40 Hz

During REM sleep the body is paralysed but the eyes are not. A lucid dreamer who has agreed in advance to look hard left-right-left-right can therefore send a signal OUT of the dream and onto the polygraph, timestamped, while the EEG independently confirms REM. That is why lucid dreaming is a scientific object rather than a claim.

1. Before sleep, the dreamer and the experimenter agree a signal — the standard one is two full left-right eye sweeps, LRLR.
2. The sleeper is recorded with EOG (eye movement), EEG (brain state) and EMG (muscle tone) simultaneously.
3. On becoming lucid, the dreamer performs the agreed eye movements as a deliberate dream action.
4. The signal appears in the EOG trace during unambiguous REM, with the chin EMG flat. The dreamer was asleep, and reported it deliberately.
5. Because the signal is timestamped, anything else the dreamer does can now be timed against physiology — this is the technique every lab result on this page depends on.

Evidence. La Berge, Nagel, Dement and Zarcone (1981, Perceptual and Motor Skills 52:727-732) verified lucid dreaming in five selected subjects who signalled by prearranged dream actions during unequivocal REM sleep. Keith Hearne had recorded ocular signalling from a lucid dreamer independently in 1975 at the University of Hull; LaBerge’s group published the peer-reviewed paradigm. The eyes work because REM atonia spares the extraocular muscles — the one motor channel a dreaming body leaves open.

The honest framing: this proves the dreamer was awake INSIDE sleep and could act on a prior intention. It does not adjudicate what the dream is, and no experiment on this page does.

La Berge et al. 1981, Perceptual and Motor Skills — DOI 10.2466/pms.1981.52.3.727 · PubMed 24171230

Four independent laboratories, working separately and converging, put questions to sleeping people and got answers back. Arithmetic, yes/no, sensory discrimination — answered from inside verified REM sleep, in real time, by eye movements and facial muscles.

1. The dreamer signals lucidity with the LRLR paradigm above. The lab confirms REM from the EEG.
2. The experimenter speaks a question aloud — for example a small addition or subtraction with an answer between 1 and 6, or a yes/no question.
3. The dreamer answers by counting out eye movements, or by contracting the smile muscle for yes and the frown muscle for no.
4. On waking, the dreamer usually reports having received the question inside the dream — sometimes as a voice from outside, sometimes woven into the dream scene, sometimes with details that diverge from the recording.

Evidence. Konkoly and colleagues (2021, Current Biology 31:1417-1427.e6) ran 36 participants across four laboratories in the USA, Germany, France and the Netherlands, including one participant with narcolepsy type 1. Six participants achieved documented two-way exchanges. Across 158 trials in signal-verified lucid REM: 29 correct (18.4%), 5 incorrect (3.2%), 28 ambiguous (17.7%) and 96 with no response (60.8%). The low response rate is the honest headline — most questions got nothing. But 29 correct against 5 incorrect is not noise, and the four labs used different methods and different populations and all four got it.

What this established, stated precisely: a sleeping brain can perceive an external signal, comprehend it, compute on it and issue a voluntary motor response — all while asleep and dreaming. It is an interface, not a seance, and the paper claims nothing more than that. What a person MEETS in a dream is not a question this experiment is able to ask.

Konkoly et al. 2021, Current Biology — DOI 10.1016/j.cub.2021.01.026 · Open access — PMC8162929

REM atonia persisting into waking awareness. You are awake, you cannot move, and roughly a third of episodes come with a felt presence in the room. It is frightening, it is harmless, it ends by itself, and it is the single most likely thing to happen to someone practising WILD. Teaching WILD without teaching this would be the omission that does the damage.

