduces IIT's predictions about split-brain patients and the cerebellum. Because IIT rejects computational functionalism, he adds, it is vulnerable to David Chalmers's "fading" and "dancing" qualia thought experiments. Drawing on the 2014 exchange between Scott Aaronson and Tononi, in which Tononi accepted that a large grid of logic gates could exceed a brain's Φ and maintained that it would be conscious, Cerullo concludes that IIT describes at most a noncognitive "protoconsciousness", a form of partial panexperientialism. He recalls that after Tononi's talk at the 2014 Science of Consciousness meeting, Ned Block remarked: "You have a theory of something, I am just not sure what it is."
2015 (November)	Event	New York University's Center for Mind, Brain and Consciousness, co-directed by David Chalmers and Ned Block, co-hosts a two-day workshop on integrated information theory, at which Giulio Tononi, Christof Koch, Chalmers, and ten other speakers present to about 30 scientists and philosophers. After Tononi's 90-minute tutorial and Koch's talk, Scott Aaronson, styling himself the "official IIT skeptic", reprises his argument that simple error-correcting systems can have enormous Φ. He charges IIT's proponents with invoking common sense when it suits them (the unconscious cerebellum) but rejecting it when it does not (the conscious grid). Tononi replies that Aaronson critiqued an outdated version of Φ and that IIT is a work in progress, while physicist Max Tegmark presents half a dozen alternative definitions of Φ, which Aaronson says his analysis also covers. Participants are particularly puzzled by the exclusion postulate. Reporting for Scientific American's blog network, science journalist John Horgan writes that the workshop strengthened his mysterianism. He endorses John Searle's 2013 objection that information presupposes a conscious observer, and predicts that IIT's panpsychist implications will limit its popularity.
2015 (November 19)	Experiment	Simone Sarasso, Marcello Massimini, Giulio Tononi and colleagues publish "Consciousness and Complexity during Unresponsiveness Induced by Propofol, Xenon, and Ketamine" in Current Biology. General anesthesia is usually equated with loss of consciousness, but subjects can be disconnected from the environment while still having experiences, as in dreaming, and ketamine in particular produces "ketamine dreams". In 18 healthy volunteers, the authors measure the complexity of the cortical response to transcranial magnetic stimulation, the approach behind the perturbational complexity index, and collect reports on emergence. Under propofol and xenon, stimulation evokes low-complexity responses (local under propofol, global but stereotypical under xenon), and subjects report no experience. Under ketamine, responses show "a wakefulness-like, complex spatiotemporal activation pattern", and subjects report "long, vivid dreams unrelated to the external environment". The results suggest that brain complexity "may be sensitive to the presence of disconnected consciousness in subjects who are considered unconscious based on behavioral responses". The study extends earlier work showing reduced effective connectivity under midazolam, and the same issue carries a commentary on the findings.
2016 (January 11)	Measure	Physicist Max Tegmark of MIT posts "Improved Measures of Integrated Information" on arXiv, following his 2015 proposal of consciousness as a state of matter. Noting that current methods of measuring integrated information in real systems are "computationally unfeasible", a problem Adam Barrett and Anil Seth had addressed in 2011, he classifies existing and new measures by their desirable properties. He presents a taxonomy in which each Φ measure is defined by three choices: a factorization method (5 options), the probability distributions to compare (3×4 options), and the measure used to compare them (7 options). Of the hundreds of possible combinations, only "a mere handful" satisfy a minimum of attractive properties, and some of these turn out to be identical. He derives exact and approximate formulas that can be applied to laboratory data "without posing unreasonable computational demands". The paper would be published in PLOS Computational Biology in November 2016. A systematic comparison of candidate Φ measures by Pedro Mediano and colleagues would follow in 2018.
2016 (May 26)	Publication	Giulio Tononi, his Wisconsin colleague Melanie Boly, Marcello Massimini of the University of Milan, and Christof Koch, now at the Allen Institute for Brain Science, publish "Integrated information theory: from consciousness to its physical substrate", an Opinion article in Nature Reviews Neuroscience. The article presents the theory in its axiomatic form to a broad neuroscience readership. It starts from five essential properties of experience, called axioms (intrinsic existence, composition, information, integration, and exclusion), and derives postulates that any physical substrate of consciousness must satisfy, concluding that this substrate must be a maximum of intrinsic cause–effect power. On this basis, the authors argue that the theory can determine, in principle, both the quality and the quantity of experience. They note that it leads to some counterintuitive predictions and argue that it can be used to develop new tools for assessing consciousness in patients unable to communicate. The work is supported in part by the Templeton World Charity Foundation, which would later fund an adversarial collaboration testing IIT against a rival theory.
2016 (November)	Measure	Silvia Casarotto, Marcello Massimini, Giulio Tononi and colleagues publish "Stratification of unresponsive patients by an independently validated index of brain complexity" in Annals of Neurology. It is a large validation of the perturbational complexity index (PCI), which measures the complexity of the cortical response to transcranial magnetic stimulation recorded with EEG and which the group introduced in 2013. To get around the absence of subjective reports from unresponsive patients, the authors first calibrate PCI on a benchmark population of 150 healthy controls and communicative brain-injured subjects, whose conscious, disconnected-conscious and unconscious states can be confirmed by report. A single empirical cutoff separates conscious from unconscious conditions with 100% sensitivity and specificity. Applied to 81 non-communicative patients with disorders of consciousness, the cutoff detects 94.7% of those in a minimally conscious state. It also flags 9 of 43 patients in a vegetative state with high PCI values, a subgroup that the authors suggest "may retain a capacity for consciousness that is not expressed in behavior".
2016 (December 6)	Measure	Masafumi Oizumi of RIKEN, Naotsugu Tsuchiya and mathematician Shun-ichi Amari publish "Unified framework for information integration based on information geometry" in the Proceedings of the National Academy of Sciences. They note that the original measure of integrated information is "severely limited" in its application to experimental data, and that alternatives proposed under more general conditions, including the time-series measures of Adam Barrett and Anil Seth, suffer from mathematical problems: overestimation when separately measured influences are summed, and confounding by non-causal influences. They derive a measure from three postulates: the strength of influences is the minimized difference between a system's actual probability distribution and a "disconnected model", disconnection is defined by a Markov condition, and the difference is measured by the Kullback–Leibler divergence. The resulting "geometric integrated information" (ΦG) is never larger than the mutual information between a system's past and present states, unlike a sum of transfer entropies or stochastic interaction. For Gaussian variables, ΦG has a closed form consistent with multivariate Granger causality. Amari and colleagues would extend the framework in 2017.
2017 (June 5)	Experiment	Chen Song, Andrew Haun and Giulio Tononi of the University of Wisconsin–Madison publish "Plasticity in the Structure of Visual Space" in eNeuro. Drawing on IIT, they hypothesize that the strength of lateral connections between neurons in visual cortex determines how related locations in the visual field feel, so strengthening those connections should shrink perceived distance. Thirty participants passively view two points flashed in synchrony, a Hebbian protocol known to strengthen lateral connections in cat visual cortex. Afterwards the perceived distance between the points contracts by up to 7.2%, with the effect greatest at a separation corresponding to about 12 millimeters in primary visual cortex and weaker at shorter or longer ones. Untrained locations just beyond the trained ones also contract, as the hypothesis predicts. The authors acknowledge that they cannot rule out changes in feedforward input, attention or feedback from frontal and parietal cortex, and a reviewer asks them to weigh explanations other than IIT. Two years later, Haun and Tononi would give a formal IIT account of spatial experience predicting that changes in connectivity alone should warp experienced space.
2017 (September 7)	Measure	Mathematician Shun-ichi Amari and neuroscientists Naotsugu Tsuchiya and Masafumi Oizumi post "Geometry of Information Integration" on arXiv. By this point multiple measures of integrated information have been proposed, each quantifying in its own way how much information is lost when a system is split into parts and the information transmission between them is removed. Using information geometry, the authors analyze four of these measures and clarify how they relate: two define the split model on dually flat manifolds and two on curved manifolds. They show that the measures form a hierarchy and give explicit expressions for each. The paper offers a unifying mathematical view of a growing family of approximations, needed because, as Balduzzi and Tononi had acknowledged in 2008, exact integrated information cannot feasibly be computed for systems as large as the brain.
2017 (October 4)	Publication	Melanie Boly, Marcello Massimini, Naotsugu Tsuchiya, Bradley Postle, Christof Koch and Giulio Tononi publish "Are the Neural Correlates of Consciousness in the Front or in the Back of the Cerebral Cortex? Clinical and Neuroimaging Evidence" in the Journal of Neuroscience, as one side of a "Dual Perspectives" exchange. Reviewing lesion, stimulation and neuroimaging studies, they argue that posterior cortex, rather than prefrontal cortex, is the main substrate of conscious contents. Several patients with extensive frontal damage kept normal consciousness, stimulating posterior cortex more reliably alters experience, and dream studies point to a parieto-occipital "posterior hot zone". Frontal activity, they suggest, often reflects task and report demands, and frontal regions act mainly as background conditions. The companion piece, by Brian Odegaard, Robert Knight and Hakwan Lau, defends a role for prefrontal cortex. The front-versus-back question would become a central test in the first adversarial collaboration between IIT and the global neuronal workspace theory.
2018 (March)	Adversarial collaboration	The Templeton World Charity Foundation holds the first workshop of its Accelerating Research on Consciousness initiative, a commitment of $30 million built around what the foundation calls "Structured Adversarial Collaboration". The meeting takes place at the Allen Institute for Brain Science, hosted by Christof Koch. It brings together proponents and critics of integrated information theory, the global neuronal workspace theory and higher-order theory. Participants include Giulio Tononi, Stanislas Dehaene, Hakwan Lau, David Chalmers, Lucia Melloni, Liad Mudrik and Michael Pitts. A proposal arising from the workshop would be funded in May 2019 as an adversarial collaboration between IIT and the global neuronal workspace theory, the project later known as COGITATE, whose first results would be presented in 2023.
