[Book I, ch. 4 — “That Also the Earth, Taken as A Whole, is Sensibly Spherical”] Now, that also the earth taken as a whole is sensibly spherical, we could most likely think out in this way. For again it is possible to see that the sun and moon and the other stars do not rise and set at the same time for every observer on the earth, but always earlier for those living towards the orient and later for those living towards the occident. For we find that the phenomena of eclipses taking place at the same time, especially those of the moon, are not recorded at the same hours for everyone … And since the differences in the hours is found to be proportional to the distances between the places, one would reasonably suppose the surface of the earth spherical … For, if it were concave, the rising stars would appear first to people towards the occident; and if it were flat, the stars would rise and set for all people together and at the same time; and if it were a pyramid, a cube, or any other polygonal figure, they would again appear at the same time for all observers on the same straight line. But none of these things appears to happen. … Again, whenever we sail towards mountains or any high places from whatever angle and in whatever direction, we see their bulk little by little increasing as if they were arising from the sea, whereas before they seemed submerged because of the curvature of the water's surface. [Book I, ch. 7 — “That the Earth Does Not in any Way Move Locally”] Now some people, although they have nothing to oppose to these arguments, agree on something, as they think, more plausible. And it seems to them there is nothing against their supposing, for instance, the heavens immobile and the earth as turning on the same axis from west to east very nearly one revolution a day … But it has escaped their notice that, indeed, as far as the appearances of the stars are concerned, nothing would perhaps keep things from being in accordance with this simpler conjecture, but that in the light of what happens around us in the air such a notion would seem altogether absurd. … never would a cloud be seen to move toward the east nor anything else that flew or was thrown into the air.

— Almagest (c. 150 CE), Book I, chs. 4 and 7 (trans. R. Catesby Taliaferro)

Claudius Ptolemy worked at Alexandria around 150 CE. The Almagest is the most successful scientific book ever written by duration of service: its planetary tables were the working standard in Greek, Arabic and Latin for roughly fourteen centuries. It is not folklore. It is a quantitative, geometrical, tested model.

First, the irony. Book I ch. 4 is one of antiquity's cleanest arguments that the Earth is a sphere, argued exactly the way the modern globe is argued: eclipses timed at different longitudes give different local hours, and the offset scales with distance; a mountain rises out of the sea as you approach it. Ptolemy rules out the alternatives by name, and the flat option is dispatched in a single clause — “if it were flat, the stars would rise and set for all people together and at the same time.” The list's largest pre-modern authority refutes the list's headline claim on his fourth page.

Second, the concession. In ch. 7 Ptolemy considers a rotating Earth and grants — in his own words — that “as far as the appearances of the stars are concerned, nothing would perhaps keep things from being in accordance with this simpler conjecture.” He states the non-discrimination point himself. He rejects rotation not from the sky but from terrestrial mechanics: a thrown object or a cloud would be left behind. That argument is the direct ancestor of ARG-A10 and ARG-A17, and it is the one part of the chapter that is wrong.

SURFACE (weak). “These items only describe how things look.” True but empty; every observation describes how things look. Stated this way the objection would equally dismiss the observations that did settle the question.

DEEPER. Ptolemaic astronomy was real science, not a placeholder. It was quantitative, it made forward predictions of eclipses and planetary positions to stated accuracy, it was corrected against observation across generations, and it worked well enough to navigate and keep calendars for fourteen hundred years. Dismissing it as superstition misrepresents the history and makes the eventual correction look like a change of fashion.

KERNEL. The strongest form is epistemic, and it is correct. “Saving the appearances” was a serious methodological programme: build the geometry that reproduces the data and remain agnostic about what is physically moving. For the daily rotation of the sky, the geocentric and heliocentric accounts are observationally equivalent — they predict the same appearance to arbitrary precision, so no amount of watching the sky turn can decide between them. It follows that anyone asserting Earth's motion as established fact before the first genuinely discriminating measurement was going beyond the evidence available to them. Ptolemy says as much in ch. 7. That is a real concession and this page grants it in full.

Why it doesn’t save the claim.

The kernel is granted and it still leaves the argument where it started, because it is an argument about a set of observations, not about the world. Equivalence is two-edged: if the daily turning of the sky is consistent with a rotating sky, it is equally consistent with a rotating Earth. An observation compatible with both models is evidence for neither. Twenty items of it is still evidence for neither — restating a non-discriminating observation twenty times in different vocabulary does not accumulate into a discriminating one.

So these items cannot do the work the list needs. They are offered as proofs; they are shared premises. Both sides of the dispute predicted every one of them before the dispute began.

And the equivalence was never permanent. It held because instruments were not yet good enough, and it ended when they became good enough — by measurement, on dated occasions, not by anyone deciding the matter.

Take the items at face value first, because most are simply correct. The celestial sphere does appear to rotate about the observer. Sidereal rotation is constant to high precision at 23h 56m 04s. Solar declination repeats annually, the ecliptic tilt is stable on human timescales, precession is real, solar noon is symmetric, and the Ptolemaic apparatus did reproduce retrograde motion. Nothing there needs correcting.

Not all twenty are observations, though, and the exceptions should be marked rather than waved through with the rest. “Day–night cycle from firmament rotation”, “Precession from dome rotation” and “Seasonal stars via dome” do not describe the sky; they explain it, by a mechanism, and the mechanism is false rather than merely non-discriminating. A single dome turning over a plane hands every observer the same centre of rotation. The southern sky has a centre of its own, with its own circumpolar trails about it, and no rotation of one surface over one plane delivers both at once — which is the argument set out at B08. Two others, “Lunar phases explainable without orbit” and “Analemma lamp artifact”, are likewise model claims rather than observations, and the analemma one is answered below.

Three items deserve specific credit rather than rebuttal. Libration is a genuine perspective and orbital effect, reaching 7°54′ in longitude from orbital eccentricity and 6°50′ in latitude from the Moon's axial tilt, which together let us see about 59% of the lunar surface — as stated, that item is right. The analemma is a real figure-eight with an exact and unmysterious cause: the north–south extent is the Sun's changing declination from the 23.44° axial tilt, and the east–west width is the equation of time, an eccentricity term of amplitude 7.66 minutes plus an obliquity term of amplitude 9.87 minutes, running from +16m 25s around 3 November to −14m 15s around 11 February. And birds do navigate by the sky: Emlen's planetarium work with indigo buntings established a stellar compass calibrated on the rotational centre of the night sky rather than on individual stars. Real biology with no cosmological content — a bird orienting on the pole would orient identically under either model.

The relevant question is not whether these observations are true. It is whether any distinguishes a turning sky from a turning Earth. None does, and that is not a criticism of them. It is what Ptolemy himself says in Book I ch. 7. On this point the page agrees with Ptolemy.

What broke the equivalence was measurement, and each break has a date. 1610: Galileo observed Venus through its full cycle of phases, including gibbous and near-full. In the Ptolemaic ordering Venus rides an epicycle between Earth and Sun and can never present a fully lit face; the full sequence is impossible in that model. (Precision matters: this refutes Ptolemy, not Tycho, whose arrangement predicts the same phases — and the list's geocentric bulk is Tychonian, so it survives this particular test.) 1729: Bradley, observing γ Draconis, announced stellar aberration — an annual displacement of every star by up to 20.4955″, in phase with Earth's velocity rather than its position. That is a direct signature of the observer's motion, and it is the first observation a stationary Earth cannot accommodate. 1838: Bessel measured the parallax of 61 Cygni at 0.3136″ ± 0.0136″; the modern value is 0.28718″. 1851: Foucault hung a 28 kg bob on a 67 m wire in the Panthéon and the swing plane precessed at about 11.3° per hour, matching Ωsinφ at the latitude of Paris — Earth's rotation made visible in a room, with no sky involved at all.

Retrograde motion deserves separate treatment, because it is where the difference between fitting and explaining is sharpest. Ptolemy's epicycles reproduce retrograde loops accurately. But to make them come out at the right times he had to impose a condition by hand: for each superior planet, the epicycle's period is locked to one year and its radius stays parallel to the Earth–Sun line. Nothing in the model says why the Sun should be written into the motion of Mars, Jupiter and Saturn at all — it is a parameter fixed to match the data. In the heliocentric account the same fact is not a parameter but a consequence: retrograde is what overtaking looks like from the inside lane, so it must occur when Earth passes the planet — exactly opposition — and the planet must be nearest and brightest precisely then. One model absorbs the correlation; the other predicts it without being asked. Both save the appearances; only one accounts for them.

So the verdict on this cluster is not that its members are false. It is that they are shared. They were the common ground of the dispute, and the dispute was settled elsewhere. Roughly ninety items across this list have this same shape. They are best read not as claims to be refuted but as a description of the sky that both parties already agreed on.

The list says

Celestial sphere rotation about observer. / Day–night cycle from firmament rotation. / Precession from dome rotation. / Seasonal stars via dome.

The source says

Almagest I.7: “as far as the appearances of the stars are concerned, nothing would perhaps keep things from being in accordance with this simpler conjecture” — the conjecture being that the earth turns. And I.4: “if it were flat, the stars would rise and set for all people together and at the same time.”

A concession was re-used as a proof. Ptolemy’s chapter I.7 says close to the opposite of what twenty proof-items need it to say. He states the rotating-Earth hypothesis and concedes it: as far as the stars go, nothing would keep things from being in accordance with the simpler conjecture. On his own account the sky cannot decide. He rejects rotation on other grounds — clouds and thrown objects would be left behind — and that is the one part of the chapter that is wrong. The list takes the very appearances he declared non-discriminating and enters them as evidence, twenty times over.

A second drift rides along, and the enum has no clean word for it. Items 34, 39 and 210 credit him with a “firmament” and a “dome.” Ptolemy has neither. Three chapters earlier he proves the Earth spherical and disposes of the flat option in a single clause. unsourced_addition understates that: he does not merely omit the dome, he refutes the cosmology it belongs to.

Our refutation answers Ptolemy, agreeing with him that these observations were the shared premises of the dispute, and dating the four measurements that ended the equivalence he was right to assert.

[1865, Section II — the seed of the cluster] "It is therefore demanded that if the Earth had a motion upon axes from west to east, and a ball, instead of being dropped down a mine or allowed to fall from the mast head of a ship, be shot upwards into the air; from the moment of its beginning to descend the surface of the Earth would turn from under its direction, and it would fall behind or to the west of its line of descent. On making the experiment no such effect is observed, and therefore the conclusion is unavoidable, that the Earth DOES NOT MOVE UPON AXES!" [1881, ch. III — the atmospheric form, ancestor of all 13 items] "Take a large grinding stone, and let the whole surface of the rim be well rubbed over with a saturated solution of phosphorus in olive oil … If it be now turned rapidly round … the phosphoric vapour, or the steam from the flannel, which surrounds it and which may be called its atmosphere — analogous to the atmosphere of the earth — will be seen to follow the direction of the revolving surface. Now the surface of the earth is very irregular in its outline … and if it is a globe revolving on its axis, with the immense velocity at the equator of more than a thousand miles an hour, it is exceedingly difficult if not altogether impossible to conceive of such a mass moving at such a rate, and yet not taking the atmosphere along with it. … Hence we are compelled to conclude that if the earth revolves, the atmosphere revolves also, and in the same direction. "If the atmosphere rushes forward from west to east continually, we are again obliged to conclude that whatever floats or is suspended in it, at any altitude, must of necessity partake of its eastward motion. … But what is the fact? If we fix upon any star as a standard or datum outside the visible atmosphere, we may sometimes observe a stratum of clouds going for hours together in a direction the very opposite to that in which the earth is supposed to be moving. … Not only may a stratum of clouds be seen moving rapidly from east to west, but at the same moment other strata may often be seen moving from north to south, and from south to north. … Such a state of the atmosphere is compatible only with the fact which other evidence has demonstrated, that the earth is at rest." [And Rowbotham's explicit rejection of the ship analogy:] "A ship with a number of passengers going rapidly in one continued direction, like the earth's atmosphere, could nevertheless have upon its deck a number of distinctly and variously moving objects, like the clouds in the atmosphere. … but the passengers are sentient beings, having within themselves the power of distinct and independent motions: the clouds are the reverse; and here the parallelism fails."

— Earth Not a Globe (1865), Zetetic Astronomy: Earth Not a Globe! — Section II, “The Earth No Axial or Orbital Motion” (1865); expanded into the explicitly atmospheric form in the enlarged 3rd ed. (1881), ch. III, “Experiment 4” and following

All thirteen items descend from these pages, and Rowbotham's chain of reasoning is more careful than most of his successors'. He grants that a rotating Earth would carry its atmosphere with it — his grinding-stone experiment is offered as proof that it would. His objection is the next step: if the whole envelope is swept eastward at over a thousand miles an hour, then everything suspended in it must be swept eastward too, like cork on a moving stream. Clouds drifting east-to-west, or several strata drifting in different directions at once, therefore look to him like a fatal counter-observation.

The lineage runs much further back. Ptolemy raises the identical objection in the Almagest I.7, around 150 CE: if the Earth turned, “never would a cloud be seen to move toward the east nor anything else that flew or was thrown into the air.” Ptolemy then considers the co-rotating-air reply and dismisses it. Rowbotham's 1881 pages reproduce both halves of Ptolemy's argument — the objection and the refusal of the co-rotation answer — with no evidence he had read the Almagest. It is a natural argument that intelligent people have reached independently for nearly two millennia, which is exactly why it deserves a careful answer rather than a dismissal.

The surface version — “the atmosphere is just dragged along” — is weak, and it invites Rowbotham's follow-up: dragged by what? A fifty-mile shell of gas resting on a rough surface is not a viscous coupling problem with an obvious answer, and Rowbotham is right that friction alone is a poor mechanism.

The kernel underneath is a genuine problem in physics: how is angular momentum communicated to and maintained in a fluid envelope, and why does that envelope not lag or lead the solid body beneath it? This is not a settled triviality. It is the reason atmospheric superrotation on Venus remains an active research question seventy years after its discovery: Venus's cloud-top atmosphere circles the planet in about four Earth days while the solid planet takes 243 — roughly sixty times faster — and identifying the momentum pump that sustains it required a dedicated orbiter (JAXA's Akatsuki) and produced a Science paper in 2020 attributing it to thermal tides. Titan's stratosphere superrotates too. So “why should a fluid envelope share the rotation of the body it sits on?” is a real question with a non-obvious answer that differs from planet to planet. Rowbotham deserves credit for noticing that the coupling needs an account.

Why it doesn’t save the claim.

The kernel is a good question; the prediction Rowbotham derived from it fails in three separable ways.

First, the mechanism is misidentified. The atmosphere is not dragged by friction from a rough surface. It is gravitationally bound, it formed with the Earth, and it has shared the Earth's angular momentum since. Nothing needs to grip it.

Second, the observation Rowbotham treats as decisive — strata moving in different directions, sometimes east-to-west — is not a residual of failed coupling. Bulk co-rotation at the equator is about 465 m/s. Ordinary surface winds are 0–30 m/s, and even jet-stream cores are 30–100 m/s. These are differential motions of a few per cent riding on a co-rotating carrier. A stratum drifting westward relative to the ground is a 10 m/s perturbation on a 465 m/s carrier, not a counter-example to the carrier.

Third, and decisively, the kernel question has an empirical answer that runs the wrong way for the claim. If Rowbotham had asked what observable signature a rotating frame would leave in the wind field, rather than assuming it must be a permanent westward gale, he would have been asking the question Hadley, Coriolis and Ferrel were answering in the same century. The signature is a latitude-dependent deflection that reverses sign at the equator, and it is present in every wind system he cites.

Start where Rowbotham starts, because it is not where the list puts him. He does not claim the air cannot move with the Earth. He claims the opposite, in terms: “we are compelled to conclude that if the earth revolves, the atmosphere revolves also, and in the same direction” — and he builds an experiment with a grinding stone to demonstrate it. His argument is the next step: that granting co-rotation, no residual effect is felt, and that this absence is what tells against the motion. So the question is not whether the atmosphere is bound to the Earth. It is whether a rotating Earth should leave a residue an observer standing on it could detect. It should, it does, and it was measured.

First, why the obvious residue is absent — answered in 1632. The reply predates Earth Not a Globe by 233 years. In the Second Dialogue of Galileo's Dialogo, Salviati states the principle: “Whatsoever motion may be ascribed to the Earth, it is necessary that it be to us, (as inhabitants upon it, and consequently partakers of the same) altogether imperceptible … so long as we have regard onely to terrestrial things.” Galileo dramatises it with the sealed cabin below decks, where flies fly and water drips straight down and the ship's motion cannot be detected from inside — and the point of putting the observer below decks is that the enclosed air participates in the motion — a premise Rowbotham grants, so it is common ground rather than a rebuttal. What Galileo supplies that Rowbotham does not accept is the consequence: uniform shared motion is undetectable from inside, so its absence from everyday experience is exactly what a moving Earth predicts.

Galileo also answers Rowbotham's specific air-gun experiment, in advance and almost verbatim: “those who make the Earth moveable, answer, that the piece, and the ball that is in it, partake of the same motion with the Earth … And the same you see to ensue, in making the experiment in a ship with a bullet shot upwards perpendicularly with a Crosse-bow, which returneth to the same place whether the ship doth move, or stand still.” Rowbotham performed that experiment in the 1860s, got Galileo's predicted result, and read it as a refutation.

The centrepiece: the deflection reverses across the equator. The strongest evidence against this cluster is the evidence the cluster itself points at. Earth's rotation produces the Coriolis effect, which deflects moving air to the right in the Northern Hemisphere and to the left in the Southern. The consequences are not subtle. The trade winds blow toward the west in both hemispheres but arrive from opposite corners: from the northeast in the north, from the southeast in the south. Cyclonic storms follow the same rule — counter-clockwise in the north, clockwise in the south.

This reversal is fatal to the sun-driven-only alternative the items propose (Jet streams sun-driven, Trade winds sun-driven, Wind bands sun-synced). Solar heating is a thermal forcing. It is approximately symmetric about the thermal equator and it has no handedness: it can drive air from hot to cold, but it contains nothing that could make air veer clockwise on one side of a line and anticlockwise on the other. Only a rotation axis defines a chirality, and only a rotation axis makes that chirality flip where the axial component of the local vertical flips. A sun-driven-only model does not merely fail to predict the reversal; it has no term from which a reversal could come.

The item that concedes the case. Cloud opposite spin — the observation that storm systems rotate one way north of the equator and the other way south — is the hemispheric reversal, offered as though it were a difficulty. It is not a difficulty; it is the fingerprint. Rowbotham does the same thing in 1881: to support his case he quotes Sir James Clark Ross on upper cloud strata “moving in an exactly opposite direction to that of the wind — a circumstance which is frequently recorded in our meteorological journal both in the north-east and south-east trades.” He names the north-east and south-east trades, in the same sentence, without registering that their mirrored orientation is the thing needing explanation.

The Coriolis parameter and the calm belts. The strength of the deflection is f = 2Ωsinφ. Sinφ is zero at the equator and maximal at the poles, so a rotating Earth predicts a band along the equator in which rotational deflection vanishes. That band exists: NOAA notes that “right at the equator there is almost no wind at all — an area sometimes called the doldrums.” The prediction goes further and is testable: with f→0 there is no torque to organise convection into a rotating vortex, and NOAA's Hurricane Research Division states a storm “must be at least 300 miles from the equator in order for the Coriolis force to create the spin” — about five degrees, which is where the observed tropical-cyclone exclusion zone sits. Equatorial calm belts is therefore not a problem for rotation. It is one of rotation's sharper confirmed predictions, listed on the wrong side of the ledger.

Jet streams and pressure bands. Jet streams sun-driven is half right, and the omitted half is the rotation. Jets are driven by horizontal temperature gradients, but the reason that gradient produces a narrow west-to-east ribbon rather than a broad poleward drift is angular momentum in a rotating frame. NOAA states it directly: “For air moving toward the poles, the Earth's rotational velocity decreases beneath it, but the air itself retains its eastward momentum. This results in wind that moves faster than the Earth rotates.” That sentence is a description of the atmosphere carrying the planet's rotational momentum — the thing the cluster says does not happen. The same applies to Constant air pressure bands and Atmosphere concentric layers: the subtropical high near 30° and the subpolar low near 60° are surface signatures of the three-cell structure, and the reason there are three cells per hemisphere rather than one is that rotation limits how far a thermally direct cell can extend before conservation of angular momentum breaks it. On a non-rotating Earth you would expect single pole-to-equator cells and no banded pressure structure.

Superrotation, misapplied. Atmospheric superrotation contradiction uses a real technical term for the opposite of what it means. Superrotation is an atmosphere rotating faster than the body beneath it — Venus's cloud deck at roughly sixty times the planet's rate, Titan's stratosphere at 100–200 m/s. Earth's atmosphere does not significantly superrotate. The term presupposes a co-rotating baseline from which to depart, so invoking it concedes the framework the argument is trying to deny.

Airglow. No annual airglow wind cannot be assessed as stated. Airglow is real — faint chemiluminescent emission concentrated near 90 km, routinely used to measure upper-atmosphere winds — but the item specifies no quantity, no predicted magnitude, no observation and no instrument. It should be marked underspecified rather than refuted.

Routine measurement. The Earth-rotation effect on air is not inferred; it is measured daily. Foucault demonstrated rotation directly in Paris on 3 February 1851, and the demonstration is quantitative: the plane of swing precesses at Ωsinφ — half the Coriolis parameter — so a full turn takes about 32 hours at the latitude of Paris, slows towards the equator, and never completes there at all. Rowbotham devotes several pages of the 1865 edition to collecting newspaper complaints about Foucault's pendulum, which tells us he understood how much rested on it. Every operational weather model carries an explicit Coriolis term, and forecasts fail without it.

The list says

Atmospheric coupling unexplained. / No wind drag from 67,000 mph orbit. / No visible rotation of atmosphere or clouds.

The source says

“it is exceedingly difficult if not altogether impossible to conceive of such a mass moving at such a rate, and yet not taking the atmosphere along with it. … Hence we are compelled to conclude that if the earth revolves, the atmosphere revolves also, and in the same direction.” — and, at the end of the argument, “Such a state of the atmosphere is compatible only with the fact which other evidence has demonstrated, that the earth is at rest.”

The item states the opposite of its source. Rowbotham is blunter than the list, not more careful: his conclusion is that the observed sky is compatible only with a stationary Earth. So the drift is not in the strength. It is in the premise. Rowbotham grants co-rotation and spends a page establishing it — the grinding-stone experiment exists to show that a rugged spinning globe must carry its air — and his objection begins only afterwards: if the whole envelope streams east at 1,042 miles an hour, everything floating in it should stream east too, and clouds visibly do not. The list inverts that into “atmospheric coupling unexplained” and a drag from the 67,000 mph orbit, a figure and a mechanism absent from his atmospheric argument, which concerns axial rotation only.

This bears on our own text. The refutation opens by answering the fragment — Galileo’s sealed cabin, the air is inside the cabin, the atmosphere co-rotates — and Rowbotham agrees with every word of it. The part that answers him is the arithmetic, a 10 m/s stratum riding a 465 m/s carrier, and the hemispheric reversal of the Coriolis deflection that his own citation of Ross’s north-east and south-east trades hands us.

Tycho Brahe, Kepler, and others, rejected the Copernican theory, principally on account of the failure to detect displacement or parallax of the fixed stars. [p. 82] Dr. Bradley declared that what many had called “parallax,” was merely “aberration.” But “Dr. Brinkley, in 1810, from his observations with a very fine circle in the Royal Observatory of Dublin, thought he had detected a parallax of 1″ in the bright star Lyra (corresponding to an annual displacement of 2″). This, however, proved to be illusory; and it was not till the year 1839, that Mr. Henderson …” [pp. 82–83] The parallax thus assigned α Centauri, is so very nearly a whole second in amount (0″.98), that we may speak of it as such. … “Professor Bessel made the parallax of a star in the constellation Cygnus to be 0″.35.” [pp. 83–84] … it is only necessary to state as an absolute truth the result of actual experiment, that, a given fixed star will, when observed from the two ends of a base line of not more than three feet, give a parallax equal to that which it is said is observed only from the two extremities of the earth's orbit. [p. 86] It is useless to say, in explanation, that this very minute displacement, is owing to the almost infinite distance of the fixed stars; because the very same stars show an equal degree of parallax from a very minute base line. [p. 86]

— Earth Not a Globe (1865), 3rd ed., rev. and enl. (London: Day, 1881), ch. III “The Earth No Axial or Orbital Motion”, pp. 82–87. Not in the 1865 first book edition — see gloss.

The edition is not the one we credited, and the correction cuts both ways. Our cluster record attributes this material to Rowbotham 1865. It is not there. Two independent full-text queries against Project Gutenberg #69892 — the 1865 Simpkin, Marshall first book edition — return the string “parallax” twice, both times as the author’s byline. That edition’s SECTION II, The Earth no Axial or Orbital Motion, contains the vertical air-gun test and the two-parallel-tubes test and no discussion of annual parallax, of Tycho, Kepler, Bradley, Brinkley, Henderson or Bessel, and no statement about the distance or apparent size of the stars. The whole argument above is an addition of the 1881 third edition, which grew the book from 221 to 430 pages.

That matters, because our own rule is that charging an author with ignoring data is a serious fault unless the data was available to them, then. Dating the passage to 1881 removes any excuse: it puts Rowbotham 43 years after Bessel, writing with Henderson’s and Bessel’s numbers open in front of him — he prints them — and it is in that setting that he answers them with an experiment whose apparatus cannot resolve the angle in dispute.

Note what he does not say. He does not say parallax is unobserved. He recites the failures (Tycho, Kepler, Brinkley’s illusory second of arc), then the successes, then reassigns the successes to a three-foot base line. The compressed list items say the opposite of this, and so does the movement’s modern descendant: Eric Dubay’s proof 19 asserts that “after 190,000,000 miles of supposed orbit around the Sun, not a single inch of parallax can be detected in the stars, proving we have not moved at all.” Rowbotham’s own chapter contains the inches.

And the pseudonym line is a pun, not a fact. Our basis string says he “took his pseudonym from the thing he denied.” He adopted “Parallax” by the end of 1849, dropping an earlier pseudonym, and Schadewald reads the choice as the ordinary optical sense of the word — the apparent change in a thing’s position when the viewer moves — which is the mechanism his entire perspective account of sunset runs on. He named himself after a principle he used, three decades before he wrote about the astronomical measurement, and when he finally did write about it he conceded the observations. The sentence is too good to be true and should come off the page.

SURFACE (weak — do not use). “They think parallax has never been measured.” Aimed at the fragment, and it misses the source by a wide margin. Rowbotham quotes the measurements. Anyone who opens with this is answering Dubay’s 2015 paraphrase and will be shown the 1881 text.

DEEPER. Two of the six items are simply true, and the strongest one is a genuine classical objection. “No naked-eye parallax” is correct: the largest annual parallax of any star is Proxima’s 768.07 mas, about 0.77 arcsec, and the unaided eye resolves roughly one arcminute — eighty times coarser. “Lack of stellar parallax in early measurements” is also correct, and it is not a small fact: from Copernicus to Bessel, some three centuries, the predicted effect was looked for and not found, and every claimed detection before 1837 collapsed. Hooke, Flamsteed and Bradley all tried; Bradley’s search produced aberration and nutation instead of parallax; Calandrelli got four arcseconds for Vega and Brinkley one arcsecond for the same star in 1810, and both were illusory. This is exactly why Ptolemy and, far more rigorously, Tycho Brahe rejected a moving Earth.

KERNEL. Two things, and the second is the sharper.

First, Tycho’s version was a valid deduction from a false premise, and the false premise was an instrument. Tycho could see that stars appeared to have real angular diameters. If a star subtends a measurable disc and must also be far enough away to show no parallax, then it is physically enormous — on Graney’s rendering of Locher, who cites Tycho, any first-magnitude star “well larger than the whole orbit of the Earth”, and Ingoli’s version has the fixed stars “surpass or equal the size of the orbit circle of the Earth itself.” That is a real physical absurdity derived by correct reasoning from an observation anyone could repeat. The observation was wrong, and not because Tycho was careless: the apparent discs are diffraction and atmospheric seeing, the point-spread function of the eye and later of the telescope, and nobody could have known that before Airy’s On the Diffraction of an Object-Glass with Circular Aperture in 1835. The star-size argument is therefore not a stupid argument. It is a dated one, and the date is legible.

Second, Rowbotham identified the correct discriminating test. This is the best thing in the cluster and the list has thrown it away. He understood, correctly, that a measured stellar displacement is evidence for the Earth’s orbit only if the displacement is caused by the base line — and therefore that the way to settle it is to change the base line and see whether the answer changes. That is a properly zetetic instinct and it is the right experiment. His claim is that a three-foot base line yields the same parallax as the orbit, which would show the effect has nothing to do with the Earth’s position and everything to do with the observer’s. If that were so, the entire distance ladder would fall in an afternoon.

Why it doesn’t save the claim.

Because the base-line test has been run, continuously, for 188 years, and it answers the other way. Parallax scales linearly with the base line and inversely with distance; that is not an assumption of the model but the definition of the quantity. 61 Cygni A shows 285.9949 ± 0.0599 mas on a one-astronomical-unit base line (Gaia DR3). Three feet is 6.1 × 10⁻¹² of an astronomical unit, so on Rowbotham’s base line the same star must shift by about 1.7 × 10⁻¹² arcsec — roughly eleven million times smaller than Gaia’s median uncertainty of 0.02–0.03 mas. His “equal degree of parallax from a very minute base line” is not a marginal error. It is off by eleven orders of magnitude, and the only observation he offers for it is a pair of hand-aligned tubes on a wooden frame — a strange foundation in a book whose method is to take nothing for granted.

Because the naked-eye point dates the argument instead of supporting it. “No naked-eye parallax” is true and inert. It is the observation that every annual parallax is smaller than one arcsecond, which is what a heliocentric model with distant stars predicts. It stopped being an objection the moment the instruments crossed the threshold, and we can name the years: Struve on Vega in 1837, Bessel on 61 Cygni in 1838, Henderson on α Centauri in 1839.

And because the false premise underneath the strongest version has been measured. Real stellar angular diameters are milliarcseconds. The largest yet found, R Doradus, is 0.057 ± 0.005 arcsec — about one thousandth of the eye’s resolution limit, and thirty per cent larger than Betelgeuse, which held the record for 75 years. Nobody has ever seen a stellar disc with the naked eye or through a Victorian telescope. What they saw was their own optics.

1. Answer the 1881 text, because the fragment is not what its author said. Rowbotham’s chapter concedes more than the list does. He recites the history of failure — Tycho, Kepler, Bradley’s reassignment of “what many had called ‘parallax’” to aberration, Brinkley’s 1810 arcsecond for Vega that “proved to be illusory” — and then quotes the successes: 0″.98 for α Centauri, 0″.35 for Bessel’s star in Cygnus. His position is not that the displacement is absent. It is that the displacement is real and has a different cause. That is a better argument than the one on the list, and it is the one we have to beat.

2. The chronology, stated precisely, because it decides how much blame is fair. Bessel published 61 Cygni in 1838; Struve had published Vega in 1837 and Henderson α Centauri in 1839. Take the three dates in Rowbotham’s career separately.

1849, the sixteen-page pamphlet, and the year he adopted the name “Parallax”. Eleven years after Bessel — but scepticism was still defensible, and we should say so. Three results existed, all at the edge of what heliometers and filar micrometers could do, and they did not agree well. Struve got 0.125″ for Vega in 1837 and then revised it to 0.261″ in 1840, which is twice the modern 0.130″ and a move away from the truth. Henderson published one arcsecond for α Centauri against a modern 0.75″. Bessel’s 0.3136″ for 61 Cygni is about ten per cent above the modern 0.2860″, and a later reduction of his own data moved it further out, to 0.360″. Behind them lay a century of retracted claims. A well-informed sceptic in 1849 could reasonably have said: these are three heroic measurements at the limit of the art, they disagree, and claims of this kind have collapsed before. He would have been in a shrinking minority — the Royal Astronomical Society had already given Bessel its gold medal, and John Herschel called the result “the greatest and most glorious triumph which practical astronomy has ever witnessed” — but the position was available. The 1849 pamphlet does not take it; it does not discuss the matter at all.

1865, the first book edition. Twenty-seven years after Bessel, and still nothing. Its Section II, with the same title the 1881 chapter carries, argues against orbital motion from an air-gun and from two parallel tubes, and never mentions annual parallax. By this date the objection was strained: the results had been confirmed by other observers on other stars with other instruments, and the disagreements had begun to close rather than widen. But silence is not an error, and we should not score it as one.

1881, the third edition. Forty-three years after Bessel, and this is where the fault is. He now engages, prints the numbers, and answers them with a factual claim — three feet gives the same parallax as the orbit — introduced with the words “it is only necessary to state as an absolute truth the result of actual experiment.” The experiment is there, and it is the weak point. Seven pages earlier, at pp. 79–81 of the same chapter, he sets out the apparatus: two carefully-bored metallic tubes “not less than six feet in length”, placed “one yard asunder” on a wooden frame and adjusted so that their axes of vision “shall be perfectly parallel to each other”, directed at a notable fixed star a few seconds before its meridian time, with an observer at each tube knocking the moment the star enters his field. His reported result is that “a distinct period of time will elapse between the signals given” — the same star is “not visible at the same moment by two parallel lines of sight … when only one yard asunder.” One yard is three feet, so the p. 86 sentence we quote is a back-reference to that trial rather than a bare assertion.

The trouble is what the rig can resolve. Over a one-yard base line 61 Cygni is displaced by about 1.7 × 10⁻¹² arcsec, while a naked eye behind an unmagnifying six-foot tube resolves no better than about a minute of arc — some thirteen orders of magnitude coarser. Two genuinely parallel tubes a yard apart must therefore show the star at the same instant. A lag between the knocks measures one quantity only: the residual angle between the two axes, which the star’s diurnal motion — up to about 15 arcsec per second of time — turns into a delay long enough to hear. Rowbotham sees the mechanism and reads it the other way, noting that “a slight inclination of the tube, B, C, towards the first tube A, S, would be required for the star, S, to be seen through both tubes at the same instant.” He names no star, no date, no observer and no measured angle, and he reports no alignment control — reversing the frame end for end, or repeating the trial on a second star, would have separated the sky from the woodwork. In a book whose founding rule is to proceed “only by inquiry; to take nothing for granted,” the one check that would have told him which of the two he was reading is the one not made. This is not anachronism, it is the opposite: he had the data, he quoted the data, and he set it aside on the authority of an instrument that reads out its own construction error.

3. The classical objection was rigorous, and it died of measurement. Ptolemy’s reason for a stationary Earth, and far more sharply Tycho’s, was that no stellar displacement is seen. Tycho’s version is the one that deserves respect: stars appeared to have angular diameters, so a star distant enough to hide its parallax would have to be physically vast — larger than the whole orbit of the Earth, on Graney’s account of Locher; Ingoli, following Tycho, has the fixed stars “surpass or equal the size of the orbit circle of the Earth itself.” Valid reasoning, absurd conclusion, and therefore — correctly, by the standards of the time — a reason to reject the premise. The premise that was false was the observation. Stars have no visible discs. What Tycho measured with the eye and what Galileo, Hevelius and Cassini measured with early telescopes (roughly 4–6 arcsec for Sirius and for “an average fixed star”) was the instrument’s point-spread function: diffraction, plus seeing, plus the eye’s own optics. Airy published the theory in 1835 — the image of a star “will not be a point but a bright circle surrounded by a series of bright rings” — and Graney and Grayson trace how slowly even professional astronomers absorbed it. True stellar diameters are three orders of magnitude smaller than the eye can resolve: the largest ever measured, R Doradus, is 0.057 arcsec against the eye’s roughly 60 arcsec. Item 8, “Constant stellar angular sizes”, is therefore false twice over — the naked-eye sizes are not the stars, and the real sizes are not constant, running from under a milliarcsecond to 57 mas across the interferometric catalogues.

4. What parallax actually is, and why it discriminates. This is the part the cluster never engages. Parallax is not “a small wobble astronomers report.” It is a geometric consequence of the observer changing position, and a moving Earth predicts four specific things about it, all of which are observed:

• It exists at all, and its size is set by the base line — one astronomical unit by convention, the full orbital diameter over six months.
• Its period is exactly one year, and its phase is locked to the Earth’s position in orbit. Nothing in a stationary-Earth account explains why the displacement should return to itself annually rather than daily, or on any other period.
• The parallax ellipse is oriented: its major axis lies parallel to the ecliptic and its shape depends on the star’s ecliptic latitude, a star at the ecliptic pole tracing a circle and one on the ecliptic tracing a line. That is the shape of the Earth’s orbit projected onto the sky, and it is measured.
• Its amplitude goes as 1/distance, and the distances are checkable independently — by VLBI radio astrometry, by eclipsing binaries, by cluster main-sequence fitting, by Cepheids. They agree.

Modern astrometry does not measure a displacement and call it parallax. Gaia fits five parameters at once — two positions, two proper-motion components and one parallax — to years of observations, precisely so that the annual periodic term is separated from the secular linear one. Gaia DR3 publishes that solution for 1,467,744,818 sources out of 1,811,709,771 total, with median parallax uncertainties of 0.02–0.03 mas for G < 15, 0.07 mas at G = 17 and 0.5 mas at G = 20. 61 Cygni, Bessel’s star, now reads 285.9949 ± 0.0599 mas (A) and 286.0054 ± 0.0289 mas (B): 11.404 light years. Bessel, with a heliometer in 1838, got 0.3136″ — ten per cent high and unmistakably the same quantity. Proxima Centauri reads 768.0665 ± 0.0499 mas, 4.2465 light years.

5. Item 103, “Inconsistent stellar parallax”, points at something real. Answer the real version. Parallax measurements have disagreed, sometimes publicly and embarrassingly. Struve doubled his own Vega value between 1837 and 1840. And the best modern case is the Pleiades: Hipparcos returned about 115–120 pc against a long-standing ~130 pc from every other method, a discrepancy serious enough to call stellar-evolution models into question. It was not waved away. Melis and colleagues settled it in 2014 by VLBI — a completely different instrument and technique — at 136.2 ± 1.2 pc, and Gaia DR2 then gave 136.67 ± 0.04 pc from a weighted mean of 1,595 stars. Hipparcos was the outlier, for reasons internal to its data reduction. Gaia’s own systematics are published in the same spirit: Lindegren and colleagues report that quasar parallaxes, which should average zero, are offset by “a few tens of microarcsec”, varying with magnitude, colour and ecliptic latitude, and they publish the correction. That is the difference between the two enterprises. The inconsistencies are real, they are found by the people who made the measurements, they are quantified, and they are two to four orders of magnitude smaller than the parallaxes themselves. An inconsistency of tens of microarcseconds does not license the conclusion that a 768,000-microarcsecond parallax is imaginary.

6. Item 240, “Stellar proper motion tiny”, is a different phenomenon and it inverts. Proper motion is the star’s own secular drift across the sky; parallax is the annual periodic loop caused by our motion. They are separately fitted parameters and confusing them is a category error. Nor is proper motion zero, or uniform. Halley detected it in 1718 by comparing Sirius, Arcturus and Aldebaran with the ancient catalogues — which by itself retires any rigid firmament, since a rigid sphere cannot have its lights drift relative to one another. Barnard’s Star moves 10.3 arcsec per year, so the constellation it sits in visibly deforms within a human lifetime. Gaia DR3 publishes proper motions for the same 1.468 billion sources at 0.02–0.03 mas/yr for G < 15. And the pattern is the one the distance model predicts and the dome model has no reason to expect: the stars with the largest proper motions are the nearest ones. Barnard’s Star, with the largest known proper motion, has a parallax of 546.9759 ± 0.0401 mas — 5.96 light years. Proxima, the nearest star at 768.07 mas, moves 3.85 arcsec per year. Convert through the distance and the velocities come out ordinary: Barnard’s Star crosses the sky at about 89 km/s, a perfectly normal speed for a star in the galactic disc. On the same numbers a firmament a few thousand miles up requires those two stars to be creeping at centimetres per second while the rest of the sky holds rigid formation, and requires the correlation between apparent drift and apparent annual loop to be a coincidence. It is not a coincidence; it is the same distance appearing in both measurements.

7. What the cluster is, as a document. Six items, of which two (37 and 264) are the identical sentence — “No naked-eye parallax.” — entered twice, 227 apart. One of the three exact duplicate pairs in the whole 461-item specimen falls inside this one cluster of six. Two more items (5 and 37/264) are true statements with their scope intact, sitting in a numbered list of evidence for a stationary Earth. Two (8 and 240) have no counterpart in either text we searched — the 1865 first book edition and the 1881 third edition’s ch. III. One (103) is a fair compression of a real argument. That distribution is not what a reading of Rowbotham produces; it is what paraphrase-expansion of a handful of stock sentences produces, and the duplicate is the fingerprint.

Verdict: refuted. Parallax is measured, at 0.02 mas precision, for 1.468 billion stars, with the annual period, the ecliptic-aligned ellipse and the inverse-distance amplitude that only a moving Earth predicts. The source’s own chapter contains the measurements the list says do not exist, and the one claim that would defeat them — equal parallax from a three-foot base line — is wrong by eleven orders of magnitude, and the experiment offered for it measures the alignment of his own tubes.

The list says

Lack of stellar parallax in early measurements. / Constant stellar angular sizes. / No naked-eye parallax. / Inconsistent stellar parallax. / Stellar proper motion tiny. / No naked-eye parallax. — items 5, 8, 37, 103, 240, 264; 37 and 264 are the same sentence entered twice.

The source says

“The parallax thus assigned α Centauri, is so very nearly a whole second in amount (0″.98), that we may speak of it as such. … ‘Professor Bessel made the parallax of a star in the constellation Cygnus to be 0″.35.’” … “it is only necessary to state as an absolute truth the result of actual experiment, that, a given fixed star will, when observed from the two ends of a base line of not more than three feet, give a parallax equal to that which it is said is observed only from the two extremities of the earth’s orbit.”

The qualifier was dropped.

The dominant drift is a qualifier doing no work. Items 5, 37 and 264 keep their scope — early measurements, naked-eye parallax — and as written are simply true; what changes is that a scoped historical observation is entered as a numbered proof of a stationary Earth, where the reader supplies the generalisation the sentence withholds. Note what the list does not do: no item in it denies that parallax is detected. That denial arrives downstream, in Dubay’s proof 19 — “not a single inch of parallax can be detected in the stars, proving we have not moved at all” — and that is a straight reversal of the source, which quotes two values to two decimal places, 0″.98 and 0″.35, and argues about their cause: a three-foot base line rather than a 186-million-mile one. That is a falsifiable claim about base-line scaling, and it is a different and stronger claim than “no parallax exists.” The reversal also shaped this cluster’s label and the drift type this block used to carry, both of which read the downstream slogan back onto the six items; the drift type is corrected above. A summary that overstates the items it summarises is the same fault we are documenting.

The six items, one by one, because they are not all the same type.

• Items 5, 37 and 264 — hedge_dropped shading into force_upgraded. “Lack of stellar parallax in early measurements” and “No naked-eye parallax” keep their qualifiers and are, as written, simply true. What changes is the speech act: a scoped historical observation appears in a numbered list of proofs of a stationary Earth, where the qualifier does no work and the reader supplies the missing generalisation. Nothing was misquoted; a true sentence was promoted to a proof.
• Items 8 and 240 — unsourced_addition. Neither is located in the 1865 first book edition (Project Gutenberg #69892) or in ch. III of the 1881 third edition, the two texts searched. “Constant stellar angular sizes” is Tycho’s star-size objection arriving without Tycho and without the diffraction that dissolved it; “Stellar proper motion tiny” imports a phenomenon that appears in neither of those two texts and, on inspection, one that argues against him. Rowbotham’s cosmos puts the stars a few thousand miles up; nothing in it explains why nearby stars by parallax should also be the fast-moving ones by proper motion.
• Item 103 — the one faithful compression. “Inconsistent stellar parallax” is a fair summary of Rowbotham’s actual recital of Bradley, Brinkley and the illusory arcsecond, and it points at a real feature of the record. It is answered on the merits in the refutation, not dismissed.

And one structural note. Items 37 and 264 are the identical sentence, entered twice, 227 items apart — one of only three exact duplicate pairs in the entire 461-item specimen, and it falls inside a cluster of six. Six items on one topic, of which two are the same sentence, two have no source, one is true and inert and one is faithful. That is the shape of a list assembled by fanning paraphrases out of a few stock lines, not by reading Rowbotham.

the same ether that caused the 1925 Michelson-Gale experiment to measure an ether-drift of a 24-hour period … is the same ether that causes a Foucault Pendulum at the North Pole to rotate 360 in a 24-hour period

— Galileo Was Wrong: The Church Was Right (2006), Vol. I, p. 168 — as cited by Sungenis himself in the Palm–Sungenis debate document of July 2018 (Internet Archive item RobertSungenisVsDavidPalmDebateOnGeocentrism), where the sentence is footnoted “GWWi, p. 168”. Not checked against page images of Vol. I; given the 2013 rearrangement into three volumes, the volume label needs confirmation independent of that footnote

Two reassignments run side by side in this material, and it is worth separating them before answering either, because they are different physical claims and only one of them is defensible.

The medium version is the sentence above: the pendulum turns because a rotating ether turns it, the same ether Michelson and Gale are said to have detected in 1925. This is the ancestor of the “electromagnetic” items in the cluster — a physical substance, circling the Earth once a day, pushing things about.

The frame version is the one that carries the argument, and Sungenis states it plainly in the 2018 debate document: “in 1918, Hans Thirring was the first to show how this would work. He showed, by pure mathematics, how a rotating universe would affect pendulums on Earth, satellites above the Earth, winds on Earth…” — and then, in the sentence this page's compression block turns on, that on Thirring's result those things “would behave very much like we see them behave, but not exactly.” He goes on to say the resulting centrifugal and Coriolis terms are “real forces and not fictitious”, and are “understood as gravitational forces in the geocentric system.”

Note who is speaking and what he is not saying. Sungenis and Bennett argue for a stationary spherical Earth at the centre of a rotating cosmos, not for a disc; Sungenis published a book-length attack on flat-earth cosmology, Flat Earth, Flat Wrong, in 2018. The six list items borrow his answer to Coriolis into a model he wrote a book against — and, as the refutation below works out, the answer does not survive the transfer, because the whole reason a rotating frame can reverse the deflection at the equator is that the Earth is round.

SURFACE (weak — do not use). “Coriolis is a fictitious force, so there is nothing to reassign,” or “a rotating universe is not a thing.” Both lose the exchange. “Fictitious” is a term of art meaning frame-dependent, not unreal: the deflection sinks ships, steers hurricanes and is coded into every operational weather model. And rotating cosmological solutions of the field equations exist; Gödel wrote one down in 1949.

DEEPER. Mach's principle is a genuine strand in the history of general relativity rather than a fringe borrowing, Einstein named it among the theory's motivations, and frame dragging is real and measured — Gravity Probe B returned a Lense–Thirring drift of −37.2 ± 7.2 mas/yr against a predicted −39.2. True, and incomplete, because it invites the reply that measured frame dragging is ten orders of magnitude too small to be the diurnal rotation.

KERNEL, first form: the rotating-shell result is stronger than the geocentrist usually states it. Thirring showed in 1918 that inside a slowly rotating mass shell the inertial frames are dragged, and that the interior acquires terms of Coriolis and centrifugal form. His solution was not clean — it carried a spurious extra term, and Sungenis's own “but not exactly” is an accurate report of it. The literature then fixed it: Pfister and Braun showed in 1985, in a paper titled Induction of correct centrifugal force in a rotating mass shell (Class. Quantum Grav. 2:909–918), that with the shell's interior stresses treated properly the correct centrifugal force is induced inside. Brill and Cohen had shown in 1966 (Phys. Rev. 143:1011) that as a shell's radius approaches its own Schwarzschild radius the dragging becomes complete — the interior frames rotate with the shell. Our universe's mass and radius put it in the neighbourhood of that limit. So “the turning sky turns the pendulum” is not crankery. It is a licensed reading of a theory nobody in this dispute rejects at that point.

KERNEL, second form — and this one is for the electromagnetic branch, which deserves better than the mockery it usually gets. Sibling argument A10 establishes the decisive fact: the deflection reverses sign across the equator, and a thermal driver such as solar heating has no handedness from which a reversal could come. Now ask what else in ordinary physics does. There is exactly one common candidate: the magnetic Lorentz force, qv×B, which is chiral by construction — and the vertical component of the geomagnetic field flips sign at the magnetic dip equator, close to the geographic one. Whoever first said “ocean currents are electromagnetic” had picked the only everyday force with the right qualitative signature: handed, and handed the other way south of a line near the equator. That is a better instinct than it looks.

