Integrated information theory (IIT) 4.0: Formulating the properties of phenomenal existence in physical terms

Consciousness Through Causal Structure

Integrated Information Theory (IIT) 4.0 seeks to explain both why consciousness exists and why particular experiences feel the way they do.
Starting from the essential properties of experience, IIT derives physical requirements and mathematical criteria that could, in principle, determine whether any system is conscious, to what degree, and in what way.
According to IIT, a conscious substrate must exert causal power on itself, select a specific cause and effect as an irreducible whole, and have definite boundaries and a definite grain.
The theory attributes the quality of an experience to the specific structure of irreducible causes, effects, distinctions, and relations specified by its substrate.
IIT 4.0 refines earlier formulations with a unique measure of intrinsic information and an explicit assessment of causal relations; the authors report consistency with neurological evidence and empirical support for some core principles.

The theory’s starting point is the existence of experience.

Foundations of Consciousness Theory

Integrated information theory (IIT) starts with subjective experience itself, rather than behavioral or neural correlates, and seeks to explain its properties in objective physical terms.
IIT treats the existence of experience as immediate and irrefutable, identifying five essential properties: intrinsicality, information, integration, exclusion, and composition.
Physical existence, unlike experience, is an explanatory postulate: something exists physically if it has cause–effect power that can be observed and manipulated.
IIT translates its phenomenal axioms into physical postulates and proposes that an experience is identical to the cause–effect structure unfolded from a maximal substrate, requiring no additional ingredients.
A mathematical framework makes it possible to assess the theory’s consistency, predict the quality and quantity of consciousness and its brain substrate, and infer consciousness in beings unlike ourselves.
The disclosure states that G.T. has an executive role and financial interest in a company developing a clinical consciousness-assessment device, while asserting that this creates no conflict for this publication.

Being conscious—having an experience—is understood to mean that “there is something it is like to be” [ 2 ]: something it is like to see a blue sky, hear the ocean roar, dream of a friend’s face, imagine a melody flow, contemplate a choice, or reflect on the experience one is having.

Foundations of IIT 4.0

Integrated Information Theory (IIT) aims to explain which physical systems support consciousness, why consciousness fades during dreamless sleep, and why particular experiences feel the way they do.
IIT 4.0 refines earlier versions with more precise postulates and mathematics, an Intrinsic Difference measure consistent with those postulates, and explicit assessment of causal relations.
The article serves as a reference for IIT’s updated mathematical framework, demonstrating its internal consistency and how a substrate’s cause–effect structure can be unfolded computationally.
IIT’s zeroth axiom holds that experience exists immediately and irrefutably: even doubting its existence is itself an experience.
IIT identifies five essential properties of every experience: it exists for itself, is specific, forms an irreducible whole, has definite boundaries, and contains structured distinctions and relations.

It is irrefutable because the very doubting that my experience exists is itself an experience that exists—the experience of doubting [ 1 , 3 ].

From Experience to Physics

IIT identifies five essential properties of experience: it exists for the experiencer, is specific, unified, definite in scope, and structured by distinctions and relations.
These axioms are presented as immediately evident and irrefutable; imagining a nonunitary experience instead produces separate experiences, each unified in itself.
IIT treats the axioms as jointly necessary and complete, while rejecting space, time, selfhood, intentionality, affect, and other common features as essential because experiences without them are conceivable and achievable.
To explain regularities in experience, IIT assumes an independently existing world and defines physical existence operationally as the reliable ability to be affected and produce effects.
By starting with the smallest units possessing cause–effect power, IIT translates phenomenal axioms into physical postulates that specify requirements for a substrate of consciousness.

Trying to conceive of an experience that were not unitary leads to conceiving of two separate experiences, each of which is unitary, which reaffirms the validity of the axiom.

IIT’s Conditions for Consciousness

Integrated information theory (IIT) specifies five requirements for a conscious substrate: intrinsic, specific, unitary, definite, and structured cause–effect power.
Intrinsic information measures how a system affects itself in a particular state, while integrated information (φ) measures how irreducible its cause–effect power is to separate parts.
Exclusion identifies a maximally irreducible set of units—a complex—as the substrate of consciousness; composition describes its overlapping causal distinctions and relations.
The resulting Φ-structure corresponds to the quality of consciousness, while the sum of its distinctions’ and relations’ φ values—big Φ—corresponds to its quantity.
IIT claims these physical requirements are both necessary and sufficient for consciousness and uses them to investigate its brain substrate, loss, and recovery.
IIT proposes an explanatory identity between experience and Φ-structure: every subjective property should be fully explainable through the structure’s physical cause–effect power, without additional ingredients.

