Baars’s Global Workspace (Dynamics) Theory
Baars’s Global Workspace Theory treats consciousness as a cognitive architecture in which a limited-capacity workspace broadcasts its momentary content to a vast population of specialized unconscious processors. Competition, attention, and unconscious contexts organize what becomes conscious and how it is interpreted. Built on contrastive analysis of matched conscious and unconscious processes, it explains why experience is serial and narrow while the brain is parallel and vast: limited capacity is the price of global access. Baars holds that globally broadcast information is phenomenally conscious, rejecting a separate access consciousness. Later versions, Global Workspace Dynamics and an alliance with Edelman’s Dynamic Core, locate the workspace function in cortico-thalamic binding and broadcasting without a privileged region. The theory is distinct from the Global Neuronal Workspace Theory it inspired.

Bernard Baars
Cognitive Psychologist
Bernard J. Baars (b. 1946, Amsterdam) is an American cognitive psychologist and consciousness researcher best known for originating Global Workspace Theory, presented systematically in his 1988 book A Cognitive Theory of Consciousness. His research uses contrasts between conscious and unconscious processing to investigate perception, learning, attention, memory, and voluntary control. He subsequently developed Global Workspace Dynamics and collaborated with Gerald Edelman and Joseph Gally on biological accounts of experience. Baars received his PhD in cognitive psychology from the University of California, Los Angeles; he is a co-founder of the Association for the Scientific Study of Consciousness, founding co-editor of Consciousness and Cognition, and author of In the Theater of Consciousness (1997).
Key Takeaways
Core Claim
Consciousness is the content of a limited-capacity workspace, broadcast to a vast population of specialized unconscious processors.
How It Works
Processors compete for access; the winning content is broadcast globally, recruiting interpretation, memory, and action, while unconscious contexts shape what appears.
Distinguishing Idea
Limited capacity is the price of global access, which explains why experience is serial and narrow while the brain is parallel and vast.
Brain Basis
In Global Workspace Dynamics, any region of the cortico-thalamic core can bind and broadcast; no region, prefrontal cortex included, is privileged.
Phenomenal Consciousness
Baars holds that globally broadcast information is phenomenally conscious and rejects a separate access consciousness; why broadcasting feels like anything remains disputed.
Relation to GNWT
Ancestor of, but distinct from, Dehaene and Changeux’s biologically specified Global Neuronal Workspace Theory.
Baars’s Global Workspace (Dynamics) Theory
Global Workspace Theory (GWT), proposed by the cognitive psychologist Bernard Baars in the 1980s, holds that conscious experience corresponds to the contents of a limited-capacity “global workspace” whose contents are broadcast to a vast population of specialized, unconscious processors. Framed as a cognitive architecture and presented systematically in 1988, it explains how a capacity-limited stream of experience can recruit extensive resources for interpretation, learning, memory, and action. Derived from comparison of matched conscious and unconscious processes, the theory explains why consciousness is serial, unified, and narrow while the brain is parallel, distributed, and of enormous capacity: the narrow workspace is the price of brain-wide access.
Baars treats being globally broadcast and being phenomenally experienced as one fact described in two ways, and later brain-based versions of the theory, Global Workspace Dynamics and his alliance with Edelman’s Dynamic Core, place the workspace function in the cortico-thalamic system without tying it to any single region. GWT therefore offers both a functional account of conscious access and, in these biological extensions, a proposed account of phenomenal experience, while leaving disputed whether global availability explains why processing feels like anything. GWT is the ancestor of, but is distinct from, Dehaene and Changeux’s Global Neuronal Workspace Theory, which has its own entry (Baars, 1988, 1997a, 2002; Baars et al., 2013).
Contrastive Analysis and the Limited-Capacity Paradox
Baars’s starting point is methodological. Consciousness can be studied as an experimental variable by “contrastive analysis”: sets of paired comparisons between closely matched conscious and unconscious processes, such as attended and unattended input, novel and habituated stimuli, explicit and implicit memory, or controlled and automatic skills. This is Baars’s distinctive step: to connect a cognitive architecture with experimentally established differences between conscious and unconscious cognition (Baars, 2002, 2005). He argues that such contrasts deal directly with personal phenomenology, since each can be demonstrated in one’s own experience, and that together they impose demanding empirical constraints on any theory (Baars, 1988, 1997b).