Why it doesn’t save the claim.

Against the frame version: reproducing a term by construction is not evidence for the model that reproduces it. This is the whole of it. If the rotating-universe description contains the same 2Ω×v with the same coefficient — and that is what its defenders are claiming, correctly — then every measurement of the Coriolis deflection comes out identically in both descriptions, and no such measurement can favour either. The six items in this cluster are offered as evidence against a spinning Earth. Their own best defence makes them evidence for nothing. That is a real cost and it is paid in this cluster's own currency: on the geocentrist's account, items 6, 106 and 292 stop being arguments and become bookkeeping.

And the reproduction is exact only in a limit that has a price. The clean interior result needs a shell with the right internal stresses, and completeness of dragging needs the shell at its gravitational radius. A model in which dragging is anything less than complete gives a Coriolis parameter reduced by a factor below one — and the measured parameter matches complete dragging to the precision of ring-laser gyroscopy, better than one part in 10⁹. The rotating-Earth account predicts that number with nothing to adjust. The rotating-universe account has to be tuned to the limit in which it agrees.

Against the electromagnetic version: right instinct, wrong line and wrong size by seven orders of magnitude. The line first. The magnetic dip equator is not the geographic equator, and it migrates — Yizengaw's survey records the dip equator moving north over Brazil at about 0.2° of latitude per year, some 3.85° in twenty years. The line where cyclonic organisation fails does not travel with it; NOAA's hurricane researchers state the exclusion zone as a distance from the equator, about 300 miles or five degrees, and it has stayed there. Then the size, with the inputs on the table so a reader can redo it: seawater conductivity 3.2 S/m, geomagnetic field 5×10⁻⁵ T, current speed 0.1 m/s give an induced current density σvB ≈ 1.6×10⁻⁵ A/m² and a Lorentz force density jB ≈ 8×10⁻¹⁰ N/m³, which on 1025 kg/m³ of seawater is an acceleration near 8×10⁻¹³ m/s². The Coriolis acceleration on the same water is 2Ωv sinφ ≈ 1.0×10⁻⁵ m/s². The proposed cause is about ten million times too weak.

Against both, and fatally for this list in particular: the reassignment is a globe-shaped answer. The reversal exists because on a sphere only the component of the rotation vector along the local vertical enters the horizontal deflection, and that component changes sign at the equator. Put the rotation axis through the centre of a flat disc, as a rotating dome requires, and the axis is parallel to the local vertical at every point on the disc: the Coriolis parameter is 2Ω everywhere, one sign everywhere, no equatorial calm belt and no mirrored trade winds. Sungenis's answer works only in Sungenis's cosmos.

First, the concession, and it has to be complete. In a frame rotating with angular velocity Ω relative to the local inertial frames, the equation of motion acquires the terms −2Ω×v, −Ω×(Ω×r) and −(dΩ/dt)×r. Nothing in that derivation says which body rotates. It is a statement about relative rotation, and it was so from the beginning: Coriolis's 1835 paper, Sur les équations du mouvement relatif des systèmes de corps (Journal de l'École Polytechnique XV, pp. 142–154), was about the relative motion of systems of bodies, and the Earth was not his subject. Bibnum's study of the paper notes that meteorology “is an area that was without a doubt even more foreign to the concerns of Coriolis” than the celestial mechanics that Foucault later drew out of it; the application to winds and currents is William Ferrel's, from the late 1850s, in the form — already the crux of this page — that the deflection “always deflects it to the right in the northern hemisphere.”

The two numbers, which this page got wrong once and now keeps apart. The Coriolis parameter is f = 2Ω sinφ. The precession rate of a Foucault pendulum is Ω sinφ — half of it — which is why the plane of swing takes a full sidereal day, 23 h 56 m, to come round at the pole and about 32 hours at the latitude of Paris. Earth's rotation rate Ω is 7.292×10⁻⁵ rad/s, or 15.04° per hour. A treatment that runs the two together will be caught by anyone who checks, and the factor of two is not decorative: it is the difference between the deflection of a moving parcel and the rotation of a plane.

Second, what general relativity actually grants, stated at full strength. Hans Thirring showed in 1918 that inside a rotating mass shell the inertial frames are dragged, and that the interior picks up terms with the form of the centrifugal and Coriolis forces. That result was imperfect — the interior also acquired a term with no Newtonian counterpart — and it was repaired, not by geocentrists and not by their critics, but by the relativity literature: Pfister and Braun, Induction of correct centrifugal force in a rotating mass shell, Classical and Quantum Gravity 2(6):909–918 (1985). Brill and Cohen, Rotating masses and their effect on inertial frames, Physical Review 143:1011 (1966), had already shown that dragging becomes complete for a shell near its own Schwarzschild radius. Put those together and the geocentrist's claim — that a universe rotating about a stationary Earth would produce the Coriolis and centrifugal effects we observe — is not a misunderstanding of physics. It is a reading physics supports. This is why the verdict on A09 is STANDARD PHYSICS and not REFUTED, and the page should not pretend otherwise.

Third, why that grant is worth nothing to the argument. Because it is symmetric, and the items are not. Item 106 is listed as a reason to think the Earth does not spin. But if the two descriptions contain the same term with the same coefficient, then the observation of that term discriminates between them exactly as well as a thermometer discriminates between Celsius and Fahrenheit. The strongest form of this argument is a proof that the Coriolis evidence is neutral. Neutral evidence is not evidence for the stationary side. Sungenis is careful here in a way the list is not: he does not present Thirring as a measurement, he presents him as a demonstration that the geocentric system is “the more complete system” because it needs no forces labelled fictitious. That is an argument about elegance and vocabulary. It is not an observation, and it was never offered as one.

Fourth, the vocabulary itself, because a straw man is buried in it. Calling the Coriolis term “fictitious” has never meant that nothing happens. It means the term is frame-dependent: it appears in the equations written in a rotating frame and vanishes in the equations written in an inertial one, while the physical facts — where the shell lands, which way the storm turns, how many fringes the ring laser counts — are the same either way. Sungenis's version, that these are “real forces and not fictitious” in the geocentric system, is a claim about which chart to privilege, dressed as a claim about what exists. Nobody needs to contest it to answer the argument, and contesting it is how the mainstream side usually loses this exchange.

Fifth, what the physical reading costs — and this is where the two readings part company. Following R01: (a) writing the standard solar system in Earth-fixed rotating coordinates is a change of chart, free, and carries no commitment; (b) asserting that the cosmos physically turns about a stationary Earth is a different model that owes a dynamics. On reading (b) the rotating sky inherits every irregularity of the Earth's rotation, because the two are the same relative motion measured one way or the other. So:

• The cosmos must wobble on a 433-day period. The Chandler wobble — period 433.0 ± 1.1 days — is the free nutation of a deformable spinning body, and Gross's analysis of 1985–1996 data attributes its excitation chiefly to ocean-bottom pressure fluctuations (3.45 mas² of power in the Chandler band, against 1.87 for atmospheric pressure). Under (b) the rotation axis of the universe is being nudged by pressure changes on the sea floor.
• The cosmos must speed up and slow down with the monsoon. Length of day varies by milliseconds, and EarthScope's summary puts about 90 per cent of the seasonal variation down to shifts in zonal wind patterns, with El Niño years running long and La Niña years short. A millisecond on 86,400 s is about one part in 10⁸; under (b) the whole rotating cosmos changes its rate by that fraction, twice a year, in step with the atmosphere's angular momentum, and secularly by the 2.3 ms per century that tidal friction adds.

Be exact about what this does and does not show. It is not a contradiction on reading (a), where these are coordinate descriptions of the Earth's own wobbles and the arithmetic is the same arithmetic. It is a bill on reading (b), and reading (b) is the one Sungenis takes: in the same passage he credits Thirring's rotating universe with showing how “its own nutation and precession would affect what we see on Earth.” Once the sky's nutation is doing explanatory work, the sky owes an account of why its nutation is excited by seawater.

Sixth, the electromagnetic branch, on its own terms. Items 56 and 113 replace the inertial deflection with an electromagnetic one. The instinct is sound — the Lorentz force is the one everyday force with a built-in handedness that flips near the equator — and the mechanism is real: moving seawater does generate electric currents in the geomagnetic field, and the resulting magnetic signals are measured, at roughly 1–2 nT at satellite altitude for the great current systems. What fails is the back-reaction. The force those currents exert on the water is around 8×10⁻¹³ m/s² on the numbers set out in the steelman, against 1.0×10⁻⁵ m/s² for Coriolis on the same parcel: seven orders of magnitude. And there is a cleaner test than arithmetic. An electromagnetic cause must scale with conductivity. Air at the surface conducts about 10⁻¹⁴ S/m, some fourteen orders of magnitude below seawater — yet the atmosphere is deflected with the same handedness, the same sign reversal and the same 2Ωv sinφ magnitude as the ocean. A Foucault bob of copper, brass or lead precesses at the same Ω sinφ. Material-independent and mass-proportional is the signature of an inertial term; conductivity-dependent is the signature of an electromagnetic one; the measurements say the first.

Seventh, the drones. Item 214 says there is no Coriolis correction in drone flight software. Take the factual core as given — we have not read hobby flight-controller source and are not going to assert anything about what is in it — because the inference fails whatever the code contains. Start with the size: for a drone at 10 m/s at 45° latitude the Coriolis acceleration is 2Ωv sinφ ≈ 1.0×10⁻³ m/s², about one ten-thousandth of gravity, which a quadcopter cancels with a tilt of six thousandths of a degree. Left entirely uncorrected it would move the aircraft about 5 m off track over a 100-second leg — smaller than the drift from a light breeze, and nulled by the same GPS-and-barometer feedback loop that nulls the breeze. A closed loop does not need a model of a disturbance it is already measuring. Now look at the systems that cannot close the loop, because they navigate with no external reference: strapdown inertial navigators carry Earth-rate and transport-rate terms explicitly in their mechanisation equations, and a gyrocompass finds true north by sensing Earth's rotation and nothing else — iMAR's engineering guide sets the threshold plainly, gyro drift below about 0.1°/h against Earth rate of 15.05°/h, and quotes the heading a 0.015°/h gyro can reach from that ratio. Ships have been steering by that principle since the early twentieth century. So the honest version of item 214 is: where the effect is below the noise and the loop is closed, nobody codes it; where it is above the noise or the loop is open, everybody does — which is what a real 2Ωv term predicts, and what a non-existent one does not. Compare A14 on artillery and A19 on gyrocompasses, where the same trade appears.

Eighth, the cluster contradicts itself, and the contradiction is not ours to resolve. Items 6, 106 and 292 say the Coriolis effects are real, are exactly as observed, and are produced by the turning heavens. Items 56 and 113 say they are produced by electromagnetism instead; item 214 says they are not produced at all. These cannot all be held. On the geocentric account the deflection is a genuine consequence of relative rotation and its magnitude is 2Ωv sinφ — so Sungenis's items refute the electromagnetic items, and both refute the drone item. This is the same structure R01 found on covariance: the list carries an argument and its own answer side by side, and a defender has to choose which one to keep.

Verdict: standard physics. The frame version of this argument is a legitimate restatement inside a theory nobody here disputes, it produces the observed Coriolis terms because it was built to, and for exactly that reason it produces no observation the rotating-Earth model does not already produce. It is not evidence; on its own showing it cannot be. What it does have is a price list — a cosmos whose spin rate tracks the seasonal winds and whose axis is nudged by pressure on the sea floor — and a shape requirement its borrowers do not meet, since the equatorial reversal that makes the whole phenomenon interesting exists only because the local vertical changes its angle to the rotation axis, which is a thing that happens on a sphere and not on a disc.

The list says

Coriolis equivalence with sky rotation.

The source says

“Thirring discovered that pendulums, satellites and winds would behave very much like we see them behave, but not exactly.”

The qualifier was dropped. The item asserts equivalence. The source, in Sungenis's own words in the July 2018 Palm–Sungenis debate document, asserts near-equivalence and says so twice in one sentence: very much like, and but not exactly. That qualification is not modesty, it is accurate physics history — Thirring's 1918 interior solution carries a term with no Newtonian counterpart, and the coefficients are not those of a rotating frame. What makes this case unusual is that the hedge became droppable after the fact: Pfister and Braun's 1985 paper, titled Induction of correct centrifugal force in a rotating mass shell, showed that a shell with its interior stresses handled properly induces the correct centrifugal force, and Brill and Cohen's 1966 result gives complete dragging in the limit of a shell at its gravitational radius. So the flat list item is closer to today's literature than the hedged source is — and it is still a worse argument, because the exactness is bought by moving to a solution that requires tuned interior stresses and a shell at its gravitational radius, which is a larger commitment, not a smaller one. The refutation above is aimed at the strong modern version, not at Thirring's 1918 approximation, which would be the easy target and the wrong one.

A second pattern, which the single-value enum cannot carry. Six items sit in this cluster and they do not share a lineage. Items 6, 106 and 292 are the reassignment as Sungenis argues it. Items 56 and 113 (“electromagnetic”) and 214 (drones) assert something his argument needs to be false, and we could not locate them in the Sungenis text we were able to search — which was limited: both archive renderings truncate, the Vol. I text stopping inside chapter 1 at printed p. 22 and the Vol. II scan at the chapter-10 contents, so this is a statement about our search and not about the books. Item 214 also has a dating problem independent of any search: consumer camera drones postdate a 2006 publication. Recorded here rather than acted on; the cluster record was left untouched.

Proposed instead: STANDARD PHYSICS for items 6, 106 and 292; REFUTED for items 56 and 113; MISLEADING for item 214

The cluster name joins two reassignments with an 'or', and they do not share a verdict. Reassigning the Coriolis terms to a rotating universe is standard physics: general relativity supports it, it reproduces the observed terms by construction, and the objection to it is that it is not evidence rather than that it is wrong. Reassigning them to electromagnetism is a different claim with a different fate — it is quantitatively dead, about seven orders of magnitude short, and it predicts a sign-reversal line on the migrating magnetic dip equator rather than on the geographic one. Item 214 is neither: its factual core may well be true and its inference does not follow, which is what MISLEADING is for on this page. Holding all six under STANDARD PHYSICS reads as though the page had blessed the electromagnetic claim, and it also hides the more interesting finding, which is that the two halves of the cluster contradict each other. This is a proposal only: clusters.py was not edited, and whether the answer is a verdict change or a cluster split is an operator's call.

The verdict on the scorecard is unchanged until the question is settled. Publishing the disagreement is the point: a rubric that never produces dissent is not being applied.

Michelson-Gale detected the ether moving past the Earth's surface at 2% of the rotation speed. While the Michelson-Morley experiment detected no heliocentric movement, the Michelson-Gale experiment measured either the effect of the Earth's rotation or the ether's rotation around the Earth.

— Galileo Was Wrong: The Church Was Right (2006), Vol. II, ch. 10 (“Technical and Summary Analysis of Geocentrism”), section “Michelson-Gale: Claims and Responses”, pp. 189–192, at p. 191 — 7th ed., 2013

These are Robert Sungenis and Robert Bennett's words, from the summary of their Michelson–Gale section in Galileo Was Wrong: The Church Was Right, Vol. II (7th ed., 2013), the volume carrying chapters 7–13. The book is the direct source of the Michelson–Gale and Sagnac material circulating in modern proof-lists, generally via Walter van der Kamp, who reframed the interferometry results for the Tychonian Society before Sungenis systematised them.

Two features are worth noting before anything else. First, the authors do not misreport the experiment. They state plainly on the same page that “unlike the Michelson-Morley experiment, the Michelson-Gale experiment did not produce null results,” and that the displacement “was closely related to the rotational velocity of the Earth.” Second, the sentence quoted here is a disjunction, not an assertion: the experiment measured “either the effect of the Earth's rotation or the ether's rotation around the Earth.” The argument is about frame ambiguity, not about the instrument reading zero.

What the cited paper actually reports: Michelson, Gale and Pearson, “The Effect of the Earth's Rotation on the Velocity of Light, Part II,” Astrophysical Journal 61 (1925) 140–145, ran a light path around a rectangle 2010 ft east–west by 1113 ft north–south at latitude 41°46′. They computed the displacement a rotating Earth should produce as 0.236 ± 0.002 fringes and observed 0.230 ± 0.005, concluding that “the observed and calculated shifts agree within the limits of observational error.” The predicted figure is not a free parameter: 4AΩsinφ/λc, with the known sidereal rate Ω = 7.292 × 10⁻⁵ rad/s and λ ≈ 570 nm, returns 0.236. The number the apparatus was checked against was the Earth's rotation rate, and it came back.

The weak version of the criticism — that they simply misread the paper — is not available here. They report the positive result, they report its magnitude, and the geocentrist Malcolm Bowden puts the concession in print in a mainstream creationist journal: “The 0.46 km/s speed of the rotation of the Earth (at the equator) relative to the aether was measured to within 2% of the known speed” (Journal of Creation 16(2), 2002, p. 80). Any rebuttal opening with an accusation that they ignored the fringe shift is answering a claim nobody made.

A better version — that they conflate rotation with orbital motion — is true of some downstream list items but not of the source. Sungenis and Bennett's own “Claims and Responses” section shows they know precisely what quantity was detected; their reply is that “it is only the relative rotation between Earth and cosmos that was detected.” That is a considered position, not a confusion.

The strongest form is this, and it should be conceded fully. There is a genuine and deep asymmetry in physics between uniform velocity and rotation, and the geocentrist argument has its finger on it. No local experiment can detect uniform translational motion; that is the relativity principle itself, and Michelson–Morley's null is its evidence. Rotation is different. A Sagnac interferometer, a Foucault pendulum, a mechanical gyroscope and a ring laser all return a frame-independent answer to the question “are you turning?” Rotation is absolute in a way velocity is not, in Newtonian mechanics (Newton's bucket) and in relativity alike. This was genuinely contentious: Georges Sagnac presented his 1913 result as evidence for a luminiferous medium and against the emerging relativity, and the “Sagnac paradox” required careful work to resolve. The resolution — that a rotating frame is non-inertial and admits no global synchronisation, so the Sagnac term is a geometric holonomy of the metric rather than a violation of c — is correct but was not obvious in advance. They are pointing at a place where the physics really does treat rotation as a thing in itself.

Why it doesn’t save the claim.

The asymmetry is real, and it runs the wrong way for the claim.

Consider the counterfactual. If rotation were as undetectable as uniform velocity, then “the Earth does not rotate” would be a permanently safe coordinate choice, immune to any experiment. That is precisely the shelter the general-covariance and no-absolute-motion arguments elsewhere on this list try to build. But the Michelson–Gale argument abandons that shelter on purpose. It stakes its case on rotation being physically detectable by a local instrument. Once that is granted, the instrument is entitled to give an answer, and the answer is a number: Ωsinφ at latitude 41°46′, which is the Earth's sidereal rate. You cannot borrow the authority of an absolute-rotation measurement and then decline the reading.

The remaining move, taken explicitly in the source, is to say the absolute rotation belongs to the cosmos rather than to us. Note what that costs. It converts the item from an experimental result into a relabelling that leaves every measured quantity untouched. Nothing in the fringe count changes; nothing new is predicted; no observation is forbidden. It is no longer a proof that the Earth is stationary — and by their own hedged phrasing the authors do not present it as one. The list items rendering this as “Michelson–Gale ether drift detection” and “Sagnac effect consistent with stationary Earth” are firmer than the source they descend from.

And the relabelling does not generalise. A rotating optical medium is a hypothesis about light. It has no purchase on a swinging pendulum, a deflected air mass, or a spinning mechanical gyroscope, none of which couple to an optical medium at all. To cover those you must promote the claim to “the local inertial frame rotates relative to the ground at 15.041° per hour” — which is what “the Earth rotates” means operationally. At that point the disagreement is about naming, not about the world, and Gerardus Bouw, the movement's only credentialed astronomer, had already reached that conclusion in Geocentricity (1992).

The experiment is a positive detection, and that is why the list wants it. The interferometry the list leans on elsewhere is null results — Michelson–Morley, Airy's failure — and Michelson–Gale–Pearson 1925 is cited precisely because the apparatus registered a real signal instead. That signal is the Earth's rotation. They laid a closed light path around a rectangle 2010 ft east–west by 1113 ft north–south near Clearing, Illinois, at latitude 41°46′, calculated in advance that a rotating Earth would displace the fringes by 0.236 ± 0.002, and observed 0.230 ± 0.005. The instrument recovered the known rotation rate to within a couple of per cent, which is why Bowden, writing as a geocentrist, states in print that the figure “was measured to within 2% of the known speed.”

What the Sagnac effect is, and why it is not an ether detector. Send two beams in opposite directions around a closed loop of enclosed area A. If the loop rotates with angular velocity Ω, the beams do not return together: Δt = 4(A·Ω)/c², giving a phase shift proportional to the enclosed area and the rotation rate. The quantity in the formula is the rotation of the apparatus with respect to the local inertial frame. No medium appears in it. The result follows from special relativity, from general relativity, and from a flat-space time-of-flight argument, without any luminiferous medium being assumed. Sungenis and Bennett derive it on the same page and note, accurately, that this result “is obtained without explicitly invoking an ether, Lorentz transformations or General Relativity.” That is the whole difficulty for the argument: an equation with no ether in it cannot be evidence for an ether. Nor does the effect contradict the invariance of c, which is a statement about local inertial frames; a rotating frame is not one, cannot be globally synchronised, and the Sagnac term is the measure of exactly that failure.

Scope limit, stated plainly. Because the effect depends on A·Ω, Michelson–Gale is sensitive to rotation and to nothing else. Uniform translation of the whole apparatus produces no first-order Sagnac shift. This experiment therefore says nothing whatever about whether the Earth orbits the Sun, and should not be presented as refuting geocentrism entire. What it refutes is the proposition that the Earth does not turn on its axis — one of the two things modern geocentrism asserts, and the one this item was recruited to support. The orbital half is settled by other evidence entirely: stellar parallax, aberration, radial-velocity and timing data.

The escape, and its real strength. The geocentrist reply is that the medium, or the whole cosmos tied to it, rotates around a stationary Earth once per sidereal day, leaving the relative rotation — and therefore the fringe shift — unchanged. Taken against Michelson–Gale alone, this is unfalsifiable, and it is worth saying so rather than pretending otherwise. An interferometer measures a relative angular velocity and does not carry a label saying which body owns it.

Why it closes in combination. The escape is tailored to an optical experiment, and the same rotation rate is returned by instruments with no optical content at all. A Foucault pendulum is a mass on a wire. Coriolis deflection governs the circulation of the atmosphere and oceans. A gyrocompass is a spinning flywheel. Each independently yields Ω = 7.292 × 10⁻⁵ rad/s, 15.041° per hour sidereal — the value the light ring gave. A hypothesis about a rotating optical medium has no mechanism by which to deflect a pendulum bob or a hurricane. Broadening it until it does means asserting that the local inertial frame itself turns relative to the ground at the sidereal rate, and that assertion is not distinguishable from the ordinary account; it is the ordinary account with the labels exchanged. Note also that the rotating shell in such a model must be tuned to reproduce Ω exactly, making it a fit to the measurement rather than a prediction of it.

One of their stated evasions has since expired. The book's second response to “the experiment shows the Earth is rotating” is that the 1925 precision “could not distinguish a 24-hour solar period from a period 4 minutes shorter.” That was a fair point about 1925 instrumentation. It is no longer available. Large ring lasers are the same physics with six orders of magnitude more resolution: the G ring laser at Wettzell tracks Earth's rotation rate below one part in 10⁹, resolving not merely sidereal versus solar but sub-millisecond length-of-day variations and polar motion — agreeing with independent VLBI determinations of the same wobbles. The four-minute question the source treats as beyond reach is now resolved routinely, by an instrument whose operating equation is the Sagnac formula.

GPS runs on the Earth's rotation being real. The Sagnac correction in GPS is computed in an Earth-Centred Inertial frame for a structural reason. As Neil Ashby sets out in Relativity in the Global Positioning System, synchronisation “is performed in the local inertial frame, or ECI coordinate system,” because attempts to synchronise within the rotating Earth-fixed frame are not transitive: “path-dependent discrepancies in the rotating frame are thus inescapable.” Carrying a synchronisation eastward around the equator and back accumulates a 207.4 ns discrepancy against the starting clock. That number is set by the Earth's rotation rate and the enclosed area; on a non-rotating Earth it would be zero.

The movement's own instruments agree. In Behind the Curve (2018), Bob Knodel described putting a ring-laser gyroscope to the question: “when we turned on that gyroscope we found that we were picking up a drift. A 15 degree per hour drift.” That is 360° divided by 24 hours. His response was to look for ways to shield or explain the reading rather than accept it. The instrument was a Sagnac interferometer, the same class of device as Michelson and Gale's, and it gave the same answer at the first attempt.

The list says

Michelson–Gale ether drift detection. / Sagnac effect consistent with stationary Earth. / Sagnac proves apparatus rotation.

The source says

“Michelson-Gale detected the ether moving past the Earth’s surface at 2% of the rotation speed. … the Michelson-Gale experiment measured either the effect of the Earth’s rotation or the ether’s rotation around the Earth.”

A concession was re-used as a proof. Sungenis and Bennett do not misreport the experiment. They say on the same pages that Michelson–Gale “did not produce null results” and that the displacement was closely related to the Earth’s rotational velocity, and their summary sentence is a disjunction: the apparatus measured either the Earth’s rotation or the ether’s rotation around it. That is a statement about what an interferometer cannot decide. It is not a claim that the Earth is at rest, and the section heading in their own text — Sidereal Relative Rotation — says as much.

The list keeps the caution and changes the job. “Sagnac effect consistent with stationary Earth” is a fair rendering of one branch of that disjunction, but it is entered as one of 461 numbered proofs, where consistent with is counted as evidence for. A live option has been promoted to a demonstration without a word being strengthened. Item 281, “Sagnac proves apparatus rotation,” is correct physics pointing the other way, and it sits in the same list unremarked.

The refutation above answers the disjunction, not the fragment. It concedes outright that Michelson–Gale alone cannot assign the rotation to one body, and puts the weight on the instruments a rotating optical medium cannot reach: the pendulum, the gyrocompass, and Coriolis circulation.

What we found is, is when we turned on that gyroscope we found that we were picking up a drift. A 15 degree per hour drift. … We obviously were not willing to accept that, and so we started looking for ways to disprove it was actually registering the motion of the Earth.

— Behind the Curve (dir. Daniel J. Clark) (2018), Bob Knodel speaking on camera in the ring laser gyroscope sequence. Transcribed here from two press accounts of the film — Newsweek, 15 February 2019, for the first sentence, and JOE.co.uk for the second — which differ in small ways. No timecoded viewing of the film was made for this entry and the surrounding edit was not verified.

The provenance of this passage is unusual and worth stating before anything else is built on it. It is not a publication. It is a man speaking in a documentary someone else made — Behind the Curve, directed by Daniel J. Clark, festival premiere 30 April 2018, US release 15 November 2018, Netflix February 2019 — and the only reason the sentence exists at all is that a camera crew was present. Bob Knodel is living, and everything below treats these words as an honest description of what an apparatus displayed and what was done next. This page makes no claim about why.

One line is deliberately left out. The most quoted sentence from this sequence is about what the instrument cost. It is easy to find and it is not reproduced here, because the only use for it is to invite a reader to infer a motive, and inferring motive from a filmed remark is the move this review exists to object to when it is done to us.

The number. 360° ÷ 24 h is 15.000° per hour exactly; the sidereal rate — the one a gyroscope actually senses, because it measures rotation with respect to the local inertial frame rather than with respect to the Sun — is Ω = 7.292115 × 10⁻⁵ rad s⁻¹, or 15.041° per hour. The two differ by 0.27 per cent. The figure as spoken carries two significant figures and cannot separate them, so nothing here claims that this apparatus resolved sidereal from solar. Large ring lasers do resolve it, routinely, and that matters for a separate reason set out under ARG-A02.

Why the reading is not nothing. A ring laser gyroscope has no spinning rotor and no bearings to bind. Two laser beams counter-propagate around a closed optical cavity, and rotation of the cavity with respect to the local inertial frame separates their frequencies by Δf = 4A·Ω / λL, with A the enclosed area, L the perimeter and λ the vacuum wavelength. Navigation-grade units of this class are specified at bias uncertainties better than 0.01° per hour — some fifteen hundred times smaller than the figure reported here. No make, model, input-axis orientation or uncertainty is given in the sequences the press transcriptions cover, so the reading is treated throughout as reported speech and never as a measurement this page relies on.

What happened next, from a secondary account. Jay L. Wile's contemporaneous review of the film describes the follow-up: the instrument was enclosed in a container excluding magnetic fields, on the hypothesis that the dome of stars was influencing it in some way; the same result came back; and a chamber made of bismuth was being sought next. That is a secondary account of a film sequence, not a laboratory record, and it is used below only for the shape of the follow-up, never for a number.

What we could not reach. No write-up of this gyroscope data authored by Knodel himself was located. The GlobeBusters broadcast archive was not searched for this entry. Unreached is not the same as non-existent, and nothing in this treatment turns on the assumption that no such record exists.

SURFACE (weak — do not use). “They do not understand gyroscopes.” This is the easy bust and it is false here, conspicuously so. The argument identifies the correct instrument, the correct physical quantity and a prediction the two models disagree about by the whole of it: a gyroscope bolted to a stationary Earth senses zero rotation with respect to inertial space, and one bolted to a turning Earth senses 15.041° per hour. Somebody then went and bought one. Anyone who opens with the surface reading has conceded the more interesting ground before the exchange starts.

DEEPER. The mechanical version of the claim rests on something true. A toy gyroscope, or a cheap mechanical one on jewelled bearings, genuinely cannot resolve 15° per hour: static friction in the bearings and imperfect balancing swamp a rate of four thousandths of a degree per second, and the resulting real wander is larger and less predictable than the signal being looked for. Videos making exactly this demonstration were circulating and being picked apart on Metabunk in March 2016, two years before the film. Item 19 — “Gyroscopes stable absent recalibration” — is therefore a true report about a class of instrument. It is incomplete rather than wrong: it is a fact about sensitivity, not a fact about the Earth.

KERNEL. The strongest form of this argument is not about gyroscopes failing. It is items 225 and 374 — “Ring laser gyro corrections”, “INS Earth updates” — and they are correct about deployed hardware. A strapdown inertial navigation system does not present the Earth's rotation to its user as a measurement. It supplies an Earth-rate term, computed from a stored latitude and a fixed constant, and subtracts it from the sensed body rates before the navigation solution is formed; the solution then accumulates integration drift and has its position periodically corrected by an external system, in practice GPS. Which means the popular debunk — “the inertial platform in your airliner proves the Earth spins” — is, on that framing, circular: the rotation was an input. The argument has found a real hole in a commonly offered proof. And the right way to close a hole like that is exactly what was done here: obtain the sensor, bypass the navigation solution, and read the raw sensed rate. That is a better experiment than most of this list contains, and better than the objection's usual critics propose.

Why it doesn’t save the claim.

Because a correction is derivable only from a quantity that is measurable, and this particular correction wears its measurement on its face.

It has a latitude signature. The Earth-rate term applied to a directional gyro is not a constant fudge; it is 15 × sin(latitude) degrees per hour, sign flipping between hemispheres. Mechanical heading indicators carry a latitude nut, set on the ground, whose entire purpose is to induce an equal and opposite real wander in the rotor. A screw calibrated to a trigonometric function of where you are standing is not a patch applied to save a theory; it is a fit to a rotation vector with a direction in space. A stationary Earth supplies no reason for that term to exist, and no reason at all for it to vary as the sine of latitude.

The same physics finds north from scratch, with nothing fed in. A gyrotheodolite is given no heading, no almanac and no satellite: it senses the horizontal component of the Earth's rotation vector and precesses into the meridian, to within about ten arc seconds. It is standard equipment in mine surveying and tunnel engineering precisely where star sights and GPS do not reach, and it was used to align the Channel Tunnel bores. Its documented failure mode is the tell: it is not used within roughly 15° of either pole, because there the angle between the Earth's rotation axis and the local vertical becomes too small for the horizontal component to be recovered. An instrument that extracts a direction out of nothing, and that fails exactly where the vector it is extracting goes flat, is not running on an assumption.

And in this cluster's own case the experiment was performed. The kernel says a deployed system presupposes the rotation rather than measuring it. The correct response is to strip the presupposition out and read the sensor directly. That was done, on camera, and the film's own record of the reading is the rate the rotating model predicts. The kernel is a genuine objection to a sloppy proof; it is not an objection to the Earth turning, and the one test it licenses has been run.

First, the concession, because it is the honest place to start. An inertial navigation system in an aircraft is not an independent demonstration of the Earth's rotation. Its mechanisation subtracts an Earth-rate term derived from a stored latitude before it forms a navigation solution; the solution suffers integration drift, because small errors in sensed acceleration and angular rate are integrated into progressively larger errors in velocity and position; and it is therefore periodically corrected from an external source, normally GPS. Anyone who answers this cluster by pointing at an airliner has argued in a circle, and items 225 and 374 have caught them at it. This page does not make that argument.

Second, what a ring laser gyroscope measures. Not motion relative to the ground, not motion relative to a medium, and not motion relative to the instrument's own past: rotation with respect to the local inertial frame. Two beams counter-propagate around a closed cavity; rotating the cavity separates their frequencies by Δf = 4A·Ω / λL. There is no free rotor, so the bearing-friction objection that defeats a mechanical gyroscope — the true objection behind item 19 — has nothing to act on. The underlying effect is the Sagnac effect, and the fact that it contains no medium in its statement is worked through at ARG-A02 rather than repeated here.

Third, the filmed reading, with its limits stated plainly. The figure spoken in Behind the Curve is 15 degrees per hour. In the sequences the press transcriptions cover it arrives with no make, no model, no input-axis orientation, no uncertainty, no duration and no repeat, and it is not evidence this page leans on. Note in particular that the axis geometry is unresolved: a three-axis unit reports the magnitude of the rotation vector, the full 15.041° per hour at any latitude, whereas a single-axis unit with its input axis vertical reads only Ω sin φ. Nothing in this treatment depends on which it was. The reading is evidence about the argument, not about the Earth. The Earth's rotation is established by the instruments in the next paragraph.

Fourth, where the evidence actually is. Léon Foucault built the first gyroscope in 1852 as a follow-up to his pendulum, for the express purpose of demonstrating the Earth's rotation, and coined the word for it — gyros and skopein, to view the rotation. Michelson, Gale and Pearson closed a light path around a rectangle at Clearing, Illinois in 1925, predicted a fringe displacement of 0.236 ± 0.002 for a rotating Earth and measured 0.230 ± 0.005 (ARG-A02). A 1 m × 1 m ring laser in Christchurch resolved the Earth's rotation in 1992, and because sensitivity grows as the square of the ring size, the 4 m × 4 m G ring laser at the Geodetic Observatory Wettzell does far better: at latitude 49.145° with a 632.8 nm helium–neon laser, the formula above returns a Sagnac beat of about 348 Hz for the Earth's rotation alone — that figure is our arithmetic from the published expression and the instrument's stated geometry, not a quoted measurement. Di Virgilio and colleagues report that G's Allan deviation drops below one part in 10⁹ of the Earth's angular rate after about 10⁴ s of integration. Nine digits. That instrument tracks length-of-day variation and polar motion, and agrees with VLBI determinations of the same wobbles made by an entirely different technique.

Fifth — and this is the crux — the reading was tested against the wrong hypothesis. “Drift” is the correct term of art: in gyro engineering it names the instrument's own bias. So the first objection to a 15°-per-hour reading is the right one, and it is this is my bias, not the Earth. What separates the two is geometry, and it is free. A bias is fixed in the instrument's body frame; the Earth's rotation is fixed in inertial space, along the celestial pole. Three consequences follow, and each is a test that costs nothing.

(a) Reverse the azimuth. Turn the unit 180° about the local vertical and re-run. The sensed Earth-rate component reverses sign; the bias does not. This two-position comparison is standard practice, and it separates the two terms outright.

(b) Tilt the input axis. With the axis vertical the sensed component is Ω sin φ; swing it to horizontal-east and the component goes to zero, because the input axis is then perpendicular to the Earth's rotation vector. A bias is indifferent to which way the box is pointed.

(c) Change latitude. Carry the same unit north or south and the vertical-axis reading scales as sin φ. A bias travels unchanged.

Magnetic shielding is not on that list, and could not have been. The Sagnac output depends on the enclosed area, the wavelength and the rotation rate; an external field is not a term in it. But the shielding run deserves credit rather than ridicule, because it was a control and it did its job: it tested a specific hypothesis — an outside influence coupling into the apparatus — and, on the secondary account we have, returned the same number, which excluded exactly what it was built to exclude. That is a result. Hunting systematics after an unwelcome reading is what every laboratory does and it is not a criticism.

Sixth, the four remaining items, answered from the hardware.

Item 19, “Gyroscopes stable absent recalibration.” The opposite is in every pilot's training material. A mechanical heading indicator wanders at an Earth-rate of 15 × sin(latitude) degrees per hour, is trimmed on the ground with a latitude nut, and is manually realigned against the magnetic compass every ten to fifteen minutes in routine in-flight checks; neglecting it is a recognised source of navigation error. An attitude indicator carries an erection mechanism that continuously applies force to return the gyro to the vertical, because the local vertical is not a fixed direction in space. Aircraft gyroscopes are not stable absent recalibration; recalibrating them is a checklist item.

Item 225, “Ring laser gyro corrections.” True, and the correction is the Earth's rotation. Its magnitude is not a free parameter chosen to make the numbers work: it is fixed by a rate measured independently, to nine digits, by the large ring lasers above, and its dependence on latitude is a signature no stationary model supplies a reason for.

Item 374, “INS Earth updates.” Conceded in the first paragraph, and it does not reach the claim. Position is externally updated; heading is not. An inertial platform obtains true north during alignment by sensing the horizontal component of the Earth's rotation vector, which is the same operation a gyrotheodolite performs underground with no satellite and no star in view.

Item 112, “Non-rotation gyroscope studies.” This names a genre rather than a result, and no study answering to the description was located in the searches run for this entry, which are the ones listed in the sources below. The nearest artefact located is a video of a mechanical-gyroscope demonstration in circulation by March 2016; the failure mode of that demonstration is set out in the steelman above and it is a real one.

Seventh, what the verdict claims, and what it withholds. Self-contradicted here is a claim about an argument, not about a person. It means this: the argument nominates a test, the test is the correct one, it was designed and run by the person our records name as the argument's originator, and the instrument returned the value the argument requires to be zero. No inference is drawn from that about anyone's sincerity, competence or state of mind — and the competence point runs the other way, since buying a navigation-grade instrument and running the measurement is more than the argument's distributors did, and more than most items on this list have behind them.

Eighth, the one thing the design was missing, stated as a method problem. The apparatus was right, the quantity was right, the prediction was unambiguous and the follow-up control was reasonable. What is missing from the sequences the press transcriptions cover is an acceptance criterion — a statement, made before the run, of which reading would count as confirmation and which as refutation. We have not viewed the full film and cannot say whether one was given elsewhere in it. Without one, no reading can be decisive in either direction, and the experiment's own result has nowhere to go. That is the whole of the finding: the instrument reported, the protocol had no place to put it. The repair is cheap and is written above — reverse the azimuth, tilt the input axis, move the box a few degrees of latitude, and say in advance what each outcome will mean.

The list says

Gyroscope anomalies indicating no rotation.

The source says

“when we turned on that gyroscope we found that we were picking up a drift. A 15 degree per hour drift.”

The item states the opposite of its source.

This is the flattest reversal on the board, and it is worth being precise about what is being compared. The item asserts a gyroscope anomaly indicating no rotation. The only source text available to hold it against is the filmed statement of the person our records name as the argument's originator, and that statement reports a gyroscope indicating rotation, at the rate the rotating model predicts. Same instrument, same experiment, opposite sign.

The reversal is carried by item 12 alone, and the cluster is named after it. That matters, because the other four do not all drift the same way and it would be dishonest to score them as though they did. Item 19 is a faithful compression of a true claim about cheap mechanical gyroscopes, whose bearing friction really does swamp a 15°-per-hour signal. Items 225 and 374 are faithful compressions of a correct engineering objection about how deployed inertial systems obtain their Earth-rate term, and they are the ones the refutation above spends most of its length on, for exactly the reason the hedge rule exists: they are the version worth beating. Item 112 names a genre rather than stating a result, so there is nothing in it to be faithful or unfaithful to.

The gap is the finding, and it runs the usual direction. A cluster whose best items are a careful objection to a circular proof arrives on the list under a headline asserting a result that is not obtained anywhere in the sequences the press transcriptions cover. What circulates is item 12. What was actually said is the quotation above.

If Earth and its atmosphere were constantly spinning eastwards over 1000mph, then the average commercial airliner traveling 500mph should never be able to reach its Eastward destinations before they come speeding up from behind! Likewise Westward destinations should be arrived at thrice the speed, but this is not the case.

— 200 Proofs Earth Is Not a Spinning Ball (2015), 200 Proofs Earth Is Not a Spinning Ball, numbered proofs 25–27, quoted from the Internet Archive copy of the print edition (item 200-proofs-the-earth-is-not-a-spinning-ball; PDF text layer, sixth and seventh pages of the list). That edition carries no printed page numbers in its text layer, so the proof numbers are the locator. Proof 26 is a block quotation from Gabrielle Henriet, Heaven and Earth — see the gloss

Read the last clause before anything else. Dubay’s empirical claim is narrow and, as far as it goes, true: westward flights are not arrived at three times the speed of eastward ones. Across proofs 25 to 27 that is the whole of the observation offered. The list’s items 110 and 236 — “Flight symmetry east/west”, “Flight times symmetric” — assert something else, and something checkable: that the two directions take the same time. They do not, and the compression block below carries that as a finding.

Where the aeroplane version comes from. The next proof in the same run, number 26, is a block quotation from Heaven and Earth by Gabrielle Henriet, and that book is the earliest text located that argues from aircraft speeds at all. Henriet’s chapter II, “On the fact that the earth does not rotate”, on the spread paginated 10–11 in both copies consulted: “Thus, if the earth rotates, as it is said, at 1,000 kilometers an hour, and a plane flies in the same direction at only 500 kilometers, it is obvious that its place of destination will be farther removed every minute.” An aircraft matching the rotation eastward would “remain suspended in mid-air over the spot from which it took off”; landing is difficult to picture on an airfield “slipping away at the rate of 1,000 kilometers an hour”. The chapter ends with a challenge that this page takes seriously enough to answer literally: it would be useful, Henriet writes, to know what people who fly think of the rotation of the earth. Dating is unsettled — one reference work gives 1956 (Holborn Publishing Co., London), another 1958; the scan used here has no date on its title page, only “Translated from the French”, the price, and the printer’s imprint at Arundel, Sussex. Mid-1950s is as tight as this pass could make it.

A word that changed on the way through. Henriet writes kilometers at every occurrence — four of them inside the block Dubay reproduces, and a fifth in the airfield sentence just past its end, in both copies checked. Dubay’s proof 26 prints miles, all four times, inside the quotation marks. The substitution is not visible to a reader, and it is not obviously careless either, because both versions happen to be roughly right: 1,000 km/h is the Earth’s surface speed at about 51.5°N, the latitude of London, to within four per cent; 1,000 mph is the equatorial figure to within four per cent. The claim is rescaled by a factor of 1.61 and its arithmetic still closes. Proof 26 also drops a sentence from the middle of the passage without an ellipsis.

The Victorian ancestry, which is Carpenter’s and not the vehicle you would expect. East/west symmetry as a numbered proof is in One Hundred Proofs (1885) twice over, and Carpenter’s own index names it both times. Proof 42, indexed “Projectiles—firing east or west”, and quoted with its conditional intact: “If the Earth went through space at the rate of eleven-hundred miles in a minute of time, as astronomers teach us, in a particular direction, there would unquestionably be a difference in the result of firing off a projectile in that direction and in a direction the opposite of that one. But as, in fact, there is not the slightest difference in any such case…” Then proof 44, the same argument restated with a figure: “since the Earth is said to move at the rate of nineteen miles in a second of time, ‘from west to east,’ it would make all the difference imaginable if the gun were fired in an opposite direction. But, as, in practice, there is not the slightest difference, whichever way the thing may be done…” Proof 54 is not an east/west comparison at all but the same decoupling argument with the vehicle changed — and it is the one that comes nearest to flight: “The aeronaut is able to start in his balloon and remain for hours in the air, at an elevation of several miles, and come down again in the same county or parish from which he ascended.” Behind Carpenter stands the air-gun of Earth Not a Globe, Section II — the ball that should fall “considerably more than one statute mile to the west” and lands at the muzzle. That argument is ARG-A10’s, and it is genuinely Rowbotham’s; the aeroplane is a later vehicle bolted onto it, and the man who built the chassis died nineteen years before the Wright brothers flew.

One thing the 1865 text does say about travelling east and west. Considering the day gained or lost on a circumnavigation, Rowbotham concludes that it is “no more favourable to the idea of rotundity than it is to the opposite fact that the earth is a plane; as both forms will permit of the same effect.” On the nearest question he addresses, the founder of the zetetic lane calls east/west travel asymmetry non-discriminating — which is the opposite of the use four items on this list make of it. (Searched: the Project Gutenberg #69892 text of the 1865 Simpkin, Marshall printing, for balloon, aeronaut, fly, flight, east and west. An aeroplane argument is not located in that text; later editions were not reached in this pass.)

SURFACE (weak — do not use). “The atmosphere rotates with the Earth, so it carries the aircraft along.” This loses, and it loses to a five-year-old’s objection: the air demonstrably does not move with the ground, which is what wind is, and an aircraft carried by an air mass is carried at the air mass’s speed, not the ground’s. Dubay has the caricature pre-loaded at proof 23 — gravity “magically and inexplicably” dragging the lower atmosphere “in perfect synchronization” up to some undetermined height — and anyone who opens with the fluid instead of the frame walks into it.

DEEPER. Galilean invariance: the aircraft, the runway and the air all share the Earth’s rotational velocity before the wheels leave the ground, and nothing in the flight removes it, so ground speed is a relation between aircraft and ground and the shared term cancels. True, and answered in 1632 in the Second Day of Galileo’s Dialogo. Incomplete, because it invites the obvious reply: if the shared motion cancels out of everything, you have not shown that it exists — which is ARG-R03’s move, and a good one.

KERNEL. Henriet has asked for the right thing. His argument is not really about aerodynamics; it is a demand for an observable that would distinguish the frames — the discrepancy between an aircraft’s motion through the air and its motion over the ground — and he is right that on a rotating Earth with a stationary atmosphere that discrepancy would be enormous and permanent and easterly. He is also right that flight is the place to look, because an aircraft is the one everyday vehicle that is mechanically decoupled from the surface for hours at a time. And his closing line is a fair challenge rather than a taunt: ask the people who fly. His number is not sloppy either — 1,000 km/h is the surface speed at British latitudes to four per cent. The strongest form of this cluster is therefore: an aeroplane is the natural apparatus for detecting the Earth’s rotation, aviation has run that apparatus continuously for a century, and if the rotation were real the industry would have had to build something to cope with it.

Why it doesn’t save the claim.

Because the industry did, and the answer to Henriet’s challenge is a printed procedure rather than an argument.

The discrepancy he asks about is computed on every flight. An inertial reference unit outputs, among other things, wind speed and direction, and it gets them by differencing two independent sensor chains: inertial ground velocity against true airspeed from the air data computers. That readout is Henriet’s quantity. If the ground were sliding eastward under a still atmosphere at 1,000 km/h, every airliner in the world would display a permanent 1,000 km/h easterly. They display tens to a few hundred kilometres per hour, in a direction that changes with the weather.

And the residual rotation terms he is implicitly betting against are not zero. They are in the navigation equations by name — Earth rate, transport rate, Coriolis — they are the reason a navigation-grade gyro is required at all, and one of them is measured before pushback: the alignment derives the aircraft’s latitude from the sensed rotation and refuses to complete if it disagrees with what the crew typed in. The kernel is exact and it points the other way: aviation is the best-instrumented place to look for the Earth’s rotation, which is why aviation found it.

Start with what is conceded, because two things in this cluster are true. A pilot makes no correction for the Earth’s rotation in the sense the argument means — there is no rotation knob, no westward drift allowance in the flight plan, nothing in the cockpit that answers to “spin”. And westward flights are not three times faster than eastward ones, which is the only empirical claim made in the three proofs quoted here. Both are granted at full strength. The cluster fails anyway, on the two things it adds to them: that flight times are symmetric, and that nothing in aviation compensates for the rotation.

