THE ENTANGLED LOOP: The argument in three steps
The trilogy makes one argument from three angles. The first paper unifies three mathematical languages of brain dynamics. The second draws out what this implies for how consciousness is built. The third tests the framework against adolescent mental health.
Each step opens with an executive view that stands alone. The rest unpacks how it works, at whatever depth you want.
Step one: Three mathematical languages, one operator
Executive view
For more than a decade, brain scientists have used three different mathematical languages to describe what the brain is doing. Each captures something real. Each has its own community. Each tells a partial story.
The first paper of the trilogy shows that the three languages are not three independent methods. They are three readings of one underlying mathematical object, the natural vibrational structure of the brain’s wiring. Three stories, told three ways, of the same thing.
This unification has one principal control parameter, called the spectral gap. The gap works in a brain the way the silence between a choir’s fundamental and the polyphony of higher voices works in a piece of music. A clear gap lets the polyphony sing. A blurred gap dissolves it into noise.
Humans, more than any other species so far studied, have a small set of unusually long-range connections that pull the spectral gap open. These long-range connections are evolutionarily recent and disproportionately influential. They are, in a precise sense, what gives humans the architectural margin to sustain rich conscious experience on a twenty-watt energy budget.
We test this picture with a pharmacological experiment using LSD. The prediction the unification makes is that any perturbation that re-tunes the underlying operator should shift all three of the brain’s apparently independent signatures in lockstep, through a single coupling parameter. It does. The unification holds.
Why this is needed
Three apparently separate mathematical traditions have grown up around how to describe brain activity. Each captures one part of the picture.
The first tradition reads activity in space. It builds on the idea that the brain’s wiring imposes a set of natural patterns, in the way that a metal plate’s shape imposes a set of natural vibration modes, and that any pattern of cortical activity can be decomposed into a weighted sum of these natural modes.
The second tradition reads activity in time. It treats each brain region as a self-sustaining rhythmic oscillator, and tracks the multi-scale patterns of local synchrony that flow across the brain’s surface, the way a flock of starlings forms and dissolves coordinated waves.
The third tradition reads activity in spacetime interference. It is the most mathematically unusual of the three, and it captures the structured ways in which the brain’s natural modes can interfere with one another to bind perceptual, affective and inferential content into one coherent experience.
If the three traditions were really independent, the brain would be three separate puzzles. The unification shows that they are three views of one puzzle. This matters because it changes the question. Instead of asking which framework is right, the field can ask what controls the underlying structure that all three frameworks read.
The three languages: A guided tour
The three subsections that follow take each language in turn. The opening line gives the picture in one breath. The rest unpacks it.
Connectome harmonics: The Chladni plate of the cortex
The brain has its own natural vibration modes, set by the geometry of its wiring. Any pattern of cortical activity can be decomposed into a weighted sum of these modes.
In 1787, Ernst Chladni sprinkled fine sand on a metal plate and bowed its edge. Wherever the plate vibrated, the sand was thrown clear. Wherever the plate stayed still, the sand collected. At each note the plate liked to sing, the sand traced out an intricate pattern of curves and dots, with mathematically precise symmetry. The patterns are the plate’s own natural vibration modes, set by its shape.
The connectome harmonics are exactly the same idea, but for the brain instead of a metal plate. The “shape” that sets the brain’s natural modes is not its physical outline but the geometry of how its regions connect to each other through white-matter tracts. The first nontrivial mode happens to coincide with the principal organisation of the cortex, the gradient from primary sensory areas at one end to the most associative areas at the other. Higher modes add progressively finer spatial detail. Together they form an alphabet in which any pattern of cortical activity can be spelled out.
What changes between brain states, between people, and between brains in health and disease, is not whether the cortex has these natural modes but which of them are loudest. The harmonic decomposition reads brain activity in the connectome’s own language, rather than imposing a coordinate system from outside.
Turbulence in coupled oscillators: The murmuration of starlings
Each brain region is a rhythmic oscillator coupled through the connectome. The turbulence framework tracks how these oscillators fall into and out of multi-scale patterns of local synchrony, in the way a flock of starlings falls into and out of coordinated wavefronts.
Watch a flock of starlings at dusk. Each bird flaps at its own rhythm but is also responding to the few birds nearest it. From these simple local rules, the flock produces astonishing multi-scale patterns. Small groups turn together. Larger regions form wavefronts. Sometimes the whole flock briefly aligns into a single ribbon. The patterns of who is moving in concert with whom keep shifting. The coherence sits at many scales at once.
