Program D · 2 sections
What differs between psychedelic-labeled sessions and ordinary days once the confound between them is faced, and what differs between the four classes.
Replicated within archiveThe first headline, more alpha under psychedelics, was eye closure (6.5× within one session); with eye state matched, alpha frequency orders LSD > DMT > ayahuasca > psilocybin within one brain.
16 — The confound Retrospective · exploratory
Analysis basis 56-recording cohort; 4 legacy + 11 fixed-block sessions for the eye-closure contrast · dataset release 2026.10.10 · last recomputed in analysis release 4.26 · current
This began as the most exciting finding in the archive: every psychedelic session showed dramatically more alpha power than every non-psychedelic one — the opposite of what the published literature reports. It survived about an hour of scrutiny. What replaced it is more useful than the finding would have been.
In August the same session recorded nine minutes eyes-open watching a show, then ten minutes eyes-closed listening to music. Posterior alpha went from 8 µV² to 54 µV². That one manipulation is larger than any psychedelic effect ever published — and the twenty recordings then labeled non-psychedelic (original 32-recording cohort; three were later reclassified as DMT) are overwhelmingly office work at a screen, while the twelve sessions then labeled acute are overwhelmingly resting, outdoors, in the evening, with eyes often closed.
The psychedelic label and the eye state were almost perfectly confounded. Measured against that same person's own eyes-closed baseline, eleven of the twelve sessions labeled acute at the time of that analysis come in lower, at a median of 0.61×.
| Comparison | vs own eyes-closed baseline | Direction |
|---|---|---|
| Ayahuasca experiment 005, early trip | 0.17× | −83% |
| LSD experiment 008, early trip | 0.53× | −48% |
| LSD experiment 011, early trip | 0.77× | −23% |
Hold eye state constant and the archive replicates the canonical alpha-suppression finding rather than contradicting it. The apparent reversal came entirely from comparing psychedelic-at-rest against non-psychedelic-at-work — the same shape as Simpson's paradox, where a within-group effect reverses when groups are pooled.
The consequence runs through everything. If the non-psychedelic comparison group is wrong, then every acute-versus-non-psychedelic contrast in the archive is contaminated the same way. So each measure was re-run against the four non-psychedelic recordings that are genuinely restful — evenings, eyes closed, high alpha — rather than against desk work.
Most of the "psychedelic signature" dissolved. Alpha power share is lower under psychedelics than during non-psychedelic rest. Spectral entropy barely moves. Standing-wave organization, harmonic bicoherence and global synchrony are indistinguishable. What remains, directionally, is a short list — and with only four resting non-psychedelic recordings, none of it reaches significance.
It was built on four resting recordings. Three turned out to be DMT sessions. The reason "psychedelic looks like non-psychedelic rest" was largely that three quarters of the rest group was psychedelic — a mislabeling, not a physiological equivalence.
The chart below is kept for the record, grayed. Do not read it as current. With one in-cohort non-psychedelic resting recording left, the contrast cannot be computed, and it stays withdrawn until there are enough confirmed non-psychedelic resting sessions to rebuild it.
What the correction does not touch: the decisive evidence in this section never used that group. It rests on one confirmed-non-psychedelic session — Resting baseline 001 — recorded eyes-open then eyes-closed minutes apart. That 8 µV² → 54 µV² swing is intact, as are the three eye-state-matched early-trip comparisons and the fact that eleven of the twelve sessions labeled acute at the time of that analysis sit below that eyes-closed reference. If anything the corrected picture is cleaner: the non-psychedelic eyes-closed baseline holds the highest posterior alpha in the whole archive, higher than any psychedelic session.
The within-session eye-state effect is no longer one observation. The fixed block that opens every session since v3 shows posterior alpha rising on eye closure in 14 of 14 channels in ten of the eleven scored sessions, by 2.0× to 13.7× (§18); the exception is Daily-life recording 005, the morning after the psilocybin session, where posterior alpha fell on eye closure (0.62×, 1 channel up) and the block cannot be trusted as labeled; with the v4 classifier the same effect is visible in every recording that contains both states. The 11-of-12 result above is untouched. The withdrawn psychedelic-versus-rest contrast stays withdrawn: the archive still holds only 146 minutes of non-psychedelic eyes-closed EEG outside stimulation (§25).
