The Verification Venue · a folk belief, put on the instrument
Much Ado About the Full Moon (Except Your Sleep)
The word lunatic is a bet the sky lost. Pool the studies and the full moon moves crime, births, and emergency rooms by essentially nothing. The honest twist: it does nudge one thing (when you fall asleep), and the reason kills the folklore, because the driver is moonlight, not gravity.
≤ 1% of the variance
No, the full moon does not affect behavior. Across 37 pooled studies of psychiatric admissions, crisis calls, homicides, other crime, and suicides, lunar phase explained no more than 1% of the variance (Rotton and Kelly, 1985): a near-perfect null that failed replicability, significance, and predictability. The lone real exception is sleep, and it runs on light, not tides.
People have been sure of this one for two thousand years. Luna gave us lunatic; the werewolf rises on the full moon; nurses and police still trade full-moon-shift stories. The belief is old, vivid, and self-reinforcing. It is also, when you actually count, not there. Rather than take that on faith, this page hands you the two pieces you need to check it yourself: a moon-phase engine that needs no data, and 15 years of real daily births to point it at.
A moon phase you can check
First, the honest half of the folklore: the phase is real and exactly computable. The routine below is the Meeus phase-angle series (mean elongation of the Moon from the Sun, plus the main periodic corrections). It takes a date and returns the illuminated fraction of the disc, 0 at new moon, 1 at full. No data, no lookup, no fudge. Pick a full moon you remember and watch it land near 100%.
Try 2024-09-18 (a full moon), 2024-10-02 (a new moon), or your own birthday. The number is computed from first principles, not read from a table.
These are published full/new-moon dates. The engine reproduces them; that is the proof the phase column is real before we correlate anything with it.
Now hunt for the spike
Here are 5,479 real days of US births (Social Security Administration counts, 2000 through 2014). Births are a clean null test: if the moon pulled on bodies you would expect it here, and multiple large studies find it does not. The page assigns every one of those days a lunar phase with the engine above, sorts them into eight phase bins from new to full and back, and draws the mean births per bin. Believers expect a tower at "full." Watch the bars stay flat.
Full-moon bin vs the rest
–
a rounding error, not a spike
Variance explained by lunar phase (η²)
–
Rotton & Kelly ceiling: ≤ 1%
Correlation r (= √η²)
–
their finding: r ≤ ~0.01
Where the "spike" comes from: the weekend
The belief is not stupid; it is a misread calendar. The meta-analysis names the real culprit: people fail to control for the weekly cycle. In this exact births data the day of the week is enormous (weekday inductions and C-sections pile births onto Tuesday through Friday and empty out the weekend), while lunar phase is a whisper. Compare the two variance bars.
"Count honestly": every day binned by phase, no cherry-picking. "Count like a believer": keep only the full-moon nights that happened to fall on a busy weekday, then compare to the overall average, exactly the selective bookkeeping the meta-analysis blames for the myth.
Full-moon nights vs average
–
honest count: no effect
Same, after controlling for day of week
–
the bump evaporates
Flip to Count like a believer and a +12% full-moon "surge" appears out of thin air. It is entirely the weekday effect: full moons that happened to fall Tuesday-through-Friday inherit the busy-weekday birth rate. Control for the day of the week and it collapses to a fraction of a percent, the same near-zero the honest count gave. That is the whole trick of lunar lore in one toggle: an uncontrolled weekly cycle wearing a lunar mask.
The one real exception: sleep
Here is where the honest answer stops saying "no" flatly. Behavior is null, but sleep is not, and it is the twist that makes this the complete answer rather than a debunk. Casiraghi and colleagues (Science Advances, 2021) tracked sleep by wrist actigraphy in three Toba/Qom communities in Argentina that differ in access to electric light: one rural with no electricity, one rural with limited electric light, and one urban. In all three, people fell asleep later and slept less in the roughly three to five nights before a full moon. The same oscillation turned up again in an independent group of urban college students in Seattle, which rules out anything peculiar to one culture or place.
Slide across the three communities. Both low-light rural settings show a large oscillation; it is clearly smaller in the city, the fingerprint of a light mechanism. (The two rural values sit inside each other's confidence intervals, so read the contrast as rural versus urban, not a strict three-step ladder.)
A reconstruction of the paper's fitted sleep-onset oscillation, amplitudes set to its reported per-community values (about 6.4 min in the city, 10 to 12 min in the two rural communities). Peak sits before full moon, when a bright moon is already up just after dusk.