This is what coupled-oscillator turbulence in the brain looks like. Each brain region is, in the framework, a self-sustaining rhythmic oscillator, much like a single bird. The connectome is the network of who is connected to whom. At any instant, local regions of the cortex are oscillating in coordinated phases, others are out of phase, and the multi-scale pattern of synchrony keeps shifting. Turbulence in this framework measures how rich and variable this pattern is.
A note on what this is not. This is not fluid turbulence. There is no flow of momentum, no Kolmogorov cascade, no Reynolds number. What is shared with fluid turbulence is the mathematics of multi-scale variability and the look of metastable, scale-spanning coherence. The substrate is coupled phase oscillators, not a fluid, and the brain is not flowing.
Complex harmonics: The Schrödinger reading
The complex harmonics form, CHARM for short, is the Schrödinger reading of the same operator whose heat reading is the standard harmonic decomposition. It makes interference between the brain’s natural modes structurally visible.
This construct is the easiest to misread, so it deserves some care.
In ordinary quantum mechanics, the Schrödinger equation governs how the wavefunction of a quantum system evolves in time. It is what tells us how electrons move in atoms, how photons evolve in cavities, how particles in superposition spread and interfere before any measurement is made.
None of that is what we mean here. The brain we are describing is a classical system of coupled oscillators, not a quantum system. There is no Hilbert space, no measurement collapse, no microscopic quantum mechanics in neural tissue. What is borrowed from quantum mechanics is exclusively the mathematical structure of evolution that preserves energy rather than dissipating it. The mathematics travels; the metaphysical commitments do not.
The substantive move is the following. The standard way of evolving the brain’s natural modes treats them like heat spreading through a metal bar. The modes decay over time. Higher modes decay faster. Interference between modes is washed out by the decay before it can do any work. CHARM replaces this heat-like evolution with the conservative one captured by the Schrödinger equation, through a mathematical move known as a Wick rotation. The modes no longer decay. They keep their energy, and they carry complex phases that let them interfere coherently with one another. CHARM is the Schrödinger reading of the same operator whose heat reading gives the standard harmonics. It makes interference visible where dissipation had washed it out.
One operator, three readings
The three apparently separate mathematical languages are three operations on one underlying object, the connectome’s natural vibrational structure.
The connectome harmonics are the eigenmodes of this structure. Turbulence in coupled oscillators is read through what is called its resolvent. CHARM is its unitary propagator. Three operations, one operator.
The bridge that makes this exact, rather than approximate, is a well-known mathematical fact. The “exponential distance rule” that has long been used to describe how brain regions communicate over distance is, mathematically, precisely the Green’s function of a screened Laplacian. That is, the operator whose eigenmodes are the harmonics is the same operator whose resolvent is the turbulence smoothing kernel. Two of the three languages are reading the same object before the unification is even noticed. The third, CHARM, completes the picture.
Reading the operator from data: Whole-brain models
The empirical data we have are BOLD signals from functional MRI. The operator we want to read lives inside the data rather than on its surface. This is the role of whole-brain modelling, the framework described in the open-access book Whole-brain modelling: Cartography of the dynamics of mind by Deco and Kringelbach (Oxford University Press, 2025). The book is freely downloadable.
A whole-brain model is a simulation of brain activity built from three ingredients. The first is a parcellation, the choice of how to divide the brain into regions that will serve as the model’s nodes. Common choices for the human cortex include the Desikan-Killiany 68-regions parcellation and the Schaefer parcellations at various scales (100, 200, 400 and 1000 regions). Species-appropriate alternatives exist for non-human primates and mice. The parcellation determines what counts as a region in the model, and a finer parcellation gives a more granular operator. The second is the anatomical wiring between these regions, typically estimated from diffusion-MRI tractography. The third is a model of what each individual brain region does on its own, the local dynamics.
Local dynamics. The local node model is a free choice within the framework. Available options include spiking neuron models, dynamic mean field models with biophysical detail (sometimes including AMPA, GABA and NMDA synapses and PET-derived neurotransmitter receptor maps for the modulation of regional gain), neural mass models such as Wilson-Cowan, and reduced models such as the Hopf normal form, the Stuart-Landau oscillator. Each is suited to different scales of question. The trilogy uses Stuart-Landau oscillators throughout, because at their natural working point they are equivalent to a biophysically realistic dynamic mean field of quadratic integrate-and-fire neurons, and they admit a linearisation that yields analytical results.