2018 (June 25)	Measure	Pedro Mediano, Anil Seth, and Adam Barrett post "Measuring Integrated Information: Comparison of Candidate Measures in Theory and Simulation" on arXiv; the paper would later appear in the journal Entropy. IIT is built around measures of how much more information a system generates than the sum of its parts, but the authors note that little is known about how the several existing candidate measures of integrated information (Φ) compare on non-trivial networks. They give intuitive descriptions of six distinct measures and test them in simulations of eight-node networks with Gaussian linear autoregressive dynamics. The results show a striking diversity: no two measures agree consistently across all analyses, and only some appear to genuinely reflect dynamical complexity, in the sense of simultaneous segregation and integration among system components. The study extends Seth and Barrett's 2011 comparison of causal density and integrated information, and complements the previous year's mathematical comparison of four measures by Amari, Tsuchiya, and Oizumi. Its results underscore that operationalizing IIT depends on which measure is chosen.
2018 (June 29)	Critique	Philosopher Tim Bayne publishes "On the axiomatic foundations of the integrated information theory of consciousness" in Neuroscience of Consciousness. Calling IIT one of the most influential scientific theories of consciousness and a guiding framework for attempts to build a consciousness meter, Bayne asks whether its foundations are secure. He notes that the IIT literature describes the postulates as "derived" from the axioms, yet never weighs rival explanations, as an abductive inference would require. Taking the five axioms in turn, he argues that none can play its intended role. Some are self-evident but impose no substantive constraint, such as the claim that each experience is "the particular way it is". Others constrain theories but are not self-evident, such as the claim that consciousness is purely intrinsic, which externalist philosophers reject. He cites associative agnosia against the integration axiom, and vague contents, like the indeterminate number of a tiger's visible stripes, against exclusion. Bayne proposes instead a "natural kind" approach modeled on the study of diseases such as hepatitis, within which, he suggests, IIT could still be developed. IIT 4.0 would later describe the step from axioms to postulates as an "inference to a good explanation".
2018 (July 5)	Response	Philosopher Francis Fallon of St. John's University in New York publishes "Integrated Information Theory, Searle, and the Arbitrariness Question" in the Review of Philosophy and Psychology, answering John Searle's 2013 attack on IIT. Searle had objected that IIT's appeal to information relies on observer-relative features. Fallon replies that this misses the point: IIT's integrated information is "intrinsic, the opposite of observer-relative". He goes further, arguing that Searle overlooks how IIT could extend his own view. Searle holds that consciousness is caused by the brain's specific causal powers, which "don't consist in mere information-processing or computation", but never says what those powers are; IIT agrees and identifies them with the generation of integrated information. Fallon concludes that IIT "renews a fundamentally ontological challenge to information-processing and computational theories of mind". The article would appear in the journal's 2020 volume.
2018 (July 26)	Measure	William Mayner, Giulio Tononi, and colleagues publish "PyPhi: A toolbox for integrated information theory" in PLOS Computational Biology. The paper describes an open-source Python package that implements IIT's framework and unfolds the full cause–effect structure of discrete dynamical systems of binary elements. Released on GitHub under the GNU GPLv3 license, with online documentation and a web-based graphical interface, the software serves as an up-to-date reference implementation of IIT's formalisms. Its authors note that it has also been applied in research on complexity, emergence, and certain biological questions. The paper is candid about limits: because the numbers of states, subsystems, mechanisms, purviews, and partitions all grow exponentially, computation time scales as O(n53n), restricting practical analysis to about 10 to 12 elements. A chain of seven stochastic majority gates takes roughly 2.75 hours on 32 processor cores. The authors also warn that the analysis assumes Markovian systems with conditionally independent elements, assumptions that may not hold for data such as EEG recordings. PyPhi would later be used to compute the worked examples in the 2023 IIT 4.0 paper.
2018 (November 22)	Response	Philosophers Olimpia Lombardi of CONICET and the University of Buenos Aires and Cristian López of the University of Lausanne publish "What Does 'Information' Mean in Integrated Information Theory?" in a special issue of Entropy on IIT. Critics such as John Searle and Michael Cerullo had assumed that any talk of information must follow Claude Shannon's theory, which describes statistical dependence from an observer's extrinsic perspective. The authors argue instead that IIT's information should be read through a "causal–manipulabilist" view. On this view, information is defined by the causal links revealed through interventions, following the theories of causation of James Woodward (philosopher) and Judea Pearl. On this reading, IIT's appeal to cause-effect power need not commit it to metaphysical causal powers, only to an operational criterion that empirical science can use. Objections such as IIT's failure to solve the hard problem, they argue, treat it as a philosophical proposal rather than a scientific theory. They conclude that it "deserves to be seriously considered both from an empirical and from a conceptual viewpoint".
2019 (January)	Critique	Philosopher Adam Pautz of Brown University publishes "What is the Integrated Information Theory of Consciousness?: A Catalogue of Questions" in the Journal of Consciousness Studies. Unlike most critics, he describes his goal as "modest": "I will not argue that the integrated information theory (IIT) is false." Instead he raises a series of basic questions about what the theory actually claims. His conclusion is that "as long as proponents of IIT do not address these questions, they have not put a clear theory on the table that can be evaluated as true or false." The paper joins Tim Bayne's 2018 challenge to IIT's axioms in questioning the theory's foundations and content rather than its empirical predictions.
2019 (January)	Critique	Philosopher Hedda Hassel Mørch publishes "Is Consciousness Intrinsic?: A Problem for the Integrated Information Theory" in the Journal of Consciousness Studies. IIT identifies consciousness with maximal integrated information, and its exclusion postulate holds that only the system with the highest Φ among overlapping candidates is conscious. Mørch formulates what she calls the intrinsicality problem: consciousness is an intrinsic property, but maximal Φ is extrinsic, since whether a system's Φ is maximal depends on the systems that overlap it, so the two cannot be identical. The objection is pointed because IIT itself counts intrinsic existence among the essential properties of experience. Mørch argues, however, that the problem is not unique to IIT. It follows from a trilemma facing almost any theory of consciousness: three claims are jointly inconsistent, namely that consciousness is intrinsic, that conscious systems do not overlap (as in Peter Unger's problem of the many), and that consciousness is constituted by more fundamental properties. She suggests that IIT's best option may be to reject the last claim, reductionism, rather than intrinsicality or non-overlap.
2019 (February 12)	Critique	Adam Barrett and Pedro Mediano post "The Phi measure of integrated information is not well-defined for general physical systems" on arXiv. They note that, according to IIT, consciousness is "a fundamental observer-independent property of physical systems", and Φ is identical to its quantity or level. For this to be plausible, they argue, "there should be no alternative formulae for Phi consistent with the axioms of IIT, and there should not be cases of Phi being ill-defined". The paper presents three ways in which Φ, in its current formulation, fails to meet these standards, and discusses how the problem might be addressed. It extends concerns Barrett had raised with Anil Seth in 2011 about the stability and physical meaning of Φ, and follows Mediano, Seth and Barrett's 2018 comparison of candidate measures, which found that they behave inconsistently. The paper appears in the same Journal of Consciousness Studies issue as Hedda Hassel Mørch's critique of IIT's claim that consciousness is intrinsic.
2019 (May 2)	Theory	Larissa Albantakis, William Marshall, Erik Hoel and Giulio Tononi publish "What Caused What? A Quantitative Account of Actual Causation Using Dynamical Causal Networks" in Entropy (journal), as part of a special issue on integrated information theory. They carry IIT's formal tools beyond consciousness to a long-standing problem in philosophy and computer science: given a specific transition between two states of a system, such as a neural network or brain circuit, which parts actually caused which? They start from five requirements adapted from IIT's postulates: realization, composition, information, integration and exclusion. From these they build a framework that considers all counterfactual states of a transition and quantifies the strength of every actual cause and effect, including causes made of several variables jointly. They apply it to classic paradoxes such as overdetermination and prevention, and to a toy image classifier, citing the difficulty of explaining the moves of AlphaGo. The account applies to discrete Markovian systems. The authors note that it cannot yet address causal dependencies in non-Markovian ones, a limit critics would later stress for IIT more broadly.
2019 (May 9)	Critique	Adrien Doerig, Aaron Schurger, Kathryn Hess, and Michael Herzog publish "The unfolding argument: Why IIT and other causal structure theories cannot explain consciousness" in Consciousness and Cognition. They distinguish theories that explain consciousness through brain functions, which many patterns of connectivity could realize, from causal structure theories such as IIT and Recurrent Processing Theory, which identify consciousness with causal structure itself, so that feedforward systems are never conscious and feedback systems always are. Drawing on theorems from the theory of computation, they argue that any recurrent network can be "unfolded" into a feedforward network with identical input–output behavior, so no experiment relying on reports could tell the two apart. Their conclusion is that causal structure theories are either false or outside the realm of science. Replies follow, including a 2020 call by Naotsugu Tsuchiya, Thomas Andrillon, and Andrew Haun to move "beyond functionalism/behaviorism". In 2022, Herzog, Schurger, and Doerig respond that first-person experience cannot rescue such theories: on their reading, IIT makes consciousness fully dissociated from anything measurable and therefore epiphenomenal, a position they call "dissociative epiphenomenalism".
2019 (May 15)	Measure	Renzo Comolatti, Adenauer Casali, Marcello Massimini, Giulio Tononi and colleagues publish "A fast and general method to empirically estimate the complexity of brain responses to transcranial and intracranial stimulations" in Brain Stimulation. The perturbational complexity index (PCI), introduced in 2013 to assess "the capacity of thalamocortical circuits to engage in complex patterns of causal interactions", had proved highly accurate in detecting consciousness in brain-injured patients. However, it required elaborate setups and offline processing and was limited to transcranial magnetic stimulation with high-density EEG. The authors introduce PCIST, based on dimensionality reduction and the quantification of state transitions in evoked responses. Validated on recordings from 108 healthy subjects and 108 brain-injured patients, it matches the original index in accuracy while being computed "in less than a second" with a simpler setup. Applied to intracranial recordings from nine epilepsy patients receiving single-pulse electrical stimulation, it detects a systematic reduction of complexity during sleep. The authors present it as a step toward "a reliable and fast clinical tool for the bedside assessment of consciousness". Comolatti would later work on IIT's account of temporal experience.
