From Reducing Valve to REBUS: Huxley's Neuroscience, Seventy Years On

“Huxley proposed a mechanism in 1954. Carhart-Harris and Friston named it in 2019. The interval between them was sixty-five years of neuroscience arriving at the same place.”

In May 1953, Aldous Huxley took mescaline under Humphry Osmond's supervision and spent eight hours observing the world with an attention he had not previously been able to bring to it. The flowers on his table were not simply flowers, they were flowers as flowers actually were, without the filtering by biological utility and social convention that ordinarily reduced them to objects with names and functions.

To explain this, Huxley reached for Henri Bergson's 1896 proposal that the brain's primary function is not to produce consciousness but to restrict it. The brain, Bergson argued, is a reducing valve, it filters out most of what could be perceived, allowing through only what is relevant to survival and action. Huxley borrowed C.D. Broad's elaboration of this hypothesis and gave it the name it has carried since: the reducing valve.

Mescaline, in his account, impaired the valve. More of Mind at Large, the field of consciousness that the valve ordinarily suppressed, was allowed through. The result was not hallucination but perception: not the addition of something foreign but the removal of a filter that had been eliminating most of what was there.

For sixty-five years, this remained philosophy and phenomenology, coherent, intuitively compelling, and without rigorous neuroscientific support. The tools to test it didn't exist.

In 2019, Robin Carhart-Harris and Karl Friston published a paper in Pharmacological Reviews proposing the REBUS model, Relaxed Beliefs Under Psychedelics. Their framework, derived from the predictive coding theory of brain function, proposes that the brain operates as a hierarchical inference machine: it generates predictions about what it expects to perceive (top-down priors) and compares them against incoming sensory signals (bottom-up prediction errors). When predictions are accurate, sensory signals are suppressed, they don't need to reach higher processing because they match what the brain already predicted.

Psychedelic compounds, in the REBUS model, reduce the 'precision weighting' of these top-down priors. The predictions become less certain, less heavily weighted, which means they suppress incoming sensory information less aggressively. More bottom-up signal reaches awareness. The brain's filtering function is relaxed.

This is Bergson's reducing valve, formalized in the vocabulary of computational neuroscience, derived from fMRI studies of psilocybin and LSD, and expressed as a falsifiable hypothesis rather than a philosophical intuition.

The convergence between Huxley's 1954 proposal and the 2019 neuroscience is not coincidental, Huxley was describing a real phenomenological event, and his philosophical framework was an attempt to capture something that was genuinely happening in the brain. But the specificity of the confirmation is remarkable. Carhart-Harris and Friston are not simply saying that Huxley was approximately right. They are saying that the mechanism he described, impairment of the brain's filtering function, resulting in richer but less organized perceptual experience, is precisely what the neuroimaging evidence shows.

The default mode network (DMN) is the system most associated with self-referential processing, mind-wandering, and what neuroscientists call 'top-down' modulation of sensory experience. It is, in functional terms, the neural correlate of Huxley's reducing valve, the system that normally constrains perception within the boundaries of what is already known and expected. Under psychedelics, the DMN shows systematic suppression in fMRI studies. Its internal connectivity decreases. Its ability to modulate other networks is compromised.

The result, as Huxley described and as the clinical literature consistently confirms, is not chaos, it is a different kind of order. Subjects do not report random sensory noise; they report heightened salience of specific features of experience, a quality of direct presence with ordinary objects, and an inability to maintain the usual narrative framing that filters experience through its relevance to the self.

The self, in Huxley's phenomenological terms and in the REBUS framework's technical terms, is a product of the filtering process, the prediction-generating system's attempt to maintain a stable model of what things are in relation to the organism. When the filter relaxes, the self relaxes with it. Ego dissolution is not a bizarre drug effect. It is what happens to the self-model when the system that maintains it is temporarily suspended.

For clinicians, this has direct implications. Depression, in the predictive coding framework, has been proposed to involve over-rigid top-down priors, the system is too certain about negative predictions, and bottom-up signals that contradict them are discounted. Psilocybin-assisted therapy, by temporarily relaxing these priors, may create a window in which the predictions can be revised, in which the certainty that the world is hostile, that the self is defective, that recovery is impossible, can be loosened sufficiently to allow new experience to register.

This is what the clinical evidence suggests, and it is what Huxley's philosophical framework, worked out in an afternoon with mescaline and Bergson, predicted. The neuroscience has arrived at a more detailed account of the same territory he mapped with a metaphor. The metaphor was accurate.

Sources

• Huxley, The Doors of Perception (1954)
• Carhart-Harris & Friston, REBUS and the Anarchic Brain, Pharmacological Reviews (2019)
• Millière et al., Front. Psychol. (2018)
• Carhart-Harris et al., Default Mode Network Modulation by Psychedelics (2022) [Source not found; under review]
• Bergson, Matter and Memory (1896)