Berjaga-Buisan and Kringelbach Track Consciousness Through a Thermodynamic Violation
A result from statistical mechanics, not neuroscience, is the newest tool for telling a conscious brain from an unconscious one. Tomas Berjaga-Buisan, Morten Kringelbach, Gustavo Deco, and ten co-authors report in Cell Reports that violations of the fluctuation-dissipation theorem, a rule that connects a system’s spontaneous fluctuations to how it responds to a perturbation, track conscious state across humans and rodents, wakefulness, anesthesia, and disorders of consciousness. The paper, “Thermodynamics of consciousness, non-equilibrium brain dynamics track conscious states,” appears in Cell Reports, Volume 45, Issue 8, article 117782, published 25 August 2026 under a CC BY licence, DOI 10.1016/j.celrep.2026.117782.
What the Fluctuation-Dissipation Theorem Measures
The fluctuation-dissipation theorem is a foundation of equilibrium statistical mechanics. It states that for a system at thermal equilibrium, the size of its spontaneous, random fluctuations is fixed by the same physical quantity that governs how strongly it responds when something perturbs it. Push on the system and measure the response, or leave it alone and measure how much it jitters on its own, and the two measurements are related by a single, predictable ratio.
That ratio breaks down the moment a system is driven away from equilibrium. A living brain is exactly this kind of driven system. It burns metabolic energy continuously to maintain its activity, so it is never at rest in the thermodynamic sense even when the person carrying it is sitting still. Kringelbach and Deco’s prior work, including the 2017 finding that resting brain dynamics sit at the edge of a Hopf bifurcation and the 2024 “Thermodynamics of Mind” framework, had already argued that this distance from equilibrium is not incidental noise but a structured property worth measuring directly. The new paper turns that argument into a number, calculating how far the observed relationship between fluctuation and response departs from the equilibrium prediction, and asking whether that departure changes with conscious state.
How the Whole-Brain Model Finds a Violation
The method does not require an actual perturbation, which is the practical break from the technique it is measured against. The perturbational complexity index needs a transcranial magnetic stimulation pulse and a subsequent EEG recording of how the brain’s activity ripples and settles afterward. Berjaga-Buisan and colleagues instead fit a generative whole-brain model, a mean-field description of large-scale neural populations, to spontaneous, unperturbed signals, whether resting-state fMRI in humans or local field potentials in rodents. The fitted model then predicts what the brain’s response to a hypothetical perturbation would look like, and the fluctuation-dissipation theorem gives a way to check that prediction against the spontaneous fluctuations actually recorded. A violation of the theorem’s expected ratio is the signature the paper tracks.
The same framework was applied across four conditions: normal wakefulness, general anesthesia, patients with disorders of consciousness such as unresponsive wakefulness syndrome, and anesthetized rodents recorded with intracranial electrodes. Testing the measure in a second species, using a different recording modality, is what lets the authors argue the effect is a property of non-equilibrium brain dynamics generally rather than an artifact of human fMRI preprocessing.
Fitting a whole-brain model this way means specifying how populations of neurons in each brain region couple to their neighbors, then adjusting those coupling parameters until the model’s simulated activity reproduces the statistical structure of the recorded signal. Once that fit is good, the model itself becomes the object the fluctuation-dissipation calculation is applied to. The theorem’s prediction is computed from the model’s dynamics, and the violation is the gap between what the model predicts a perturbation response should be and what the theorem would require if the same system were at equilibrium. Nothing about this step touches the patient or animal a second time, which is the entire practical advantage over PCI.
A Result Built By Both Measurement Traditions
Two of the paper’s thirteen authors, Marcello Massimini and Simone Sarasso, are members of the Milan group that built PCI in the first place, developing the TMS-EEG protocol and the Lempel-Ziv complexity calculation the index depends on. Their presence on this paper means the same research effort that produced the invasive gold standard also produced and validated the non-invasive measure checked against it. That is a stronger form of cross-validation than an outside group replicating a result, since it removes any question of whether the two labs were applying PCI consistently when they compared it to the new measure.
