Catching Consciousness in the Act

Theories of consciousness are cheap to propose and brutally hard to test. Consciousness doesn't leave fingerprints the way a heartbeat or a muscle twitch does — you can't point a sensor directly at "experience." So researchers have had to get creative, building indirect methods that compare brain activity when something is consciously experienced against brain activity when the same information is processed without awareness.

This article is a companion to an earlier piece on theories of consciousness and the brain structures they propose. Here, the focus shifts from theory to evidence: what methods do researchers actually use, and what have specific experiments found? Where relevant, each experiment is tied back to the theories it supports, challenges, or complicates.

Part 1: The General Toolkit

Methods used to study consciousness

Before getting into individual experiments, it helps to know the main instruments in the toolbox. Most consciousness experiments are some combination of these.

1. Scalp EEG (electroencephalography)
Electrodes on the scalp record the brain's electrical activity with excellent timing precision (millisecond-scale) but coarse spatial resolution. EEG is the workhorse for measuring event-related potentials (ERPs) — characteristic voltage patterns that appear a fixed time after a stimulus. Two ERP components come up constantly in consciousness research: the visual awareness negativity (VAN), an early signal over posterior electrodes, and the P3b, a later, broader positive wave linked to global broadcasting of information.
2. Functional MRI (fMRI)
fMRI tracks blood-oxygen changes as an indirect proxy for neural activity, giving excellent spatial detail (down to a few millimeters) but poor timing resolution (activity is smeared over seconds). It's the main tool for locating where conscious content is represented, and it's central to clinical work on disorders of consciousness.
3. Intracranial recordings (iEEG / ECoG / single-unit)
In patients undergoing surgery for epilepsy, electrodes are sometimes implanted directly on or inside the brain to localize seizures — a rare opportunity to record human brain activity with both excellent timing and spatial precision, occasionally down to single neurons. Similar recordings are done in animals, especially macaque monkeys.
4. MEG (magnetoencephalography)
MEG measures the tiny magnetic fields produced by neural activity, offering a useful middle ground: better spatial resolution than scalp EEG, while keeping EEG's excellent timing precision.
5. TMS-EEG and the Perturbational Complexity Index (PCI)
Transcranial magnetic stimulation (TMS) sends a brief magnetic pulse into the cortex from outside the skull, and EEG records how the brain "echoes" that pulse. The key insight, formalized by Marcello Massimini and colleagues, is that a conscious brain responds in a complex, differentiated, but still integrated pattern, while an unconscious brain responds simply, or not at all. This is scored as the Perturbational Complexity Index (PCI) — sometimes nicknamed "zap and zip," since the scoring algorithm compresses the EEG response the way a zip file compresses data, and a less compressible (more complex) response means a higher PCI.
6. Direct electrical brain stimulation
In patients with implanted electrodes, clinicians can stimulate small brain regions directly and observe effects on behavior and reported experience — a rare, causal way to test whether a structure is necessary for some aspect of consciousness.
7. Pharmacological manipulation
Anesthetic drugs (propofol, sevoflurane, ketamine, xenon) reliably and reversibly switch consciousness off, making them a valuable experimental "dial." Psychedelics (psilocybin, LSD) alter — rather than remove — conscious experience, offering a different kind of manipulation.
8. Lesion and case studies
Naturally occurring brain damage — strokes, tumors, surgical resections, or rare developmental conditions — offers "experiments" nature performs on its own. Results have to be interpreted cautiously, but they can reveal what a structure is or isn't necessary for.
9. Bistable perception and masking paradigms
When a stimulus is ambiguous (as in binocular rivalry, where each eye is shown a different image) or presented briefly and then masked, perception can flip between "seen" and "not seen" even though the raw sensory input stays the same — letting researchers compare identical input under different conscious outcomes.
10. No-report paradigms
A major recent methodological shift. Traditionally, participants report what they perceived (a button press, a word), but the report itself involves extra brain activity that has nothing to do with the experience and can contaminate results. No-report paradigms infer experience from indirect measures — eye movements, pupil size — without requiring an explicit report.
11. Disorders-of-consciousness assessment
Patients with severe brain injury can end up in a coma, an unresponsive wakefulness syndrome, or a minimally conscious state. Because behavioral tests alone can misdiagnose these states, researchers combine bedside scales with neuroimaging or PCI to look for covert awareness that behavior alone would miss.
12. Sleep and dream research
Waking people up at different points in the sleep cycle and asking whether they were dreaming turns sleep into a natural, repeatable on/off switch for conscious experience, one that doesn't require any drug or injury.
13. Computational and complexity measures
Beyond PCI, researchers compute mathematical measures of signal diversity and integration directly from resting EEG or fMRI data (such as Lempel-Ziv complexity, or various approximations of IIT's Φ), aiming for scalable, drug- and injury-free markers of the "capacity for consciousness."