That is, there must be a one-to-one correspondence between the way the experience feels and the way distinctions and relations are structured.

IIT’s Methodological Foundations

IIT judges explanations by their scope, unity, precision, internal and broader consistency, simplicity, and testable predictions; IIT 4.0 seeks mathematically consistent formulations of its central postulates.
Although IIT begins with subjective experience, it endorses realism: a world exists independently of experience, and this assumption explains more than solipsism.
Operational physicalism identifies physical existence with reliable cause–effect power, established by observing and manipulating a substrate’s units and ideally represented by a transition probability matrix.
Operational reductionism aims to explain physical reality entirely through causal interactions among the smallest observable and manipulable units, ultimately eliminating the need to posit intrinsic properties or laws.
To explain experience, IIT evaluates existence from the entity’s own intrinsic perspective: information depends on its actual state and possible states, while external units serve as background conditions rather than parts of the entity.
The intrinsic perspective creates a trade-off: adding units can increase causal informativeness while diluting selectivity. IIT 4.0 formalizes these constraints to characterize a substrate’s cause–effect structure in a given state.

Eventually, this would leave neither room nor need to assume intrinsic properties or laws.

Formalizing Consciousness Through Causality

IIT identifies candidate substrates of consciousness by determining their maximal cause–effect states and measuring their integrated information, or system phi.
A system qualifies as a substrate of consciousness only if its integrated information is maximal compared with all competing systems that share its units.
The selected substrate specifies a Φ-structure composed of cause–effect distinctions and relations among them; this structure represents the quality of experience, while the sum of its components’ integrated information represents its quantity.
IIT 4.0 remains a work in progress but aims to characterize Φ-structures completely using measures that uniquely satisfy its postulates.
IIT equates physical existence with causal power and models substrates as finite, discrete-time stochastic systems whose units are conditionally independent given the preceding system state.
The analysis begins with a transition probability matrix describing the system’s causal evolution under interventions that initialize it in every possible state, rather than merely recording observed transitions.

IIT takes physical existence as synonymous with having cause–effect power, the ability to take and make a difference.

Identifying Consciousness Through Causation

Integrated information theory (IIT) evaluates potential substrates of consciousness through their cause–effect power, using causal models that capture relevant features rather than every physical detail.
Existence requires a system to both take and make a difference, while intrinsicality requires it to exert cause–effect power over itself; units outside the candidate system serve as background conditions.
The information postulate requires a system to specify a particular cause state and effect state, selected by maximizing intrinsic information.
Integration requires the system’s cause–effect specification to be irreducible; system integrated information is measured using a minimum partition and the lesser of its cause and effect integrated information.
Exclusion identifies a definite substrate, or complex, by selecting the candidate with maximal integrated information among overlapping systems; in the example, the two-unit system aB outranks both its subset and its three-unit superset.
The formalism is illustrated with simple simulated networks, not actual physical substrates, and ultimately tests whether cause–effect power is intrinsic, specific, integrated, definite, and structured.

According to IIT, existence understood as cause–effect power requires the capacity to both take and make a difference (see Box 2 , Principle of being).

Defining Consciousness Through Causation

Intrinsicality requires a system to exert cause–effect power within itself; units outside the system are treated as causally inert background conditions.
The information postulate requires the current system state to specify a maximal cause–effect state. Intrinsic information combines informativeness, or deviation from chance, with selectivity, or concentration of causal power on a particular state.
Larger systems face a tension between expansion and dilution: adding units may increase informativeness while reducing selectivity, so greater size does not necessarily yield greater intrinsic information.
Integration requires irreducible cause–effect power. Integrated information measures the loss of intrinsic information under the least-disruptive partition and is sensitive to weak or one-directional connections between parts.
Exclusion identifies a definite substrate of consciousness, called a complex, by maximizing integrated information among overlapping candidate systems; units, updates, and states must also have a definite grain.
Once a complex is identified, composition requires examining all its subsets to fully characterize its cause–effect structure.

Integrated information is highly sensitive to the presence of fault lines —partitions that separate parts of a system that interact weakly or directionally [ 13 ].

