Sejnowski’s Computational Two Tiers and Traveling Waves

Computational neurobiologist Terrence J. Sejnowski’s theory of consciousness is a computational framework that focuses on the global dynamics of spiking neurons and traveling waves. He proposes that consciousness arises from a sophisticated, "two-tier" organizational structure within the cerebral cortex, where high-level cognition coexists with, but remains functionally distinct from, standard sensorimotor processing (Sejnowski, 2015, 2026).

Sejnowski’s theory descends directly from the research program initiated by Francis Crick, and then transforms the 20th-century focus on visual awareness into a 21st-century focus on "two-tier" cortical processing and high-precision temporal coding. Sejnowski suggests that “With a deeper understanding of the brain mechanisms that govern perception, decision-making, and planning, the problem of consciousness could disappear like the Cheshire cat,” leaving behind only the "broad grin" of scientific understanding (Sejnowski, 2015).

Two-Tier Architecture and Spike-Timing Precision

The foundation of Sejnowski’s theory is a radical reorganization of cerebrocortical function via a "two-tier" hypothesis. For much of the last century, neuroscience relied on the "neuronal rate coding" model, which suggests that neurons convey information primarily through their average firing rates. While this "Tier 1" sensorimotor tier is efficient for fast, real-time interactions with the environment—averaging a firing rate of approximately 1 Hz (0.1 Hz to 10 Hz) during routine tasks and more than 100 Hz (even up to 200 Hz) during active movement or sensory stimulation—it is insufficient to explain the complexities of high-level cognition and self-generated thought. Sejnowski proposes a second, "Tier 2" cognitive tier that "rides astride" this sensorimotor network. This second tier utilizes millisecond-precision spike timing and temporary synaptic plasticity to support cognitive processing and long-term working memory (Sejnowski, 2026).

Sejnowski says, “The discovery of millisecond-precision spike initiation in cortical neurons was unexpected (Mainen and Sejnowski, 1995). Even more striking was the precision of spiking in vivo, in response to rapidly fluctuating sensory inputs, suggesting that neural circuits could preserve and manipulate sensory information through spike timing. High temporal resolution enables a broader range of neural codes. The relative timing of spikes between presynaptic and postsynaptic neurons in the millisecond range triggers spike-timing-dependent plasticity (STDP)” (Sejnowski, 2026).

Sejnowski’s move toward this temporal precision was supported by the discovery that cortical pyramidal neurons are remarkably reliable when responding to fluctuating inputs. Reflecting on the experimental freedom that led to this insight, he notes: "One of the advantages of having a wet lab is that you do not have to wait for someone else to do your experiment.”

By injecting "frozen noise" into neurons, Sejnowski’s lab demonstrated that spike initiation is surprisingly robust, organized by a post-inhibitory rebound mechanism. In this model, inhibitory inputs from parvalbumin-positive (PV) basket cells hyperpolarize the neuron, de-inactivating sodium channels and priming the cell for a precisely timed spike. This precision is not mere noise; it is the fundamental language of the cognitive tier, allowing the brain to perform complex computations that rate codes alone cannot sustain (Sejnowski, 2026).

Traveling Waves as a Spacetime Code for Context

A central pillar of Sejnowski's theory is the role of traveling waves in organizing these precisely timed spikes into a "spacetime population code". Unlike synchronous oscillations, where neurons fire at the exact same moment, traveling waves are "wave packets" that propagate across the cortex, mixing spatial and temporal information. Sejnowski draws a powerful analogy between these dynamics and holography, where a three-dimensional object is reconstructed from a two-dimensional photographic plate. In the brain’s "holographic" representation, time serves as one of the critical dimensions, allowing a single population of neurons to represent a window of time rather than just a fleeting moment (Sejnowski, 2026).

These waves provide the necessary temporal context for consciousness by extending the brain’s integration window. As information moves up the cortical hierarchy, this window expands from 100 milliseconds in the primary visual cortex to as much as 10 seconds in the prefrontal cortex. Sejnowski argues that this mechanism mirrors the "self-attention" found in modern AI Transformers, where the meaning of a word is extracted from its relationship to other words across a sentence. He observes: "Spacetime codes in cortical association areas can represent, within a single population, a window of time spanning both signals, like words in a sentence" (Sejnowski, 2026).