1. The prediction has the wrong sign

Take the source at its own strength. Henriet’s aircraft flies east into a receding destination and west with 1,000 km/h added; Dubay’s version has westward destinations reached at thrice the speed. Both make eastbound the disadvantaged direction. On the mid-latitude routes this argument is always made about, eastbound is the fast direction, and has been since the routes existed.

The clean instance: British Airways flight BA112, a Boeing 747-400, New York JFK to London Heathrow on 9 February 2020, 4 hours 56 minutes for 5,554 km — an average ground speed of 1,126 km/h and a peak of 1,327 km/h, against a cruise airspeed of about 933 km/h. The aircraft covered ground faster than it moved through the air, going east, in the direction Henriet says an aircraft cannot make progress at all. The same day, westbound flights were routing north of the jet stream over Greenland; some, Flightradar24 reported, were adding more than sixty minutes, while others that managed to avoid the jet stream entirely were in some cases arriving early. The routine version of the same fact, without the storm, is the jet stream’s standing effect on scheduling: eastbound long-haul flights tend to run shorter than their westbound returns, on long routes by an hour or more.

So the answer to items 110 and 236 is not that the asymmetry is smaller than claimed. It is that the asymmetry is real, published in every timetable, and pointed the other way.

2. The asymmetry reverses with latitude, which is the discriminating part

A ground-slip term has one sign everywhere: whatever the Earth is doing, it is doing it in the same direction at every latitude. The observed asymmetry does not behave like that. Below about 30° the prevailing surface winds are the easterly trades; above it they are the westerlies, and the mid-latitude jet stream sits in that belt. Both are products of the same rotation: air moving equatorward in the Hadley circulation is “deflected toward the west in both hemispheres by the Coriolis effect”, and air moving poleward is deflected the other way, which is why the mid-latitude belt blows from the southwest.

Sailors were exploiting the reversal four centuries ago. Urdaneta’s tornaviaje of 1565 established the Pacific round trip that the Manila galleons then ran for 250 years: west in the tropical trades, home eastward at high latitude in the westerlies. The east/west travel asymmetry this cluster denies is not merely real, it is old enough to have organised the world’s shipping three centuries before Rowbotham, and its latitude structure is a rotation signature that no slippage between ground and air can imitate.

3. The compensation exists, and it has a procedure and a duration

Item 13 is the interesting one, because it is answerable from a manual. An airliner’s inertial reference system aligns on the ramp before every flight, and here is what the alignment is: “The alignment computations use the basic premise that the only accelerations during alignment are due to the earth’s gravity; the only motion during alignment is due to the earth’s rotation.” The accelerometers find local vertical from gravity; then “the laser gyro sensed earth rate components are used to establish the heading of the airplane”, and “Earth rate sensing by the laser gyros allows the IRU to determine initial latitude”, which “is compared to the crew entered latitude” — a comparison that “must be favorable to complete the alignment period”. Minimum align time: ten minutes.

Read that again as a measurement rather than as avionics. Before the aircraft moves, its instruments determine which way is north and what latitude it is at, using nothing but gravity and the rotation of the Earth — and then check the answer against the crew. On a stationary Earth the gyros would see nothing to work with and the procedure would have no content. The signal is not marginal, either: Earth rate is 15.041°/hr (the sidereal day is 23 h 56 m 4 s), of which 9.36°/hr is horizontal at the latitude of London, while Honeywell’s GG1320AN — the industry-standard navigation-grade ring laser gyro — has a typical bias stability of 0.0035°/hr. The thing being measured is about four thousand times the instrument’s noise floor. This is the same quantity, and the same instrument class — a navigation-grade ring laser gyro — that ARG-A07 records a flat-earth researcher reading 15°/hr on camera, where the finding is the missing acceptance criterion rather than the reading.

The rotation then stays in the equations for the rest of the flight. The non-inertial terms a navigation-grade system must compensate are named in the professional literature as Earth rate, transport rate and Coriolis acceleration — and the same source notes the converse, that in cheap sensors “Earth rate (15 deg/hr) stays below gyro errors and cannot be measured reliably”, which is precisely why cheap sensors cannot navigate. The size of the Coriolis term at cruise: 2Ωv sin φ = 0.026 m/s² at 45° and 250 m/s. Dropped from the mechanisation, an acceleration that size integrates to a position error of order ½at² — something like 170 km after an hour, before any of the error-bounding behaviour of a real navigator is considered. It is not dropped.

4. Two more residues that were measured, one of them from an aircraft

Weight. The vertical companion of the Coriolis term is the Eötvös effect: 2Ωv cos φ, which at 45° and 250 m/s is 0.026 m/s² — 2,578 mGal, about 0.26% of g, and reversing sign between east and west for a swing of some 5,150 mGal. Baron Roland von Eötvös noticed the discrepancy in shipboard gravimetry and it was confirmed in 1908 by sending two ships across the Black Sea in opposite directions. Airborne gravity surveying inherited the problem wholesale — Harlan’s 1968 paper in the Journal of Geophysical Research is titled, flatly, “Eötvös corrections for airborne gravimetry”. An aircraft flying east weighs measurably less than the same aircraft flying west, and the survey industry has spent sixty years subtracting the difference.

Time. In October 1971 Hafele and Keating put four caesium clocks on scheduled commercial flights and flew them round the world twice, once east and once west. Their abstract reports “directionally dependent time differences”: the flying clocks “lost 59 ± 10 nanoseconds during the eastward trip and gained 273 ± 7 nanoseconds during the westward trip”, against predictions of −40 ± 23 and +275 ± 21. That 332-nanosecond split between the two directions is the east/west asymmetry of air travel, measured to the nanosecond, on ordinary airline tickets — and it exists only because the ground the aircraft take off from is itself rotating relative to the local inertial frame. Absent that rotation the direction of travel would carry no distinction in principle, and there would be nothing for that split to come from. The experiment has been repeated by the National Physical Laboratory twice, London–Washington in 1996 (39 ± 2 ns against 39.8 predicted) and round the world in 2010 (230 ± 20 against 246 ± 3).

5. What the frame argument does and does not have to do

The reason an aircraft does not need to chase its destination is not that the atmosphere drags it. It is that ground speed is a relation between the aircraft and the ground, and both carry the same rotational velocity into the flight — the point Galileo made with a ship’s sealed cabin in 1632 and which ARG-A10 works through in Rowbotham’s own words, since Rowbotham granted that the atmosphere turns with the Earth and built an experiment to show it. That is the whole of the frame answer, and on its own it would leave ARG-R03’s reply standing: cancelling terms prove nothing either way.

Which is why the load in this entry is carried by the terms that do not cancel. Coriolis, Eötvös, the sensed Earth rate in the alignment, the directional split in the clocks: four quantities that are zero on a stationary Earth under a fixed heaven, are not zero, and are each the size a rotating Earth predicts. The list asks whether aviation compensates for the spin. It does — in the alignment procedure, in the mechanisation equations, in the gravimeter’s correction table and in the timing budget — and the compensations are the measurement.

And here is the concession those four terms owe, made where a reader can see it. Each of them measures rotation relative to the local inertial frame, which is all any of them can do. If instead the heavens turn about a fixed Earth and the distant matter carries the local inertial frame round with it, the Earth is still turning with respect to that frame: the gyro still senses 15°/hr, the Coriolis and Eötvös terms are still there, and the clocks still split. That reply is a real one, it is the strongest available to this cluster, and this page takes it seriously at ARG-R01 and ARG-R03 rather than here.

What it does not do is rescue the two claims actually on the list. “Flight times symmetric” is false in the published timetables whatever defines the inertial frame, and “aircraft navigation not compensating for spin” is false in the alignment procedure whatever is rotating relative to what — a procedure that derives latitude from a sensed rotation is compensating for one, on any account of what is going round. A defender who reaches for general covariance to save those two sentences has conceded them.

The list says

Aircraft navigation not compensating for spin. / Plane flight times consistent with rest Earth. / Flight symmetry east/west. / Flight times symmetric.

The source says

“…the average commercial airliner traveling 500mph should never be able to reach its Eastward destinations before they come speeding up from behind! Likewise Westward destinations should be arrived at thrice the speed, but this is not the case.”

The source does not contain this. The source’s empirical claim is true; the list’s is false. That is the finding, and it is the reverse of this project’s usual one. Dubay denies a threefold westward advantage, which is correct — no such thing exists. Henriet, quoted in the next proof, offers no empirical observation in the passage Dubay reproduces: it is an a priori argument about what would follow if the ground moved, and what flight times actually do is not stated anywhere in it. Items 110 and 236 supply the missing premise on the sources’ behalf — “Flight symmetry east/west”, “Flight times symmetric” — and it is refutable from a timetable. Item 13’s nearest source text is proof 27 (landing on a moving runway) and Henriet’s airfield “slipping away”; the phrase “navigation not compensating” is the list’s own.

unsourced_addition is recorded, with a qualification the enum cannot carry. Directional symmetry itself is in this literature, asserted as observed fact, with a different vehicle: Dubay’s proof 24, one item before the passage quoted above, has it that “regardless of which direction cannons are fired, the distance covered is always the same”, and Carpenter’s proofs 42 and 44 say the same of projectiles. What is not located anywhere in the texts read for this entry — Dubay’s numbered proofs, Henriet’s chapter II, Carpenter 1885 — is the claim asserted of flight times: the one vehicle whose east/west asymmetry is printed in every timetable, and the vehicle all four list items name. The addition is therefore a change of vehicle rather than an invention from nothing, which makes scope_widened a defensible alternative reading: “not thrice as fast”, said of aircraft, and “always the same”, said of cannonballs, widened into “flight times symmetric”. A reader who prefers it has both texts above to judge from.

The sharper drift happened one link earlier, and the enum has no word for it. Dubay’s proof 26 quotes Henriet inside quotation marks and prints miles four times where Henriet wrote kilometers — verified in two independent copies of Henriet and in two text layers of the Internet Archive scan of Dubay — while dropping a sentence from the middle without an ellipsis. The claim is rescaled by 1.61 in transit. It is invisible because both numbers survive the change: 1,000 km/h is the Earth’s surface speed at the latitude of London to within four per cent, and 1,000 mph is the equatorial figure to within four per cent. This is not the list compressing a source; it is one source restating another and the number moving, which is the same process one link upstream.

The refutation answers the source, not the fragment. It takes Henriet’s claim at his own strength — that an eastbound aircraft could not cover ground — and answers it with an eastbound aircraft covering ground faster than it moved through the air, 1,126 km/h averaged across the crossing against a 933 km/h airspeed; it takes his challenge to ask people who fly and answers it with their alignment procedure. What the fragment adds, the symmetry of the timetables, is refuted separately, because the fragment is what circulates.

Returning to the pendulum, it will be found to be equally unsatisfactory as a proof of this peculiar rotundity of the Earth. It is argued that as the length of a seconds pendulum at the equator is 39,027 inches, and 39,197 inches at the north pole, that the Earth must be a globe, having a less diameter through its axis than through its equator. But this proceeds upon the assumption that the Earth is a globe having a “centre of attraction of gravitation,” towards which all bodies gravitate or fall … It should also be first proved that no other cause could operate besides greater proximity to the centre of gravity, to produce the variable oscillations of a pendulum. This not being attempted, the whole subject must be condemned as logically insufficient, irregular, and worthless for its intended purpose. Many philosophers have ascribed the alterations in the oscillations of a pendulum to the diminished temperature of the northern centre. … Thus there are two distinct and tangible causes which necessarily operate to produce the variable oscillations of a pendulum, without supposing any distortion in the supposed rotundity of the Earth.

— Earth Not a Globe (1865), Section I, in the discussion of the oblate spheroid and the pendulum, from the Project Gutenberg transcription of the 1865 first book edition (ebook #69892, “Original publication: United Kingdom: Simpkin, Marshall, and co., 1865”). The same argument, with a third alternative cause added, is at ch. XIV “Variability of Pendulum Vibrations” of the 1881 third edition

He prints the measurement. The two figures in that sentence — 39.027 inches at the equator, 39.197 at the north pole; the comma is 1865 typesetting for a decimal point — are the equatorial relief. The length a pendulum must have to beat seconds is L = g/π², so a table of pendulum lengths by latitude is a table of gravity by latitude. Rowbotham does not dispute the numbers. He disputes what they show. Two of the four items in this cluster — “No felt centrifugal force” and “No equatorial centrifugal relief” — deny the thing his own page tabulates.

What his argument actually is, and it is not a silly one. Three moves. (i) The inference from a shorter pendulum to a shorter polar radius presupposes a centre of attraction and therefore a globe. (ii) Nobody has shown that no other cause could produce the variation. (iii) Two other causes are available: the pendulum rod contracts in the colder north, and the colder northern air is denser and resists the bob differently. The 1881 third edition adds a third — “electric and magnetic states of the atmosphere.” He italicises assumption, is and no other; the emphasis is his.

The alternative causes are testable, and he supplies the numbers to test them with. On the same pages he quotes a temperature table from Sir Richard Phillips’s Million of Facts, p. 475 (mean annual temperature 84.2 °F at the equator, 0 °F at the pole) and a thermal-expansion figure from Noad’s Lectures on Chemistry, p. 41 (“a change of temperature equal to 30° Fah. will alter its length 1/5000th part…”). Those two quotations decide the question against him, and the refutation below does the multiplication.

The other half of this cluster comes from a different book and a different century. Items 55 and 105 concede a bulge and dispute its cause; Rowbotham denies the bulge is established at all, and argues from the disagreement among meridian-arc measurements that the Earth might as well be an oblong spheroid. The “another cause” formulation is the Machian one, and the earliest text located that carries it in the form the list uses is Sungenis and Bennett, Galileo Was Wrong. In the Vol. I scan (archive.org item GallileoWasWrong), Bennett’s ch. 12 opens at printed p. 710 by setting out the geokinetic case he means to answer — “there are three geokinetic claims for terrestrial motion”, of which the second is that centrifugal forces cause the polar flattening and equatorial bulge and “This also explains why the acceleration of gravity is less at the equator” — and he answers it at p. 711 not by denying any of that but by writing that the effects “depend on the assumption that the inertial effects can only be caused by the Earth’s rotation.” At ch. 4, p. 204 the same volume says the oblateness proves “that there is a force causing their effect, not that a rotation of the Earth is the force.” Both books grant the phenomena. No assertion of items 180 or 257 was located in the passages of Vol. I read for this entry — pp. 49, 204, 239, 460, 699, 710–711 and Appendix 7 — and the two sentences that come nearest are both answered below: the satellite-photograph parenthesis at p. 699, and the remark at p. 711 that in the rotating-shell model “an object at the center of the hollow sphere will not be affected by the inertial forces … but not the Earth itself, if it is centrally located”, which pulls against the concession a few lines above it.

The first of those two lines, in full, because it runs the other way and it is fair to note it. At printed p. 699, in ch. 11 on Hildegardian geocentrism, a parenthesis has it that the precession is attributed to the equatorial bulge “(even though satellite photographs of the Earth do not show an equatorial bulge)”. As a remark about photographs that is true and uninteresting: the flattening is 1/298, so on an image 1,000 pixels across the polar diameter is about three pixels shorter than the equatorial one. As an argument it does not survive the same volume’s GPS appendix — Appendix 7 in this scan, at printed p. 1028 — which prints the WGS84 flattening as 1/298.257223563 without demur.

What this passage is being cited as. The earliest text located that carries the pendulum-and-oblateness argument in the form the list compresses, and it confirms the record’s edition: the material is in the 1865 first book edition, not only in the enlarged 1881 third. It is an ancestor of two of the four items and it is not evidence of origination for the other two.

SURFACE (weak — do not use). “You can feel it, and the bulge proves rotation.” Both halves lose. You cannot feel a third of one per cent of your weight, and saying so hands over the only part of the item that is straightforwardly true. And the bulge on its own proves that something is flattening the Earth, which is what the geocentric source says too, in nearly those words.

DEEPER. Rowbotham is right that the pendulum alone underdetermines the figure of the Earth. A latitude variation in g is a fact about the gravity field, not about the shape; deriving the shape needs an assumption about how mass is distributed inside, and that is exactly what Newton and Huygens disagreed about — 1/230 against 1/578, from the same data. He is also right that the 18th- and 19th-century meridian arcs disagreed badly, and right that the British arcs he tabulates come out as if the Earth were oblong. Those are real published results, quoted accurately.

KERNEL. The strongest form is not Rowbotham’s at all; it is the one Galileo Was Wrong takes, and it is Mach’s. The equatorial bulge and the latitude variation of gravity do not by themselves distinguish a rotating Earth in a fixed sky from a fixed Earth in a rotating sky. That is not a flat-earth evasion invented for the purpose: Mach wrote it, Einstein endorsed it in 1911 and again in the 1920 Leyden address, and it is a theorem — Thirring showed in 1918 that a rotating mass shell induces Coriolis and centrifugal terms in its interior, and Pfister and Braun (Class. Quantum Grav. 2:909, 1985) showed that with the shell’s stresses handled correctly the induced centrifugal force comes out exactly right. Sungenis and Bennett quote Mach’s sentence in full at their ch. 7, p. 460, qualifying clause included, and their conclusion is precisely calibrated: the phenomena prove “that there is a force causing their effect, not that a rotation of the Earth is the force.” On the narrow question of what the bulge proves about whose rotation, that sentence is correct.

Why it doesn’t save the claim.

Because it concedes everything the four items deny, and buys nothing back.

Take the kernel at full strength and follow it. If the sky turns instead of the Earth, the relative rotation rate is the same — one turn per sidereal day, ω = 7.292115 × 10⁻⁵ rad/s — because it is the relative rate that both descriptions have to reproduce. Every number then comes out identical: the same 0.0339 m/s² of centrifugal acceleration at the equator, the same 21.4 km bulge, the same 0.53% weight difference between pole and equator. Mach’s point is that the two descriptions are the same physics; it is not that one of them has no bulge. So the Machian argument predicts the equatorial relief that items 180 and 257 report as absent. It is not an alternative to the measurement. It is an alternative label on the measurement.

And Rowbotham’s half is not underdetermination but arithmetic. He does not stop at “the pendulum alone cannot settle the figure”, which would have been defensible. He asserts two specific causes, and the numbers on his own page sink both: the expansion coefficient he quotes makes the first about eight times too small, and the air-pump experiment he cites for the second makes it, read at its most generous, some twenty times too small. A hypothesis that names its own mechanism has given up the protection that underdetermination affords.

Start with the two concessions, because both are permanent. First: you cannot feel it. The centrifugal term at the equator is 0.0339 m/s², which is 0.35% of gravity, and no human perceives that. Anyone who answers this item by claiming the effect is perceptible has lost the exchange and deserved to. Second: the bulge and the latitude variation of gravity do not, by themselves, tell you whether the Earth turns or the sky does. That is Mach’s observation, Einstein agreed with it, and the geocentric source states it accurately. This page is not going to pretend either point away.

What the verdict ranges over. Not “the bulge has been proved to come from the Earth’s own rotation rather than the universe’s.” The cluster’s four items make two claims, and they are answered separately: that no centrifugal effect is detectable (items 180, 257), which measurement contradicts; and that the bulge has some other cause (items 55, 105), which turns out to change no measured quantity at all.

1. The effect is not felt. It is weighed, and it has been since 1672

Jean Richer took a pendulum clock from Paris to Cayenne, at 4.9°N, in 1672 and found it losing time — Newton records the rate as 2^m 28^s a day — and had to shorten the bob by 1¼ lignes to make it beat seconds again. Huygens read that as the centrifugal effect of rotation reducing apparent gravity near the equator; Newton read it as the equator being further from the centre. Both were partly right, and the argument between them is what made the figure of the Earth a research programme.

Modern numbers, from the GRS80 reference ellipsoid, recomputed here 2026-08-10 so a reader can redo them:

• Normal gravity at the equator 9.7803268 m/s², at the pole 9.8321864 m/s². Difference 0.0518596 m/s², or 0.5302%.
• Centrifugal acceleration at the equator ω²a = (7.292115 × 10⁻⁵)² × 6 378 137 m = 0.0339157 m/s² — 65.4% of that difference. The remaining 34.6% is the shape term: the equator sits 21.4 km further from the centre.
• In practical units: 100.00 kg weighed at the pole reads 99.47 kg at the equator. That is half a kilogram, and an ordinary digital scale would show it.

It is not measured with bathroom scales. The International Gravity Formula — g(φ) = 9.780327 (1 + 0.0053024 sin²φ − 0.0000058 sin²2φ) — is the first correction applied to every gravity survey ever run for minerals, oil or geodesy, and the 0.0053024 in it is the number above. A field gravimeter resolves about 0.01 mGal; the centrifugal term is 3,392 mGal, some 340,000 times larger. Absolute gravimeters of the FG5 class do not use a pendulum at all — they drop a corner cube in vacuum and time it interferometrically against a rubidium clock, with an accuracy the manufacturer states as “2 µGal (observed agreement between FG5-X instruments)”. That instrument has no rod to expand and no air to resist it, and it returns the same latitude dependence. So do superconducting gravimeters, which are finer still.

The rotation is not merely detected, it is used. The Earth’s surface moves east at ωa = 465 m/s at the equator, and that velocity is a free contribution to any eastward launch: 463 m/s at Kourou (5.2°N), 409 at Cape Canaveral (28.5°N), 325 at Baikonur (45.6°N). The differences are large enough to drive the siting of launch complexes. And the Eötvös correction — a ship or aircraft moving east weighs less, by 2ωvcos φ — is applied as routine in marine and airborne gravimetry: at 10 m/s eastward on the equator it is 146 mGal, more than a hundred times the survey noise floor. Set ω to zero and every one of those corrections becomes a systematic error that nobody has ever had to make.

2. Rowbotham’s alternative causes, tested with Rowbotham’s own coefficients

This is where the source’s version of the argument, which is the one that has to be answered, comes apart — and it comes apart on its own page. He offers temperature and air density as the causes of the pendulum’s variation. He then quotes the two numbers needed to check the first one.

From Phillips’s Million of Facts he takes mean annual temperature falling from 84.2 °F at the equator to 0 °F at the pole. From Noad’s Lectures on Chemistry he takes the expansion of a pendulum: 1/5000 of its length for 30 °F, that is 1/150,000 per degree. Multiply: 84.2 ÷ 150,000 = 5.6 × 10⁻⁴, or 0.056%. The variation he has to explain, from his own two pendulum figures, is (39.197 − 39.027) ÷ 39.027 = 0.436%. Thermal expansion is short by a factor of 7.8. Against the modern value for the same quantity, 0.530%, it is short by 9.4.

The coefficient is not the problem. 1/150,000 per °F is 6.7 × 10⁻⁶/°F, against a modern figure for steel of 11.7 × 10⁻⁶/°C = 6.5 × 10⁻⁶/°F. He quoted a good number and did not multiply it out.

His second cause fails the same way, and it should be tested with the experiment he names, not with a substitute. The mechanism he states is resistance, not buoyancy: “if the pendulum vibrates in the air, which is colder and therefore denser in the north than at the equator, it must be more or less resisted in its passage through it”, and his authority is Derham, who compared arcs of vibration in an air-pump receiver at different densities and found the rate differing by “two seconds in an hour when the vibrations were longest”. Take Derham’s number at face value and in Rowbotham’s favour. Two seconds in 3,600 is a fractional change in period of 5.6 × 10⁻⁴, and since T ∝ √L that is 1.1 × 10⁻³ on the length of a seconds pendulum — but that is the whole atmosphere against a vacuum. What Rowbotham needs is the equator-to-pole difference, and on his own temperature table (84.2 °F against 0 °F) air at constant pressure is denser at the pole by about 18%. Eighteen per cent of 0.11% is 0.02%, against the 0.436% he has to explain: short by a factor of about twenty, on the most favourable reading available to him. Computed instead from the physics — buoyancy plus added mass shift the apparent g by roughly 1.5 ρ_air/ρ_bob, some 2 × 10⁻⁴ for a brass bob in full air, of which the latitude variation is again about a fifth — it is short by of order a hundred. And Derham’s experiment measures the wrong thing for the purpose: damping changes the amplitude, and amplitude reaches the period only through the circular-error term, which is smaller still. And the instrument that settles it has no air in it at all: the FG5 above returns the same latitude dependence with its mass falling in vacuum.

The 1881 edition enlarges the argument and quotes the answer to it. By the third edition (printed pp. 185–186) Rowbotham has added a survey of General Sabine’s pendulum campaign — 131 observations, from 79° 49′ north to South Shetland at 62° 56′ south — and concludes from the 23 of them that departed markedly from the computed values that “the assumption of Sir Isaac Newton that the earth is an oblate spheroid, is not confirmed by experiments made with the pendulum.” He then quotes Sabine’s own account of why those 23 departed: the discrepancies are “due in a far greater degree to local peculiarities than to what may be more strictly called errors of observation”, and Francis Baily’s judgement that a pendulum’s vibrations “are powerfully affected, in many places, by the local attraction of the substratum on which it is swung”. That is the correct explanation, printed in his own text: the residual scatter about the latitude law is local geology, which is what a gravimeter is for. It is the same move the project records at ARG-B06, where he reprints the Encyclopædia Britannica’s article on levelling — the correction for the difference between the true and the apparent level, which is the curvature term — and then denies the curvature it corrects for.

And here is the part that decides it. Take L = g/π² and the GRS80 gravity values, and the seconds pendulum should be 39.014 inches at the equator and 39.221 at the pole. Rowbotham printed 39.027 and 39.197. His table agrees with what a rotating oblate Earth predicts to within 0.034% at one end and 0.061% at the other — while the effect he was explaining away is 0.436%. He had the right answer on the page in front of him and offered a mechanism eight times too small to displace it.

3. The meridian arcs: his facts are largely right, and they point the other way

The disagreements he tabulates are real. The Ordnance Survey series he quotes from Hugh Murray’s Encyclopaedia of Geography does show degrees getting shorter going north over southern England, which is the signature of an oblong figure. And the Lapland degree he quotes, 57,422 toises, is the published Maupertuis figure, which did not survive later scrutiny intact: he prints the Swedish re-measurement’s verdict himself, that the French had given the degree there “196 toises more than the true length”.

His Peru degree does not check out. He gives “the first degree of the meridian from the equator as 56,653 toises”. The published Peruvian arc — 176,945 toises over 3° 7′ 3″, in the 1911 Britannica article cited below — works out at about 56,760 toises to the degree, and reducing that to the equator moves it by only some ten toises. His figure is roughly a hundred toises, about 200 m, low. That is a second corrupt number of the same kind as the Huygens ratio below, and it pulls the same way: it widens the disagreement his argument runs on.

What produces those discrepancies is now the most useful thing about them. A meridian arc is measured by combining a triangulated ground distance with astronomically observed latitudes, and astronomical latitude is referred to the local plumb line — which is pulled sideways by nearby mass. Short arcs are therefore dominated by local gravity anomalies rather than by the Earth’s overall figure, which is exactly why the small English arcs scatter into apparent oblongness. The most famous instance is the one that founded a field: in the Great Trigonometrical Survey of India the latitude difference between Kaliana and Kalianpur came out 5.24″ smaller geodetically than astronomically, and when Pratt computed what the Himalaya ought to do to the plumb line he got 15.885″, more than three times the observed value. Pratt (1855) published the discrepancy and said he could not explain it; Airy (1855) proposed that mountains float on lighter roots, and isostasy was born. The anomaly Rowbotham reads as evidence that the figure of the Earth cannot be measured is the anomaly that told geophysicists what mountains are made of.

Two further points of detail, since he rests weight on both. Maupertuis’s Lapland result implied a flattening near 1/179 — far more oblate than the truth — so the later Swedish re-measurement in Bothnia, which shortened his arc, moved the computed flattening toward the modern 1/298.257 and not toward zero. (Svanberg’s own published flattening was not located in the sources reached here; the direction follows from the geometry and is stated as our inference.) And the Huygens figure he cites — “577 to 875, or a difference of about one-third of the whole diameter”, in both the 1865 and the 1881 editions — is a corrupt number: Huygens’s ratio was 577:578, a flattening near 1/578, and 578 has plainly been scrambled into 875, with the “one-third” then computed off the corruption. That matters because the disagreement is his whole argument: 577 : 875 implies a flattening of (875 − 577)/875 = 0.34 — which is where his “about one-third” comes from — against Huygens’s actual 1/578 = 0.0017, so the figure he prints is about 197 times the one Huygens computed. There is no reason to think he did it deliberately; he took a number from a secondary source and did not check it against the arithmetic he printed beside it.

The underlying point is the one his framing misses. Newton got 1/230 and Huygens 1/578 from the same observations because they assumed different things about the density inside the Earth. The spread was not a failure of the method; it was the method telling them that the flattening measures the interior. Which is what Clairaut then proved.

4. One number has to do two jobs, and it does

Clairaut’s theorem relates the shape of a rotating equilibrium figure to the gravity on it: (g_p − g_e)/g_e = (5/2)m − f, where f is the flattening and m = ω²a/g_e is the ratio of centrifugal to gravitational acceleration at the equator. Put the measured numbers in. m = 0.0339157/9.7803268 = 0.00346775. With f = 1/298.257222101, (5/2)m − f = 0.0053166. The observed gravity ratio is 0.0053024. They agree to 0.27%, which is the size of the second-order terms Clairaut’s first-order theorem drops.

That is the check the “other cause” has to pass. The flattening and the pole-to-equator gravity difference are measured by completely different means — satellite geodesy and orbit perturbations for one, gravimeters for the other — and a single parameter, the rotation rate, ties them together to three decimal places. That rotation rate is itself measured independently, by the length of the sidereal day, by ring-laser gyroscopes, and by very-long-baseline interferometry against quasars. Any proposed alternative cause of the bulge inherits an obligation to reproduce both numbers and their relationship, using something other than ω. Nothing on offer does.

5. The rotating universe reproduces all of it, which is the problem

The geocentric alternative is not a different value of anything. Mach’s claim, endorsed by Einstein and made rigorous for a rotating mass shell by Thirring (1918, corrected 1921) and definitively by Pfister and Braun (1985), is that a universe rotating about a fixed Earth induces the same Coriolis and centrifugal fields. Take it seriously and the consequence is immediate: the relative rotation rate must still be one turn per sidereal day, so the induced centrifugal field is still ω² times distance from the axis, so the equilibrium figure is still flattened by 1/298 and gravity is still 0.53% weaker at the equator. The re-description predicts the equatorial relief that two of these four items deny. It is the same spacetime in different coordinates, and no observable moves.

What it does cost is a global inertial frame: with the whole sky turning once a sidereal day, the coordinate tangential speed reaches c at c/ω = 4.11 × 10¹² m, about 27.5 AU — outside Uranus, inside Neptune. General relativity permits that as a coordinate effect and nothing local exceeds c, which is why this page treats the frame question at ARG-R01 and the Machian machinery at ARG-R02 and ARG-R05 rather than here. For this cluster the relevant fact is narrower and harder: whichever description you adopt, the bulge and the weight difference are exactly where the measurements find them.

There is a further wrinkle inside the source’s own statement of the model. At p. 711, a few lines below the Response quoted above, Bennett reports from Barbour and Bertotti’s rotating-shell result that an object at the centre of the hollow sphere is not affected by the inertial forces, and infers that the space around the Earth shows the effects but not the Earth itself. That is true of a point at the centre, where the net centrifugal force vanishes. It is not true of a body 6,371 km in radius: the centrifugal potential grows as the square of the distance from the axis, and it is that variation across the body — not the force at its centre — that raises an equatorial bulge. On the version of the model as stated, the effect the argument is trying to re-explain would not appear at all.

6. The bulge is not rigid, and the part that is unexplained does not help

Item 105 asserts a stable equatorial bulge; no wording for it was located in either source, so what follows answers the claim as the list states it. The bulge is not rigid, and the ways in which it moves are among the better-measured things in geophysics — but the honest thing to say first is how big those movements are. The solid Earth flexes about 30 cm twice a day under lunar and solar tides: one part in 70,000 of the 21.4 km bulge itself. The dynamical oblateness J₂ was decreasing secularly through the 1980s and 1990s as the crust rebounds from the last ice age, and then reversed — Cox and Chao reported the turn around 1998 in Science 297:831 (2002), Dickey and colleagues argued later the same year (Science 298:1975) that ocean and ice mass redistribution accounts for it, and the decadal behaviour is still an active question — and that secular trend is a fractional change of order 10⁻⁸ a year, which is sub-millimetre a year if it is carried onto the polar–equatorial radius difference.

Those are small numbers, and saying so is the point rather than a concession. They are the size a responsive rotational figure should show: a body that visibly deforms under the tides, and whose oblateness tracks the redistribution of ice and water on it, is a body whose 21.4 km flattening is a deformation under load rather than a fixed shape. What none of them is, is an opening for a non-rotational cause. And the weight of item 105 is on bulge, not on stable: what the argument has to displace is a flattening tied to the rotation rate by Clairaut’s theorem to three parts in a thousand, and centimetres of tide do not displace it. This page is not going to answer a four-word item by pointing at the fourth decimal place.

And the bulge is not in hydrostatic equilibrium. Chambat, Ricard and Valette (Geophysical Journal International 183:727, 2010), correcting Nakiboglu’s standard calculation, put it plainly: “The difference between the polar and equatorial radii appears to be 113 ± 1 m (instead of 98 m) larger than the hydrostatic value.” That is a real excess — 0.53% of the 21.4 km bulge — and it is not fully accounted for. It is attributed to density structure in the mantle and to delayed viscous relaxation from Pleistocene deglaciation, and that attribution is a modelling result rather than a closed case. This page will keep saying so.

It is also no use to the argument. The excess is half a per cent; the rotational term supplies the other 99.5%, and the excess is defined relative to the hydrostatic figure a rotating Earth predicts. A residual measured against a model is not evidence against the model that generated the residual.

7. What is left, stated without decoration

The centrifugal effect at the equator is 0.35% of gravity: imperceptible, and measured continuously by instruments a hundred thousand times more sensitive than it needs. The flattening and the gravity difference are tied together by the rotation rate through Clairaut’s theorem to better than three parts in a thousand. Rowbotham’s alternative causes are about eight and about twenty times too small, on the coefficient and the experiment he printed himself. The geocentric alternative cause is not an alternative to any measurement; it is the same physics in rotating coordinates, and it predicts the equatorial relief this cluster reports as missing. Two of these four items deny a phenomenon that both of the books behind this cluster affirm.

The list says

No equatorial centrifugal relief.

The source says

Rowbotham: “It is argued that as the length of a seconds pendulum at the equator is 39,027 inches, and 39,197 inches at the north pole, that the Earth must be a globe … It should also be first proved that no other cause could operate.” — Sungenis & Bennett, Vol. I ch. 12, p. 710, stating the geokinetic claim they are about to answer: centrifugal forces produce “the polar flattening and equatorial bulge. This also explains why the acceleration of gravity is less at the equator” — and answering it at p. 711 not by denying those effects but by writing that they “depend on the assumption that the inertial effects can only be caused by the Earth’s rotation.”

The item states the opposite of its source. Both books affirm the effect these items deny. Rowbotham prints the equatorial relief as a table — a seconds pendulum of 39.027 inches at the equator against 39.197 at the pole — and then argues about its cause. Sungenis and Bennett set the phenomenon out as the claim they mean to answer and then answer it by relocating the cause to a rotating universe, writing at Vol. I p. 711 that the effects “depend on the assumption that the inertial effects can only be caused by the Earth’s rotation”, and at ch. 4, p. 204 that the oblateness proves “that there is a force causing their effect, not that a rotation of the Earth is the force.” The same volume quotes Bertrand Russell, at p. 49, on “the flattening of the Earth at the poles, and the fact that bodies are heavier there than at the equator”, and glosses Einstein at p. 239 as holding that “the Earth’s poles would flatten from either reference frame”. An argument about attribution has arrived on the list as an assertion of absence.

The nearest thing to a source for item 180, and it is in the same chapter. A few paragraphs below the Response, on that same p. 711, Bennett restates Barbour and Bertotti’s rotating-shell result and adds that “an object at the center of the hollow sphere will not be affected by the inertial forces … but not the Earth itself, if it is centrally located”. Read flat, that is a source statement that no centrifugal effect acts on the Earth — close to item 180, and in tension with the concession he has just made a few lines above. It is answered on the merits in section 5 of the refutation: the cancellation holds at a point at the centre, and it is the variation of the centrifugal potential across a body 6,371 km in radius, not the force at its centre, that raises a bulge. Recorded here because a scoped claim about what the sources do and do not assert has to carry its own counter-instance.

The four items do not agree with each other. Items 55 and 105 concede a bulge and dispute its cause; items 180 and 257 deny that there is anything to attribute. One phrasing is worth flagging as genuinely ambiguous: “No equatorial centrifugal relief” can be read as “there is no reduction in weight at the equator”, which is false and is what its neighbour item 180 commits it to, or as “the reduction is not centrifugal in origin”, which is Rowbotham’s actual position. Four words cannot carry that distinction, and the distinction is the whole argument. reversed is recorded for the first reading, which is the one the list’s own neighbouring item forces and the one a reader arrives with; the second reading is answered on the merits in section 5 of the refutation.

The refutation answers the sources, not the fragments. Against Rowbotham it grants that the pendulum alone underdetermines the figure of the Earth, and then uses the expansion coefficient and the temperature range he quotes on his own page to show his named alternative cause is about eight times too small. Against Sungenis and Bennett it grants Mach, Einstein and the rotating-shell theorem at full strength, and puts the weight on the consequence: the same relative rotation rate produces the same bulge and the same 0.53% weight difference, so the alternative cause predicts precisely what items 180 and 257 report as absent.

One thing this entry does not claim. At Vol. I p. 204 the book quotes Mach up to “and so on…”, stopping where the qualifying clause begins. That looks like a dropped hedge and it is not: at ch. 7, p. 460 the same volume prints the sentence in full, clause included, together with Mach’s statement that he prefers to reformulate the law of inertia rather than accept absolute motion. On this argument the books are the careful party and the list is not.

Four one-line items, three instruments, and no author. The search for one was run in full and it ended in an answer rather than a shrug, so here is exactly what was read and where it stopped — including the places a reader who knows better should push back.

The specimen cites nothing. Re-fetched 2026-08-10: the page is headed “THE EARTH IS FLAT AND NON ROTATING” and carries 461 numbered lines with no footnote, link or attribution against any of them. Items 232, 249, 250 and 398 read in their entirety: “Gravimeter constancy.” “Pendulum building sensitivity.” “Torsion balance null variation.” “Pendulum clocks stable.” That is the whole of the argument as published.

The Victorian trunk of the tradition. Rowbotham’s Zetetic Astronomy: Earth Not a Globe! (1865 first book edition, Project Gutenberg transcription #69892) returns 78 hits for “pendulum” and none at all for “gravimet-”, “torsion” or “Cavendish”. Carpenter’s One Hundred Proofs (1885, Gutenberg #55387) returns seven for “pendulum”, all of them proof 73 and the index line, and zero for the other three words. Both pendulum discussions are about Foucault precession, which this review scores at ARG-A06.

And the seventy-eight hits point the wrong way for the list. This is the finding worth the search. In the 1865 edition, in the sequence of anti-rotundity arguments that follows the Polaris passage in Section I, Rowbotham takes up precisely the phenomenon item 398 denies — and concedes it. He sets the globe-side argument out in its own terms — that “because a pendulum vibrates more rapidly in the northern region than at the equator, the Earth is thereby proved to be a globe” — and at no point disputes the phenomenon; he prints the figures: “the length of a seconds pendulum at the equator is 39,027 inches, and 39,197 inches at the north pole” (the commas are the printer’s; read 39.027 and 39.197). His objection is not that the variation is absent. It is that the inference to a spheroid “proceeds upon the assumption that the Earth is a globe having a ‘centre of attraction of gravitation’”, and that “it should also be first proved that no other cause could operate besides greater proximity to the centre of gravity”. He then supplies two other causes — thermal expansion of the rod, and the density of the air — and concludes that “there are two distinct and tangible causes which necessarily operate to produce the variable oscillations of a pendulum, without supposing any distortion in the supposed rotundity of the Earth”. The founding text of the flat-earth tradition agrees that pendulum clocks are not stable, and argues about why. The refutation below answers that argument, at its own strength, in his own numbers, at §2. It is the nearest thing to a source this cluster has, and it is its negation — which is why nobody is credited here, and why item 398 has not been moved into ARG-A13, where Rowbotham’s real argument lives.

The modern compilers. Dubay’s 200 Proofs Earth Is Not a Spinning Ball (archive.org item 200proofsearthisnotaspinningballericdubay, djvu text) gives nine hits for “pendulum”, all inside proof 140, and none for “gravimet-”, “torsion” or “Cavendish”. Proof 140 is a Foucault claim — the swing “depends on 1) the initial force beginning its swing and, 2) the ball-and-socket joint used” — which is the honest ancestor of item 249 on one reading of that item, and is answered in §3 of the refutation. Edward Hendrie’s The Greatest Lie on Earth (2016), the largest post-2015 compendium with a physics apparatus, returns zero hits for all five of “pendulum”, “Foucault”, “gravimeter”, “torsion” and “Cavendish” in the archive.org OCR text of that scan.

The geocentric side, where the instruments actually appear. Sungenis and Bennett’s Galileo Was Wrong is the one place in this literature where torsion balances are discussed at length, and it discusses them twice, in opposite directions from the list. In Vol. I, chapter 12 — a catalogue of ether-drift experiments, each with a “Geocentrism’s Response” appended — Trouton–Noble 1903 appears at printed p. 855: a charged capacitor hung on a torsion fibre, which returned a null for the Earth’s motion. That is the only torsion-balance null on point anywhere in the reachable geocentric corpus, and the book declines to bank it. Its response, on p. 856, reads in full: “Only light and gases show ether effects; the experiment was incapable of achieving ether detection unless a charged gas is used between the plates.” The same section records that the null “was repeated in experiments by Chase in 1927 and Hayden in 1994” and that such results “are now thought to be consistent with Special Relativity”. Meanwhile Vol. II runs the opposite argument: at printed p. 27 and again in Appendix 1 at p. 640 it cites torsion-balance and pendulum anomalies — Long’s 1976 inverse-square discrepancy “to the tune of 0.37%”, the eclipse pendulum measurements at Physical Review D3, 823, the Holding and Tuck mineshaft determination of G — as positive evidence for a LeSagean geocentric dynamics. Searching the OCR text of both volumes for “gravimeter” and “gravimetr-” returns zero hits in either.

The one live modern argument, which is not this one. The Flat Earth Society wiki does have pages on gravimetry, and their claim is worth stating accurately because it is better than item 232. It is not that gravimeters read constant. It is that they are long-period seismometers whose published output is reached through a stack of modelled corrections — latitude, free-air, terrain — and the page quotes a training manual to the effect that “Gravimeters do not give direct measurements of gravity; rather, a meter reading is taken which is then multiplied by an instrumental calibration factor.” That is the circularity charge, it is the strongest version of the position, and it is answered in §5 of the refutation rather than here.

Why this is not filed as older than the movement. That state records where an argument came from, and awarding it needs an old text that contains the argument. Two of the three instruments settle themselves: the torsion-balance nulls these items invoke are Eötvös’s of the 1890s and Trouton and Noble’s of 1903, and the superconducting gravimeter is of the 1960s, all later than the movement’s founding texts, so no older text can be arguing from them. The third does not — precision pendulum work is far older than the movement, Richer in 1672 and Kater in 1818, and §2 of the refutation rests on it — but that literature exists because it found the variation, and the earliest text of the movement to take it up, Rowbotham in 1865, concedes the variation as well. We found nothing older asserting the constancy item 398 asserts. That is a report on the corpora listed above rather than the result of a systematic search of pre-1849 literature, and it is the ground on which this state was rejected: there is nothing in the works listed above to inherit from.

An honest note on our limits. No source found means we did not find one, not that none exists. Four noun phrases with no verb between them can begin in a livestream, a comment thread or a caption, leaving nothing to search; and the corpora above were read as OCR transcriptions rather than print copies, so a hit could have been missed to a scanning error. A reader who can point us at someone who actually made this argument — in print, on air, anywhere datable — will improve this entry, and we will publish the correction.

SURFACE (weak — do not use). “Gravimeters vary all the time, so the items are simply false.” This wins a quarter of the argument and loses the rest in public. It is true of item 232 and item 398. It is not true of item 250, because torsion-balance tests of the equivalence principle really do return nulls and are celebrated for it; and it is not true of item 249 on its natural reading, because a Foucault pendulum really is sensitive to how it is hung and how it is started. Two of the four sentences in this cluster are correct as written. Anyone who opens by calling all four false has handed a defender two free recoveries.

DEEPER. The nulls are extraordinary and they are real. A continuously rotating torsion balance at Washington compared the fall of beryllium and titanium and found the Eötvös parameter to be η = (0.3 ± 1.8) × 10⁻¹³, with differential accelerations in any space-fixed direction bounded below 8.8 × 10⁻¹⁵ m/s². That is one of the most precise mechanical measurements ever made, and it found nothing. A defender who says only this has said nothing a physicist would contest.

KERNEL. The strongest form drops the false items and keeps the true one, and it is genuinely uncomfortable: the most sensitive mechanical instruments ever built have been pointed at this question for a century and a quarter, and the headline result is a null. Eötvös got a null in 1909. Trouton and Noble got a null in 1903 and Hayden got it again in 1994. The rotating balance gets a null today at one part in 10¹³. And a gravimeter sitting in a vault registers precisely nothing attributable to a planet supposedly travelling at 30 kilometres a second. You keep telling us the Earth is racing through space. The best instruments in the building cannot feel it. Every factual component of that is correct, and the last sentence is correct in a way this page has to concede in its own voice rather than talk around.

Why it doesn’t save the claim.

Because the nulls are nulls of the wrong quantity, and one of them is a null that only exists because the Earth spins.

Take the gravimeter first, and concede it fully. A gravimeter shows no signal from the Earth’s orbital velocity, and it should not. The Earth is in free fall around the Sun: the Sun’s pull at our distance is 5.9 × 10⁻³ m/s², and the orbital acceleration cancels it to first order everywhere in the laboratory. What survives is the gradient across the Earth — the solar tide, about 5 × 10⁻⁷ m/s², roughly 50 µGal — and gravimeters measure that, twice a day, on schedule. A steady velocity produces no force on anything; that has been the answer since Galileo put the sailor below decks, and ARG-A17 scores it separately. Insisting an instrument should register a velocity is asking it to violate the principle that makes it work.

Now the torsion balance, which is where the true thing points the other way. Eötvös’s balance compares two bodies of different composition hanging on one beam. It can distinguish them only because gravity pulls toward the Earth’s centre while the centrifugal effect of rotation pulls perpendicular to the spin axis, so the two do not act along the same line, and a difference in the ratio of inertial to gravitational mass would twist the beam. The size of that twist is set by the horizontal component of the centrifugal acceleration, ω²R cos φ sin φ, which at the latitude of Budapest is 0.0339 × cos 47.5° × sin 47.5° = 0.0169 m/s², about 1.7 × 10⁻³ of g (recomputed here 2026-08-10). Set the Earth’s rotation to zero and that term is identically zero. The beam has nothing to twist it, in either direction, whatever the test masses are made of. On a stationary Earth the Eötvös experiment is not a null result. It is no experiment. The movement’s favourite instrument is sensitive at all only on the hypothesis it is being cited against.

The modern version keeps the same dependence and adds a second one. The rotating balance quotes its bound on differential accelerations “in any direction” in a space-fixed frame — one defined by the stars — and the route from the laboratory to that frame is a rotation at the Earth’s rate. The null is stated in coordinates that presuppose the turning the item denies.

What this verdict ranges over, stated before anything else. Not “these instruments are useless”, and not “all four sentences are false”. Two of the four are true as written: torsion-balance tests of the equivalence principle do return nulls, and pendulum apparatus is genuinely sensitive to how it is built and started. The two that are false — gravimeter constancy and pendulum clocks stable — are contradicted below by named measurements. What fails across all four is the step from an instrument reading to a conclusion about the Earth, and that step is never taken anywhere in the specimen. Four noun phrases are not an argument, which is why the verdict is not demonstrated rather than refuted: there is no inference here to be wrong. REFUTED was weighed and rejected for exactly the reason a defender would give — it would be right about half this cluster and wrong about the other half.

1. Gravimeters are not constant, and the largest thing that moves them is the Earth’s spin

Start with the number every gravity survey begins from. The International Gravity Formula puts g at 9.7803253359 m/s² on the equator and 9.8321849378 m/s² at the pole — a spread of 0.53%, which is thousands of times the resolution of a field instrument. Of that, the centrifugal term accounts for ω²a = (7.292115 × 10⁻⁵)² × 6,378,137 = 0.0339 m/s², or 0.347% of g (recomputed here 2026-08-10); the remainder comes from the equatorial bulge, which is itself a consequence of the rotation. ARG-A13 scores the bulge on its own.