Composing Causal Experience

IIT holds that explaining the structure of experience requires examining the causal powers of system subsets and their relations, not just individual units or the system as a whole.
A subset qualifies as a mechanism when it both takes and makes a difference within the system, specifying cause and effect states over subsets called purviews; together these form a causal distinction.
Each mechanism’s cause–effect state maximizes intrinsic information, while its selected purviews maximize irreducibility. Only distinctions congruent with the complex’s overall cause–effect state belong to its structure.
Distinctions with congruently overlapping cause or effect states are connected by causal relations. These distinctions and relations form a Φ-structure, whose total integrated information is the sum of their individual contributions.
The formal IIT analysis first identifies complexes by finding maxima of system integrated information, then unfolds each complex’s distinctions and relations to characterize its structure.
IIT’s principle of being equates existence with cause–effect power: physically, something must be able to take and make a difference to exist.

In other words, in physical, operational terms, to exist requires being able to take and make a difference.

IIT’s Principles of Existence

IIT links existence to causal power, echoing Plato’s Eleatic principle and Dharmakīrti, but requires an entity both to affect and to be affected rather than merely either one.
The principle of maximal existence selects the overlapping candidate complex with the greatest system integrated information (φ s), excluding competing overlapping sets with lower values.
Maximal existence also helps determine a complex’s constituent units, their grain, and the distinctions and relations composing its cause–effect structure.
The complementary principle of minimal existence measures irreducibility at the weakest partition and takes the lesser of cause-side and effect-side irreducibility.
To identify substrates of consciousness, IIT evaluates subsets of a substrate as candidate systems and causally marginalizes background units so their contributions do not directly count toward the system’s intrinsic cause–effect power.

The principle states that, when it comes to a requirement for existence, nothing exists more than the least it exists .

Measuring Intrinsic Causal Power

Separate transition probability matrices are constructed for evaluating a system’s causes and effects, rendering background units causally inert and eliminating correlations they introduce.
Effect evaluation fixes background units according to the universe’s current state, while cause evaluation uses that state to infer a probability distribution over their possible prior states.
Intrinsic information measures how strongly a system in its current state selects a particular cause or effect state over itself; the states maximizing this measure form its maximal cause–effect state.
Integrated information measures how irreducible the maximal cause–effect state is under the directional partition that disrupts it least.
Systems with maximal integrated information relative to overlapping competitors qualify as maximal substrates, or complexes.
Intrinsic information combines selectivity and informativeness using interventional probabilities; unconstrained effect probabilities average equally over all possible current system states.

In this way, we quantify the cause–effect power of a system from its intrinsic perspective, rather than from the perspective of an outside observer (see Box 1 ).

Intrinsic Information Trade-offs

Intrinsic cause information uses interventional probabilities and Bayes’ rule with a uniform prior to assess which possible past states produced the system’s current state.
Informativeness measures how much a state raises an effect’s probability—or a cause raises the current state’s probability—above chance; both comparisons use forward probabilities.
Selectivity measures how strongly probability is concentrated on a particular cause or effect state, using backward probabilities for causes and forward probabilities for effects.
Intrinsic information combines informativeness and selectivity, creating a trade-off: larger systems can deviate further from chance, but their information may be diluted across more possible states.
Indeterminism reduces selectivity over effects, while degeneracy reduces selectivity over causes, even in deterministic systems. Intrinsic information is measured in intrinsic bits, or ibits, to distinguish it from typically additive information measures.

For example, as shown in [ 12 ], if an unconstrained unit is added to a fully specified unit, intrinsic information does not just stay the same, but decreases exponentially.

Measuring Intrinsic Causal Power

Intrinsic information combines informativeness and selectivity to measure how a system, in a particular state, takes and makes a difference to itself; its unit, the ibit, is a probability-weighted information value.
The principle of maximal existence selects the specific cause and effect states that maximize intrinsic information, rather than treating all possible states equally.
To assess intrinsic causal power, probabilities must come from the system’s interventional transition probabilities with a uniform prior over possible system states, not simply from observed dynamics.
A system’s intrinsic cause and effect states need not match what actually occurred before or will occur afterward, because actual dynamics can depend on external influences.
Intrinsic difference uniquely satisfies three requirements: detecting causal differences, increasing with noiseless expansion and decreasing with signal-free expansion, and capturing the causal power of specific states.
Effect-side intrinsic information is formally an intrinsic difference between constrained and unconstrained probability distributions; cause-side information uses Bayes’ rule and is not strictly equivalent, although it retains the same selectivity–informativeness structure.