The confound of §16 has so far been argued; it can now be measured. A classifier trained on the fixed eyes-open / eyes-closed blocks of seven sessions — three features of posterior alpha, chosen over EEGNet and four EEG foundation models (LaBraM24, CBraMod25, EEGPT, BIOT) that scored the same in cross-validation but broke on older recordings — calls held-out sessions with 96.6% balanced accuracy, reads 0.0% eyes-closed across 170 minutes of desk work and 92% across the closed-eye sweeps. Applied to the archive it says what §16 suspected: the psychedelic sessions are a median 34% eyes-closed, the ordinary recordings 1.2%, and eye state alone separates the two groups with an AUC of 0.83.
The next step is to re-test every psychedelic-versus-ordinary contrast on this site using eyes-open windows only, where both groups have plenty of data (about 1,500 and 2,200 clean minutes), with a pre-stated rule for what counts as surviving. The eyes-closed comparison cannot be made at all yet: the archive holds 146 minutes of non-psychedelic eyes-closed EEG, nearly all of it under stimulation. Ten twenty-minute eyes-closed rest sessions would multiply that fifteen-fold, and are queued. Two further designs follow from §19: intensity levels in random order, to separate light level from elapsed time, and the fine sweep — 0.05 Hz steps around the predicted resonance — to measure a line the echo says is about 1 Hz wide.
17 — Four psychedelics compared Retrospective · exploratory
Analysis basis 56 for the frequency ladder and the recurring features (6 / 7 / 4 / 2 by class against 37 non-psychedelic); the within-session signal-diversity excursions are the September analysis on 41 · dataset release 2026.10.10 · last recomputed in analysis release 4.26 · current
LSD, DMT, ayahuasca and psilocybin, recorded from the same skull, the same fourteen electrode positions and the same eight months. Between-subject studies cannot ask what separates these compounds without the person separating them too. This archive can hold the person constant — and the comparison turns out to be the better-controlled half of the site, though not a controlled one.
The largest effect in this archive is not a psychedelic. It is eye closure, which moves posterior alpha by a factor of six within a single session, and it is confounded with the psychedelic label: acute sessions are evening, resting, mostly eyes closed, while the non-psychedelic recordings are largely desk work. That is the confound that dismantled the archive's first headline, and §16 takes it apart in full. It sits between the acute group and the rest of the archive. Within the acute group it is much reduced, not removed: the 16 sessions share the headset, the posture and broadly the hour, but the v4 eye-state labels show they are not matched on eye closure, which runs from 4% of clean time (psilocybin) to 67% (the second ayahuasca night). So the differences between the four classes below are the firmer half of this section; the similarities, which are measured against the non-psychedelic baseline, inherit the eye-state confound in full and should be read as "what these sessions look like", not "what the psychedelics do".
Sample sizes are LSD 6, DMT 7, ayahuasca 4, psilocybin 2, against 37 non-psychedelic recordings. A test is reported only between LSD and DMT, the two classes with more than five sessions. Ayahuasca, with four, and psilocybin, with two, are observations. Session length differs by class — DMT sessions are minutes, LSD sessions are hours — and that is not controlled.
Signal diversity is the measure the entropic-brain account rests on, so it is worth separating by compound. Whole-session medians are not the way to do it here: they are dominated by how much of each session was spent with the eyes closed, which runs from 4% to 67% across these sessions. Each session is therefore used as its own control — the largest smoothed rise in spatiotemporal Lempel–Ziv over that session’s own first fifth — and the identical peak-picking is run on the non-psychedelic sessions, because a maximum climbs with session length and noise whether or not anything happened. That null is a rise of +0.0775.
