Photographs settle the argument first. Shoot the full moon the moment it clears the rooftops, then shoot it again hours later with the same zoom setting, and the two images match: the horizon moon is the same width as the overhead moon, pixel for pixel. NASA's own explainer on the moon illusion says exactly this, and adds the conclusion the comparison forces — the effect is an illusion rooted in the way brains process visual information, not something the atmosphere is doing to the light. If anything, the air slightly squashes the horizon moon rather than magnifying it, as a NASA scientist put it in a recent interview, because refraction compresses the vertical angle near the horizon. What your eye reports and what the lens records diverge, and the divergence is the story. Photographing our nearest neighbor has other traps — why moon photos come out as white blobs is a separate exposure problem — but for measuring the moon's apparent size, a camera is a trustworthy referee.
A brain that rescales, not a sky that inflates
The leading explanations share one idea: your visual system does not passively record the moon's angular size; it interprets it. The horizon comes loaded with distance cues — buildings, trees, terrain fading into haze — and against those cues the brain treats the horizon moon as a far object that still looms large, and scales its apparent size accordingly. Look straight up and the cues vanish: the sky behaves like a surface with no depth information, and the same angular patch reads smaller. Researchers have formalized versions of this account for decades; the classic papers by Lloyd Kaufman and Irvin Rock in the 1960s, published in Science, argued for an apparent-distance mechanism and launched the modern experimental literature. The intuition to carry away is that size perception is a computed quantity, stitched together from angle and estimated distance — and at the horizon, the stitching gets extra material to work with.
Here is the mechanism in its plainest form. The dome of sky overhead does not feel like a hemisphere; it feels flattened, with the zenith seeming nearer than the horizon. If the moon spans the same angle in both places but the brain has tagged the horizon as farther away, then the only way to reconcile same-angle-with-greater-distance is to enlarge the perceived object — the same trick that makes a distant person who subtends a small angle still look person-sized. Distance cues at the horizon supply the tag; the empty zenith supplies nothing, so the computation runs unconstrained and the moon shrinks toward whatever a bare patch of brightness at an unknown distance deserves. The account predicts the odd details neatly, among them that the illusion weakens when the terrain is masked and strengthens the longer the vista — which is precisely what the between-the-legs and paper-tube improvisations exploit.
One corollary trips up photographers hoping to keep the bigness. The illusion lives in the interpretation stage, not the light, so no camera setting preserves it: a wide shot of the moon over a skyline renders an honest small disc, and the dramatic horizon photographs you admire are usually long lenses compressing a real scene, not a captured illusion. The angle itself is modest either way — NASA's fingernail comparison at arm's length is roughly the moon's true width on the sky — and once a lens records that angle faithfully, the picture tells the geometric truth your evening brain would not.
Two thousand years without a verdict
Here is the honest part. Aristotle wrote about the effect, and scientists are still designing experiments to pin down exactly which mechanism carries how much weight. NASA is direct about the state of play: there is still no satisfying scientific explanation for exactly why we see it, and no consensus among the competing accounts. Candidate explanations trade on the flattened-dome shape the sky appears to have, on the contrast between the moon and the terrain around it, on eye and head posture, or on combinations of all three. Treat any single tidy mechanism you meet online with suspicion; the phenomenon is real and robust, and the accounting of it is genuinely unfinished.
The ten-second test
NASA suggests a small experiment that costs nothing. Hold up your outstretched index finger next to the horizon moon: your fingernail and the moon should be roughly the same size. Look at the moon through a paper tube, or — less elegantly — bend over and view the sky backward between your legs, and the moon will seem nowhere near as big as it had seemed, because the trick strips away the surrounding distance cues your brain was using. The photo comparison is the ironclad version: same camera, same zoom, horizon versus midnight, and the prints agree with each other. Notice what all three tests have in common — they interrupt the interpretation rather than the light. Judging an attribute with the reference removed is also a fair description of what memory does to a color; the Color Memory game shows how quickly an exact percept decays into a summary once the original is out of sight.
The moon illusion is worth keeping precisely because it is unfixable by knowing about it. The horizon moon will look huge next full moon, and you, informed reader, will know the pixels say otherwise, and it will look huge anyway. Perception runs its estimates whether or not you have read the lab reports. The sky is simply the largest place where that quiet fact about yourself is on public display — twice, once at sunset, once at sunrise, every day the moon is up to show it.