Because consciousness is the way it is, the formulation of its properties in physical, operational terms should be unique and based on quantities that uniquely satisfy the postulates [ 12 , 32 ].

Measuring Causal Irreducibility

A candidate system can support consciousness only if its cause–effect power is irreducible: its units must form one system rather than independently existing subsets.
Directional partitions divide the system into nonempty, nonoverlapping parts and replace selected connections between them—inputs, outputs, or both—with independent noise.
Directional cuts test a key requirement of system membership: every subset must both affect and be affected by the rest of the system.
Partitioned transition probability matrices represent the disrupted system by causally marginalizing the cut connections while retaining the intact connections.
Integrated effect and cause information measure how much a partition weakens the system’s specification of its selected effect and cause states. Negative integrated effect values are set to zero, and ties between maximally informative states are resolved by maximizing integrated information.

In other words, to be a part of a system, a subset of units must be able to interact with the rest of the system in both directions (cause and effect).

Integration and Conscious Complexes

System integrated information is determined by the weaker of its cause-side and effect-side irreducibility, evaluated across its minimum partition.
The minimum partition is selected using normalized integrated information, allowing fair comparisons across partitions of different sizes and revealing genuine weak links rather than favoring cuts around individual units.
After selecting the minimum partition, its unnormalized integrated information measures the absolute loss of intrinsic information caused by that cut, quantifying the system’s irreducible cause–effect power over itself.
The exclusion postulate identifies a definite substrate of consciousness: among overlapping candidates, the system with the greatest integrated information forms a complex, while weaker overlapping candidates are excluded.
Complexes are identified recursively by selecting the strongest remaining candidate and removing its units from subsequent searches; removed units still provide background conditions, and overlapping ties require choosing the next-best unique system.

among candidate systems competing over the same substrate with respect to an essential requirement for existence, in this case irreducibility, the one that exists is the one that exists the most.

Finding Maximal Unit Grains

Selecting systems with maximal integrated information relative to all overlapping systems partitions the universe into a disjoint, exhaustive set of complexes.
Integrated information theory must consider both spatial and temporal grains, ranging from brain regions to subatomic constituents and from minutes to microseconds.
The exclusion postulate requires a definite grain: the winning scale is the one that maximizes the irreducible existence of the complex.
Macro units combine micro units through state mappings, but must themselves be maximally irreducible within their own substrates to qualify as genuine units.
The chosen collection of units must maximize the integrated information of the larger complex, rather than maximize each unit’s own existence.
Finding the optimal grain requires an iterative search over candidate substrates, state mappings, and intermediate scales until a global maximum is established.

Otherwise it would be possible to “make something out of nothing.”

Unfolding Causal Distinctions

Once a maximal substrate and its cause–effect state are identified, IIT unfolds their causal power into a structure of distinctions and relations.
A causal distinction consists of a mechanism in its current state, its maximal cause and effect over subsets of system units, and a positive value of irreducibility.
Identifying distinctions requires selecting cause and effect states using intrinsic information, assessing integration under a minimum partition, and choosing purviews with maximal irreducibility.
A mechanism contributes a distinction only if its maximal cause–effect state is congruent with the system’s overall maximal cause–effect state.
To isolate a mechanism’s causal contribution, other system units are causally marginalized using a uniform distribution; product probabilities further prevent externally induced correlations from being mistaken for the mechanism’s effects.

Otherwise, X might introduce correlations in Z that would be wrongly considered as effects of M .

Intrinsic Causal Information

IIT constructs cause and effect probabilities as products of individual-unit probabilities, preserving conditional independence through independent noise applied to connections.
This approach differs from standard information-flow analyses: it treats variables as physical units that can be independently observed and manipulated, and uses interventional probabilities throughout.
Intrinsic information combines informativeness—how much a mechanism or cause raises a state's probability above chance—with selectivity—how strongly causal power is concentrated on that particular state.
The principle of maximal existence selects the purview state with maximal intrinsic effect information, although this state need not be unique.
Indeterminism and degeneracy reduce intrinsic information, while enlarging a purview creates a trade-off: informativeness tends to increase, but selectivity tends to decrease.
After identifying a mechanism's specific effect, IIT evaluates integrated information to determine whether that effect is irreducible.

