Debunks Reality Denial Bedford Level

Testing the Bedford Level Experiment

The Bedford Level experiment occupies a unique place in the history of flat-Earth belief. Modern arguments about lasers over lakes, distant skylines, canals, and “water finding its level” are descendants of a test Samuel Birley Rowbotham began promoting in the nineteenth century on a remarkably straight stretch of the Old Bedford River in the English Fens. Rowbotham reported that a boat remained visible for six miles even though, on a spherical Earth, it should have been substantially hidden. From this he concluded that the canal surface was flat and that the accepted curvature of Earth did not exist.

That observation became one of the foundations of Zetetic Astronomy: Earth Not a Globe. It also produced one of the most famous public scientific wagers of the Victorian era. In 1870, flat-Earth supporter John Hampden offered £500 to anyone who could demonstrate the convexity of a body of water. Alfred Russel Wallace—the naturalist independently famous for formulating natural selection and a man with practical surveying experience—accepted. Wallace changed the experimental design. Instead of placing his telescope only inches above the water and asking whether a distant object remained visible, he elevated the sight line and introduced a midpoint marker. The midpoint appeared above the straight line joining the two endpoints by approximately the amount expected from Earth curvature, reduced somewhat by refraction. A contemporary report in Nature described the result.[1]

The historical story is colorful, but the physics is more important. The Bedford Level controversy is not a case in which one side “believed the globe” and the other side “trusted observation.” Both sides observed the same canal. The crucial difference was experimental design. Rowbotham chose a line of sight extremely close to water, where terrestrial refraction is strongest, and treated visibility itself as proof of flatness. Wallace used three elevated points whose relative geometry could reveal convexity directly.

The core finding: Rowbotham’s original observation does not need to be fabricated for his conclusion to be wrong. A boat can genuinely remain visible farther than unrefracted geometry predicts when the optical path is bent by a strong near-water refractive gradient. Rowbotham’s setup was extraordinarily sensitive to exactly that effect. Wallace’s redesigned experiment reduced the problem and changed the observable from “can I still see it?” to “are three equal-height points actually collinear?” They were not.

Contents

1. Why the Old Bedford River Was Chosen

The Old Bedford River is part of the drainage works of the Fens in eastern England. One stretch near Welney runs unusually straight for about six miles. That geometry made it attractive to nineteenth-century experimenters because a long, calm water surface seemed to offer a natural reference against which curvature could be tested.

For a visual experiment, the location has obvious advantages. There are no mountain ridges blocking the sight line. The route is nearly straight. Water offers a surface whose mean elevation is governed by gravity. Six miles is long enough for the expected curvature to become measurable in feet rather than fractions of an inch.

But the location also has one enormous disadvantage: the optical path is low over water.

A long sight line grazing a canal passes through air whose temperature and density can change rapidly with height. Water has a large thermal inertia compared with air. During some conditions the air immediately above the surface can be cooler and denser than air only a few feet higher. This creates an atmospheric refractive-index gradient capable of bending nearly horizontal light downward.

The very feature that made Bedford Level attractive as a geometric test therefore made it dangerous as an uncontrolled optical test.

This dual character explains why opposite conclusions could be reported from the same canal. The geometry of the water did not change. The experimental sensitivity to the atmosphere did.

2. Samuel Rowbotham and the 1838 Experiment

Samuel Birley Rowbotham, later writing under the pseudonym “Parallax,” made the Bedford Level experiment central to his flat-Earth system. In Zetetic Astronomy, he described an observation beginning at a point he called Welche’s Dam and extending toward Welney Bridge, six statute miles away.[2]

His arrangement was simple:

Rowbotham reported that the flag and boat remained visible through the journey. In the later edition of his book available today, he gives the flag top as about five feet above the water and the observer only eight inches above it.[2]

He then compared this observation with the expected curvature of a globe and concluded that the boat should have disappeared. Since it did not, he argued that the water surface could not be convex.

It is important to state the issue accurately. There is no need to accuse Rowbotham of inventing the observation. The physics allows an observation of that kind under sufficiently strong terrestrial refraction. The error lies in treating one visual outcome as uniquely diagnostic of surface geometry.

3. What Rowbotham Reported

Rowbotham’s own reasoning is explicit. With his line of sight only eight inches above the water, he argued that the geometric horizon should occur roughly a mile away. Beyond that horizon, the curved surface should increasingly obscure the boat.

He calculated that by the time the boat reached the far bridge, the top of its flag should be more than eleven feet below the visible horizon if Earth had the accepted curvature. Yet, according to his report, the boat and flag remained distinctly visible.[2]

From this he drew a strong conclusion: the water must have remained effectively parallel with his line of sight for the full six miles.

This conclusion would be warranted only if the optical ray from the boat to the telescope were known to be a straight line.

That assumption is exactly what fails near the water surface.

An optical observation tells us the path along which light reached the instrument. It does not tell us that the path was straight unless the refractive medium is sufficiently uniform. If the ray curves downward, an object lying below the unrefracted geometric horizon can remain visible.

Therefore Rowbotham’s experiment contained two unknowns—surface geometry and ray geometry—but attempted to solve for the first while silently fixing the second.