2019 (June)	Measure	Miguel Aguilera and Ezequiel Di Paolo publish "Integrated information in the thermodynamic limit" in Neural Networks (journal), addressing a basic limitation of integrated information theory. Its computational cost "generally precludes applications beyond relatively small models", so it is unclear how integration behaves as systems grow or interact with an environment. After revising some of the theory's assumptions, including the time span over which integration is measured and the treatment of elements outside a system, they compute modified measures of integration in infinitely large kinetic Ising models using mean-field approximations. They report "for the first time" how such measures scale, finding that integration diverges at certain critical points. Comparing these divergences across parts of a system lets them draw the boundary between an integrated unit and its environment. They also present a model that adapts to keep its integration despite environmental change. The work extends IIT's measures to large systems, at the cost of departing from the theory's exact formulation. Aguilera would extend the analysis to large finite networks in December 2019.
2019 (October)	Adversarial collaboration	The Templeton World Charity Foundation holds a second workshop at the Allen Institute for Brain Science, again hosted by Christof Koch, this time on testing integrated information theory against the global neuronal workspace theory in animal models. Participants include Giulio Tononi, Lionel Naccache, Pieter Roelfsema, Yuval Nir, Anne Churchland, Zachary Mainen, Massimo Scanziani and Shawn Olsen. The foundation would fund a resulting proposal in December 2021: an adversarial collaboration between the two theories in animal models, extending to non-human brains the approach of the human study funded after the first workshop in 2018. The resulting study protocol would be published in February 2026.
2019 (November 27)	Theory	Andrew Haun and Giulio Tononi publish "Why Does Space Feel the Way it Does? Towards a Principled Account of Spatial Experience" in a special issue of Entropy devoted to integrated information theory. Earlier work had concentrated on the quantity of consciousness; this paper uses the cause-effect structures formalized in IIT 3.0 to address its quality. The authors choose space because, unlike color or pain, it is "penetrable by introspection". They argue that for the visual canvas to feel extended, it must be composed of countless "spots" related to one another through connection, fusion and inclusion, from which regions, locations, sizes, boundaries and distances follow. Simulating small grid-like networks, they show that the distinctions and relations these networks specify mirror those properties, while a network with the same units but randomly scrambled connections specifies no extended distinctions at all. The results support the hypothesis that spatial experience depends on cortical areas with grid-like connectivity, and they predict "that changes in connectivity, even in the absence of changes in activity, should lead to a warping of experienced space". The approach would be carried into IIT 4.0 and extended to the experience of time.
2019 (December 5)	Measure	Miguel Aguilera of the University of the Basque Country and the University of Zaragoza publishes "Scaling Behaviour and Critical Phase Transitions in Integrated Information Theory" in a special issue of Entropy on IIT. Because of its computational cost, IIT 3.0 is difficult to apply to systems "larger than a dozen units", and it is not well known how integrated information scales with size. Extending his earlier mean-field work on infinite networks to large finite ones, Aguilera uses a simple model of a critical phase transition, an infinite-range kinetic Ising model with homogeneous couplings, as a benchmark against which to test the theory's design choices. He examines how integration is measured over time, how cause repertoires are computed, which distance measure is used and how the environment is treated. He concludes that some of IIT 3.0's mathematical definitions "fail to capture critical phase transitions" and proposes a reformulation of the assumptions behind integrated information measures.
2019	Publication	Christof Koch publishes The Feeling of Life Itself: Why Consciousness Is Widespread but Can't Be Computed with MIT Press, a book-length case for integrated information theory, which a reviewer calls "the beating heart of this book". Koch presents IIT as a theory that "starts with experience and asks how matter must be organized to support the mental", and argues that consciousness is graded with integrated information, so that many other creatures share it to some degree. His central claim concerns machines: "Consciousness is not a clever algorithm. Its beating heart is causal power upon itself, not computation", and such causal power "cannot be simulated. Not now nor in the future". Applied to digital computers, IIT implies that their maximally irreducible cause-effect power is "tiny and independent of the software running on the computer", whatever a simulation does. The position contrasts with Koch and Giulio Tononi's 2008 answer to whether machines can be conscious, which was "Yes". The book closes with a chapter on the ethics of a world in which experience is widespread.
2020 (January 28)	Response	Naotsugu Tsuchiya, Thomas Andrillon and Andrew Haun publish "A reply to 'the unfolding argument': Beyond functionalism/behaviorism and towards a science of causal structure theories of consciousness" in Consciousness and Cognition. It is among the first replies to the 2019 argument by Adrien Doerig and colleagues that IIT and other causal structure theories cannot be tested scientifically, because a recurrent network can be "unfolded" into a feedforward one with identical input-output behavior but different predicted consciousness. As the title signals, the authors argue that the unfolding argument presupposes a functionalist or behaviorist view, in which consciousness can be inferred only from a system's input-output behavior, and they defend a science of causal structure theories instead. Johannes Kleiner would publish a separate reply later in 2020. In 2022, Doerig, Michael Herzog and Aaron Schurger would answer this reply directly, arguing that first-person experience cannot rescue causal structure theories from the argument.
2020 (June 12)	Critique	Jake R. Hanson and Sara Imari Walker of Arizona State University post "Formalizing Falsification for Theories of Consciousness Across Computational Hierarchies" on arXiv. They build on the 2019 unfolding argument, which held that IIT lets functionally identical systems differ in predicted consciousness and so may be falsified without any experiment, but note that the debate has stayed "too abstract to readily determine the scope of their validity". To make it concrete, they describe two functionally equivalent machines buildable with table-top electronics: isomorphic digital circuits, one with feedback and one without. Using these, they show that IIT is "simultaneously falsified" at the level of the finite-state automaton and "unfalsifiable" at the level of the combinatorial state automaton. They propose a general criterion: to avoid being unfalsifiable or already falsified, a theory of consciousness must be invariant to changes that leave the inference procedure fixed at a given level of a computational hierarchy. The paper would appear in Neuroscience of Consciousness in 2021.
2020 (October)	Response	Mathematician Johannes Kleiner publishes "Brain states matter. A reply to the unfolding argument" in Consciousness and Cognition, answering the 2019 claim by Adrien Doerig and colleagues that IIT and other causal structure theories "are either false or outside the realm of science". Kleiner restates the unfolding argument in formal terms and proves that its premises imply "a much stronger result": the problem would hold for "almost all theories of consciousness" that depend non-trivially on physical systems, including the global neuronal workspace theory and functionalist and representationalist models, not only causal structure theories. He then argues that one premise is unwarranted, namely "that measures of brain activity cannot be used in an empirical test of theories of consciousness", and that without it the argument "ceases to work". Kleiner would extend the analysis with Erik Hoel in 2021. In 2022, Doerig, Michael Herzog and colleagues would observe that all sides of the debate largely agree the argument holds if consciousness science rests strictly on input-output experiments, so most replies, including Kleiner's, hold that first-person experience must also count, a move they reject.
2020 (December 30)	Critique	Philosopher Matthias Michel and neuroscientist Hakwan Lau publish "On the dangers of conflating strong and weak versions of a theory of consciousness" in the open-access journal Philosophy and the Mind Sciences. They observe that some proponents of IIT hold strong views on the neural correlates of consciousness: that large swathes of the neocortex, the cerebellum, at least some sensory cortices, and the limbic system are all not essential for any form of conscious experience. Michel and Lau argue that this is no accident. Conflating a strong version of the theory with a weaker, empirically grounded one, they contend, has led these researchers to adopt definitions of the neural correlates of consciousness inconsistent with their own earlier definitions, "inadvertently betraying the promises of an otherwise fruitful empirical endeavour". Their reference list draws on the full critique literature to that point, from Aaronson's 2014 blog posts to Bayne, Cerullo, and the unfolding argument. In 2022, Pedro Mediano, Anil Seth, and colleagues would build on the strong/weak distinction, recasting it as a constructive division of labor.
2021 (February 26)	Experiment	Angus Leung, Dror Cohen, Bruno van Swinderen and Naotsugu Tsuchiya publish in PLOS Computational Biology a rare empirical application of IIT 3.0's cause–effect structure to neural recordings. Using local field potentials from 15 channels spanning the brains of fruit flies, recorded during wakefulness and isoflurane anesthesia, they estimate what they call "integrated information structures": the set of φ values for every mechanism within four-channel systems. They find that integrated interactions during wakefulness "collapsed to isolated clusters of interactions" under anesthesia, uniformly across the brain. These structures also discriminated the two states more accurately than system-level Φ, the "scalar summary" that IIT champions as its main measure. Because such structures cannot arise in purely feedforward systems, the results also challenge the view of insect brains as largely feedforward. The authors caution that they do not apply every aspect of IIT. They infer causation from observed transitions rather than perturbations, and the findings "cannot directly confirm or invalidate IIT as a theory of consciousness", since the nature of insect consciousness remains unclear.
2021	Publication	Christof Koch publishes "Reflections of a natural scientist on panpsychism" in the Journal of Consciousness Studies. Admitting a long-standing "secret crush" on panpsychism, Koch writes that IIT shares several of its intuitions: consciousness is an intrinsic, fundamental property of reality, is graded, and is present in small amounts in simple physical systems. Unlike panpsychism, he argues, IIT specifies which systems are and are not conscious and addresses the combination problem through its exclusion postulate, which determines whether aggregates such as ant colonies, beehives, or murmurating flocks form a single conscious entity. He adds that IIT explains why consciousness is adaptive, and faults panpsychism and other philosophical "isms" for failing to offer a sustained research program that yields new experiments. Against philosopher Philip Goff's claim that quantitative science cannot capture qualitative experience, Koch points to IIT's mathematical account of how the experience of empty two-dimensional space corresponds to the causal powers of a simple grid-like network, work published by Andrew Haun and Giulio Tononi in 2019. The piece restates the position Koch and Tononi had set out six years earlier.