On top of that sits a signal nobody can operate a gravimeter without modelling. The solid Earth tide swings local gravity by 100 to 300 µGal depending on the phase of the Moon, twice a day, every day. It is not a delicate inference from a specialist instrument: a chip-scale MEMS gravimeter, a device etched in silicon, tracked nineteen days of it with a correlation of 0.975 against the theoretical tide. The theoretical tide is computed from the positions of the Moon and the Sun. A reading that agrees with an ephemeris to that tolerance is the opposite of a constant.

And then there is the signal that settles the matter, because it exists for one reason only. The Earth’s rotation axis wanders within the planet — the Chandler wobble, about 435 days. As it wanders, the centrifugal pseudo-force at a fixed station changes, and the station’s gravity changes with it. Superconducting gravimeters of the Global Geodynamics Project network see this, and the quantity they report is a measured admittance rather than an assumed one: stacking decade-long records gives a gravimetric factor at the Chandler frequency of 1.118 ± 0.016 in amplitude with a phase of −0.45° ± 0.66°, which the authors note is “smaller in amplitude than expected”. Two things follow. The pole tide is a gravity signal generated by the variation of a centrifugal term, so it cannot exist at all on a non-rotating Earth. And the number came out disagreeing with the prediction, which is not what a fitted parameter does.

2. Pendulum clocks: the instability is the founding datum of the whole subject

In 1672 Jean Richer took a pendulum clock, regulated in Paris, to Cayenne at about five degrees north. Newton reported that it went “slower than it ought in respect of the mean motion of the sun at the rate of 2^m 28^s a day”, and that to beat seconds there the pendulum had to be made 1.25 lignes — 2.256 mm — shorter than at Paris. Huygens read it as the centrifugal effect of the Earth’s rotation reducing apparent gravity near the equator; Newton read it as the equatorial bulge. Both readings are readings of a spinning planet, and the disagreement between them is the reason the eighteenth century sent expeditions to Lapland and Peru. Item 398 asserts the stability of an instrument whose instability is the single observation from which the figure of the Earth was first inferred.

By 1818 the effect was metrology. Kater’s reversible pendulum gave the seconds pendulum at London, at sea level, at 62 °F, swinging in vacuum, as 39.1386 inches, with a scatter from the mean of 0.00028 inches — a precision in g of about 7 mGal. Pendulum parties then carried Kater instruments from the tropics to the Arctic for the express purpose of measuring how much the reading changed with latitude, and published the ellipticity they got out of it.

And the finest pendulum clocks felt the Moon. In 1929, before spring gravimeters were measuring Earth tides at all, a Shortt–Synchronome free-pendulum clock was run against a timing system with a least count of a millisecond, and Brown and Brouwer’s analysis of the record found a diurnal sinusoid of amplitude 0.097 × 10⁻³ s and a semidiurnal one of 0.161 × 10⁻³ s — against a predicted semidiurnal amplitude of 0.178 × 10⁻³ s, ten per cent larger and almost exactly in phase. Duncan Agnew’s history of these measurements draws the conclusion plainly: clock-based systems, though noisier than spring gravimeters, “were an early form of an absolute gravimeter that could indeed observe Earth tides”. A pendulum clock detected the tidal pull of the Moon on its own bob, in 1929, at the level of a sixth of a millisecond. That is the instrument item 398 calls stable.

And the one argument in this tradition that engages the evidence concedes the instability and disputes the cause — then fails in its own author’s numbers. Rowbotham accepts the latitude variation of the seconds pendulum and reattributes it to thermal expansion of the rod and to the density of the air. That reattribution is ARG-A13’s material and is scored there; it is answered here because it is the nearest thing this cluster has to a source, and because it is the negation of item 398. Take his figures. What he must explain is (39.197 − 39.027)/39.027 = 4.36 × 10⁻³ in length, and since the period goes as √L, half of that in rate: 188 seconds a day. His own footnote, quoting Noad, prices the thermal mechanism — 30 °F alters the length by one part in 5,000 and costs “8 seconds per day” — and his own table gives his equator-to-pole temperature span as 84.2 °F to 0 °F, so the whole journey buys (84.2/30) × 1/5,000 = 5.6 × 10⁻⁴ of length, about 24 seconds a day. He is short by a factor of about eight. The air is worse: removing all of the atmosphere is worth 0.5 × (1.2/8,500) × 86,400 ≈ 6 seconds a day for a brass bob, and the equator-to-pole density difference is a fraction of that. Then the datum he quotes closes it without arithmetic. The London length he takes from Noad, 39.13929 inches, is specified “in vacuo … at the temperature of 62°” — it is Kater’s reduction, the two agreeing to about one part in 50,000 — so both of the causes he offers had already been held fixed, by definition, in the number he is explaining. He prints that specification himself, so this is not a matter of his being unaware of the corrections. (Recomputed here 2026-08-11, in his units, from the 1865 text.)

3. “Pendulum building sensitivity” is true, and it is the field’s own point

Item 249 admits two readings — that pendulums are sensitive to the building they hang in, or that pendulums can be built to great sensitivity — and the first is the one with a tradition behind it. Dubay’s proof 140 says the behaviour of a Foucault pendulum “depends on 1) the initial force beginning its swing and, 2) the ball-and-socket joint used which most-readily facilitates circular motion over any other”, and Carpenter’s proof 73 makes the Victorian version. On the physics, they are right, and we are not going to pretend otherwise. A Foucault pendulum “requires care to set up because imprecise construction can cause additional veering which masks the terrestrial effect”; a geometrical imperfection or elasticity in the wire can beat two horizontal modes against each other; an elliptical swing precesses on its own. This is why the bob is released by burning a thread rather than by hand, and why a Charron ring is fitted to bleed off the ellipticity.

What follows is the opposite of what the item wants. A demonstration whose systematics you have to engineer away is a demonstration, and the movement is entitled to say so. The measurement is elsewhere, and it has no thread and no bearing: a ring-laser gyroscope has no moving parts, and the G instrument at Wettzell measures the Earth’s rotation rate directly, its Allan deviation falling below one part in 10⁹ after about 10⁴ s of integration. Ring-laser data of this kind have been combined with VLBI observations to sharpen the Earth rotation parameters — polar motion and δUT1 — that VLBI alone derives from sparse networks, though the ring lasers are not yet at VLBI’s accuracy. ARG-A07 and ARG-A19 follow that through. Conceding that a museum pendulum is fiddly costs this page nothing, because the pendulum stopped being the evidence decades ago.

There is also an internal tension the list never resolves. Item 11 of the same specimen — scored at ARG-A06 — treats the Foucault precession as a real, regular phenomenon requiring an explanation, and explains it by a rotating firmament. Item 249, read the first way, treats the same precession as an artefact of the mounting. The list needs the precession to be real when it is being reassigned to the sky and unreal when it is being dismissed.

4. The torsion-balance null is real. It is also the wrong null, and it is one the geocentric literature has already spent

Torsion balances return nulls, and the best ones are magnificent. A continuously rotating balance comparing beryllium and titanium reports η = (0.3 ± 1.8) × 10⁻¹³ and bounds differential accelerations in any space-fixed direction below 8.8 × 10⁻¹⁵ m/s². Nobody disputes the null; the question is what it is a null of. It is a null of a difference between two materials, and the channel through which that difference could show up is the horizontal component of the centrifugal acceleration — 1.7 × 10⁻³ of g at Eötvös’s latitude, and exactly zero if the Earth does not turn. Cite the null and you have cited an instrument whose sensitivity is a function of ω.

Meanwhile the one torsion-balance experiment in this literature that was aimed at the Earth’s motion has been disclaimed by the people citing it. Trouton and Noble hung a charged capacitor on a torsion fibre in 1903 to look for a turning couple from motion through the ether, and found none. Galileo Was Wrong Vol. I sets the experiment out at printed p. 855 and answers it on p. 856 in its own voice: “Only light and gases show ether effects; the experiment was incapable of achieving ether detection unless a charged gas is used between the plates.” A null from an apparatus your own source calls incapable is not evidence, and the book is right to say so; the same section notes the result “was repeated in experiments by Chase in 1927 and Hayden in 1994” and is “now thought to be consistent with Special Relativity”.

Worse for the item, the same authors argue the reverse elsewhere in the same work. Vol. II’s chapter on the cause of gravity, at printed p. 27, and its Appendix 1, at p. 640, recruit torsion-balance and pendulum anomalies as positive evidence for a LeSagean geocentric dynamics: Long’s 1976 report of a discrepancy in the inverse-square law “to the tune of 0.37%”, the eclipse pendulum measurements at Physical Review D3, 823, the Holding and Tuck mineshaft determination of G. A tradition cannot run “the sensitive instruments show nothing, which proves us right” and “the sensitive instruments show anomalies, which proves us right” on the same page of the same shelf.

For completeness, and because it is the honest way to hold this ground: the most interesting claim ever made linking torsion balances to the Earth’s rotation ran in the geocentrists’ favour and did not survive. Anderson and colleagues reported in 2015 that published measurements of G oscillate with a 5.9-year period in phase with length-of-day variations. Schlamminger, Gundlach and Newman showed that corrections and additions to the G data “significantly weaken the correlation”, and Pitkin’s Bayesian reanalysis found a model with an extra unknown noise component favoured over any sinusoid by factors of order e³⁰. We would have had to publish that had it held. It did not.

5. The circularity charge, which is the best objection here and is answerable

The strongest version of item 232 is not in the specimen at all; it is on the Flat Earth Society’s own wiki, and it runs: a raw gravimeter reading means nothing until you have applied a latitude correction, a free-air correction, a terrain correction and a drift correction, every one of them derived from the model whose truth is in question. You are reading the globe back out of numbers you put the globe into.

That deserves a direct answer, and there is one. Take the Eötvös correction, which is the rotation term applied to every gravity measurement made from a moving platform. Its size is 2ωv cos φ: at 45° latitude, a ship steaming east at ten knots reads 53 mGal light, and an aircraft at 200 m/s reads about 2,060 mGal light (recomputed here 2026-08-10). Marine and airborne surveys work at the milligal level, so this term is tens to thousands of times the signal being mapped. Set ω to zero and it vanishes identically, and every marine gravity survey ever flown or sailed would need no such term at all.

Now the part that answers the circularity. The Eötvös effect was not imposed as a correction. It was found as an unexplained discrepancy. A German team from the Geodetic Institute of Potsdam took gravity measurements aboard ships in the Atlantic, Indian and Pacific in the early 1900s, and the readings came out lower when the vessel was moving east and higher when it was moving west. Eötvös identified the cause, and in 1908 two ships were run in opposite directions on the Black Sea to test it. The prediction held. That is the shape of a discovery, not of a fitted parameter: the data arrived first, from people who were not looking for it, and the term that explains it is proportional to the rate at which the Earth turns.

6. What is conceded, without decoration

A gravimeter shows nothing attributable to the Earth’s 30 km/s orbital velocity. That is true, it will stay true however good the instruments get, and it is a prediction of the moving-Earth account rather than a difficulty for it — the laboratory is in free fall, the Sun’s 5.9 × 10⁻³ m/s² pull is cancelled by the orbital acceleration, and only the gradient survives. The same non-discrimination applies to a uniform velocity of any kind, which is ARG-A17’s subject and was Galileo’s in 1632. What the instruments do see is rotation, and they see it four separate ways: in the 0.53% latitude gradient of g, in the pole tide as the spin axis wanders, in the Eötvös term on every moving platform, and in the very existence of a signal channel in the Eötvös balance. Three sentences of this cluster describe instruments that register the Earth’s rotation, and one describes an instrument that cannot register a velocity because nothing can.

There is no original to hold these four lines against, and the search that established it is set out in full under “No original to quote” above: the specimen carries no citation for any of its 461 items; Rowbotham 1865 and Carpenter 1885 discuss pendulums only as Foucault precession, and neither “gravimet-” nor “torsion” nor “Cavendish” is located in the full text of either; Dubay’s 200 Proofs is the same picture; Hendrie’s 2016 compendium returns zero hits for all five terms in the OCR text of that scan; and both volumes of Galileo Was Wrong return zero hits for “gravimeter”. The hedge rule has nothing to bite on, and that is itself the result — where an argument has an author we can show the list hardening a hedge, and here there is no hedge because there is no sentence upstream of the fragment.

But the search returned something sharper than a blank, and it runs in three directions at once. On the pendulum: the founding text of the tradition states the reverse of item 398. Rowbotham concedes that the seconds pendulum is longer at the pole than at the equator, prints both figures, and argues about the cause rather than the fact. On the torsion balance: the only null on point in the reachable geocentric corpus — Trouton–Noble 1903, at Galileo Was Wrong Vol. I printed p. 855 — is disclaimed on the next page by the book that reports it, as an experiment “incapable of achieving ether detection”. And on the instruments generally: the same authors’ Vol. II argues at printed pp. 27 and 640 that torsion balances and pendulums show anomalous variation, and treats that as evidence for geocentrism. So the compressed items are not firmer versions of something upstream; on three of four they point where the upstream literature declines to go.

One further gap belongs here rather than in the refutation, because it is a property of the compression itself: “Pendulum building sensitivity” is four words that will not resolve. It reads either as “pendulums are sensitive to the building” — the Dubay and Carpenter claim, which is true and which the refutation grants at §3 — or as “pendulums can be built to great sensitivity”, which is the reading our own cluster name silently adopts. An item that cannot be pinned to one assertion cannot have drifted from a source, and cannot be answered without answering both. §3 answers both.

Bearing this fact in mind, that there exists a continual pressure of the atmosphere upon the Earth, and associating it with the fact that the Earth is a vast plane “stretched out upon the waters,” and it will be seen that it must of necessity slightly fluctuate, or slowly rise and fall in the water. As by the action of the atmosphere the Earth is slowly depressed, the water moves towards the receding shores and produces the flood tide; and when by the reaction of the resisting oceanic medium the Earth gradually ascends the waters recede, and the ebb tide is produced. This is the general cause of tides. Whatever peculiarities are observable they may be traced to the reaction of channels, bays, headlands, and other local causes. …A large and heavily laden vessel will make several fluctuations in a minute of time; the Earth once only in about twelve hours, because it is proportionately larger.

— Earth Not a Globe (1865), Section 10, “Cause of Tides”, in the 1865 book edition, as transcribed in Project Gutenberg ebook #69892 (produced from an Internet Archive scan; the quotation was checked character-by-character against that transcription on 2026-08-10, and not against a print copy). The enlarged third edition of 1881 carries the same argument as Chapter XII, “The Cause of Tides”, with further material; the 1881 wording quoted in the gloss below is from the sacred-texts.com HTML of that edition.

This is a mechanism, and it is a specific one. Rowbotham does not stop at denying the Moon. He names a driver (the weight of the atmosphere), a restoring force (the buoyant reaction of the water beneath), and a period fixed by the size of the floating body — and he then derives consequences from it, on the same two pages: high and low water arrive simultaneously on meridians 180° apart; there are no tides in lakes and inland seas, “which being contained within and fluctuating with the Earth cannot therefore show a relative change in the altitude of the surface”; inland wells rise and fall because they connect to the sea “by subterranean passages”; and the Earth carries a “vibratory or tremulous motion” detectable in a spirit level. Those are four testable statements, which is more than the one-line item preserves and considerably more than it is usually given credit for.

The chain forward, with pages. Thomas Winship reproduces the conclusion at Zetetic Cosmogony (1899) p. 131 — “Tides are caused by the gentle and gradual rise and fall of the earth on the bosom of the mighty deep” — and David Wardlaw Scott quotes Winship in turn at Terra Firma (1901), ch. XIII sec. 2, pp. 258–261. Eric Dubay carries Rowbotham’s lunar-mass objection into 200 Proofs (2015) as proofs 117 and 118 without the mechanism attached. The Flat Earth Society wiki’s Tides page, as served on 2026-08-10, carries exactly one footnote, and it is “Rowbotham, Samuel Birley. Chapter XII., Earth Not a Globe., Second Ed., p.160.” — so the sentence that page offers about Newton is sourced to a flat-earth pamphlet rather than to Newton. This passage is cited here as the earliest located ancestor of the tide material, not as the origin of these four items: the vocabulary they use — tidal nodes, resonance, seiches, bores — is not located in any of those four texts.

Two hedges travel with it and neither survives into the list. Winship, on the same page 131: “The moon is the TIMEKEEPER for the tides, nothing more. The ‘phase’ of the moon tells what kind of a tide may be expected, but she does not and cannot ‘attract’ either the solid body of the earth or the waters.” Scott, at the close of his account of the cause: “By what means the tides are so regularly affected is at present only conjectured; possibly it may be by atmospheric pressure on the waters of the Great Deep, and perhaps even the Moon itself, as suggested by the late Dr. Rowbotham, ‘may influence the atmosphere … and indirectly the rise and fall of the Earth in the waters.’ Of this we cannot now be sure.” The canon’s own two carriers concede that the Moon’s phase predicts the tide, and that the cause is conjectural.

But read the page before, because it cuts the other way. The timekeeper sentence is a concession Winship attaches at the end; the flat assertion sits on p. 130, in his own voice, and it is the one item 252 preserves: “But the tides are not uniform. At Port Natal the rise and fall is about six feet, while at Beira, about 600 miles up the coast, the rise and fall is 26 feet. This effectually settles the matter that the moon has no influence on the tides.” He then prints Melville’s Tahiti observation as proof of it. So the Moon–tide mismatch of item 252 is Winship’s claim and not the list’s: what travels is the assertion, and what is left behind is the qualification he put next to it.

And the evidence Winship printed is the answer to him. Directly above his conclusion on p. 131 he reprints a report from the Leicester Daily Post of 25 August 1892 on Ernst von Rebeur-Paschwitz’s horizontal-pendulum observations at Wilhelmshaven and Potsdam, with the Tenerife runs of December 1890 and April 1891, and sets the finding in capitals: “IT WAS ESTABLISHED THAT THE EARTH RISES AND FALLS LIKE THE OCEAN OR THE ATMOSPHERE.” The clipping he chose says the oscillations correspond “with the course of the moon”. Those observations are the first detection of the solid Earth tide — Fréchet and Rivera, Journal of Seismology 16:315–343 (2012): “he succeeded in observing the luni-solar tide for the first time, by recording and measuring its effect on the apparent ground tilt.” The ground does rise and fall, by about 38 cm in the M2 component and some 55 cm at the equator, and it does so on the Moon’s clock.

The other strand, and its one sentence. The geocentric literature reaches this topic once. In Galileo Was Wrong (2006 PDF, 1,160 pp.), ch. 12, printed p. 794, Robert Bennett’s scored table of experiments carries a row for Galileo’s tidal argument, and the Geocentric Response cell proposes that “the ether flow varies with latitude, which is the direct cause of the two tidal bulges” — under an explicit “Caution” that the lunar cause “may be only be indirect”, and immediately after the concession that the Moon’s motion around the Earth is the one motion on which geocentrism and modern physics agree. The row is scored 0 in every column, against a printed key in which 0 means “neutral or does not apply”. That single hedged sentence is the closest thing to a sky-driven tidal mechanism located in any of the tide passages searched for this entry, and it is not a torque.

SURFACE (weak — do not use). “The Moon causes the tides; this is settled; next.” Anyone who opens this way is about to be shown a cotidal chart. The picture implied by that sentence — two bulges of water standing under the Moon and sweeping westward around the globe — is not what the ocean does, it is not what oceanographers teach, and it has been known to be wrong since Laplace. Newton’s equilibrium theory predicts a tide of well under half a metre; the Bay of Fundy gets fifteen. Leading with the bulge concedes the field.

DEEPER. All four observations in this cluster are correct and none is obscure. There really are points in every ocean basin where the tidal range falls to nearly nothing while the tide rotates around them. Basins really do resonate, and the resonance is what produces the world’s great tidal ranges. Lakes really do oscillate for hours after a squall with no celestial help at all. And high water really does not arrive when the Moon is overhead: the lag runs from about three hours in deep water to eight at New York, and in places it exceeds a day. A defender who says only this has said nothing an oceanographer would dispute.

KERNEL. The strongest form is sharper than any of that, and it is the version worth answering. It runs: the local behaviour of the tide is very nearly decoupled from the Moon’s instantaneous position. The phase is set by the shape of the basin, not by where the Moon is; the amplitude is set by how near the basin’s natural period lies to the forcing period; and the decoupling can go all the way — at Tahiti the tide follows the Sun and not the Moon, with high water day after day at about noon and midnight, which is what Melville reported in 1847 and what the Encyclopædia Britannica prints today. If the phase, the amplitude and in the limiting case even the driver can all be supplied by the water and the basin, what is the Moon still doing in the account? Every clause of that is true, and the last question is a fair one.

Why it doesn’t save the claim.

Because what the basin controls is phase and amplitude, and what it cannot touch is frequency. A driven oscillator can lag its driver by any angle you like, and if its natural period is longer than the forcing period it will invert outright and run half a cycle behind — which is precisely why high water arrives hours after the Moon’s transit and why the lag differs from port to port. What no forced system does is oscillate at a frequency its driver does not supply.

So look at the frequencies. A. T. Doodson expanded the tide-generating potential into 388 harmonic constituents in 1921, and he built them out of six astronomical arguments — mean lunar time, the Moon’s longitude, the Sun’s longitude, the longitude of lunar perigee, the longitude of the ascending lunar node, the longitude of perihelion — taken from the ephemerides, before any tide gauge was consulted. Harmonic analysis of a real record returns power at those lines and, to first order, nowhere else. The largest is M₂ at 12.4206012 h, which is half a lunar day. The second is S₂ at exactly 12 h, which is half a solar day.

Now take the kernel’s own best case. Tahiti follows the Sun because it sits near a place where the lunar line nearly vanishes, leaving the solar line behind — and the solar line is 12.0000 h, which is why high water lands at the same clock time every day. The most extreme instance of the tide “not following the Moon” turns out to be the ocean sorting itself into exactly the two frequencies the Sun and the Moon provide, and choosing the other one. That is not the tide escaping the sky. That is the tide being nothing but the sky, filtered through a basin.

The concessions first, because there are four and they are not small. Newton’s equilibrium theory of tides is quantitatively wrong — it yields a tide of under half a metre where fifteen metres occur — and Rowbotham and Scott are right that this was the weak joint in the Principia. Laplace’s tidal equations, which replaced it in 1775, have no general closed-form solution, so every global tide model is a numerical one that assimilates satellite altimetry. A quarter to a third of the tidal energy budget was only located in 2000, and where it finally goes in the abyss is argued about now. And the appearance of a tidal bore on a given morning, in the words of the UK National Tidal and Sea Level Facility, “cannot be predicted with certainty”. Any answer to this cluster that begins by pretending the tides are a solved problem deserves to lose.

What the verdict ranges over. Not “the tides are simple.” The claim on the table is that four named phenomena — nodes, resonance, seiches, bore timing — count against a lunar cause and for a sky-driven or Earth-driven one. Three of the four are the machinery by which lunar forcing is converted into what a tide gauge actually records, and the fourth is a control that separates the two hypotheses cleanly and comes down against the list.

1. The force in question does not lift water, and the objection that it cannot is aimed at nobody

Rowbotham’s 1881 chapter asks “how can the lesser power overcome the greater?”, and Dubay’s proof 117 restates it: the Earth is 87 times the more massive, so the Moon cannot pull the ocean up. Nothing in tidal theory claims it can. The tide-raising acceleration is a difference — the Moon pulls the near ocean slightly harder than it pulls the Earth’s centre, and the far ocean slightly less — and it falls off as the cube of distance rather than the square. Recomputed here on 2026-08-10: 2GM_☽R_⊕/d³ = 2 × 6.674×10⁻¹¹ × 7.342×10²² × 6.371×10⁶ / (3.844×10⁸)³ = 1.10 × 10⁻⁶ m/s², which is 1.1 × 10⁻⁷ of surface gravity. It could not raise a bathtub. What it does is act sideways: the horizontal component sweeps water across thousands of kilometres of ocean floor into a heap, and it is the heap, not the lift, that arrives at the coast.

That same arithmetic gives the one number in this subject that is a genuine prediction rather than a fit. The solar tidal acceleration is 5.05 × 10⁻⁷ m/s², so the Sun’s tide-raising force is 46% of the Moon’s — a figure that comes out of two masses and two distances measured by astrometry, with no reference to any sea. It is why the semidiurnal solar constituent S₂ is roughly half M₂ in the open ocean, and why the equilibrium estimate of spring range against neap range, (M₂ + S₂)/(M₂ − S₂), comes out at 2.7. Get the Moon’s distance wrong and that ratio breaks.

2. Tidal nodes are a fingerprint of the Earth’s rotation

An amphidromic point — the item’s “tidal node” — is a location where high and low water differ by almost nothing while the cotidal lines pinwheel around it, so that every hour of the tidal cycle occurs somewhere on the circle at once. William Whewell arrived at them by extrapolating cotidal lines inward from the North Sea coasts and finding that the lines had to meet.

Why they exist is the point. A tidal wave running up a channel on a rotating planet is deflected sideways by the Coriolis force, piling against one wall on the way in and the other on the way out; the two Kelvin waves interfere and the result is a rotary system with a null at its centre. The direction of the pinwheel is not free: it “is in the direction of the Coriolis force; anticlockwise in the northern hemisphere and clockwise in the southern”. Set the Earth’s rotation to zero and there is no Coriolis deflection, no Kelvin wave, and no amphidromic point — the cotidal charts the item points at would be blank. Item 233 offers as evidence against terrestrial motion a structure whose existence, and whose handedness in each hemisphere separately, is produced by terrestrial motion.

3. Tidal resonance is lunar-forced in the literature’s own sentence

Resonance is real and it is what makes the great tides great. The classic case is the Bay of Fundy, and the classic paper is Garrett’s in Nature in 1972, whose abstract reads in full: “Tidal ratios outside and inside the Bay of Fundy/Gulf of Maine system fall into two distinct groups. The period of the free mode oscillation in the bay is 13.3 h, and this is forced by the lunar M₂ tide of the North Atlantic to produce the unusually high tides recorded in the area.”

Read the second sentence twice. The bay supplies a natural period of 13.3 h. It supplies no oscillation at all until something drives it, and the thing that drives it is named. Resonance is an amplifier, and an amplifier does not choose the station: it takes what the aerial gives it and makes it louder. Fundy’s free mode at 13.3 h sits close enough to 12.42 h to give a large response, and that near-miss is the whole explanation of the range Garrett calls “unusually high”. What the bay cannot do is generate a 12.42 h signal on its own, because 12.42 h is not its period. Item 245 takes the amplifier for the transmitter.

4. Seiches are the control, and the control comes out the other way

A seiche is a free standing wave in an enclosed basin: tip the water with a squall or a pressure gradient, remove the forcing, and the basin sloshes at its own period until friction kills it. The period follows Merian’s formula T = 2L/√(gh), which contains nothing but the length and depth of the lake. Computed here from mean depths: Lake Geneva, 73 km long and about 154 m deep, gives 62 minutes against Britannica’s observed uninodal period of about 74 minutes — the right size, with the shortfall attributable to using a mean depth for a basin that is not a rectangle. Lake Erie, 388 km long and about 19 m deep, gives 15.8 hours. Two lakes, two periods, differing by a factor of fifteen, each set by its own geometry.

Now put the ocean beside them. Harmonic analysis of tide records from basins of wildly different length, depth and shape returns a line at 12.4206012 h in all of them — the same period, to seven figures, in basins whose free periods have nothing in common. A free oscillation cannot do that; a lake answers only to its own geometry. Only a driven system can, and the driver has to be something every basin shares.

NOAA makes the same comparison from the other end, and its verdict is worth quoting because the movement’s claim is that lakes have no tides. They do. “True tides — changes in water level caused by the gravitational forces of the sun and moon — do occur in a semi-diurnal (twice daily) pattern on the Great Lakes”, but the spring tide is “less than five centimeters (two inches) in height”, because the tide-raising force acts across a whole ocean and a lake is not wide enough to accumulate much of a difference. And the wind-driven seiche, whose half-period of four to seven hours is “very similar to the six-hour time period of the tides on the ocean”, is “frequently mistaken for a tide”. Item 251 is that mistake, presented as a discovery.

5. Bore timing: the lag is the theory, and the bore is locked to the Moon’s phase

High water does not arrive under the Moon. The delay is called the lunitidal interval, its value at new and full moon has been tabulated for centuries as the “establishment of the port” — “the time that elapses between the moon’s transit across the local meridian at new or full moon at a given place and the time of the next high water at that place” — and mariners used it to predict high water before tide tables existed, by taking their port’s fixed figure and adding the Moon’s daily fifty-minute slip. A quantity that is a stable property of each harbour, published in almanacs, and computed from the Moon’s transit, is not evidence that the tide has come loose from the Moon. It is the phase lag of a forced oscillator, and it is the observation Laplace’s dynamic theory was built to account for.

Bores make the lunar lock about as visible as it gets. They occur only on the flood, only where a big range funnels into a narrowing shallow channel, and, in the National Tidal and Sea Level Facility’s words, the Severn bore is “generally at its best when a very high spring tide is predicted”, with the spectacular ones “at spring tides in the months around the equinoxes”. Spring tides are by definition the tides of new and full moon, when the solar and lunar constituents come into phase. So the most famous bore in Britain runs best on the days the Sun, Earth and Moon line up, and the arrival times published for it up and down the river are quoted as offsets from the predicted high water at Avonmouth — a prediction made from the harmonic constituents of exactly this astronomy.

The single hardest case is Tahiti, where the tide really does ignore the Moon — Britannica: “the tide at Tahiti, for example, follows the Sun and not the Moon — the time of high water occurring, day after day, at about midnight and noon and that of low water at about 6 am and 6 pm.” Winship printed that observation at p. 131 of Zetetic Cosmogony, quoting Melville’s Omoo, and treated it as decisive. It is decisive, in the other direction. Tahiti sits where the lunar semidiurnal tide is nearly extinguished, so what remains is the solar semidiurnal tide, whose period is exactly 12 h 00 min — which is why high water is at the same hour every day. The explanation of the anomaly is items 233 and 245: an amphidromic null, plus the relative response of the basin at the two forcing frequencies. Three of the four items in this cluster are the anomaly, the explanation of the anomaly, and the mechanism of the explanation, filed separately as three proofs against it.

6. The frequencies settle it, and this is where the list’s own mechanism dies

Whatever drives the tides has to supply the periods that are in the record. Take the two largest. M₂ = 12.4206012 h, half a lunar day. S₂ = 12.0000 h, half a solar day. A sky that turns once a day — the rotating firmament of item 233, or the ether flow of the geocentric strand, or anything else keyed to the diurnal circuit of the heavens — can supply 12.0000 h, or 11.9672 h if you key it to the stars. Neither of those is 12 h 25 min. That is not a small discrepancy: a tide on the 12.0000 h clock and a tide on the 12.4206 h clock drift into complete opposition in 7.38 days and back into step in 14.765 days.

And that predicted beat is in the record. The fortnightly swing between spring and neap range is the beat of those two lines against each other; recomputed here, 1/(1/12.0000 − 1/12.4206012) = 354.367 h = 14.7653 days, against half a synodic month of 29.530588/2 = 14.7653 days. Six significant figures. Be exact about what that shows and what it does not: given that M₂ is half a lunar day and S₂ half a solar day, the agreement is an algebraic identity, not an independent measurement. The evidence is that the ocean has a fortnightly modulation of that period at all, and that its maxima fall on the days of new and full moon rather than at some arbitrary phase. A sky-driven tide has one clock. The sea keeps two, and the second one is the Moon’s.

The record contains slower lunar signatures too, and they are not subtle. The Moon’s orbital plane precesses with an 18.61-year period, which modulates the tidal constituents; Peng and colleagues examined tide-gauge records worldwide and found that cycle changes high water levels by up to 30 cm. A whirlpool at the pole, an atmosphere pressing on a floating plate, or an ether flowing with latitude has no reason to know about the regression of the lunar nodes.

There is exactly one way for a sky-driven model to produce 12 h 25 min, and it is fatal to take it. The Moon circuits the sky once every 24 h 50 min — slower than everything else up there — so a mechanism keyed to the Moon’s passage overhead would deliver the right period at once. That is the escape, and it is not new: David Wardlaw Scott took it in 1901, suggesting that the Moon “may influence the atmosphere, increasing or diminishing its barometric pressure, and indirectly the rise and fall of the Earth in the waters”, and adding in the same breath “Of this we cannot now be sure”. But whoever takes it has conceded the case. The whole point of these four items is that the tide is not keyed to the Moon; a model that reproduces the tidal period by keying it to the Moon has agreed that the Moon sets the beat and is arguing only about the coupling. There is no third option, because there is nothing else in the sky with a period of 24 h 50 min.

7. The proposed alternatives, answered on their own hedged terms

The floating plane. Rowbotham’s model is the only fully worked non-lunar mechanism in the tradition, and it has to be met where its author left it defended. His stated consequences are narrower than they are usually made to sound. He does not predict high water simultaneously along a coastline; what he writes is “The simultaneous ebb and flow upon meridians 180° apart” — and that is not a discriminating prediction at all, because the equilibrium tide of the theory he is rejecting also puts high water at both ends of the Earth–Moon line at once. And he pre-empts local disagreement in the same paragraph, tracing “Whatever peculiarities are observable” to “the reaction of channels, bays, headlands, and other local causes”.

Take that escape at face value — it is his, and he is entitled to it — and it costs him the model’s only quantitative content. Once channel reaction may supply any phase at any place, the floating plane predicts nothing whatever about when high water arrives anywhere. What is left in its own voice is the period, fixed by the size of the floating body: a laden ship fluctuates several times a minute, “the Earth once only in about twelve hours, because it is proportionately larger”. That is the prediction, and it is the one that fails. The ocean’s largest constituent sits at 12 h 25 min, in records from basins of every size and shape, and its range swells and slackens on a 14.77-day cycle whose maxima fall on new and full moon. The size of a plate and the buoyancy of the deep contain no route to either number, and no appeal to headlands can reach a period that is the same in every basin. His remaining consequence fails on its own terms too: he predicts no tides in lakes and inland seas, and the Great Lakes carry a measured semidiurnal astronomical tide of under 5 cm — small, as an ocean-scale differential force acting across a small basin should be, but present, and keeping the sky’s time rather than the basin’s.

The tradition’s own favourite datum belongs here as illustration and not as the refutation, and the distinction matters. Scott quotes Captain George Peacock reporting that at Holyhaven “the tide is actually falling, and running down rapidly, while, at the same moment, it is running up rapidly at London Bridge”, that the St Lawrence carries “four high waters and three low waters…at the same time”, and that the Amazon carries six or seven. Those are the signature of a progressive wave travelling up an estuary, which is what dynamic theory says a tide is. But an estuary and a river are exactly the class Rowbotham scoped out, so they cannot be turned against him as a disproof without doing to him what this project objects to being done to it. What they do show is how much the escape concedes: Scott introduces them by granting that tides “vary greatly in height, owing chiefly to the different configurations of the adjoining lands”, which hands the basin precisely the role dynamic theory gives it, and leaves the plate with nothing to explain but a period it gets wrong.

The part of the model that is right is the part its authors did not realise they had. The Earth does rise and fall. The solid ground flexes with an M₂ vertical amplitude of 384.83 mm, reaching about 55 cm at the equator, and it was first detected by the very observations Winship reprinted — Rebeur-Paschwitz’s horizontal pendulums at Wilhelmshaven, Potsdam and Tenerife between 1889 and 1891, which Fréchet and Rivera record as the occasion on which “he succeeded in observing the luni-solar tide for the first time”. The clipping Winship chose to print in capitals says the ground oscillations correspond with the course of the Moon. It is the correct observation attached to the wrong cause, and the instrument that made it was reading the Moon.

The latitude-varying ether. The geocentric proposal at printed p. 794 of Galileo Was Wrong is that “the ether flow varies with latitude, which is the direct cause of the two tidal bulges”. Take it seriously and it does not produce a tide. A flow that is a function of latitude alone is axially symmetric about the Earth’s axis and, in the Earth-fixed frame the model insists on, steady. A steady axisymmetric field at a fixed station is a constant: it can give a permanent equatorial swelling, and it cannot give an oscillation twice a day, because there is nothing in it that varies as the station goes round. To get two bulges sweeping past an observer you need something that breaks the axial symmetry and moves relative to the ground — which is to say, you need a Moon. The sentence is also hedged twice over: it is prefaced “Caution”, it says the lunar cause “may be only be indirect”, and it comes immediately after the concession that the Moon’s motion around the Earth is the one motion on which geocentrism and modern physics agree.

8. What is left, stated without decoration

The tides are a hard, partly open problem in physical oceanography, and this page will not pretend otherwise: no closed-form solution, models that assimilate data, agreement with pelagic gauges of about 0.9 cm in the deep ocean but 5–6.5 cm on shelves and coasts, and roughly a terawatt of dissipation whose fate in the abyss is unsettled. What is not open is the driver. The frequencies in the record are the Sun’s and the Moon’s, down to the 18.61-year regression of the lunar nodes; the amplitude ratio between them matches an inverse-cube law computed from astrometry; the nodes and the rotary sense around them are made by the Earth’s spin; the resonances are described in the primary literature as lunar-forced; and the solid Earth itself rises 38 cm on the same clock. Nothing here discriminates between a moving Earth and a stationary one in the list’s favour. Two of the four items — the nodes and the seiches — discriminate in the other direction: one requires the Earth to turn, and the other shows what a basin does when nothing in the sky is driving it.

The list says

Tidal resonance geocentric.

The source says

“…attributing the tides to the Moon’s motion around the Earth. (The only motion with which Geocentrism and modern physics agree is the motion of the Moon around the Earth.) Caution: Even here the lunar cause of tides may be only be indirect; the ether flow varies with latitude, which is the direct cause of the two tidal bulges!”

A concession was re-used as a proof. The source scores this argument as worth nothing to geocentrism, and the list publishes it as a proof. The cell above is the Geocentric Response in Robert Bennett’s scored table of experiments at printed p. 794 of the 2006 Galileo Was Wrong PDF. The row it belongs to — Galileo’s tidal argument — is marked 0 in every column, against a key printed on the same pages in which 0 means “neutral or does not apply”. Twenty-six pages later the same chapter says outright that “the Galilean arguments for the phases of Venus, the moons of Jupiter and the tidal flows support both heliocentric and geocentric views”. A row its own author scored as non-discriminating arrives on the list as item 245. (The doubled “be” is in the original.) The identification is ours, and it is an inference. The p. 794 row is the one place in the located tide passages where the book scores a tidal argument, but its cell says nothing about resonance — it answers Galileo’s sloshing-vessel argument — and the one extended resonance passage located in that PDF, Hoyle on the atmosphere at printed pp. 633–634, is about tidal braking of the Earth’s rotation and is treated in the length-of-day cluster. If a reader rejects the identification, the consequence is that no source was located for item 245 either, which is a finding of the same weight.

Each of the other three items drifts differently, which is why one enum value cannot carry the block. Item 233, “Tidal nodes sky torque”: a sky- or firmament-driven torque on the oceans is not located in any of the tide passages searched for this entry — Rowbotham 1865 Section 10, Winship 1899 pp. 130–131, Scott 1901 pp. 258–261, Dubay 2015 proofs 117–118, the Flat Earth Society wiki’s Tides page as served on 2026-08-10, or any of the 25 pages of the 2006 PDF carrying the strings “tide” or “tidal”. The closest text is the ether sentence quoted above, which is a flow and not a torque, and which is hedged. On the located record the phrase is the list’s own; that is an unsourced_addition. Item 252, “Tidal bores moon visual timing”: this one does not drift on the mismatch. Winship asserts it flatly, in his own voice, on p. 130 — “But the tides are not uniform. At Port Natal the rise and fall is about six feet, while at Beira, about 600 miles up the coast, the rise and fall is 26 feet. This effectually settles the matter that the moon has no influence on the tides.” — prints Melville’s Tahiti observation on p. 131 as proof of it, and Dubay’s proof 118 carries it forward as “the Earth’s tides vary greatly and do not follow the Moon”. What the list drops is the concession Winship attaches at the end of the same passage, that the Moon is nonetheless “the TIMEKEEPER for the tides…The ‘phase’ of the moon tells what kind of a tide may be expected”, and Scott’s “at present only conjectured…Of this we cannot now be sure.” That is hedge_dropped, not a reversal. The bores are a separate matter: no tidal-bore passage is located in any of the texts listed above. Item 251, “Seiche oscillations periodic”: seiches are not located in any of those texts either; the term enters here from oceanography, not from the tradition.

The refutation above answers the sources, not the fragments. It takes Rowbotham’s mechanism at full strength — atmospheric pressure depressing a floating plane, with the consequences he derives from it and the local-causes qualifier he attaches — and answers it where that qualifier cannot reach, at the period; it takes Bennett’s ether sentence at full strength and shows that a latitude-only flow is steady and axisymmetric in the frame his model requires, so it cannot oscillate twice a day; and it concedes what Rowbotham and Scott got right about Newton’s equilibrium theory before answering them. The gap is still worth publishing, because the fragments are what circulates: the tradition’s own texts call the Moon the timekeeper, call the mechanism conjectural, and score the geocentric version neutral, and none of those three qualifications survives into the four one-line items.

Proposed instead: REFUTED

NOT DEMONSTRATED rests on the basis line “No mechanism specified”, and the texts located for this pass do not support that. Rowbotham's 1865 Section 10 specifies a mechanism in detail — atmospheric pressure depressing a plane earth floating on the deep, with the buoyant reaction restoring it — and derives four consequences from it: simultaneous high water on meridians 180 degrees apart, no tides in lakes, tidal wells fed by subterranean passages, and a period of about twelve hours set by the size of the floating body. Winship 1899 pp. 130-131 and Scott 1901 pp. 258-261 restate it; the Christian Flat Earth Ministry page of 2015 adds a second mechanism, a polar vortex reversing every six hours; and Sungenis and Bennett offer a third, an ether flow varying with latitude. The problem is not that no mechanism was proposed. It is that four incompatible ones were, and that each makes a checkable prediction which fails even after its author's own qualifier is granted. Rowbotham scopes local disagreement out in advance — “Whatever peculiarities are observable they may be traced to the reaction of channels, bays, headlands, and other local causes” — so the estuary and river data the tradition itself likes to cite, at Holyhaven, on the St Lawrence and on the Amazon, cannot be used against him and are not used against him here. What the qualifier cannot reach is the period, which his model fixes by the size of the floating body at “once only in about twelve hours”: the same 12 h 25 min line turns up in tide records from every basin whatever its size or shape, the range peaks at new and full moon, and neither a plate, a vortex nor an axisymmetric ether flow has any route to that. Lakes do carry a measured semidiurnal astronomical tide under five centimetres, which he predicted they would not. Separately, items 251 and 252 are not mechanism claims at all but observational ones — about seiches, and about high water not arriving under the Moon — and both are answered by measurement rather than left undemonstrated. Where a stated mechanism makes a wrong prediction, the rubric's word is REFUTED. If a reviewer instead rejects the identification of item 233's “sky torque” with the Sungenis and Bennett ether sentence, the consequence is that no source was located for the cluster's name, which argues for renaming the cluster rather than for keeping the present verdict.

The verdict on the scorecard is unchanged until the question is settled. Publishing the disagreement is the point: a rubric that never produces dissent is not being applied.

If Earth and its atmosphere were constantly spinning eastwards over 1000mph then North/South facing cannons should establish a control while East-firing cannonballs should fall significantly farther than all others while West-firing cannonballs should fall significantly closer. In actual fact, however, regardless of which direction cannons are fired, the distance covered is always the same.

— 200 Proofs Earth Is Not a Spinning Ball (2015), Proof 24 of 200 Proofs Earth Is Not a Spinning Ball, transcribed from the archive.org OCR text of the PDF (item 200ProofsEarthIsNotASpinningBall), where the numeral is rendered “lOOOmph”. Identified by its neighbours: proof 23 is the atmosphere-dragging item and proof 25 is the eastbound airliner. Not checked against the 2018 print edition, which may renumber

Read what the claim commits to. It is not vague. It names a control (north/south fire), a signal (an east–west range difference), and a result (there is none). That is a testable proposition about gunnery, and it is the only member of this cluster’s ancestry that specifies the rotation rather than the Earth’s motion in general. Everything below answers it at that strength.

Where the word “artillery” comes from, because it is not from here. Dubay writes “cannons”. The list writes “artillery”, twice, and the vocabulary tracks a second and later book: Edward Hendrie, The Greatest Lie on Earth (2016), chapter 9, “No Coriolis Effect Proves a Stationary Earth”, at printed pp. 105–114, with the passage the items compress at pp. 113–114. Hendrie states it without a hedge anywhere in it: “you will look in vain for any mention of Coriolis effect in any military artillery or sniper instruction manual”, and that in all the wars of history “no soldier has ever been instructed to consider the Coriolis effect of a spinning Earth” when laying a gun. He adds a personal warrant — he is a former federal firearms instructor and has never seen a round affected by it, nor read of it in a firearms manual.

The footnotes are the shortest route to the answer. That passage carries two notes. One cites FM 23-10, Sniper Training (17 August 1994) and a Navy SEAL sniper program. The other, attached to a claim about every war ever fought, cites four artillery manuals: Bethel’s Modern Artillery in the Field (1911), Instruction for Field Artillery (1860), the Artillerist’s Manual (1863) and Roberts’s Hand-Book of Artillery (1863). Three of the four are American Civil War-era and the newest is from 1911 — that is, the evidence offered for the state of gunnery is a shelf assembled entirely before the ranges at which the effect matters were reached.

The argument is far older than either book, and its ancestor is better made. In the Almagestum Novum (Bologna, 1651), the Jesuit astronomer Giovanni Battista Riccioli set out 77 arguments against the motion of the Earth; numbers 17 and 19 are cannon arguments. If the Earth turned, he reasoned, a ball fired toward the pole would be carried off, because “on parallels nearer the poles, the ground moves more slowly, whereas on parallels nearer the equator, the ground moves more rapidly”, and a ball fired north would strike its target more weakly than one fired east. He credits the cannon experiment to Tycho and argument 19 to his colleague Grimaldi; Dechales repeated it in 1674. Riccioli derived his effect from theory rather than asserting it from a shelf of manuals, and Christopher Graney, whose English rendition is used here, puts the identification carefully: the argument appears to be an early description of what is now called the Coriolis effect. Two centuries before Coriolis, a geocentrist had the physics of this cluster essentially right.

What this passage is being cited as. The reachable modern statement of the rotation-specific ballistics claim, in the compilation this 461-item list is padded out of. It is not evidence of origination: the argument is older than anyone on the People tab, and no single author is credited for it here.

SURFACE (weak — do not use). “Snipers correct for Coriolis, so the claim is simply false.” This loses on the defender’s home ground. Searching the archive.org OCR text of FM 23-10, Sniper Training (1994) — the manual Hendrie’s own footnote cites — returns no occurrence of the word Coriolis, and none of the word rotation in any relevant sense. Anyone who opens with this is contradicted by the document the other side supplied.

DEEPER. “Long-range artillery corrects for it.” True, and still incomplete, because it invites the obvious reply: your effect is conveniently absent everywhere an ordinary person could check it, and present only in tables produced by the same institutions you are asking me to trust about everything else. A defender who is quick will get there in one move.

KERNEL. The strongest form is Riccioli’s, not Hendrie’s, and it is a real argument. It runs: a rotating Earth makes a specific, derivable prediction about where a cannon ball lands — I can derive it myself, without instruments, from the fact that the ground moves faster near the equator than near the pole. Gunners are extraordinarily good at their trade; skilled artillerymen can put a shot down the mouth of an enemy cannon. If the prediction were real they would have had to notice. They have not. The Copernican reply is that the difference falls below what we can measure — but that reply can be made about any prediction whatever, and a theory rescued that way is not being tested. Every step of that is honest, and the last sentence is a serious point about falsifiability that the seventeenth century had no way to settle.

Why it doesn’t save the claim.

Because the “too small to measure” reply was not a rescue — it was a quantitative prediction with a date on it, and the date arrived.

The deflection is not a free parameter. On a flat trajectory it is D = Ω sin φ × R × T: the Earth’s rotation rate, the sine of the latitude, the range and the time of flight, and nothing else. That formula says the effect must be invisible at Riccioli’s ranges and unavoidable at long ones, and it says by how much. Riccioli’s own case is a ball crossing 250 paces in two seconds; at Bologna’s latitude that is a deflection of three to four centimetres, on a shot from a smoothbore whose round-to-round scatter was measured in metres. He was not wrong to see the effect. He was wrong about one number, and the Copernicans he dismissed had that number right.

