What Science Knows — and Does Not Know — About Consciousness
What can we actually measure about consciousness? This article separates robust neuroscientific findings, competing theories, clinical limits and the unresolved problem of subjective experience.
Consciousness is an unusual object of science: every awake person knows it from the first-person perspective, yet a researcher must study it from the outside—through reports, behaviour and measurements of brain activity. Progress is therefore real, but different from a problem whose object can simply be placed on a scale. Science can now investigate the conditions, correlates and some mechanisms of conscious states with considerable sophistication, but it does not have a generally accepted explanation of why and how particular physical processes are associated with subjective experience.
This article is therefore a source audit, not a manifesto for one theory. It separates three levels: what the data show, how competing theories interpret those data, and which metaphysical conclusions would require additional argument. The dependence of conscious states on brain function is empirically very strong; that fact alone does not settle what consciousness ultimately is in an ontological sense.
The safest summary is: we know enough about consciousness to test and reject some claims, but not enough to declare any current theory the final solution.
First we have to say what is being measured
The word consciousness combines several related questions. One is whether an organism is in a state in which it has experiences at all; another is what the content of experience is; a third is whether that content is available for report, reasoning, decision-making and behavioural control. Wakefulness, responsiveness, attention and consciousness often overlap, but they are not synonyms.
Contemporary theories therefore do not always explain exactly the same target phenomenon. Some focus mainly on conscious access and the global availability of information, others on phenomenal content—on there being something it is like to see or feel—and still others on physical-system properties proposed to be associated with the existence of experience.
This is more than a terminological detail. If two theories explain different aspects of consciousness, their apparent conflict may be incomplete; if they make competing predictions about the same measurable phenomenon, they can be compared directly.
Sleep, general anaesthesia, epileptic seizures, brain injury, electrical or magnetic stimulation and pharmacological interventions can systematically alter the level or content of conscious experience. This is one of the strongest foundations of consciousness neuroscience: when particular brain processes change, conscious state changes in predictable ways as well.
But a correlation or even a manipulation effect is not yet a complete theory. Showing that a neural system is necessary for some form of experience does not automatically tell us whether its activity is consciousness itself, causes it, enables it or is one component of a broader mechanism.
Mind, Brain and Consciousness: Three Things We Often Confuse therefore separated brain, mind and consciousness. This article takes the next step: empirical findings strongly constrain possible explanations, but they do not jump directly from neurobiological dependence to a final metaphysical conclusion.
Researchers study the neural correlates of consciousness (NCCs): minimal neural mechanisms associated with a specific conscious experience or with consciousness more generally. It is important to distinguish background conditions that enable consciousness from mechanisms that specify a particular content—such as a face, pain or a sound.
This helps explain why it is misleading to search for a single anatomical 'consciousness spot'. Brainstem and subcortical systems are important for wakefulness and arousal, while cortical patterns participate in specific experiential contents. Consciousness is a question about the organisation and dynamics of a system, not merely one anatomical switch.
The term NCC is deliberately modest: a correlate is a candidate necessary or sufficient neural basis, not yet a complete explanation of why that basis carries experience.
Sleep and anaesthesia are natural laboratories for changing consciousness
General anaesthesia is especially useful because the same person can be followed from responsive wakefulness into a state without ordinary responsiveness and back again. Anaesthetics do not simply 'turn the brain off'; they change communication, integration and large-scale network dynamics, with effects that vary by drug and depth of anaesthesia.
Sleep likewise shows that wakefulness and experience are not the same thing. Dreaming can occur in REM and also in NREM sleep, so external unresponsiveness does not necessarily imply a total absence of experience. Awakening studies have linked reported dream experience to particular activity patterns, especially in posterior cortical regions.
These cases break the simple equation 'eyes open = conscious, eyes closed = unconscious'. Conscious state has to be inferred from multiple kinds of evidence.