2021 (April 17)	Critique	Johannes Kleiner and Erik Hoel publish "Falsification and consciousness" in Neuroscience of Consciousness. They call IIT "the first well-formalized theory of consciousness", which is why it has drawn "the most in-depth responses, expansions, and criticisms", among them the 2019 unfolding argument, and they set out to ask when any theory of consciousness can be falsified. Testing a theory means comparing the experience it predicts from a system's internal observables, such as brain imaging data, with the experience an experimenter infers from report or behavior. Through a "substitution argument", they show that if prediction and inference are independent, as the field's reliance on report implies, "any minimally informative theory of consciousness is automatically falsified". This affects most contemporary theories, including IIT and the global neuronal workspace theory, not IIT alone. If prediction and inference are strictly dependent, as in behaviorist theories, the theory becomes unfalsifiable. The authors explore possible ways out of the dilemma. Hoel had co-authored IIT's 2013 work on causal emergence.
2021 (May 19)	Critique	Bjorn Merker, Kenneth Williford and David Rudrauf publish "The integrated information theory of consciousness: A case of mistaken identity" as a target article in Behavioral and Brain Sciences. They examine the construct validity of Φ by analyzing its formal properties, its relation to key aspects of consciousness, and how it varies with relevant empirical circumstances. They conclude that what Φ actually measures is "the causal efficacy with which differentiated networks accomplish global information transfer", and that identifying this with consciousness is mistaken. The misidentification would require attributing consciousness to "large-scale electrical power grids, gene-regulation networks, some electronic circuit boards, and social networks". Rather than treating this as a disconfirmation, they argue, IIT "embraces it", drifting "toward the orbit of panpsychist ideation". They instead propose what function the cortical combination of differentiation and integration actually performs in humans. Following the journal's open peer review format, the article appears with 23 commentaries and the authors' response, and it is published in the journal's 2022 volume. It extends earlier objections that IIT attributes consciousness to simple systems, notably Scott Aaronson's of 2014.
2021 (August 30)	Publication	Simone Sarasso, Adenauer Casali, Silvia Casarotto, Mario Rosanova, Corrado Sinigaglia and Marcello Massimini publish "Consciousness and complexity: a consilience of evidence" in Neuroscience of Consciousness. They argue that a surge of studies has converged on complexity-related measures as reliable markers of consciousness across sleep, anesthesia, hallucinatory states, coma and related disorders, whereas other candidate markers such as the P3b wave, metabolic rates and EEG power show conflicting results. These measures were "independently proposed by researchers endorsing disparate frameworks", but the authors trace their common denominator to a principle Tononi and Edelman put forward in 1998: consciousness requires the coexistence of functional integration and differentiation. They describe IIT as "the direct evolution" of that dynamic core hypothesis, and note that the global neuronal workspace, the free energy principle and other frameworks have since embraced complexity too. They propose a provisional taxonomy of the measures and flag open problems, including where in the brain complexity is generated and how to tell apart observational and perturbational estimates. Critics would later argue that such measures are compatible with rival theories and are proxies rather than approximations of Φ.
2021 (September 21)	Theory	Matteo Grasso, Andrew Haun and Giulio Tononi publish "Of maps and grids" in Neuroscience of Consciousness, a companion to Francesco Ellia and colleagues' paper on the "fallacy of misplaced objectivity". They model two simple eye-movement systems that represent a stimulus's position and move the eye to fixate it. One is "grid-like", with lateral connections, resembling posterior cortex; the other is "map-like", without them, resembling pretectal nuclei. Standard neuroscience tools show the two to be functionally equivalent, with input-output functions correlated at ρ = 0.9715. Unfolding their cause-effect structures with IIT, however, shows that the grid forms a single entity with 79 causal distinctions whose relations account for the extendedness of experienced space, while the map breaks down into seven independent units, so that "there 'cannot be' any phenomenology for a map". The authors conclude: "We must be grids, not maps." They predict that scrambling a grid's input and output connections in a compensatory way would leave function almost unchanged but make a moving dot appear at random locations. The dissociation of function from experience is precisely what the 2019 unfolding argument had called untestable.
2021 (September 23)	Theory	Matteo Grasso, Larissa Albantakis, Jonathan Lang and Giulio Tononi publish "Causal reductionism and causal structures" in Nature Neuroscience. They target causal reductionism, "the widespread assumption that there is no room for additional causes once we have accounted for all elementary mechanisms within a system". In neuroscience this means that once every neuron's firing is explained, nothing causal is left to account for. The authors argue that this intuition rests on "an implicit, unexamined notion of causation that conflates causation with prediction". Using a simple model organism, they show that causal reductionism "cannot provide a complete and coherent account of 'what caused what'", and they outline an explicit, operational approach to analyzing causal structures instead. The paper continues the IIT group's argument, begun with causal emergence in 2013, that higher-level causes need not reduce to micro-level ones, a view central to IIT's claim that consciousness is the intrinsic cause-effect structure of a system rather than the activity of its parts. Tononi would apply a similar intrinsic view of causation to free will in 2022.
2021 (September 30)	Critique	In an interview with Quanta Magazine timed to the publication of his book Being You: A New Science of Consciousness, neuroscientist Anil Seth of the University of Sussex sets out his mixed view of integrated information theory. The book presents perception and the self as a "controlled hallucination" generated by a predictive brain, and recounts how measures of consciousness independent of wakefulness have allowed locked-in patients to be recognized as conscious. In the interview, Seth finds "some bits of IIT promising, but not others". He credits Gerald Edelman and Giulio Tononi's observation that conscious experiences are both informative and integrated, and calls the theory's explanation of why the cerebellum contributes little to consciousness "a very convincing one". What he finds less promising is the claim "that integrated information actually is consciousness". This claim, he says, "makes it almost impossible to measure for any nontrivial system" and implies that "consciousness is sort of everywhere". Seth would later co-author the 2022 case for a "weak" version of IIT and the Sussex group's 2026 assessment of the theory.
2021 (October 7)	Response	Psychologist Marius Usher publishes "Refuting the unfolding-argument on the irrelevance of causal structure to consciousness" in Consciousness and Cognition. The 2019 unfolding argument held that because any recurrent network's input-output function can be approximated by an "equivalent" feedforward network, theories tying consciousness to causal structure, IIT among them, are unfalsifiable. Unlike replies that appeal to first-person experience, Usher answers on computational and neuroscientific grounds. He argues that "a robust functional equivalence" between feedforward and recurrent networks is supported neither by the mathematics of the universal approximation theorem nor by data from cognitive neuroscience. Such equivalence, he holds, can apply only to "static functions between input/output layers and not to the temporal patterns or to the network's reactions to structural perturbations". He also reviews evidence that consciousness has functional characteristics, such as flexible control of behavior, mediated by interacting top-down and bottom-up processes. In 2022, Michael Herzog, Adrien Doerig and colleagues would reply that the specific networks Usher discusses are not in fact functionally equivalent. His point about structural perturbations would be taken up in a 2026 analysis of plasticity.
2021 (October 15)	Theory	Francesco Ellia, Giulio Tononi and eight colleagues from the IIT group publish "Consciousness and the fallacy of misplaced objectivity" in Neuroscience of Consciousness. They accept that behavioral, functional and neural correlates are essential tools for studying consciousness, but warn against the "fallacy of misplaced objectivity": the assumption "that only objective properties should and can be accounted for objectively through science". What needs explaining, they argue, is "what experience is intrinsically", not "just what we can do with it extrinsically", and it must be explained, "otherwise the way experience feels would turn out to be magical rather than physical". Drawing on IIT, they argue that an "objective science of the subjective" can account in strictly physical terms both for the properties essential to every experience and for what makes particular experiences feel as they do, illustrated with the extendedness of spatial experience. The paper appears in the same journal and year as critiques arguing that theories like IIT cannot be falsified through report and behavior.
2022 (January 12)	Critique	Michael Herzog of the École Polytechnique Fédérale de Lausanne, Aaron Schurger of Chapman University, and Adrien Doerig of the Donders Institute publish "First-person experience cannot rescue causal structure theories from the unfolding argument" in Consciousness and Cognition, answering critics of their 2019 unfolding argument. They identify the crux of the debate as whether first-person experience, independent of third-person data, can ground a theory of consciousness, and argue that it cannot: science relies on measurement, and experience is measurable only through reports, brain activity, and the relation between them. No experiment based on third-person data, they contend, can therefore test IIT as a theory of consciousness: "IIT may be a good theory of something, but not of consciousness." Because IIT dissociates consciousness from anything measurable, they argue, it makes both the level and the content of consciousness epiphenomenal, a "dissociative epiphenomenalism" in which even reports of experience cannot be trusted. Since such reports are the foundation of IIT's axioms, they conclude that accepting IIT leads to rejecting its own axioms.
2022 (January 21)	Critique	Aaron Schurger, a co-author of the 2019 unfolding argument, and Michael Graziano publish "Consciousness explained or described?" in a Neuroscience of Consciousness special issue on theories of consciousness. They argue that most modern accounts, including the global neuronal workspace theory, higher-order thought theory and IIT, are not theories but "(competing) laws of consciousness": like Newton's law of gravity, they describe what they cannot explain. For IIT, "the step from Φ to subjective experience is a leap of faith". IIT's axioms are "asserted without proof or empirical evidence other than an appeal to introspection", and its identity claim lacks the third-person data that support identities such as water and H2O. The clinical success of the perturbational complexity index does not settle the matter either, since an accurate "consciousness meter" might track perception or another function tightly coupled with consciousness. The authors stress that they are "neither condemning nor even criticizing" these accounts, and argue that Graziano's attention schema theory is a genuine explanatory theory because it explains why people believe they have subjective experience.