Moreover, the product of selectivity and informativeness leads to a tension between expansion and dilution: larger purviews tend to increase informativeness because conditional probabilities will deviate more from chance, but they also tend to decrease selectivity because of the larger repertoire of states.

Defining Irreducible Causal Distinctions

Disintegrating partitions test mechanism–purview pairs by splitting a mechanism into independent parts or severing its connections to the purview; unlike system partitions, they assess causal power within an already defined system.
A mechanism’s effect is reducible if any allowed partition leaves its effect probability unchanged or increases it. Integrated effect information is evaluated using the minimum partition, with cause information defined analogously.
Intrinsic information identifies the cause or effect state a mechanism specifies, while integrated information measures how irreducibly it specifies that state.
Under exclusion, each mechanism selects the cause and effect purviews with maximal integrated information; the causal distinction’s irreducibility is the smaller of its integrated cause and effect information.
Unfolding a system’s full cause–effect structure requires evaluating every subset of its units. Subsets with positive distinction irreducibility are candidates, but must also satisfy system-level intrinsicality and information requirements.

Again, φ ( m , Z ) is a quantifier of irreducible existence.

Intrinsic Causal Handles

A system’s causal distinctions must specify causes and effects consistent with the cause–effect state selected by the system as a whole.
Causal distinctions are irreducible “handles” linking intrinsic causes to intrinsic effects over subsets of a system; identifying them requires evaluating every subset.
From the system’s intrinsic perspective, causes and effects excluded from its cause–effect structure do not exist, even if an external observer can identify them.
Distinguishing a group does not imply distinguishing its individual members, and distinguishing individual members does not imply distinguishing them together as a group.
Distinguishing individuals and their whole group is insufficient for sequential ordering: appropriate pairwise distinctions, such as ab and bc but not ac, are also needed.
Causal relations capture overlaps among the causes and effects of distinctions, revealing structure that individual distinctions do not inherently possess.

Altogether, distinctions can be thought of as irreducible “handles” through which the system can take and make a difference to itself by linking an intrinsic cause to an intrinsic effect over subsets of itself.

Binding Causal Distinctions

Relations bind distinctions within a complex by specifying shared units in matching states across their causes, effects, or both.
This binding has its own irreducibility, measured by φr > 0; it captures causal structure that cannot be represented by a higher-order distinction alone.
Each relation has “faces” representing the different combinations of causes and effects that overlap congruently; each face’s maximal overlap defines its purview.
A relation’s degree counts the distinctions it binds, whereas a face’s degree counts its participating cause and effect purviews. A single distinction can also form a self-relation between its cause and effect.
Overlapping purviews can support multiple faces: the illustrated pair of distinctions has nine because all four cause and effect purviews agree on the same unit and state.
Neither a single undifferentiated effect nor multiple non-overlapping effects support relations; relations arise when congruent overlaps bind effects into a structured whole.

Relations thus reflect how the cause–effect power of its distinctions is “bound together” within a complex.

Binding Causal Distinctions

Causal relations bind distinctions through causes and effects that specify shared units in matching states, forming an ordered structure of overlaps.
Relations inherit existence from their distinctions’ cause–effect power and intrinsicality from causes and effects specified within the substrate; matching states make these relations specific.
A relation’s irreducibility is assessed by “unbinding” its distinctions, assuming each distinction’s integrated information is distributed uniformly across its unique cause and effect purview units.
Relation integrated information combines the size of the joint overlap with the smallest per-unit information contribution among its distinctions, ensuring that a relation cannot be stronger than its weakest distinction.
Maximal congruent overlap maximizes relation integrated information. When every distinction’s cause and effect fully overlap those of the others, the relation reaches the integrated information of its weakest distinction.

Defining φ r in this way guarantees that the integrated information of a relation cannot exceed the integrated information of its weakest distinction.

Consciousness as Causal Structure

A cause–effect structure consists of causal distinctions and the relations binding them; when specified by a maximal substrate, or complex, it is called a Φ-structure.
Structure integrated information, Φ, is the sum of the integrated information of all distinctions and relations, rather than a measure computed by partitioning the structure as a whole.
Meaningful substructures called Φ-folds describe individual distinctions and their relational contexts, groups associated with subsets of units, or highly interrelated distinctions constituting content.
IIT distinguishes the externally observable substrate from the intrinsic entity: a complex exists for itself as its fully unfolded causal structure.
According to IIT, experience is identical to a Φ-structure: the structure specifies experiential quality, while its Φ value specifies the quantity of intrinsic content, without additional ingredients.
The forthcoming examples examine how connectivity and activity shape consciousness, and why functionally equivalent systems may nevertheless differ in their Φ-structures and consciousness.