Fitting. The whole-brain model is fitted to the empirical brain data by iteratively adjusting the coupling between regions, starting from the raw anatomy. The fitting objective is to match both the standard correlations between regions and a set of time-shifted correlations that capture which regions tend to lead and which tend to follow.
Why the time-shifted correlations matter. Ordinary correlation is symmetric. A’s correlation with B equals B’s correlation with A, so a fit based on correlation alone cannot tell whether A drives B or B drives A. Time-shifted correlations are inherently asymmetric, because A’s effect on B in the future is generally not the same as B’s effect on A in the future. Including them in the fit breaks the symmetry of the coupling matrix and produces a directed, hierarchical effective coupling rather than a symmetric undirected one.
Generative effective connectivity. The matrix that comes out of this procedure is called the generative effective connectivity, or GEC. Each individual whose scan is fitted gets their own. GEC is neither the raw anatomy nor the empirical correlations. It is the coupling the model needs in order to reproduce what the brain actually generates, given that brain’s anatomy and the modulatory state it is currently in. The unification’s operator is read from this individualised, directed GEC, and that is what makes the framework empirically tractable rather than abstract.
The punchline: The spectral gap, long-range connections, and what makes humans human
The unification has one principal control parameter, the spectral gap. Re-tune the gap and you re-tune all three readings of the operator at once. Humans have a small set of unusual long-range exceptions to EDR that open the gap, and that is the structural signature of what allows rich conscious experience.
The intuition for the spectral gap is musical rather than numerical.
Think of the operator as a choir. Every choir has a fundamental tone, the deepest sustained note from which the other voices take their bearings. Above the fundamental sits the polyphony, the rich texture of higher voices in counterpoint. The spectral gap is the silence between the fundamental and the lowest of the upper voices. A clear gap lets you hear the fundamental as the anchor of the music. A blurred gap dissolves the architecture of the piece into a wash.
For the connectome, the fundamental is the principal mode, the unimodal-to-transmodal gradient that organises everything else. The polyphony above is the rich set of higher harmonics in which structured patterns of activity, and ultimately conscious content, are carried. The gap between them is what makes the polyphony intelligible rather than chaotic.
Re-tune the gap, and all three readings of the operator re-tune at once. The hierarchy carried by the harmonics deepens or shallows. The richness of multi-scale phase coherence carried by turbulence rises or falls. The interference between modes carried by CHARM strengthens or weakens. One knob, three signatures.
Long-range connections, and why they matter for humans
The brain’s connections mostly follow the exponential distance rule: Each region talks mostly to its neighbours, with strength falling off sharply with distance. But humans, more than any other species so far studied, have a small set of unusual connections that violate this rule. They are long-range exceptions, like express trains threaded through a network of local stops. They are evolutionarily recent and disproportionately influential.
When the brain’s effective coupling is reconstructed by fitting a whole-brain model to data, the spectral gap of this reconstructed operator opens up relative to the raw anatomy. The fundamental separates more cleanly from the polyphony above it. The modes become well-defined enough to interfere coherently rather than chaotically. The hierarchy deepens. The long-range exceptions are part of what does this work, by breaking the local-only character of the exponential distance rule and creating the modal separation that lets the polyphony become legible.
The evolutionary picture is then clear. To support rich conscious experience on a small metabolic budget, evolution had to find a way to open the gap, so that the polyphony above the fundamental could carry structured information rather than dissolve into noise. The long-range exceptions, particularly amplified in humans, are how that gap was opened. They are why we, more than the species whose connectomes follow the exponential distance rule strictly, can sustain the structured quantum-like binding inference that the Entangled Loop describes as architecturally central. They are, in a precise sense, the structural signature of what makes humans human.
Testing the unification with LSD
The unification makes a falsifiable prediction. A pharmacological perturbation should re-tune the operator through one scalar coupling, and two mathematically independent functional domains should shift together at the same coupling value. They do.
If the three languages really are three readings of one operator, then any structural perturbation that re-tunes the operator must shift all three signatures in lockstep, through a single coupling parameter. The paper tests this with a pharmacological perturbation by LSD, applied through a map of 5-HT2A receptor density.
