№ 175 · astronomy

The Moon on the horizon only looks huge

A full Moon rising behind houses or hills can look enormous, far bigger than the same Moon a few hours later, high in the sky. Nothing about the Moon has changed. Measure it and the rising Moon is, if anything, a little smaller, and the air squashes it slightly from top to bottom. The giant Moon is made in the head of the person looking. How the brain does it has been argued over for centuries, and it is still not settled.

Predict first

You photograph the full Moon just as it rises, then again when it is high, with the same camera and lens. Measured on the photos, the rising Moon's disc is:

Check your answer

A little smaller, and squashed. When the Moon is low you look at it across the Earth, so it is farther away than when it is overhead, and its disc is narrower by about 1.7%. The air does not magnify it. It bends the light from the lower edge more than from the upper edge, which lifts the bottom of the disc and flattens it. The big rising Moon is not on the photo.

Measure it, and the horizon Moon shrinks

The Moon's distance is measured from the centre of the Earth, but you stand on the surface, about 6,371 km from the centre. When the Moon is straight overhead, you are that much closer to it. When it is on the horizon, you look at it sideways across the Earth and gain almost none of that. NASA's Moon fact sheet, kept by David Williams, gives the Moon's average distance as 384,400 km and its radius as 1,737.4 km. With those numbers the Moon overhead is about 378,000 km away and its disc is 31.6 arcminutes across, a little more than half a degree. On the horizon it is about 384,400 km away and 31.1 arcminutes across. The horizon Moon is about 1.7% farther away, and its disc is narrower by the same share.

The air changes its shape, too. Air bends light coming in at a shallow angle, so everything near the horizon looks a little higher than it really is. Teresa Wilson's 2018 dissertation on sunrise and sunset timing explains that the standard value for this lift at the horizon is 34 arcminutes, more than the width of the Moon itself. The lift falls off quickly with height, so the lower edge of a rising Moon is lifted more than its upper edge, and the disc looks flattened. Wilson describes exactly this for the rising and setting Sun. With a simple formula for the lift that G. G. Bennett published in 1982, which Wilson's dissertation sets out, a Moon whose lower edge just touches a flat horizon is lifted into an oval only about 84% as tall as it is wide.

So a camera shows the rising Moon slightly narrower and noticeably flatter than the high Moon. The air is a squasher, not a magnifier.

Interactive What a camera records as the Moon climbs. Set how high the Moon's lower edge is above a flat horizon and how far away the Moon is in its orbit. The figure works out your distance to the Moon from where you stand on Earth's surface, the width of its disc, and how much the air bends light from its lower edge more than from its upper edge. The right panel draws the disc as a long telephoto lens would record it, with a dashed circle for the same Moon straight overhead. Press Watch it rise to see the disc change as it climbs. Every control works with Tab, Enter and the arrow keys.

lower edge above horizon0.0°

31.08′ wide, 26.0′ tall

flat horizon marker not to scale telephoto view, 4 px per arcminute dashed: same Moon overhead 30′ (half a degree) how big each change is, in diameter this disc vs overhead 1.7% near vs far in orbit 11.6% what eyes report 50 to 100% bigger 0 to 100%
Lower edge on the horizon.

How big the illusion is

The size people report is far larger than any of these measured changes. In a 1994 paper Cornelis Plug and Helen Ross pulled together the observations of the real Moon and found that horizon objects look about 1.5 to 2.0 times as wide as the same objects high in the sky. Lloyd Kaufman and James Kaufman write that the horizon Moon may look nearly twice the diameter of the elevated Moon. Plug and Ross also note that the effect varies, especially with how much detail the horizon scene holds.

For comparison, the Moon's orbit is not a circle. At its average nearest point, 363,300 km by the fact sheet, it is about 12% wider than at its average farthest, 405,500 km. That real change in size is much larger than the horizon-to-overhead change. The illusion is several times bigger still.

Size depends on how far away you think it is

Your eye records only how wide an angle an object covers. A car 10 metres away and a car 100 metres away look the same size to you, though the far one covers a tenth of the angle, because your brain takes distance into account. Kaufman and Kaufman call this size constancy, and it only works if the brain has some sense of the distance.

The older of the two main explanations today uses exactly this. Karl Hipius and David Kornreich trace it back to the scholar Ibn al-Haytham in the 11th century. Seen across a stretch of land, with fields, trees and houses getting smaller toward the horizon, the Moon is treated as something very far away. Seen high in an empty sky, there is nothing to say how far away it is, and it is treated as nearer. An object that covers the same angle but is taken to be farther away must be bigger, so the horizon Moon looks larger. In 1962 Kaufman and Irvin Rock found that the terrain matters most: what counts is seeing ground in the direction of one Moon and none in the direction of the other, while the angle at which you raise your eyes has only a weak effect.

There is a catch, which Kaufman and Kaufman call the size–distance paradox. Ask people which Moon looks closer, and most say the big horizon Moon. That is the opposite of what the theory seems to need.

