What the horizon does
A ship leaves harbour and vanishes bottom-first. A shoreline you can see from one kilometre looks the same from fifty. This file records what the horizon is seen to do, and what geodesy calculates it should do.
Stand on a sea wall and watch a boat head out. It does not dwindle to a dot. It goes in a fixed order: first the hull, then the waterline, then the deck line, then the cabins, then the top of the mast, and last of all a sliver of rigging. Bring binoculars up and the process reverses, but not in the way a shrinking object should. The ship does not simply grow back into view from the middle out. It returns from the top down, the upper works appearing first over a line of water that is plain to see and plainly in the way.
Then notice the second half of the same picture, which people who have watched the first half for years often have not. The horizon is level with your eye. Climb a cliff, a mast or a fuselage and it does not fall away below you; it rises with you and stays level with your gaze, while the ground and the water beneath it spread outward. From a low pier the line sits a few kilometres off. From an airliner it is a wide, faintly bowed band a long way out. Whatever the horizon is, it behaves like a limit set by the eye rather than a wall standing at a fixed distance.
Distant objects are cut off from the bottom upward and restored from the top downward when magnified, over a visible intervening stretch of water or ground. The horizon remains at eye level as the observer's height changes, and a level line of sight carried across several kilometres of still water arrives at the far end without the fall the globe model calculates. Neither statement is exotic; both can be watched in an afternoon.
Why the order of disappearance matters
Distance alone does not explain the sequence. Anything merely too far to resolve shrinks toward the middle of its own outline and fades, top and bottom going together. The hull-first disappearance is a different event: something removes the bottom of the object while leaving the top in place, along a straight, sharp, steadily advancing line that coincides with the horizon.
Two explanations are on offer. One says a surface has risen between the boat and the eye: the water bends away and its near edge stands as an obstacle. The other says light from the lower part of the boat never reaches the eye in a straight line, because the air along the way is not uniform. Both accounts agree on the appearance and disagree on the cause.
The Bedford Level and the difficulty of proof
The classic attempt to settle it was made on the Old Bedford River, a straight artificial channel driven across the Cambridgeshire fens. A long, uninterrupted, still reach of water is about the best natural test-bed anyone has for a level sightline: no hills to confuse the picture, no surf, and a length of several miles in a single straight run.
Samuel Birley Rowbotham reported his tests there in Zetetic Astronomy: Earth Not a Globe, first issued as a pamphlet in the 1860s and expanded later. His claim was that a mark set a few feet above the water — a boat's flag, or later a disc on a pole — stayed visible through a telescope along the whole length of the reach, when a water surface bulging away between the two points should have hidden it behind the intervening curvature. Rowbotham treated this as decisive, and it is still the experiment most often cited by the modern revival.
It was contested at the time on the ground that matters most: refraction. In 1870 the naturalist Alfred Russel Wallace took a wager on the same stretch of water, using a telescope with cross-hairs set at a fixed level and a marker on a bridge parapet, and argued that the alignment came out as a flat surface predicts. The dispute then turned on the state of the air, because warm air lying over cold water bends light and can lift a distant object several minutes of arc above where geometry puts it.
What the globe model predicts
The mainstream figure is straightforward. On a sphere of radius R, a level line departing from the surface falls away at the rate d² / 2R. At the Earth's mean radius of about 6,371 km that is a little under eight centimetres in the first kilometre — imperceptible — and then it compounds, because the term is squared.
| Distance from the observer | Predicted fall of a level line | What that looks like |
|---|---|---|
| 1 km | 0.08 m | Nothing the eye can detect |
| 5 km | 2.0 m | The height of a tall person, wholly hidden |
| 10 km | 7.8 m | About two storeys of a house |
| 20 km | 31 m | A ten-storey building |
| 50 km | 196 m | Two thirds of the height of a 60-storey tower |
| 100 km | 785 m | More than twice the height of the tallest building on earth |
| 200 km | 3,140 m | A mountain range, if the far end were at sea level |
The figures assume a straight level line and the mean radius. They are the arithmetic the globe model commits to, and the numbers any flat-earth sightline argument has to beat. Below about five kilometres the two models are hard to separate by eye, which is why the interesting tests are all long ones.
Geodesy accepts that the horizon is a real limit set by the curvature of the surface, and that its distance and its angle below the horizontal are predictable from the observer's height. The angle of dip is θ = arccos(R / (R + h)), or close to √(2h / R) radians: about 2.7 minutes of arc from a 2 m eye height and about 6.1 minutes from 10 m. Nautical tables, correcting for refraction, give the visible dip as roughly 1.76 minutes of arc multiplied by the square root of the height in metres — about 5.6 minutes at 10 m, so a little less than the geometric figure because the atmosphere raises the apparent horizon slightly.
The mainstream account of hull-first disappearance is that the surface is genuinely convex and the near water cuts the line of sight. On magnified sightings that seem to restore a hidden hull, it points to refraction: light travelling along kilometres of water passes through air whose temperature varies with height above the surface, so the ray curves. Under a temperature inversion, when a layer of warm air sits over cold water, the curvature is magnified and a distant object can be lifted far above its geometric position — the effect that produces looming and superior mirages at sea. Surveyors and geodesists have measured this effect rather than assumed it: the standard combined curvature-and-refraction correction used in spirit levelling is about 0.0675 metres per square kilometre of distance, which is the geometric 7.85 cm per km² reduced by roughly fourteen per cent. Terrestrial refraction coefficients are tabulated, typically near 0.13 under ordinary conditions along the ground and higher in strong inversions.
The evidence that cuts the other way
Honesty requires the strongest point against the flat reading to be stated as clearly as the rest. Navigators measure the dip. A sextant observation of the horizon from a ship's bridge gives an angle below the horizontal that grows with height in close agreement with the standard formula, and the refraction correction built into nautical tables came from that body of measurement rather than from theory. If the horizon were simply the optical limit of a level plane, a sextant pointed at it from the mast and from the deck would report the same angle.
Long-lens photographs taken across large lakes are the other contested ground. The skyline of Toronto seen across Lake Ontario from the far shore, roughly fifty kilometres away, is the usual example: the tower there rises a little over five hundred and fifty metres to the tip of its antenna, so parts of the city stay geometrically visible even on a curved surface, and the argument is over which floors should be hidden and whether a stretch of water in the middle of the frame is refraction or clear sky. Similar disputes surround the Chicago skyline seen across Lake Michigan. We record these as live disagreements, not settled proofs.
Where this leaves the argument
The flat reading is this: the horizon is the edge of the level surface itself, and the bottom-first disappearance of a ship is the near water standing in the way of the far. It stays at eye level because a plane viewed from any height meets the eye along that same line, and it never falls away because there is nothing to fall away down. The argument then lives or dies on the long sightline tests.
The globe reading is that the dip is measured, tabulated and used in navigation, and that the anomalies are atmospheric and quantifiable. A reader who wants to weigh this should go to Try it yourself, where the level and laser tests over still water are set out with the equipment each one needs, and then on to water finding its level and the two skies.
This page presents the flat-earth interpretation as its subject matter. It is a belief-based reading of the observations, not a scientific finding — see About for how we frame claims and for the mainstream sources.