4. What Globe Geometry Predicts

For a spherical Earth of radius R, the approximate distance to the geometric horizon from observer height h is:

d ≈ √(2Rh)

For an eye only eight inches, or about 0.67 feet, above the water, the unrefracted geometric horizon is only about one mile away.

A five-foot-high target has a horizon of only a few miles from its own top. Add the observer and target horizon distances and the unrefracted mutual visibility distance remains substantially shorter than six miles.

This means Rowbotham correctly identified a real geometric issue: in a straight-ray spherical model, a five-foot flag should not remain completely visible to an observer only eight inches above the water at six miles.

This point should be granted, because otherwise the actual flaw in the experiment is obscured.

The correct scientific response is not “Rowbotham did the curvature arithmetic wrong, therefore nothing happened.” The deeper response is: his observation, if accurately reported, is evidence that the optical path was not the unrefracted straight ray he assumed.

And that is physically plausible in exactly the environment he chose.

5. The Eight-Inch Observer Is the Critical Detail

The experiment would have been much less vulnerable to refraction if the line of sight had been elevated several meters above the canal. Rowbotham instead positioned his observation only inches above the surface.

That places the ray inside the atmospheric boundary layer immediately over the water.

The problem is not merely that “refraction exists.” The problem is that vertical temperature gradients are strongest and most variable near surfaces. Water can remain cool while afternoon air above it warms. A cool dense layer then hugs the canal beneath warmer air. The refractive index is slightly greater in the cooler denser layer. A nearly horizontal ray can curve downward toward it.

Rowbotham’s published description says the test occurred on a summer afternoon with bright sunshine.[2] That is not evidence by itself that a specific refractive profile existed, but it highlights exactly why contemporaneous temperature measurements would have been essential.

At eight inches above water, the line of sight was not simply measuring curvature. It was measuring curvature plus the most unstable part of terrestrial refraction.

Modern geodetic leveling practice deliberately tries to avoid this regime. NOAA’s National Geodetic Survey discusses atmospheric refraction as a significant source of leveling error and emphasizes sight-height and meteorological control.[3]

6. The Experiment Was Built Inside a Refractive Boundary Layer

Terrestrial refraction occurs because the refractive index of air is not perfectly uniform.

Air density depends on temperature, pressure, humidity, and composition. The refractive index of ordinary air is only slightly greater than one, but over miles even tiny vertical changes can bend a nearly horizontal ray appreciably.

If refractive index decreases with height in the appropriate way, light curves downward. This makes distant objects appear higher than their purely geometric positions. Under stronger gradients, objects that are geometrically below the horizon may become visible.

Maritime navigation has recognized this for generations. NOAA’s U.S. Coast Pilot describes looming and superior mirage conditions in which objects can appear elevated and objects normally below the horizon can be brought into view.[4]

Rowbotham’s observed boat is exactly the class of object such refraction can affect.

The canal therefore does not present a binary choice:

boat visible → flat Earth
boat hidden → globe

The actual possibilities include:

boat visible → flat geometry OR curved geometry + downward-bending ray

Once more than one physical model predicts the observation, the experiment must be redesigned to distinguish them.

7. Rowbotham’s Refraction Argument Was Physically Wrong

Rowbotham knew that critics invoked refraction, and he attempted to answer them.

In a later section of Zetetic Astronomy, he argued that refraction occurs only when light passes from one medium into another—for example from air into water—and that if both observer and object are surrounded by the same medium, such as air, there can be no refraction.[5]

This is incorrect.

Refraction does not require a sharp interface between substances with two discrete refractive indices. Light bends continuously through a medium whose refractive index changes continuously with position.

The atmosphere provides exactly such a medium. A vertical density gradient can cause a curved ray even though the light remains entirely within air.

This is standard atmospheric optics. The familiar inferior mirage over hot pavement occurs because a refractive-index gradient exists within air itself. Superior mirages and looming occur under opposite thermal structures. Radar propagation is likewise refracted continuously through atmospheric gradients; NOAA describes subrefraction, superrefraction, and ducting caused by changes in atmospheric refractivity.[6]

Rowbotham therefore rejected the correct explanation on the basis of an incorrect theory of refraction.

8. Light Can Refract Continuously Within the Same Atmosphere

The underlying optics can be understood without advanced mathematics.

Imagine the atmosphere divided into many extremely thin horizontal layers. Each layer has a refractive index infinitesimally different from the one above it. At each boundary the ray bends by an infinitesimal amount. Make the layers thinner and thinner and the path becomes a smooth curve.

There is no physical requirement that the layers be made of different chemical substances. Temperature alone can alter density enough to change refractive index.

A continuously graded optical fiber uses the same basic idea deliberately: refractive index varies continuously across the material and bends rays smoothly.

Earth’s atmosphere is a natural graded-index medium.

The sign of the bending depends on the gradient. Hot ground commonly produces air that is warmer and less dense immediately above the surface, generating the classic “water on the road” inferior mirage. Cold water beneath warm air can produce the opposite structure and bend rays downward, producing superior mirages and looming.