A single scalar coupling, when fitted, predicts both the multi-scale shift in turbulence and the macroscale redistribution of harmonic energy. These two effects live in mathematically independent domains. They have no reason to coincide at the same coupling value unless the underlying operator is one rather than three. They coincide.
[Read the One Operator paper.](papers/#one-operator)
Step two: An architecture of consciousness
Executive view
If three of the principal mathematical languages of brain dynamics are three readings of one operator, what follows for the brain that uses them?
The trilogy’s second paper offers an architectural theory of consciousness, the Entangled Loop. It builds on the Beautiful Loop framework of Laukkonen, Friston and Chandaria (2025), in which conscious experience is a self-evidencing inference in which a brain models the world and itself in the same loop.
The Entangled Loop adds three architectural features. All three are derived from a single thermodynamic constraint, the requirement to run a rich conscious experience on twenty watts. They are not three features that happen to coexist. They are three readings of one solution that evolution found to a hard problem.
The three features are:
1. Emotion as the foundation. Emotion is evolution’s solution to a hard computational problem. To survive, a brain has to navigate a vast space of possibilities faster than any serial enumeration would allow. Emotion compresses that space into a graded summary that lives in the body. Far from being a content that consciousness happens to represent, emotion is the architectural ground from which the rest of the world model takes its shape.
2. Quantum-like binding inference. Cognitive science has settled, over the last twenty years, around the idea that the brain is a Bayesian inference machine. It holds prior beliefs, sees evidence, and updates the priors in proportion to how informative the evidence is. The Entangled Loop accepts the Bayesian frame and extends it. The binding inference by which conscious content is unified across the brain operates in a regime mathematically richer than Bayesian, one that allows the brain to evaluate many hypotheses in parallel through interference between modes. Classical Bayesian inference is recovered as the special case where this interference vanishes. The quantum-like regime is a strict superset.
3. Hierarchical orchestration. Above the basic computational machinery, the architecture has an integrative level that listens widely and coordinates the rest. The framework calls it the council of wise elders. It is a function, not a fixed anatomy. It listens, integrates, and rebalances when one voice dominates or another falls silent.
The thermodynamic argument that ties them together is the deepest claim of the paper, and what makes the architecture more than a stipulation. And once the architecture is understood, a different question becomes possible: What is it for? The framework’s answer is flourishing.
Emotion: The evolutionary and computational foundation
Emotion is the evolutionary and computational foundation of consciousness in mammalian brains, not merely one of its contents. It is the compression that lets a brain navigate the world fast enough to keep its owner alive.
In most current theories of consciousness, emotion is treated as a content the conscious mind happens to represent. The Entangled Loop reverses this. Emotion is not painted onto a neutral cognitive substrate. It is the architectural ground from which the rest of the world model takes its shape.
The argument is computational at root. The world arrives at the senses far faster than any serial chain of neurons could analyse option by option, and an organism deciding what to do has to decide quickly or not at all. Evolution found a compression trick. The torrent of incoming signals is collapsed to two axes: Whether the thing is good for me or bad for me, and how strongly the body should respond. Together these two axes define what affective neuroscience calls core affect.
The compression is computed by the brain. It manifests in the body, in the rising heart rate, the prickling skin, the catch in the breath, the gut tightening or releasing, but emotion is not located in the body. The body is one of the channels through which the brain runs the compression and one of the streams the brain reads back as part of completing it. William James proposed in 1884 that what we call emotion is the perception of bodily change. Modern affective neuroscience has refined this, with the qualification that core affect is itself architectural. Core affect is a feature of the brain’s computational organisation, not a derivative of body monitoring, and the body is a participant in its operation rather than its seat.
The distinction between emotion and feeling matters here. Emotion is the compressive operation the brain runs as the architecture does its valenced work. Feeling is what arises in consciousness when this operation surfaces, broadcast through the integrative level of the architecture into the unified field we call experience. The two are usually conflated in everyday language but the order matters. The architecture compresses. The body changes. The compression broadcasts. The feeling forms on top of work the architecture has already done.
The pleasure cycle
The architecture runs through a characteristic three-phase cycle. Each phase has a felt quality, a brain signature and a function.
Wanting is the anticipatory drive that orients an organism toward something it has identified as good for it. It is what most people call desire. The lean of attention, the readiness to act, the sense of pull toward a goal.
Liking is the consummatory pleasure at the moment of getting it. It is what most people call pleasure proper. The warmth of the bite, the relief of the embrace, the hedonic peak.
