Tsao’s Compositional World Model of Symbol-Building Consciousness

Neuroscientist Doris Tsao, now Chief Scientist of Astera Neuro, leads a program built on the “concrete hypothesis” that “consciousness is the process by which the brain builds symbols”—“stable, persistent representations of the world and the self” that the brain composes “into an internal model of reality”—and on the conviction that consciousness, “a property of certain forms of organized matter,” can be not only explained but engineered by writing such symbols into the brain to evoke specific experiences (Astera Neuro, n.d.). In other words, Tsao account proposes that phenomenal consciousness consists in the brain’s construction and composition of persistent representations of the world and self. Its central hypothesis identifies experience with an organized representational process, while its experimental strategy seeks to discover that process’s elementary units and combinatorial rules by reading and manipulating neural activity.

Her most recent experimental contribution, with Janis Hesse and colleagues, shows that, during a constant stimulus and stable percept, the same neurons in inferotemporal and prefrontal cortex alternate roughly fifteen times per second between micro-epochs representing a consciously seen image and micro-epochs representing a suppressed one, so that conscious access requires “temporal selection beyond localization.” Thus, cortical populations interleave conscious and suppressed visual representations, constraining this broader account: detailed neural coding alone does not suffice for awareness; the theory must also explain how selected contents enter a persistent experienced world (Astera, n.d.-a, n.d.-b; Hesse et al., 2026).

On the combined picture, phenomenal consciousness is a physical, constructive process: the brain builds, selects, and composes object representations into a world-model, and what is experienced is that model, not the world directly. Astera’s stated mission is “to take consciousness from evolved to engineered,” and its method is mechanistic, aiming to move neuroscience “from observing correlates of experience to precisely generating specific percepts.”

Consciousness as Organized Matter

Astera Neuro’s statement opens with an ontological commitment: “Consciousness is not separate from the physical world, but is a property of certain forms of organized matter,” and “understanding how matter becomes conscious is the scientific challenge of our time” (Astera Neuro, n.d.). It holds that “systems neuroscience is finally ready to reveal how the brain generates conscious experience.” Tsao’s own remarks pair this physicalism with acceptance of experience as a given. She holds that “consciousness itself is something given”—“you just have to accept that certain complex systems are conscious”—while expecting science to discover “exactly how a set of neurons needs to be configured to be conscious or not” (Tsao, 2024). The existence of experience is thus accepted as a datum, not explained away, while its conditions are treated as fully tractable to neural mechanism.

Symbols, Objects, and the World-Model

The central hypothesis locates experience in representational construction. “Conscious experience begins with a stable, coherent model of the physical world, comprising persistent objects, and the relations between them,” and the brain’s object-tracking machinery, which separates what an object is from where it is, lies “at the foundation of all conscious experience” (Astera Neuro, n.d.). This continues Tsao’s theoretical work with Thomas Tsao (her father), arguing that the visual environment itself contains complete information for perceiving objects “as discrete, persistent units,” recoverable through a topological surface representation without learning (Tsao and Tsao, 2022). Thought is held to reuse the same machinery: abstract symbols flexibly recombined, with the claim that “the generativity of language and reasoning is inherited from the compositional structure of the physical world itself.” The view converges with Tsao’s account of perception as generative: “everything that we’re conscious of is generated by the brain,” and what we are aware of is “the top down process that actually recreates the world” (Tsao, 2024). Phenomenal content, on this picture, is the brain’s composed model of persistent objects, self, and their relations.

Temporal Selection in Cortex

The rivalry experiments address a narrower question: which of several competing representations becomes conscious. High-level visual neurons track the dominant percept during binocular rivalry (Sheinberg and Logothetis, 1997), yet consistently less strongly than when images are shown alone; with a no-report paradigm, Hesse and Tsao found that face cells encode both perceived and suppressed stimuli (Hesse and Tsao, 2020). Prefrontal recordings had suggested a purer correlate (Panagiotaropoulos et al., 2012; Kapoor et al., 2022), and the adversarial test of Global Workspace and Integrated Information predictions did not settle on a decisive locus (Cogitate Consortium, 2025). Using Neuropixels recordings from face patches ML and AM and from prefrontal cortex in two macaques fixating without report during binocular flash suppression, with percepts established in fourteen human observers, Hesse and colleagues partitioned single-trial activity into two states using face cells’ linear axis code (Chang and Tsao, 2017). One state reconstructed the seen face almost as well as responses to it shown alone; the other reconstructed the suppressed face with only slightly lower fidelity. States lasted about 60 ms, the same neurons served both, and cross-validation and model comparison favored coordinated alternation over averaging. The conscious state carried tuning gain about one-fifth higher and was linearly decodable, and prefrontal cortex showed the same structure (Hesse et al., 2026).

Routing, Access, and Phenomenality

On this account, suppressed content is not, as the view Tsao challenges holds, a residue that weakens toward the front of the brain and fails “to ignite” (Tsao, 2026a), but a full representation in the same feature space, alternating with the conscious one; weak rivalry modulation is an artifact of averaging across the two. The authors propose that a signature of conscious epochs determines whether content is dynamically routed to systems for report, introspection, and behavioral control, and, speculatively, that suppressed epochs let the brain explore alternative interpretations while conscious epochs express a working hypothesis (Hesse et al., 2026). They situate this within a broader principle of temporal multiplexing (Fiebelkorn and Kastner, 2019; Jezek et al., 2011; She et al., 2024). The paper explicitly sets aside why neural activity feels like anything, and Tsao has cautioned that it does not show that “consciousness has a frame rate” (Tsao, 2026b), although she has suggested elsewhere that conscious perception is sampled discretely, “just like a movie strip” (Tsao, 2024). The experimental work thus speaks to access; the symbol-building hypothesis carries the claim about phenomenal consciousness.