Once this is understood, Rowbotham’s premise that “same medium means no refraction” collapses.

9. Looming Can Reveal Geometrically Hidden Objects

Looming is not merely a visual distortion that makes an already visible object look funny.

Under suitable atmospheric conditions, downward-bending rays can reach an observer from an object that lies geometrically below the unrefracted horizon. The observer sees the arriving ray and mentally extrapolates it backward in a straight line, causing the image to appear higher than the object’s true geometric position.

NOAA navigation material explicitly warns that objects below the normal horizon may become visible during looming conditions.[4]

This directly answers one of the most common defenses of the Bedford experiment:

“Refraction might distort the boat, but it cannot make something hidden behind curvature visible.”

It can. That is one of the defining effects of strong terrestrial refraction.

And because Rowbotham placed his telescope exceptionally close to the water, he maximized the possibility that the ray would stay inside the refractive layer for much of the six-mile path.

10. Visibility Alone Cannot Distinguish Flatness from Refraction

The methodological problem can be expressed formally.

Let G represent surface geometry and A represent atmospheric refractive structure. The observed apparent elevation E of a target depends on both:

E = f(G, A)

Rowbotham measured E but assumed A = 0 or negligible. He then solved for G.

If A is unknown, the problem is underdetermined. Different combinations of geometry and refraction can give the same apparent target position.

This is why “I could still see the boat” is weaker than it sounds.

One needs another independent constraint. Wallace’s midpoint-marker design supplied such a constraint by comparing three positions instead of merely detecting one distant target.

A modern version could go farther: use multiple marker heights, reciprocal observations, simultaneous temperature profiles, surveyed elevations, and ray tracing.

11. Rowbotham’s Flags and Theodolite Experiments

Rowbotham did not stop with one boat observation. Zetetic Astronomy presents multiple experiments along the Bedford Level, including rows of flags and the use of a theodolite.[7]

His basic idea remained the same: establish objects at nominally equal heights above the water, view them along the canal, and interpret their apparent alignment as evidence that the water surface remained horizontally straight.

These tests are more interesting than caricatures of Rowbotham sometimes acknowledge. He was attempting empirical geometry rather than merely citing scripture or intuition.

But the tests inherited the same conceptual problems:

Rowbotham’s fourth experiment even discussed “forward leveling” along the canal and concluded that repeated level observations demonstrated a globally straight water surface.[8]

But repeated local horizontals on a curved equipotential surface do not all share one global direction. Each setup establishes a new local tangent.

12. The Misunderstanding of “Level”

This issue became central in the later dispute between Wallace and William Carpenter.

In everyday language, “level” often means “flat.” In surveying, the terms are more precise.

A horizontal line at one point is perpendicular to the local vertical, where the local vertical is defined by gravity. It is a straight tangent direction.

A level surface is an equipotential surface. On a massive planet, local vertical directions at different places are not parallel. The resulting level surface is curved.

Therefore a surveyor can repeatedly level an instrument along a route and trace a curved level surface through a sequence of local tangents.

Rowbotham regarded this as circular reasoning: if each setup is re-leveled, the surveyor has simply “invented” curvature by changing the line. But the changing local vertical is itself measurable with plumb lines and gravitational instruments.

NOAA’s National Geodetic Survey defines physical heights and level surfaces in terms of Earth’s gravity field and the geoid.[9]

So “water is level” and “water lies on one infinite Euclidean plane” are not equivalent statements.

13. John Hampden’s £500 Challenge

By 1870 Rowbotham’s ideas had attracted supporters, including John Hampden.

Hampden publicly offered a £500 challenge to anyone who could demonstrate the convexity of a canal, river, lake, or similar body of water. Wallace’s surviving correspondence records his response. On January 15, 1870, Wallace offered to stake an equal amount and proposed showing and measuring the convexity of a canal or lake.[10]

The total stake became £1,000, an enormous amount of money at the time.

Wallace initially suggested another site, including Bala Lake, but the Bedford River was ultimately chosen. Wallace later objected to claims that the canal had been his own preferred site, stating that Hampden had suggested it and he had accepted.[11]

The choice had symbolic importance. Bedford Level was not merely any canal. It was the site Rowbotham had made famous as experimental proof of flatness.

14. Why Alfred Russel Wallace Accepted

Wallace is best known as the naturalist who independently developed the theory of evolution by natural selection at roughly the same time as Charles Darwin. Less relevant to popular memory but highly relevant here, Wallace also had practical surveying experience.

He believed the issue could be settled by a straightforward geometric demonstration. In later correspondence he acknowledged that accepting the wager had been unwise because the aftermath became expensive and personally miserable, but his experimental objective was clear: display the convexity of the water visibly and quantitatively.[12]

Wallace understood the vulnerability of near-water observations to refraction. Instead of looking only inches above the canal, he placed the experiment roughly thirteen feet above the surface.

More importantly, he changed what was measured.

Rowbotham asked: can I see the distant object?

Wallace asked: if three markers are all the same height above the water at the beginning, midpoint, and end of a six-mile stretch, do they actually lie on one straight line?

That is a fundamentally stronger geometric test.