An experiment with artificial moons

In 2000 Lloyd and James Kaufman tested the two kinds of explanation against each other. On a hilltop on Long Island, people looked through a half-silvered mirror at two artificial moons projected onto the real sky, each about 0.6 degrees wide. The moons could be set just above a distant horizon of hazy hills, or 45 degrees up in an empty sky. Each eye saw a slightly different image, so a viewer could move one moon nearer in depth, the way two eyes see depth in the real world. Each person's task was to set the movable moon halfway between themselves and the other one.

If the horizon Moon is treated as farther away, the halfway point should be farther away too. That is what happened. On average people put the halfway point for the horizon moons about 36 metres away and for the elevated moons about 8.6 metres away, roughly 4.2 times farther. All five subjects put the horizon halfway point farther, by factors from 1.58 to 6.48. Kaufman and Kaufman concluded that the visual system treats the horizon Moon as though it were much farther away, and that is why it looks bigger.

They also found that the simple rule of size growing in step with distance, known as Emmert's law, did not hold over these long distances. Their explanation of the paradox is that distance cues set the size the Moon is seen at, and then people use that size when asked how far away it is. A bigger-looking Moon gets judged as closer, even though the visual system had treated it as farther.

The other camp, and why it is still open

Not everyone accepts this. Kaufman and Kaufman set out the rival view, which runs the other way: the high Moon first looks smaller, for some other reason, and that smaller size then makes it seem farther away. Proposed reasons include the eyes focusing or turning inward slightly when they look up into an empty sky, and the Moon looking small next to the wide empty sky around it. Plug and Ross argued that the illusion is a change in how big an angle the Moon seems to cover, made up of several effects added together: the size of things around it, how far the eyes turn inward, the position of the eyes and head, the haze of distance, and colour. Joseph Antonides and Toshiro Kubota proposed in 2013 that the brain treats the sky as a surface and the Moon as an object in front of it, and enlarges the Moon to settle the conflict with what two eyes report.

A 2023 study by Hipius and Kornreich tested the shape people give the sky. Over open water in New York State, 30 people pointed to the spot in the sky they felt was directly above landmarks from 3 metres to 12.6 km away. The answers did not trace out the flattened dome that the old explanation assumes. The sky people pointed to never came down to meet the horizon, which left a band of roughly the lowest 20 degrees of sky with no matching point on the ground. That band is where the illusion is strongest. Hipius and Kornreich also collect two odd facts from earlier work: the illusion disappears when the Moon is viewed upside down, and it gets weaker when you look with only one eye. They describe the illusion as still not fully explained.

You can test the first part yourself. Hold a small coin or a pencil eraser at arm's length next to the rising Moon, and again when it is high. The Moon covers the same width both times. Or take two photos with the same lens and compare them. The giant Moon disappears; only the feeling stays.

In short

The rising Moon is not bigger. It is about 1.7% farther away than the Moon overhead, so its disc is a little narrower, and the air squashes it from top to bottom. People still see it as 1.5 to 2 times wider. The older of the two main explanations is that the brain treats a Moon seen over distant land as farther away than a Moon in an empty sky, and so sees it as bigger. Kaufman and Kaufman's artificial-moon experiment supports that. Other researchers argue for different or added causes, and the question is still open.

Where this comes from

  1. The Celestial Vault is Not a Dome: Implications for the Moon Illusion linked only, not reproduced
    Karl Hipius and David A. Kornreich · arXiv:2306.13656 · 2023
    arxiv.org/abs/2306.13656
  2. Binocular disparity as an explanation for the moon illusion linked only, not reproduced
    Joseph Antonides and Toshiro Kubota · arXiv:1301.2715 · 2013
    arxiv.org/abs/1301.2715
  3. Explaining the moon illusion (PNAS 97, 500-505) linked only, not reproduced
    Lloyd Kaufman & James H. Kaufman · 2000
    pmc.ncbi.nlm.nih.gov/articles/PMC26692/
  4. The natural moon illusion: a multifactor angular account (Perception 23, 321-333) linked only, not reproduced
    Cornelis Plug & Helen E. Ross · 1994
    doi.org/10.1068/p230321
  5. The moon illusion, I (Science 136, 953-961) linked only, not reproduced
    Lloyd Kaufman & Irvin Rock · 1962
    doi.org/10.1126/science.136.3520.953
  6. Moon Fact Sheet linked only, not reproduced
    David R. Williams, NASA Goddard Space Flight Center · 2024
    nssdc.gsfc.nasa.gov/planetary/factsheet/moonfact.html
  7. Evaluating the Effectiveness of Current Atmospheric Refraction Models in Predicting Sunrise and Sunset Times (PhD dissertation, Michigan Technological University) linked only, not reproduced
    Teresa A. Wilson · 2018
    doi.org/10.37099/mtu.dc.etdr/697
  8. The calculation of astronomical refraction in marine navigation (Journal of Navigation 35, 255-259) linked only, not reproduced
    G. G. Bennett · 1982
    doi.org/10.1017/S0373463300022037