Peak onset delay (before full moon)
–
vs your own monthly average
Across-cycle sleep swing
46–58 min
reported total-sleep difference (all sites)
Driver
evening moonlight
not gravity, not tides
Why moonlight, and not the moon's pull
The tell is where the effect is strong: it is larger in the two rural communities, where electric light is scarce, than in the city. Gravity does not care whether your village has a power line. Light does. The moon's tidal force on a human body is physically negligible (a person standing next to you pulls on you far harder than the Moon does), so tides and "lunar force" are ruled out. What is left is the obvious thing: in the nights before a full moon the Moon is already bright and high in the early evening, so there is extra light in the hours after dusk, and extra evening light delays sleep onset. Same reason a phone screen keeps you up.
Footnote, not headline: a small 2013 lab study (Cajochen et al., ~20 min less sleep and ~30% less deep sleep near full moon) is often quoted, but it was a tiny sample and Cordi et al. failed to replicate it in 2014. The load-bearing evidence is the large field study above; treat the lab figures as suggestive, not settled.
The whole answer, in one table
| Outcome | Lunar effect? | Best evidence |
|---|---|---|
| Psychiatric admissions | No (≤1% variance) | Rotton & Kelly 1985 (37 studies) |
| Homicides & other crime | No (≤1% variance) | Rotton & Kelly 1985 |
| Crisis-center calls | No (≤1% variance) | Rotton & Kelly 1985 |
| Suicides | No (≤1% variance) | Rotton & Kelly 1985 |
| Births (timing) | No (η² ≈ 0.01% here) | This page + large birth studies |
| ER load | No reliable effect | CMAJ 2005 review |
| Sleep onset & duration | Yes, small & real | Casiraghi et al. 2021 (field, 3 communities) |
Everything on the behavior side is a null of at most one percent. The only "yes" is sleep, it is small (tens of minutes of timing, not a personality change), and its mechanism is a lamp in the sky, not a force on your blood.
The check: every number recomputed in front of you
Nothing here is a stored figure. The phase comes from the Meeus series; the variance figures come from binning the 5,479 real daily birth counts embedded in this page. Live values:
Free choices & uncertainty. The phase engine samples each day at noon UTC and uses eight phase bins; the null holds for other bin counts and sampling times (the effect is simply absent). Births are the null demonstrator here, not proof about every outcome, but they agree with the behavior meta-analysis and with dedicated birth-timing studies. "≤1%" is Rotton and Kelly's own ceiling: a handful of significant relations did surface across 37 studies, but none exceeded ~1% of variance, so this is a near-perfect null, not literally zero. The +12% believer bump is real arithmetic on a cherry-picked subset; it is an artifact of the weekly cycle, shown here to be dismantled. The sleep curve is a stylized reconstruction of Casiraghi et al. 2021 (direction and reported amplitudes), not their raw actigraphy. The offline gate reproduces the phase computation, the variance figures, and the sleep direction two ways: node research/does-the-full-moon-affect-behavior/verify-does-the-full-moon-affect-behavior.mjs.
What is exactly true here, and what is a model
Exactly computed. The illuminated fraction of the Moon for any date, from the Meeus phase-angle series; it reproduces published full and new moon dates to well within a day. The variance figures (η² for lunar phase and for day of week) are ordinary one-way ANOVA on the embedded real birth counts, recomputed live. The +12% believer bump and its collapse under day-of-week control are exact arithmetic on that same data.
The behavior null. Rotton and Kelly (1985) pooled 37 published and unpublished studies of lunar phase against mental-hospital admissions, psychiatric disturbances, crisis calls, homicides, other crimes, and suicides, and concluded lunar phase "accounted for no more than 1% of the variance," failing replicability, statistical significance, and predictability. Later reviews (CMAJ 2005) concur: no reliable lunar effect on crime, births, ER load, or psychiatric episodes. This page does not re-derive their pooled estimate; it demonstrates the same null on an independent real series (births) and reproduces their named confound (the weekly cycle).
The sleep exception (a real effect, a modeled curve). Casiraghi et al. (2021) found, by actigraphy across three Toba/Qom communities in Argentina (rural no-electricity, rural limited-light, and urban), that sleep onset is later and duration shorter in the ~3–5 nights before full moon, with total sleep differing 46–58 min across the cycle and bedtimes shifting ~30 min. Their fitted onset-delay amplitudes were about 10.0 min (rural, no electricity), 12.1 min (rural, limited light), and 6.4 min (urban); a separate Seattle student cohort showed the same oscillation on a standardized (z-score) scale. The confidence intervals for the three Argentine communities overlap, so read the contrast as the two low-light rural communities larger than the city, not as a clean three-step ranking. That rural-versus-urban contrast is the evidence for an evening-moonlight mechanism rather than gravity or tides. The on-page curve is a sinusoid tuned to those reported numbers; read it for direction and scale, not as raw data.
Mechanism discipline. The Moon's tidal force on a human body is negligible; do not read the sleep effect as gravity, tides, or any "lunar force." Nothing here supports the moon moving crime, births, or emergency-room chaos. The etymology (luna → lunatic) and werewolf lore are the belief's origin, not evidence.