15. Wallace’s 1870 Experimental Design

The 1870 arrangement is described in Wallace’s correspondence, contemporary reports, and later summaries.

The six-mile stretch was bounded by two bridges. The parapet at Welney Bridge was about thirteen feet above the water. At the far bridge, Wallace installed a white sheet with a horizontal black band arranged at a corresponding elevation. At roughly the midpoint, three miles from each end, he erected a pole carrying red discs. The upper disc was placed at the same nominal height above the water as the relevant sighting marks at the endpoints.[13]

A large telescope was positioned at one endpoint.

If the canal surface were an infinite straight plane and all three targets were truly the same height above that plane, the telescope, midpoint disc, and distant mark should lie in one straight line. The midpoint disc should project onto the distant mark.

If the canal followed a convex Earth, the midpoint point at equal height above the water would lie above the straight chord joining the two endpoints.

This prediction is visual and directional. The middle marker should appear displaced relative to the distant endpoint.

16. Why the Midpoint Marker Changes Everything

The three-point design is the conceptual heart of the experiment.

Take three points A, B, and C separated along a curved surface, with B halfway between A and C. Raise each point by the same small height above the local surface.

On a plane, A, B, and C are collinear if the heights are equal.

On a sphere, they are not.

The straight line from A to C forms a chord above or through the curved geometry depending on the exact elevated configuration. The midpoint B follows the surface and therefore lies farther from that chord than it would on a plane.

For a symmetric six-mile baseline, the midpoint displacement is on the order of six feet before refraction is considered.

This is far more informative than simply asking whether C is visible from A. A downward-refracted ray can restore visibility of C, but the relative position of B gives another geometric constraint.

It does not make refraction disappear. It makes the experiment less degenerate.

17. The Expected Midpoint Bulge

For a circular arc of radius R with total arc length approximately L, the midpoint sagitta relative to the chord joining the endpoints is approximately:

s ≈ L² / (8R)

For L = 6 miles and Earth radius approximately 3,959 miles:

s ≈ 6² / (8 × 3959) miles ≈ 0.001137 miles ≈ 6.0 feet

This is the simple unrefracted spherical prediction for an exact six-mile symmetric interval. Historical values in the Wallace accounts vary around roughly five to six feet because the actual geometry, marker heights, distances, and refraction corrections were not the idealized numbers in this calculation.

Nature summarized Wallace’s prediction as the central signal appearing elevated by about five feet above the line joining the outer points.[1]

Wallace later described the observed curvature in his correspondence as approximately five and a half feet at the middle signal in the large-telescope view and argued that the remaining difference from the unrefracted theoretical value was compatible with refraction.[14]

The key result is not whether one historical estimate was 5.0, 5.5, or 5.7 feet. It is the sign and order of magnitude: the middle point did not line up as a planar model required.

18. What Wallace Actually Observed

The observation showed the midpoint marker displaced relative to the far marker in the direction expected from convexity.

The contemporary Nature report described the three signals as being at equal elevation above the water and reported Wallace’s expectation that the center should appear approximately five feet above the line joining the others on a curved Earth.[1]

Wallace’s own published response to Carpenter stated that the three equal-height points did not coincide when viewed from an endpoint. He argued that their relative displacement was close to the amount expected for a circle of Earth’s accepted radius.[14]

This is sometimes misrepresented as Wallace simply tilting the telescope until he “made” the middle marker look high. But the large-telescope test did not require the telescope itself to define a global horizontal. The collinearity question is independent: three physical points either lie on one straight optical line or they do not.

If the middle point visually lies above the line from the telescope to the distant point, the three are not collinear.

A flat-water interpretation then needs some other effect to produce that non-collinearity.

19. Reciprocal Observations Strengthened the Result

Wallace did not rely only on one viewing direction.

His correspondence states that views from both ends of the six-mile interval agreed closely.[14]

This matters because an unnoticed fixed tilt or endpoint setup error may reverse or change character when the observation direction is reversed. Reciprocal measurements are a standard way to diagnose systematic error.

The experiment was not modern metrology, and nineteenth-century documentation does not give the atmospheric profiling we would demand today. But using the geometry in both directions was methodologically stronger than a one-way “I can still see the boat” observation.

Modern surveying uses reciprocal or balanced methods for the same reason: systematics that contaminate one path can often be reduced when the geometry is reversed.

20. The Spirit-Level Telescope Test

In addition to the large telescope, Wallace used a spirit-level telescope to establish the local horizontal direction at an endpoint.

This second test addressed criticism that the ordinary telescope had not itself been leveled.

A local horizontal is tangent to the level surface at the observation point. On a curved Earth, equal-height points farther away should fall progressively below this tangent direction, subject to refraction.

Wallace argued that the observed positions relative to the level telescope’s crosshair were consistent with this behavior.[14]

This distinction became tangled in the subsequent debate because Carpenter used “straight” and “level” in ways Wallace regarded as confused.

The conceptual separation is simple:

21. William Carpenter’s Objection

William Carpenter, a supporter of the flat-Earth position and Hampden’s referee, disputed Wallace’s interpretation.

