06. Relaxing the Predictive Constraint

A cartographer redrawing an old map does not throw away everything on it. Rivers stay where rivers are. Coastlines stay where coastlines are. But a border that has been assumed fixed for years — one drawn from an old survey, never revisited — can turn out to be wrong, and until someone redraws it, every traveler who trusts the map walks into the wrong country. One influential idea about how psychedelic drugs affect mood suggests something similar happens inside the brain: not a chemical top-up, but a temporary loosening of an old, over-fixed map, giving the brain a chance to notice the borders it drew long ago and never checked again.

The brain, in one leading framework, is not a passive receiver of sensory information. It is a prediction engine. It builds a model of the world — and of the self — and constantly compares incoming signals against that model, adjusting the model when the mismatch is large enough. Depression, in this view, is not merely low mood; it is a model that has become too rigid. High-level beliefs — "nothing will improve," "I am a burden," "this feeling is permanent" — get assigned so much confidence, so much weight, that the brain stops updating them even when incoming evidence disagrees. The map is treated as more reliable than the territory.

Robin Carhart-Harris, working first at Imperial College London and later at UCSF, had spent years studying what psychedelic drugs do to the brain's overall pattern of activity, describing an increase in what he called neural entropy — a rise in the brain's moment-to-moment unpredictability, alongside a measurable loosening of the usual segregation between brain networks that normally stay fairly separate. In 2019, Carhart-Harris and Karl Friston, a theoretical neuroscientist at UCL known for the broader "free energy" framework of brain function, combined these threads into a model they named REBUS — relaxed beliefs under psychedelics. The proposal: serotonergic psychedelics reduce the precision, the confidence weighting, assigned to the brain's high-level prior beliefs, loosening their grip and allowing suppressed signals from deeper, more emotional brain systems to come through more strongly than usual. It is worth being clear about what this is and is not. REBUS is a theoretical, integrative framework, not a single experiment with a number attached — a lens built from prior imaging data, offered as the interpretive frame for understanding why a psychedelic experience might loosen an entrenched depressive thought pattern. And it is not settled: by 2023, some of the same authors had already published a revision, moving the model toward what they called REBAS — revised beliefs after psychedelics — an acknowledgment that the original formulation needed updating. It remains an active, debated theory, offered honestly as such.

The theory offers an explanation for why the experience itself, not just the molecule, might matter — why the same drug given in a sterile, unsupported setting and given with careful psychological preparation and support might leave behind a different result, since a temporarily loosened set of beliefs still has to be reorganized into something once the loosening ends, and how that reorganization happens is not automatic.

The theory's real test came from actual trials, and the first major one delivered a result that ran directly against the popular narrative it later got folded into.

What the 2021 trial actually found

In 2021, Carhart-Harris and colleagues published, in the New England Journal of Medicine, a double-blind, six-week trial comparing psilocybin — the compound in certain mushrooms — against escitalopram, a standard SSRI antidepressant already in wide clinical use. One group received two sessions of 25 milligrams of psilocybin, alongside psychological support and placebo capsules matched to look like the SSRI; the other received daily escitalopram alongside two sessions of a placebo version of the psilocybin session. The trial's pre-registered primary outcome — the single measure the whole study was designed and powered to answer — was the change in a standard depression scale, the QIDS-SR-16, at six weeks.

On that primary measure, there was no statistically significant difference between the two treatments. Not "psilocybin won." Not "psilocybin lost." Statistically indistinguishable from an SSRI already trusted by regulators and prescribed for decades. Several secondary measures did favor psilocybin, and the trial's authors and the broader research community discussed those secondary findings as promising — but the headline that circulated in general press coverage, "psilocybin beats SSRIs for depression," was not what the trial's own pre-specified primary result showed. It showed non-inferiority to an existing treatment, on a small, six-week study — itself a meaningful finding, since it suggested psilocybin was at minimum competitive with the standard of care, but a considerably more modest claim than the one that reached the public. A rigorous account of the trial has to hold both facts in the same sentence: real promise on secondary measures, and no significant advantage on the primary endpoint the trial was actually designed to test.