Part 2: Specific Experiments and What They Found

A. Timing and the "ignition" question

Sergent & Dehaene (2004), attentional blink

Using a rapid stream of letters, participants sometimes missed a target because attention was still occupied by an earlier one (the "attentional blink"). Having participants rate the visibility of the target on a continuous scale, the researchers found that conscious perception behaved in an all-or-none, nonlinear fashion: near the threshold, participants tended to report either seeing the letter clearly or not at all, rarely something in between.

Relation to theory: This nonlinear, threshold-like jump is exactly what Global Neuronal Workspace theory predicts as "ignition" — a sudden, self-amplifying switch into a globally broadcast state, rather than a gradual ramp-up. It's frequently cited as behavioral support for GNW.

Libet's readiness potential (1983, and much follow-up work)

Though originally designed to study the timing of voluntary decisions rather than consciousness itself, Benjamin Libet's finding that a slow buildup of brain activity (the "readiness potential") precedes a person's reported conscious decision to move became deeply entangled with consciousness research, because it raises the question of when, exactly, a conscious "moment" occurs relative to the brain activity that produces it.

Relation to theory: This doesn't directly support or reject a specific structural theory, but it fed into GNW's emphasis on late-stage, effortful "ignition" as a marker of the conscious moment, and remains a touchstone (and a subject of methodological critique) in debates about what a conscious report is actually measuring.

B. Masking and thresholds

Del Cul, Baillet & Dehaene (2007), visual masking

Participants viewed a target letter followed by a mask at varying delays, making the target harder or easier to consciously see. EEG showed that as visibility crossed the threshold from "unseen" to "seen," a late, widespread wave of activity appeared abruptly, rather than growing gradually — extending notably into frontal and parietal regions.

Relation to theory: Like the attentional blink study above, this supports GNW's nonlinear ignition prediction, and specifically implicates the frontoparietal network GNW proposes.

C. Bistable perception (rivalry) studies

Logothetis and colleagues, binocular rivalry in monkeys (1990s)

When a monkey's two eyes are shown different images, its perception flips back and forth between them even though the physical input never changes. Recording single neurons, Logothetis found that most neurons in early visual cortex kept firing in relation to the physical stimulus, but neurons further along the visual pathway, in higher areas like inferior temporal cortex, tracked what the animal was perceiving — switching their firing with the animal's reported percept.

Relation to theory: This was some of the earliest strong evidence that conscious content tracks higher-level, more integrated representations rather than raw sensory input — consistent with, though not decisive for, both Global Workspace and posterior "hot zone" accounts, and an important building block for later no-report designs.

Frässle et al. (2014), no-report binocular rivalry

Using eye-movement patterns (optokinetic nystagmus) instead of button presses to infer which image a person was perceiving during rivalry, this fMRI study found that prefrontal cortex activity linked to perceptual switches largely disappeared once the need to report was removed, while activity in visual and parietal areas remained.

Relation to theory: A direct challenge to the idea that prefrontal activity is a necessary marker of consciousness itself. It suggests earlier studies may have measured the act of reporting a percept, not the percept — a serious methodological worry for both GNW and Higher-Order theories. (Not all follow-up studies agree; some single-neuron recordings in prefrontal cortex during no-report rivalry have found content-related activity persists, so this remains actively debated.)

D. Direct brain stimulation

Koubeissi et al. (2014), claustrum stimulation

In a woman undergoing epilepsy monitoring, electrical stimulation near her claustrum reliably and reversibly stopped her responsiveness: she stared blankly, stopped reading mid-sentence, and had no memory of the episode once stimulation ended — while stimulation of other nearby regions had no such effect.

Relation to theory: Widely cited as support for the idea (floated earlier by Francis Crick and Christof Koch) that the claustrum may act as a coordinating hub that binds distributed cortical activity into one unified experience. An important causal data point, but from a single patient with abnormal, seizure-prone tissue — subsequent claustrum stimulation studies in other patients have not always reproduced the effect as dramatically.

Central thalamic deep brain stimulation (Schiff and colleagues; Tasserie et al., 2022, in anesthetized macaques)

Electrically stimulating the central thalamus in anesthetized monkeys reliably restored arousal, wake-like brain dynamics, and frontoparietal coherence — effects not seen when nearby thalamic sites were stimulated instead. Related human work has used the same approach as an experimental therapy in some minimally conscious patients.

