Reading passage
The human brain contains roughly 86 billion neurons, yet consciousness feels strikingly unified: at any moment we experience one coherent scene, not billions of separate processes. Explaining how the brain binds distributed processing into a single stream of awareness is among the oldest puzzles in neuroscience. One family of theories, known as global workspace models, proposes that consciousness arises when selected information gains access to a brain-wide communication system, while the vast majority of neural activity remains unconscious, automatic and local.
The idea entered modern psychology through the work of Bernard Baars in the 1980s. Baars compared the mind to a theatre: consciousness is the brightly lit stage where a single actor performs, while a huge cast of unconscious specialists works backstage and in the darkened auditorium, shaping what appears in the spotlight. The stage is small, reflecting the narrow capacity of conscious awareness, but whatever appears on it can be seen by the whole theatre. In cognitive terms, conscious content becomes available to memory, language, planning and decision systems at once.
Stanislas Dehaene and his colleagues transformed this metaphor into a biological mechanism, the global neuronal workspace. In their account, specialised long-range neurons, concentrated in prefrontal and parietal cortex, form a network able to broadcast a single piece of information across the whole brain. When a stimulus is weak or unattended, its activity fades locally within sensory areas. When it crosses a threshold, a sudden non-linear amplification, which Dehaene calls ignition, recruits the workspace and makes the stimulus consciously available. Consciousness, on this view, is brain-wide information sharing.
The model makes sharp, testable predictions, and many have been borne out. In masking experiments, a word flashed briefly and followed by visual noise can fail to reach awareness, and brain imaging shows its processing then remains confined to posterior sensory cortex. When the same word is consciously seen, a wave of activation spreads to frontal and parietal regions about 300 milliseconds after the stimulus, producing a characteristic electrical signature known as the P3 wave. Studies of the attentional blink and of inattentional blindness reveal the same pattern: identical sensory input, different workspace access.
More recent work has refined the picture. So-called no-report paradigms ask participants not to name what they see, allowing researchers to separate the neural correlates of consciousness itself from the machinery of verbal report. Such studies suggest that some fronto-parietal activity may reflect task demands rather than experience. Lesion studies and recordings in split-brain patients probe how much broadcasting is enough, and large-scale computational models now simulate workspace dynamics in artificial networks, generating predictions that can be compared with human brain data.
The workspace model also faces serious rivals. Local recurrence theories, associated with Victor Lamme, argue that recurrent processing within sensory areas may suffice for visual experience, without any fronto-parietal broadcast, while integrated information theory ties consciousness to the causal structure of a system rather than to its communication patterns. To arbitrate, an adversarial collaboration known as COGITATE, funded by the Templeton Foundation, pitted the global neuronal workspace against integrated information theory in pre-registered experiments. The results, announced in 2023, were mixed: each theory scored some hits and some misses, and neither emerged fully vindicated.
Whatever its final standing, the workspace framework has already proved clinically fertile. It offers a natural account of anesthesia, in which long-range communication collapses while local processing persists, and it has inspired tools for assessing disorders of consciousness. In patients diagnosed as being in a vegetative state, brain imaging guided by workspace ideas has detected signs of covert awareness in a significant minority, with direct consequences for prognosis and care. Engineers, meanwhile, have borrowed the architecture, building artificial systems that route selected information through a central bottleneck.
Critics note that the model explains access consciousness, the availability of information for reasoning and report, more convincingly than it explains phenomenal feel, the raw texture of experience. That objection may ultimately define its limits. For now, the global workspace remains what a good scientific model should be: a fruitful, falsifiable map of how a conscious brain might work, and a standard against which every rival theory must measure itself.