Wallace’s letters in The Field responded point by point. He accused Carpenter of treating “straight line” and “level line” as interchangeable and of misunderstanding both surveying and the geometry of the three points.[14]

Carpenter emphasized that the telescope itself and the markers had particular apparent positions within the field of view. Wallace replied that the decisive point was simpler: if the three equal-height points were on a true straight line, the nearer marker should coincide with or cover the farther one when seen from the third point. It did not.

This is ordinary collinearity geometry. You do not need to know which direction the line points in order to test whether three points are on the same straight line.

Wallace even challenged Carpenter to arrange three ordinary objects in a straight line elsewhere and make the middle one appear displaced in the same way merely by choosing a different telescope orientation.

22. Straight Line Versus Level Line

This semantic issue continues to confuse modern discussions.

Imagine three stakes, each exactly ten feet above a vast spherical water surface, separated by miles.

The tops of the stakes trace a circle concentric with the water surface. They are all the same level height above the water locally.

They are not all on one straight line.

At short distance the difference is tiny. Over six miles it grows to feet.

Therefore “all markers were the same height above water” does not imply “all marker tops should visually align” unless one has already assumed that the water surface is planar.

This is exactly why the Wallace experiment is a test rather than a tautology.

23. The Contemporary Nature Report

The April 7, 1870 issue of Nature published a brief report titled “The Rotundity of the Earth.” It stated that six miles of Bedford Level were used and that three signals, approximately 13 feet 4 inches above the water, were placed three miles apart. Wallace expected the center signal to appear roughly five feet above the line joining the other two, whereas Hampden’s position required them to align.[1]

The importance of this source is chronological. It is not a modern retrospective written after flat-Earth arguments reappeared on the internet. It is a contemporary scientific report describing the experiment while the controversy was still active.

Nature returned to the controversy later in 1870 after Hampden and Rowbotham’s supporters continued their public attacks.[15]

One can disagree with the tone of those Victorian reports. The geometry does not depend on the tone.

24. Winning the Experiment Was Not the End of the Dispute

The referee associated with Wallace concluded that the experiment demonstrated curvature. Hampden’s side did not accept the verdict.

The dispute rapidly moved away from pure geometry into accusations of fraud, pamphlets, legal actions, and personal harassment.

Wallace later described the entire affair as one of the worst mistakes of his life. A 1908 article in Science reviewing the experiment quoted Wallace’s recollection that the lawsuits, prosecutions, settlements, experimental expenses, and years of harassment cost him more than the money he had initially received.[16]

The episode is sometimes retold as though Hampden’s refusal itself proves the scientific case. It does not. Personal conduct is not evidence for Earth curvature.

The evidence lies in the experiment.

The aftermath matters mainly because it explains why so much surviving material is polemical and why the story became culturally famous.

Another source of confusion is the later legal status of the wager.

Wallace was initially awarded the stake under the agreed refereeing process. Subsequent litigation over the wager and Hampden’s withdrawal created legal complications, and Wallace ultimately did not simply retain the money as an uncontested prize.

That has sometimes been rewritten as “a court ruled Wallace’s curvature experiment was wrong.” It did not.

A court deciding whether a wager is legally recoverable is not measuring the canal.

Likewise, Hampden’s later convictions or legal troubles do not prove the Earth is spherical.

Scientific and legal questions should be separated:

Only the first bears directly on Earth curvature.

26. Henry Yule Oldham’s Later Repetition

The Bedford Level did not disappear from scientific discussion after Wallace.

In 1901, geographer Henry Yule Oldham presented another experimental demonstration of Earth curvature along the Bedford Level to the British Association for the Advancement of Science. Contemporary references describe his work as a photographic re-measurement of the curvature.[17]

Photography was an important improvement because it reduced reliance on competing verbal descriptions of what each observer claimed to see through a telescope.

The existence of later repetitions also undermines the notion that Wallace’s result was a one-off trick dependent on a secret telescope alignment.

A good historical treatment should nevertheless avoid overselling any nineteenth- or early-twentieth-century experiment. Modern geodesy has far more powerful methods available today.

Bedford Level is valuable primarily because it illustrates experimental design and the history of flat-Earth reasoning, not because modern science depends on it to establish Earth’s shape.

27. Why Wallace’s Design Is Better Than Rowbotham’s

The improvements can be summarized directly.

FeatureRowbothamWallaceWhy it matters
Sight-line heightAbout 8 inches above water in the famous boat testAbout 13 feet above waterHigher rays are less dominated by the immediate water boundary layer.
ObservableWhether a distant boat/flag remained visibleRelative alignment of three equal-height pointsThree-point geometry provides a direct curvature signature.
Midpoint constraintNoneMarker near midpointDistinguishes a convex arc from a straight plane more directly.
Viewing directionPrimarily one-way observationObservations from both endsReciprocity helps expose setup errors.
Level referenceLow telescope and later theodolite argumentsAdditional spirit-level telescope testGives an independent local-horizontal reference.
Refraction awarenessIncorrectly denied continuous atmospheric refractionExplicitly allowed for refractive reductionAtmospheric optics is essential over water.