Clinical studies of severe brain injury have shown that a person can fail to produce an observable response to a command even while brain activity indicates that the command was processed and a task was attempted. This phenomenon is called cognitive motor dissociation.
A large multicentre study published in the New England Journal of Medicine in 2024 detected task responses with fMRI, EEG or both in 60 of 241 participants who showed no observable behavioural response to commands—about 25 percent. This does not mean that one quarter of all unresponsive patients in every setting are necessarily conscious; the cohort was clinically selected and the methods have limitations. It does show that behavioural unresponsiveness is not a complete test for the absence of conscious processing.
That has direct medical and ethical consequences: where possible, borderline cases benefit from combining standardised behavioural assessment with additional neurophysiological methods.
Promising state indices exist, but there is no universal 'consciousness meter'
One influential approach combines transcranial magnetic stimulation with EEG. A brief external pulse perturbs cortical activity, and researchers measure how complex and widespread the resulting response becomes. This led to the Perturbational Complexity Index (PCI), which has distinguished several conscious and unconscious conditions in research settings without requiring a direct behavioural report.
Such indices are an important advance because they do not infer consciousness solely from movement or speech. Yet no current number is a universal device that directly measures subjective experience across all humans, animals or artificial systems.
A measure is always tied to a model, a population and validation data. Good classification of known states is not the same as solving the definition of consciousness.
The simplest way to test conscious perception is to ask, 'Did you see it?' But answering requires attention, working memory, a decision and a motor or verbal response. If we then measure activity associated with the answer, part of that activity may belong to reporting consciousness, not to the experience itself.
This motivated no-report approaches, which infer perceived content from eye movements, pupil responses, neural signals or other indirect measures. They can remove some confounds, but they are not a magical solution: once report is removed, it becomes harder to independently verify what the participant actually experienced.
That methodological tension is fundamental. Consciousness is private experience, while science requires publicly measurable data; paradigms must therefore keep asking what exactly their operational measure captures.
Global Neuronal Workspace: consciousness as globally available information
Global Neuronal Workspace Theory (GNWT) proposes that many specialised processes remain local and unconscious, while some information becomes consciously accessible when it is sufficiently amplified and broadly broadcast through a distributed brain network. It then becomes available to memory, decision-making, report and planning.
The theory has generated measurable predictions about sudden 'ignition', long-range connectivity and the role of frontoparietal or prefrontal regions. Its strength is the way it links conscious access to the flexible availability of information across cognitive systems.
An open question is whether global availability explains access and report only, or also the phenomenal fact that anything is experienced in the first place. That question is part of the broader disagreement among theories.
Integrated Information Theory (IIT) begins from proposed properties of experience and tries to derive the properties a physical system would need to support such experience. It emphasises differentiation—many possible distinct states—and integration, in which the whole is not reducible to independent parts.
IIT is more ambitious than a simple search for correlates because it aims to say which physical systems are conscious and why. That also makes its claims harder to test empirically and has generated intense debate about measurement, formalism and the relationship between the theory itself and neuroscientific proxies for integrated information.
For this article, the important point is that IIT is a serious and influential research theory, not an established fact. Its formal constructs should not be equated with every generic measure of brain 'complexity'.
Other theories place consciousness elsewhere in the processing chain
Higher-order theories emphasise that a state becomes conscious when it is appropriately represented at a higher order—roughly, when the system not only represents the world but in some sense also represents its own mental state. Recurrent Processing Theory gives more weight to feedback and recurrent loops within sensory systems, rather than requiring global broadcasting for consciousness.
Predictive or generative processing is another influential family of ideas, but the general principle of prediction does not by itself constitute one universally accepted theory of consciousness. Different theorists connect it to consciousness in different ways.
Reviews of the field therefore describe a family of competing theories that differ in what they aim to explain, which mechanism they treat as essential and whether their predictions are specific enough for decisive experiments.
In 2025, the COGITATE Consortium published in Nature one of the most ambitious direct comparisons of consciousness theories to date. Proponents of IIT and GNWT, together with theory-neutral researchers, preregistered divergent predictions and then studied 256 participants using fMRI, MEG and intracranial EEG.