Engineering Experience

What distinguishes the program is its proposed test, testing the compositional hypothesis by writing predicted patterns into neural circuits and assessing whether the corresponding experiences result. Successful theory should predict what a written neural pattern manipulation makes a subject experience, rather than merely decode what an external stimulus has elicited (Astera, n.d.-b). Rather than correlating activity with reports, it seeks causal tools for high-density, chronic read-write access at single-neuron resolution, “so that we can write sequences of symbols into the brain to evoke specific experiences,” first in model systems and eventually in people (Astera Neuro, n.d.).

The program extends to artificial systems, asking “whether such systems can experience anything at all” and “how to know whether any system, biological or artificial, is conscious at all,” and to whether “a mind can be extended, merged, or carried onto a new substrate” (Astera Neuro, n.d.). Consistently, Tsao regards monkey visual experience as very likely similar to ours and holds that anything with all our behaviors is likely conscious (Tsao, 2024).

There is a relevant experimental precedent: face-patch microstimulation altered macaques’ performance in perceptual matching tasks involving faces and some nonface objects, supporting a causal contribution to perception (Moeller et al., 2017). Such interventions do not establish a complete neural syntax, however. The stronger program requires controlled, generalizable construction of content according to independently specified representational rules.

Explanatory Scope and Status

The account now reaches beyond locating neural correlates: it proposes an ontology of organized physical representation, a compositional architecture, and a causal standard for testing content generation. Nevertheless, symbol construction must be distinguished from unconscious representation, and compositional capacity from phenomenality. A decisive theory must specify which physical organization makes representations experienced rather than merely processed.

The alternation findings remain a preprint concerning selected visual populations, without trial-by-trial perceptual reports from the recorded macaques. Detailed reconstructions pooled state-conditioned activity across trials and sessions. Moreover, that paper explicitly brackets why neural activity feels like anything (Hesse et al., 2026). The broader Astera program aims to explain and engineer experience; the present evidence supports components of that undertaking without establishing its central constitutive claim.

References

Astera Institute. (2025, December 10). Neuroscientist Doris Tsao joins Astera to lead its new neuroscience program. https://astera.org/neuroscientist-doris-tsao-joins-astera-to-lead-its-new-neuroscience-program/

Astera Neuro. (n.d.-a). Neuro: Understanding and engineering consciousness. Astera Institute. Retrieved October 27, 2026, from https://astera.org/neuroscience/

Astera. (n.d.-b). Neuroscientist Doris Tsao joins Astera to lead its new neuroscience program. Program statement.

Chang, L., and Tsao, D. Y. (2017). The code for facial identity in the primate brain. Cell, 169(6), 1013–1028.

Cogitate Consortium, et al. (2025). Adversarial testing of global neuronal workspace and integrated information theories of consciousness. Nature, 642, 133–142.

Fiebelkorn, I. C., and Kastner, S. (2019). A rhythmic theory of attention. Trends in Cognitive Sciences, 23(2), 87–101.

Hesse, J. K., Lanfranchi, F. F., McGrory, C., Shi, Y., Engel, T. A., and Tsao, D. Y. (2026). Rapid alternation of conscious and suppressed stimulus representations in the primate brain. bioRxiv (posted October 16, 2026; not peer reviewed). https://doi.org/10.64898/2026.10.15.759299

Hesse, J. K., and Tsao, D. Y. (2020). A new no-report paradigm reveals that face cells encode both consciously perceived and suppressed stimuli. eLife, 9, e58360.

Jezek, K., Henriksen, E. J., Treves, A., Moser, E. I., and Moser, M.-B. (2011). Theta-paced flickering between place-cell maps in the hippocampus. Nature, 478, 246–249.

Kapoor, V., Dwarakanath, A., Safavi, S., Werner, J., Besserve, M., Panagiotaropoulos, T. I., and Logothetis, N. K. (2022). Decoding internally generated transitions of conscious contents in the prefrontal cortex without subjective reports. Nature Communications, 13, 1535.

Panagiotaropoulos, T. I., Deco, G., Kapoor, V., and Logothetis, N. K. (2012). Neuronal discharges and gamma oscillations explicitly reflect visual consciousness in the lateral prefrontal cortex. Neuron, 74(5), 924–935.

She, L., Benna, M. K., Shi, Y., Fusi, S., and Tsao, D. Y. (2024). Temporal multiplexing of perception and memory codes in IT cortex. Nature, 629, 861–868.

Sheinberg, D. L., and Logothetis, N. K. (1997). The role of temporal cortical areas in perceptual organization. Proceedings of the National Academy of Sciences, 94(7), 3408–3413.

Tsao, D. (2024, July 29). Doris Tsao on how the brain turns vision into the world [Interview by S. Carroll]. Mindscape, Episode 284. https://www.preposterousuniverse.com/podcast/2024/07/29/284-doris-tsao-on-how-the-brain-turns-vision-into-the-world/

Tsao, D. Y. [@doristsao]. (2026a, October 17). “1/ We discovered something totally unexpected about the neural structure of consciousness…” [Thread]. X. https://threadreaderapp.com/thread/2111316522098733078.html

Tsao, D. Y. [@doristsao]. (2026b, October). “The outpouring of interest in our paper has been … a bit insane. May I make four points?” [Post]. X. https://x.com/doristsao/status/2111859732806074812

Tsao, T., and Tsao, D. Y. (2022). A topological solution to object segmentation and tracking. Proceedings of the National Academy of Sciences, 119(41), e2204248119.