Tsao Lab: Brain Toggles Seen and Unseen Faces 15x a Second

Neural activity in the monkey brain alternates about 15 times a second between representing what is consciously seen and what is suppressed from awareness, according to a bioRxiv preprint posted on 16 October 2026. The team is led by Janis Hesse and Doris Tsao of UC Berkeley.

Tsao wrote on X on 17 October that the finding “overturns decades of thinking on its biological basis and brings us significantly closer to the recipe for generating conscious awareness.” The paper has not been peer reviewed.

Business Pill · PRETRAINING

A Business Engineer explainer on how an AI model learns during pretraining. It explains a general AI concept and says nothing about the brain or about this preprint.

What the Team Recorded

The authors recorded from two rhesus macaque monkeys with Neuropixels probes placed in face patches of inferotemporal cortex and in two regions of prefrontal cortex, according to the paper.

The monkeys viewed a binocular flash suppression display: one face is shown to one eye, then a second face is flashed to the other eye, and the flashed image is the one perceived while the first is suppressed from awareness. The paper says the team first verified the effect in 14 human participants, who reported by button press what they saw.

Tsao’s thread describes recording “responses from thousands of neurons in inferotemporal and prefrontal cortex” while presenting “thousands of flash suppression stimulus pairs.”

What Was Posted on X

The first post of Doris Tsao’s 12-post thread on 17 October, announcing the preprint led by Janis Hesse and linking it on bioRxiv. The significance claims in the piece are hers and are quoted from this thread.

Tsao’s summary post in the thread: neural activity across inferotemporal and prefrontal cortex alternates between a conscious mode and an unconscious mode about 15 times every second.

What They Found

The abstract says that during a constant stimulus and percept, “the brain rapidly alternates, about 15 times per second, between conscious and suppressed micro-epochs.” Tsao’s thread puts the switch at “every ~ 60 ms.”

During the conscious micro-epochs, the paper reports, the team could reconstruct the consciously perceived face from neural activity; during the suppressed micro-epochs the same signal tracked the suppressed face, allowing reconstruction of the image the subject was not aware of.

At the level of single cells, the paper says, neurons responded selectively to the conscious stimulus in one kind of epoch and to the suppressed stimulus in the other. Tsao wrote that “the same population of neurons represent both what you’re aware of and what you’re not, at different moments of time.”

The discussion section says these switches in neural activity happen “over an order of magnitude faster (15 Hz) than perceptual switches (~0.5 Hz).” The authors add that prefrontal cortex showed the same interleaved structure.

Why the Authors Say It Matters

The paper opens by noting that earlier studies found high-level visual cortex only weakly modulated by conscious perception, “yet a pure signal of conscious perception has not been found anywhere in the brain.”

The authors write that their findings “expose a limitation of searching for the neural correlates of consciousness within the average activity of specific brain regions”, and the abstract concludes that “conscious access requires temporal selection beyond localization.”

In her thread, Tsao called the team’s result “a pure correlate of conscious awareness in the brain during rivalry,” and wrote that it makes it tractable “to understand how a pattern of neural activity accesses consciousness.”

What Remains Open

The paper names the next question itself: “why do representations in some micro-epochs reach awareness, whereas others do not?”

It also notes that the two kinds of epoch can be told apart by a linear decoder, which the authors say suggests the activity carries information about whether content is destined for awareness.

The Structural Read

As we read it, the preprint resolves a measurement puzzle the authors set out themselves: average responses during rivalry were only weakly modulated by conscious perception, and the paper reports that splitting activity into its two alternating states recovers a strong signal for each face.

As we read it, the reconstruction result is what makes the claim testable: the paper reports that both the consciously perceived face and the suppressed face could be reconstructed from population activity, depending on which micro-epoch the activity came from.

As we read it, the open question is now narrower: the paper asks why representations in some micro-epochs reach awareness and others do not, and notes that a linear decoder can tell the two kinds of epoch apart.

Three Implications

CONSCIOUSNESS RESEARCH The paper says the findings expose a limitation of searching for the neural correlates of consciousness within the average activity of specific brain regions.

BRAIN DECODING The paper reports reconstructing both the consciously perceived and the suppressed face from neural activity recorded with Neuropixels probes.

READERS OF THE PREPRINT The work is posted on bioRxiv and has not been certified by peer review; the recordings come from two rhesus macaque monkeys.

What Is Not Established

The work is a preprint on bioRxiv and has not been certified by peer review. The recordings come from two monkeys, with the perceptual effect of the display verified separately in human participants.

We read the preprint in full and Tsao’s thread on 17 October 2026; we did not contact the authors.

The Bottom Line

A Berkeley-led team reports that, in the monkey brain, the same neurons alternate about 15 times a second between representing a seen face and a suppressed one, and that both images can be reconstructed from that activity.

Tsao calls it a step toward the recipe for conscious awareness; the paper itself is unreviewed and names how content reaches awareness as the open question.