So the argument carries its own falsification schedule. As soon as ranges and times of flight grew, the term crossed the dispersion of the weapon — and at that point gunners started correcting for it, which is exactly what the theory Riccioli was attacking predicts and what his own theory forbids. By 1918 the Paris Gun was throwing shells 120 km in about three minutes, where the same formula gives well over a kilometre of drift. The tables followed the arithmetic, not the other way round.

Start with what is conceded, because two of the three concessions are permanent.

One. The word “Coriolis” does not occur anywhere in the text layer of the April 2016 edition of the US Army’s cannon gunnery manual searched here. That manual calls the thing rotation of the earth throughout, as does FM 6-40 in 1945. What the rest of Army doctrine calls it was not checked for this entry.

Two. The small-arms half of the claim is true. Searching the archive.org OCR text of FM 23-10, Sniper Training (17 August 1994) — the manual the source’s own footnote cites — returns no occurrence of the word Coriolis. A firearms instructor who says he has never seen a round visibly affected by the Earth’s rotation, and never heard it discussed on a range, is reporting something accurate.

Three. The popular-science passage the source attacks deserves it. The National Geographic explanation quoted in that chapter says that a pilot flying “in a straight line” from Portland, Oregon would end up near New York or Pennsylvania. That is a great circle drawn on a Mercator projection being reported as a Coriolis deflection — a conflation the Coriolis-force reference article cited below warns against explicitly. Attacking it is fair.

What the verdict ranges over. Not “the effect is large”, and not “everyone who points a weapon corrects for it.” The cluster’s claim is that gunnery ignores the Earth’s rotation, and that this is evidence the Earth does not turn. Both halves fail, and they fail for the same reason: the size of the correction is fixed by the rotation rate, and gunners apply it exactly where that size exceeds the accuracy of the weapon.

1. The table exists, and it is indexed by latitude

Open TC 3-09.81, Field Artillery Manual Cannon Gunnery, Headquarters, Department of the Army, dated 13 April 2016 — the same year the source’s book was published. Paragraph 3-63 lists the deviations from standard conditions that firing data must be corrected for. Under range effects: muzzle velocity, projectile weight, range wind, air temperature, air density, rotation of the earth, propellant temperature. Under deflection effects: drift, crosswind, rotation of the earth.

Chapter 7 says how. Table H gives “the correction to range in meters for the rotation of the earth at 0° latitude”, entered with the range and with “the exact azimuth (to the nearest mil) to the target”, and multiplied by a factor taken from a latitude table beneath it. Table I gives “the correction to deflection in mils, for the rotation of the earth”, and there are “tables for every 10° latitude starting from 0° north or south latitude to 70° north or south latitude”. Northern latitudes are entered along the top of the table; for southern latitudes “you enter from the bottom”. The manual’s own worked example instructs the student to “Determine corrections for azimuth to compensate for rotation of the earth from Table I” and then to “Select the appropriate Table I on the basis of latitude (30°N).”

Read that structure rather than just the fact of it. The correction depends on where on the Earth the gun is standing, on which way it is pointing, and it is entered from opposite edges of the table in the two hemispheres. Those are three dependencies that a rotating sphere predicts and that nothing else in the gunnery problem — not wind, not air density, not propellant temperature, not the rifling — produces.

And this is not new doctrine. FM 6-40, Field Artillery, Gunnery, War Department, 1945, already tells the computer that firing-table factors correct for, among other things, “rotation of the earth tables for direction and range”, and instructs him to list the “altitude (and latitude) of battery, direction of fire” and to take “range and deflection effects due to rotation of the earth for the latitude of the battery (long range weapons only).” That parenthesis is not a hedge in our favour; it is the whole physics of the case, printed in 1945, and section 3 below is about it.

2. The passage above is the manual’s own worked example

The claim in the passage is that east-firing and west-firing cannon show no range difference. Paragraph 7-22 of the 2016 manual sets out the theory behind Table H in precisely those terms: because of the rotation of the earth “a point on the equator has an eastward linear velocity of approximately 457 meters per second”, decreasing to zero at either pole; a gun on the equator firing east will have its projectile “impact east of the target (over the target in this case)”, and one firing west will impact “short of the target”. Firing east goes long, firing west goes short, and the manual tabulates by how much. That is the passage’s experiment, performed, quantified and issued as doctrine.

The size is the point. Two different things are being run together in the claim, and separating them settles it:

• The version the passage argues against. If the ground’s 1,000 mph (447 m/s) were added to an eastward shot and subtracted from a westward one, then at a howitzer’s muzzle velocity of roughly 684 m/s, with range going as the square of velocity, the east/west range ratio would be about 23 to 1. Nothing of the sort is observed. The passage is right, and the reason is Galilean: gun, ball, target and air all share the rotation, so it cancels out of the leading term. That has been the answer since 1632 and this review carries it at ARG-A17.
• The version that survives. What does not cancel is the small residual from the rotation of the frame itself. For a 155 mm shell out to 22 km with a 65-second time of flight at 45° latitude, the vertical component works out to an acceleration of about 0.035 m/s², which shifts the range by roughly 78 m — about 0.36% of it, in opposite directions for eastward and westward fire. That is Table H. Put the two side by side at the same range: the naive version predicts an eastward shot going about 60 km against a westward 2.6 km, an east/west difference of tens of kilometres, where the real difference is the 78 m taken twice — about 150 m, smaller by a factor of several hundred. And still far above what a battery can afford to leave out.

(Both figures recomputed here on 2026-08-11 from the standard expressions; the lateral formula used throughout is cross-checked below.)

One limit on all of this, stated rather than buried: Table H and Table I are reproduced in the manual as figure images and were not read here. The numbers above are computed from the standard expressions and cross-checked in section 3 against an independently published rifle figure — not against the printed tables. If someone reads those tables and the values disagree with the arithmetic, this section is wrong and should be corrected against them.

3. Why the rifleman is right and the gunner is not

One expression explains the entire pattern of who corrects and who does not. For a flat trajectory the lateral deflection is

D = Ω sin φ × R × T,   Ω = 7.292115 × 10⁻⁵ rad/s

— the rotation rate, the latitude, the range and the time of flight. It is checked here against an independently published figure rather than asserted: for a 1,000-yard northward shot at Sacramento’s latitude the standard number is about 2.8 inches, and the expression returns 7.1 cm for a 1.7-second time of flight. Now run it across the cases:

Six millimetres at 300 m is inside the group of any rifle and any shooter; it is not that instructors suppress it, it is that it is smaller than the thing they are trying to teach. Seven centimetres at 1,000 yards is at the edge — comparable to a very good rifle’s dispersion, which is why it is long-range shooters and not riflemen generally who bother with it. A hundred and thirty-nine metres at 30 km is larger than the target, the battery and the safety fan.

So the claim is a description of the low-range end of that table, presented as a fact about the world. And the 1945 manual states the cut-off in its own words — the earth-rotation tables are for “long range weapons only.” The correction appears exactly where Ω sin φ R T crosses the accuracy of the weapon. No conspiracy, no fudge factor, and no discretion: a gunner who omitted it at 30 km would miss, and one who applied it at 300 m would be adjusting by less than the width of the bullet.

4. Answering the strongest form — Riccioli’s, not the list’s

Riccioli put the argument better in 1651 than the modern texts do, and he anticipated the reply. His rendition records it: the Copernicans either deny the effect “or concede it but claim that the differences in trajectory fall below our ability to measure”, and he judged that “the argument is strong, and this response is not.” He also gave the case its parameters — a ball of sixty or eighty pounds crossing 250 paces in two human pulsebeats.

Put his own numbers into the expression above. Two hundred and fifty paces is somewhere around 310–370 m; two seconds of flight; Bologna is at 44.5° north. The deflection is three to four centimetres. Riccioli’s guns scattered by metres. The Copernicans were not evading him; they were right, and they were right by a factor of about a hundred, and the only way to have known that in 1651 was to do an arithmetic nobody could yet do — the mathematics that puts a number on the deflection was published by Coriolis in 1835, and Riccioli died in 1671.

That is the whole history of this argument in one sentence: it was a good argument that was refuted by the growth of gun ranges. Reviving it in 2016 requires the ranges to have stopped growing, and they did not.

5. What the claim would have to explain

First, the strongest version of the reply: that the tables are decoration on a stationary-Earth solution. The point is a real one and the manual concedes half of it in print — the list of standard conditions the firing tables are built on includes NO ROTATION OF THE EARTH. The ballistic solution genuinely is computed on a motionless Earth and corrected afterwards. But the same list includes NO WIND and a propellant temperature of 70°F, and nobody reads it as a claim that wind does not exist or that powder is always at seventy degrees. The paragraph that introduces the deviations says so in terms: actual firing conditions “will never equate to standard conditions.” A baseline is not a denial. So the question is not whether the baseline is stationary — it is — but what the corrections applied to it depend on.

Suppose gunnery really did work on a stationary Earth and the tables were institutional decoration. Then a defender owes an account of four things at once: why the correction is a function of the battery’s latitude; why it is a function of the azimuth of fire; why the deflection table is entered from the opposite edge in the southern hemisphere; and why its magnitude agrees with a number computed from nothing but the length of the day. Wind depends on none of those. Air density depends on none of them. Rifling drift is a constant handed sideways, the same at every latitude. There is no second candidate that produces a latitude-and-azimuth-dependent, hemisphere-reversing correction of exactly the computed size, and none of the sources traced for this entry proposes one.

6. What is left, stated without decoration

The rotation of the earth is a named, tabulated correction in US Army cannon gunnery in 1945 and in 2016. It is indexed by latitude and azimuth, reverses across the equator, and is published in mils — the artillery angular unit that one of these three list items names. The claim that it is absent is true only of small arms, where the arithmetic says it should be, and is stated in the sources as a claim about everything. This argument was better made in 1651, and it lost to nothing more exotic than longer guns.

The list says

Ballistics ignore rotation. / Artillery stationary calculations. / Artillery geocentric mils.

The source says

“In actual fact, however, regardless of which direction cannons are fired, the distance covered is always the same.” (Dubay, proof 24) — and “you will look in vain for any mention of Coriolis effect in any military artillery or sniper instruction manual … no soldier has ever been instructed to consider the Coriolis effect of a spinning Earth” (Hendrie 2016, ch. 9, p. 113).

The source does not contain this. Take the middle item first, because it is the unusual result. “Ballistics ignore rotation” does not overstate its sources. If anything the traffic runs the other way: Dubay asserts the range is “always the same” whatever the direction of fire, and Hendrie asserts it of any military manual and of every war ever fought, with a personal warrant attached. There is no hedge here to drop. This review has found the compressed version firmer than the original in almost every argument it has checked; on this one the compression is the mildest statement in the chain, and the honest finding is that the source is where the overreach lives.

Of the other two items, one is an addition and one is not. “Artillery geocentric mils” is the list’s own vocabulary: the word mils is not located anywhere in the archive.org OCR text of either book searched for this entry, and whatever the phrase was meant to convey — the wording does not settle it — it did not come from these sources. “Artillery stationary calculations” is not an addition. It is not in proof 24 of 200 Proofs, which is where a check against Dubay alone would stop; it is in Hendrie, at p. 113, where the artillery officers who on a spinning Earth would be “trained to consider the spin of the earth in making their calculations for accurate firing” are said not to be. Four words for one sentence, and nothing added.

And this is the part worth the reader’s time. Both statements are true, and both convict the cluster they sit in. Firing tables really are computed on a stationary Earth: NO ROTATION OF THE EARTH is one of the standard conditions listed in figure 7-1 of TC 3-09.81 (2016), in the same list as NO WIND, which nobody reads as a denial that wind exists. And the artillery angular unit really is the mil: it is the unit in which Table I publishes the correction to deflection for the rotation of the earth, tabulated for every 10° of latitude from 0° to 70° north or south. The two details that make the claim sound technical are the two details that answer it.

The refutation answers the sources, not the fragments: it grants Dubay’s east–west null at full strength and explains it (Galilean: the naive version predicts an east/west range ratio near 23:1, where the real residual shifts the range by about 0.36%), grants Hendrie’s sniper manual and his firearms testimony outright, grants that the word Coriolis is not located in the 2016 artillery manual searched for this entry, and then answers the strongest version of the argument — Riccioli’s 1651 one — on its own ground.

The aeronaut is able to start in his balloon and remain for hours in the air, at an elevation of several miles, and come down again in the same county or parish from which he ascended. Now, unless the Earth drag the balloon along with it in its nineteen-miles-a-second motion, it must be left far behind, in space: but, since balloons have never been known thus to be left, it is a proof that the Earth, does not move, and, therefore, a proof that the Earth is not a globe.

— One Hundred Proofs that the Earth Is Not a Globe (1885), Proof 54, carried in Carpenter's own index of proof titles as “54 Balloons not left behind.” Text as transcribed at Project Gutenberg ebook #55387 (Baltimore: printed and published by the author, 1885). The stray comma in “the Earth, does not move” stands as it does in that transcription and was not checked against a printed copy. Companion: proof 49, the same argument run on clouds instead of balloons.

The premise is in the sentence, and so is its dismissal. Carpenter does not overlook the possibility that the air travels with the Earth — he names it, in the one clause the whole proof turns on: “unless the Earth drag the balloon along with it”. Then he disposes of it in the same breath, on the ground that balloons are not observed to be left behind, which is the observation the condition predicts. Quoting him up to the word unless and stopping would be the same trick the list plays in the other direction, so the whole sentence is above and the conclusion is as flat as he wrote it. His error is not a missing consideration; it is treating the confirmation of a premise as its refutation.

Nineteen miles a second is the orbit’s number, attached to the spin’s direction. Carpenter’s figure belongs to the Earth’s revolution around the Sun (about 18.5 mi/s; his 19 is right), but he attaches it to a west-to-east motion — proof 44 has “the Earth… said to move at the rate of nineteen miles in a second of time, ‘from west to east’”, and proof 49 has the globe “revolving through space from west to east at the rate of nineteen miles in a second”, the identical argument run on clouds moving “west to east being as frequent a direction as any other”. So he is not separating revolution from rotation at all; he is arguing against the motion of the Earth undifferentiated. The modern versions retarget the same picture at the daily spin specifically, at over a thousand miles an hour, and it is the modern target the list inherits: items 108 and 227 sit in the rotation lane.

The modern statements the list actually tracks. Eric Dubay, 200 Proofs Earth Is Not a Spinning Ball (2015), proof 21 — quoted in full in the compression block below — has helicopters and hot-air balloons hovering over the surface and waiting for their destinations to arrive, on the condition that the Earth spins eastward at over a thousand miles an hour. Proof 22 puts a number on it from the Red Bull Stratos jump: Baumgartner, “spending 3 hours ascending over New Mexico” — his figure; the real ascent was about two and a half — “should have landed 2500 miles West into the Pacific Ocean but instead landed a few dozen miles East of the take-off point.” Proof 158 runs it the other way, granting co-rotation for the sake of argument: if the air is carried round, the higher layers must move faster, so “hot-air balloons would also be forced steadily faster Eastwards as they ascended through the ever increasing atmospheric speeds.” That one is answered in section 2 below, because the answer to it is not the answer to proof 21. For item 237 the ancestors are proofs 124 and 125 — balloon footage showing “a clear hot-spot reflecting on the clouds directly below the Sun’s spotlight-like influence”, and sunbeam angles traced back to a source “relatively close to Earth just above the clouds”. Quotations from the Internet Archive OCR text of that book (item 200ProofsEarthIsNotASpinningBall), read 2026-08-11.

An older sighting, and it belongs to the other side. Searching the 1865 Earth Not a Globe (Simpkin, Marshall; Project Gutenberg #69892) for balloon returns thirteen occurrences: eleven in one run in the section on the apparent concavity of the sea seen from the car, one in Rowbotham’s perspective explanation of the setting Sun (used in section 3 below), and one inside a block quotation from the Literary Gazette of 1851, reporting Foucault’s pendulum, where the Gazette’s writer remarks that anyone proposing to demonstrate the Earth’s rotation by direct experiment would once have been thought “as mad as if he were to have proposed reviving Bishop Wilkins’s notable plan for going to the North American colonies in a few hours, by rising in a balloon from the Earth and gently floating in the air until the Earth, in its diurnal rotation, have turned the desired quarter towards the suspended æronaut”. The earliest text located that puts a balloon and the Earth’s rotation in one sentence is therefore a Copernican using the idea as a byword for absurdity. The attribution to Bishop John Wilkins is the Gazette’s and was not checked against Wilkins’s own works, which matters, because Wilkins argued for the Earth’s motion; nothing in this entry rests on it.

What this passage is cited as. The earliest documented statement of the cluster’s drift half, in the flat-earth literature, that this pass could locate — an ancestor, not a demonstration of origination. Whether the specimen list took it from Carpenter, from Dubay, or from the ambient stock was not established, and the cluster record still carries no originator.

SURFACE (weak — do not use). “Balloons drift with the wind, so the question is silly.” This concedes the argument’s premise without noticing. The wind is the residual; the claim is about the other 388 metres per second, and answering “wind” invites the immediate and fair reply that we have explained a 10 m/s effect and said nothing about a 400 m/s one.

DEEPER. “Gravity holds the atmosphere, so it co-rotates.” True, and incomplete in a way a defender will find at once: gravity is a central force and exerts no torque about the rotation axis, so it cannot by itself impart rotation to the air. Stopping here makes co-rotation look like an assumption bolted on to save the model.

KERNEL. The real insight in the argument is the choice of instrument, and it is a good one. A free balloon is the closest thing in ordinary experience to a Lagrangian tracer: once it is aloft it has no traction on the ground, no rails, no thrust, nothing to hold it in place. If the ground were sliding beneath the air, a balloon is precisely the object that would reveal it, and the aeronauts of the 1860s were among the first people in a position to look. The instinct is sound and the experiment was worth doing. What it establishes is that the air shares the Earth’s rotation to within the wind speed — and then, when the same instrument is read more carefully, that the wind itself is organised by that rotation.

Why it doesn’t save the claim.

Because the co-rotation is not a rescue hypothesis, it is the reading the instrument returns, and it comes with a magnitude. At Roswell’s latitude the ground and the air move east at 388 m/s; the winds a radiosonde measures are tens of m/s. The balloon experiment’s actual result is that the air travels with the Earth to about two per cent, and the two per cent is the weather.

And the residual is not noise. It has the structure rotation predicts and nothing else predicts: mid-latitude upper winds that blow west-to-east, at a speed recoverable from the Earth’s rotation rate and the measured pole-to-equator temperature difference; circulation around a low that runs one way north of the equator and the other way south of it; and an angular-momentum budget that closes against the solid Earth, so that the length of the day itself wobbles by about a third of a millisecond a year in step with the winds. The instinct — put up a free-floating object and see what the rotation does to it — was right. The reading is positive.

Concede the observation first, because it is correct. Balloons do come back down near where they went up. Felix Baumgartner rode a helium balloon out of Roswell, New Mexico on 14 October 2012, launched at 09:30 MDT, spent roughly two and a half hours ascending to a ratified 38,969.4 m (127,851 ft), and landed on his feet in eastern New Mexico at 12:17 MDT; the capsule came down about 70.5 km (43.8 mi) east of him. Nobody on either side of this disputes any of that. The whole argument lives in what follows from it.

1. The number the argument needs, and the number the physics gives

At Roswell’s latitude the Earth’s surface travels east at ΩRcosφ = 388 m/s, or 868 mph — and so does the air, and so does anything released into it. A balloon does not have to be dragged along; it is already moving, and it keeps moving, for the same reason a dropped coin lands at your feet on a moving train. Dubay’s figure of about 2,500 miles is his own stated three hours of ascent run at roughly that co-rotation speed — 868 mph for three hours is 2,604 miles — and on the real ascent time of about two and a half hours the same reasoning gives 2,169 miles. His round number is not exactly reconstructible and nothing here turns on which reading is taken: either way it is the distance the argument predicts and the observation refuses.

But the honest answer is not zero, and saying zero is where this rebuttal usually goes wrong. Lift a parcel of air straight up while it conserves its angular momentum about the Earth’s axis and it must fall behind the ground beneath it, because it now sits on a larger circle. For a rise to 38,969 m at 33.4° N the arithmetic is

u = Ωcosφ [R²/(R+z) − (R+z)] = −4.7 m/s,

a westward lag of about 43 km over a two-and-a-half-hour ascent — the same order as the drift actually observed, and about one part in eighty of what the argument requires (43 km against 2,169 miles, or 3,491 km). (Reproduced here 2026-08-11 with Ω = 7.292115 × 10⁻⁵ rad/s and R = 6,371 km.) The real atmosphere stays closer to solid-body rotation than that, because pressure gradients and turbulent mixing keep redistributing angular momentum, and what survives as a net displacement is simply the wind. So the flat-earth expectation is not wrong by being non-zero. It is wrong by a factor of about eighty — nearly two orders of magnitude — in a direction the weather decides.

2. The balloons are a rotation detector, and they read positive

Item 227 names weather balloons specifically, so take it up on that. Radiosondes drift a long way: “a sonde may drift several hundred kilometers during the 90- to 120-minute flight”, far enough that the same article notes it could in principle disturb model initialisation — though it goes on to report the effect as showing up at most locally, in stratospheric jet-stream regions. That is the answer to “balloon drift anomaly” on its own terms, and against Carpenter’s proof 54 and Dubay’s proofs 21 and 22 it points the wrong way twice over: the drift is a few hundred kilometres rather than a few thousand, and at mid-latitudes it is predominantly eastward — the direction an Earth turning east under a stationary atmosphere could not produce.

Dubay’s proof 158 predicts eastward too, and that has to be answered rather than skated over. 158 grants co-rotation for the sake of argument and infers that “hot-air balloons would also be forced steadily faster Eastwards as they ascended through the ever increasing atmospheric speeds” — so pointing at eastward drift and calling it a refutation would be conceding his point in a louder voice. The error in 158 is a confusion of two speeds. Solid-body co-rotation means uniform angular velocity: a balloon at 39 km is indeed carried east faster in absolute terms than the ground, by Ωzcosφ ≈ 2.4 m/s, and it is carried east not at all relative to the ground directly beneath it, because the two keep the same longitude. A faster absolute speed aloft is not a displacement over the surface, and only a displacement would sweep a balloon anywhere.

What is left over runs the other way. The one real altitude effect — angular momentum conserved on a rising parcel, section 1 — is a westward lag of 4.7 m/s, the opposite sign to 158’s prediction. So the eastward drift that is actually observed cannot be his effect, and it is not: it is thermal wind, and it is identifiable as such by structure he has no way to produce. It varies with latitude, it reverses its circulation sense between the hemispheres, and it does not increase monotonically with height — the flow peaks near the tropopause and weakens above it, which is the opposite of “ever increasing atmospheric speeds”.

Why eastward, and why at that speed? Because of the rotation. The mid-latitude westerlies are what the thermal-wind relation gives when a pole-to-equator temperature gradient sits on a rotating planet: the vertical shear of the geostrophic wind is ∂u/∂z = (g/fT) (∂T/∂y), and f = 2Ωsinφ carries the Earth’s rotation rate inside it. With representative mid-latitude values — f = 1.03 × 10⁻⁴ s⁻¹, about 40 K of pole-to-equator contrast over 5,000 km, T ≈ 250 K — the shear is 3.0 × 10⁻³ s⁻¹, giving about 30 m/s over ten kilometres of altitude: the jet stream, at roughly its observed strength, from the Earth’s spin and a thermometer. (Order-of-magnitude, recomputed here 2026-08-11.) Set Ω to zero and there is no f, no geostrophic balance, no thermal wind and no jet — and no parameter left to fit one with.

Two further readings from the same instrument, which is worth stressing: the same instrument. The sense of circulation around a low reverses between the hemispheres, which is a sign change in 2Ωsinφ and has no counterpart in any stationary model. And the air’s angular momentum is exchanged with the ground rather than conjured: “There is now general agreement that most of the changes in LOD on time scales from weeks to a few years are excited by changes in AAM” — the length of the day varying with an annual amplitude of 0.34 ms and a semiannual amplitude of 0.29 ms, in step with atmospheric angular momentum computed from the winds those balloons measure. On a stationary Earth there is no length of day to vary and nothing to trade.

The lineage disagrees with itself here, and it is worth seeing. Dubay’s proof 21 needs the air not to travel with the Earth; his proofs 23 and 158 attack the idea that it could, 158 by granting it for the sake of argument. But the largest single originator on this list went the other way: our ARG-A10 records Rowbotham, in the 1881 third edition, granting that if the Earth turns then the atmosphere turns with it and in the same direction, and building his argument from inside that concession. (That quotation is A10’s; the 1881 text was not reachable this pass and it is not re-quoted here.) The same 461-item list carries both, as items in the same lane, without noticing that they require opposite atmospheres.

3. The sun-angle half

Item 237 descends from two claims about photographs: that balloon footage shows a bright hot-spot on the cloud deck beneath the Sun, and that sunbeams slanting through cloud gaps converge on a source just above the clouds. Take them in order.

Converging rays are perspective. Parallel lines receding from an observer appear to meet at a vanishing point; that is why railway tracks close. The decisive observation is the one the argument cannot use: anticrepuscular rays, which appear on the opposite side of the sky and converge a second time at the antisolar point — they are “essentially parallel, but appear to converge toward the antisolar point, the vanishing point, due to a visual illusion from linear perspective.” A lamp hanging just above the clouds can put a convergence point above the clouds. It cannot also put one below the horizon behind you, 180° away, and the same evening sky routinely shows both.

And the law being appealed to is the tradition’s own. The perspective principle that answers proof 125 is not an outside import: Samuel Rowbotham builds his account of sunrise and sunset on it in 1865, and reaches for a balloon to illustrate it. Apparent descent with distance, he writes, “arises from a simple and everywhere visible law of perspective”, of which one instance is that “when a balloon sails from an observer without increasing or decreasing its altitude, it appears gradually to approach the horizon”. He is right about the law and uses it to keep his Sun a few thousand miles up. But it is a single law with a single consequence, and applied to a beam of light it produces convergence at a vanishing point whether or not anything is there — which is why it also produces the second convergence, at the antisolar point, that no near source can account for.

The hot-spot moves with the observer. A bright patch on a cloud deck or a water surface beneath the Sun is specular and forward-scattered glare, and its position is fixed by the geometry between the observer, the surface and the source — the same effect as the glitter path on the sea, which follows you along the beach. A genuinely local source would do something different and checkable: it would illuminate a pool of cloud fixed beneath itself, brighter than everything around it, and a second balloon a few hundred kilometres away would photograph that pool sitting under the Sun rather than under itself. What the footage shows is the glare tracking the camera.

Three quantitative tests finish it, all using the flat literature’s own figures for the Sun (about 32 miles across, a few thousand miles up):

• Angular size. Those numbers give a disc 36.7′ wide overhead, shrinking to 18.9′ for an observer 5,000 miles away as the Sun sinks toward the horizon — a halving, every day, visible to anyone with a filter and a camera. The measured Sun runs 0.527° to 0.545° (31.6′ to 32.7′), a 3.4% variation over the year, tracking the Earth’s orbital eccentricity, with no diurnal shrinkage at all.
• Inverse square. A source 5,000 km up is 1.89 times further from a station 8,000 km downrange than from the point beneath it, so normal-incidence sunlight there would be 3.56 times weaker at the same instant. Top-of-atmosphere irradiance measured from orbit is about 1,361 W/m² irrespective of where the measurement is made.
• The balloon measurement item 237 names. This is the one the item promises and the one worth running. Rising 38.97 km toward a Sun 5,000 km up raises the observer by a measurable fraction of the way to it: the Sun’s elevation angle shifts by about 13 arcminutes — 0.22°, roughly 40% of a solar diameter, obvious in a single frame. For a Sun at 1.496 × 10⁸ km the same ascent shifts it by 0.054 arcseconds, which nothing on a balloon can see. A high-altitude balloon carrying a clock and a camera settles the question in one flight, and the two models differ by a factor of about fifteen thousand. That measurement is not carried by any of the three list items, and it is not located in the 200 Proofs text searched here.

4. What the verdict ranges over

Not “balloons do not drift” — they drift hundreds of kilometres and the drift is measured, published and corrected for. Not “the atmosphere co-rotates perfectly” — it does not, and the departure is the entire science of meteorology. The claim being refuted is the inference: that because a balloon is not left thousands of miles behind, the ground beneath it is not moving. The premise that defeats that inference was written into the argument’s own oldest statement, in 1885, as the condition its author needed to exclude and did not.

The list says

“High-altitude balloons drift anomaly.” / “Weather balloons drift anomaly.” / “Balloon sun-angle mismatch.”

The source says

Carpenter 1885, proof 54: “unless the Earth drag the balloon along with it in its nineteen-miles-a-second motion, it must be left far behind, in space: but, since balloons have never been known thus to be left, it is a proof that the Earth, does not move”. Dubay 2015, proof 21: “If Earth were truly constantly spinning Eastwards at over 1000mph, helicopters and hot-air balloons should be able to simply hover … and wait for their destinations to come to them!”

The qualifier was dropped.

Both documented ancestors argue from a conditional, and both report the same observation: balloons behave normally. Carpenter’s hinges on “unless the Earth drag the balloon along with it”; Dubay’s on “If Earth were truly constantly spinning Eastwards…”. Neither claims to have measured an anomalous drift. Their claim is that a drift which ought to appear does not.

The list nominalises that into a noun: “Weather balloons drift anomaly.” Four words, no conditional, and the grammar of a reported measurement. A reader meeting item 227 in a list of 461 numbered proofs will take it for a finding in balloon data and go looking for the dataset — and the sources it descends from never had one, because their argument was about an expectation, not a record. That is the gap, and it is the reason the compression matters even though the underlying argument is answered above on its own terms.

The finding is narrower than “careful author traduced”, and it has to be said so. Carpenter names the escape clause and then dismisses it in the same sentence, ending with the conclusion stated flat: the Earth “does not move”. Quoting him up to the word unless and stopping would be the hedge rule violated in our favour. What the compression actually deletes is not his caution but the content of his condition — and that content is the load-bearing premise, and it is true. The 1885 sentence tells a reader exactly which fact would break the proof; the four-word item does not, and cannot.

A second shift, recorded but not the one the enum names. Carpenter’s number is the orbital one — “its nineteen-miles-a-second motion” — but he attaches it to a west-to-east direction (proofs 44 and 49), so he is not separating revolution from rotation at all; he is arguing against the Earth’s motion undifferentiated. The modern versions retarget the same picture at the daily spin specifically, at over a thousand miles an hour (Dubay 21, 22, 158), and it is that version items 108 and 227 sit downstream of. That is a sharpening between two sources rather than between a source and the list, which is why drift_type records the hedge and this paragraph records the rest.

Proposed instead: REFUTED

The record carries NOT DEMONSTRATED on the basis 'No dataset cited.' Both halves of that are now wrong. The argument is traceable — Carpenter 1885 proof 54, indexed 'Balloons not left behind', restated as Dubay 2015 proofs 21, 22 and 158, with the sun-angle item descending from Dubay 124 and 125 — and the observation it rests on is real and accurately reported: balloons do come down near where they went up, and Felix Baumgartner did land in eastern New Mexico rather than in the Pacific. Nothing is missing that a dataset would supply. What fails is the inference from the observation, which is the definition the project uses for REFUTED, and which is how the neighbouring inertial arguments from the same author are already scored (A08, A14, A17 all REFUTED). NOT DEMONSTRATED is reserved for items that cite no measurement at all (A15, A20), and applying it here understates the case in our own favour by implying we have not answered the claim when we have. Proposed: REFUTED, with the basis line rewritten to name the co-rotating atmosphere and the measured drift rather than a missing dataset.

The verdict on the scorecard is unchanged until the question is settled. Publishing the disagreement is the point: a rubric that never produces dissent is not being applied.

This motion of the firmament is evidenced in the Sagnac effect, the well-known Coriolis forces, and by geosynchronous satellites (or, in a more Tychonian vein, geostationary satellites). In the geocentric model, we agree that if the heavens ceased their rotation, the satellites would fall to the earth.

— Galileo Was Wrong: The Church Was Right (2006), Vol. II, Appendix 1, in the reprinted essay “Geocentricity's Critics Refuse to Do Their Homework” by Martin Selbrede, at printed p. 648 of the archive.org OCR text (item GalileoWasWrongTheChurchSungenisRobertA.Bennett4276; the project records this scan as Vol. II, 7th ed., 2013). The OCR prints “geostationarv” for “geostationary”, corrected here. Not checked against a print copy; a second archive scan (item GallileoWasWrong) paginates overlapping material differently

Read the second sentence with the first. The claim is evidential — satellites evidence the firmament’s daily rotation — and it is stated flatly, so nobody should pretend the source only ever denied that satellites prove anything. But it arrives attached to a conditional that names what the model is resting on: the satellite stays up because the heavens turn, and would fall if they stopped.

The essay does not leave the concession there, and neither should we. The excerpt above stops one sentence short, because carrying on would take it past the 60-word ceiling this project keeps for quoting an in-copyright book; the sentence it stops short of is Selbrede turning the concession round: “But when the heavens are postulated to be in motion, it is Dr. Nieto’s equations that are deficient, not ours.” That is fair as far as it goes. The conditional is counterfactual in his model, and he is answering a critic who, on his account of him, had argued that a satellite could not stay up over a fixed Earth at all. The objection here is not that the conditional is false. It is that the one quantity the whole arrangement then hangs on — the rate at which the heavens turn relative to the ground — is an input: read off the rotating-globe solution and put in by hand, together with the equator and the height that go with it. That is section 1 of the refutation below, and it is where this entry puts its weight.

Whose argument this is, and where it came from. The essay is Martin Selbrede’s, written against a critique that Gary North commissioned from Michael Martin Nieto of Los Alamos; Sungenis and Bennett reprint it as Appendix 1 and, in their own footnotes and bibliography, give the original as The Chalcedon Report, 1994, pp. 11–12. Its physics is not homemade. It runs on Einstein’s 1913 letter to Mach, on Thirring’s 1918 rotating-shell paper, on Orwig’s Phys. Rev. D 18:1757 (1978) and on Ø. Grøn and E. Eriksen, “Translational inertial dragging”, General Relativity and Gravitation 21:105–124 (1989). Every one of those exists, and every bibliographic detail the book prints for them checks out — checked here against Crossref. Said at its true strength and no higher: where the book gives an article page range it is right (Grøn and Eriksen at 105–124), but it gives no volume number for Orwig at all — Physical Review D, 1757–1763, 1978, in the Vol. I footnote and again in the Vol. I bibliography — Thirring it gives by volume and opening page rather than by range, and the page numbers it prints inside Brill and Cohen and, in Vol. I, inside Grøn and Eriksen are quote locations, not article ranges. The papers themselves were not opened for this entry, so those internal locations are the book’s; what can be said for them is that the Grøn and Eriksen quotations appear word for word in two different editions of the book, Vol. I’s chapter 10 footnotes and this appendix. The footnote at Vol. I p. 607 credits Selbrede with supplying the references.

Three restrictions the source states and the list does not inherit. At printed p. 635 of the same essay: geostationary satellites can sit “only over the equator, and at the same prescribed height as that indicated by the Newtonian methods Dr. North favors” — the equator, and Newton’s own number. At p. 637 the essay states its thesis outright: “it is impossible to launch an attack on geocentricity on the basis of general relativity, by definition.” And at p. 648, the conditional quoted above. A result that holds by definition, at a height somebody else’s theory already fixed, over an equator, is a demonstration that the two descriptions agree. That is what the essay set out to show, and it shows it.

The book’s other satellite passages, for anyone following the chain. Its opening chapter (Vol. I scan, ch. 1, printed p. 37) lists geosynchronous satellites among phenomena which “do not prove, in the least, the heliocentric system” — the negative form. Vol. I ch. 7, printed p. 468, quotes the geosynchronous satellite “hovering with no visible means of support!” from Dennis Sciama’s The Unity of the Universe (1959), where it is an illustration of why Newton’s second law needs an inertial frame. Vol. I ch. 11, printed p. 662, has Helmut Posch glossing a vision of Hildegard of Bingen: “Therefore, geostationary satellites travel against the rotation of space in order to appear stationary [to us].” The sentence quoted above is the technical version of the same move, and it is the one worth answering.

What this passage is being cited as. The earliest text located in this review that states the argument in the form the three items take, with a date and a named author. It is an ancestor, and the essay itself declines to be an origin: the equatorial restriction, Selbrede writes, “has been asserted in books, in journals, on audiotapes, and videotapes”, and the material was already on a videotape sent to his critic in 1992. Origination is not established here and is not claimed.

SURFACE (weak — do not use). “If the Earth were not turning, a geostationary satellite would have nothing holding it up, so it would fall.” This is exactly the objection the source was written to answer, and it loses to Einstein’s own words. Einstein to Mach, 25 June 1913: rotate a heavy shell of matter about an axis through its centre and “a Coriolis force arises in the interior of the shell, that is, the plane of a Foucault pendulum is dragged around.” There is no free win here.

DEEPER. “It is only a change of coordinates.” True, and incomplete, because it concedes the point in the form the defender wants: yes, the Earth-fixed rotating frame is a legitimate chart, satellite operations use one (ITRF/ECEF) every day, and general covariance guarantees the field equations hold in it. Stopping there invites the reply then you agree with us.

KERNEL. The specific true thing this tradition found is a real and respectable result, and it is stronger than the version in the book. A rotating mass shell does induce centrifugal- and Coriolis-like fields in its interior (Thirring 1918); in the limit where the shell approaches its gravitational radius the interior inertial frames are dragged round rigidly with it (Brill & Cohen 1966; Orwig 1978; quoted by Grøn and Eriksen as “perfect dragging”). And the numbers are not absurd. For an interior dragging coefficient d₁ = 4α(2−α)/((1+α)(3−α)) with α = GM/2Rc², perfect dragging needs α = 1; a Hubble sphere at critical density gives α = 1/4 exactly (the H² cancels) and d₁ ≈ 0.51 — the right order, off by a factor of two. That is a static-shell estimate and not a cosmological calculation, and it is offered as nothing more; but it is why Einstein, Sciama and a good deal of subsequent literature took cosmological inertial dragging seriously, and it is why the satellite question has a real answer inside a geocentric chart.

Why it doesn’t save the claim.

Because perfect dragging is the statement that the two descriptions cannot be told apart. It is not a rival account that happens to fit; it is the condition under which no observation distinguishes “Earth turns in a fixed cosmos” from “cosmos turns about a fixed Earth”. The better the dragging works, the less there is to detect — which is why the essay can say its conclusion follows “by definition”. A conclusion that follows by definition is not evidence for anything, and cannot be item 15 in a list of proofs.

And notice what has to be imported before the dragging account can start. The shell has to rotate at one particular rate, once per sidereal day. The satellite has to sit over the equator — a plane defined by that rotation. Its radius has to be 42,164 km from the Earth’s centre, which is (GM/ω²)^1/3, the Newtonian answer, as the essay concedes in as many words. Nothing in the Machian picture supplies any of those three; they are read off the rotating-globe solution first and then re-described. The geocentric frame is a translation, and it is a translation that arrives after the original.

The concession comes first, because it is large and it is permanent. A geostationary satellite is at rest in the Earth-fixed rotating frame. That frame is a legitimate coordinate system; general relativity places no bar on writing the field equations in it; a rotating shell of matter really does induce Coriolis-like and centrifugal-like fields inside itself, as Thirring showed in 1918 and as Brill and Cohen, Orwig, and Grøn and Eriksen developed afterwards. Every paper the source cites for this exists — Thirring 1918, Brill and Cohen 1966, Orwig 1978, Grøn and Eriksen 1989 — and every bibliographic detail the book prints for them checks out against Crossref: the journals, the years, Thirring’s volume and opening page, Brill and Cohen’s volume and issue, and the two article page ranges it gives, Grøn and Eriksen at 105–124 and Orwig at 1757–1763. Two things it does not give, said here so that the compliment is the right size: no volume number for Orwig in the Vol. I footnote, the Vol. I bibliography or the reprinted essay, the three places it cites him; and, inside Brill and Cohen and inside Grøn and Eriksen, page numbers that locate the quotations rather than the articles. Those we did not confirm, because the papers themselves were not opened for this entry. Nothing the book prints is wrong. Anyone answering this argument by claiming the citations are invented, or that a stationary-Earth chart is forbidden, is going to lose the exchange in five minutes.

What the verdict ranges over. Not “the satellite would fall.” The claim under review is that satellite behaviour is evidence for a rotating firmament about a fixed Earth. It is not, and the reason is visible in the source’s own sentence: the satellite stays up because the heavens turn, at a rate that has to be put in by hand, over an equator that rotation defines, at a height Newton’s arithmetic already fixed.

1. The whole argument is one number, and both models share it

A body in a circular orbit at radius r keeps station over a point on the ground when GM/r² = ω²r, so r = (GM/ω²)^1/3. With GM = 3.986004418 × 10¹⁴ m³/s² and ω = 7.2921150 × 10⁻⁵ rad/s this returns 42,164 km from the Earth’s centre — 35,786 km of altitude, orbital speed 3.075 km/s, period 1,436 minutes. (Recomputed here 2026-08-11; the published figures are 42,164 km, 35,786 km and 3.07 km/s.) Rewrite the same physics in coordinates rotating with the ground and the orbital term becomes a centrifugal term; the satellite is then at rest, held by gravity against that term. In the geocentric telling the centrifugal term is supplied by the turning cosmos instead of by the choice of frame. The three tellings differ in what they say ω belongs to. The number is the same in all three, and so is the orbit.

Which is why the source concedes the number. Selbrede, on the same essay’s printed p. 635: stable geostationary satellites sit “only over the equator, and at the same prescribed height as that indicated by the Newtonian methods Dr. North favors.” A model that reproduces its rival’s number, at its rival’s location, has demonstrated agreement. It has not produced a measurement of its own.

One detail worth having, since the list trades on the word stationary. The station-keeping radius is set by the sidereal day, 86,164.1 s, not the solar day. Redo the arithmetic with a 24-hour period and you get 42,241 km — 77 km too high, and a satellite parked there drifts west of its slot by 0.985° a day, a full circuit of the Earth in a year. So what a geostationary satellite keeps station with is the sky, not the Sun. That does not discriminate between a turning Earth and a turning firmament, since the Tychonic firmament turns on the sidereal period too, and this page is not going to pretend it does. It does show what the number encodes: a rotation rate measured against the stars, which is the one quantity everybody in this argument agrees about.

2. Where the argument’s own machinery comes from

The image of the satellite “hovering with no visible means of support” is not a geocentrist’s invention and the book does not pretend otherwise: at printed p. 468 of the Vol. I scan it is a block quotation from Dennis Sciama’s The Unity of the Universe (1959), where the point is that Newton’s second law only holds if accelerations are measured in an inertial frame, and the satellite is the vivid case. Sciama’s conclusion is that inertial frames are what need explaining — the argument that leads to Machian cosmology, not away from a moving Earth. The same chain runs through Einstein’s 1913 letter to Mach and Thirring’s 1918 paper. This tradition is quoting the mainstream Machian literature, accurately, and then reporting the equivalence it establishes as a win for one side of it.

One link in that chain the book gets backwards, and it matters because a mechanism is built on it. Thirring’s 1918 interior field contains, besides the Coriolis and centrifugal terms, an extra axial term — the 2(ω·r)ω piece — with no Newtonian counterpart. Vol. I, printed pp. 611–612, treats that term as a discovery, and proposes it as the cause of the 23.5° obliquity: the axial force “keeps bringing the universe back to the equatorial” plane. But the extra term is an artefact of holding the shell rigid, and the title of the paper that removed it says as much: Pfister and Braun, “Induction of correct centrifugal force in a rotating mass shell”, Classical and Quantum Gravity 2:909–918 (1985). Let the shell deviate from a precisely spherical shape and let its mass density vary — that is, let it take the shape its own rotation gives it — and the interior can be made flat, with the correct centrifugal force induced and no axial residue to build a mechanism on. Thirring’s main result survives, and the geocentric use of it survives with it. The obliquity mechanism does not: it is built on the part of the 1918 calculation that a later paper removed.

3. The satellite is not hovering; it is being flown

“Stationary” is doing rhetorical work the orbit cannot support. A geostationary satellite left alone does not stay put. Lunar and solar gravity tilt its orbital plane, and an uncorrected satellite reaches an inclination of about 15° in 26.5 years; holding inclination near zero costs roughly 50 m/s of delta-v every year. The Earth’s equator is slightly elliptical, so the satellite also drifts in longitude towards one of two stable points at 75.3°E and 108°W (with unstable points at 165.3°E and 14.7°W); correcting that costs up to about 2 m/s a year. Solar radiation pressure adds more. Every one of those terms is Newtonian celestial mechanics of an oblate, rotating globe perturbed by the Moon and the Sun; each is computed and budgeted before launch, and burned as propellant across the satellite’s working life. The Earth-fixed description inherits all of it unchanged and contributes none of it. That is the shape of the whole argument in one example: it can re-express the answer, and in the geocentric material read for this review it does not derive one.

The same holds for the frame itself. Satellite operations do run in Earth-fixed coordinates, and converting between them and the inertial frame requires Earth-orientation parameters that have to be measured, continuously, because the rotation is not uniform: the excess length of the mean solar day over 86,400 SI seconds ran between about 0.25 and 1 ms across 1999–2010, and on 29 June 2022 the day came in 1.59 ms short. The IERS publishes UT1−UTC and polar motion for exactly this reason. Read geocentrically, the entire firmament speeds up and slows down by parts in 10⁸, in step with terrestrial weather and the Earth’s core, and the rate has to be re-measured and re-broadcast every day so that spacecraft can be flown. That is not a contradiction — the chart absorbs it, as charts do. It is a statement about which end of the arrangement the explanations are coming from.

4. What the argument costs its own list

Follow the concession at p. 635 to where it lands. Geostationary orbit exists only over the equator, and only at 42,164 km from the Earth’s centre. An equator is the great circle a rotation axis defines on a sphere; a radius from the centre is a distance from the middle of a solid body. Item 15 is proof number fifteen in a list which elsewhere asserts that long-distance views show a flat horizon (item 43), that radar horizons are flat (217), that mining surveys assume a plane (395) and that railways make no allowance for curvature (386). The argument at item 15 is a spherical-Earth argument, and it is being used to support the claim that the Earth is not a sphere. Its authors would say so themselves: the Tychonian tradition holds the Earth to be a globe that does not move, and the satellite reasoning here is unusable without one.

5. What is left, stated without decoration

The Earth-fixed frame is legitimate, the dragging literature is real, and the source’s use of it is careful enough that its strongest sentence is a claim of exact equivalence: no attack on geocentricity can be launched from general relativity “by definition”. Agreed — and the trade runs both ways, which is the part that does not travel. You can have the coordinate freedom or you can have the evidence; the freedom is granted precisely because nothing follows from it. Every quantity the satellite argument uses — the rotation rate, the equator, the 42,164 km — is imported from the description it means to displace, and the satellite that is supposed to be hovering unsupported is in fact being held in a slot, against the Moon and the Sun and the Earth’s own lumpy equator, at a cost of about fifty metres per second a year.

The list says

Geostationary satellites fit rotating sky model.

The source says

“This motion of the firmament is evidenced in the Sagnac effect, the well-known Coriolis forces, and by geosynchronous satellites … In the geocentric model, we agree that if the heavens ceased their rotation, the satellites would fall to the earth.” (p. 648) — and, as the essay’s own thesis about what any of this can be used for: “it is impossible to launch an attack on geocentricity on the basis of general relativity, by definition.” (p. 637)

A concession was re-used as a proof. On wording, this compression is unusually faithful, and the entry says so rather than manufacturing a gap. The source asserts the evidential claim flatly — satellites evidence the firmament’s rotation — so it will not do to answer only the book’s milder opening chapter, where geosynchronous satellites are listed among phenomena that “do not prove, in the least, the heliocentric system” (Vol. I scan, ch. 1, printed p. 37). And item 109 goes further in the source’s favour than the source needed: “Geostationary satellites reinterpretation” concedes in one word that this is a re-description. A reader could reasonably score that as no drift at all.

The drift is in the speech act and in three dropped restrictions. The essay is a defence — it exists to answer two critics who argued that geosynchronous satellites are impossible on a fixed Earth — and its thesis is equivalence: “it is impossible to launch an attack on geocentricity on the basis of general relativity, by definition” (printed p. 637). A permission that holds by definition arrives on the list as proof item 15 of 461. That is the R01 move exactly: the wording barely shifts and the speech act does. Travelling with it, and lost: the equator — geostationary satellites sit “only over the equator, and at the same prescribed height as that indicated by the Newtonian methods Dr. North favors” (p. 635); the height — conceded in that same clause to be Newton’s own number, 42,164 km from the Earth’s centre; and the conditional — the satellites would fall if the heavens stopped, so the entire support is the assumed cosmic rotation rate.