Relation to theory: Strongly supports thalamocortical theories that treat the thalamus, rather than cortex alone, as a critical hub, and aligns with GNW's emphasis on restoring frontoparietal communication specifically.

E. Anesthesia and the Perturbational Complexity Index

Casali et al. (2013), introducing PCI

Using TMS-EEG across wakefulness, dreamless sleep, several anesthetics (including the dissociative drug ketamine), and patients with disorders of consciousness, this study showed PCI values reliably dropped during unconsciousness and rose again during consciousness — including during ketamine anesthesia and REM sleep, states where standard EEG can look "activated" and thus misleadingly wake-like.

Relation to theory: PCI was explicitly inspired by IIT's core idea that consciousness requires both integration and differentiation of information. Its success as a bedside marker is often cited as indirect support for IIT's general framework, even though PCI doesn't require computing IIT's full mathematical measure (Φ), which remains practically incalculable for a real brain.

Casarotto et al. (2016) and later clinical validation studies

Applying PCI to patients with chronic disorders of consciousness found it correctly classified unresponsive wakefulness syndrome versus minimally conscious state with high sensitivity, and identified some behaviorally unresponsive patients whose PCI values fell in the "conscious" range — suggesting covert awareness undetected by bedside exams.

Relation to theory: Beyond its clinical value, this reinforces the idea that a whole-brain measure of complexity can track consciousness more reliably than surface behavior — a point IIT emphasizes, but which is, in practice, theory-agnostic and useful regardless of which theory turns out to be right.

F. Neuroimaging in disorders of consciousness

Owen et al. (2006), "Detecting Awareness in the Vegetative State"

A patient who met every clinical criterion for vegetative state (no behavioral signs of awareness) was asked, while in an fMRI scanner, to imagine playing tennis or to imagine walking through the rooms of her home. Her brain activated the same specific regions seen in healthy volunteers performing the same mental tasks — despite showing no outward behavioral response at all.

Relation to theory: This landmark study didn't set out to test a specific theory of consciousness, but it fundamentally reshaped the field by proving that behavior can radically underestimate awareness. It also gave researchers a tool — deliberate, willed mental imagery — later used to communicate with some non-responsive patients (Monti et al., 2010), and strongly motivated the shift toward brain-based rather than purely behavioral diagnosis captured in methods like PCI.

G. Lesion evidence

Weiskrantz and colleagues, blindsight

Patients with damage restricted to primary visual cortex (V1) sometimes retain a startling ability to correctly guess the location, motion, or orientation of visual objects in their "blind" field — far better than chance — while firmly denying any conscious visual experience of them.

Relation to theory: One of the most-cited pieces of evidence that a functioning V1, and the local recurrent processing that depends on it, may be necessary for conscious visual experience specifically — a key prediction of Recurrent Processing Theory. Also widely discussed within the Global Workspace framework, since without functioning V1 there is no input to "ignite" for broadcast. The two accounts don't fully agree on why blindsight happens, which makes it a genuinely contested case rather than a clean win for either theory.

Merker (2007) and Aleman & Merker (2014), hydranencephalic children

Children born with hydranencephaly — where cerebral hemispheres are largely replaced by fluid, leaving mainly the brainstem and some diencephalon intact — nonetheless show behavior that looks affectively rich and situationally responsive: smiling and laughing during play with familiar caregivers, crying to aversive stimuli, and showing preferences, despite having little or no functioning cerebral cortex.

Relation to theory: One of the most direct challenges to every cortex-centered theory on this list — GNW, IIT, Higher-Order theories, and Recurrent Processing Theory all lean heavily on cortical mechanisms. Merker argues instead for a brainstem-based origin of basic (affective) consciousness. Critics counter that what these children display may be simpler, more reflexive responsiveness rather than full phenomenal experience — the debate remains unresolved, and much hinges on how "consciousness" is defined in the first place.

H. Sleep and dreaming

Siclari et al. (2017), "The neural correlates of dreaming"

Using high-density EEG (256 channels), researchers repeatedly woke sleepers during both REM and non-REM (NREM) sleep and asked whether they had been dreaming. Regardless of sleep stage, dream reports were consistently linked to a local drop in low-frequency activity within a posterior cortical "hot zone" spanning occipital, parietal, and posterior cingulate regions — and monitoring this zone in real time could predict, above chance, whether a person would report a dream upon waking.