None of these improvements makes Wallace infallible. They simply make his experiment better at separating competing explanations.

28. Wallace’s Experiment Was Not Perfect

A rigorous modern analysis should resist turning Wallace into a scientific saint.

His experiment had limitations.

There was no dense vertical meteorological profile along the entire canal. Refraction was estimated rather than reconstructed from modern radiosonde-style or distributed sensor data. Water levels and marker elevations were measured with nineteenth-century methods. Documentation was filtered through adversarial participants and contemporary publications.

The atmosphere still affected the observation at thirteen feet above water. Raising the ray reduced the worst near-surface effects; it did not eliminate refraction.

Wallace’s expected numerical curvature also depended on exact spacing and marker geometry, which varied slightly among historical descriptions.

These limitations matter if one wants a precision measurement of Earth radius from the 1870 data.

They do not erase the basic qualitative result: the three equal-height marker points were observed not to be collinear in the way a flat-plane interpretation required, and their departure was in the direction and approximate magnitude expected from convex curvature.

29. What a Modern Bedford Level Test Should Do

A modern replication could be dramatically better.

The goal should be stated quantitatively before the experiment: compare the expected apparent position of multiple surveyed markers under a planar model and an ellipsoidal-Earth model with measured refraction.

A strong setup would include:

  1. survey-grade coordinates for all stations;
  2. orthometric heights tied to a known vertical datum;
  3. multiple markers at fixed known heights;
  4. optical paths several meters above the water where possible;
  5. simultaneous observations from both ends;
  6. calibrated total stations or telescopes;
  7. temperature sensors at multiple heights;
  8. water-temperature measurements;
  9. pressure, humidity, wind, and solar-radiation measurements;
  10. continuous imaging rather than selected snapshots;
  11. precomputed flat and globe predictions;
  12. a complete uncertainty budget.

Better still, repeat the experiment under several thermal conditions. Curvature should remain fixed. Refraction should change.

If the apparent midpoint displacement varies with atmospheric profile around a stable geometric baseline, the two effects can be separated empirically.

30. Why Simply Replacing the Telescope with a Laser Does Not Solve It

Modern flat-Earth demonstrations often assume a laser removes the ambiguity that affected Bedford Level.

It does not.

A laser beam still travels through the same refractive atmosphere. A low laser over cold water can be bent downward just as light from a distant boat can. Real laser beams also diverge, have finite width, wander in turbulence, and can be detected through scattered light.

The laser therefore adds new measurement problems unless the beam is carefully characterized.

A laser can be an excellent instrument in a controlled geodetic experiment. It is not a magic “straightness machine” whose presence makes atmospheric optics disappear.

For a full analysis of that class of test, see Testing Laser-over-Water Experiments.

31. What the Flat Model Predicts

The classical Bedford flat model is simple.

If the canal surface is one Euclidean plane and three marker centers are placed exactly the same perpendicular height above it at equal intervals, the three marker centers are collinear.

From an eye located at the first marker’s center, the midpoint marker should overlap the distant endpoint marker, ignoring ordinary perspective size differences.

No “middle hump” should appear because the base surface has no convexity.

A flat model can invoke atmospheric refraction to alter the apparent positions. But once it does, it must calculate that alteration rather than using refraction only when convenient.

Moreover, if a flat model requires a specific refractive distortion to mimic the systematic convex signature across reciprocal observations, that prediction can be measured meteorologically.

The flat hypothesis is therefore testable if stated quantitatively.

32. What the Globe Model Predicts

For an idealized six-mile circular arc of Earth radius, the midpoint of an equal-height marker arc lies about six feet above the straight chord between the endpoints.

Atmospheric refraction generally reduces the apparent amount under ordinary downward-bending terrestrial-refraction conditions because the ray itself curves somewhat toward the surface.

Thus the expected optical signature is a midpoint displacement somewhat less than the pure geometric sagitta, with the exact apparent amount dependent on the refractive profile.

That is approximately what Wallace reported.

The strength of the model is not that one nineteenth-century sight line happens to resemble a diagram. Earth curvature is independently measured through geodesy, satellite tracking, leveling networks, radar propagation, celestial observations, gravimetry, and global navigation systems.

The Bedford test is one local manifestation of that larger geometry.

33. Common Objections Tested One by One

“Rowbotham saw the boat with his own eyes. Observation beats theory.”

The observation can be genuine. What does not follow is the interpretation. The apparent position of an object depends on both surface geometry and the optical path through the atmosphere. Strong terrestrial refraction can reveal objects below the unrefracted geometric horizon.

“Refraction cannot happen if both observer and boat are in air.”

False. A refractive-index gradient within air bends light continuously. Mirages and looming are examples. Rowbotham explicitly denied this mechanism, but modern atmospheric optics measures it routinely.

“Wallace raised the telescope because he was hiding curvature.”

Raising the sight line reduces the influence of the strongest near-water refractive gradient. It improves the experiment. The three-marker geometry also creates a direct collinearity test.

“If the middle marker looked high, Wallace simply tilted the telescope.”