The restriction that does not survive the journey at all. An equator and a radius from a centre are properties of a globe, and the tradition the passage belongs to says so openly. Items 43, 217, 386 and 395 of the same list assert that the surface is flat. The compression therefore does something the seven-value enum has no word for: it moves a claim into a document whose headline its own premises deny. force_upgraded is recorded because it is the plainest and most checkable of the three, and the reader has both texts above.

The refutation answers the source, not the fragment: it grants the Earth-fixed frame, grants that the cited relativity papers are real and that the book’s bibliographic details for them check out, grants that a rotating shell induces the right forces, and puts the weight on what the source itself concedes — the equator, the Newtonian radius, and a conclusion that holds by definition.

In fact, this experiment was called “Airy's failure,” because it contradicted the heliocentric metaphysics. The term “Airy's failure” gives psychological insight to the thoughts of the experimenters during this era.

— Galileo Was Wrong: The Church Was Right (2006), Vol. I, p. 853 of the Internet Archive scan (item GallileoWasWrong); printed page approx. 841, not yet checked against a print copy

Sungenis and Bennett inherit the phrase from van der Kamp, whose 1988 book carries it in the subtitle and as a chapter title (“The Unfailing Import of Airy's Failure”, p. 52). Note the passive voice — was called — and the claim about what Airy's contemporaries thought. Both are checkable, and both fail. Airy's own paper is titled neutrally: “On a supposed alteration in the amount of Astronomical Aberration of light, produced by the passage of light through a considerable thickness of Refracting Medium,” Proc. Roy. Soc. London (1871), pp. 35–39. The word “failure” is not in that title. The phrase is internal to the movement: its earliest documented use is van der Kamp's 1988 subtitle, and Bouw's obituary of him uses it as settled vocabulary, crediting van der Kamp with “pioneering work in pointing out the geocentric nature of Airy's failure”.

The experiment is real, the null is real, and the reasoning behind it was sound for its moment. Wilhelm Klinkerfues, working from Thomas Young's stationary-aether theory, had argued that if aberration is caused by the Earth ploughing through a static aether, then slowing the light down inside the telescope should change the angle you must tilt it. Airy built the apparatus, mounted it in 1870, observed γ Draconis for two years, and found no change. Van der Kamp is also right about something more general and more serious: a null result cannot by itself distinguish “there is no motion” from “there is motion plus a compensating effect.” That is a real problem in philosophy of science, and it is the same underdetermination point Duhem and Quine made in more respectable company.

Why it doesn’t save the claim. Because the null was predicted in advance, twice, by theories that assume the Earth moves — which is precisely the case underdetermination does not cover. Fresnel's dragging coefficient, published decades before Airy and already confirmed by Fizeau, says aether inside a medium of refractive index n is dragged at v(1 − 1/n²), exactly cancelling the slower light speed and restoring the vacuum angle. Special relativity gets the same result more cleanly: aberration is a transformation of the light ray's direction between frames, fixed by the relative velocity of source and observer, so a refracting medium downstream of that transformation cannot alter it. A result that two competing theories both predicted, and that a third (stationary aether) got wrong, is evidence against the third — not for a stationary Earth.

Three things have to be separated, because the argument depends on running them together.

1. What Airy measured. A null — no change in the aberration angle with a water-filled tube. Uncontested by anyone.

2. What the null rules out. Klinkerfues's stationary-aether prediction. That is a hypothesis about the aether, not about the Earth. Airy's result killed one aether model and left the Earth's motion untouched.

3. What aberration itself implies. The annual ~20.5″ ellipse Bradley discovered in 1729 tracks the observer's velocity: it follows the Earth's orbit, and it is the same ~20.5″ for every star, whatever that star's distance, proper motion or direction of travel. Van der Kamp did not miss this. His answer was to move the sky instead — the Earth at rest, with “the starry dome … revolving relative to us” (De Labore Solis, p. 40) — and to keep the dome's tangential speed below c he had to shrink the cosmos: Bouw's obituary records that “Walter believed [aberration] implied a small universe 60 light days in radius.” That is the price of the model, and it is where the refutation lands. A revolving sky generates aberration from the sources' own tangential speeds, which on a rigidly turning sky grow with distance, so getting the same angle for every star requires every star to sit at the same distance. The measured parallaxes say otherwise: they run from Proxima Centauri's 0.768″ down to microarcseconds, which puts Proxima alone about 1,550 light-days away, and Gaia DR3 publishes them for about 1.47 billion sources. Aberration is flat across that whole range and parallax is not — which is what a moving observer predicts and a moving sky cannot deliver. So the argument takes an experiment that measures a consequence of the Earth's orbital motion, shows the magnitude is medium-independent, and reports this as evidence the Earth does not move.

The rebranding is the tell. An experiment whose result matched the standing prediction is not a failure in any sense a physicist would recognise; it is a confirmation. The word does no work except to make a confirmation sound like a scandal — and it entered the literature 117 years after the experiment, from outside physics.

The list says

Airy's failure to detect starlight motion.

The source says

“this experiment was called ‘Airy's failure,’ because it contradicted the heliocentric metaphysics. The term … gives psychological insight to the thoughts of the experimenters during this era.”

The kind of claim changed. The source makes a historiographical claim in the passive voice — that the experiment was called a failure, and that the name tells us something about how its contemporaries felt. The list item makes a physical claim: that Airy failed to detect starlight motion. Those are different assertions and only the second is false on its face, because Airy detected stellar aberration perfectly well; what he found was that filling the tube with water did not change it. The refutation above answers the source's version — the underdetermination argument van der Kamp actually makes, which is a real argument — and not the fragment. But the fragment is what travels, and it converts a claim about Victorian sentiment into a claim about optics that the source never made.

Michelson proved this in two ways. The first was by the Michelson-Gale experiment in 1925 that measured the same absolute motion that Sagnac discovered in 1913; the second, by the Michelson-Pease-Pearson experiment which showed an ether drift against the Earth, and that the speed of light was affected by it.

— Galileo Was Wrong: The Church Was Right (2006), Vol. I, ch. 6 “What Did Michelson-Morley Actually Demonstrate?”, section “The Dayton Miller Experiments”, at printed p. 387 of the archive.org OCR text (item GallileoWasWrong, the CD-ROM issue whose title page reads Volume I / The Scientific Evidence). The same sentences appear in the seventh-edition scan at ch. 5, printed p. 662. Not checked against a print copy

Read that against the item before anything else. The item says null. The source says the experiment showed an ether drift. This is not a hedge that has been dropped in transit; it is the other side of the question, and the list has taken the side the physics literature is on.

The book uses the word “null” of this experiment once in the text reachable here, and puts it in quotation marks. At printed p. 353, in a paragraph listing the repetitions that followed 1905, Michelson “teamed up with F. G. Pease and F. Pearson and declared again that he produced a ‘null’ result”. The scare quotes are the book’s. Thirty pages later it explains what it thinks really happened: the 1928 run, it says, was the one where “Michelson indeed found significant fringe shifting” (p. 386), and Michelson then “obfuscates his results” by comparing them to a supposed 300 km/s motion of the solar system. So a reader who took “null” from p. 353 took a word the same chapter spends thirty pages retracting.

Where the claim comes from, link by link. The reading is not original to this book and the book does not pretend it is. Its footnotes 751 and 754 cite James DeMeo, “Dayton Miller’s Ether-Drift Experiments: A Fresh Look”, and the narrative here follows DeMeo’s almost sentence for sentence — the three attempts, the 22-metre and 32-metre paths, the same block quotation of Michelson, the same 20 km/s. DeMeo’s own caption calls it “their successful detection of an ether-drift of some unspecified quantity just under 20 km/sec.” Behind DeMeo stands Héctor Muñera’s 1998 Apeiron paper, quoted at length in this book’s footnote 689, which rereads the classic null results as small positives. DeMeo is not a geocentrist and Muñera is not arguing for a stationary Earth; the geocentric conclusion is added here, at this link in the chain, which is why the argument is recorded under this book and not under theirs.

One number, three answers, in one chapter. The sentence everybody is quoting is a bound: “The results gave no displacement as great as one-fifteenth of that to be expected on the supposition of an effect due to a motion of the solar system of three hundred kilometers per second.” The main text divides, across pp. 386–387: “if one multiplies his ‘three hundred kilometers per second’ by ‘one-fifteenth,’ the result is 20 km/sec”. Footnote 689, at the foot of p. 353, quotes Muñera taking the square root instead — “the corresponding solar velocity is then 300(1/15)^1/2 = 77.5 km/s, which is not null by any means” — and reporting 42.4 km/s for the JOSA printing’s one-fiftieth, which is 300/√50 and so carries the 300 km/s basis over to the second printing; Swenson’s history attaches a different velocity basis to that printing, and neither 1929 paper was reached here to settle which is right. Footnote 759 passes on Galaev’s 6,000 m/s, which is the same division done on one-fiftieth. Three velocities, three cited authorities, no reconciliation. Which arithmetic is right is settled in the refutation; that the book prints all three is visible without any physics at all.

What this passage is being cited as. The earliest text located that puts this experiment into a geocentric argument. A keyword search of the De Labore Solis PDF at geocentricity.com returned zero hits for Pease and zero for Pearson on 2026-08-10; Bouw’s Geocentricity (1992) was not reachable. So this is the earliest located, not the first, and the specimen carries no citation for item 10 — it carries none for any of its 461 — so which text its compiler read is not something this page can show.

SURFACE (weak — do not use). “The 1929 experiment was a null, so the item is simply correct and there is nothing to discuss.” That concedes the interesting half of the case and walks past the fact that the tradition the item comes from says the opposite. It also invites the reply that a “null” in this literature never meant a measured zero — which is true.

DEEPER. The experiment really did not return zero, and nobody claimed it did. What Michelson, Pease and Pearson published was a ratio: no displacement as great as one-fifteenth of the expected one. Reported limits are not measured zeros, the two 1929 papers do not give the same limit — Nature in January said one-fifteenth, JOSA in March said one-fiftieth — and a reader entitled to ask what the residual actually was is asking a fair question that the two-page papers do not answer.

KERNEL. The strongest form is about the shielding, and it was put by Dayton Miller at the time rather than invented later. Miller’s claim was an entrained ether: one dragged along near massive bodies, so that a drift would show up out in the open air on a mountain and be suppressed indoors. On that hypothesis the 1929 apparatus was the worst possible test, and this is not hindsight — it is where the instrument actually was. Loyd Swenson’s history has the team, in the summer of 1928, removing the superstructure and putting the apparatus “in a well of the Pasadena Laboratory” inside an improvised constant-temperature room. Miller’s printed objection follows exactly, hedges and all: “… it would seem that such massive and opaque shielding is not justifiable … it would seem to be essential that there should be the least possible obstruction between the free ether and the light path …” A defender who says your best null came out of a hole in the ground under a temperature-controlled ceiling, and the one instrument that was in the open air is the one that kept showing something has said something true, specific and awkward.

Why it doesn’t save the claim.

Because the objection was acted on — though not, in the end, by the run that went up the mountain — and because the version of it in this book contradicts the version that makes it good.

Acted on, and the loose end named first: in the summer of 1930 Pease and Pearson had the instrument transported up the mountain and installed in the base of the 100-inch telescope, the elevation Miller had asked for, though still not the open air he wanted (Swenson, p. 225). No result from that installation is reported in Swenson’s account of it, and his note on the same page records correspondence of January 1930 indicating that Michelson had by then lost interest and that the latest observational data was “not suitable for publication”. So the mountain run is not available to be quoted against Miller, and this page does not quote it. What did answer the objection independently was Georg Joos: an automated 21-metre interferometer, built at the Zeiss works in Jena in direct response to Miller’s 1925 announcement, reported in September 1930 after a year of running, which bounded any effect below a thousandth of a fringe width and any aether wind under 1.5 km/s. The entrainment hypothesis got its independent replication at full instrumental strength, and it did not produce Miller’s signal.

Contradicted: the entrainment argument works only if the 1929 run was shielded and low. The book needs the opposite. Its explanation of why the third attempt supposedly showed something is that “This higher altitude and longer light-path came closer to Miller’s specifications” (p. 386) — and there was no higher altitude. The observatory’s laboratory is in Pasadena. A defender may have the shielding point or the higher-altitude point; the two cancel, and only one of them is in the source.

Three things have to be kept apart here, because the item and its source disagree about the first of them. What the 1929 papers reported. What the geocentric literature says they reported. And what an instrument of that design could ever have said about the Earth’s rotation, which is the word the item actually uses.

1. The reported quantity is a limit, and the chain converted it into a measurement

The sentence everyone quotes reads: “The results gave no displacement as great as one-fifteenth of that to be expected on the supposition of an effect due to a motion of the solar system of three hundred kilometers per second.” That is the ordinary grammar of an experimental bound — no displacement as great as — and it is how limits have been stated since limits have been stated. DeMeo’s figure caption renders it as “their successful detection of an ether-drift of some unspecified quantity just under 20 km/sec”, and Galileo Was Wrong takes that over: the experiment “showed an ether drift against the Earth”. An upper bound of X is not a detection at X. If it were, every non-detection in physics would be its own discovery, reported at the threshold.

The 20 km/s is also arithmetically wrong, and the book contains the correction. A Michelson-Morley fringe displacement goes as the square of the speed — Δ = 2Lv²/λc² — so a shift one-fifteenth of the shift expected for 300 km/s corresponds to 300/√15 = 77.5 km/s, not 300/15 = 20. That is exactly the calculation the book’s own footnote 689 reproduces from Muñera, alongside 42.4 km/s for the JOSA printing’s one-fiftieth; footnote 759 then passes on Galaev’s 6,000 m/s, which is 300/50 divided linearly again. One sentence of Michelson’s, three incompatible velocities, all three printed in the same chapter as support for the same conclusion. Only one of them is even the right kind of arithmetic, and the one the main text builds its match to Kennedy-Thorndike on is not it.

On the discrepancy itself, the documentation is older than the argument, and it is messier than either side has use for. Loyd Swenson’s history reports that the two 1929 papers did not agree: Nature on 19 January carried a note that no displacement as great as one-fifteenth of the expected one was found, while in March the Journal of the Optical Society of America reported that “nothing was observed within one-fiftieth of the expected shift, based on Sternberg’s estimate of the solar system’s resultant velocity of about three thousand kilometers per second” (p. 222; Swenson names Gustaf Strömberg, the Mount Wilson astronomer who set the observing schedule with Miller, at p. 221). That trailing clause matters here, because it attaches the second ratio to a different velocity basis from the three hundred km/s the Nature sentence names — and Muñera’s 42.4 km/s for the JOSA printing is 300/√50, which carries the first basis over to the second. Neither 1929 printing was reached for this entry: the ratios and the bases are as reported by Swenson and by Muñera, who do not agree, and nothing on this page turns on which of them has it right. What both do settle is that the book’s footnote 753 — the one-fiftieth figure “appears to be in error” — sits badly beside its own footnote 689, which quotes Muñera reporting both printings and converting both, though Muñera too breaks off to wonder in parenthesis whether the JOSA figure is a misprint. Swenson, writing in 1972, asked whether the change was a refinement of the data or a change of judgment, concluded that “to most physicists these figures were clearly null results either way”, and added that “Miller or his partisans could capitalize on such discrepancies” — twenty-six years before Muñera and thirty-four before this book.

2. The higher altitude did not happen

The book’s explanation of why the third attempt allegedly showed something is causal and checkable: the run moved to a “well-sheltered basement room of the Mount Wilson laboratory”, the light path went to 52 metres, and “This higher altitude and longer light-path came closer to Miller’s specifications.” The path length is right. The altitude is a misreading of a name. The Mount Wilson Observatory’s laboratory and optical shops were in Pasadena, on the valley floor, and that is where this instrument was built — on the seven-thousand-pound cast-iron bedplate that had been used to polish the 100-inch Hooker mirror. Swenson records that in the summer of 1928, still losing the fight against temperature drift and asymmetrical strains, the team stripped off the superstructure, sank the basic apparatus into a well of the Pasadena laboratory, improvised a constant-temperature room, floated the optics on a mercury bath, enlarged the path length to eighty-five feet, and roofed the whole assembly over so that the observer could sit above the ceiling while the sealed system rotated beneath him (pp. 220–221). The instrument was carried up the mountain in the summer of 1930, into the base of the 100-inch telescope — eighteen months after the paper the argument rests on.

So the third run was not nearer to Miller’s conditions than the first two. It was further away: deeper, more thermally sealed, and 1,500 metres lower than Miller’s canvas house on the summit. Which is a real point for the other side — it is Miller’s own objection, and this page grants it in the steelman — but it is the opposite of the reason the book gives, and the two cannot both be run.

3. Michelson announced the result at a press conference

The framing around all of this is that Michelson knew and concealed: he “obfuscates his results”, he was “too blinded by whatever was prohibiting him from telling the whole truth”, he “went to his grave” unaware of what he had proved. What Michelson did with the result was to report it to the Optical Society of America at the Michelson Meeting in November 1928 and then take questions from reporters. The New York Times of 3 November 1928 quotes him: “the results of my experiment conducted with greater scientific care, improved apparatus and refined technique, with the intention of eliminating every possible source of error, are again negative.... It is for physicists to study and explain these results and reconcile them with the existence of the hypothetical ether.” Miller was in the room, said his own experiments had been conducted “in the honest hope of arriving at a negative result also”, conceded that periodic temperature fluctuations might account for his positives, and maintained them anyway. That is what the disagreement looked like from inside: two men stating their results and their doubts in public, on the same afternoon.

4. The word in the item is “rotation”, and this instrument has no purchase on it

A Michelson interferometer sends each beam out and back along its own arm. The closed circuit formed by the two interfering paths therefore encloses no area, and the first-order rotation term — the Sagnac term, which is proportional to the enclosed area times the rotation rate — cancels. All that remains of the Earth’s spin is second order in the surface speed. Put the numbers in. At Pasadena’s latitude the surface moves at ΩRcosφ = 384.5 m/s; at the summit, 384.1 m/s. With the 1928 path length of 25.9 m that is

Δ = 2Lv²/λc² = 1.5 × 10⁻⁴ fringes,

about one sixty-seventh of the 0.01 fringe the experiment could bound. (Recomputed here 2026-08-10; the same formula returns 0.91 fringes for 30 km/s at that path length, which is the 0.9 the standard comparison tables give, so the constants are the tables’.) A null rotation result at this instrument is not evidence about the Earth. It is what every party to the dispute should have predicted, and nobody in 1929 claimed otherwise.

Michelson knew this, which is why his rotation experiment looked nothing like this one. He had proposed a closed-circuit test in Philosophical Magazine in 1904, and in 1925 he and Henry Gale, assisted by Fred Pearson — the same Fred Pearson — laid out an evacuated pipe rectangle 2,010 by 1,113 feet at Clearing, Illinois, enclosing 0.208 km², because area is the thing a rotation measurement needs and a laboratory interferometer has none. The predicted shift for a rotating Earth was 0.236 ± 0.002 fringes and the observed shift was 0.230 ± 0.005. Recomputing 4AΩsinφ/λc from the published dimensions and latitude gives 0.2363 — the prediction was not fitted to the data. Within four years, two Michelson interferometers with two different geometries gave the two answers each geometry can give: the one enclosing no area saw no rotation, and the one enclosing a fifth of a square kilometre measured it, the paper reporting that the observed displacement agreed with the computed value “within the limits of experimental error”.

5. The geocentric explanation cannot produce the residuals it is invoked to explain

Take the book’s own mechanism seriously, because it states one. Footnote 752: “It is precisely the rotation of the ether every 24 hours that accounts for the small positive results of all the interferometer experiments at the surface of the Earth.” An ether circling a fixed Earth once a sidereal day passes a Pasadena laboratory at the same 384.5 m/s that a rotating Earth carries the laboratory through a fixed ether — the relative velocity is the same vector, which is exactly why no laboratory interferometer can tell the two apart. But the same identity fixes the size. The predicted shift is the same 1.5 × 10⁻⁴ fringes. Miller’s claimed drift of about 10 km/s is 26 times that speed and therefore 676 times that fringe shift; the 20 km/s the book infers for Michelson is 2,705 times it. The hypothesis offered to explain the small positive results under-predicts them by between two and three orders of magnitude in velocity. A defender can reach for a partially entrained ether with a velocity gradient, and the book quotes Galaev doing exactly that — but an entrained ether is a different explanation, it is Miller’s and not the 24-hour one, and it predicts the effect grows with altitude, which returns us to the fact that the 1928 run was in a well in Pasadena.

6. What has happened since, on both instruments

Both lines were continued and both kept giving the same answers. On the translation side: Joos at Jena in 1930 bounded any drift below a thousandth of a fringe, under 1.5 km/s; the modern descendants are rotating optical cavities, which bound the anisotropy of the speed of light at Δc/c ≈ 10⁻¹⁷ (Herrmann et al. 2009) and 10⁻¹⁸ (Nagel et al. 2015). Set that against 1929 in the same units rather than in fringes: a bound of v ≤ 3 km/s is a bound on the fractional anisotropy of about (v/c)² ≈ 10⁻¹⁰, so the modern experiments are seven and eight orders of magnitude past it — still null, still exactly what relativity predicts for an Earth in motion. On the rotation side: the ring laser at Wettzell, the direct descendant of the 1925 rectangle, resolves the Earth’s rotation rate to better than one part in 10⁹ after about 10⁴ seconds of integration (Eur. Phys. J. C 82, 2022); a 120-day continuous run of a large ring laser resolved length-of-day fluctuations to a few milliseconds, about five parts in 10⁹, at three hours per data point (Schreiber et al. 2023); and during the CONT17 campaign the Wettzell ring laser was run alongside very-long-baseline interferometry, where folding the two together improved the δUT1 and polar-motion solutions rather than contradicting them (Böhm et al. 2019). The instrument that cannot see rotation still cannot. The instrument that can, does, continuously, to nine figures.

7. What the verdict ranges over

Not “the 1929 experiment was a fraud” and not “the residual was exactly zero”. The experiment was careful, it reported a limit rather than a zero, and the two papers stating that limit disagree with each other — all of which is conceded above. What fails is the use. As the list states it, the item describes a null in an instrument that could not have detected the Earth’s rotation on any cosmology, and so discriminates nothing. As its source states it, the claim is a detection, and it is built on an upper limit read as a measurement, a linear conversion of a square-law quantity, and an altitude the apparatus did not have. The two readings are each other’s contradiction, and the list is holding the one its own authority spent a chapter arguing against.

The list says

Michelson–Pease null rotation detection.

The source says

“… the second, by the Michelson-Pease-Pearson experiment which showed an ether drift against the Earth, and that the speed of light was affected by it.”

The item states the opposite of its source. The gap here runs the other way from the usual one. On most arguments on this page the list is firmer than its source. On this one the list is the more orthodox party: it calls the 1929 experiment a null, which is what the physics literature calls it, while the text our record names as the argument's origin says the experiment “showed an ether drift against the Earth” and spends thirty pages arguing that the null was manufactured. The one place the reachable text applies the word “null” to this experiment, at printed p. 353, it is inside the book's own quotation marks.

Two smaller gaps travel with it. The word “rotation” is the list's. The book treats this experiment as a Michelson-Morley repetition about translation through the ether, and puts rotation elsewhere — in Michelson-Gale and Sagnac, where it concedes the measurement and reassigns the motion to the ether. No claim that this experiment measured the Earth's rotation is located in the archive.org OCR's pages on Michelson-Pease-Pearson (printed pp. 385–387) or at the “null” mention on p. 353. Rotation does appear on those pages, but as the book's own explanation rather than as the experiment's subject: footnote 752, at the foot of p. 387, credits “the rotation of the ether every 24 hours” with the small positive results of all the interferometer experiments — a mechanism answered on its own numbers in the refutation. And the item is one item. It is the only member of this cluster, which is worth saying on a page about how 461 items compress: a claim can circulate as a single line with no relatives and still carry a reading of the primary literature that its own tradition rejects.

reversed is recorded because that is what the comparison shows, but the enum was built on the assumption that the item is a derivative of the source, and here the derivative agrees with the papers and contradicts the source — a shape none of the seven values quite names. The specimen carries no citation for item 10, so which text its compiler read is not something this page can establish; both texts are printed above and the reader can see the distance between them.

The refutation answers the source, not the fragment. It takes the book's claim at full strength — a detection of an ether drift, at 20 km/s, in a run said to have been made at higher altitude — and answers each part: the quoted sentence states an upper bound rather than a measurement; the conversion of that bound is linear where the fringe shift goes as the square of the speed, as the book's own footnote 689 shows by doing it correctly; and the 1928 apparatus sat in a well of the observatory's Pasadena laboratory, going up the mountain only in the summer of 1930. The list's own phrasing is answered separately, in section 4, where the arithmetic shows that an instrument of that geometry has a rotation signal about seventy times below the limit it could set.

Proposed instead: SELF-CONTRADICTED

STANDARD PHYSICS is defensible on the experiment and is what A01 carries for the 1887 original: a real, explained, non-discriminating null. But the legend defines SELF-CONTRADICTED as the case where "the claim's own source, or another item on the same list, points the other way", and that is this item exactly. The item asserts a null; the work our record names as its originator asserts at printed p. 387 that the same experiment "showed an ether drift against the Earth", with no hedge in the sentence, and uses "null" of it once, in its own scare quotes, in a paragraph it then spends thirty pages retracting. Filing it as STANDARD PHYSICS records the least interesting true thing about it and loses the provenance finding, which is the product of this review. THE WEAK POINT, STATED SO IT CANNOT BE MINED: the challenge depends on the originator attribution, and the specimen carries no citation for item 10, so we cannot show the compiler read this book rather than a textbook table that calls the result null. If an operator judges that too thin to carry "the claim's own source", STANDARD PHYSICS should stand and nothing else in this entry changes: the refutation answers both readings in separate sections and does not depend on which verdict chip sits above it.

The verdict on the scorecard is unchanged until the question is settled. Publishing the disagreement is the point: a rubric that never produces dissent is not being applied.

The Sagnac effect used in laser gyroscopes and the precession of mechanical gyrocompasses indicate the Earth is spinning. … Implicitly denied is the equally valid premise that the rotation of the external world, the universe, can cause the very same inertial forces — centripetal and Coriolis. That premise is known as Mach's Principle.

— Galileo Was Wrong: The Church Was Right (2006), Vol. I, ch. 12 “Technical and Summary Analysis of Geocentric Cosmology”: the third numbered geokinetic claim at printed p. 710, and the response to Claim #1 at printed p. 711, of the Internet Archive OCR (item GallileoWasWrong). In this scan the printed folio sits at the foot of its page, so the text between markers 709 and 710 is p. 710; the volume's own contents page, which puts the chapter's first section at 710, confirms it. The same passage is reprinted in the 2013 seventh edition at Vol. II, ch. 10 (opening at printed p. 157), the claim and response at pp. 158–159 (item GalileoWasWrongTheChurchSungenisRobertA.Bennett4276); both scans were read for this entry. No print copy was consulted.

Read who is speaking. The first sentence is not the book’s claim. It is the book stating the case against itself — the third of three “geokinetic claims” it lists at p. 710 before answering them — and it states that case accurately and without sneering. The second sentence is the answer, from the response to Claim #1 on the facing page. Between them they contain the whole of what the source holds about gyroscopic instruments, and it is not what item 229 holds.

What the source actually claims: a not-disproven, not a result. Two pages on, at p. 712: “No measurement of absolute or preferred rotation has been made to test whether the Earth is rotating or its surroundings. Until such a test is performed, Mach’s principle is a valid statement; it has not been disproven experimentally.” And in the Sagnac chapter, numbered point 16 at p. 742: “By Mach’s principle the Sagnac effect cannot distinguish between whether the Earth actually rotates and the ether is at rest, or the Earth is at rest and the ether whirls around it.” That is an indistinguishability claim, and on its own terms it is correct. Point 15, immediately above it, is even-handed in the same way: “A free gyroscope can be used to measure the rotation of the gimbal mounting; a Sagnac interferometer measures its angular velocity with respect to the local inertial (Geocentric) frame.”

But it does not stop at the disjunction, and neither should we. It would be the same trick in reverse to quote the book up to “cannot distinguish” and present it as neutral. Later in the same chapter, answering the claim that Michelson–Gale shows the Earth rotating with respect to the heavens, the third numbered response at p. 746 drops the hedge entirely: “Just as the free mechanical motion of the Foucault pendulum defined a plane of motion relative to the rotating heavens, the free motion of the Michelson-Gale light ring defined a plane of radiation relative to the same heavens.” The heavens rotate; that is the book’s position, stated flat. The refutation below answers that, at that strength.

The detail on the same page that the list could not carry. Immediately before those responses, at the head of that same p. 746, the book derives the Michelson–Gale time difference as Δt = 4Aω sin φ/c² — the bare form Δt = 4Aω/c² closes the previous page — and explains the latitude factor geometrically: the Earth’s axis of rotation projects onto the apparatus at an angle corresponding to the latitude; at the equator the polar axis lies parallel to the loop and there is no effect; at either pole it is perpendicular and the effect is maximal. That explanation is available only if the plane of the apparatus tips as you carry it over the surface — which is to say only on a globe. Sungenis and Bennett are geocentrists, not flat-earthers, and their spherical Earth is load-bearing in their own arithmetic.

The objection the book prints against itself. At p. 311, footnote 615, the book reproduces — in its own words, “Sciama quotes Eddington’s objection to Mach” — the sentence “We do not believe that if the heavenly bodies were all annihilated it would upset the gyrocompass.” That footnote hangs off a passage about relativity wanting things both ways, and this entry takes no view on what the book concludes from it. It is quoted for one thing only: a century ago, both sides of the Mach argument already understood the gyrocompass to be where the question sits.

What this passage is being cited as. The earliest text located that carries this content in the form the item uses. It is an ancestor and not evidence of origination. The compressed word sky-locked is not located in the OCR of either Sungenis volume; the word gyro is not located anywhere in the archive.org OCR of Marshall Hall’s The Earth Is Not Moving (1991), where the corresponding mechanical argument is about the Foucault pendulum. And the fact the item leans on — that a free gyroscope holds its axis against the stars rather than the ground — is ordinary navigation physics that predates the movement: Léon Foucault built the gyroscope in 1852 to demonstrate it. Our record credits nobody with this argument, and this entry does not change that.

SURFACE (weak — do not use). “The gyrocompass finds north by sensing the Earth’s rotation, so it proves the Earth rotates.” This loses, and it loses to the source’s own page. A gyrocompass senses rotation with respect to the local inertial frame. Nothing in the instrument tells you whether that frame is fixed and the Earth turning, or the Earth fixed and the frame dragged round by a turning cosmos. Anyone opening with the surface version will be shown Mach’s principle and will have nothing to say.

DEEPER. The item’s premise is literally true of a free gyroscope, and it is the mainstream reference works that say so. Wikipedia’s own article on the gyrocompass puts it this way: a spun-up wheel “will normally maintain its original orientation to a fixed point in outer space (not to a fixed point on Earth).” That is sky-locked, in three words, from a source nobody in this argument would call partisan. It is why an aircraft’s directional gyro wanders against the compass card and gets realigned every ten to fifteen minutes, and why a mechanical heading indicator carries a latitude nut — see ARG-A07. An instrument that has to be dragged back into line with the ground several times an hour is not obviously an instrument that is referenced to the ground.

KERNEL. The strongest form is the source’s, and it is real physics rather than rhetoric. Mach’s principle says the inertial field is set by the mass distribution of the universe; a rotating shell of distant matter therefore produces, inside itself, the same Coriolis and centrifugal terms as a rotating observer inside a still one. This is not a fringe conjecture. It is a general-relativistic calculation with a hundred years of literature behind it — Hans Thirring did the rotating-shell interior in 1918, Brill and Cohen sharpened it in Physical Review 143:1011 (1966), and Gravity Probe B measured frame dragging around the actual Earth in 2011 to about 19 per cent. So the kernel is: every gyroscopic instrument on the planet measures relative rotation between the Earth and the cosmos, the physics of dragged inertial frames is mainstream and measured, and a gyroscopic reading cannot apportion that relative rotation between the two bodies. All of that is correct, and it is why this page does not argue that a gyrocompass proves the Earth spins.

Why it doesn’t save the claim.

Because what Mach’s principle buys is an equivalence between rotations. It buys nothing whatever about shape — and the list this item appears on is a flat-earth list.

Grant the whole Machian package. The inertial field inside the shell is then the same field an observer would see if the shell stood still and everything inside it turned; that is the point of the calculation. Now ask what a gyroscopic instrument standing on a surface inside that field actually reads. The answer depends on one thing only: the angle between the local vertical at the instrument and the rotation axis — local vertical meaning the plumb line, apparent gravity, since that is what the instrument’s pendulous mass answers to. On a sphere that angle is the co-latitude, and it sweeps through a right angle between equator and pole, which is why the free-gyro azimuth drift goes as sin φ and the tilt rate — at a given heading, and with it the gyrocompass’s whole north-seeking torque — goes as cos φ. On a flat Earth whose down is perpendicular to the plane the local vertical is a single direction shared by every point on it, so that angle is the same everywhere, and no gyroscopic quantity can vary from place to place at all. The one escape — a self-gravitating disc whose plumb line tilts outward from the centre — is answered in section 3 of the refutation, and it is answered by pointing out what it costs: an apparent vertical that swings with position is a curved equipotential, which is the curvature the model was built to avoid.

That conclusion does not care which body is turning, and it does not care which way the axis points. It is a fact about the surface. So the source’s best argument, taken at full strength and conceded in full, delivers a spherical stationary Earth — which is what Sungenis and Bennett actually hold, and what their own Michelson–Gale derivation on p. 746 requires. Transplanted onto a flat Earth it stops working, and it stops working on the one quantity every ship’s bridge in the world measures daily.

Start with the concession, because it is large and it is permanent. A free gyroscope is sky-locked. That is not a flat-earth talking point, it is how the instrument works, and the mainstream reference on the gyrocompass says so in one sentence: a spun-up wheel “will normally maintain its original orientation to a fixed point in outer space (not to a fixed point on Earth).” Second concession, bigger: no gyroscopic instrument can tell you whether the Earth turns or the cosmos turns around it. It measures rotation relative to the local inertial frame, and on a Machian model the frame is dragged by the turning cosmos and every reading comes out identical. The source makes that argument, it is right, and this page grants it whole. Anyone answering this item with “a gyrocompass cannot work unless the Earth spins” has skipped the step the source will not let them skip, and will lose the exchange on it.

1. What the instrument is, since the item names it and the name matters

A gyrocompass is not a free gyroscope. It is a free gyroscope that has been deliberately broken as a free gyroscope, and the breaking is the entire mechanism. The rotor is constrained so its spin axis stays roughly in the horizontal plane and is made bottom-heavy — gravity control, by pendulous mass or by a mercury ballistic. Now the Earth’s rotation carries the local horizontal round underneath a spin axis that is trying to hold still against the stars; one end of the axis rises; the weight applies a torque; and a gyroscope answers a torque by precessing at right angles to it, so the axis swings in azimuth. The swing is towards the meridian, it overshoots, and damping brings it to rest pointing true north, typically over some hours. The oscillation period is tuned to the Schuler period of 84.4 minutes so that the ship’s own accelerations do not throw it off.

So the item has named the one gyroscopic instrument in the world that is engineered not to be sky-locked. Its whole design is the conversion of a sky-locked axis into an Earth-locked one, and the conversion is powered by the relative rotation of the two. ARG-A07 handles the free-gyro and ring-laser items; this one is about the machine that eats the drift.

2. Two numbers, and they are the argument

Resolve the rotation vector Ω — magnitude 7.292115 × 10⁻⁵ rad s⁻¹, or 15.041° per hour — into the local vertical and the local horizontal at the instrument. Two rates follow, and both are in the training material a watchkeeping officer sits an exam on:

azimuth drift of a free gyro = 15° × sin φ per hour
tilt rate of a horizontal axis = 15° × cos φ × sin (azimuth) per hour

(The training tables round the constant to 15°; the rate the instrument actually senses is the sidereal one, 15.041° per hour, a difference of 0.27 per cent that matters for the large ring lasers at ARG-A02 and not for a ship’s compass.)

The second of those is the gyrocompass’s power supply: no tilt, no torque, no north-seeking. Hence the standard statement of the directive force — it “is maximum at the equator and decreases to zero at the poles.” A gyrocompass is at its best on the equator and gives up entirely at the pole, which is the reverse of the intuition most people arrive with, and it is checkable by anyone who has stood a watch.

Notice what the two rates are reading. They are not two independent empirical curves fitted to data. Ω has a vertical component Ω sin φ and a horizontal component Ω cos φ, and those two sum in quadrature — sin² φ + cos² φ = 1 — to the same 15.041° per hour at every latitude on Earth, a magnitude known independently of any of these readings. The drift rate reads the vertical component off directly. The tilt rate reads the horizontal one, cut down by sin (azimuth), because only the part of the horizontal component lying at right angles to the rotor axis shows up as tilt at all: a rotation about the spin axis itself moves nothing. Lay the axis east–west and the tilt rate is the whole horizontal component; lay it north–south and the tilt rate is zero while the drift rate is untouched. The only thing that changes from place to place is how that fixed vector is split between vertical and horizontal — and that split is nothing but the angle between the local vertical and the rotation axis. The equator is the degenerate case worth checking by hand: there Ω lies in the horizontal plane pointing due north, so a free gyro aimed north is aligned with it and holds still against both the sky and the ground at once, while a gyro aimed east tilts at the maximum rate. Both facts fall out of the same decomposition.

3. The step that flat and round part company on

On a sphere, the local vertical is the radius, so the angle between it and the polar axis is the co-latitude and sweeps through 90° from equator to pole. That is where sin φ and cos φ come from, and there are no free parameters in it.

On a plane, the local vertical is one and the same direction at every point of the surface — which is an assumption about the model and not a definition, so it is named and defended in the next paragraph. The angle it makes with a fixed rotation axis is therefore identical everywhere, whatever that axis is and whichever body is doing the rotating. A rotating dome, a rotating disc, an ether whirl about the pole star — it makes no difference to this, because nothing in the reasoning depends on what is producing the rotation. It follows that on a flat Earth no gyroscopic quantity can depend on where you are standing. Free-gyro drift would be the same at Singapore and Reykjavik; the tilt rate would be the same; the gyrocompass’s directive force would be the same. If the dome turns about an axis perpendicular to the plane, that common tilt rate is zero, and no gyrocompass would settle anywhere on Earth.

Name the assumption, because it is carrying the paragraph. “The local vertical” here means the direction the plumb line points — apparent gravity, which is what a pendulous mass or a mercury ballistic actually answers to — and not the geometric normal to the surface. On any flat model that keeps down perpendicular to the plane the two coincide and the argument above is closed. The escape is to separate them: let the disc gravitate on its own account, and the plumb line tilts inward as you move out from the centre, so the angle it makes with a fixed axis does vary with position after all. That escape buys the latitude dependence by buying curvature under another name. A surface whose apparent vertical swings through a right angle between centre and rim is a curved equipotential, and the ocean lying on it follows the equipotential — which is to say the sea surface is curved, which is the proposition being denied. It also fails to deliver the reversal. Take the rotation to be about the axis through the centre of the disc, which is what every flat cosmology on offer proposes, the stars being observed to wheel about Polaris: the Foucault sense is fixed by the sign of the component of Ω along the local up, that component is Ω times the cosine of the plumb line’s tilt, and a tilt of less than a right angle cannot make a cosine negative. To flip the sign the local up must turn through more than 90° — which is a sphere, and is where section 5’s hemispheric reversal comes from.

Turn that around and it is a positive result rather than a refutation: a gyrocompass measures the angle between the local vertical and the Earth’s rotation axis, from inside a closed steel box, with no horizon in view and no photograph involved. Watching that angle change as you sail is watching the surface curve.

4. The paper trail, which is in the regulations and the sales literature

None of the above is a theoretical nicety maintained by cosmologists. It is written into the documents that govern commercial shipping.

The international standard. IMO Resolution A.424(XI), Performance standards for gyro-compasses, requires that the compass “settle within six hours in latitudes of up to 60°”, and specifies the tolerances as trigonometric functions of where the ship is: the settle point error shall “not exceed ± 0.75 x secant latitude”, and the residual steady state error after speed and course correction at twenty knots ± 0.25 × secant latitude. Secant latitude is 1/cos φ — precisely the reciprocal of the directive force. The standard also requires that “means should be provided for correcting the errors induced by speed and latitude.” A regulator writing rules for a flat Earth has no reason to reach for a trigonometric function of position, and no reason to stop the standard at 60°.

The datasheet. Sperry Marine’s NAVIGAT X MK 2, a spinning-mass gyrocompass sold with twin rotors at 19,000 rpm, publishes its accuracy the same way: static error “< 0.1° secant latitude”, dynamic “< 0.4° secant latitude”. The globe’s trigonometry is in the specification a buyer holds the manufacturer to.

The correction table. A gyrocompass on a moving ship settles off true north, because the vessel’s own northward motion adds to the Earth-rate the instrument is chasing. The correction, in the standard maritime form, is tan δ = −V cos C / (900 cos φ + V sin C), with V in knots and C the course. The 900 is the eastward speed of the ground at the equator, and the cos φ scales it to your latitude. Look at where that number comes from: the equator is 21,600 nautical miles round, because a nautical mile is one minute of arc of a great circle and there are 360 × 60 of them; 21,600 divided by 24 hours is 900 knots exactly. The constant in every ship’s speed-error table is the Earth’s rotation expressed in units of the Earth’s own curvature. (Against the stars the figure is 902.5 knots; the 0.3% difference sits far inside the instrument’s error budget and nothing here rests on it.)

5. Its twin, which points the other way and settles the sign

The Foucault pendulum is the same vector read on its other component: its precession goes as sin φ, it turns clockwise in the northern hemisphere and counterclockwise in the southern, and at the equator “the plane of oscillation remains fixed relative to Earth.” Gyrocompass directive force peaks exactly where the pendulum stops, and stops exactly where the pendulum is fastest. Two instruments of different physics, complementary across the whole range of latitude, summing to one constant vector. The hemispheric reversal is the part no single dome rotation can produce: one rotation about one axis over one plane has one sense, everywhere on the plane. See ARG-A06.

6. What the list says about this two hundred items earlier

Take the item at its word. Sky-locked means the axis holds against the heavens and therefore turns, once a day, with respect to the deck it is bolted to. Item 229 is thus a concession that a gyroscopic instrument registers a daily rotation between ground and sky — and item 225, “Ring laser gyro corrections”, concedes the same thing from the engineering side, since a correction is only needed if there is something to correct for. Now item 12, on the same list: “Gyroscope anomalies indicating no rotation.” And item 224, five lines above item 229 in the same run of instrument fragments: “No rotation proof mechanical.” Item 12 denies exactly what 229 and 225 assume. Under every model anyone in this argument holds, the ground and the heavens turn relative to one another once a day — otherwise there is no day — so an instrument cannot be holding still against both, and a list cannot have it both ways either.

Item 224 taken slowly, because four words will carry two readings and we should not simply take the convenient one. Read “No rotation proof mechanical” in the geocentric idiom — there is no mechanical proof that the Earth in particular turns — and it does not contradict item 229 at all. It is this page’s own opening concession, printed on the list as though it were a proof; and on that reading the list has published item 229 as a proof of a claim its own item 224 says no mechanical instrument can establish. Read it the other way — no mechanical instrument detects any rotation — and it contradicts 229 head-on. The four words give the reader nothing to choose between the readings, and neither reading leaves item 229 standing as a proof. This entry therefore rests nothing on item 224.

One more of these items has a defence, and it should be stated. Item 19, “Gyroscopes stable absent recalibration”, is a fair report about cheap mechanical gyroscopes, whose bearing friction really does swamp a rate of four thousandths of a degree per second — that is worked through at ARG-A07 and it is not counted here. What survives, and survives whichever way item 224 is read, is the contradiction between item 12 on one side and items 225 and 229 on the other. It does not turn on instrument grade either, because item 12 is a claim about a reading — anomalies indicating no rotation — and the person our own records name as the originator of that argument is filmed reporting the opposite reading off his own ring laser gyroscope, a drift of some fifteen degrees an hour, which is set out with its limits at ARG-A07. The list carries both halves and offers each as a proof. That is what the verdict chip on this entry names, and it is checkable in a minute against the specimen itself.

7. What the verdict does and does not range over

It does not claim that the gyrocompass settles whether the Earth turns or the heavens do. It does not, the source says it does not, and the source is right; that question is argued at ARG-R01 and ARG-A22, on different ground. What is claimed here is narrower and harder: the item asserts, of a specific instrument, a behaviour that instrument is built to prevent; the latitude signature of the thing it actually does is a measurement of the surface’s shape and is unavailable on a plane whatever rotates; and the list carries, elsewhere in its own numbering, items that flatly contradict it. The argument arrives on a flat-earth list out of a geocentric book whose own Michelson–Gale derivation, four pages later, needs the globe to produce its sin φ. It is the lineage problem this review exists to document, in a single four-word item.

The list says

Gyrocompass sky-locked.

The source says

“The Sagnac effect used in laser gyroscopes and the precession of mechanical gyrocompasses indicate the Earth is spinning.” — p. 710, the book setting out the case it is about to answer. … “No measurement of absolute or preferred rotation has been made … Mach's principle … has not been disproven experimentally.” — p. 712. … And then, later in the same chapter and with no hedge left in it: “Just as the free mechanical motion of the Foucault pendulum defined a plane of motion relative to the rotating heavens …” — p. 746, the sentence completing with the Michelson–Gale light ring defining a plane of radiation relative to those same heavens.

A concession was re-used as a proof.

The wording barely moves and the speech act changes completely. In the source, the first sentence is the opposing side’s claim, set out at p. 710 so that it can be answered; the answer, running from p. 711, is that nobody can tell which body is turning and that Mach’s principle “has not been disproven experimentally”. A not-disproven is not a result. The list keeps one horn of an explicit indistinguishability and publishes it as an affirmative fact about an instrument — the force_upgraded pattern already worked at ARG-R01, where a concession its own author called “forcing an open door” arrives on a list as a proof.

The book does not stop at the not-disproven, which is why the third quotation is printed above. At p. 746 it says flatly that the free mechanical motion of the Foucault pendulum, and the free motion of the Michelson–Gale light ring, each define a plane relative to the rotating heavens. Ending our own blockquote at the hedge would have been the hedge rule broken on our side, and it would have made this drift look larger than the evidence supports. It does not dissolve the drift. What the book states flatly it states of free instruments — a swinging pendulum, a light ring — while item 229 asserts it of the gyrocompass, the one instrument here that is deliberately built, by gravity control and damping, to stop being free and to hold on the ground instead. The upgrade survives with the flat statement in view, and the finding is worth more for having been made against the book’s strongest sentence rather than its most cautious one.

A second drift travels with it that the seven values have no word for. Sungenis and Bennett are geocentrists: their Earth is a globe, it simply does not move. Later in the same chapter, at p. 746, their own derivation of the Michelson–Gale effect is Δt = 4Aω sin φ/c², and they explain the sin φ by the apparatus tipping relative to the polar axis as it is carried over the surface — zero at the equator, maximum at the poles. That reasoning needs the sphere. The item removes the sphere and keeps the conclusion. It is not a hedge dropped, not a scope widened and not a category shifted: it is a conclusion lifted out of the model that made it true. force_upgraded is recorded because it is the plainest and most checkable of the two, and both texts are printed above for the reader to judge.

The gap is the finding, and here it costs the list its own coherence. The source’s claim — gyroscopic instruments measure relative rotation and cannot apportion it — is true, careful, and survives this page intact. The compressed version is the one that circulates, and it asserts something about a gyrocompass that is false of a gyrocompass, on a list that elsewhere asserts the opposite of it in items 12 and 19. The refutation above answers the source at the source’s strength, by conceding the Machian point entirely and moving to the one thing Mach does not cover; the drift is published here because the reader meets the four words, not the chapter.

Either coordinate system could be used with equal justification. The two sentences: “the sun is at rest and the Earth moves,” or “the sun moves and the Earth is at rest,” would simply mean two different conventions concerning two different coordinate systems.

— Galileo Was Wrong: The Church Was Right (2006), Vol. I, Introduction, p. 9 (with n. 22)

This is the load-bearing sentence for the whole cluster, and the first thing to say about it is that Sungenis and Bennett did not write it. It appears in their Introduction as a quotation, footnoted (n. 22) to Albert Einstein and Leopold Infeld, The Evolution of Physics (1938/1966), p. 212. On the same page they also quote Poincaré on the undetectability of uniform translation, and Bellarmine on saving the appearances. The Introduction assembles a conventionalist case: that the Ptolemy–Copernicus dispute is about description rather than fact.