2022 (February 24)	Adversarial collaboration	The University of Amsterdam announces a $3.6 million, three-year grant from the Templeton World Charity Foundation's "Accelerating Research on Consciousness" program for INTREPID, an adversarial collaboration pitting integrated information theory against two predictive processing theories. Neurorepresentationalism is represented by project director Cyriel Pennartz, active inference by Karl Friston, and IIT by Giulio Tononi. It follows the foundation's first collaboration, which set IIT against global neuronal workspace theory. Experiments led from Amsterdam, Melbourne, Glasgow and Madison, each replicated in a second laboratory, target incompatible predictions. One, coordinated by Umberto Olcese, tests whether inactive neurons contribute to conscious experience, as IIT predicts and its rivals deny. The theory leads commit to accepting the results "irrespective of whether these results support or challenge" their own theory. The consortium would preregister its predictions in 2024 and, in 2026, publish a joint review in which each theory states its confidence in each prediction, with results to be pooled under a Bayesian scheme.
2022 (March 11)	Experiment	A team including Yong Wang and Xiaoli Li publishes "Application of Fast Perturbational Complexity Index to the Diagnosis and Prognosis for Disorders of Consciousness" in IEEE Transactions on Neural Systems and Rehabilitation Engineering. Noting that TMS–EEG is an effective tool for measuring the level of consciousness, but that an accurate method for quantifying TMS-evoked activity is still lacking, the study applies a fast variant of the perturbational complexity index (PCIst) to recordings from 30 healthy participants and 181 patients with disorders of consciousness. PCIst distinguishes healthy participants from patients in a minimally conscious state and patients with unresponsive wakefulness syndrome, at low computational cost and even with data down-sampled to 250 Hz. Using one-year follow-up scores on the Coma Recovery Scale-Revised for 114 of the patients, the authors find that PCIst in the 9–12 Hz band performs best for both diagnosis and prognosis. Carried out by researchers outside the groups that introduced PCI, the study tests the approach on a patient sample far larger than that of the original 2013 paper.
2022 (March 23)	Response	Giulio Tononi, Melanie Boly, Matteo Grasso, Jeremiah Hendren, Bjorn Juel, William Mayner, William Marshall and Christof Koch publish "IIT, half masked and half disfigured", the lead commentary on Bjorn Merker, Kenneth Williford and David Rudrauf's Behavioral and Brain Sciences target article "The integrated information theory of consciousness: A case of mistaken identity". That article had argued that Φ measures the efficiency of global information transfer in differentiated networks, not consciousness. The IIT authors charge that the target article "misrepresents the foundations" of the theory and "ignores many essential publications", and they restate IIT's axioms and postulates to correct "major misconceptions". The commentary explains why IIT starts from phenomenology rather than from brain function, and why it predicts that only select physical substrates can support consciousness. It stresses that IIT's account of experience, a cause–effect structure quantified by integrated information, "has nothing to do with 'information transfer'". The commentary is one of 23 on the target article, to which Merker and colleagues reply in an authors' response.
2022 (June 4)	Publication	Giulio Tononi, Christof Koch, and four Wisconsin colleagues, including Larissa Albantakis and Melanie Boly, post "Only what exists can cause: An intrinsic powers view of free will" on arXiv, an essay that applies integrated information theory to the problem of free will. According to IIT, the presence of consciousness corresponds to a maximum of cause–effect power in the brain, and the way an experience feels to how that cause–effect power is structured. If the theory is right, the authors argue, people have free will in a genuine sense: they have alternatives, reasons, and values; they make decisions; and they, not their neurons or atoms, are the cause of their willed actions and bear responsibility for them. The argument rests on what the authors call an "intrinsic powers ontology": what exists absolutely, in physical terms, are intrinsic entities, and only what exists can cause. The essay extends IIT's claim to describe what truly exists into ethics and agency. It would remain under revision for years, reaching a fifth version in September 2026.
2022 (August)	Critique	Pedro Mediano, Fernando Rosas, Daniel Bor, Anil Seth, and Adam Barrett publish "The strength of weak integrated information theory", an Opinion article in Trends in Cognitive Sciences. Noting that IIT divides researchers, they propose separating two "flavours" of the theory. Strong IIT identifies consciousness with maxima of integrated information in any physical system. Weak IIT, without claiming an identity, treats measures of integration and differentiation as explanatory correlates of consciousness in brains. The authors argue that the formula of strong IIT is intractable, not well defined for non-Markovian systems such as the brain, and internally inconsistent, since the exclusion postulate requires comparisons with other systems while intrinsicality forbids them. Existing evidence, they contend, supports only weak IIT. The perturbational complexity index, for instance, tracks conscious level well but fails several of strong IIT's postulates, and no mathematical argument links it to ΦMax. They cite a survey in which measures of neural integration are the most favoured indicators of consciousness, while IIT itself is judged less promising than rival theories. As a path forward, they propose integrated information decomposition, which separates multiple modes of information dynamics.
2023 (February 10)	Adversarial collaboration	Lucia Melloni, Liad Mudrik, Michael Pitts, Giulio Tononi and colleagues publish in PLOS ONE the study protocol of the adversarial collaboration between integrated information theory and the global neuronal workspace theory, later known as COGITATE. It grew out of a meeting at the Allen Institute for Brain Science in early 2018 and is funded by the Templeton World Charity Foundation. The experiments and predictions had been preregistered on the Open Science Framework. Proponents of both theories endorse five contrasting predictions about conscious vision: where its neural correlates lie (posterior "hot zone" for IIT, prefrontal involvement for GNW), where its content can be decoded, whether activity persists for the duration of an experience (as IIT predicts) or appears as brief "ignitions" (as GNW predicts), the role of pre-stimulus activity, and which areas synchronize. Theory-neutral teams in six laboratories would collect the data using fMRI, M-EEG and intracranial EEG, with built-in replication. The outcomes "should either support, refute, or challenge these theories". First results would be presented in June 2023 and published in Nature in 2025.
2023 (June 19)	Critique	Michele Farisco of Uppsala University and Jean-Pierre Changeux of the Pasteur Institute, who introduced the global neuronal workspace theory with Stanislas Dehaene and Michel Kerszberg in 1998, publish in Neuroscience of Consciousness an analytical comparison between that theory and the perturbational complexity index. The index was introduced within the framework of IIT, and its success in clinical settings has often been cited in IIT's favor. The authors argue that the notion of brain complexity behind PCI fits the workspace theory's core claim that conscious processing depends on long-range connections among cortical regions, with global amplification, propagation and integration of signals. They conclude that the two are "fundamentally compatible, even though there are some partial disagreements". They cite earlier arguments that PCI lacks a clear theoretical link to IIT and so cannot confirm it. They also note that IIT's complexity-based measures predict only the level of consciousness, not its contents, and that attempts to link complexity to spatial experience remain at an early stage.
2023 (June 23)	Adversarial collaboration	Initial results of the first adversarial collaboration funded through the Templeton World Charity Foundation's Accelerating Research on Consciousness initiative, a $30 million program, are presented at the annual meeting of the Association for the Scientific Study of Consciousness. Led by Lucia Melloni of the Max Planck Institute for Empirical Aesthetics, the project pits IIT against the Global Neuronal Workspace theory. Six independent laboratories ran experiments to test predictions agreed in advance by both camps, following a method pioneered by psychologist Daniel Kahneman. IIT locates the substrate of consciousness in a posterior cortical "hot zone", while the workspace theory centers on prefrontal networks that broadcast information across the brain. The results match neither theory perfectly. Activity in the posterior cortex is consistent with IIT, but the predicted sustained synchronization between brain regions does not appear; some of the workspace theory's prefrontal predictions hold, but not all. Three judges conclude that IIT fared slightly better. The foundation, which describes the findings as "groundbreaking", stresses that neither theory has been ruled out and that further experiments are under way.
2023 (June 23)	Event	At the annual meeting of the Association for the Scientific Study of Consciousness in New York City, neuroscientist Christof Koch and philosopher David Chalmers settle a 25-year wager. In 1998, Koch had bet Chalmers that "the mechanism by which the brain's neurons produce consciousness would be discovered by 2023". Both now agree publicly that it remains "an ongoing quest", and Chalmers is declared the winner. Nature reports the outcome as "philosopher 1, neuroscientist 0". The settlement comes at the same meeting where the first results of the adversarial collaboration between integrated information theory and the global neuronal workspace theory are presented, results that support neither theory decisively. Koch, IIT's most prominent advocate after Tononi, had embraced the theory in 2012.
2023 (August)	Publication	Patrick Butlin, Robert Long and 17 co-authors, including Yoshua Bengio, Jonathan Birch, Liad Mudrik and Eric Schwitzgebel, post "Consciousness in Artificial Intelligence: Insights from the Science of Consciousness" on arXiv. The report derives "indicator properties" of consciousness from scientific theories, including recurrent processing, global workspace, higher-order, predictive processing and attention schema theories, and uses them to assess recent AI systems. It concludes that "no current AI systems are conscious", but that there are no obvious technical barriers to building systems that satisfy the indicators. Integrated information theory is deliberately left out, "because it is not compatible with computational functionalism", the report's working assumption. The authors note that IIT's proponents hold that a system running the same algorithm as the human brain would not be conscious if its components were of the wrong kind, so IIT "does not imply that some AI systems built on conventional hardware would be better candidates for consciousness than others". They consider "weak IIT", which links integration measures to global states of consciousness, to have limited implications for AI. IIT researchers would formalize their opposing view in 2024.
2023 (September 15)	Critique	A group of 124 researchers, signing as "IIT-Concerned", posts "The Integrated Information Theory of Consciousness as Pseudoscience" on the preprint server PsyArXiv. The letter was drafted by ten researchers, among them Stephen Fleming, Chris Frith, Hakwan Lau, Joseph LeDoux, and Matthias Michel. Signatories include philosophers Daniel Dennett, Patricia Churchland, and Keith Frankish, AI researcher Yoshua Bengio, Bernard Baars (originator of the global workspace theory), and Adrien Doerig, Michael Herzog, and Aaron Schurger, authors of the unfolding argument. Responding to news coverage of the June adversarial collaboration results, which had described IIT as a "leading" and empirically tested theory, the authors argue that the experiments tested only idiosyncratic predictions not logically related to IIT's core. They cite IIT's implications that an inactive grid of logic gates, organoids, early fetuses, and perhaps plants could be conscious. Until the theory as a whole is empirically testable, they conclude, "the pseudoscience label should indeed apply", warning of consequences for coma patients, AI regulation, stem cell research, and abortion. Within days, Nature reports an "uproar" in the field: some researchers argue over the label, while others worry that it will increase polarization in a discipline that has long grappled with issues of credibility.
