Moon illusion
A Socratic walk-through of the Moon illusion — reasoned out one step at a time, not lectured.
The question we started with
THE QUESTION #Why does the Moon look enormous near the horizon and small overhead, when a photograph shows the same size in both?
A full Moon rising behind rooftops looks enormous, and everyone has felt it. Four hours later, high overhead, the same Moon looks like a coin. Photograph both with the same lens and the two discs measure the same.
Most explanations of this begin confidently. That confidence is the interesting part, because after two thousand years of argument the honest answer is that we do not know which mechanism is responsible — and the most useful thing this walk-through can do is show where each candidate fails.
Reasoning it through
REASONING #Nail down the measurement first, because everything downstream depends on it. Hold a small coin at arm's length against the rising Moon, then against the high Moon. It covers the same amount both times. The angular diameter is about half a degree, and it does not change with altitude.
Now kill the folk account — that the atmosphere magnifies it — with arithmetic, twice over, and in both cases the sums point the wrong way.
First, refraction. Air bends light more strongly the closer to the horizon you look, and it lifts the Moon's lower limb more than its upper. The disc is therefore squashed vertically near the horizon; at the true horizon the flattening is a substantial fraction of the disc's own half-degree. Refraction makes the horizon Moon smaller in one dimension, the opposite of the illusion, and it is plainly visible in photographs of moonrise.
Second, geometry. When the Moon is overhead you stand one Earth radius nearer to it than an observer at the planet's centre; when it is on the horizon you do not. Earth's radius is 6,371 km against a mean lunar distance of 384,400 km, and 6,371 / 384,400 is 0.0166. So the horizon Moon is about 1.7 per cent further away, and its angular diameter 1.7 per cent smaller. Real, measurable, and against the illusion.
So nothing about the light changed. Whatever changed is in the observer — and that is where agreement stops.
The oldest account, going back to Ptolemy, is apparent distance. The sky is perceived not as a hemisphere but as a flattened dome, so a point on the horizon is registered as further away than a point overhead. If two things subtend the same angle and one is taken to be further, the further one must be bigger — an inference the visual system makes constantly and correctly for objects on the ground. It predicts something true: the illusion is strongest over a terrain-filled horizon. Its embarrassment is blunt. Ask people which Moon looks nearer and they overwhelmingly say the horizon one. The account must then reply that the registered distance is unconscious and contradicts the conscious report — which may be right, but makes the central variable something no one can measure directly.
The rival account is angular size contrast. The horizon Moon sits among rooftops, trees and hills that subtend small angles; the overhead Moon floats in an empty field, and a disc looks large among small things, as in the Ebbinghaus figure. This needs nothing unconscious. Its embarrassment is equally blunt: the illusion is reported over a bare sea horizon and by pilots at altitude, with no small comparison objects present at all.
A third observation fits neither. Bend over and view the rising Moon between your legs, and for many people the illusion shrinks or vanishes — pointing at head and eye posture, or at how the visual system encodes "up".
The load-bearing claim here is the narrowest one: the illusion is in the observer, not the light. It is falsifiable in an afternoon. Take a calibrated angular measurement — a sextant, a fixed-focal-length camera, a coin at a marked distance — of the rising and the high Moon. Find a genuinely larger angular diameter at the horizon and everything above collapses. In two millennia nobody has.
The analogy
THE ANALOGY #Your eye never receives a size, only an angle. To turn one into the other it must supply a distance, the way a price in a foreign currency means nothing until you apply an exchange rate. The angle is the number on the tag; the assumed distance is the rate; the size you experience is the converted price. Nothing was wrong with the tag.
an exchange rate is a published number you could look up and correct, whereas the visual system's rate is not available to introspection at all — which is precisely why an observer can report the horizon Moon as both larger and nearer without ever noticing that the two reports contradict each other.
Clarifying the model
THE MODEL #Three refinements hold the argument together.
The magnitude is not a fixed quantity. The illusion is usually reported at something like one and a half to two times (recalled as the typical range), but the figure depends on how you ask — matching an adjustable disc, drawing, verbal estimate and comparison through an aperture all give different answers. That instability is itself evidence: what is measured is a judgement, not a stable percept with one value. Nor is the effect peculiar to the Moon; the Sun shows it, constellations show it, and it can be produced with an artificial disc, so any account leaning on something lunar is answering the wrong question.
And I should falsify my own headline. Many researchers do not think this is open: Kaufman and Rock's experiments of the early 1960s, and later stereoscopic measurements, are widely read as establishing the apparent-distance account, and that evidence is real. What covers the gap is the scope of the claim. "Unexplained" here does not mean "no evidence exists"; it means no single account yet predicts the whole pattern — sea horizon, altitude reports, the between-the-legs reversal and the contradictory distance judgements together. The likeliest resolution is that several cues contribute with weightings that shift with the scene, which satisfies neither camp.
The nearby piece Linear perspective is worth reading against this one. It shows depth cues doing their job correctly on a flat surface; this shows the same machinery running on a scene where its assumptions fail, returning an answer the geometry does not support. Same apparatus, opposite verdict — and its caution that perspective is only one cue among several is exactly why no single-cue account of the Moon works either.
A picture of it
THE PICTURE #How to readThis borrows the class notation for something it was not built for — each box is a rival account rather than a type of object, so read the notation loosely. The top box is what every candidate must accommodate: the measurements nobody disputes. In each account below it, a plus line is what that account gets right and a minus line is the observation that defeats it. The dashed arrows all point the same way, at the evidence, and the point of the picture is that no box has only plus lines.
What became clearer
WHAT CLEARED #The Moon does not change size, and neither does the light: refraction and geometry both make the horizon Moon marginally smaller, so the whole effect is manufactured downstream of the eye. What is unsettled is which piece of that machinery does it. The apparent-distance account is contradicted by what observers say about distance; the size-contrast account by the illusion surviving over an empty sea. Knowing precisely which observation each account cannot accommodate is a firmer position than a confident answer would be.
Where to go next
ONWARD #- Why the Ebbinghaus and Ponzo figures produce size errors on a flat page, and whether the same mechanism could be at work in the sky.
- How anyone tests a perceived distance to an object nobody can reach.
Key terms
TERMS #| Term | What it means |
|---|---|
| Angular diameter | how wide something appears measured as an angle; the Moon's is about half a degree, whatever its altitude. |
| Size-distance invariance | the rule that perceived size follows from angular size together with perceived distance. |
| Size-distance paradox | the finding that observers call the horizon Moon both larger and nearer, which the apparent-distance account cannot take at face value. |
Every term the collection defines is gathered in the glossary.