The contrasts yield a puzzle. Conscious contents are few, internally consistent, and serial, whereas unconscious processing is massively parallel, specialized, and efficient. Yet conscious events have disproportionately wide consequences: only conscious information can be reported, voluntarily acted upon, and used to recruit knowledge that is otherwise inaccessible. The question that drives the theory is how a serial, integrated, and very limited stream of consciousness can arise in a nervous system that is mostly unconscious, distributed, and parallel (Baars et al., 2021). Baars’s summary answer is that “consciousness creates global access” (Baars, 1997b).
Thus, whereas the nervous system performs many complex operations simultaneously without awareness, the “global workspace” provides a temporary means of coordinating these distributed capacities.
Specialist processes compete and cooperate to establish a dominant content. The selected “winner” is broadcast and becomes “globally available,” enabling other systems to interpret it, associate it with memories, assess its significance, or use it to organize behavior. A conscious encounter with an unfamiliar object can therefore recruit expertise whose location and operations the subject does not consciously know. Consciousness concerns the shared result, not awareness of all the computations producing or responding to it.
“Global” means availability across a sufficiently extensive range of functions; it does not require every neuron to receive identical information or every recipient to act. The functional center of exchange need not occupy a single anatomical center. Likewise, limited capacity concerns the dominant organized episode, which can contain multiple related features, rather than an invariant quota of isolated sensations.
The Workspace Architecture
The term “global workspace” comes from artificial intelligence, where blackboard architectures allow large collections of specialized programs to cooperate on problems none can solve alone by posting messages to a shared, fleeting memory. Baars adapted this design as a model of human cognition built from three constructs. Specialized processors are unconscious, relatively autonomous experts (for syntax, face recognition, motor control, and so on). The global workspace is a central exchange of very limited capacity; processors or coalitions of processors compete for access to it, and the winning content is broadcast to the entire system. Contexts are stable coalitions of unconscious processors, among them perceptual, conceptual, and goal contexts, that shape and constrain what can become conscious without being conscious themselves (Baars, 1988).
On this account, a content is conscious when it occupies the workspace and is broadcast, and Baars adds that conscious contents must be internally consistent and informative: fully predictable input fades from consciousness, as in habituation, because it no longer triggers adaptation in the receiving processors. Limited capacity is therefore not a design defect but a design feature. Because only one consistent content can be broadcast at a time, the system pays for global access with a narrow momentary focus (Baars, 1988; Baars et al., 2013).
The Theater Metaphor and the Functions of Consciousness
Baars’s best-known image is the theater. Conscious contents are events on a stage lit by the spotlight of attention; the audience is the vast array of unconscious processors that receive the broadcast; and behind the scenes, contextual systems such as goals and the self act as directors that shape what appears on stage. Baars distinguishes this from the “Cartesian theater” criticized by Dennett: there is no single point where everything comes together and no inner observer (i.e., no homunculus, no “soul”), only a distributed audience of unconscious specialists (Baars, 1997a, 1997b).
The architecture assigns consciousness definite work. Global broadcast recruits unconscious resources to deal with novel or ambiguous situations, enables learning, supports error detection and voluntary control, and gives access to the self as an interpreting context. Baars later condensed these into the “conscious access hypothesis”: consciousness is the gateway that allows otherwise separate brain functions to exchange information, a claim he linked to working memory, whose active components he takes to be conscious (Baars, 2002; Baars & Franklin, 2003).
Brain Basis: From ERTAS to Global Workspace Dynamics
While Baars’s 1988 theory was psychological, he proposed a neural candidate from the outset: an “extended reticular-thalamic activating system” (ERTAS), comprising the brainstem reticular formation, the thalamus and its reticular nucleus, and the diffuse thalamocortical projections, which could distribute information widely across the cortex. A more explicit neural model followed (Newman & Baars, 1993). By the early 2000s, Baars read the accumulating brain-imaging literature as confirming the central prediction that conscious, as opposed to matched unconscious, stimuli evoke widespread activity beyond sensory cortex (Baars, 2002, 2005).