2023 (September–November)	Response	The "pseudoscience" letter draws immediate reactions. Christof Koch says that IIT "is a theory, of course, and therefore may be empirically wrong", but that it makes its assumptions very clear, for example that consciousness has a physical basis and can be mathematically measured. David Chalmers, who had helped design the adversarial collaboration, writes that IIT "has many problems, but 'pseudoscience' is like dropping a nuclear bomb over a regional dispute", calling the label disproportionate, unsupported by good reasoning, and damaging to the whole field. Hakwan Lau, a lead co-author of the letter, defends it in an extended preprint, "What is a pseudoscience of consciousness?". He argues that it is already false to describe IIT, "a panpsychist theory", as empirically tested in any meaningful way, and that the goal of falsifying theories is "unrealistic, given where the field is at the moment". Physicist and neuroscientist Alex Gomez-Marin, writing in eNeuro, counters that "cancel culture has unfortunately landed in the sciences", and calls the pseudoscience label "a pseudo-argument akin to name-calling".
2023 (October 16)	Critique	Ignacio Cea, Niccolo Negro and Camilo Miguel Signorelli publish "The Fundamental Tension in Integrated Information Theory 4.0's Realist Idealism" in Entropy, examining the metaphysics of the theory's latest version. IIT 4.0 "openly rejects the mainstream view that consciousness is generated by the brain", treating consciousness as ontologically primary and the physical domain as merely "operational". The authors argue this position is "underdeveloped and is not rigorously formulated", inviting misinterpretation. They reconstruct it as a combination of phenomenal primitivism, reductionism about Φ-structures and their substrates, and eliminativism about non-conscious physical entities such as, presumably, atoms, neurons and rocks, which on this reading do not truly exist. That idealism sits uneasily with IIT's simultaneous claim that "something exists (and persists) independently of our own experience". After rejecting three possible resolutions, they propose reading IIT's realism as asserting only that other experiences exist beyond one's own. They call this a "non-solipsistic idealist realism", which by David Chalmers's distinction counts as a realist idealism.
2023 (October 17)	Theory	A team at the University of Wisconsin led by Larissa Albantakis, with Giulio Tononi as senior author, publishes "Integrated information theory (IIT) 4.0: Formulating the properties of phenomenal existence in physical terms" in PLOS Computational Biology. Presented as the successor to IIT 1.0 (2003–2004), 2.0 (2008–2009), and 3.0 (2014), the paper takes the existence of experience as a "zeroth" axiom and identifies five essential properties: intrinsicality, information, integration, exclusion, and composition. Each is restated as a postulate about the cause–effect power a physical substrate must have. New elements include an intrinsic difference measure, said to be uniquely consistent with the postulates, and the explicit assessment of causal relations. The theory proposes an explanatory identity: an experience is the Φ-structure of distinctions and relations unfolded from a maximal substrate. Arguing that functionally equivalent systems need not be phenomenally equivalent, the authors conclude that digital computers running even artificial general intelligence programs might experience nothing, since consciousness is about being, not doing. They concede that exhaustive calculation is infeasible for realistic systems. Tononi discloses a financial interest in Intrinsic Powers, Inc., a company developing a clinical device to assess consciousness.
2024 (January 23)	Publication	Tobias A. Wagner-Altendorf, of the Munich School of Philosophy, publishes "Progress in Understanding Consciousness? Easy and Hard Problems, and Philosophical and Empirical Perspectives" in Acta Analytica. It takes stock more than 25 years after David Chalmers separated the "easy" problems of consciousness from the hard problem. He files integrated information theory under progress on the "empirical" problem, beside the global neuronal workspace, recurrent processing and dendritic integration theories. He calls it "a somewhat more abstract and mathematical approach" in line with the emphasis on posterior cortical areas, and notes that recent versions are "explicitly based on connectivity". He cites the COGITATE adversarial collaboration as an example of rivals testing diverging predictions. On the philosophical side, he takes panpsychism as his test case and finds only weak progress. He concludes that "empirical progress and philosophical progress on consciousness are essentially uncoupled" and calls Chalmers's prediction of progress on the hard problem "overly optimistic". His classification contrasts with IIT's own framing, which starts from the essential properties of experience rather than from neural correlates.
2024 (January 26)	Publication	Robert Lawrence Kuhn, creator and host of the television series Closer To Truth, publishes "A landscape of consciousness: Toward a taxonomy of explanations and implications", a review of more than 140 pages in Progress in Biophysics and Molecular Biology. It arrays theories of consciousness on a "roughly physicalist-to-nonphysicalist landscape", with categories including materialism theories, non-reductive physicalism, quantum theories, panpsychisms, monisms, dualisms and idealisms. Integrated information theory gets a top-level category of its own, set between the quantum theories and panpsychisms rather than among the neurobiological or informational materialist theories. Each explanation is "self-described by its adherents", with minimal critique and "no attempt to adjudicate among theories". Kuhn also assesses each family's implications for meaning and value, AI consciousness, virtual immortality and survival beyond death. The INTREPID adversarial collaboration would cite the survey in its 2026 joint review as evidence of how theories of consciousness have proliferated.
2024 (May 7)	Publication	Neuroscientist Christof Koch, formerly president of the Allen Institute for Brain Science, publishes Then I Am Myself the World: What Consciousness Is and How to Expand It with Basic Books. According to the publisher, the book's heart is integrated information theory, the model Giulio Tononi first proposed in the early 2000s, presented as "the idea that the essence of consciousness is the ability to exert causal power over itself". Koch rejects the view of a person as a "Turing machine made flesh" running consciousness as software. He extends consciousness to other animals, in degrees set by how richly their neurons interconnect. He uses the theory to address when a fetus becomes self-aware and what near-death and psychedelic experiences (including his own) reveal. He also explains why generative AI may "ultimately be able to do the very thing we can do, yet never feel any of it": simulating a brain on a computer, he argues, no more makes it conscious than a simulated black hole sucks anything in. Reviewing it for Science News, Fred Schwaller calls the book "a compelling and surprisingly light read".
2024 (November 30)	Theory	Renzo Comolatti, Matteo Grasso and Giulio Tononi post "Why does time feel the way it does? Towards a principled account of temporal experience" on arXiv, extending to time the approach Andrew Haun and Tononi had applied to space in 2019. Through introspection they characterize the conscious present as an extended window composed of "moments" with four properties: each moment is directed, pointing away from itself, and moments include, connect to and fuse with one another in a directed way, from the now toward the then. Using the formalism of IIT 4.0, they show that small "directed grids", unlike the undirected grids that account for space, specify cause-effect structures with exactly these properties. On this view, the feeling of flow corresponds to a directed structure that is "static" rather than to a process unfolding in clock time: "Time, on this view, is not a process in clock time but a structure specified by the system's current state". The authors argue that this explains why experienced order can be edited or extrapolated relative to clock time, as in some auditory, language and motion illusions, and the paper would appear in iScience in August 2025.
2024 (December 5)	Theory	Graham Findlay, William Marshall, Larissa Albantakis, Isaac David, William Mayner, Christof Koch and Giulio Tononi post "Dissociating Artificial Intelligence from Artificial Consciousness" on arXiv. Prompted by suggestions that artificial general intelligence is within reach, they ask whether a computer functionally equivalent to a human would also have experiences. Using IIT, they analyze pairs of systems built from simple Boolean units, in which a basic stored-program computer simulates the other system "with full functional equivalence". They conclude that two systems "can be functionally equivalent without being phenomenally equivalent", whatever the simulated system's function, and that a digital computer could simulate human behavior, "possibly even by simulating the neurons in our brain", without replicating human experience. This, they write, "contrasts sharply with computational functionalism", the thesis that performing the right computations is necessary and sufficient for consciousness. The paper formalizes the position Koch had set out for a general audience in 2019. Critics had cited the same dissociation between function and experience as grounds to doubt that IIT can be tested, in the unfolding argument of 2019.
2024 (December 14)	Theory	Alireza Zaeemzadeh and Giulio Tononi post "Shannon information and integrated information: message and meaning" on arXiv. Claude Shannon's 1948 theory of communication had quantified information by the probability of messages, without regard to their meaning. Yet, the authors note, computational approaches regularly appeal to "information processing" to explain how meaning is encoded and decoded in natural and artificial systems. They contrast Shannon's theory with IIT, which "considers meaning as integrated information and characterizes it as a structure, rather than as a message or code". On this view, the cause-effect structure unfolded from a substrate in a given state "fully defines the intrinsic meaning of an experience and its contents". It follows that meaning can be communicated only if "the cause-effect structures of sender and receiver must be similar". The paper develops a distinction IIT had drawn from the start: Tononi and Olaf Sporns had noted in 2003 that communication theory measures how information is transmitted but not how a system integrates it, and in 2008 Tononi and Christof Koch described Φ as an intrinsic property, different from Shannon information.
2025 (February 27)	Adversarial collaboration	Kengo Takahashi, Umberto Olcese, Rafael Yuste, Karl Friston, Giulio Tononi, Cyriel Pennartz and colleagues of the INTREPID consortium publish in PLOS One the study protocol for its mouse experiment. The experiment addresses whether conscious experience depends on neuronal activity, as predictive processing theories hold, or on the causal structure of neural circuits, as IIT holds. Drawing on its account of spatial experience as generated by grid-like retinotopic networks, IIT offers what the authors call "a counterintuitive prediction": inactivating a large population of already silent neurons in visual cortex should alter the cause–effect structure and shrink perceived space. Head-fixed mice will report whether a pair of visual patches is centered left or right, while patterned optogenetics silences portions of the left primary visual cortex that show no stimulus-evoked responses in the first 200 milliseconds. IIT predicts a "functional hemineglect", even though the neurons involved fire little or not at all. Neurorepresentationalism and active inference instead hold that only active neurons contribute to experience. The consortium would set out each theory's stated confidence in this and its other predictions in a 2026 joint review.