Relation to theory: Lines up strikingly well with IIT's proposed posterior "hot zone" as a core substrate of conscious content, and is one of the more direct pieces of support IIT has received from outside the anesthesia literature — notably, it comes from Giulio Tononi's own research group. It also complicates GNW somewhat, since vivid dream experiences can apparently occur without a "waking-style" frontoparietal ignition.

I. The big head-to-head test

Cogitate Consortium (2023 preprint; 2025, Nature), adversarial collaboration

In the most ambitious direct test to date, researchers aligned with GNW and researchers aligned with IIT jointly designed a shared experiment (viewing streams of faces, objects, letters, and false fonts) and agreed in advance on what pattern of results would count as support for each theory, then tested it using fMRI, MEG, and intracranial recordings in over 250 participants across multiple labs.

Relation to theory: The results were a genuine split decision. Decoding of consciously perceived content was strongest in posterior regions, aligning with IIT's spatial prediction — but the sustained, synchronized posterior activity IIT specifically predicted did not clearly appear. GNW's prediction of a late frontoparietal "ignition" received partial support, but the theory's expectation of strong, sustained content representation in prefrontal cortex was not clearly borne out either. Both theories had core predictions confirmed and other core predictions challenged — a rare case of a large, pre-registered experiment producing an honestly mixed verdict rather than a clean win.

Summary Table: Experiment, Method, and Theoretical Relevance

Experiment	Method	Structure implicated	Theory supported / challenged
Sergent & Dehaene (2004), attentional blink	EEG, behavioral rating	Frontoparietal "ignition"	Supports GNW
Del Cul et al. (2007), visual masking	EEG	Frontoparietal cortex	Supports GNW
Logothetis, binocular rivalry (monkeys)	Single-unit recording	Higher visual cortex (IT)	Broadly consistent with GNW / posterior views
Frässle et al. (2014), no-report rivalry	fMRI, eye tracking	Prefrontal vs. posterior/parietal	Challenges GNW & Higher-Order theories
Koubeissi et al. (2014), claustrum stimulation	Direct electrical stimulation	Claustrum	Supports claustrum-as-hub / binding proposals
Tasserie et al. (2022), central thalamus DBS	DBS + fMRI (macaques)	Central thalamus	Supports thalamocortical theories, GNW
Casali et al. (2013), PCI	TMS-EEG	Whole-cortex integration/differentiation	Inspired by, and supportive of, IIT's framework
Casarotto et al. (2016), PCI in DOC patients	TMS-EEG	Whole-cortex integration/differentiation	Clinical support for complexity-based (IIT-adjacent) markers
Owen et al. (2006), vegetative-state fMRI	fMRI, mental imagery	Motor, parahippocampal, parietal cortex	Reshaped diagnostic practice; motivates brain-based measures generally
Weiskrantz, blindsight	Lesion / case studies	Primary visual cortex (V1)	Supports Recurrent Processing Theory; debated within GNW
Merker (2007), hydranencephaly	Case studies	Brainstem / upper diencephalon	Challenges cortex-centered theories broadly
Siclari et al. (2017), dreaming	High-density EEG	Posterior cortical "hot zone"	Supports IIT's posterior hot-zone proposal
Cogitate Consortium (2025)	fMRI, MEG, iEEG	Posterior cortex & frontoparietal cortex	Partial support and partial challenge for both IIT and GNW

The honest takeaway

No single experiment on this list has settled the matter, and taken together they don't even point in one consistent direction — some favor posterior cortex, some favor frontoparietal networks, some point to the thalamus, and a couple point away from the cortex altogether. What's genuinely improved isn't consensus; it's rigor. Methods like no-report paradigms and PCI were built specifically to close loopholes that let earlier results be explained away, and collaborations like Cogitate show that competing camps can now agree, in writing, on what would prove them wrong. That's slow, unglamorous progress — but in a field this contested, it may be the only kind that counts.

Key sources cited above: Sergent & Dehaene (2004, Psychological Science); Del Cul, Baillet & Dehaene (2007, PLoS Biology); Logothetis & Schall (1989, Science); Frässle et al. (2014, Journal of Neuroscience); Koubeissi, Bartolomei, Beltagy & Picard (2014, Epilepsy & Behavior); Tasserie et al. (2022, Science Advances); Casali et al. (2013, Science Translational Medicine); Casarotto et al. (2016, Annals of Neurology); Owen et al. (2006, Science); Monti et al. (2010, NEJM); Weiskrantz, Barbur & Sahraie (1995, PNAS); Merker (2007, Behavioral and Brain Sciences); Aleman & Merker (2014, Acta Paediatrica); Siclari et al. (2017, Nature Neuroscience); Cogitate Consortium (2025, Nature).