The current brain-based formulation is Global Workspace Dynamics. Here the workspace is not an anatomical hub but a dynamic capacity of the cortico-thalamic core for “binding and broadcasting.” Conscious contents can arise in any region of that core when competing signal streams settle into a winning coalition, which then ignites an any-to-many broadcast lasting roughly 100–200 milliseconds, possibly carried by cross-frequency phase coupling. Sensory percepts are proposed to bind and broadcast from posterior cortex, and non-sensory “feelings of knowing” from frontotemporal regions (Baars et al., 2013). Baars has accordingly denied that GWT requires prefrontal cortex for conscious perception, stating that the theory neither asserts nor denies the dominance of any anatomical region and that static divisions between prefrontal and posterior cortex are superseded by the dynamic connectome (Baars et al., 2021).
A parallel development is the convergence with Edelman’s Neural Darwinism. Edelman, Gally, and Baars (2011) argue that the two hypotheses are complementary: reentrant signaling in the thalamocortical Dynamic Core supplies the mechanism for workspace events, and a global broadcast can be regarded as a momentary snapshot of Dynamic Core activity. The joint account is offered as strictly biological, shaped by evolution, development, and neuromodulatory value systems, and it concludes that there is sufficient evidence to consider building a conscious artifact: a brain-based device incorporating a global workspace, a dynamic core, and a value system.
Phenomenal Consciousness and the Hard Problem
GWT is often described as a theory of access, not of phenomenal consciousness. Baars rejects the division. Responding to Block’s proposal that phenomenal and access consciousness have distinct neural correlates as well as phenomenological aspects, Baars and Laureys (2005) argue that there is one correlate, not two: information that is globally accessed is phenomenally conscious, and the specific contents of experience depend on the cortical regions that supply them. In Global Workspace Dynamics, the point is made directly: conscious contents, or qualia, reflect their sources in cortex (Baars et al., 2013). What Baars does not offer is an account of why broadcast information should be experienced at all; the theory takes reportable experience as its datum and explains its conditions, limits, and functions.
The fullest treatment of phenomenal experience under Baars’s name is his paper with Edelman and Gally. There, qualia are identified with the higher-order discriminations entailed by Dynamic Core activity (the experience of blue is distinguishable from that of warmth, and so on); they are private because they arise within a single embodied phenotype; and they are not themselves causal, the underlying neural processes being what does the causal work. On this view, the hard problem “does not require a solution, but rather, a cure” (Edelman et al., 2011). The claim that consciousness is not causal sits in some tension with the functional role that GWT elsewhere assigns to conscious broadcast, a tension the authors address by treating talk of conscious causation as a reliable shorthand for the causal efficacy of the entailing neural events.
Computational Implementations and the Relation to GNWT
Because GWT specifies an architecture, it lends itself to engineering implementation. Stan Franklin’s IDA and LIDA agents realize the workspace cycle in software, and Baars and Franklin used them to model how conscious events mobilize unconscious routines in working memory (Baars & Franklin, 2003). Shanahan and Baars (2005) proposed that a workspace architecture addresses the frame problem. The Blums’ Conscious Turing Machine formalizes the theory in theoretical computer science, and VanRullen and Kanai (2021) outline a deep-learning “global latent workspace.” These developments are why GWT figures centrally in discussions of machine consciousness.
They are also why the relation to the Global Neuronal Workspace Theory is contested. Baars counts the Dehaene–Changeux model as a member of a workspace “family” of theories (Baars et al., 2021). Changeux and Farisco (2026) resist the assimilation, describing GWT as a psychological, disembodied model open to functionalist interpretation and contrasting its diffuse reticular-thalamic substrate with the specified network of long-range cortical pyramidal neurons, inhibitory interneurons, and neuromodulatory inputs of their own model. Farisco recommends treating the two as separate theories that share “the high-level characterization of conscious processing as access of information to a central, global workspace” while differing anatomically and architecturally (Farisco, personal communication, 2026). Mashour et al. (2020) add that Baars’s original instantiation did not distinguish the level of consciousness, controlled by the reticular formation, from its content. Baars’s later insistence on a strictly biological, cortico-thalamic account narrows this gap.