2025 (March 10)	Response	Nature Neuroscience publishes three companion Comment pieces on IIT's scientific status. In "What makes a theory of consciousness unscientific?", the "IIT-Concerned" group behind the 2023 letter argues that IIT "is indeed unscientific because its core claims are untestable even in principle". In reply, Giulio Tononi, Christof Koch, Marcello Massimini, and 19 co-authors publish "Consciousness or pseudo-consciousness? A clash of two paradigms". They argue that IIT "starts from consciousness, which is subjective, and accounts for its presence and quality in objective, testable terms", and that the pseudoscience charge reflects "a crisis in the dominant computational-functionalist paradigm". They present IIT as a "consciousness-first" paradigm, as opposed to computational functionalism, on which computers that replicate human functions would at most be "pseudo-conscious". A third piece, by Alex Gomez-Marin and Anil Seth, "A science of consciousness beyond pseudo-science and pseudo-consciousness", rejects both labels; Seth discloses that he sits on the steering committee of INTREPID, a second Templeton-funded adversarial collaboration contrasting IIT with active inference and neuro-representationalism. The IIT authors disclose that Tononi and Koch sit on the board of Intrinsic Powers Inc., in which they and Massimini hold a financial interest.
2025 (April 30)	Adversarial collaboration	The Cogitate Consortium publishes "Adversarial testing of global neuronal workspace and integrated information theories of consciousness" in Nature, with theory leaders including Giulio Tononi, Christof Koch, and Stanislas Dehaene among its co-authors. Theory-neutral teams at twelve institutions on three continents tested three preregistered predictions, with pass/fail criteria agreed in advance, in 256 participants using fMRI, magnetoencephalography, and intracranial EEG. Information about conscious content appeared in visual, ventrotemporal, and inferior frontal cortex, in line with some predictions of each theory. For IIT, the most direct challenge is the absence of sustained synchronization within the posterior cortex, which contradicts its claim that network connectivity specifies consciousness. For the workspace theory, the prefrontal cortex showed no "ignition" when a stimulus disappeared and carried little information about stimulus identity or orientation. The authors stress that they tested the theories' proposed brain implementations rather than their mathematical cores, so proponents could revise the former while keeping the latter. Rather than issue a verdict, they invite readers to weigh the evidence, publishing the adversaries' own interpretations alongside the consortium's.
2025 (May 6)	Publication	Nature (journal) publishes an editorial, "Make science more collegial: why the time for 'adversarial collaboration' has come", alongside the COGITATE consortium's paper testing integrated information theory against the global neuronal workspace theory. Recalling Daniel Kahneman, who pioneered adversarial collaborations to combat what he called "angry science", the editors describe how principal investigators for each theory and an independent mediator agreed on a design involving six laboratories and 256 participants. Neither theory was fully validated, a finding that each side's investigators accepted. The editors then turn to the wider field. After the preliminary results appeared in 2023, an open letter described IIT as a pseudoscience, and "such language has no place in a process designed to establish working relationships between competing groups." The best way for such collaborations to advance science, they argue, is for all concerned "to accept that there is a chance that they could be wrong". The journal also publishes the peer reviewers' reports and the authors' responses. A correction on May 22 would amend the editorial's description of the letter's signatories.
2025 (October)	Theory	Andrew Haun and Giulio Tononi publish "The unfathomable richness of seeing" in Trends in Cognitive Sciences, challenging what they call a common persuasion among experts: that visual experience is sparse and its richness illusory. They distinguish three levels of visual phenomenology: high-level object and scene categories, mid-level feature groupings, and a fundamental spatial field "composed of spots and their spatial relations". Seeing an object, they argue, requires seeing the groupings that compose it, and seeing groupings requires seeing the space that grounds them. Hence "much of what we see cannot be used, reported, or remembered. And yet we see it." They also separate phenomenal vision (what it is like to see) from functional vision (what vision lets us do), holding that the former "must be vastly richer". The argument builds on the authors' earlier IIT account of spatial experience as an intrinsically structured field. Its split between experience and function echoes IIT's view that consciousness does not follow from what a system does. It runs against "sparse" views that limit experience to what can be accessed, such as Michael Cohen and colleagues' 2016 estimate of the bandwidth of perceptual experience.
2025 (December 4)	Adversarial collaboration	Jonathan Robinson, Andrew Corcoran, Anil Seth, Karl Friston, Giulio Tononi, Jakob Hohwy and colleagues of the INTREPID consortium publish in PLOS One the study protocol for its experiment centered on active inference rather than IIT. The active inference account holds that active inference is necessary for changes in conscious content, a claim neither IIT nor neurorepresentationalism makes. The protocol adapts the motion-induced blindness paradigm. In an "active" condition, participants look toward a target after it vanishes from awareness. In a "passive" condition, they keep fixation while the display is moved to replay their own earlier eye movements. Reappearance of the target is compared across conditions, and the experiment will be run at Monash University in Melbourne and replicated at Friedrich Schiller University Jena. Active inference predicts faster reappearance with active looking, with high confidence. IIT and neurorepresentationalism predict no systematic difference or a shorter time, and IIT stakes this with lower confidence. Slower reappearance under active looking would therefore cost active inference the most.
2026 (January 29)	Adversarial collaboration	Clement Abbatecola, Lars Muckli, Giulio Tononi, Karl Friston, Cyriel Pennartz, Patrick Cavanagh and colleagues of the INTREPID consortium publish in PLOS One the study protocol for its human blind spot experiment, to be run in Glasgow and Toronto. The blind spot's cortical representation lacks feedforward input from one eye. According to IIT, such "non-activatable" regions create a different cause–effect structure, so space perceived across the blind spot should be altered. The two predictive processing accounts, active inference and neurorepresentationalism, hold instead that internal models "will accommodate structural deviations" based on available sensory evidence, so judgments should change little beyond reduced precision. Participants will judge distances and areas that either span the blind spot or not, without stimulating it directly, as well as illusory motion near it, with each eye. The authors model psychometric functions for perceived versus objective space and present simulated results matching each theory's predictions. A companion protocol would test patients with cortical scotomas, and the consortium would pool all its experiments under a Bayesian scheme in a 2026 joint review.
2026 (February 25)	Adversarial collaboration	Matilda Gibbons, Ethan McBride, Yuri Saalmann and colleagues publish in PLOS One the study protocol for an adversarial collaboration testing integrated information theory against the global neuronal workspace theory in animals. It is funded by the Templeton World Charity Foundation and was shaped at the foundation's October 2019 workshop at the Allen Institute for Brain Science. Theory-neutral experimentalists will test the proponents' predictions about the location, timing and connectivity of the neural correlates of consciousness. Rhesus macaques at the University of Wisconsin–Madison and mice at the Allen Institute and the University of Pennsylvania will view or hear stimuli of varying duration in a go/no-go task that controls for report, while Neuropixels probes record from sensory, parietal and prefrontal cortex. IIT predicts that the correlates lie mainly in posterior cortex, with its lattice-like connectivity, and persist for as long as a stimulus is perceived. The workspace theory predicts prefrontal and parietal involvement, with brief "ignition" at onset. To test these predictions causally, prefrontal activity will be disrupted by electrical stimulation in monkeys and by optogenetic silencing in mice.
2026 (April 13)	Critique	Adam Barrett, Pedro Mediano, Fernando Rosas, Daniel Bor, Anil Seth and colleagues, mostly from the Sussex Centre for Consciousness Science, post "Integrated information theory: the good, the bad and the misunderstood" on arXiv, aiming to remedy "misunderstandings and misrepresentations" in the debate. They credit IIT with inspiring the perturbational complexity index and with linking neural architectures to the experience of space and time. They also reject two common objections. First, high Φ is not synonymous with "more consciousness": an inactive grid of logic gates, the example Scott Aaronson had raised in 2014, would entail only "a phenomenologically minimal form of conscious experience". Second, IIT's distinctive panpsychism, a "tiling" of space and time with substrates of consciousness, "is not problematic for the theory". But they reiterate that Φ "is not well-defined for real physical systems, and has not been computed on any real physical system", that measures applied to data are "proxies" rather than approximations, and that compatibility with fundamental physics would require reformulating IIT in terms of continuous fields.
2026 (June 16)	Critique	Vikas O'Reilly-Shah, Alessandro Selvitella, and Aaron Schurger publish "A caveat regarding the unfolding argument: implications of plasticity" in Neuroscience of Consciousness. The unfolding argument, introduced in 2019 by Doerig, Schurger, and colleagues, holds that because any recurrent neural network can be "unfolded" into a functionally equivalent feedforward network, theories that tie consciousness to causal structure, such as IIT, are either false or unscientific. The new paper offers mathematical proofs that this equivalence breaks down when a recurrent network's connection weights change rapidly through plasticity: its input–output function drifts over time, retains traces of perturbations indefinitely, and outgrows any finite set of static networks. The authors argue that this restores empirical testability to theories that build in fast plasticity, while stressing that it lends no theory credibility. For IIT specifically, they conclude that the original challenge stands, since IIT's causal structure can still be expressed in an unfolded network, and suggest that IIT could escape only by incorporating plasticity for principled rather than ad hoc reasons. Schurger is a co-author of both the original argument and this caveat.
2026 (August)	Adversarial collaboration	Andrew Corcoran, Andrew Haun, Reinder Dorman, Giulio Tononi, Karl Friston, Cyriel Pennartz and the INTREPID Consortium publish "Integrated information and predictive processing theories of consciousness: An adversarial collaborative review" in Neuroscience & Biobehavioral Reviews. INTREPID, funded under the Templeton World Charity Foundation's "Accelerating Research on Consciousness" initiative, pits IIT against two predictive processing theories: Pennartz's neurorepresentationalism and Friston's active inference account. The review lays out the rival predictions, preregistered in 2024, for a series of multi-site experiments. In mice, IIT predicts that optogenetically inactivating visual cortex neurons that are already inactive or minimally active will bias behavior as in hemineglect, while both rivals predict no change. In humans, IIT predicts that stimuli spanning the blind spot or a cortical scotoma will look closer together. Each theory also states how confident it is in each prediction, and results are to be pooled in a Bayesian scheme that scores the evidence each theory accumulates, since "no single experiment is decisive". The authors concede that the approach will not settle every controversy, as theories may accommodate unexpected results through auxiliary hypotheses.