By IIT, a complex S * does not exist as such, but exists “unfolded” into its Φ -structure, with all the causal distinctions and relations that compose it.

Connectivity Shapes Integrated Information

The examples use binary units with logistic activation functions; increasing the slope parameter k makes their behavior more deterministic and less noisy.
Five six-unit networks are compared in the same state, showing how connectivity determines which subsets form irreducible complexes and what cause–effect structures they specify.
Bottleneck and modular architectures fragment into smaller complexes; the modules have low structure integrated information (Φ) because they specify few distinctions and relations.
A directed six-unit cycle forms an integrated complex, yet its cause–effect structure contains only first-order distinctions and few relations, showing that system integration alone does not ensure high Φ.
A specialized lattice specifies 27 distinctions and over 1.5 million relations, yielding Φ = 11,452 ibits; a slight connectivity change instead selects a smaller complex and excludes the full system despite its positive system integrated information.

The full system is not maximally irreducible and is excluded as a complex, despite its positive φ s value (indicated in gray).

Architecture Limits Integrated Information

A bottleneck network with high degeneracy and moderate indeterminism breaks into small complexes: one two-unit complex and four individual-unit “monads,” limiting the richness of its Φ-structure.
Many possible causes and effects reduce selectivity and intrinsic information in larger subsets. The example suggests that high Φ requires specialized units that interact reliably, characteristics attributed to the cerebral cortex.
Weak connections between strongly connected modules create causal fault lines, dividing a six-unit network into three small complexes with low Φ. The text relates this pattern to the highly modular cerebellum, which does not contribute to consciousness despite its large neuron count.
Causal fault lines can also arise physiologically: during early slow-wave sleep, neuronal bistability can render cortical connections ineffective, coinciding with loss of consciousness.
A directed copy cycle forms a six-unit complex but has low structural integrated information because it supports only first-order distinctions and few relations. Thus, a large, irreducible substrate does not necessarily yield a rich Φ-structure.

In this case, the weak inter-module connections are clear fault lines.

Connectivity and Conscious Substrates

Directed cycles lack cross-connections, limiting the richness of their Φ-structures; adding more units to a cycle does not increase its system integrated information.
Brain circuits linking cortex, basal ganglia, and thalamus support important cognitive functions but, according to the theory presented, may not contribute directly to conscious experience.
Specialized lattices with overlapping inputs and outputs can generate many distinctions and causal relations, yielding much richer Φ-structures than simple cycles.
The dense, divergent-convergent organization of posterior cerebral cortex makes it a plausible candidate for the substrate of human consciousness.
The exclusion principle selects the maximally irreducible subsystem rather than every integrated whole: a four-unit subset can qualify as a complex even when its encompassing six-unit network is highly integrated.
Consequently, integration across the brain and body does not imply that all of them constitute one conscious substrate; a posterior cortical subset may have greater integrated information than the larger system.

The brain as a whole is undoubtedly integrated (not to mention that it is integrated with the body as a whole), and neural “traffic” is heavy throughout.

Inactivity Versus Inactivation

Connectivity strongly shapes which groups of units form conscious complexes and the integrated-information structures they specify.
A substrate’s causal powers depend on its current state: changing even one unit can alter many distinctions, relations, and their irreducibility.
In the five-unit example, switching unit E from ON to OFF preserves the complex and its 23 distinctions, but changes their causes and effects and reduces relations, lowering total integrated information.
Fixing unit E permanently OFF removes its causal contribution entirely: the complex shrinks to four units and 14 distinctions, while integrated information falls from 22.26 to 3.35 ibits.
A single unit contributes not merely a bit of information but an extensive substructure called a compound Φ-fold; changing one neural unit can therefore reshape an entire portion of experience.

In other words, what a single unit contributes to intrinsic existence is not some small “bit” of information. Instead, a unit contributes an entire sub-structure, composed of a very large number of distinctions and relations.