A larger trial the following year moved the question from "does it work at all" toward "at what cost." In 2022, COMPASS Pathways, an industry sponsor, working with academic investigators including researchers at King's College London, published a phase 2b trial in 233 patients with treatment-resistant depression — patients for whom multiple prior treatments had already failed. Participants were randomized to a single session of 25 milligrams, 10 milligrams, or 1 milligram of a proprietary psilocybin formulation, each with psychological support. Only the 25-milligram dose produced a statistically significant reduction in depression scores compared to the 1-milligram group, at three weeks, with the difference sustained through 12 weeks of follow-up — a genuine, dose-dependent signal for the higher dose.

The same trial reported adverse events in 179 of the 233 participants — 77 percent — including headache, nausea, and dizziness. And suicidal ideation, suicidal behavior, or self-injury occurred across all three dose groups, including the lowest, 1-milligram group that was meant to function as something close to an inert control. That last detail complicates the comparison at its foundation: if the "control" dose is not fully inert — if it still produces some psychoactive or expectancy effect — then the true gap between active treatment and no treatment is harder to pin down than the headline comparison between 25 milligrams and 1 milligram suggests.

The adverse-event rate and the suicidality signal across all arms are the trial's own most frequently cited caveats, not something uncovered by outside critics after the fact. A trial that finds a real, dose-dependent antidepressant signal and a 77 percent adverse-event rate, in the same 233 patients, has shown something more complicated than either "it works" or "it's dangerous" alone.

Running underneath both trials is the same structural problem that trailed ketamine in the previous chapter. Psilocybin's subjective effects are not subtle — altered perception, a departure from ordinary consciousness — and a participant experiencing them can generally tell, with reasonable confidence, that they did not receive the inert comparator. In a trial specifically designed to try to preserve blinding, roughly 80 percent of participants correctly guessed that they had received the active, moderate-or-higher dose. This is not one dissenting critic's complaint; it recurs across independent systematic reviews of the psilocybin depression literature, alongside other compounding concerns — small sample sizes, elements of open-label design, and waitlist-control structures that make an inert comparison group hard to construct in the first place. A 2025 methodological commentary described functional unblinding as an open, unsolved problem in psychedelic trial design, one the field has not yet built its way around. The consequence is not that these trials found nothing real. It is that no current trial design can fully separate a pharmacological antidepressant effect from the effect of a participant knowing, unmistakably, that something happened.

What stands after the caveats: a theoretical model, still being revised, for why loosening the brain's most confident beliefs might ease a depression built from beliefs held too rigidly; a large trial that failed to beat a standard antidepressant on its own primary measure, despite headlines to the contrary; a second, larger trial with a genuine dose-dependent signal purchased at a 77 percent adverse-event rate and a suicidality signal in every arm; and, threading through all of it, a blinding problem the field itself has not resolved.

Overturned if: a trial with a validated, effective blinding method — one where participants could not reliably distinguish active psilocybin from placebo — replicated a statistically significant antidepressant effect at a magnitude comparable to the unblinded trials, showing the benefit does not depend on participants knowing which arm they were in.

Sources

• 2019 · Carhart-Harris RL, Friston KJ, "REBUS and the Anarchic Brain: Toward a Unified Model of the Brain Action of Psychedelics," Pharmacological Reviews 71(3):316-344 — Wikipedia overview
• 2021 · Carhart-Harris R, Giribaldi B, Watts R, et al., "Trial of Psilocybin versus Escitalopram for Depression," New England Journal of Medicine 384:1402-1411 — NEJM
• 2022 · Goodwin GM et al., "Single-Dose Psilocybin for a Treatment-Resistant Episode of Major Depression," New England Journal of Medicine 387:1637-1648 — NEJM
• 2025 · "Commentary on Methods for Addressing Functional Unblinding and Mechanistic Uncertainty in Clinical Trials of Psychedelic-Assisted Treatments," Current Treatment Options in Psychiatry — Springer