The results found information about conscious content in posterior and some frontal regions, with some observations aligning with predictions from both theories. At the same time, the lack of predicted sustained synchronisation within posterior cortex challenged a key IIT prediction, while limited prefrontal effects and the absence of some expected ignition events challenged important GNWT predictions.
The most important conclusion is not 'both theories are false'. It is that when theories are forced to make preregistered, divergent predictions, data can genuinely reduce the space of viable explanations. The study substantially challenged both theories without settling the consciousness problem.
So what has actually become more secure?
Despite disagreement, several points are relatively well established. Human conscious experience is tightly linked to organised brain activity; it is not identical to simple sensory stimulation or a motor response; it can change without major changes in the external environment; and it depends on dynamic interactions among multiple neural systems rather than on one isolated neuron.
It has also become clear that researchers must separate prerequisites, correlates and consequences of conscious experience. Attention can prepare information, working memory can maintain it and report can express it—but none should automatically be identified with consciousness itself. Attention: The Gateway of Conscious Experience made this point for attention and Memory Is Not a Recording: How We Reconstruct the Past for memory.
Progress can therefore look less spectacular than headline claims suggest. Scientifically, it is already important to know which simple equations fail.
Even if we knew every neural event associated with seeing red, feeling pain or having a sense of self, a further scientific and philosophical question would remain: why is this processing associated with experience at all? David Chalmers popularised this as the 'hard problem' of consciousness. Not all researchers accept his framing or his conclusion that a special kind of explanation is required, but the explanatory transition remains central to the debate.
Current theories offer different answers: global broadcasting, recurrent processing, higher-order representation, integrated causal structure, or combinations involving generative and embodied processes. None has yet accumulated the combination of unique predictions, repeated confirmation and broad expert convergence that would allow the field to say the problem is solved.
It is therefore fair to say that neuroscience is making rapid progress on the neural conditions and functional properties of conscious processing, while the relation between physical description and subjectivity itself remains open.
The knowledge boundary is clearest where there is no report
In a healthy adult, a brain measure can be compared with a report. The problem is harder in an infant, an animal, a severely brain-injured person or an artificial system. We then infer from homology, behaviour, anatomy, dynamics, perturbation responses and theory-derived criteria.
This creates a risk of circular reasoning: if a theory is used to decide who is conscious and those same cases are then used to prove the theory, there is no independent test. That is why multimethod approaches and experiments comparing different theories on the same data are especially valuable.
There is currently no generally accepted diagnostic test of consciousness for artificial intelligence either. Linguistically persuasive output or complex behaviour is not by itself a direct measurement of subjective experience; equally, absence of human-like behaviour is not sufficient for an automatic conclusion that experience is absent.
We do not have to choose between two extremes: 'science knows nothing about consciousness' and 'science has proved that consciousness is just X'. Both are too crude. We have robust experimental paradigms, clinically useful measurements and theories that make genuine predictions; we also have methodological confounds, partly mismatched definitions and an unresolved explanatory problem.
This is a good example of the THY-REALITY method. Finding: conscious states change systematically with brain states and can be partially tracked with behavioural and neurophysiological measures. Theory: competing models attempt to explain which mechanisms are essential. Metaphysics: claims that consciousness is fundamental to the universe, merely an epiphenomenon of matter, a quantum phenomenon or something else require additional arguments and are not direct outputs of a single EEG or fMRI study.
This article therefore ends where a good source audit should end: with a clear boundary between what we can currently measure, what we can meaningfully test in theory, and what we still cannot explain. Ignorance at the final level does not erase knowledge at the first two—and knowledge at the first two does not license us to declare the final level solved.
Sources and further reading
- Seth, A. K. & Bayne, T. (2022). Theories of consciousness. Nature Reviews Neuroscience 23, 439–452 — comparative review of higher-order, global workspace, re-entry/predictive processing and integrated information approaches, emphasizing incomplete convergence and the need for discriminating tests.