Being Is Not Doing

IIT distinguishes an inactive unit, which retains causal potential and shapes experience, from an inactivated unit, which loses its causal power and no longer contributes to experience.
Inactivating a unit removes or changes the distinctions and relations involving it, shrinking the complex that constitutes the conscious substrate.
A fully inactive but functioning substrate could still support a richly structured experience, such as empty space or “pure presence,” whereas a fully inactivated substrate cannot.
According to IIT, consciousness depends on a system’s internal causal structure rather than its input–output behavior; identical functions need not imply identical experiences.
Three networks that count eight positive inputs exhibit identical external behavior and global state-transition patterns, yet have markedly different Φ-structures: one has high integrated information, another has less, and the third splits into three minimal complexes.

Consciousness and functional equivalence: Being is not doing

Function Versus Conscious Experience

Three systems perform the same coin-counting function and follow equivalent global state transitions, yet their different internal connections produce different cause–effect structures.
Two systems form three-unit complexes, but their structure integrated information differs markedly: Φ = 21.01 ibits versus Φ = 3.64 ibits. The third, feed-forward system is reducible and separates into three small complexes.
According to integrated information theory (IIT), equivalent behavior and global dynamics do not imply equivalent conscious experience: intrinsic causal organization matters.
Because feed-forward architectures can reproduce the behavior of discrete recurrent systems, IIT implies that even highly complex behavior can be performed without consciousness.
Digital computers could potentially emulate every human function without having human-like experiences; observable dynamics alone cannot fully reveal a system’s intrinsic cause–effect structure.

In other words, they are equivalent in what they do extrinsically, but not in what they are intrinsically.

Consciousness as Intrinsic Causation

Integrated Information Theory (IIT) begins with the existence and essential properties of experience, translating these axioms into physical postulates intended to specify necessary and sufficient conditions for consciousness.
A conscious substrate must exert causal power upon itself and specify informative, selective, irreducible, definite, and structured causes and effects; its subsets together constitute a cause–effect structure called a Φ-structure.
Consciousness depends on intrinsic causal powers rather than ongoing activity alone: even a dynamically inactive system at a fixed point could experience “pure presence.”
Indeterminism, degeneracy, and weak integration limit which substrates qualify as conscious complexes; even maximally integrated systems can have extremely sparse Φ-structures.
IIT proposes that a universal substrate resolves into disjoint, non-overlapping complexes, and that every property of experience is identical to a property of a qualifying substrate’s cause–effect structure.
Unlike accounts in which consciousness emerges from matter, IIT gives experience conceptual priority and identifies subjective intrinsic existence with causally defined, objectively observable existence.

For example, a system in a fixed point is dynamically “dead” (and “does” nothing), but it may be phenomenally quite “alive,” for example, experiencing “pure presence” (see (14) in S1 Notes ).

Consciousness Beyond Computation

Integrated information theory (IIT) identifies experience with the Φ-structure specified by a physical substrate in its current state, rather than with the substrate itself; small physical changes can therefore produce major changes in experience.
Functionally equivalent systems need not have equivalent experiences: IIT holds that physical architecture, not behavior or computation alone, determines consciousness.
Under IIT, even super-intelligent artificial systems could experience little or nothing if their hardware is feed-forward or contains critical bottlenecks, despite behaving indistinguishably from humans.
Exhaustively calculating IIT’s structures for realistic systems is impractical because of nested combinatorial explosions; complete analyses are currently limited to idealized systems with only a few units.
Approximations and bounded estimates are important research directions, while IIT-inspired complexity measures have already supported its prediction that consciousness is associated with higher Φ in wakefulness and dreaming than in dreamless sleep or anesthesia.

Said otherwise, consciousness is about being, not doing [ 1 , 29 , 55 ].

Testing Consciousness and Meaning

Integrated Information Theory (IIT) predicts that the brain’s conscious substrate, its boundaries, and its unit scale should correspond to the system with maximal integrated information (φ s), matching clinical and experimental evidence.
IIT distinguishes inactivity from inactivation: mostly inactive constituents can still support conscious contents, whereas fully inactivated constituents cannot; their complete inactivation should produce unawareness that anything is missing.
The quality of experience depends on the substrate’s Φ-structure and architecture. Two-dimensional neural grids offer an account of spatial extendedness, while explanations of experienced time and objects remain under development.
Changes in neural connection strength should alter experience even without changes in neural activity; similarities between experiences should likewise reflect similarities between their underlying Φ-structures.
Although many predictions sound qualitative, they rely on quantitative analysis. Further validation requires ongoing exchange among phenomenology, theory, and neuroscience, alongside refinement of unresolved mathematical choices.
IIT identifies meaning with feeling and with substructures of a current Φ-structure, making all meaning intrinsic—even in dreams. Further work seeks to explain how intrinsic meanings capture relevant causal features of the environment.