The list items are considerably firmer than the source. Sungenis and Bennett are making a philosophical argument about the covariance of physical law. In this Introduction they do not argue that GPS's use of ECEF coordinates, or a nautical almanac's geocentric columns, constitute evidence that the Earth does not move. Vol. II does take up the GPS, at ch. 10, and Sungenis returned to it in 2010; what he claims there for the Earth-centred frame is that it is easier to compute in, which is a different claim and is answered in the refutation. That inference — from engineering practice to cosmological fact — is a downstream compression, and it is weaker than what the book says. Einstein's claim in context concerns the form of physical laws under coordinate transformation; it is not a claim that a rotating and a non-rotating frame are dynamically indistinguishable, and Einstein plainly did not think that.

SURFACE (weak). “They mistake a coordinate choice for a claim about reality.” True but nearly empty. It answers a slogan, not a position.

DEEPER. The frames really are conventional. General relativity is generally covariant; the laws can be written in any coordinate system, including one rigidly attached to a point on the Earth's surface. No experiment returns the value “origin.” Choosing the solar-system barycentre over the Earth's centre of mass is a choice made for tractability, and tractability is not truth.

KERNEL — the operationalist argument, which deserves to be met head-on. Every measurement humans have ever made was made from the Earth. The ICRF is defined by quasar positions determined by Earth-based VLBI. The barycentric ephemerides are fitted to observations taken from Earth's surface, or from spacecraft tracked from it. The barycentric frame is therefore not something we found; it is something we constructed from topocentric data by applying a transformation we chose. If the epistemic root of every frame is an Earth-bound observation, in what sense is a frame whose origin lies somewhere nobody has ever stood more real than the one where the observations were actually taken? This is Poincaré's conventionalism, and it is a serious position seriously held. The 28 items can be read as an empirical demonstration of it: when people need to do something rather than philosophise, they revert to the frame the measurements were made in — in geodesy, seismology, aviation and volcanology alike. That convergence is a real fact and it wants an explanation.

We concede all of this without reservation.

Why it doesn’t save the claim.

The conventionality of the origin is not the conventionality of the dynamics. This is the hinge. You may put the origin anywhere, and nothing physical changes. What you may not do is thereby make the fictitious-force terms disappear. Write Newton's laws in a frame rigidly attached to the Earth's surface and they are false as stated: you must add a centrifugal term and a Coriolis term to recover agreement with observation. Write them in a frame that does not rotate with respect to the distant matter of the universe and you need add nothing. That asymmetry does not depend on where the origin is. It divides all frames into two classes, and the division is measurable, not stipulated.

The coefficient of those extra terms is a number, ω, and it is the same ω that appears as a defining constant of WGS 84, as the rate parameter in the IERS Earth Rotation Angle, as the Sagnac term in GNSS pseudorange processing, and as the reading of a ring laser gyroscope in a sealed room with no view of the sky. Conventionalism is a thesis about description. It cannot make a measured number go away.

There is a second reason the kernel does not deliver, and it decides this cluster. Even granting operationalism in full, the argument fails on its own evidence. If the practical primacy of Earth-fixed coordinates were evidence of a fixed Earth, we would expect those coordinate systems to require no rotation parameter. Instead every one of them has Earth's rotation rate written into its definition, and several break by tens of metres if that rate is set to zero. The items point at the artefacts; the artefacts contain the answer.

First, what the book actually argues — because it is not what the list says. Sungenis and Bennett argue conventionalism: that the Ptolemy–Copernicus dispute is a dispute about description rather than about fact, and that the geocentric description is therefore admissible. They argue it by quoting Einstein and Infeld, Poincaré and Bellarmine. In the passage this cluster rests on — Vol. I, Introduction, p. 9 with n. 22 — they do not cite GPS, WGS 84, ITRF, seismic hypocentres or a nautical almanac: no instrument is named at all. Elsewhere they do, and that has to be said plainly. Vol. II gives a section of ch. 10 to the “Global Positioning System: Claims and Responses” (p. 204), with a further appendix on the GPS satellites; and Sungenis’s 2010 reply to Phil Plait of Discover makes this cluster’s move with the instrument named: “Using the ECI frame (Earth Centered Inertial Frame) is much easier than the Solar Barycentric frame. If you don’t believe me I suggest you ask the engineers who put up the GPS.” So the GPS and ECI items do have an ancestor in the authors’ own writing. An earlier version of this section said the case was never restated with an instrument named, and that was wrong; it is withdrawn here. What the 2010 sentence claims is ease of use. It does not claim fixity, and the gap between “easier to compute in” and “therefore at rest” is the whole of this cluster. So the argument still has to be met where it is made, and where it is made it is conventionalist.

The Einstein sentence, with its condition restored. The quotation is real and it is quoted accurately, but it arrives on the list stripped of the clause that governs it. Einstein and Infeld are making a conditional: if the laws of physics can be formulated for arbitrary frames of reference, then the old struggle between the two systems would be meaningless, and either coordinate system could be used with equal justification. That is a claim about what follows from general covariance. It is not a claim that the Earth is at rest, and Einstein plainly did not hold that it is.

Now the answer: a conventional origin is not a conventional dynamics. This is the hinge, and it survives granting the conventionalist everything they ask for. Put the origin wherever you like — nothing physical changes. What you cannot do is thereby make the fictitious-force terms disappear. Write Newton’s laws in a frame rigidly attached to the Earth’s surface and they are false as stated: you must add a centrifugal term and a Coriolis term to recover agreement with observation. Write them in a frame that does not rotate with respect to the distant matter of the universe and you need add nothing. That asymmetry has nothing to do with where the origin sits. It sorts all frames into two classes, and the sorting is measured, not stipulated.

The coefficient of those extra terms is a number, ω. It is the same ω that appears as a defining constant of WGS 84, as the rate parameter in the IERS Earth Rotation Angle, as the Sagnac term in GNSS pseudorange processing, and as the reading of a ring laser gyroscope in a sealed room with no view of the sky. Conventionalism is a thesis about description. It cannot make a measured number go away.

And the conventionalist reading costs its holder the rest of the list. If the choice really is a convention with nothing to decide between the options, then no experiment can decide it either — which retires Michelson–Gale, Sagnac, “Airy’s failure”, Michelson–Morley and the microwave-background alignments in a single stroke, since each of those asserts that a measurement came out the geocentrist’s way. The same trade is set out at R01. It is available, it is honest, and the list cannot afford it.

The twenty-eight items are, in the main, the list’s addition, not the authors’. Everything above answers the book. What follows answers the items, which are a different thing and need saying so. Each is a bare noun phrase — “WGS84 ECEF.”, “LiDAR ECEF.” — and none of them contains an argument. The premise that turns them into one is supplied by the page they sit on, headed “435 Pieces of Evidence The Earth is Not A Spinning Ball”. Twenty-eight technical artefacts have been added to the source’s account by later hands — excepting the GPS and ECI ones, which the authors do take up elsewhere, as above — and the inference attached to them — a practical system uses Earth-fixed coordinates, therefore the Earth is fixed — is strictly weaker than the conventionalism it replaced.

We answer them anyway, because they are what circulates and what a reader arrives with. The technical range creates an impression of independent lines converging; in fact every item has the identical logical form, so twenty-eight restatements of one invalid inference are not twenty-eight pieces of evidence. They sort into seven kinds — and the striking thing, which the compilers cannot have intended, is how many of them carry the Earth’s rotation inside their own definitions.

Kind 1 — the standard names its own rotation rate. WGS84 ECEF, geodesy Earth-centered, GPS Earth-centered, radar ECEF, LiDAR ECEF, drone RTK, GNSS pseudorange. WGS 84 has exactly four defining parameters, and NGA lists them as: semi-major axis a = 6378137.0 m; reciprocal flattening 1/f = 298.257223563; geocentric gravitational constant GM = 3986004.418 × 10⁸ m³/s²; and the angular velocity of the Earth, ω = 7292115 × 10⁻¹¹ rad/s. Two of the four bear directly on the question. The flattening is there because an oblate spheroid — TR8350.2 calls the reference figure “a geocentric ellipsoid of revolution” — is the equilibrium shape of a self-gravitating body that spins; a non-rotating fluid Earth would be round. And ω is not derived, inferred or optional: it is one of the four constants from which the rest of the datum is computed. TR8350.2 §3.2.4 defines it as the rate of “a standard Earth rotating with a constant angular velocity,” and adds that “the actual angular velocity of the Earth fluctuates with time.” The coordinate system offered as proof that the Earth does not turn is defined by how fast it turns, and carries a footnote about the irregularities in that turning.

Kind 2 — the item is about a transformation that exists only because of rotation. ECI-ECEF transforms, navigation inertial Earth base, airliner FMS Earth grid, ephemeris time. IERS Conventions (2010) ch. 5 gives the terrestrial-to-celestial transformation as [GCRS] = Q(t) R(t) W(t) [ITRS], where W(t) handles polar motion, Q(t) precession-nutation, and R(t) is a rotation through the Earth Rotation Angle, ERA(T_u) = 2π(0.7790572732640 + 1.00273781191135448 T_u). On a non-rotating Earth, R(t) would be the identity matrix, Q(t) and W(t) unnecessary, and ECI and ECEF the same frame — there would be no transformation to name. The rate constant repays reading: 1.00273781191135448 revolutions per UT1 day is the sidereal-to-solar ratio, and the excess 0.00273781 is precisely the one extra rotation per year that an orbiting body accumulates. An item citing “ECI-ECEF transforms” as evidence for a fixed Earth is citing a matrix whose entire content is the Earth's rotation and orbit.

Kind 3 — clocks, where rotation is a first-order effect. Ashby's Living Reviews in Relativity survey is explicit that GPS synchronisation is carried out in the Earth-Centred Inertial frame precisely because synchronisation in the rotating frame is not self-consistent: “Path-dependent discrepancies in the rotating frame are thus inescapable … while synchronization in the underlying inertial frame … is self-consistent.” For an eastward equatorial circumnavigation the discrepancy is 207.4 ns. In receiver processing the same physics appears as the Sagnac correction: Farrell and Hu (J. Geodesy 98:102, 2024) note the ECEF frame rotates during signal propagation, and that neglecting it puts the computed satellite position in error by over 30 metres, with roughly 20 metres of user-position shift at mid-latitudes. Every one of these numbers is exactly zero on a stationary Earth.

Kind 4 — the frame requires a continuously measured rotation series. The ITRF is not usable alone; the transformation to the celestial frame requires the Earth Orientation Parameters — polar motion, UT1−UTC, length of day, celestial pole offsets. These are not computed from theory. They are measured continuously by VLBI on quasars, satellite and lunar laser ranging, and GNSS, because, as the US Naval Observatory puts it, “the rotational speed of the Earth remains essentially unpredictable in nature due to incompletely understood variations,” with length-of-day departures of 0.001 to 0.002 seconds. The Earth-fixed frame presented as evidence of fixity is maintained by an international service whose full-time job is publishing bulletins of how the Earth's rotation changed last week.

Kind 5 — the pipeline runs the other way. JPL's DE440/DE441 equations of motion are integrated in a barycentric frame tied to the ICRS, in barycentric dynamical time. The Horizons documentation states that all underlying calculations are done in the reference frame of the planetary ephemeris, taken as indistinguishable from the ICRF, and results are then transformed to whatever geocentric or topocentric output the user requests. The geocentric columns in a nautical almanac are the product of a barycentric integration, not an alternative to one. The item worded “barycentric transforms yield observer ephemerides” describes this sequence correctly and draws the reverse conclusion from it. The parallax item inverts itself more sharply still: the parallax equation begins from the Earth because the Earth moves, the baseline being the diameter of its orbit. No motion, no baseline, no equation.

Kind 6 — self-refuting on its own geometry. A great circle is by definition the intersection of a sphere with a plane through the sphere's centre; great-circle distance is a quantity that exists only on a sphere. Airline and marine route planning uses it because the Earth is a globe, and the modern refinement goes further in the same direction: Karney's geodesic algorithms (J. Geodesy 87:43–55, 2013) compute shortest paths on an ellipsoid of revolution — the sphere corrected for the very flattening that rotation produces. There is no reading of “great-circle math” that supports a flat or non-rotating Earth; the item names the sphere in its own title.

Kind 7 — seismology and infrasound. Hypocentres are reported in Earth-fixed latitude, longitude and depth, but the location is obtained by inverting travel times against a velocity model — and the standard models, PREM and ak135, are spherically symmetric, radially stratified models of a layered globe. Dziewonski and Anderson's PREM specifies an Earth radius of 6371 km and divides the interior into ocean, crust, lithosphere, low-velocity zone, transition zone, lower mantle, outer core and inner core. You cannot locate an earthquake in Earth-fixed coordinates without assuming a spherical Earth with a liquid outer core.

And the honest remainder. Several items are simply true statements about engineering convenience and should be granted without qualification: marine sextant work is done in a geocentric-topocentric framework; naked-eye positional astronomy is topocentric; VOR bearings are referenced to the ground; “centre equals measurement origin” is a fair description of how origins get picked. You do navigate in the frame you stand in, and nobody disputes it. These items are accurate. What they are not is evidence about the Earth's motion — any more than a ship's captain plotting positions relative to his own vessel is evidence that the vessel is at rest.

The list says

Modern astronomy uses Earth-centered coordinates. / GPS Earth-centered. / Marine sextant geocentric. / Radar ECEF coordinates. / Nav systems Earth-fixed. / Satellite ground tracks Earth-based. / Solar tables geocentric. / GPS Earth-centered corrections. / ECI/ECEF transforms geocentric. / Navigation inertial Earth base. / Barycentric transforms yield observer ephemerides. / Ephemeris time Earth-defined. / Parallax equations Earth start. / Center equals measurement origin. / Metrology standard Earth clocks. / Naked-eye astronomy topocentric. / Geodesy Earth-centered. / WGS84 ECEF. / ITRF Earth crust reference. / Airliner FMS Earth grid. / VOR network Earth bearings. / Great-circle math. / Nautical almanac geocentric. / Drone RTK Earth grid. / LiDAR ECEF. / Seismology Earth-fixed. / Volcano infrasound triangulation. / GNSS pseudorange Earth clocks.

The source says

“Either coordinate system could be used with equal justification. The two sentences: ‘the sun is at rest and the Earth moves,’ or ‘the sun moves and the Earth is at rest,’ would simply mean two different conventions concerning two different coordinate systems.” — quoted by Sungenis & Bennett from Einstein & Infeld, where it stands under a conditional: If this can be done, our difficulties will be over … the struggle “would then be quite meaningless.”

The source does not contain this.

The list supplies both halves of the inference. Sungenis and Bennett argue conventionalism — that the Ptolemy–Copernicus dispute is about description rather than fact — and they argue it by quoting Einstein and Infeld, Poincaré and Bellarmine. They nowhere argue that GPS, WGS 84, ITRF, seismic hypocentres or a nautical almanac are evidence that the Earth does not move, and Sungenis’s own later restatements of the case reach for the Einstein quotation without naming a single instrument. Twenty-eight technical artefacts have been added to the source’s account. The premise that turns them into an argument comes from the page they sit on, which is headed “435 Pieces of Evidence The Earth is Not A Spinning Ball”; the items themselves are bare noun phrases. Two smaller drifts ride along. The Einstein sentence is conditional in its original — if the laws can be written for arbitrary frames, the old struggle would then be meaningless — and arrives here unconditioned; and a permission is again doing the work of a proof, the pattern already recorded at R01. Our refutation was rewritten on 2026-08-05 because of this finding. It now answers the book first — the Einstein sentence with its condition restored, and the reply that a conventional origin is not a conventional dynamics — and reaches the twenty-eight artefacts only afterwards, under a heading saying whose they are.

The system may have a symmetry which matches that of the coordinate system and simplifies the mathematical clutter used in its description…. Nevertheless, any reasonable coordinate system may be used. The weave pattern of a net does not determine the shape of the objects that are put into it.

— Galileo Was Wrong: The Church Was Right (2006), Vol. II (7th ed., 2013, chs 7–13), ch. 10, section “General covariance”, p. 171 of the archive.org scan; the operative technical derivation is nine pages earlier, at p. 161

This is the whole of ARG-R06 in three sentences, and it is correct. The section around it is an accurate textbook statement: general covariance is the invariance of the form of physical laws under arbitrary coordinate transformations, all physical theories admit a generally covariant formulation, and a coordinate system is a description rather than a constituent of the thing described. Nothing here needs correcting.

What is worth noticing is where the net metaphor points. If the weave of the coordinate net does not determine the shape of what is put into it, then laying an Earth-centred net over the solar system does not make the Earth the centre of anything either. The metaphor is an argument for the irrelevance of coordinate choice, and irrelevance is symmetric.

The actual physics sits at p. 161, where the authors write the Lagrangian for a test body in a frame with the Earth spinning at angular velocity ω, obtain the equation of motion containing the Coriolis term −2m(ω×v) and the centrifugal term −m[ω×(ω×r)], then write the corresponding Lagrangian with a stationary Earth and a rotating star shell generating a field they label B, and observe that the two equations of motion coincide when ω = B/2. That is a clean and honest demonstration — of a change of coordinates. It is also the source of the terms that make the rotation measurable.

SURFACE (the version to avoid). “Fifteen pieces of jargon dressed over one point.” Descriptively true and argumentatively worthless. It invites the reader to dismiss rather than to check, and it does not engage the reason a technically literate person finds this cluster persuasive.

DEEPER. Coordinate freedom is genuine. You can write mechanics, electrodynamics and general relativity in an Earth-fixed rotating chart and compute correct answers for eclipses, occultations and satellite clock rates. Working astronomers do this routinely; ephemerides are published in Earth-centred frames because that is where the observers are.

KERNEL — and it belongs to Erich Kretschmann. In 1916 Einstein presented general covariance as the foundational principle of general relativity, the thing that generalised the relativity of motion beyond inertial frames. In 1917 Kretschmann replied that the demand places no constraint on physics at all: any spacetime theory can be given a generally covariant formulation if one is willing to do the reformulating work. Special relativity can be written in arbitrary coordinates; Newtonian gravitation was later given a fully geometric, generally covariant form by Cartan. If every theory can wear the costume, wearing it distinguishes nothing. Einstein conceded a great deal of this, and the resulting dispute ran for eight decades.

So the serious position underneath R06 is not a misunderstanding. It is this: the formalism of general relativity, considered as formalism, does not single out a preferred centre or a preferred state of rotation, and anyone who claims the equations by themselves refute a stationary Earth is claiming more than Kretschmann allows. That objection is roughly 110 years old, it was pressed by a professional physicist against Einstein directly, and Einstein had to answer it. Conceding it costs us nothing; refusing costs us the reader.

Why it doesn’t save the claim.

Kretschmann's blade has two edges and the second is fatal here. If general covariance has no physical content, it has none to lend geocentrism. An argument that the formalism is neutral cannot then be cited as evidence that the formalism favours you. At best it establishes that this question is not settled by the choice of coordinates — which is what everyone on the other side has been saying.

What does the discriminating work in general relativity is not covariance but the equivalence principle together with a dynamical metric. Spacetime geometry is not a passive labelling scheme; it is a physical field with its own degrees of freedom, sourced by stress-energy, carrying its own initial-value problem, and detectable in its own right — gravitational waves are that field oscillating. Coordinate labels can be changed at will; the metric field cannot. Write the metric in an Earth-fixed rotating chart and it acquires off-diagonal time–angle components. Those components are not bookkeeping. They are the Coriolis and centrifugal effects, and they are measured: by Foucault pendulums, by the Michelson–Gale fringe shift, by the Wettzell ring laser tracking length-of-day variation below one part in 10⁹, and by the 207.4 ns path-dependence GPS must correct for around an eastward circumnavigation. The authors' own p. 161 derivation produces precisely these terms and names them.

A change of chart is free; the field that appears in the chart is not, and it is locally measurable.

What the source says is correct, and that is why the verdict is what it is. The passage this cluster rests on states that a system may have a symmetry matching that of the coordinate system, which simplifies the description — and that nevertheless any reasonable coordinate system may be used. That is right. It is what every general relativity course teaches, it is disputed by nobody, and the authors state it accurately. The verdict standard physics attaches to that sentence.

Why being right does not help them. Kretschmann’s 1917 objection has two edges and the second is fatal here: if general covariance carries no physical content, it has none to lend geocentrism either. An argument that the formalism is neutral cannot then be cited as evidence that the formalism favours you. At best it establishes that the question is not settled by a choice of coordinates — which is what the other side has been saying throughout.

What does the discriminating work in general relativity is not covariance but the equivalence principle together with a dynamical metric. Spacetime geometry is not a passive labelling scheme; it is a physical field with its own degrees of freedom, sourced by stress-energy, carrying its own initial-value problem, and detectable in its own right — gravitational waves are that field oscillating. Coordinate labels can be changed at will; the metric field cannot. Write the metric in an Earth-fixed rotating chart and it acquires off-diagonal time–angle components. Those are not bookkeeping. They are the Coriolis and centrifugal effects, and they are measured: by Foucault pendulums, by the Michelson–Gale fringe shift, by the Wettzell ring laser tracking length-of-day variation below one part in 10⁹, and by the 207.4 ns path-dependence GPS must correct for around an eastward circumnavigation. The authors’ own derivation, nine pages before the passage quoted above, produces precisely these terms and names them. A change of chart is free; the field that appears in the chart is not, and it is locally measurable.

Six of the fifteen item titles could not be located in the volume we were able to search. This needs stating before anything is refuted, because the rest of this section would otherwise attribute to two named authors a vocabulary we cannot show they used. A full-text search of that scan — chs 7–13 of Vol. II of the seventh edition (2013) — returns zero occurrences of “Noether”, zero of “Fermat”, and zero of “light cone”. “Gauge” occurs only as Ron Hatch’s clock-rate scale factor, never as gauge symmetry. “Renormalization” occurs four times, all as a complaint that infinities in quantum field theory are swept under the rug, never in connection with frames. “Spontaneous symmetry breaking” does not occur in it. And “anthropic” occurs three times with the authors attacking the anthropic principle, which would make the item “Anthropic coordinate justification” a statement of the reverse of their position.

The caveats travel with the finding. The scan is one volume of a multi-volume work; OCR quality is variable, rendering “All” as “A11” and omega as “co”; and we were unable to re-run the search independently. The volume label carried an error of our own, and it has been corrected: the archive item our source list points at is titled Galileo Was Wrong (2), and its title and copyright pages give Vol. II of the seventh edition, 2013, chapters 7 to 13 — not volume I, as this entry cited it until now. The passage locator and the source line have been changed to match the scan. That correction narrows the finding rather than strengthening it: Volume I, which our cluster record names as the source, we have not been able to search at all, no full text of it having been located. The defensible statement is therefore not located in chs 7–13 of the scanned Vol. II, not the authors never wrote this. On the anthropic item specifically there is contrary testimony — Peter Woit, writing on the film The Principle, describes Sungenis as taking anthropic cosmology to support his view — so we leave that one unsettled rather than claim a reversal. These vocabularies do circulate in downstream forum material, and that is the likeliest route by which they reached the list.

So what follows is addressed to the items, whose author we do not know. They are answered because they circulate, not because they are the book’s. Fifteen vocabularies create the impression of fifteen convergent lines of support; they are one line, restated, and an argument that fails once fails fifteen times.

Kind 1 — correct physics, inference does not follow (nine of the fifteen). “Metric tensors Earth-centered”, “Covariance of tensor math”, “Gauge choice alters story only”, “Geodesics drawn from observers”, “Variational minimal paths Earth frame”, “Hamiltonian Earth-origin simplicity”, “Boundary conditions Earth surface”, “Conservation in Earth frame” and “Fermat with rotating medium” are one statement: the equations can be written in any coordinates. Yes. Tensor equations are covariant by construction — that is the definition of a tensor equation, not a discovery about the Earth. Geodesics are fixed by the metric and the affine connection; the observer who draws them contributes nothing to where they run. Fermat's principle in a moving medium is real optics (it is how Fresnel dragging is described), but it concerns light in matter, not the arrangement of the solar system. The Hamiltonian item is weakest, because it appeals to simplicity — and on the page facing the covariance passage the authors reject simplicity as a criterion, arguing that Occam's razor can conceal complexity beneath a veneer. If simplicity is not evidence when it favours general relativity, an Earth-origin Hamiltonian being tidier is not evidence either.

Kind 2 — a term imported from a domain where it means something else. “Gauge freedom equality” treats gauge freedom as a synonym for coordinate freedom. There is a real family resemblance worth stating honestly: both are redundancies in a description, and in the constrained-Hamiltonian treatment of general relativity diffeomorphism invariance is handled as a gauge freedom. But the gauge symmetries of the Standard Model act on internal field space, not on spacetime. The U(1) of electromagnetism rotates the phase of the electron field; the SU(3) of quantum chromodynamics rotates colour. Performing one does not move you to a different origin, frame or state of rotation, and only gauge-invariant quantities are observables.

Kind 3 — a term whose actual content runs the opposite way. “Noether charges Earth-defined” and “Symmetry breaking selects Earth” invert their own physics. Noether's first theorem derives conserved quantities from continuous symmetries of the action: time-translation invariance yields energy, spatial-translation yields momentum, rotational yields angular momentum. The direction of derivation is symmetry → charge. Choosing an Earth-centred origin does not create a symmetry, and no charge follows from a choice of origin. The deeper point cuts harder: Noether's second theorem applies when the invariance is local rather than global — exactly the case in a generally covariant theory — and there the would-be conservation laws collapse into identities satisfied by the field equations rather than into independent charges. This is why gravitational energy in general relativity has no local density; the standard constructions are pseudotensors, and a pseudotensor can be made to vanish at any chosen point by a choice of coordinates. Of all these vocabularies, Noether's is the one that most emphatically refuses to attach a conserved charge to a coordinate origin. Spontaneous symmetry breaking is likewise about a field's ground state failing to share the symmetry of its Lagrangian; the asymmetry lives in field space. It does not select spatial locations.

Kind 4 — equivocation on a word. “Renormalization frame Earth lab” appears to trade on the word “frame”. The renormalization scale μ_R is an energy, quoted in GeV; the Particle Data Group describes it as unphysical and states that physical observables do not depend on unphysical scales. It has no spatial orientation, no origin and no state of motion. Even granting the item its best reading, it would establish that the choice makes no difference — the opposite of what a preferred-frame argument needs.

Kind 5 — the item with genuine physics in it. “Light cone tilt reinterpretation” is the only member of this cluster pointing at something real. Light cones do tilt. Inside the ergosphere of a rotating black hole the tilt is severe enough that no observer can remain static relative to infinity. In rotating coordinates the Langevin metric has a g_tt coefficient proportional to (1 − ω²r²/c²), which vanishes at r = c/ω — the light cylinder, a radius that also does real work in pulsar magnetosphere physics. Beyond it no observer can be at rest in those coordinates. But read what that says. It is a statement about the coordinate patch, not about the world: coordinate velocities in a non-inertial chart are not bounded by c; only locally measured velocities are. A chart failing at r = c/ω is a chart of limited domain, exactly as Schwarzschild coordinates fail at the horizon while the spacetime does not. And the direction of the difficulty matters: this is a problem for models in which the universe physically rotates about a stationary Earth, since there the distant stars traverse their circuits once a day and those beyond r = c/ω move superluminally. The standard picture has no such object.

What the cluster amounts to. One correct sentence from the source, answered at R01 where it is made most carefully, plus fourteen restatements and six titles we cannot trace to the volume we could search. R06 adds terminology and no new premise. The count is an artefact of the thesaurus, not of the evidence.

The list says

Metric tensors Earth-centered. / Fermat with rotating medium. / Gauge freedom equality. / Covariance of tensor math. / Hamiltonian Earth-origin simplicity. / Anthropic coordinate justification. / Variational minimal paths Earth frame. / Light cone tilt reinterpretation. / Symmetry breaking selects Earth. / Noether charges Earth-defined. / Conservation in Earth frame. / Boundary conditions Earth surface. / Gauge choice alters story only. / Renormalization frame Earth lab. / Geodesics drawn from observers.

The source says

“The system may have a symmetry which matches that of the coordinate system and simplifies the mathematical clutter used in its description…. Nevertheless, any reasonable coordinate system may be used. The weave pattern of a net does not determine the shape of the objects that are put into it.” (Galileo Was Wrong Vol. I, ch. 10, “General covariance”)

The source does not contain this.

The cited source says one thing, and says it correctly: any reasonable coordinate system may be used. What it does not supply is the vocabulary. The full-text search logged in this review’s corrections file found no occurrence of Noether, Fermat or light cone in the scanned volume, gauge only as Ron Hatch’s clock-rate scale factor, and renormalization only as a complaint about infinities in quantum field theory. Six of the fifteen titles are on that showing not the authors’ claims at all but names attached to their position by later hands. We could not re-run the search independently: every route open to us either refused the file or returned a truncated view whose silence proves nothing — and the archive item cited in our own source list is labelled Galileo Was Wrong (2), not volume I, which should be resolved before the count is leaned on. A second, weaker drift sits inside the first: “Anthropic coordinate justification” recruits a principle the authors are recorded as attacking, which would make that one item reversed; but Woit, writing on the film The Principle, describes Sungenis as taking anthropic cosmology to support him, so we leave it unsettled. Our refutation was rewritten on 2026-08-05 because of this finding. It now answers the one sentence the source does supply, states the six unlocatable titles and the caveats on that count before refuting anything, and addresses the remaining items as claims of unknown authorship rather than the authors’.

the reigning relativity can indeed not pillory an Earth-centered cosmology. … I can stick to my geocentric guns. If Einstein is right the Tychonian quest amounts simply to forcing an open door.

— De Labore Solis: Airy's Failure Reconsidered (1988), section “Testing Einstein! Why? He Can't be Wrong!”, p. 67

This is the founder of the modern Tychonian movement conceding, in his own book, that the relativistic version of his case is a hollow win. Van der Kamp has just canvassed the standard reading — he quotes Martin Gardner's “from the standpoint of relativity the choice of reference frame is arbitrary” — and accepts that on that reading nothing in relativity can rule an Earth-centred cosmology out. His own word for the result is forcing an open door: if the door is already open, walking through it proves nothing about the room.

His response is to reject the door. Later in the same book he calls “the general theory … in its present form untenable” and says relativity's experimental support is “obtained by affirming the consequent and is therefore not in the least compelling.” He wanted a control experiment on a fast-moving platform — a modus tollens, in his phrasing “If P, then Q”, “If no Q, then no P” — and framed the enterprise through Popper and Herbert Dingle. That logical point is sound as far as it goes: confirmation is not proof.

A stationary Earth granted by covariance is granted to everyone, on the same terms, for free. Sungenis and Bennett later promoted that consolation prize to a headline, and the 461-item lists compress it further into flat assertions such as “Relativity permits stationary Earth frame” — losing the qualification their own founder insisted on.

SURFACE (weak — do not use). “They do not understand relativity.” False here, and anyone who runs it will lose the exchange. Sungenis and Bennett's own gloss — general covariance is “the invariance of the form of physical laws under arbitrary coordinate transformations”, and “all physical theories such as mechanics and electrodynamics must necessarily have a generally covariant formulation” — is textbook-correct, and their charge that people who insist astronomy must be done heliocentrically confuse “the intrinsic physical properties and relationships of a system” with “the arbitrary mathematical coordinate system used to describe” it is a fair hit on a common popular error.

DEEPER. The underdetermination is genuine, not rhetorical. Kinematically, “the Earth rotates once per sidereal day relative to the local inertial structure” and “the local inertial structure rotates once per sidereal day relative to the Earth” describe the same relative motion. Mach took that seriously, Einstein named Mach's principle among the theory's motivations, and Thirring's 1918 result showed that a rotating mass shell does drag inertial frames inside it — the beginning of a mechanism rather than a hand-wave.

KERNEL. The strongest form is not a physics claim at all but a claim about what physics can decide. General covariance is real. Kretschmann's objection to reading physical content into it is real, has stood since 1917, and is now mainstream — Norton notes that many modern treatments of general relativity “no longer even mention a principle or requirement of general covariance.” And Gerardus Bouw, the movement's only credentialed astronomer, reached the honest end point in print: replying to Faulkner on his own society's site he grants that “the geocentric and heliocentric models can both account for the observed motions of the universe,” quotes Hoyle that the difference “is one of relative motion only, and that such a difference has no physical significance,” and refers the decision to Scripture. That is a philosopher-of-physics argument plus a candid statement of where the choice actually gets made. Concede every word of it.

Why it doesn’t save the claim.

Because the kernel's conclusion is the coordinates do not settle it, and the claim on the list is the Earth is stationary. Those are not the same sentence, and the gap between them is where the whole argument lives.

Kretschmann's blade cuts both ways, and the second cut is the one that matters. If general covariance places no constraint on physical content — because special relativity, Newtonian gravitation and anything else can be written in arbitrary coordinates — then it constrains nothing in favour of a stationary Earth either. A principle that permits every cosmology confers no distinction on this one. Van der Kamp put it more sharply than any critic: an open door.

And Bouw's version of the kernel is not free. Observational equivalence is symmetric. If no measurement discriminates, then no measurement is evidence — which retires the interferometry, aberration and microwave-background items that make up most of the rest of this list.

First, the concession, without hedging. General relativity is a generally covariant theory. Its field equations are tensor equations, and a tensor equation that holds in one smooth coordinate system holds in every one related to it by a diffeomorphism. There is no preferred frame in general relativity. You may adopt coordinates rigidly attached to the rotating Earth, integrate the field equations there, and obtain correct predictions for every observable; geodesy, satellite navigation and VLBI do it as routine business. Anyone answering this argument by asserting that physics forbids an Earth-fixed frame is wrong.

Second, what general covariance is worth. Erich Kretschmann's 1917 reply to Einstein is the standard answer and it has never been overturned. Norton's summary: Kretschmann held that in demanding general covariance “Einstein had placed no constraint on the physical content of his theory. He had merely challenged his mathematical ingenuity”, since “any spacetime theory could be given a generally covariant formulation as long as we are prepared to put sufficient energy into the task of reformulating it.” Norton then works the example: special relativity, ordinarily written with the Minkowski line element in standard coordinates, becomes a generally covariant theory the moment one writes ds² = g_ik dxⁱ dx^k with the field equation R_iklm = 0 and geodesic equations of motion. Special relativity has an absolute inertial structure and is still generally covariant. So covariance cannot be what tells you whether a theory has absolute structure — which is what this argument needs it to tell you.

Norton's own refinement is worth stating rather than hiding. He argues that once a theory has been given a generally covariant formulation, that covariance does express something: a gauge freedom, a redundancy in the description. That is still a fact about descriptions. A redundancy in how you label events is not a fact about which events happen.

Third — and this is the crux — relabelling is not a rival model. Two quite different things go under the same slogan, and the argument runs on the equivocation between them.

(a) The chart change. Take the standard solar system — one spacetime, one metric, one set of worldlines — and describe it in coordinates fixed to the Earth. Nothing physical has been altered. Every point-coincidence is preserved: the same eclipses at the same events, the same pulsar arrival times, the same ring-laser fringe count, the same 207.4 ns of accumulated Sagnac offset. This is what covariance guarantees. It proves nothing about the Earth, because the identical move yields a Jupiter-fixed or a Ceres-fixed chart on the same terms, and nobody takes those for cosmologies.

(b) The physical claim. That the Earth is truly non-rotating with respect to the local inertial structure, and that the sun, planets, stars and distant galaxies physically circle it once a sidereal day. This is a different model of the world, not a re-labelling of the standard one. It has different content, so it needs its own dynamics: a matter distribution, a stress-energy tensor, a solution, and a demonstration that it reproduces the observations. Covariance supplies none of that. It tells you that whatever the true solution is, you may write it in any coordinates; it does not tell you which solution is true. The argument moves from (a) to (b) as though the licence for the first carried over to the second, and it does not.

Fourth, what does discriminate. Not the coordinates — the metric and the inertial compass. Write the Earth-fixed rotating chart honestly and the metric acquires off-diagonal time-angle components. Those terms are not decoration; they are the Coriolis and centrifugal effects, and they are locally measurable. Ashby's review derives the resulting path dependence directly: in the rotating frame, synchronisation around a closed equatorial loop fails to close by 207.4 ns, and “path-dependent discrepancies in the rotating frame are thus inescapable”, which is why GPS synchronises the whole constellation in the Earth-Centred Inertial frame instead. Foucault, Michelson–Gale–Pearson and modern ring lasers measure the same quantity by other means: the G ring laser at Wettzell resolves Earth's angular rate to better than one part in 10⁹, tracking the Chandler wobble, diurnal polar motion and solid-Earth tides. Note what this does and does not show. It is not “the Earth-fixed chart is illegal.” It is that rotation relative to the local inertial structure is a coordinate-free invariant, it is non-zero for the Earth, and it has been measured to nine digits. Model (b) says that quantity is zero — a claim about the world, which the world has answered.

Fifth, what the physical version costs. A rigidly co-rotating cosmos puts everything beyond r = c/ω — about 4.1×10¹² m, roughly 27.5 AU, so from about Neptune outward — at a coordinate speed exceeding c. Be fair about this: it is not an instant refutation. Coordinate velocities in a non-inertial chart are not bounded by c, and no light cone is violated by the arithmetic alone. But under reading (a) the figure is an artefact of the labels and carries no commitment, while under reading (b) it describes real matter really moving, and a mechanism is owed. Sungenis and Bennett's published answer is the decisive datum: asked how stars can circle faster than light, they reply that superluminality “is inherent in Galilean relativity, which has no limit for the velocity of physical objects”, and elsewhere in the same volume they identify the Earth-Centred Inertial frame as “one universal absolute preferred frame in which c is isotropic.” That is a preferred-frame ether theory. It may be argued on its own merits, but it is the negation of the covariance argument, not an extension of it. The two cannot be run together: one says no frame is preferred, the other says one is.

Sixth, the concession is the argument's own end state — and it is expensive. Faulkner's review states the position plainly: “Mathematically, the essential difference between the heliocentric and Tychonian models is a co-ordinate change from the Sun to the Earth.” Bouw does not dispute it; he grants that the two models “can both account for the observed motions of the universe”, endorses Hoyle's judgement that the difference “has no physical significance”, and refers the decision to Scripture. Follow that out. Equivalence runs both ways: if no observation can discriminate, then Michelson–Gale cannot be evidence for a stationary Earth, and neither can the Sagnac effect, “Airy's failure”, Michelson–Morley, Miller's drift or the microwave-background alignments. Every one of those items asserts that an experiment came out the geocentrist's way, which presupposes that experiments can decide. R01 says they cannot. The list runs an argument and its own refutation side by side, and the defender must pick one. Bouw picked, the honest way: on grounds outside physics.

Seventh, the genuine GR neighbours. Rotating solutions of the field equations exist and should not be waved away. Gödel's 1949 solution is a real, exact, rotating cosmology — but it does not expand and contains closed timelike curves through every event, so it is not the universe we observe. Frame dragging is real and measured: Gravity Probe B returned a Lense–Thirring drift of −37.2 ± 7.2 mas/yr against a predicted −39.2. That is a genuine confirmation of Machian-flavoured effects in general relativity, and it is roughly ten orders of magnitude too small to be the diurnal rotation — about 1×10⁻⁹ degrees per hour against 15 degrees per hour. Machian sympathy is not a completed dynamics, and none of the geocentric works read for this review — van der Kamp 1967, Bouw Geocentricity 1992, Sungenis & Bennett Galileo Was Wrong — supplies one.

Verdict: standard physics. The true content of the argument — the field equations may be written in Earth-centred coordinates — is taught in every general relativity course and disputed by nobody. It is a statement about the freedom to relabel events. It does not become a statement about which events occur, it does not distinguish the Earth from any other body one might fix a chart to, and the moment its defenders press it into a physical rotating-cosmos model they trade covariance for a preferred-frame ether and forfeit the rest of the list with it.

The list says

Relativity permits stationary Earth frame.

The source says

“the reigning relativity can indeed not pillory an Earth-centered cosmology. … If Einstein is right the Tychonian quest amounts simply to forcing an open door.”

A concession was re-used as a proof. This is the sharpest case on the page, because the wording barely drifts at all — “relativity permits a stationary-Earth frame” is a fair paraphrase of “relativity cannot pillory an Earth-centred cosmology.” What changes is what the sentence is for. In van der Kamp it is a concession, and he says so in the same breath: a win by relativity is forcing an open door, and an open door proves nothing about the room. On the list the identical proposition appears as proof item 26, in a numbered list of evidence for a stationary Earth. A permission has been promoted to a demonstration without a word being altered. The refutation above answers van der Kamp's version — which is why it concedes the covariance point outright instead of disputing it, and puts the weight on what the concession costs the other 150-odd items that need experiments to be decisive.

Bertrand Russel's contention that the observable phenomena will be the same whether the Earth rotates or the Heavens revolve, as well as Fred Hoyle's declaration that the geocentric view is as good as anybody else's, but not better, they are only tenable if certain presuppositions are assumed to be self-evident. Which they are not!

— De Labore Solis: Airy's Failure Reconsidered (1988), p. 62 of the PDF scan at geocentricity.com, in the discussion of Russell and Hoyle; retrieved twice on separately worded passes over the same scan, not checked against a print copy

Read who is speaking. The proposition the list credits to van der Kamp is, in his own book, other people’s — Russell’s and Hoyle’s — and he names them in order to say their claims will not stand on their own. (“Russel’s” is the spelling in the scan.) He quotes Russell earlier, at pp. 12–13, in the famous form: whether the Earth rotates west to east or the heavens revolve east to west, “the observable phenomena will be exactly the same”, which Russell calls a defect in Newtonian dynamics because “an empirical science ought not to contain a metaphysical assumption, which can never be proved or disproved by observation.” Russell is a hostile witness and that is exactly why van der Kamp wants him. But wanting a hostile witness is not the same as owning his conclusion, and by p. 62 van der Kamp is refusing to own it.

What he wanted instead. The book opens with a demand for a test, not a stalemate. He asks for “at least one control experiment” (p. 6), sets it out as a modus tollens — if the speed of light measured from a fast-moving platform comes out at the earthly c, “he stands vindicated”, and if it changes with the platform’s speed, “then he will be discredited” (p. 7) — and warns against reading too much into any null: “if in the Sahara no icefields can be found, this observation does not thereby prove that icefields exist nowhere” (p. 6). His relativistic argument runs under an antecedent he rejects, “If Einstein is right …”, and that sentence and its cost are worked in full at ARG-R01.

On the work cited. Our cluster record for R03 credits van der Kamp’s earlier booklet The Heart of the Matter (1968). No copy of that booklet was reachable from here, so this treatment quotes and cites only De Labore Solis (1988) — the published book, and the text later geocentrist writers work from. What is documented about the 1968 item is Bouw’s obituary account: a rough draft privately circulated in 1967, printed in January 1968, which in Bouw’s words “went nowhere fast.”

SURFACE (weak — do not use). “Of course absolute motion is detectable — look at a Foucault pendulum.” This loses the exchange in one move, because a pendulum detects rotation, not velocity through space, and anyone who conflates the two has conceded that they do not know what the relativity principle says. Equally weak: “Michelson–Morley was a failed experiment.” It was a successful one, and its result is the claim under discussion.

DEEPER. The proposition is simply true, and it is not a hedge or a shrug. No experiment has detected uniform motion relative to a substrate, and the modern versions are brutal about it: rotating optical cavities bound a direction-dependence of the speed of light at the 10⁻¹⁷ level (Herrmann et al., Phys. Rev. D 80:105011, 2009) and at 10⁻¹⁸ (Nagel et al., Nature Communications 6:8174, 2015). A defender of the list who says only this has said something correct that no physicist will dispute.

KERNEL. The strongest form is not the list’s and it is not a physics claim at all. It is van der Kamp’s, and it is a claim about what a null result is worth. He accepted that the observations tie; he refused to accept that a tie in the observations settles what is the case; and he asked for the experiment that would break the tie rather than treating the tie as a resting place. His Sahara line is a compact statement of a real point in philosophy of science — the underdetermination of theory by evidence — and his framing of the test as a modus tollens rather than a confirmation is Popper’s point, correctly applied. Add Russell as the hostile witness: an empirical science ought not to carry an assumption that observation can neither prove nor disprove. That is a serious complaint, seriously made, by people with no interest in geocentrism. Concede all of it.

Why it doesn’t save the claim.

Because Russell’s complaint was answered, and the answer was to delete absolute space, not to put the Earth in the middle of it. The metaphysical assumption Russell objected to in Newtonian dynamics is the one relativity removed: after 1905 there is no quantity “the Earth’s absolute velocity” that is hidden from us, because there is no such quantity to have a value. That is why the principle cannot be turned into a positive result for anybody. “The Earth is at absolute rest” is not an unproven-but-live possibility on this theory; it is a sentence with nothing to be true of, and it is exactly as available to Jupiter.

And the principle is scoped, in the wording of the man who named it. Poincaré’s 1904 St Louis lecture states it as the laws being the same for a stationary observer and for one “carried along in a uniform motion of translation, so that we have no means, and can have none, of determining whether or not we are being carried along in such a motion.” Such a motion. Rotation is not a uniform translation, acceleration is not a uniform translation, and neither was ever inside the shield. The one claim geocentrism has to make about the Earth — that it does not turn — falls in the part of kinematics the principle leaves exposed.

Finally, van der Kamp’s own standard convicts the list rather than the critics. He said a null result does not license a universal negative. The list’s items 29, 88 and 317 are universal negatives built from null results.

First, the concession, and it is total. There is no experiment that reveals motion relative to absolute space, and the attempts to find one are among the most refined measurements physics has. Michelson and Morley set the pattern in 1887; Kennedy and Thorndike added the velocity-dependent version in 1932; the modern descendants use rotating cryogenic optical cavities and bound a direction-dependence of the speed of light at the 10⁻¹⁷ level (Herrmann et al. 2009) and a Lorentz-violating anisotropy at 10⁻¹⁸ (Nagel et al. 2015). Nothing has turned up. This is not an embarrassment being managed — it is among the most stringently confirmed symmetries in physics, and it is the reason the theory has the shape it has. Anyone answering this argument by claiming that absolute motion has been measured is wrong, and will deserve to lose the exchange.

Second, the word the principle turns on. Poincaré, naming it in 1904: the laws are the same for a stationary observer as for one “carried along in a uniform motion of translation, so that we have no means, and can have none, of determining whether or not we are being carried along in such a motion.” Einstein’s 1905 postulate is scoped the same way, to frames in which the equations of mechanics hold — inertial frames. A rotating frame is not one. This is not a loophole discovered by critics; it is the load-bearing distinction of the whole subject, the one Newton built the bucket to illustrate, and it is why the theory that abolished absolute velocity left absolute rotation standing as a local, coordinate-free fact about a body.

So the quantity to look at is the Earth’s rotation relative to the local compass of inertia — what gyroscopes point at — and it is measurable in a closed room with the blinds down. Foucault did it in 1851 with a wire. Michelson, Gale and Pearson did it optically in 1925, predicting a 0.236 fringe shift and measuring 0.230 ± 0.005 (that experiment is ARG-A02, where the list cites it backwards). The G ring laser at Wettzell now resolves the Earth’s rate to better than one part in 10⁹, tracking the Chandler wobble and solid-Earth tides as they move it. Every one of these instruments answers the question without reference to the sky. The relativity principle has nothing to say about them, by its own terms.

Be fair about the one reply that engages. A defender can say that the compass of inertia is itself set by the matter of the universe, so a cosmos turning once a day would drag the gyroscopes with it and the ring laser would read 15.04°/h either way. That is Mach’s principle, it is a real research question, Thirring proved a version of it for a rotating shell in 1918, and it converts R03 from a claim about what experiments can decide into a claim about dynamics — a stress-energy tensor, a solution, a fit. The dynamical version is argued in full at ARG-R01 and is not re-run here. The point that belongs to R03 is narrower and harder: the moment the defence becomes dynamical it is outside the relativity principle’s protection. A theory that says the sky physically drags the local inertial frame is a theory with consequences, and consequences are the thing R03 was invoked to say we could never have.

Third, the revolution, which is the one item here that makes a checkable claim. Item 88 says there is no proof of revolution. The Earth’s orbital motion is not merely detected; it is used as a working instrument. Every precision Doppler spectrum taken anywhere in the world has the observer’s velocity about the Solar-System barycentre subtracted before the data are used, and the standard for that correction is accurate to 1 cm/s (Wright & Eastman, PASP 126:838, 2014). If the Earth held still, that subtraction would be a fiction, and the fiction would have to be the reason exoplanet signals stay coherent across years of observation. In the microwave background the same 30 km/s shows up as the orbital dipole, and Planck used it as the absolute photometric calibrator for its channels from 30 to 353 GHz. The Earth’s revolution is a metrological standard. The classical versions of this point — aberration and stellar parallax — are at ARG-A03 and ARG-A05.