Satiety is the post-consummatory sense of completion that lets the cycle close. It is what most people call contentment. The fullness that signals enough, the calm that releases attention back to other things.
Above the basic three-phase cycle sits a higher-order sense of significance, the recognition that the cycle has been worth running. This higher-order layer is where meaning lives, where the cycle of momentary pleasure connects to a longer sense of a life well lived. The framework calls this flourishing, and we will return to it as the practical horizon of the architecture.
Quantum-like binding inference
The brain’s binding of conscious content uses the mathematics of interference between modes. Classical Bayesian inference is recovered as the special case where the interference vanishes.
A word first on Bayesian inference, since the framework presupposes it. Bayesian inference is the mathematical rule by which a reasoner updates beliefs in light of evidence. The reasoner starts with a prior belief about how the world might be. Each new piece of evidence is weighted by how informative it is, and the prior is revised into a posterior. Cognitive science has, over the last two decades, increasingly converged on the idea that the brain is essentially a Bayesian inference machine of this kind. Predictions are run against incoming signals, and the model is updated whenever the predictions miss. The Entangled Loop accepts the Bayesian frame and extends it.
The phrase “quantum-like” needs care. It does not refer to quantum biology, to microscopic quantum effects in neural tissue, or to the Penrose-Hameroff proposal that consciousness involves quantum mechanics in microtubules. The brain in this framework is a classical system of coupled oscillators. The label “quantum-like” is borrowed from a specific tradition in the foundations of probability, in which classical systems can produce interference effects and violations of the law of total probability that mathematically resemble quantum probability.
What the framework claims is that the brain’s binding inference, the process by which conscious content is unified across regions, operates in this quantum-like regime. Many hypotheses can be evaluated in parallel, by being superposed in a single phase field whose interference structure carries the answer, rather than being computed one after another. Classical Bayesian inference is recovered as the limit where the interference terms vanish. The quantum-like regime is a strict superset. It does everything Bayes does, plus interference effects that no finite classical circuit can reproduce, at lower thermodynamic cost.
Hierarchical orchestration: A council of wise elders
Above the basic computational machinery sits an integrative level that listens widely and rebalances. The Entangled Loop calls it the council of wise elders. The framework needs this level as a function, not as an anatomy.
Every functioning culture has a group of senior figures whose role is not to do everything themselves but to coordinate everyone else: To listen widely, weigh competing voices, integrate information from across the community and shape the collective response. The brain has a functional analogue. A higher level of organisation listens to many parts of the architecture at once, integrates what it hears, and rebalances the system when one voice dominates or another falls silent.
This higher level is the council of wise elders. What matters for the framework is the architectural fact that such an integrative level exists, not the specific anatomy that implements it. The substrate may turn out to involve particular regions, particular dynamics or particular relationships between regions and dynamics. Different research traditions have proposed different candidates and the empirical identification is still being worked out. The framework requires the function, not the anatomy. When the council does its work, conscious experience is coherent and flexible. When it falters, the dysregulation that follows is one of the architectural signatures of neuropsychiatric disorder.
Thermodynamics: The one constraint that selects all three
Emotion as architectural foundation, quantum-like inference and hierarchical orchestration are not three separate desirable features that happened to coevolve. They are three readings of one architectural solution to running a conscious system on twenty watts.
The brain runs on approximately twenty watts, less than a dim incandescent bulb. Every architectural feature the brain has, from the layered cortex to the long-range connections, has been selected under this constraint. Anything that wastes energy is selected against. Anything that delivers more computation per joule, or more computation per millisecond, is selected for.
A classical Bayesian system that needs to evaluate many hypotheses must compute each one in turn, paying a thermodynamic cost for each computation in sequence. A quantum-like system can superpose many hypotheses in a single phase field and let them interfere with each other coherently, evaluating the whole space in parallel through the structure of the interference. The result is more computation per joule and more computation per millisecond. Exactly what a brain on twenty watts, making decisions in real time, requires.
The trilogy quantifies this argument through the cost-of-cognition framework developed in earlier work, in which the thermodynamic cost of computation can be measured directly from the structure of the brain’s dynamics. Quantum-like dynamics are demonstrably more efficient than the classical alternatives, particularly at the integrative level of the architecture where binding inference happens.