Telescope tilt can move the whole field of view, but it cannot make three truly collinear points become noncollinear. The key large-telescope observation was the relative position of the midpoint and far marker.

“The telescope was not level, so the experiment is invalid.”

A level telescope is needed to measure depression relative to local horizontal. It is not needed merely to determine whether three points are collinear. Wallace also performed a separate spirit-level telescope observation.

“The middle marker should look lower, not higher, on a globe.”

Relative to the straight chord joining two equal-height endpoints on a convex arc, the midpoint lies outward—above that chord. Which direction is “up” in an inverted telescope image is a separate optical-display question.

“All three markers were the same height above water, so they must line up.”

Only if the water surface is planar. On a sphere, equal-height points follow a concentric curve and are not collinear over long distances. That is precisely what the experiment tests.

“Nature only assumed Wallace was right because scientists already believed in a globe.”

The historical report is not the proof. The geometry is independently reproducible. The value of the Nature report is that it documents the setup and dispute contemporaneously.

“Hampden’s side disagreed, so the result was inconclusive.”

Disagreement by itself does not make measurements inconclusive. One must evaluate the stated observations and geometry. Carpenter’s published objections are available, as are Wallace’s replies.

“Wallace lost the court case, so he lost the experiment.”

The legal enforceability of a wager is a different question from the shape of the canal surface. Courts do not change optical geometry.

“Rowbotham repeated the experiment many times.”

Repeated observations using the same confounded low sight line can repeatedly measure the same refractive phenomenon. Replication is valuable only if the design distinguishes the competing explanations.

“Water cannot curve because water always seeks level.”

In physical geodesy, a level surface is an equipotential surface perpendicular to local gravity. Local horizontal directions change across Earth. A large level surface can therefore be curved.

“If refraction follows Earth curvature, that is an impossible coincidence.”

Refraction does not always follow Earth curvature. It varies with atmospheric conditions. Sometimes it bends less, sometimes more, sometimes upward, and sometimes enough to create looming or ducts. The variability is measurable and is exactly why low water tests can produce different apparent results.

“The canal itself is engineered flat.”

Engineers can control water levels and gradients locally, but “flat” in canal operation does not mean an infinite Euclidean plane independent of gravity. Water surfaces follow local gravitational equipotentials subject to small hydrodynamic gradients.

“Wallace predicted the result before observing it, so he was biased.”

Pre-stating a prediction is a strength in experimental science. A test is more informative when competing models make different predictions before the data are seen.

“Wallace was a famous scientist, therefore the establishment simply sided with him.”

Authority is irrelevant to the geometry. Wallace’s fame does not make the midpoint displacement true. The setup can be reproduced by anyone with sufficiently controlled instruments.

“A six-foot bulge over six miles sounds too large.”

Using the sagitta approximation s ≈ L²/(8R) with L = 6 miles and R ≈ 3,959 miles gives about six feet. Large planetary radius does not imply zero local curvature; the effect grows with the square of baseline length.

“If the middle is six feet high, people should visibly see Earth curving everywhere.”

The six feet is measured relative to a straight chord stretching three miles to each side, not as a six-foot hill in the water. Local slope remains extremely small. Human vision without a reference line cannot easily detect such gradual curvature.

“Oldham repeating Wallace proves nothing because they were both globe believers.”

Belief does not determine marker geometry. The correct question is whether the procedure and data can be independently reproduced. Modern instruments can test the same prediction without relying on either man.

34. The Deeper Methodological Lesson

The Bedford Level experiment is valuable today less because anyone needs a Victorian canal to prove Earth is round than because it is an unusually clear example of experimental underdetermination.

Rowbotham observed something real—or at least something entirely physically plausible—and assigned it to the wrong cause.

That happens in science constantly.

A strange observation does not automatically validate the explanation attached to it.

To distinguish causes, experiments need controls.

Rowbotham’s test contained two unknowns:

He assumed the second and claimed to measure the first.

Wallace improved the experiment by adding another geometric feature—the midpoint—and by moving the sight line farther from the strongest surface refractive layer.

A modern investigator can improve it further by actually measuring the refractive profile.

This progression is what good experimental science looks like. You identify a confounder and redesign the test so the hypotheses make distinct predictions.

The worst response would be to preserve a nineteenth-century experimental flaw because its result is rhetorically useful.

Conclusion

The Bedford Level experiment did not prove Earth is flat. It demonstrated how easily a long near-water optical observation can confuse atmospheric refraction with geometry—and how a better-designed three-point experiment reveals the curvature the first setup could not isolate.

Samuel Rowbotham’s original observation was not foolish in conception. A six-mile straight canal is a natural place to ask whether a water surface curves. He also correctly recognized that an observer only inches above the water should lose a short target behind the geometric horizon on a spherical Earth.

His fatal mistake was optical.

Rowbotham treated the ray from boat to telescope as straight. When critics invoked terrestrial refraction, he argued that refraction could occur only when light crossed from one medium into another. That is physically false. A continuously varying refractive index within the atmosphere bends rays continuously. Near cold water beneath warmer air, the bending can be downward. Looming can lift objects into view that are geometrically below the unrefracted horizon.