According to IIT, the meaning of an experience is its feeling—whether those of spatial extendedness, of temporal flow, or of objects, to name but a few (“the meaning is the feeling”).

IIT’s Implications and Limits

Integrated Information Theory (IIT) must explain how shared, objectively validated knowledge can arise despite meaning’s intrinsic and partly individual nature; empirical brain evidence is essential to establishing its credibility.
IIT predicts that typical computer architectures cannot support consciousness merely by reproducing human behavior, while animals with very different behavior may be highly conscious if their brains have suitable architectures.
The authors argue that IIT implies genuine free will: conscious individuals, rather than their neurons, truly cause their willed actions and bear responsibility for the consequences.
IIT aims to develop a unified account of nature grounded in intrinsic existence, linking consciousness with actual causation, meaningful information, and the evolution of causal powers.
The theory remains vulnerable to challenges concerning its compatibility with physics, the soundness and uniqueness of its axioms, and—most importantly—empirical findings about consciousness and its brain substrate.

Because only what truly exists (intrinsically, for itself) can truly cause, we, rather than our neurons, cause our willed actions and are responsible for their consequences [ 18 ].

Consciousness Theory References

This bibliography traces integrated information theory from early formulations and measures of information integration to later accounts of consciousness and its physical substrate.
Referenced work develops mathematical treatments of integrated information, including discrete dynamical systems, mechanism-level and system-level measures, and intrinsic information.
Several citations connect the theory to subjective experience, including qualia and the distinctive character of spatial experience.
Other references explore causation, causal reductionism, intrinsic ontology, and free will, extending the discussion beyond consciousness alone.
The bibliography includes work distinguishing intelligence from consciousness in artificial systems and measuring integrated information in quantum mechanisms.
Empirical references address the breakdown of cortical effective connectivity during sleep and the use of brain complexity to classify unresponsive patients.

Only what exists can cause: An intrinsic view of free will.

Consciousness and Causal Architecture

This bibliography section maps research relevant to consciousness, causal structure, and neural function rather than presenting a continuous argument.
References on causal influence, information flow, and causal composition provide tools for analyzing how systems generate effects.
Studies of integrated information across scales and causal emergence explore whether macro-level organization can surpass micro-level descriptions in causal power.
The list includes both a critique of the general applicability of integrated information measures and computational resources such as the PyPhi toolbox.
Neuroscience references examine functional specialization, visual coding, and reliable neural activity amid noise, alongside clinical cases of people living without a cerebellum.
Works on Plato and Dharmakīrti, plus unpublished studies of time’s apparent flow and conceptual binding, broaden the bibliography’s philosophical and phenomenological scope.

Can the macro beat the micro?

Consciousness Research References

This section is a bibliography rather than a narrative, bringing together neuroscience, computational modeling, and mathematical approaches to consciousness.
Cited neuroscience studies examine cortical responses during non-REM sleep and after focal brain injury, alongside evidence connecting consciousness with neural complexity.
Several references address brain circuit organization and the debate over whether the neural correlates of consciousness lie in the front or back of the cerebral cortex.
Theoretical works explore integrated information, intrinsic cause-effect power, and the evolution of integrated causal structures in artificial agents.
Other citations concern falsifying theories of consciousness, alternative computational accounts, and connections to physics, quantum theory, and dynamical systems.

Bistability breaks-off deterministic responses to intracortical stimulation during non-REM sleep.

Integrated Information Theory Reference

The cited article presents informational structures and informational fields as a prototype for describing the postulates of Integrated Information Theory.
The article was published in the journal Entropy in 2019, volume 21, issue 5, with article number 493.
The citation includes the PubMed identifier 33267207 and links to the article, PubMed/NCBI, and Google Scholar.
This chunk contains only a bibliographic reference, so it provides no substantive explanation, findings, or evaluation of the proposed framework.

Informational Structures and Informational Fields as a Prototype for the Description of Postulates of the Integrated Information Theory.