Emotion provides the compression that keeps the decision space tractable. Quantum-like inference provides the parallel evaluation that keeps the computation fast. Hierarchical orchestration provides the coordination that keeps the whole system coherent. The three features are three readings of one physical operator unfolding under thermodynamic constraint along the arrow of time.
Flourishing: What the architecture is for
The framework’s practical horizon is flourishing, the meaningful pleasure that the architecture is capable of when its components run cleanly together. Flourishing is achievable but it is not a plateau. It is a dynamical state that has to be continually renewed.
Once the architecture is understood, a different question becomes possible. What is the architecture for?
The Entangled Loop’s answer is that the architecture is for flourishing. By flourishing the framework means something more than transient pleasure. It means the meaningful pleasure that arises when the pleasure cycle runs cleanly through the council of wise elders, with the wanting, the liking and the satiety coordinated, and the higher-order sense of significance laid over the top. Aristotle’s term eudaimonia, the life well attended by a good guiding spirit, names what is meant. The Centre for Eudaimonia and Human Flourishing at Linacre College, Oxford, where the Entangled Loop programme is based, takes its name from this Aristotelian aspiration.
The framework’s distinctive claim about flourishing is that it is intrinsically unstable. The brain’s dynamics are intrinsically unstable, and the architecture cannot hold a static state for long. Flourishing is therefore not a plateau one reaches and occupies. It is a dynamical state that must be continually renewed, lost and recovered, and this instability is not a failure but the condition of flourishing. The practical task, both clinical and personal, is therefore not to secure flourishing permanently. It is to develop the capacity for its repeated achievement.
Five reliable routes into the meaningful states that mark a flourishing life have been identified through the Centre’s work: The sensorium, the company of others, music, meditation and psychedelic experience. Each of these runs through the same architecture the Entangled Loop describes. The framework’s offer to the clinic is therefore concrete. For a given person, in a given state, the question becomes which of the routes the architecture can support, where the dysregulation lies, and how to rebalance the orchestrating council so that the pleasure cycle can run cleanly again.
[Read the Entangled Loop.](papers/#entangled-loop)
Step three: A first application to adolescent mental health
Executive view
If the architecture of consciousness is right, then disturbances of the architecture should be readable as disturbances of its measurable signatures.
The third paper of the trilogy tests this prediction in adolescent mental health. Mood and anxiety disorders typically emerge in adolescence, yet they are usually identified only after symptoms have consolidated, when intervention can only be reactive. A marker that registered the loss of healthy brain function before symptoms appeared would change this.
The paper reports such a marker, and calls it a canary in the mind, after the caged canaries once carried into coal mines that warned miners of toxic air before any human could detect it. The mathematical name is the Växjö Interference Connectivity, or VIC. It is computed from a single baseline brain scan, using the same whole-brain modelling and the same unified operator established in step one. It reads the brain’s interference fingerprint, not its average activity or its correlation structure. In a sample of 150 adolescents aged 14 to 17 from the HCP BANDA dataset (Boston Adolescent Neuroimaging of Depression and Anxiety), with both baseline resting-state fMRI and one-year follow-up clinical assessment on the Revised Children’s Anxiety and Depression Scale (RCADS), VIC predicts the follow-up score from the baseline scan alone at r = 0.60. The cortical contributions to the prediction are concentrated in the parts of the architecture associated with the integrative council of step two, the level at which the framework expects affective broadcast and modal interference to be coordinated.
The result is promising but not yet settled. The sample is internally cross-validated rather than externally replicated, and the BANDA sample was itself drawn from a larger release of 215 adolescents on motion and follow-up-availability criteria. External replication on an independent longitudinal dataset is the immediate next step.
The mechanism, which the rest of this section unpacks, is that VIC reads architectural tuning, not change. It tells you which adolescents enter the year with the architectural margin to weather it, and which do not.
Why VIC is so sensitive
VIC reads from a different mathematical channel than ordinary functional connectivity. Phase relationships between the brain’s natural modes amplify small structural changes into measurable signals, in the way that interferometric measurements outperform direct ones.
Ordinary functional connectivity is a second-order statistic of the BOLD signal, averaged over many time points and largely insensitive to the phase relationships between regions. Most of the brain’s information is averaged away by the time the correlation matrix is computed. This is why the Marek ceiling exists for functional-connectivity-based prediction.