His eight-inch-high sight line was therefore almost ideally placed to be misled by the atmosphere.

Alfred Russel Wallace’s 1870 redesign addressed both weaknesses. He raised the sight line to roughly thirteen feet and used three equal-height markers across the six-mile span. On a plane, the three points should align. On a convex surface, the middle point should lie above the straight chord joining the endpoints by roughly five to six feet, depending on exact geometry and refraction.

The midpoint appeared displaced in the predicted direction. Reciprocal views agreed. A spirit-level telescope provided an additional local-horizontal test. Contemporary scientific reporting recorded the result, and Wallace’s own correspondence explicitly addressed the geometrical objections raised by Hampden’s referee.

The subsequent wager dispute, lawsuits, accusations, and harassment are historically fascinating but scientifically secondary. Whether Wallace legally kept £500 does not alter the location of the markers.

Later re-measurement by Henry Yule Oldham again used the Bedford Level to demonstrate curvature photographically. Modern science has since moved far beyond these canal experiments; geodetic networks, satellite ranging, GNSS, gravimetry, radar propagation, and precise leveling determine Earth geometry with vastly greater accuracy.

The Bedford Level remains useful because it shows why experiment design matters more than dramatic observation.

A target being visible farther than an unrefracted geometry calculation predicts does not immediately falsify curvature. It may reveal atmospheric optics. A laser striking a distant screen does not immediately prove a straight water plane. It may reveal beam divergence, pointing error, or refractive bending. A proper experiment must separate those variables.

And this is exactly what happened historically. The earliest Bedford observation asked only whether the distant object could still be seen. The improved experiment asked whether multiple equal-height points occupied the geometry predicted by a plane or by a convex surface.

They did not behave as a plane required.

For the optical issues in modern versions of the experiment, see Testing Laser-over-Water Experiments. For the broader visibility argument, see Testing “You Can See Too Far for Earth Curvature”. For the full set of flat-Earth claims, see Testing Flat Earth.

Sources and Further Reading

  1. Nature — “The Rotundity of the Earth” (7 April 1870). Contemporary report describing Wallace’s six-mile Bedford Level test, the three equal-height signals, and the predicted central displacement.
  2. Samuel Birley Rowbotham, Zetetic Astronomy: Earth Not a Globe — Experiment 1. Rowbotham’s own description of the boat, flag, eight-inch sight line, six-mile observation, and flat-Earth conclusion.
  3. NOAA National Geodetic Survey — Geodetic Leveling, Manual NOS NGS 3. Professional treatment of curvature, terrestrial refraction, sight-height control, balanced observations, and water/valley crossings.
  4. NOAA — U.S. Coast Pilot, Atmospheric Refraction and Mirages. Describes looming and superior mirages capable of elevating images and bringing objects below the normal horizon into view.
  5. Rowbotham, Zetetic Astronomy — Experiment 9 / discussion of refraction. Primary-source example of Rowbotham’s incorrect claim that refraction requires different surrounding media.
  6. NOAA JetStream — Radar Beams and Atmospheric Refraction. Explains continuous atmospheric refraction, superrefraction, subrefraction, and ducting caused by refractivity gradients.
  7. Rowbotham, Zetetic Astronomy — List of Bedford-Level Experiments and Diagrams. Primary-source index showing the multiple canal, flag, theodolite, and leveling experiments Rowbotham used.
  8. Rowbotham — Experiment 4, Forward Leveling. Shows Rowbotham’s reasoning about successive theodolite setups and horizontal water.
  9. NOAA National Geodetic Survey — Geopotential Surfaces. Explains the geoid, gravity-defined level surfaces, and the physical meaning of level.
  10. Alfred Russel Wallace — Letters to John Hampden (1870). Preserves Wallace’s acceptance of the £500 challenge, proposed tests, and correspondence about the Bedford River experiment.
  11. Wallace — “The Bedford Level Experiments” (1872). Wallace’s correction of later claims about who selected the Bedford site and how the test was agreed upon.
  12. Wallace — Reply to Hampden’s Charges. Wallace’s retrospective account of accepting the challenge and the controversy surrounding the wager.
  13. Robert M. Brown, “Experiments on Earth Curvature,” Science (1908). Summarizes the 1870 arrangement: bridge sheet, black band, midpoint pole and discs, telescope, and aftermath.
  14. Wallace — Letters Concerning the Bedford Canal “Flat Earth” Experiment, The Field (1870). Wallace’s detailed response to Carpenter, including the three-point geometry, reciprocal observations, level telescope, and curvature estimate.
  15. Nature — “The Rotundity of the Earth” (14 July 1870). Follow-up contemporary discussion of the ongoing Bedford Level controversy and earlier tests.
  16. Science (1908) — Bedford Level Historical Review. Includes Wallace’s later assessment of the financial and personal consequences of the dispute.
  17. Henry Yule Oldham — references to the 1901 British Association curvature demonstration. Provides bibliographic references to Oldham’s photographic Bedford Level re-measurement in the British Association annual report.
  18. Documents Concerning the Bedford Canal Experiment. Historical documents concerning the wager, referees, experiments, and later legal dispute.