Thagard’s Neural Representation, Binding, Coherence, Competition

Philosopher Paul Thagard poses big questions upfront. “Why do people have conscious experiences that include perceptions such as seeing, sensations such as pain, emotions such as joy, and abstract thoughts such as self-reflection? Why is consciousness central to so much of human life, including dreams, laughter, music, religion, sports, morality, and romance? Are such experiences also possessed by other animals, plants, and robots?” (Thagard, 2025).

Thagard's theory of consciousness “attributes conscious experiences to interactions of four brain mechanisms: neural representation, binding, coherence, and competition.” It distinguishes itself from current theories in several respects, he says. “The four brain mechanisms described are empirically plausible and clearly stated. Conscious experiences emerge from their interactions in areas across the brain.”

The mechanisms, he argues, “explain not only ordinary perceptual experiences such as vision, but also the most complex kinds of conscious experience including self-valuation, dreams, humor, and religious awe.” Moreover, he adds, “A crucial but often neglected aspect of consciousness is timing, but the four mechanisms fit perfectly with recent neuroscientific findings about how time cells enable brains to track experiences” (Thagard, 2025).

Earlier Formulation: Semantic Pointer Competition

The present four-mechanism theory developed from the Semantic Pointer Competition theory of consciousness, formulated by Paul Thagard and cognitive scientist Terrence C. Stewart. Drawing on Chris Eliasmith’s account of semantic pointers (Eliasmith, 2013), Thagard and Stewart proposed that consciousness results from three interacting neural mechanisms: representation by patterns of firing in neural populations, binding of representations into more complex representations called semantic pointers, and competition among semantic pointers to capture the aspects of an organism’s current state that are most important at a given time. Semantic pointers are not conventional abstract symbols. They are compressed neural representations that can bind sensory, motor, emotional, physiological, and verbal information while retaining connections to the lower-level representations from which they were constructed (Thagard & Stewart, 2014).

On this account, a semantic pointer contributes to consciousness only when it reaches a sufficient level of neural activation and prevails in competition with alternative representations. The qualitative character of an experience depends on the particular representations bound into the winning semantic pointer: pain, for example, may combine nociceptive information, bodily changes, emotional responses, situational representations, and cognitive appraisals. Binding helps explain the experienced unity of consciousness, while competition explains selective attention and transitions from one conscious content to another. Thagard and Stewart used neurally realistic computer simulations to model qualitative differentiation, the onset and cessation of consciousness, shifts in conscious contents, varieties of consciousness, unity and disunity, and the storage and retrieval of conscious experiences. They contrasted this biologically mechanistic account with Integrated Information Theory, arguing that consciousness should be explained through specific neural operations rather than identified with an abstract quantity of integrated information (Thagard & Stewart, 2014).

Thagard’s newer theory preserves the earlier architecture of representation, binding, and competition, but broadens and develops it in several respects. Most importantly, it makes coherence a fourth explicitly identified mechanism governing how mutually supportive representations become organized into stable conscious experiences. Coherence was already implicit in the earlier model’s account of unity, mutual enhancement, and the coordination of compatible semantic pointers, but it was not designated as a separate core mechanism. The newer formulation also ranges beyond the original emphasis on semantic-pointer competition to explain complex experiences—including dreams, humor, music, self-evaluation, and religious awe—and to assess possible consciousness in nonhuman animals and artificial systems. Semantic Pointer Competition is therefore best understood not as a discarded alternative, but as the more narrowly specified computational and neurorepresentational predecessor of Thagard’s expanded theory.

Attribution and Assessment

Thagard founds his theory on strict, empirically based neuroscience. His way of thinking is exemplified by his “Attribution Procedure,” an eight-step process for using what he calls “explanatory coherence” as a touchstone to establish “whether or not an animal or machine has a mental state, property, or process.” (Thagard, 2021, pp. 13–14).

For example, he offers twelve features of intelligence (i.e., problem solving, learning, understanding, reasoning, perceiving, planning, deciding, abstracting, creating, feeling, acting, communicating) and eight mechanisms to explain these features (i.e., images, concepts, rules, analogies, emotions, language, intentional action, consciousness). “All eight of these mental mechanisms can be carried out by a common set of neural mechanisms, many of which have been modeled computationally.”

This account of twelve features and eight mechanisms, Thagard says, “yields a twenty-item checklist for assessing intelligence in bots and beasts.” A similar way of thinking he applies to consciousness, stating that consciousness results from competition among neural representations (Thagard, 2021, pp. 3–4, 50, 49).

Scope and Implications

Claiming that his theory of consciousness possesses “the accuracy and breadth of application to mark a solid advance in the grand task of explaining how and why consciousness is so central to human life,” Thagard highlights an empirically supported explanation of consciousness resulting from the four brain mechanisms (i.e., neural representation, binding, coherence, and competition); application to a broad range of conscious experiences including smell, hunger, loneliness, self-awareness, religious experience, sports performance, and romantic chemistry; use of these four brain mechanisms to generate novel theories of dreaming, humor, and musical experience; a new theory of time consciousness; assessment of consciousness in non-human animals and machines, including the new generative AI models such as ChatGPT (√).

Working together, these four brain mechanisms, Thagard says, “explain the full range of consciousness in humans and other animals, and show why plants, bacteria, and ordinary things lack consciousness.” No current computers are conscious, he asserts, using a checklist of features and mechanisms of consciousness, “but the new generative models in artificial intelligence have similar mechanisms to humans that might enable some degree of consciousness.”

He concludes with high physicalist confidence: “Consciousness does not need to be a mystery once we understand how brains build it” (Thagard, 2025).