Pooled across classes the psychedelic rise is +0.1143 against the non-psychedelic +0.0775 (p = 0.017), and once the peak is restricted to windows matching each session’s own baseline eye state it is +0.0878 against +0.0773 (p = 0.453). So the excursion is larger under psychedelics than on an ordinary day, and it is LSD that carries it — but it stops being distinguishable as soon as the comparison is held to the session’s own eye state. Eye closure moves this measure by -0.0606 per unit on its own, which is 3 times the adjusted psychedelic term, so that is the obvious thing for the raw excursion to have been tracking. Whether anything survives it needs sessions matched on eye closure by design, not after the fact.
| Measure | Non-psychedelic | LSD (6) | DMT (7) | Ayahuasca (4) | Psilocybin (2) | Class spread (range of z) |
|---|---|---|---|---|---|---|
| Posterior alpha center frequency | 9.87 | 10.53 | 10.23 | 10.14 | 10.28 | 1.3 |
| Alpha peak above background | 0.56 | 1.37 | 1.55 | 1.39 | 1.08 | 2.0 |
| Alpha harmonic phase-locking z | 3.0 | 31.2 | 19.2 | 43.4 | 31.1 | 8.8 |
| Frontal–posterior alpha wPLI | 0.046 | 0.179 | 0.105 | 0.204 | 0.145 | 6.5 |
| Frontal–posterior envelope r | 0.35 | 0.73 | 0.74 | 0.77 | 0.67 | 0.7 |
| Alpha plane-wave score | 1.9 | 6.2 | 7.6 | 9.0 | 6.2 | 2.9 |
| Alpha state switches | 6.5 | 43.9 | 52.0 | 46.1 | 26.4 | 3.7 |
| DFA exponent | 0.720 | 0.815 | 0.758 | 0.831 | 0.757 | 2.2 |
| Right-alpha fraction | 0.94 | 0.71 | 0.77 | 0.72 | 0.61 | 1.8 |
| Phase-difference entropy | 0.953 | 0.824 | 0.876 | 0.808 | 0.845 | 3.4 |
Medians. "Class spread" is the range of the four class medians expressed in non-psychedelic median-absolute-deviations, so it is comparable across measures. Right-lean loss and phase order are signed so that a larger number means further from the ordinary recording.
Four compounds, four separate literatures, and almost no direct comparison: reviews of psychedelic EEG note that no modern study has recorded LSD, psilocybin, DMT and ayahuasca in the same people, which is the gap an N-of-1 archive is unusually placed to probe. What the separate literatures agree on is a shared effect — alpha power falls, most strongly over occipital cortex, at moderate to high doses of all four1,2,3,40 — and one divergent one: LSD shifts the alpha peak frequency upward, a modulation not reported for DMT and only inconsistently for psilocybin1.
| What the published work says | Where it comes from | What this archive shows | Verdict |
|---|---|---|---|
| Alpha power falls under all four, strongest occipitally | LSD1, psilocybin2,41, DMT3, ayahuasca40 | Replicated once eye state is held constant: 11 of 12 acute sessions sit below this person's own eyes-closed baseline, at a median of 0.61× (§16). Against desk work they look higher, which is the confound, not the psychedelic. | Agrees |
| LSD raises the alpha peak frequency; other compounds do not, or not reliably | LSD MEG1 | Agrees: LSD is highest at 10.53 Hz and every class sits above the ordinary 9.87 Hz (10.28 psilocybin, 10.23 DMT, 10.14 ayahuasca). The published contrast is LSD against the rest; here LSD leads and the three other classes sit together within 0.15 Hz, no longer in the graded order of September. | Agrees and extends |
| Signal diversity rises under LSD, psilocybin and DMT | MEG and EEG4,3, and the entropic-brain account5 | Depends on the comparison. Session against session there is no increase: 7.4% below all thirty-seven non-psychedelic recordings (p = 0.013) and 0.4% below the four resting recordings (p = 0.92), the comparison the published work makes (§07). Within a session there is one: the largest excursion above a session’s own opening fifth is +0.114 under psychedelics against +0.078 on ordinary days (p = 0.017), carried by LSD (+0.130, p = 0.003). It does not survive matching on eye state (p = 0.45): the psychedelic sessions move between eyes-open and eyes-closed five times as much as an ordinary day, which is what an excursion statistic rewards. Adjusted for eye state a small positive term remains (+0.020, 95% +0.003 to +0.036; eyes-open windows 2.6% higher, p = 0.044, eyes-closed nothing), but that comparison is resting against desk work, so task differs alongside the compound. Single-channel Lempel–Ziv is lower in every class than during desk work — low-amplitude task EEG binarizes into something that looks maximally complex (§07). | Mixed |
| Theta and delta emerge at the peak of the DMT experience | Intravenous DMT3 | Not seen. Theta share falls in all four classes, and no recording in the archive has a posterior theta peak above its own 1/f background (§10). The likeliest reconciliation is timescale: that finding is a minute-scale peak effect of an intravenous bolus, while these are whole-session medians of inhaled and ingested doses. | Disagrees |