VIC sits in a different part of the observation space. It reads the same data but through a different mathematical channel, the unitary-propagator structure of the fitted operator that step one’s whole-brain model recovers. That channel is more sensitive to small structural changes than the covariance channel.
The amplification comes from the mathematics of interference itself. Interferometric measurements of small displacements, of the kind used in gravitational-wave detectors to register spacetime distortions smaller than a proton, are sensitive in a way that direct measurements of position are not. Phase relationships amplify small geometric differences into measurable phase shifts. VIC is brain interferometry in the same mathematical sense. It is not better engineering. It is better physics.
The conductor’s ear
A choir starting to drift out of tune loses its phase precision before it sings wrong notes. A trained conductor hears the drift first.
Mental health problems in adolescence are similar. The architecture begins to lose precision before the symptoms surface. The phase relationships between the modes drift slightly. The spectral gap begins to close from its healthy spacing. The long-range structure starts to be used a little less effectively. None of this shows up in mean activity or in correlation, because each individual mode is still essentially functioning. It shows up in the interference fingerprint, because the interference fingerprint is sensitive to relationships between modes rather than to modes one at a time.
VIC is the conductor’s ear. It hears the drift before the audience does, which is to say, before symptoms surface.
Architectural tuning, not change detection
VIC is not detecting drift from a previous reading. It is reading the architectural tuning of the brain at one moment, which predicts how the system will weather the coming year.
Two adolescents may have indistinguishable mean activity, indistinguishable functional connectivity and indistinguishable static anatomy, and yet have measurably different architectural tunings. One has a clean spectral gap that opens cleanly under the fitted whole-brain model. Its principal modes are well separated. Its modal interference is structured. Today, this adolescent looks fine, and is fine.
The other has a partly muddled gap, less cleanly separated modes, and less coherent interference. Today, this adolescent also looks fine. They will pass any standard clinical assessment. But their architectural margin is smaller. Their capacity to absorb the affective and developmental demands of the coming year, the disrupted sleep, the social stresses, the brain remodelling, is smaller. The first adolescent will probably weather the year. The second is at higher risk of crossing into clinically significant depression or anxiety.
VIC reads this architectural margin. It is not detecting a change, because no change has happened yet. It is detecting the tuning quality that will determine whether change happens, and how badly, over the coming year.
The result, and the caveats
Combined with connectome-based predictive modelling, VIC predicts depression and anxiety symptoms one year later in held-out HCP BANDA adolescents at r = 0.60. The result is promising but not yet settled.
The cortical contributions to the prediction concentrate at the level of the integrative council that step two identifies as the architectural centre of conscious orchestration. The specific anatomical implementation of this council is still being worked out, but the contributing areas in the present analysis are known from independent work to be involved in self-related processing and the integration of affective content.
The sample is 150 adolescents, internally cross-validated. The 150 were the subset of the BANDA 1.1 release of 215 participants who had both baseline and one-year follow-up RCADS scores and low-motion resting-state data at baseline. External replication on an independent longitudinal dataset is the immediate next step, and the candidate datasets have been identified. The framing in the paper is therefore that of a discovery sample.
The mathematical reason the result might generalise is the architectural reason. VIC is computed from the same operator the unification establishes, the same operator whose three readings can be coordinated by one perturbation in the LSD test. If that mathematical structure is real, the clinical signal should be real with it. If it is not, the clinical signal will not replicate, and that is the test that will run over the next twelve to twenty-four months.
[Read the canary in the mind paper.](papers/#canary)
What the trilogy adds up to
A mathematical unification of brain dynamics, an architectural theory of consciousness that the unification scaffolds, and a first promising application to mental health, all derived from a single thermodynamic constraint.
The three papers are designed to be read together. Each paper can fail on its own terms without bringing the others down. The unification rests on operator calculus and the LSD test, and is independent of how the canary replicates. The architectural theory rests on the unification together with a separable set of theoretical claims about consciousness, and remains a candidate framework even if the clinical signal does not generalise as hoped. The canary rests on a particular cohort and methodology, and its replication record will be tracked openly on the updates page as independent evidence comes in.
For the methodology behind the whole-brain models, see the open-access book Whole-brain modelling: Cartography of the dynamics of mind by Deco and Kringelbach (Oxford University Press, 2025).
For the most likely questions readers, journalists and reviewers will ask, including where Entangled Loop’s commitments place it in Chandaria’s five-level taxonomy of consciousness theories and what this means for AI consciousness, see the FAQ.