Scientific reference document
Drafting date: July 13, 2026
This document reflects the state of scientific knowledge at this date. Periodic updating is recommended.
Contemporary neuroscience stands at a crossroads. For a century, it has mapped cortical areas, decoded neural circuits and identified neurotransmitters. But one fundamental question resists: How does electrical activity of inert matter become lived experience? David Chalmers, in 1995, named this challenge the « Hard Problem » — distinguishing « easy » problems (measurable functions) from the « hard » problem (the subjective essence of consciousness).
This document surveys the state of objective knowledge, separates validated theories from open hypotheses, and examines the hypothesis of the brain as receiver rather than generator.
The human brain contains approximately 86 billion neurons, connected by thousands of synapses each (see R-13). Three systems dominate current research:
The prefrontal cortex and fronto-parietal network — Stanislas Dehaene and the Global Neuronal Workspace (GNW) team have localized consciousness to late and massive activation in the dorsolateral prefrontal cortex, anterior cingulate and posterior parietal lobes. When a stimulus reaches consciousness, it triggers sustained activity crossing these hubs — the famous « P3 wave » in EEG. Unconscious stimuli, meanwhile, remain local, sensory. This difference is observable, reproducible, validated by peer-reviewed publications.
Synaptic plasticity — Kandel, Bear and others have demonstrated that neural connections strengthen or weaken according to use. Memory is not stored in a single location; it emerges from synaptic weights modified by experience. This mechanism explains learning, recovery after lesion, adaptation.
Default Mode Networks — At rest, when the subject performs no directed task, a specific network activates: posterior cingulate cortex, precuneus, medial prefrontal cortex. This network shuts down as soon as external attention is required. Its role remains debated: introspection ? Self-reference ? Temporal integration ?
Despite these advances, no unique center of consciousness has been found. Progressive elimination of motor, sensory, visual cortex — even the thalamus — often leaves subjective experience intact. Patients with 50% less brain tissue function normally. Water flowing in a pipe can change course without losing its current.
Chalmers formulates the problem as follows: « Even after explaining the functional, dynamic and structural properties of the conscious mind, we can still legitimately ask: why is it conscious ? » (translation from the original English). The standard neuroscientific answer — « because these neurons are active » — does not solve the question. It sidesteps it.
Global Workspace Theory (Baars, Dehaene) — Bernard Baars (1988) proposes a theatrical metaphor: Consciousness is the projector illuminating certain contents on the stage, while the unconscious constitutes the audience and backstage. Dehaene implements it neurally: The GNW is a network broadcasting information throughout the brain — the « global workspace ». Experimental validations: Masking, attentional blink, binocular rivalry — all show that access to this global workspace predicts the ability to report, learn and voluntarily control.
Integrated Information Theory (Tononi) — Giulio Tononi (2004+) quantifies consciousness by phi (Φ), measuring causal integration of a system. A conscious system cannot be reduced to its parts — the sum must exceed the components. IIT predicts that the posterior cortex, rich in recurrent loops, contains more Φ than the cerebellum. Clinical applications: Assessment of vegetative states, anesthesia, sleep. Controversies: Measuring Φ in large networks remains mathematically complex, even impossible to calculate in practice. (see R-11)
Orchestrated Objective Reduction (Penrose, Hameroff) — Roger Penrose (physicist) and Stuart Hameroff (anesthesiologist) propose that consciousness emerges from quantum processes in neuronal microtubules. Quantum decoherence at room temperature remains Achilles' heel: BECs (see R-02) require vacuum and -273 °C to maintain coherence. The brain, at 37 °C, wet and complex, seems hostile to this state. The theory remains minority in dominant neuroscience, though not refuted.
The architecture of deep neural networks roughly mimics the brain — layers, weights, backpropagation. But differences are fundamental:
| Aspect | Biological brain | Current AI |
|---|---|---|
| Support | 86 billion neurons, 10¹⁵ synapses | GPU/TPU, silicon memory |
| Energy | ~20 Watts | ~100 kW for large models |
| Learning | Continuous plasticity, few-shot | Massive training, labeled data |
| Consciousness | Subjective experience unknown | No subjective report |
| Architecture | Distributed, redundant | Centralized, sequential |
AI simulates cognitive functions — recognition, generation, statistical reasoning. It reports no internal experience. Is this absence technical (not yet complex enough) or principled (silicon cannot have experience) ? The question remains open (see B-07).
According to Alice & Bob, the brain would not produce consciousness but receive it, like a radio receives waves. Several observations support this hypothesis (see R-08):
NDE: Signal without receiver — Bruce Greyson, for 50 years at University of Virginia, has collected thousands of cases of Near-Death Experiences. He designed a 16-question questionnaire allowing to measure and classify these experiences: a score equal to or greater than 7 confirms it is a genuine NDE. Patients describe verifiable perceptions: surgeons describing specific instruments, conversations heard out-of-body while brain activity is flat. The receiver (brain) is stopped, but the signal (consciousness) persists with precision. The dominant neuroscientific model cannot explain these cases without contradictions (see R-05).
The in-between: Effortless access — In sleep-wake transitions, hypnosis, deep meditation, information emerges without active cognitive process. Access requires no effort — it occurs. This passivity contradicts the « generator » model where the brain should actively produce information (see Seed Chap. 3 — The In-Between).
Isolation vs interaction — Isolation studies (COVID-related confinement, feral children) show rapid cognitive degradation. Conversely, schools of thought (Copenhagen, Bohr, Heisenberg) produce breakthroughs through collision of ideas. If the brain generated consciousness alone, social interaction would not be critical. Dependence on environment suggests a connected system.
| Question | Scientific consensus | State of hypothesis |
|---|---|---|
| Does consciousness emerge from the brain ? | Majority (materialism) | Dominant but contested |
| GWT vs IIT: which theory wins ? | No clear victory | Active debate, evidence on both sides |
| Orch OR: viable at 37 °C ? | Minority, high skepticism | Requires thermal protection mechanism |
| Brain = receiver ? | Marginal, non-dominant | Empirical evidence (NDE) unexplained |
| Can AI have consciousness ? | Unknown, philosophical debate | No objective measure available |
The path remains open. Neuroscience has accomplished the impossible: map the material, quantify correlates. But consciousness — this primary quality of existence — resists reduction. Perhaps a paradigm shift is needed: not to seek consciousness in the brain, but the brain in consciousness.