- Koch, C., Massimini, M., Boly, M. & Tononi, G. (2016). Neural correlates of consciousness: progress and problems. Nature Reviews Neuroscience 17, 307–321 — framework distinguishing full/content-specific NCCs, background conditions and methodological confounds.
- Mashour, G. A. (2024). Anesthesia and the neurobiology of consciousness. Neuron 112(10), 1553–1567 — current review of anaesthesia, arousal, sensory processing and large-scale network mechanisms relevant to consciousness.
- Siclari, F. et al. (2017). The neural correlates of dreaming. Nature Neuroscience 20, 872–878 — within-sleep comparison linking reported experience to posterior cortical activity and showing that dreaming is not restricted to REM sleep.
- Bodien, Y. G. et al. (2024). Cognitive Motor Dissociation in Disorders of Consciousness. New England Journal of Medicine 391, 598–608 — multicentre study detecting task-based fMRI/EEG command-following responses in 60/241 behaviorally non-responsive participants.
- Casali, A. G. et al. (2013). A theoretically based index of consciousness independent of sensory processing and behavior. Science Translational Medicine 5(198), 198ra105 — introduction and validation of the Perturbational Complexity Index across wakefulness, sleep, anaesthesia and disorders of consciousness.
- Tsuchiya, N., Wilke, M., Frässle, S. & Lamme, V. A. F. (2015). No-report paradigms: extracting the true neural correlates of consciousness. Trends in Cognitive Sciences 19, 757–770 — rationale for separating conscious experience from report-related neural activity.
- Duman, I. et al. (2022). The No-Report Paradigm: A Revolution in Consciousness Research? Frontiers in Human Neuroscience 16, 861694 — methodological critique showing that no-report paradigms solve some confounds while introducing inferential challenges of their own.
- Mashour, G. A., Roelfsema, P., Changeux, J.-P. & Dehaene, S. (2020). Conscious Processing and the Global Neuronal Workspace Hypothesis. Neuron 105, 776–798 — contemporary GNWT account of amplification, global availability and large-scale neural dynamics.
- Tononi, G., Boly, M., Massimini, M. & Koch, C. (2016). Integrated information theory: from consciousness to its physical substrate. Nature Reviews Neuroscience 17, 450–461 — influential IIT formulation linking properties of experience to proposed intrinsic causal structure.
- Lau, H. & Rosenthal, D. (2011). Empirical support for higher-order theories of conscious awareness. Trends in Cognitive Sciences 15, 365–373 — review of higher-order approaches and their empirically testable commitments.
- Lamme, V. A. F. (2006). Towards a true neural stance on consciousness. Trends in Cognitive Sciences 10, 494–501 — influential recurrent-processing-oriented account and argument for closer alignment of neural and behavioural definitions.
- Cogitate Consortium, Ferrante, O. et al. (2025). Adversarial testing of global neuronal workspace and integrated information theories of consciousness. Nature 642, 133–142 — preregistered multimodal direct comparison finding partial support but substantial challenges to key predictions of both IIT and GNWT.
- Arachchige, A. S. P. M., Svet, A. & Svet, M. (2026). Consciousness in the Brain: An Integrative Review of Contemporary Theories. Brain Sciences 16(7), 745 — recent overview noting continued theoretical plurality and lack of a comprehensive agreed neural account.
- Chalmers, D. J. (1995). Facing Up to the Problem of Consciousness. Journal of Consciousness Studies 2(3), 200–219 — classic formulation of the 'hard problem'; included as a philosophical framing, not as an empirical result or universally accepted diagnosis of the field.
- Dehaene, S., Lau, H. & Kouider, S. (2017). What is consciousness, and could machines have it? Science 358(6362), 486–492 — separates global access and self-monitoring functions and illustrates why machine-consciousness claims depend on explicit consciousness criteria rather than surface behaviour alone.