Numerical and visual data

Google Trends

The chart below shows worldwide Google Trends data for the search term "Integrated information theory" from January 2004 to October 2026. Values are relative, with 100 marking the month of peak search interest. Interest is negligible until the early 2010s, then holds at a low but steady level of about 5 for roughly a decade. A small rise in 2023 coincides with the adversarial collaboration results and the "pseudoscience" controversy. Interest climbs sharply from 2025 to a peak of 100 in 2026, before falling back to around 30.

Wikipedia views

The chart below shows Wikipedia pageviews for the article Integrated information theory on desktop, mobile-web, desktop-spider, mobile-web-spider and mobile app, from July 2015 to September 2026. Views rise gradually from about 3,000 to 5,000 a month in 2016 to around 7,000 to 9,000 by the early 2020s. A sharp spike of over 30,000 views in the second half of 2023, driven by both desktop and mobile readers, coincides with the first results of the adversarial collaboration with the global neuronal workspace theory and the letter calling IIT "pseudoscience". The larger peaks of 2025 and 2026, reaching about 46,000 views, are mostly desktop-spider traffic, which reflects automated crawlers rather than human readers.

Google scholar

The following table summarizes per-year mentions of the term "integrated information theory" on Google Scholar, from 2004 to 2025 (as of October 2026).

Year	Mentions
2004	53
2005	57
2006	61
2007	87
2008	105
2009	136
2010	113
2011	154
2012	164
2013	194
2014	260
2015	364
2016	372
2017	443
2018	547
2019	593
2020	639
2021	692
2022	694
2023	771
2024	892
2025	1970

Meta information on the timeline

How the timeline was built

The initial version of the timeline was written by Sebastian.

Check Detail construction for full timeline in timelines, Inclusion criteria for full timeline in timelines, and Representativeness of events in timelines.

Funding information for this timeline is available.

Feedback and comments

Feedback for the timeline can be provided at the following places:

• FIXME

See also

• Timeline of consciousness research
• Timeline of David Chalmers' work on consciousness
• Timeline of artificial intelligence
• Timeline of machine learning
• Timeline of anesthesiology