| Ayahuasca reduces alpha and raises gamma 75–125 minutes in | Pharmaco-EEG mapping40 | Untestable here. Scalp gamma on this headset is dominated by jaw and neck muscle and is never reported as a finding (§25). | Cannot test |
| Alpha reduction tracks the intensity of visual and ego-dissolution effects | Psilocybin, 5-HT2A-mediated41 | Untestable here. No subjective ratings were logged for any session — the gap the current protocol closes (§25). | Cannot test |
| We did not identify a published analysis of alpha waveform shape or α–2α phase coupling under psychedelics | — | All four classes grow a phase-locked harmonic at twice the alpha frequency, two-fold apart in strength. Under LSD it lands at 21–22 Hz, where a band-power analysis would report it as a beta increase (§11). | Novel observation, as far as our search found |
| We did not identify a published comparison of all four within one nervous system | — | This section. One person, 6 / 7 / 4 / 2 sessions, no blinding and no placebo — so it is an ordering to be tested, not a result that transfers. | No prior within-person comparison identified; weakest possible design |
The pattern in that table is worth naming. Where the literature describes what the four compounds share, this archive agrees only once eye state is controlled. Where it describes what separates them, the one published dissociation — LSD's faster alpha — comes out cleanest here, as a graded ladder rather than a binary. And the two things this archive can say that the literature has not asked, the harmonic and the four-way comparison, are exactly what one brain measured many times is built to see.
One caution on the alpha-power row. Recent work notes that correcting for broadband noise can flip the sign of an apparent alpha change, since a shift in the 1/f background alone moves band power. Every spectral number on this site is measured above a fitted background for that reason (§25), which is why the alpha peak height here rises in all four classes while the raw band power falls — two statements about the same recordings that are not in conflict.
Recurring features, in different amounts. Across the four labeled conditions five signal features recur: an alpha oscillation that is larger, more strongly coupled to its own second harmonic, more coupled front to back, longer in its bursts and more willing to rearrange itself across the scalp. That is a set of shared electrophysiological features in this archive, not evidence of one discrete state — the learned representation of §21 finds a weak graded direction rather than two clusters, and these features are measured against an ordinary-day group that differs in eye state and setting. What separates the classes is magnitude: how fast the oscillator runs, how strong the harmonic coupling is, how organized the phase structure becomes, and how far the front of the head is drawn into step with the back.
The frequency ladder is the cleanest result here, and it is the one that would most easily replicate elsewhere — the more so because the one published frequency dissociation, LSD against the other compounds, points the same way. A candidate mechanism is available: in a thalamocortical resonator model the resonant frequency depends on loop delay and gain, so a compound that raises gain would raise frequency (§10, §18). That is a property of the model; the data fit it but do not single it out. The ladder runs LSD, DMT, ayahuasca, psilocybin — which is neither a potency ordering nor a duration ordering, and it disagrees with the orderings the harmonic and the phase-gradient measures produce, so whatever sets it is not simply "more of it".
What would sharpen it. Three sessions per class is the minimum this comparison needs; ayahuasca now has four and psilocybin two, and the second psilocybin session moved its class median by 0.28 Hz, which is the measure of how little two sessions settle. The photic sweep of §18 is the sharper instrument: it estimates the driven alpha resonance within a session rather than inferring it from the spontaneous peak, and one sweep per class would test the ladder in an afternoon each. Until then this section is an ordering, not a result — the strongest ordering the archive has, in the one comparison it is best placed to make.
A single-subject EEG archive: 56 recordings, 100.1 hours, 19 acute sessions, four psychedelic classes, 34 photic resonance sweeps, thirteen light-noise runs, one brain, February to October 2026. All 56 recordings form the analysis cohort.