Home > Positioning > Persons > Baars

Bernard Baars (b. 1946)

Baars proposed that consciousness is not a place or a substance but a mode of distribution. Global Workspace Theory holds that the mind is a crowd of specialised processors running in parallel, none of them conscious; they compete for access to a limited-capacity global workspace, and whatever wins is broadcast brain-wide — made globally available to attention, memory, verbal report, and action. That global availability, on the theory, is what being conscious of something amounts to. The image he made famous is the theatre: a spotlight on a stage, a vast unconscious audience, and stagehands working in the dark. The architecture came from artificial intelligence — the blackboard systems of the 1970s, in which independent specialist programs shared one common data structure — carried into psychology and, in other hands, into the scanner.

Life

Born 1946 in Amsterdam; the family emigrated to California in his childhood. PhD in cognitive psychology at UCLA, trained as a psycholinguist — his early research was on speech errors, including a well-regarded experimental treatment of Freudian slips. Taught at the Wright Institute in Berkeley, where the workspace theory took shape; A Cognitive Theory of Consciousness (1988) is its founding statement, published at a time when consciousness was barely respectable as a scientific subject. Later a senior fellow in theoretical neurobiology at the Neurosciences Institute in San Diego — Gerald Edelman’s institute — placing the workspace idea alongside the selectionist programme. Co-founded the Association for the Scientific Study of Consciousness and the journal Consciousness and Cognition; as much as anyone, he helped build the field in which his theory now competes.

Global Workspace Theory

The method: contrastive analysis. Baars’s methodological contribution came first and made the theory tractable: treat consciousness as a variable, not a mystery. Compare closely matched pairs — the same stimulus seen and unseen, the same process novel and automatised, the same content attended and unattended — and whatever differs between the members of each pair is what consciousness contributes. The method turned a philosophical question into an experimental design, and the masking and attentional-blink paradigms that later carried the theory’s evidence are contrastive analysis industrialised.

The architecture. Specialised processors — for edges, phonemes, faces, syntax, threat — run in parallel, fast, and unconsciously. The workspace is a limited-capacity facility: one content at a time. Coalitions of processors compete for it; the winning content is broadcast to the whole system, recruiting whatever specialists can use it. Consciousness is the broadcast, not any processor’s private activity. The theory is functional through and through — it says what conscious access does (integration, flexible routing, novel coordination) and identifies consciousness with that function. The theatre image drew fire: Daniel Dennett had made “Cartesian theatre” the name of the error — an inner stage where experience is presented to an inner spectator — and pressed the question whether Baars’s theatre repeats it. Baars’s answer is that the workspace theatre has no spectator: the audience is the unconscious crowd of specialists, and the broadcast is the experience, not a showing to anyone. Dennett has treated broadcast-style views as close to his own fame-in-the-brain picture; how much distance remains between them is a live question in the literature.

The neural turn. Stanislas Dehaene, Jean-Pierre Changeux and colleagues built the Global Neuronal Workspace: the workspace realised in long-range excitatory neurons concentrated in prefrontal and parietal cortex, and conscious access given a measurable signature — ignition, a late (~300ms), sudden, all-or-none spread of activation, watchable in masking experiments that hold a stimulus at the edge of visibility. An unseen stimulus produces early, local activation only; a seen one adds the late global wave. Baars credits the neural version, Dehaene credits the original; the relationship is one of implementation, not rivalry.

Beyond the human case. A strand of Baars’s corpus argues that consciousness is not a uniquely human property. With David Edelman and Anil Seth he developed the hallmarks approach: identify the observable criteria that accompany consciousness in humans — the waking thalamocortical dynamics, the capacity limits, the flexible novel routing — and look for the same hallmarks in other species, including non-mammalian ones (“Criteria for consciousness in humans and other mammals,” 2005; “Identifying hallmarks of consciousness in non-mammalian species,” 2005). The line is adjacent to his Neurosciences Institute period, and it matters to the theory: animals that cannot report are where the workspace architecture must stand on its functional signatures alone.

In artificial intelligence. The workspace idea has travelled back to the field it came from. Yoshua Bengio’s “consciousness prior” is explicitly anchored in it — a low-dimensional bottleneck of globally available elements — and shared-workspace architectures have been built for coordinating neural modules. In current machine-consciousness assessments, Global Workspace Theory serves as one of the standard indicator frameworks.

The live disputes. The ignition signature is under attack at its most load-bearing point. Tsuchiya, Wilke, Frässle and Lamme (2015) argued that the late frontal component may index the act of reporting rather than awareness itself, and no-report paradigms have found that consciously seen but task-irrelevant stimuli can fail to produce the signature at all; workspace theorists reply that task-irrelevant contents may simply not be accessed. A large adversarial collaboration published in Nature in 2025, testing the neuronal workspace against integrated information theory, found prefrontal content present but weaker than the theory predicted, and the predicted second ignition at stimulus offset largely absent; both camps have published their own readings of how much damage was done.

Where Baars stops

Global Workspace Theory is, by its own construction, a theory of access: of which contents get distributed, and what their distribution does. Whether that is the whole of consciousness is the question the theory does not answer, and Baars does not claim it does. Ned Block made the gap canonical: access consciousness (a content’s availability for use in reasoning, report, and action) and phenomenal consciousness (what the experience is like) may come apart, and Sperling-style experiments — where subjects seem to see more than they can report — have been read as phenomenal experience overflowing workspace access. On that reading, the theory describes the doorway and not the room. Dehaene’s response is that reportable access is the only scientifically tractable notion; Block’s is that this defines the hard part away.

The boundary is real either way: what the workspace framework can specify is the traffic — competition, threshold, broadcast, availability. What it is like to be the system the traffic runs through is a question the programme hands on, to philosophy or to some future science, and the theory’s standing does not settle it.

Key works

See also: Edelman · Friston · Dennett · Process models of the brain