1. What is happening: the muscle paralysis that keeps you from acting out dreams has outlasted the dream. Breathing is automatic and unaffected — the diaphragm is not part of REM atonia. The chest pressure people report is a perception, not a restriction.
2. What it feels like: inability to move or speak, often a presence in the room, sometimes a weight on the chest, sometimes vivid hypnagogic figures. Cultures name the figure differently — the Old Hag, the kanashibari, the jinn — and the phenomenology is remarkably stable across all of them.
3. How it ends: on its own, in seconds to a couple of minutes, always. Trying to force a large movement tends to prolong the panic; moving one finger or one toe, or slowing the breath, usually breaks it faster.
4. If you are practising WILD deliberately, expect it, and decide in advance that you will stay still and watch rather than fight. The people who describe it as a doorway and the people who describe it as an assault are in the same physiological state.
5. When it is NOT just this: recurrent sleep paralysis with daytime sleep attacks, cataplexy or hypnagogic hallucinations can be narcolepsy. That is a diagnosis and it is made by a clinician, not by a web page.

Evidence. Sharpless and Barber (2011, Sleep Medicine Reviews 15:311-315) aggregated 35 studies, total N=36,533: at least one lifetime episode in 7.6% of the general population, 28.3% of students and 31.9% of psychiatric patients (34.6% among those with panic disorder). It is common, and it is commonest in exactly the population that reads pages like this one.

Caution. Frequent, distressing episodes, or episodes with daytime sleepiness or cataplexy, belong with a sleep clinician. Knowing the physiology reduces the fear; it does not make recurrent paralysis a thing to manage alone.

Sharpless & Barber 2011 — DOI 10.1016/j.smrv.2011.01.007 · PubMed 21571556

The eye contains a third photoreceptor class that does not contribute to what you see. It contains melanopsin, peaks in the blue at around 460–480 nm, projects to the clock in the hypothalamus, and it is the pathway by which light suppresses melatonin and shifts circadian phase. This is not contested and it is one of the cleanest stories in biology.

1. Light in the morning advances the clock; light late in the evening delays it. Timing is not a detail — it is the sign of the effect.
2. The quantity that matters is not photopic lux, which is weighted for daytime vision, but melanopic EDI, which is weighted for melanopsin. Two rooms at the same 100 lux can differ FOURFOLD in mEDI depending on the lamp.
3. The healthy-adult consensus figures — ≥250 lx mEDI in the day, <10 lx mEDI in the three hours before bed, <1 lx mEDI asleep — are in integrations/light-signal.mjs with their source, and are a consensus recommendation rather than a standard anyone is obliged to meet.
4. Nothing here is a protocol for a condition. It is a description of a sensory pathway.

Evidence. Berson, Dunn & Takao (2002, Science 295:1070–1073) showed that retinal ganglion cells are themselves photosensitive and set the circadian clock; Hattar and colleagues (2002, Science 295:1065–1070) described the melanopsin-containing cells, their architecture and projections. The human action spectrum for melatonin suppression was measured INDEPENDENTLY, TWICE, in the same year — Thapan, Arendt & Skene (2001, J Physiol 535(1):261–267) and Brainard and colleagues (2001, J Neurosci 21(16):6405–6412) — and the two agree on a peak near 460–480 nm, which is not the peak of the visual system. Lockley, Brainard & Czeisler (2003, JCEM 88(9):4502–4505) showed the same short-wavelength sensitivity for phase resetting, not only for melatonin suppression. Two independent action spectra converging is the strongest form this kind of evidence takes.

⭐ The date is the whole reconciliation. This mechanism entered the literature around 2000. Ghadiali was convicted in 1948. A tradition that said "light and colour act on the body" was directionally right about a pathway nobody had found yet, and comprehensively wrong about what it did — and being right about the first does not retroactively license the second.