The rest of the author list spans the collaboration’s usual geography for this line of work: Berjaga-Buisan, Deco, and Sanz Perl at Universitat Pompeu Fabra in Barcelona, Kringelbach at Oxford and Aarhus, Maurizio Corbetta at the University of Padova, and Maria V. Sanchez-Vives at IDIBAPS in Barcelona for the rodent electrophysiology. Corbetta and Sanchez-Vives both bring disorders-of-consciousness patient cohorts and cortical recording expertise the Barcelona modeling group does not have in-house, which is what let the paper test the measure on unresponsive wakefulness syndrome rather than stopping at anesthesia in healthy volunteers.
| Perturbational Complexity Index | Fluctuation-Dissipation Violation | |
|---|---|---|
| Input signal | TMS-evoked EEG response | Spontaneous, unperturbed activity |
| Perturbation required | Yes, external TMS pulse | No, inferred from a fitted model |
| Species validated | Humans | Humans and rodents |
| Physical grounding | Algorithmic complexity (Lempel-Ziv) | Statistical mechanics (FDT) |
What the Results Show
Across every condition tested, fluctuation-dissipation violations mirrored what PCI had already established. Violations were largest during ordinary wakefulness, and they decreased in general anesthesia and in unresponsive disorders of consciousness compared with conscious baselines. The paper reports this as a direct empirical link between the two measures, non-equilibrium dynamics extracted from spontaneous signals and perturbational complexity extracted from evoked ones, converging on the same separation between conscious and unconscious states even though they are computed from entirely different kinds of data.
The authors are careful about what this does and does not establish. It does not identify a new neural correlate of consciousness in the sense of a brain region or cell type. It shows that a specific physical quantity, borrowed from statistical mechanics and applied to a generative model of brain dynamics, moves in step with an already-validated behavioral and clinical marker, without needing the pulse that marker depends on. That is a methodological result before it is a theoretical one, and the paper positions it as opening a non-invasive route to bedside assessment rather than as settling which theory of consciousness is correct.
Where This Sits Among Physical Measures
Non-equilibrium thermodynamics is one of several efforts to ground a consciousness marker in physics rather than in information theory or global broadcast dynamics. Ephaptic field coupling and McFadden’s electromagnetic field theory both locate the relevant physical quantity in the brain’s electromagnetic field. The fluctuation-dissipation approach instead locates it in the brain’s departure from thermal equilibrium, a quantity defined for any dissipative physical system, biological or not, provided that system has a well-characterized fluctuation and response structure to compare.
Comparison to The Consciousness AI
The Consciousness AI project runs a spiking substrate on digital hardware, a mathematical object built from discrete leaky integrate-and-fire units rather than the continuous, noise-driven mean-field populations the fluctuation-dissipation framework was built to describe. Whether a spiking network run on conventional silicon exhibits an analogous violation, a measurable departure from a fluctuation-response equilibrium defined for that substrate, is not something the project has tested, and the paper itself offers no claim about non-biological systems. What the result does supply is a second, independent physical quantity, alongside PCI’s complexity measure, that any future substrate-independence argument would need to address on its own terms rather than by analogy to either one.
What This Changes and What It Leaves Open
The immediate change is practical. A marker that needs only resting-state recordings and a fitted model, not a TMS pulse and a dedicated EEG setup, is easier to deploy in settings like an intensive care unit where perturbing an already-fragile patient carries real risk. The open question is the same one that follows any new correlate of consciousness, whether the measure tracks consciousness itself or tracks a physiological state, such as level of arousal or level of metabolic activity, that happens to correlate with it in every condition tested so far. The scientific effort to define consciousness in measurable terms gains one more candidate marker from this paper, and gains it from a direction, statistical mechanics, that most of the field’s existing measures do not draw from.
Sources: Berjaga-Buisan, T., Monti, J. M., Cortada, M., Colombo, M. A., Geli, S. M., Gaglioti, G., Sarasso, S., Kringelbach, M. L., Corbetta, M., Sanchez-Vives, M. V., Massimini, M., Perl, Y. S., and Deco, G. (2026). “Thermodynamics of consciousness, non-equilibrium brain dynamics track conscious states.” Cell Reports, 45(8), article 117782. DOI 10.1016/j.celrep.2026.117782.
Researchers covered here
- Tomas Berjaga-BuisanComputational Neuroscience Group, Center for Brain and Cognition, Universitat Pompeu Fabra, BarcelonaNon-equilibrium thermodynamics of whole-brain models, fluctuation-dissipation theorem violations as a non-invasive consciousness marker
- Morten KringelbachDepartment of Psychiatry, University of Oxford; Centre for Eudaimonia and Human Flourishing, Linacre College, Oxford; Department of Clinical Medicine, Aarhus UniversityWhole-brain computational modeling of non-equilibrium brain dynamics, the thermodynamics of mind framework, hedonic and pleasure circuitry
-
Gustavo DecoICREA, Universitat Pompeu Fabra, BarcelonaWhole-brain computational models, metastability at the edge of the Hopf bifurcation, the dynamical cortical core
- Marcello MassiminiUniversity of MilanThe Perturbational Complexity Index, TMS-EEG measurement of consciousness, disorders of consciousness assessment, sleep slow waves