This Tier 2 network, organized by traveling waves and temporary synaptic plasticity (STDP), essentially creates a "global workspace" for cognition that can last for hours. This theory also offers a parsimonious explanation for the Blood Oxygen Level Dependent (BOLD) fMRI signal, which Sejnowski argues reflects the metabolic cost of this synaptic plasticity and attention rather than simple spiking activity.

As he puts it, “Cortical traveling waves have been observed across many frequency bands with high temporal precision, and neural mechanisms can plausibly enable traveling waves to trigger STDP lasting for hours in cortical neurons. This temporary cortical network, riding astride the long-term sensorimotor network, could support cognitive processing and long-term working memory” (Sejnowski, 2026).

Postdiction and the Subjective Unity of the "Now"

Perhaps the most counterintuitive aspect of Sejnowski’s theory is the assertion that conscious awareness is postdictive rather than purely predictive. He argues that our brain is constantly "revising history" to create a unified experience from the disparate, delayed signals it receives from the senses. This is evidenced by the "flash-lag effect," where a flash and a moving object at the same location appear offset because the brain’s perception depends on events occurring in the 80 milliseconds after the flash (Eagleman and Sejnowski, 2000). Sejnowski defines this temporal revision as a core feature of awareness: “The brain is postdictive rather than predictive; that is, the brain is constantly revising history to make the conscious present consistent with the future".

This ability to unify experience also depends on a specific thalamic architecture. While the "core" thalamic system supports Tier 1 sensorimotor control, the "matrix" thalamic system projects widely to the superficial layers of the cortex and thus controls Tier 2 cognitive processing. Sejnowski highlights the vital importance of this matrix system by noting that even minor lesions in the intralaminar nuclei can lead to a total loss of consciousness. Thus, consciousness is the result of a massive, global coordination of these matrix pathways, organizing self-generated activity—thinking—that continues even when external prompts stop (Sejnowski, 2026).

Thinking as Self-Generative Activity

Sejnowski’s theory of consciousness ultimately distinguishes biological consciousness from current artificial intelligence by its "self-generative" nature. While Large Language Models (LLMs) possess feedforward hierarchies that cease activity when a prompt ends, the human brain maintains massive feedback loops and recurrent connectivity that support internal attention and decision-making without external input. By implementing "System 2" capabilities—self-generative activity organized by traveling waves and STDP—it may be possible to endow future machines with a form of cognition that more closely resembles the human experience.

To Sejnowski, “Many philosophical questions about the ‘mind’ could have scientific answers if we knew more about the neural mechanisms underlying cognition. Descartes (1637) put it simply: ‘Cogito, ergo sum.’ A corollary of this dictum is: ‘Once we understand thinking, we will know ourselves.’ We should be open to more surprises as we explore the global dynamics of spiking neurons and their underlying subthreshold dynamics over a wide range of time scales.”

Ultimately, Sejnowski’s framework suggests that consciousness is the global dynamic state of a brain that is "thinking about thinking". It is a system that uses its Tier 2 network to build an internal representation from prior experiences, effectively pulling context across time to guide future actions. As he concludes, understanding the mechanical "how" of this self-generation is the key to knowing ourselves.

References

Eagleman, D. M. & Sejnowski, T. J. (2000). Motion Integration and Postdiction in Visual Awareness. Science, 287, 2036-2038.

Mainen, Z. F. & Sejnowski, T. J. (1995). Reliability of spike timing in neocortical neurons. Science, 268, 1503-1506.

Muller, L., Churchland, P. S., & Sejnowski, T. J. (2024). Transformers and cortical waves: encoders for pulling in context across time. Trends in Neurosciences, 47(10), 788-802.

Sejnowski, T. J. (2015). Consciousness. Dædalus, 144(1), 123-132.

Sejnowski, T. J. (2018). The Deep Learning Revolution. The MIT Press.

Sejnowski, T. J. (2026). Dynamical Mechanisms for Coordinating Long-term Working Memory Based on the Precision of Spike-timing in Cortical Neurons. Computational Neurobiology Laboratory.

Tiesinga, P., Fellous, J. M., & Sejnowski, T. J. (2008). Regulation of spike timing in visual cortical circuits. Nature Reviews Neuroscience, 9, 97-109.