Berson, Dunn & Takao 2002, Science 295(5557):1070–1073 — doi:10.1126/science.1067262 · Hattar et al. 2002, Science 295(5557):1065–1070 — doi:10.1126/science.1069609 · Thapan, Arendt & Skene 2001, J Physiol 535(1):261–267 — doi:10.1111/j.1469-7793.2001.t01-1-00261.x · Brainard et al. 2001, J Neurosci 21(16):6405–6412 — doi:10.1523/JNEUROSCI.21-16-06405.2001 · Lockley, Brainard & Czeisler 2003, JCEM 88(9):4502–4505 — doi:10.1210/jc.2003-030570 · Brown et al. 2022, PLOS Biology 20(3):e3001571 — the consensus thresholds

The best-supported use of light in this whole library, and the magnitude is disputed by a factor of two. Three independent meta-analyses all find bright light beats a control in winter-pattern SAD; their pooled effect sizes range from 0.84 down to 0.37. The direction is replicated. The size is not settled, and the null rows are printed here alongside the positive ones.

1. The doses studied are 10,000 lux for about 30 minutes, or 2,500 lux for about two hours, at the eye, in the morning. The 2,500 lux/2 h form is the original from Rosenthal et al. (1984); the 10,000 lux/30 min form is what most later trials used.
2. ⭐ In melanopic terms — which is what the receptor actually responds to — a "10,000-lux" box is not one dose but a range. Oldham, Oldham & Desan (2019) measured 24 commercial devices: warm fluorescent efficacy ratio (melanopic lux ÷ photopic lux) mean 0.52, cool fluorescent 0.86, white LED 1.11. Two boxes both labelled 10,000 lux therefore deliver roughly 5,200 vs 11,100 melanopic lux — more than a factor of two apart. And only SEVEN of the 24 devices met all three of their clinical adequacy criteria.
3. Timing: morning is the studied condition and the one the phase-shift hypothesis predicts. But read the primary result carefully — see the evidence note. Morning superiority appears under strict remission criteria and NOT in continuous depression scores.
4. It takes weeks, not days. In the best-blinded trial the active/placebo separation did not reach significance until the third week.
5. None of this is a protocol we are handing you for a condition you have. Winter depression is a diagnosis and its management is a conversation with a clinician; what is on this page is what the trials did and what they found.

Evidence. THE POSITIVE HALF. Golden et al. (2005, Am J Psychiatry 162(4):656–662) pooled bright light in SAD at an effect size of 0.84 (95% CI 0.60–1.08, 8 studies) and dawn simulation at 0.73 (0.37–1.08, 5 studies). Pjrek et al. (2020, Psychother Psychosom 89(1):17–24) ran a newer and larger meta-analysis restricted to randomised blind trials of ≥1,000 lx against dim light (≤400 lx) or sham ion generators: standardised mean difference −0.37 (95% CI −0.63 to −0.12; 18 studies, 610 patients) and a response risk ratio of 1.42 (95% CI 1.08–1.85; 16 studies, 559 patients). Wan et al. (2025, Medicine 104(27):e43107), a network meta-analysis of 17 RCTs and 773 patients, ranked white light best of four visible colours (SUCRA 80.7% / 81.7%) with red light not distinguishable from placebo. Eastman et al. (1998, Arch Gen Psychiatry 55(10):883), the best-placebo-controlled single trial, found by strict criteria at four weeks: 61% responded to morning light, 50% to evening light, 32% to placebo. THE NULL HALF, WHICH IS THE SAME SIZE. (1) Golden’s fourth row: bright light as an ADJUNCT to antidepressants in NON-seasonal depression came out at −0.01 (95% CI −0.36 to 0.34, 5 studies) — essentially zero, and a source that quotes the first row and drops this one is not reporting a meta-analysis. (2) Eastman’s own headline: "There were no differences among the 3 groups in expectation ratings or mean depression scores after 4 weeks of treatment" — the benefit showed up only under strict remission criteria and took at least three weeks to appear. (3) Flory, Ametepe & Bowers (2010, Psychiatry Research 177(1–2):101–108), 73 women, four arms: "For raw scale scores, neither main effects of treatment nor interactions between treatment and time were significant"; all four groups, placebos included, improved significantly. (4) Legenbauer et al. (2024, JAMA Psychiatry 81(7):655), 227 adolescents randomised to 10,000 lux vs 100 lux placebo red light as an add-on to inpatient care: both arms improved by a mean of −7.5 BDI-II points (95% CI −9.0 to −6.0, Hedges g = 0.71) and bright light had NO significant group × time effect. That population is adolescent non-seasonal depression, not adult winter SAD, and the difference matters — but it is a large, genuinely double-blind, active-placebo null and it belongs on the same page as the 0.84. (5) Mårtensson et al. (2015, J Affect Disord 182:1–7) reviewed the same literature critically and concluded the evidence is not unequivocal, because the pooled result depends heavily on which studies are selected. AND ONE STRUCTURAL FACT: there is NO Cochrane review of light therapy as acute treatment for SAD. Cochrane has published four reviews in this family and all four are about PREVENTION.

