Predictive coding

Predictive coding describes perception as an active process: rather than recording the world in full, the brain uses prior experience to anticipate what sensory information means, then adjusts its predictions against incoming evidence. The model makes ordinary waking perception less like a transparent window than a working construction. In the book, it offers a framework for understanding why a familiar face can seem insistently present where it is not, why psychedelic experience may loosen accustomed interpretations, and why claims about such experiences are difficult to separate from expectation and experimental conditions.

Perception as Prediction

The brain’s filtering is presented as a practical economy. The senses provide more information than can be processed at once, so the brain admits what seems necessary and uses prior experience to fill gaps. Perception therefore involves both incoming signals and expectations: an observer does not simply receive the world but interprets it through a model formed by what has been encountered before. The same process that makes perception efficient can make it misleading. A few visual cues suggesting a face may be enough for the brain to insist on seeing a convex face, even when the object is concave. Michael Pollan’s attempted mask experiment during a psilocybin journey serves as a concrete example of a habitual prediction pressing perception toward the familiar interpretation.

This account also complicates the distinction between ordinary perception and more unusual states. Normal waking consciousness feels reliable partly because its predictions are continually tested against the world and against memory. A person can reach toward an apparent object or check, on waking, whether a remembered event actually occurred. Such reality testing helps stabilize everyday perception; it does not make perception a literal copy of reality. The model instead describes it as a construction that is ordinarily constrained by sensory evidence and shaped for everyday survival.

Psychedelic Perception and Unstable Predictions

The book connects predictive coding to psychedelic experience through the changing balance between prior expectations and incoming sensory information. Robin Carhart-Harris’s early psilocybin brain-imaging experiment did not show the anticipated increase in brain activity. Instead, the measurements indicated reduced activity, especially in the default mode network. The book describes this network as a filter involved in regulating mental activity and information entering consciousness. These imaging findings form part of the context for the predictive account, but they do not by themselves establish every proposed explanation of perception.

Under psychedelics, the usual coordination between predictions and sensory evidence may become less stable. Sometimes expectations appear to dominate: a mind trying urgently to interpret a flood of information may impose a pattern that is not there, as when faces are perceived in clouds. At other times, the filter may admit more sensory detail than usual. People who are color-blind have reported seeing colors, and research described in the book suggests that music may be heard with greater attention to timbre, the quality that contributes to its emotional character. Carhart-Harris proposes that the balance can shift back and forth: familiar concepts may lose their authority, allowing more sensory information through, while the mind then generates new interpretations to make sense of it. Predictive coding thus helps frame psychedelic perception not as one uniform condition, but as a fluctuating contest between expectation and sensory input.

The model is related to, but should not be confused with, the book’s discussion of expectancy effects. Psychedelic experiences are highly shaped by what people have been told to expect: expectations of spirituality, temporary insanity, or access to hidden mental material can influence what participants experience. The surrounding environment matters as well. Humphry Osmond and Abram Hoffer found that an engaged, empathetic therapist could help prevent difficult sessions, and suspected that some adverse reactions might reflect the impersonal conditions of a clinical room and clinician. These observations concern psychological expectations and treatment context as well as perception. They support the importance of considering prior beliefs and surroundings, but do not establish that every reported experience can be reduced to predictive coding.

Experimental Conditions and Therapeutic Claims

The history of LSD treatment for alcohol dependence shows how experimental design can change the experience being measured. At Weyburn, the Saskatchewan treatment produced results that led the provincial government to make LSD therapy a standard option for alcoholics. In an effort to test the treatment under more controlled conditions, Toronto researchers used neutral rooms, limited clinicians’ engagement with volunteers, and focused on questionnaires; participants were sometimes restrained or blindfolded. The results did not match those from Saskatchewan, and several volunteers had terrifying experiences. Critics could reasonably point out that the treatment performed less well under the controlled conditions; its supporters could reasonably argue that the impersonal setting removed conditions essential to its success. The comparison shows why isolating a drug’s effect is difficult when the setting itself may shape the response.

The Concord Prison Experiment offers a separate warning about how predictions and hopes can affect the interpretation of results. Timothy Leary reported that ten months after release, 25 percent of the psilocybin recipients had returned to jail, compared with 80 percent of the control group. Decades later, Rick Doblin’s subject-by-subject reconstruction found no statistically significant difference in recidivism. The discrepancy does not directly test predictive coding, but it illustrates the broader difficulty of turning a compelling account of change into secure evidence. Enthusiasm, study design, and the way outcomes are assessed all matter when a treatment’s effects are interpreted.

Promise, Correlation, and Limits

Later clinical findings renewed interest without resolving those methodological problems. In the New York University and Johns Hopkins cancer trials, about 80 percent of patients showed clinically significant reductions in anxiety and depression, and the effect lasted at least six months. The intensity of reported mystical experience closely correlated with symptom reduction. The studies included only eighty participants in total, however, and the book notes that larger trials would be needed before the government could consider rescheduling psilocybin and approving treatment. A correlation between experience intensity and improvement is significant evidence to investigate, not proof of a single mechanism.

A small Imperial College feasibility study led by Carhart-Harris examined psilocybin for treatment-resistant depression. Its first twelve participants had no control group and knew they were receiving psilocybin. All showed improvement after a week, and two-thirds were depression-free; seven still had substantial benefit after three months. When the study expanded to twenty participants, six remained in remission at six months while others had relapsed to varying degrees. The treatment was reported as well tolerated, with no adverse events, but the open design and small sample limited what could be concluded.

The book therefore treats predictive coding as an illuminating framework, not a complete explanation or a clinical verdict. Psychedelic research has repeatedly been vulnerable to exaggerated hopes, expectancy effects, and the special attention that small studies can provide. Early participants may be selected for their likelihood of responding, while skilled and dedicated therapists can contribute to outcomes; promising initial signals may not persist in larger populations. The model helps explain how prior expectations and sensory evidence may interact, and why changes in perception can feel so consequential. It also clarifies why those experiences require careful study: the mind is not merely receiving evidence, and the conditions under which it interprets that evidence are part of the phenomenon.