Assessment
GWT’s influence rests on three features: an explicit method for treating consciousness as a variable, an architecture that explains the coexistence of limited conscious capacity with vast unconscious resources and competence, and predictions about widespread brain activity that later imaging work was taken to support. Its principal philosophical critics grant much of this while denying that it reaches phenomenal consciousness. Chalmers (1995) argues that nothing in the theory explains why globally accessible information is experienced; Block (2005) accepts the workspace as an account of access but not of phenomenality, to which he assigns a distinct neural basis.
Baars’s reply is that the distinction is not supported by the evidence and that a theory accounting for the contrastive data has addressed experience itself. Empirically, the anatomical openness of Global Workspace Dynamics makes it less exposed than prefrontal versions of workspace theory to disputes over the role of prefrontal cortex, but for the same reason harder to test decisively. Whether the workspace is best understood as a function that any suitably organized system could realize, or as a property of a particular biological organization, remains the open question that separates GWT from its neuronal descendant.
References
Baars, B. J. (1988). A Cognitive Theory of Consciousness. Cambridge University Press.
Baars, B. J. (1997a). In the Theater of Consciousness: The Workspace of the Mind. Oxford University Press.
Baars, B. J. (1997b). In the theatre of consciousness: Global workspace theory, a rigorous scientific theory of consciousness. Journal of Consciousness Studies, 4(4), 292–309.
Baars, B. J. (2002). The conscious access hypothesis: Origins and recent evidence. Trends in Cognitive Sciences, 6(1), 47–52.
Baars, B. J. (2005). Global workspace theory of consciousness: Toward a cognitive neuroscience of human experience. Progress in Brain Research, 150, 45–53.
Baars, B. J., & Franklin, S. (2003). How conscious experience and working memory interact. Trends in Cognitive Sciences, 7(4), 166–172.
Baars, B. J., Franklin, S., & Ramsøy, T. Z. (2013). Global workspace dynamics: Cortical “binding and propagation” enables conscious contents. Frontiers in Psychology, 4, 200.
Baars, B. J., Geld, N., & Kozma, R. (2021). Global workspace theory (GWT) and prefrontal cortex: Recent developments. Frontiers in Psychology, 12, 749868.
Baars, B. J., & Laureys, S. (2005). One, not two, neural correlates of consciousness. Trends in Cognitive Sciences, 9(6), 269.
Block, N. (2005). Two neural correlates of consciousness. Trends in Cognitive Sciences, 9(2), 46–52.
Chalmers, D. J. (1995). Facing up to the problem of consciousness. Journal of Consciousness Studies, 2(3), 200–219.
Changeux, J.-P., & Farisco, M. (2026). The Global Neuronal Workspace as a multilevel model of conscious processing. Trends in Cognitive Sciences. https://doi.org/10.1016/j.tics.2026.03.004
Edelman, G. M., Gally, J. A., & Baars, B. J. (2011). Biology of consciousness. Frontiers in Psychology, 2, 4.
Farisco, M. (2026). Personal communication to R. L. Kuhn.
Mashour, G. A., Roelfsema, P., Changeux, J.-P., & Dehaene, S. (2020). Conscious processing and the global neuronal workspace hypothesis. Neuron, 105(5), 776–798.
Newman, J., & Baars, B. J. (1993). A neural attentional model for access to consciousness: A global workspace perspective. Concepts in Neuroscience, 4(2), 255–290.
Shanahan, M., & Baars, B. (2005). Applying global workspace theory to the frame problem. Cognition, 98(2), 157–176.
VanRullen, R., & Kanai, R. (2021). Deep learning and the global workspace theory. Trends in Neurosciences, 44(9), 692–704.