⭐ ON MORNING VERSUS EVENING, READ THE PRIMARY RESULT AND NOT THE FOLKLORE. Terman, Terman & Ross (1998) randomised 158 people to six sequences of 10,000 lux for 30 min/day. Their own abstract: "Analysis of depression scale percentage change scores showed low-density ion response to be inferior to all other groups, WITH NO OTHER GROUP DIFFERENCES. ... Stringent remission criteria, however, showed significantly higher response to morning than evening light." So morning superiority is CRITERION-DEPENDENT: it is there under strict remission and absent from the continuous scores in the same trial. Eastman (1998) found the same shape — 61% / 50% / 32% under strict criteria, no difference in mean scores. The honest sentence is "morning is the studied condition and the better bet", not "morning works and evening does not". Graded STRONG on direction and replication, NOT on magnitude — the magnitude is unsettled by a factor of two and this entry prints both ends of it.

Caution. Reported side effects in the SAD trials are mostly mild and early — headache, eyestrain, jumpiness, nausea, mostly in the first days (Terman & Terman 1999, n = 83 at 10,000 lux). Two things are not mild and are for a clinician, not a page: light exposure can trigger a switch toward hypomania or mania in people with bipolar disorder, and it interacts with photosensitising medication and with pre-existing retinal disease. If any of those apply to you, this is a conversation with your doctor before it is anything else.

Golden et al. 2005, Am J Psychiatry 162(4):656–662 — doi:10.1176/appi.ajp.162.4.656 · Pjrek et al. 2020, Psychother Psychosom 89(1):17–24 — SMD −0.37 — doi:10.1159/000502891 · Wan et al. 2025, Medicine 104(27):e43107 — network meta-analysis — doi:10.1097/MD.0000000000043107 · Eastman et al. 1998, Arch Gen Psychiatry 55(10):883 — the placebo-controlled trial — doi:10.1001/archpsyc.55.10.883 · Terman, Terman & Ross 1998, Arch Gen Psychiatry 55(10):875 — n = 158, morning vs evening — doi:10.1001/archpsyc.55.10.875 · Flory, Ametepe & Bowers 2010, Psychiatry Research 177(1–2):101–108 — doi:10.1016/j.psychres.2008.08.011 · Legenbauer et al. 2024, JAMA Psychiatry 81(7):655 — the large null — doi:10.1001/jamapsychiatry.2024.0103 · Mårtensson et al. 2015, J Affect Disord 182:1–7 — the critical review — doi:10.1016/j.jad.2015.04.013 · Oldham, Oldham & Desan 2019, Psychiatr Res Clin Pract 1(2):49–57 — 24 devices measured — doi:10.1176/appi.prcp.2019.20180011 · Rosenthal et al. 1984, Arch Gen Psychiatry 41(1):72 — the founding description — doi:10.1001/archpsyc.1984.01790120076010 · Terman & Terman 1999, J Clin Psychiatry 60(11):798–808 — side effects — doi:10.4088/jcp.v60n1113