State plainly what these are and are not. They are measurements of motion relative to matter: relative to the Solar-System barycentre, relative to the frame in which the microwave background looks isotropic. None of them is a detection of absolute motion, and nothing here claims otherwise. That is the whole point. Relative motion is the observable kind — the list says so itself, in item 276 — and the Earth’s relative motion is observed, tabulated and used.

Fourth, the shield excludes the thing it was raised to protect. The geocentric thesis is not “the Earth is at absolute rest.” It is that the Sun, planets and stars physically circle the Earth once a sidereal day. That is a claim about the relative motion of matter, which is the category the principle declares observable. What is actually observed is a relative rotation of Earth and sky at about 15.04°/h, on which both models agree and which therefore discriminates between nothing; and a rotation of the Earth relative to local gyroscopes, on which they disagree, and which has been measured to nine digits. A principle whose content is “only relative motion is observable” cannot be used to protect a claim about relative motion from observation.

Fifth, the symmetry, which is the cluster’s basis and worth stating without decoration. If no experiment can establish that the Earth moves absolutely, none can establish that it is absolutely at rest. The principle is not evidence for one of the two options; it dissolves the question both options were answers to. Item 131 (“No absolute frame”) is true and equally forbids a geocentric absolute frame. Item 317 (“No global inertial frame proven”) is true in a stronger way than the list appears to intend: in a curved spacetime inertial frames are local, there is no global one, and so there is no global “at rest” for a body to occupy. That item is a true statement whose truth erases the sentence it is being marshalled to support. R01 records the movement’s own retreat from this position — Sungenis and Bennett naming the Earth-Centred Inertial frame as “one universal absolute preferred frame in which c is isotropic”, which is a preferred-frame theory and the negation of R03 rather than an extension of it.

Sixth, the cost across the rest of the list. R03 is the deepest thing in the collection and it is also the most expensive, because it is a universal solvent. If experiments cannot discriminate between the frames, then Michelson–Gale is not evidence, “Airy’s failure” is not evidence, Miller’s drift is not evidence and the microwave alignments are not evidence — and each of those is asserted elsewhere on the same list as an experiment that came out the geocentrist’s way. Bouw, the movement’s only credentialed astronomer, took the consistent route and referred the decision to Scripture (ARG-R11). The list takes neither route: it runs R03 and the experimental items side by side and asks the reader to accept both.

Seventh, the test he asked for. Van der Kamp wanted the speed of light measured from a fast platform — “e.g. a Concorde or Space shuttle” — before he would grant relativity viability. That class of experiment has a name he did not use for it: it is the Kennedy–Thorndike experiment, which tests whether the speed of light depends on the velocity of the apparatus, and its modern versions get the varying platform velocity for free from the Earth’s own rotation and orbit rather than from an aircraft. Those versions are the same cryogenic-cavity experiments quoted above. Clocks on fast platforms had already been flown in 1971, when Hafele and Keating carried caesium standards east and west around the world and found the gains and losses in agreement with the relativistic predictions to within the roughly 10% precision of the experiment — seventeen years before De Labore Solis went to press. His demand was reasonable when he framed it and had in substance been met by the time he published it.

Verdict: standard physics. The proposition is correct, it is taught, and it belongs to relativity rather than to geocentrism. It says that one particular quantity — velocity through a substrate — is not a physical quantity at all. It carries no implication that the Earth’s motion is unknowable, because rotation and acceleration were never covered by it; it does not favour the Earth, because it favours nothing; and it cannot be spent, because spending it retires most of the list that carries it.

The list says

Kinematical equivalence.

The source says

“Bertrand Russel's contention that the observable phenomena will be the same whether the Earth rotates or the Heavens revolve, as well as Fred Hoyle's declaration that the geocentric view is as good as anybody else's, but not better, they are only tenable if certain presuppositions are assumed to be self-evident. Which they are not!”

The qualifier was dropped. Three qualifications travel with this proposition in De Labore Solis and none of them survives into the eight list items. Whose claim it is: in the book the equivalence belongs to Russell and Hoyle, who are named, and van der Kamp’s sentence about them is a refusal — their claims hold only on presuppositions he rejects in the next three words. On the list the same proposition appears in his own name. The antecedent: his relativistic argument runs under “If Einstein is right …” (p. 67), and he held general relativity untenable in its present form; the list asserts the consequent with the antecedent removed. The demand: the book opens by asking for a control experiment to settle the question (pp. 6–7) and warns that a null result licenses no universal negative — “if in the Sahara no icefields can be found, this observation does not thereby prove that icefields exist nowhere” (p. 6). Items 29, 88 and 317 are universal negatives built from null results, which is the move his own line was written to block. The refutation above answers the source, not the fragment: it concedes the relativity principle outright, at the strength the modern experiments give it, and puts the weight on the scope Poincaré wrote into the sentence and on what a dynamical rotating-cosmos defence costs once it leaves that scope. This is a rare case in which compression makes the argument look stronger than its author left it: he published a request for a trial, and the list publishes a verdict. Compare ARG-R01, where the same book is compressed the other way — a concession promoted to a proof with its wording intact.

The Failure of General Relativity 334 — The Demise of Relativity Theory 441 — Chapter 10: Mathematical Models of a Geocentric Universe 590 — Absolute Rest versus Relative Motion 591 — The Gyroscopic Effect on Earth 599 — Einstein's Geocentrism 607 — Thirring's Geocentrism 611

— Galileo Was Wrong: The Church Was Right (2006), Vol. I (2006), contents pages vi–xi of the Internet Archive scan (item GallileoWasWrong), whose title page reads “Volume I: The Scientific Evidence” and whose ISBN is 0-9779640-0-0. Chapter 10 opens at printed p. 590; the section text at p. 607 was not read here

What is quoted above is the book’s own table of contents, and it is quoted because that is the part of this work we were able to read. The Internet Archive PDF renders far enough for a fetcher to return the front matter and the full contents listing and no further; the chapters themselves stop being retrievable long before p. 607. So the volume, chapter and page of this argument are recorded here and its wording is not. Anyone with the printed book should read ch. 10 pp. 590–637 and correct this entry.

Even so, the contents page settles three things a reader should have.

Where the argument lives. Chapter 10, “Mathematical Models of a Geocentric Universe” (p. 590), runs: Absolute Rest versus Relative Motion 591, The Gyroscopic Effect on Earth 599, Einstein’s Geocentrism 607, Thirring’s Geocentrism 611, then Rosser 616, Bondi 619, Brill and Cohen 623, Møller 624, Lynden-Bell 626, Barbour and Bertotti 628. That is not a list of cranks. Brill and Cohen’s “Rotating Masses and Their Effect on Inertial Frames” (Phys. Rev. 143:1011, 1966) is a real result about a real effect, and the steelman below is built out of it.

Which volume this is, in which year. In the 2006 printing these are chapters of Volume I. In the seventh edition of 2013 the work was split into three volumes, and that edition’s own chapters 7–13 make up Volume II — the identity our ARG-E03 and ARG-R01 entries had to establish the hard way. The chapters were renumbered as well as rehoused. The chapter-10 section run quoted above, from Absolute Rest versus Relative Motion onward, reappears in the 2013 Volume II as chapter 9, “Modern Science and the Acceptance of Geocentrism”, opening at p. 111, with Einstein’s Geocentrism at p. 126 and Thirring’s Geocentrism at p. 130. Chapter 10 of the 2013 Volume II is a different chapter — “Technical and Summary Analysis of Geocentrism”, p. 157, the one ARG-E13 cites — and it descends from chapter 12 of the 2006 Volume I, “Technical and Summary Analysis of Geocentric Cosmology”, p. 710. So neither chapter numbers nor page numbers carry between the two editions, and page citations in the critical literature to “vol. 1, 9th ed.” will not line up with this scan’s pagination. Ours is the 2006 first printing throughout.

What the book announces about relativity. The same volume that models a geocentric universe on Thirring and Brill–Cohen also runs sections titled The Failure of General Relativity (ch. 5, p. 334) and The Demise of Relativity Theory (ch. 7, p. 441). That is not a contradiction on the authors’ part, and it should not be reported as one: the appeal to relativity is dialectical, and Sungenis says so himself in documents that are readable. Answering Phil Plait’s Bad Astronomy post of 14 September 2010 he writes that “a geocentrist appeals to Relativity not as proof or even evidence for geocentrism” but to show that geocentrism “can have just as much scientific respectability as heliocentrism”, and adds that “the geocentrist makes his case for geocentrism on solid scientific evidence that has little or nothing to do with Relativity.” In an undated interview he puts the positive claim at its own strength: general relativity “not only supports geocentrism as a ‘relative’ alternative”. A relative alternative, offered as respectability rather than evidence, is what the list has turned into “Equivalence principle validation.”

SURFACE (weak — do not use). “They do not understand the equivalence principle.” There is no evidence for that here and the move loses the exchange. The chapter this cluster comes from is organised around Thirring, Rosser, Bondi, Brill and Cohen, Møller, Lynden-Bell, and Barbour and Bertotti — the actual general-relativity literature on rotating frames and inertial dragging. Whatever is wrong with the conclusion, the reading behind it is not imaginary.

DEEPER. “Local is not global.” True, and the whole answer in one line, but incomplete as stated, because it invites the obvious reply: where does local end? If the boundary is a hand-wave, the objection is a hand-wave. It is not a hand-wave, and the third section of the refutation gives the boundary in centimetres.

KERNEL. The specific true thing is that Einstein himself did not stop at the falling lift. In the 1916 general-relativity paper he extended the equivalence heuristic to rotating systems, treating the centrifugal field an observer finds in a rotating frame as a gravitational field he may ascribe to the rest of the matter in the universe — and that Machian instinct was later given equations. Thirring in 1918, and Brill and Cohen in 1966, showed that a rotating mass shell really does drag the inertial frames in its interior around with it, the dragging becoming complete in the limit where the shell approaches its own gravitational radius. So the geocentric position here is not “physics forbids it and we do not care”. It is: the local inertial frame is not laid down in advance, it is determined by the matter around it; there is a published mechanism by which distant matter fixes it; therefore an Earth taken as non-rotating with the cosmos turning about it is a physically discussable arrangement, not a category error. Concede every word of that.

Why it doesn’t save the claim.

Because the kernel’s own content is that the local inertial frame is a physical thing you can go and measure — and the moment it is physical rather than conventional, “does the Earth turn with respect to it?” stops being a matter of viewpoint and becomes a reading on an instrument. The equivalence principle is what makes the question well posed. It supplies the standard; the standard has been read; it reads 15.04° per hour.

And the two halves of the kernel are not the same claim. The equivalence principle fixes a local frame at each event and says nothing whatever about how the matter of the universe is arranged or moving. Frame dragging is a statement about how the matter is arranged and moving — a solution of the field equations with a stress-energy tensor, boundary conditions and a fit to observation. The geocentric conclusion needs the second and cites the first. Nothing in the equivalence principle entails that the cosmos is a rotating shell, and no amount of local validity ever adds up to a cosmological one: that is what “local” means.

Finally, the licence is not exclusive. The principle grants a local rest frame to a free-falling observer in Neptune’s atmosphere, to a probe in the Kuiper belt, and to a grain of dust in the Coma cluster, on exactly the same terms. A principle that hands a rest frame to everybody distinguishes nobody — the same symmetry that sinks the covariance argument at ARG-R01.

First, the concession, in full. The equivalence principle is true and is among the best-tested statements in physics. The MICROSCOPE satellite’s final result puts the Eötvös ratio for titanium against platinum at η = [−1.5 ± 2.3 (stat) ± 1.5 (syst)] × 10⁻¹⁵ — no violation at a part in 10¹⁵. Will’s standard statement of the Einstein equivalence principle has three parts: the weak principle holds; “the outcome of any local non-gravitational experiment is independent of the velocity of the freely-falling reference frame in which it is performed”; and “the outcome of any local non-gravitational experiment is independent of where and when in the universe it is performed.” Both operative clauses carry the same qualifier, and it is not decoration. A local rest frame is validated. The verdict on this cluster is standard physics for that reason and no other.

Second, notice who else is covered. Every freely falling observer at every event in every spacetime gets the same licence. If the equivalence principle validated a stationary Earth it would equally validate a stationary Enceladus, a stationary asteroid and a stationary hydrogen atom in intergalactic space, each entitled to declare the rest of the universe in motion about it. The principle is universal, and a universal permission confers no distinction on any particular recipient. This is the same structure as the Kretschmann objection at ARG-R01: an argument that the formalism is neutral cannot then be cited as evidence that the formalism favours you.

Third, what “local” costs, in centimetres. The falling-lift picture is exact only in a uniform field, and the Earth’s field is not uniform: it converges on a centre and falls off with distance. The residue is tidal, and it is the thing that tells you the region is not small enough. Two numbers fix the scale at the Earth’s surface. Vertically, gravity weakens with height at 2g/R ≈ 3.08 × 10⁻⁶ s⁻² — the free-air gradient geodesists carry as 0.3086 mGal per metre. Horizontally, two plumb lines both point at the Earth’s centre and so converge at g/R ≈ 1.54 × 10⁻⁶ m s⁻² for every metre of separation. Those two numbers are components of the Riemann tensor. They are the curvature, they are what geodesic deviation measures, and no change of coordinates removes them: the metric can be brought to Minkowski form and its first derivatives to zero at a point, never its second derivatives.

So the boundary of “local” is not a philosopher’s vagueness. It is set by the sensitivity of whatever you carry into the lift, and it moves as instruments improve. A chip-scale MEMS gravimeter demonstrated at 40 µGal Hz^−½ (4 × 10⁻⁷ m s⁻²) resolves the vertical gradient across a height difference of 4×10⁻⁷ ÷ 3.086×10⁻⁶ ≈ 0.13 m, and the same device recorded the solid-Earth tide — the Moon and Sun stretching the ground under a laboratory bench — which is a curvature effect being read out on a silicon chip. Thirteen centimetres. That is the size of the region over which the equivalence principle licenses you to say there is no gravity here, given a gravimeter you can hold in one hand. The claim on the list asks that the same licence be carried out to the edge of the observable universe, about 4.4 × 10²⁶ m. Between the warrant and the conclusion lie some twenty-seven orders of magnitude, and the warrant expires at the first of them.

Fourth — and this is where the argument turns over — the frame the principle supplies is not rotating. A local inertial frame is not only one in free fall; its axes are the ones a gyroscope keeps, Fermi–Walker transported along the observer’s worldline. Rotation with respect to that frame is a local, coordinate-free, physical observable, and it is not something the equivalence principle lets you talk your way out of. There is a clean technical reason why not. The centrifugal term in a rotating frame is a static field and can be traded against gravity in the equivalence-principle way; the Coriolis term cannot, because it depends on the test body’s velocity, and no static gravitational field imitates a velocity-dependent force. What can imitate it is a gravitomagnetic field — off-diagonal time–space components of the metric — and producing one requires a specified distribution of moving matter. That is a dynamical claim, owed a solution, and it is not a consequence of the equivalence principle.

Meanwhile the observable has been read, repeatedly and by unrelated methods. Foucault’s pendulum precesses at Ω sin φ, about 11.3° per hour at the latitude of Paris, in a closed room with no view of the sky. Ring laser gyroscopes at Wettzell resolve the Earth’s angular rate to better than one part in 10⁹. Gravity Probe B flew four gyroscopes precisely to test what a local inertial frame does, and returned a geodetic drift of −6,601.8 ± 18.3 mas/yr against a predicted −6,606.1, and a frame-dragging drift of −37.2 ± 7.2 against −39.2 — a confirmation of exactly the Machian-flavoured effect the steelman rests on, at roughly ten orders of magnitude too small to be a diurnal rotation. The equivalence principle does more than fail to deliver a stationary Earth. It defines the compass against which the Earth’s rotation is measured, and the compass says 15.04° per hour.

Fifth, the compass wobbles with the weather. This is the part a defender should be shown early rather than late. The Earth’s rotation rate is not constant: length-of-day varies at the millisecond level, and “shifts in zonal wind patterns and atmospheric circulation are responsible for around 90% of seasonal length of day variations”, with El Niño events associated with longer days and La Niña with shorter ones. On the reading where the Earth is truly at rest and the cosmos physically turns about it, that signal has to be transferred to the cosmos: the rotation of every distant galaxy must speed up and slow down in step with the trade winds over the Pacific, and must do so instantaneously enough to keep the sky rigid. Angular momentum exchanged between the atmosphere and the solid Earth is a local bookkeeping entry; on the geocentric reading it becomes a cosmological one.

Sixth, three different principles are being run together as one. The five items in this cluster are not five statements of a single claim. Item 316, “Equivalence local labs rest”, is the equivalence principle proper, and it keeps the qualifier that makes it true. Items 57 and 132 are the same sentence with “local” taken out, at which point they assert something that lies outside the principle’s scope. Items 28 and 130, “Einstein’s equivalence of motion” and “Einstein admission of equivalence”, are not about the equivalence principle at all: they are the general relativity of motion and Einstein’s conventionalist remark about the sun and the earth, answered at ARG-R01 and ARG-R08. Three principles, three different fates. The equivalence principle is local and true. General covariance is global and, since Kretschmann in 1917, empty of physical content. Machian dragging is dynamical, real, measured — and none of the geocentric works read for this review supplies the cosmological solution that would be needed to scale it to a daily rotation, which is where ARG-R05 and ARG-A09 take it up. Presenting the three together makes one licence look like three convergent lines of support.

Seventh, the source is more careful than the list, and the difference is the finding. The volume this cluster is credited to also contains sections titled The Failure of General Relativity and The Demise of Relativity Theory, and Sungenis has written that a geocentrist appeals to relativity “not as proof or even evidence for geocentrism”, and that the case rests on evidence “that has little or nothing to do with Relativity”. Take him at his word and this cluster is not evidence for anything, by its own originator’s account; it is a claim of respectability. Decline to take him at his word and the appeal has to be run at full strength, in which case it collapses for the reasons above. Either way the list’s flat “Equivalence principle validation” overstates the authority it is drawing on.

Verdict: standard physics. The true content of the argument — that a local frame in which the Earth is at rest is physically legitimate — is uncontroversial, taught everywhere, and denied by nobody. It is true of every body in the universe, it holds over a region currently bounded at the decimetre scale by instruments you can buy, and the frame it certifies is the non-rotating one that Foucault, Wettzell and Gravity Probe B use to measure the Earth turning.

The list says

Equivalence principle validation.

The source says

General Relativity “not only supports geocentrism as a ‘relative’ alternative, but also modified the postulates of STR” — and “a geocentrist appeals to Relativity not as proof or even evidence for geocentrism but merely to show … that … geocentrism can have just as much scientific respectability as heliocentrism.”

A concession was re-used as a proof.

The same trade as ARG-R01, one lane over. Sungenis offers relativity as a relative alternative conferring respectability, and states in as many words that it is not offered as proof or even as evidence. The list prints it as item 132, “Equivalence principle validation”, in a numbered series of evidence for a stationary Earth. Nothing has been misquoted; a permission has been promoted to a validation.

What was and was not compared. The book’s own statement of the argument, at Vol. I ch. 10 p. 607 in the 2006 printing, was not read: the Internet Archive rendering of that volume stops returning text long before it. The comparison above is therefore made against two things we could read — the volume’s contents page, which announces The Failure of General Relativity at p. 334 and The Demise of Relativity Theory at p. 441, and the author’s own published characterisations of what the appeal to relativity is for. That is enough to establish the speech act and not enough to establish the sentence. Anyone reaching p. 607 should re-run this block.

A second drift, visible inside the list itself. Item 316 reads “Equivalence local labs rest” and keeps the qualifier the principle cannot do without; items 57 and 132 drop it. One compressed line retains the scope and two delete it, which is the whole disagreement in five words. Only one drift type can be recorded, and force_upgraded is the one carrying evidence from the source’s side; the scope-widening is recorded here because a reader can check it against the corpus without leaving the page.

Considering that Lynden-Bell's paper includes ten pages of the most rigorous mathematical analyses to date of Mach's principle (i.e., that the universe in rotation around a fixed Earth equates to an Earth in rotation within a fixed universe), geocentrism has been established by the very physics that sought to dethrone it in 1905.

— Galileo Was Wrong: The Church Was Right (2006), Chapter 10, “Mathematical Models of a Geocentric Universe”, at the page footer numbered 628 in the Internet Archive OCR text of item GallileoWasWrong (title page: Volume I, The Scientific Evidence, ISBN 0-9779640-0-0). The section runs from p. 625 to p. 629 under the running head “Chapter 10 Galileo Was Wrong”, and the item's table of contents places its three headings at 625, 626 and 628. Pages verified against the OCR text only; not checked against a print copy or page images.

The sentence before it is the one that matters. A page earlier, summarising the same paper, the book reports that “The Lynden-Bell team stresses several times their general proof that the angular momentum of any closed universe is zero,” and continues: “Interestingly enough, the null value for the angular momentum will provide the fixed and undisturbed cradle for the barycenter, the Earth, and thus Mach’s principle has inadvertently vindicated geocentrism once again” (p. 627). The theorem is quoted correctly. It is the result which most directly rules out what the chapter wants: a universe whose distant matter circulates rigidly once a day carries an immense total angular momentum, and that is the quantity the theorem sets to zero. The reading offered here is that the rotating sky “generates no angular momentum to twist or rotate the Earth” — a statement about torque on one body, which is not what the proof is about.

The condition is quoted and then dropped. The book states the antecedent itself: Lynden-Bell, Katz and Bičák “completely exonerate Mach’s principle, at least, as they say, if the universe is closed” (p. 627). Their paper is titled Mach’s principle from the relativistic constraint equations and its result is that Mach’s principle follows from the constraints provided the universe is closed. The sentence quoted above, which is the one the list inherits, carries no trace of that condition — and closure is an open question the data do not currently favour: Planck’s 2018 parameters give a curvature density Ω_K = 0.0007 ± 0.0019 with lensing and BAO, consistent with flat.

Whose word “geocentrism” is. Three consecutive section headings on these pages attach it to physicists’ names — “Nightingale’s Geocentrism”, “Lynden-Bell’s Geocentrism”, “Barbour and Bertotti’s Geocentrism”. The papers under those headings are about Mach’s principle and inertial frames. In the Barbour and Bertotti extract the geocentric reading is supplied inside the quotation marks by square brackets, which by convention are the quoting author’s: “a rigid, uniform shell of mass M₀ and radius R₀ [e.g., the universe]. The test body [e.g., the Earth] is near the center of the shell” (p. 629). The shell and the test body are the authors’; the universe and the Earth are the reader’s guide.

The chain the quotations travelled. Every Barbour and Bertotti extract in this chapter is footnoted to “The Geocentric Papers,” Association for Biblical Astronomy, Cleveland, Ohio — at its pp. 88, 89 and 98 — a movement compilation published by the organisation Gerardus Bouw ran, rather than the journal. Consistent with that route, the journal volume is given as “32B” three times in this text (a footnote at p. 464, the footnote here at p. 628, and the bibliography at p. 1075) and as “38:1” once, in an appendix at p. 1027; the paper is Nuovo Cimento B 38(1):1–27, March 1977. OCR confusion of 8 for 2 could not be ruled out from the scan text alone, but it would have to have gone one way three times and the other way once.

SURFACE (weak — do not use). “Mach’s principle is philosophy, not physics.” This loses immediately. Machian frame dragging is a measured effect: Gravity Probe B flew four gyroscopes to look for it and returned a frame-dragging drift of −37.2 ± 7.2 mas/yr against a predicted −39.2, alongside a geodetic drift of −6,601.8 ± 18.3 against −6,606.1. Anyone who dismisses the whole subject has conceded that they have not read it.

DEEPER. The kinematic half of the claim is Mach’s own, in print. In The Science of Mechanics he writes that “the motions of the universe are the same whether we adopt the Ptolemaic or the Copernican mode of view. Both views are, indeed, equally correct” (p. 232), and in the Appendix he puts the bucket question himself and answers it in the same breath: “Can we fix Newton’s bucket of water, rotate the fixed stars, and then prove the absence of centrifugal forces? The experiment is impossible, the idea is meaningless, for the two cases are not, in sense-perception, distinguishable from each other.” He then states the conclusion flatly: “I accordingly regard these two cases as the same case and Newton’s distinction as an illusion” (p. 543). Quote him as printed and the defence gets stronger, not weaker — it is not a dare left hanging but a verdict. What should not be quoted is the loose paraphrase this entry carried until now, in which Mach challenges the reader to “try to fix Newton’s bucket and rotate the heaven of fixed stars”: that wording is not located anywhere in the McCormack second English edition (Open Court, 1902) or in the text of Galileo Was Wrong, and where it came from was not traced. On a page whose subject is quotations losing their qualifiers in transit, printing a variant would be the worst available error.

KERNEL. The specific true thing is that somebody has done the sums, more than once, in peer-reviewed print. Luka Popov, Eur. J. Phys. 34:383 (2013), works the Sun–Earth–Mars system twice — once Copernican, once neo-Tychonic — and shows that with a Machian pseudo-potential standing in for the accelerated distant matter, the trajectories come out the same. André Assis goes further: Relational Mechanics and Implementation of Mach’s Principle with Weber’s Gravitational Force derives the Earth’s equatorial bulge and the precession of Foucault’s pendulum from the rotation of the distant galaxies about a stationary Earth, and states the conclusion plainly: “We have found a complete equivalence between ptolemaic and copernican world systems … not only kinematically or visually, but also dynamically.” So the strongest form of R02 is not a plea for tolerance. It is: there exists a published, quantitative, relational mechanics in which an Earth at rest with the cosmos turning about it reproduces the equatorial bulge, Foucault’s precession and the Coriolis deflections. Concede every word of that; it is true.

Why it doesn’t save the claim.

Because in that same book the equivalence is handed to everything. Two paragraphs after the Ptolemaic–Copernican section, Assis writes: “As a matter of fact, any other frame of reference would be equally valid. Anyone or any arbitrary frame of reference can be considered really at rest, while the entire universe moves relative to this person according to his will.” His worked example is a rock in free fall, which may be taken as permanently at rest while the Earth and all the galaxies accelerate towards it. A theory on which the falling rock’s frame is as good as the Earth’s has not made the Earth the centre of anything. It has abolished centres. The list needs a result about the Earth and relational mechanics supplies a result about frames.

And the other great relational programme, the one Galileo Was Wrong leans on by name, forbids the thing outright. Barbour and Bertotti built their dynamics to be invariant under time-dependent rotations, and the price of that invariance is that only the zero-total-angular-momentum solutions survive; in the relativistic descendant, as the Stanford Encyclopedia puts it, “one cannot have a solution consisting of a single rotating body: the overall angular momentum of the universe must vanish.” The book’s own summary of the 1977 paper records the authors proceeding “by assuming a non-rotating universe”. So the two theories that have actually been written down disagree about the one question the cluster turns on, and the citation carries the word “relational” without carrying either theory’s answer.

First, what is conceded, and it is a lot. Machian effects are real and measured. A rotating mass does drag the local inertial frames inside and around it: Thirring wrote the shell calculation in 1918, Brill and Cohen showed in 1966 that the dragging becomes complete as a rotating shell approaches its own gravitational radius, and Gravity Probe B measured the effect in orbit — a frame-dragging drift of −37.2 ± 7.2 mas/yr against a predicted −39.2. The kinematic equivalence Mach asserted is real. And the relational programme is serious physics done by serious people, published in Nuovo Cimento, in MNRAS, in the American Journal of Physics and in the European Journal of Physics. Nothing below depends on treating any of that as fringe.

Second, “Mach’s principle” is not one proposition, and the ambiguity is doing the work. Bondi and Samuel set out eleven inequivalent statements that go under the name and score general relativity against each: Mach0, that the distant galaxies show no rotation relative to local inertial frames — an observational statement, and true; Mach3, that local inertial frames are affected by the cosmic distribution of matter — which general relativity satisfies; Mach2, that an isolated body in empty space has no inertia, and Mach7, that removing all matter removes space — which it does not satisfy. Einstein named the principle in 1918 as a goal of his theory and had recognised by that year that the theory did not deliver all of it. So a sentence of the form “Mach’s principle allows a fixed Earth” has not said anything until it says which Mach, and the versions that general relativity actually satisfies are the ones about how inertial frames are influenced, not the ones that would let a planet be exempted from turning.

Third, and this is the centre of it: the theorem quoted in support is the theorem that excludes the model. The chapter’s strongest citation is Lynden-Bell, Katz and Bičák, and what it takes from them is their “general proof that the angular momentum of any closed universe is zero”, read as providing “the fixed and undisturbed cradle for the barycenter, the Earth”. Read the theorem as written. It says that in a closed universe the total angular momentum of everything vanishes. A cosmos in which the Sun, the planets, the stars and the galaxies circulate rigidly about the Earth at 15.041° per hour has an enormous total angular momentum and there is nothing left over to cancel it. The geocentric arrangement is precisely the configuration the theorem forbids.

The relational reply to that has to be met, because it is the best move available: in a theory where only relative motion is real, who is entitled to compute “the angular momentum of the universe” in some frame outside it and wave the answer at us? The reply is that the vanishing is not a quantity imported from outside. It is a constraint on the matter itself, which is why Lynden-Bell, Katz and Bičák present their result as a vindication of Mach rather than of Newton. And Barbour and Bertotti reach the same restriction from the opposite end, with no background whatever: demand that a rigid, time-dependent rotation of the entire configuration be unobservable, and Dirac’s analysis of the resulting constraints forces the total angular momentum to zero. Bondi and Samuel, cataloguing the versions, set this down as Mach5 — “the total energy, angular and linear momentum of the universe are zero” — false in Newtonian mechanics and in asymptotically flat spacetimes, while in relativistic cosmology, they write, “it is claimed that the total angular momentum of a closed universe must vanish”; and of the relational models themselves they observe that Newtonian theory admits solutions with nonzero angular momentum, such as a solar system in an otherwise empty universe, “while relational models do not permit such solutions”. Two independent routes to the same exclusion, both of them relational, and neither needing an absolute space to state it.

What the paper offers is the opposite service: with the angular momentum distribution as observed, the local inertial frames do not rotate relative to the distant matter — which is Mach0, the empirical statement, and it is the statement that the sky and the gyroscopes agree, leaving the Earth to turn under both.

The same reversal runs through the Barbour and Bertotti material. Their dynamics is built to be invariant under time-dependent rotations of the whole configuration, and what survives that construction is only the sector with zero total angular momentum; its relativistic descendant cannot accommodate a single rotating body at all, because the universe’s overall angular momentum must vanish. The book’s own paraphrase of the 1977 paper contains the words “by assuming a non-rotating universe”. Three of the four technical authorities marshalled here are Machian relationalists, and Machian relationalism is the framework in which a rotating cosmos is hardest to write down.

Fourth, the one theory that does deliver the equivalence delivers it to everybody. Assis’s relational mechanics really does reproduce the equatorial bulge, Foucault’s precession and the Coriolis deflections from a rotation of the distant galaxies about a stationary Earth; the derivation is in chapter 18 and the conclusion is stated without hedging. It is also stated, on the same two pages, that “any other frame of reference would be equally valid” and that “anyone or any arbitrary frame of reference can be considered really at rest, while the entire universe moves relative to this person according to his will” — illustrated with a rock in free fall which may be held permanently at rest while the Earth and the galaxies accelerate upwards towards it. That is the whole content of the result: frames are cheap. It cannot be spent on the Earth without being spent on the rock, and a doctrine that makes the falling rock the centre of the universe has stopped being geocentrism.

Assis is also candid about the bill. His mechanics runs on Weber’s force law and needs the relation H₀²/ρ ≈ G, of which he writes: “We cannot say that this relation is exactly valid, due to uncertainties in the observational values”. On the deflection of starlight by the Sun he writes “we need further research in this direction before drawing final conclusions”, and on the gravitational redshift he reports the calculations with Weber’s law as still to be published as of that edition. And the cosmology comes attached: a universe infinite in space and time, not expanding, with Hubble’s law produced by light losing energy to the intergalactic medium. Adopting relational mechanics to license a stationary Earth means adopting all of that, and the rest of this list does not.

Fifth, the neo-Tychonic calculation and the assumption inside it. Popov’s paper is the best single technical item in this cluster, and it says what it is: the analysis is carried out “in the framework of Newtonian mechanics”, “the kinematical equivalence … is shown to be a consequence of the presence of pseudo-potential”, and the model is defined by “the assumption that orbits of distant masses around the Earth are synchronized with the Sun’s orbit”. That last clause is the entire annual mechanism, and it is granted rather than derived. Every distant galaxy must complete a circuit about the Earth once a year, phase-locked to where the Sun happens to be, and a second circuit once a sidereal day, with the two motions superposed and no dynamical account of what enforces either. The paper is honest about this. What it establishes is that given a pseudo-potential of the required form, the geostatic trajectories come out right — a statement about the consistency of a description, which is what the pseudo-potential was built to be.

Sixth, the arithmetic of a turning cosmos. Two numbers bound the ambition. At an angular rate of one sidereal turn per day, a body co-moving with the sky reaches the speed of light at a radius of c/Ω = 4.11 × 10¹² m, about 27.5 astronomical units — inside the orbit of Neptune. Beyond that radius, being “at rest with respect to the Earth” is not a state matter can occupy, because the Earth-fixed rotating chart stops being a reference frame there; the co-ordinates survive, the frame does not. And the observational bound on a global circulation runs the other way by a wide margin: the required rate compared to the expansion rate is Ω/H₀ ≈ 3 × 10¹³, while the Planck analysis of Saadeh and colleagues bounds the vector mode, the one associated with vorticity, at (σ_V/H)₀ < 4.7 × 10⁻¹¹ at 95% confidence, and disfavours anisotropic expansion at odds of 121,000:1. State the caveat ourselves rather than wait for it: that bound is derived within a perturbed relativistic cosmology, so a defender who rejects the framework will reject the number. The light-cylinder radius does not depend on the framework, and neither does the third section above, which is why the argument does not rest here.

Seventh, what Mach actually wrote, which cuts both ways and is worth quoting at length because it is out of copyright and the geocentric literature quotes only its first half. In the McCormack translation, pp. 231–232: “Relatively, not considering the unknown and neglected medium of space, the motions of the universe are the same whether we adopt the Ptolemaic or the Copernican mode of view. Both views are, indeed, equally correct; only the latter is more simple and more practical. The universe is not twice given, with an earth at rest and an earth in motion; but only once, with its relative motions, alone determinable. It is, accordingly, not permitted us to say how things would be if the earth did not rotate. We may interpret the one case that is given us, in different ways. If, however, we so interpret it that we come into conflict with experience, our interpretation is simply wrong.” Two sentences there do work against the use being made of him. The first denies that we are entitled to the counterfactual — “how things would be if the earth did not rotate” is exactly the sentence a geocentric proof needs and exactly the sentence Mach rules out of order. The second sets a constraint on interpretations that the pseudo-potential models have to meet like anything else.

And the earlier statement of the same idea, from 1872, is more explicit still about what the equivalence costs: “if we think of the Earth at rest and the other celestial bodies revolving round it, there is no flattening of the earth, no Foucault’s experiment, and so on — at least according to our usual conception of the law of inertia. Now, one can solve the difficulty in two ways: Either all motion is absolute, or our law of inertia is wrongly expressed … The law of inertia must be so conceived that exactly the same thing results from the second supposition as from the first.” That is a research programme announced, not a result reported. Mach is saying that a geostatic account requires a rewritten law of inertia, and proposing to write one. A hundred and fifty years later the rewriting is still contested, still incomplete in the versions that exist, and — in the version with the most mathematical development, Barbour’s — comes out against a rotating universe.

Eighth, the dilemma, which is where this ends. Take “the universe rotates about the Earth” and ask what kind of statement it is. If it is a choice of description, then relational mechanics grants it, grants it equally to the Sun, to Mars and to a falling rock, and grants nothing about which body is at the centre — the same trade made at ARG-R01 and ARG-R03. If it is a physical claim that the matter of the cosmos carries a real circulation, then it is a claim with consequences: a total angular momentum that the closed-universe theorem sets to zero, a co-rotation radius at 27.5 AU, a vorticity the microwave background constrains, and a required synchronisation of every distant galaxy with the Sun’s annual position for which the published models supply a stipulation. Both horns are in the sources this cluster cites. The chapter takes the first horn’s permission and the second horn’s conclusion, and the four list items report the result as a finding of modern physics.

This one runs the other way, and the honest record is that the list is milder than its source. Item 27 says Mach’s principle “supports fixed Earth possibility”. Chapter 10 of Galileo Was Wrong says geocentrism “has been established” and that Mach’s principle “has inadvertently vindicated” it. The seven drift_type values have no slot for a list that understates its source, and rather than force this into the nearest box it is recorded as none with the direction stated here. The refutation above answers the book’s stronger version, which is the one that needed answering.

The compression loss on this cluster is real, and it happened one link upstream. It is between the journals and Chapter 10, not between Chapter 10 and the list, and three instances are documented in the pages read for this entry. One: Lynden-Bell, Katz and Bičák prove Mach’s principle from the relativistic constraint equations if the universe is closed; the book quotes the condition on p. 627 and the sentence that travels onward carries none of it. Two: their “general proof that the angular momentum of any closed universe is zero” is quoted accurately on p. 627 and presented as providing “the fixed and undisturbed cradle” for the Earth, when a cosmos circulating once a day is the configuration that proof excludes. Three: the Barbour and Bertotti extract on p. 629 carries “[e.g., the universe]” and “[e.g., the Earth]” supplied in brackets, under a heading naming the authors as geocentrists, and the book’s own paraphrase a few lines earlier on the same page records them proceeding “by assuming a non-rotating universe”.

Why that is a finding rather than a technicality. The project’s thesis is that a claim degrades in transit and the degradation runs towards certainty. R02 is the case where the degradation can be located at a particular link: hedged, conditional results in MNRAS and Nuovo Cimento, reached through a movement compilation (“The Geocentric Papers”) rather than the journals, arrive in Chapter 10 with their antecedents stripped and their names re-labelled; the list then copies Chapter 10 faithfully, and even softens it. Blaming the list for this cluster would be blaming the wrong link.

Proposed instead: MISLEADING

STANDARD PHYSICS means “real, already explained, does not discriminate”. The first two clauses do not fit. Mach's principle is not an explained piece of standard physics but an unfinished research programme: Bondi and Samuel enumerate eleven inequivalent statements of it and general relativity satisfies some and violates others, which is what our own cluster note says when it records that the principle “is not a settled part of GR”. A verdict and a note that contradict each other on the summary line is the defect batch 8 found five times. MISLEADING — “real data, wrong conclusion made to look supported” — fits what is on the page: real, peer-reviewed papers by Lynden-Bell, Katz and Bičák and by Barbour and Bertotti, presented under the section headings “Lynden-Bell's Geocentrism” and “Barbour and Bertotti's Geocentrism”, quoted with “[e.g., the universe]” and “[e.g., the Earth]” supplied in brackets, and concluding that “geocentrism has been established”. The sharpest instance is the zero-angular-momentum theorem, quoted accurately and then read as support for the arrangement it excludes. The case for leaving the verdict alone, stated fairly: in Assis's relational mechanics the equivalence really does hold, so there is a reading on which the claim is true physics that fails to discriminate — which is exactly what STANDARD PHYSICS is for. I think that reading rescues one branch of a four-item cluster whose weight is carried by the other branch, but it is not a silly reading, and the change would move 4 items between verdict columns.

The verdict on the scorecard is unchanged until the question is settled. Publishing the disagreement is the point: a rubric that never produces dissent is not being applied.

These results, however, do not prove either General Relativity or heliocentrism. In fact, as noted above, Thirring's original 1918 model theorized the universe as a rotating shell around a fixed-Earth as opposed to a rotating Earth in a fixed-universe.

— Galileo Was Wrong: The Church Was Right (2006), Vol. I, ch. 10 “Mathematical Models of a Geocentric Universe”, section “Thirring's Geocentrism” (p. 611), at printed p. 613 of the archive.org OCR text of the CD-ROM issue (item GallileoWasWrong, the 2006 first printing). Not checked against a print copy. The 2013 Vol. II carries this section as ch. 9 at printed p. 133 with different wording — see the gloss

Read the whole paragraph and then the two pages after it, because the sentence quoted above is the careful one and it is not the book’s last word. What is being answered here is Ciufolini and Pavlis’s 2004 Nature result, which the book quotes accurately from a NASA Goddard release two paragraphs earlier — the Lense–Thirring “precession of the satellite’s node on the equatorial plane”, agreeing with general relativity “to 99% ± 5%.” The claim in the quoted sentence is negative: the measurement does not settle the question. Stopping there would be the hedge rule broken in our own favour.

Two pages later the book is flat, and it is the flat version this page answers. At printed p. 615: “We maintain, however, that the ‘hollow sphere’ is physical, and thus the recent discovery of the frame-dragging effect has a physical cause, not a ‘space-time’ cause.” A mechanism follows in the book’s own voice — the centrifugal force generated by the rotating universe is weakest near the Earth and strongest at the rim, and the Sun and planets sit at the radii where that centrifugal force balances gravity — and a page earlier, at p. 614, the residual of the rim’s motion reaching the Earth is identified with the roughly 4 km/s the early interferometer experiments were read as showing. Earlier still, at p. 610, the chapter states its conclusion about Einstein without a hedge on it: that general relativity “leads inevitably back to the ‘unthinkable’ position that the Earth is immobile in the center of the universe.” And the chapter does claim a centre and not merely a rotation: at p. 623, glossing Brill and Cohen, it concludes that “either the shell of ‘fixed stars’ must be fixed around a rotating center, or the center must be the fixed point for a revolving shell.” Those are the claims below, and section 1 answers that sentence rather than a weaker one.

The edition matters here more than usual. The 2013 three-volume rearrangement carries this section as Vol. II ch. 9, and at printed p. 133 of that scan the same paragraph has been strengthened twice. “Heliocentrism” becomes “a rotating Earth”, and a sentence is added with no counterpart at p. 613 of the 2006 scan: “In fact, the Lense-Thirring effect leans toward a rotating universe since it is easier to see how the large mass of the universe would drag the matter inside of it as opposed to the small Earth trying to accomplish a similar task.” Item 274 on the list tracks that later, positive sentence; our cluster record cites the earlier, negative one. The refutation answers both.

Whose physics the chapter is quoting. Thirring 1918 and Lense and Thirring 1918 are cited in the original German with translated titles, and quoted — including Thirring’s own remark that the required equivalence “appears to be guaranteed by the general covariance of the field equations,” and his admission that treating the heaven of fixed stars as an infinitesimally thin hollow sphere “is certainly not physical.” The rotating-cosmology citation in this chapter is Yu. N. Obukhov, “Rotation in Cosmology”, Gen. Rel. Grav. 24:121 (1992), quoted at length in a footnote at printed p. 608. All of it is real literature, reported without distortion. Gödel’s 1949 solution — which our own cluster record names as the source here — is not located in that chapter of the 2006 scan: a search of that scan’s full OCR text for every plausible spelling of the name returns three hits, two on the incompleteness theorem and one a bibliography line for a biography of Einstein. Gödel enters this material only in the 2013 Vol. II, and there not in the volume’s own text: both occurrences sit at printed p. 647, inside Appendix 1, which reprints Martin Selbrede’s writing — a longer quotation of Obukhov’s opening sentence, and a citation of a 1993 paper on synchronised frames for Gödel’s universe.

And the item-52 sentence is in a different chapter, with a rider the list drops. The equivalence claim in its plainest form is at printed p. 661, in ch. 11 “Hildegardian Geocentrism”, section “The Rotation of the Firmament”: “…geometrically speaking, there is no relative difference between a rotating universe around a fixed Earth and a rotating Earth in a fixed universe. They are, indeed, mirror images of one another. But there is only one true reality. As such, only one cosmology can be correct.” The tie is then broken, on that page, by the visions of Hildegard of Bingen. The authors are not hiding this; it is the chapter’s title. It does mean that the source itself treats the equivalence as settling nothing.

SURFACE (weak — do not use). “Frame dragging is real but far too small to swing the stars round the Earth in a day.” This loses, and it loses in one move. The measured Lense–Thirring effect is small because it is produced by the Earth, and its size bounds nothing whatever about dragging by a rotating cosmos. Brill and Cohen showed in 1966 (Phys. Rev. 143:1011) that a rotating shell’s dragging of its interior becomes complete as the shell approaches its own gravitational radius, and the observable universe is within a factor of about two of its own: 2GM/c² comes out near 2.2 × 10²⁶ m for M ~ 1.5 × 10⁵³ kg, against a comoving radius near 4.4 × 10²⁶ m. Thirring’s interior coefficient 4GM/3Rc² on the same numbers is about 0.3. Order unity, not ten orders down.

DEEPER. Rotating solutions of the field equations are not fringe. Gödel published one in 1949 in Reviews of Modern Physics; Schücking and Ozsváth built rotating Bianchi IX cosmologies through the 1960s; Obukhov’s 2000 review of the field carries several hundred references. And cosmic rotation is not a closed question in 2026: Szigeti, Szapudi, Barna and Barnaföldi proposed a slowly rotating dark-fluid variant of ΛCDM in 2025 as a route out of the Hubble tension. Anyone who says a rotating universe is forbidden by relativity has said something false and can be shown to have said it.

KERNEL. The strongest form puts those together and is genuinely uncomfortable. It runs: the equivalence is not a coordinate trick but a dynamical result — Thirring proved that a rotating shell induces Coriolis and centrifugal fields inside it, Pfister and Braun (Class. Quantum Grav. 2:909, 1985) fixed the coefficient so that the induced centrifugal force is exactly right, Brill and Cohen showed the dragging goes to completion near the gravitational radius, and our universe sits near that limit. So the sky’s rotation could in principle be the physical origin of every inertial effect we attribute to the Earth’s spin. That is not a re-labelling; it is a mechanism, and it is in the literature. Every clause of that is correct, and the last one is what the chapter under discussion is for.

Why it doesn’t save the claim.

Because the mechanism is symmetric and the model is not — and because the two moves that would break the symmetry both fail in public.

Complete dragging is the strongest form of the defence and it is also its self-liquidation. If the sky’s rotation drags the local compass of inertia exactly, then the compass and the sky turn together, the Earth turns relative to both at 7.292 × 10⁻⁵ rad/s, and every experiment returns what it returns now. That is the whole content of the claim: the same spacetime, described from a frame nailed to one planet. It yields no observation, and it cannot yield a centre, because the dragging relation contains the shell’s mass and radius and no privileged interior point.

The two ways out both fail. Make the dragging incomplete — which is what the book does when it insists the effect is “physical” rather than “space-time”, and when it wants a 4 km/s ether residual measurable at the Earth — and you now have matter rotating relative to the inertial compass at a rate the sky-wide observations bound: Saadeh et al. put (ω/H)₀ below 5.2 × 10⁻¹¹ across the Bianchi degrees of freedom, against a requirement of 3.3 × 10¹³. Or claim an asymmetry in the measurement, as the 2013 edition does — the great mass of the universe drags more easily than the small Earth — and the asymmetry has to be paid for in free parameters. General relativity predicts the measured node combination outright, 48.2 mas/yr, from the Earth’s independently measured polar moment and spin rate with nothing left to adjust; a cosmic shell dragging the interior compass has a mass, a radius and a rotation rate to choose, and must choose them so as to reproduce a number relativity computed. That is the difference between the two accounts, and it is section 3’s.

What is conceded, and it is most of the physics. General relativity admits rotating cosmological solutions; Gödel wrote one down in 1949 and the field has produced many since. Thirring’s 1918 result is real: a rotating mass shell induces Coriolis and centrifugal terms in its interior, and Pfister and Braun repaired the coefficient in 1985 so that the induced centrifugal force comes out exactly right. Brill and Cohen’s 1966 result is real: the dragging becomes complete as the shell nears its own gravitational radius. Cosmic rotation is a live research question in 2026, not a settled negative — a 2025 paper proposes a slowly rotating variant of the standard model as a route out of the Hubble tension. Nothing below treats any of that as fringe, and the book’s reporting of it is accurate.

What the verdict ranges over. Not “a rotating universe is impossible”, which would be false. The cluster’s claim is that these solutions and this measurement support a stationary Earth at the centre. Three things defeat that, and only the third depends on any disputed physics.

1. The chapter argues from rotating shells, and a shell’s centre is not a place the Earth has any claim on

Take the chapter as it is. Almost all of ch. 10 is a roll-call of named physicists enlisted for the rotating-shell picture rather than for rotating cosmologies — Thirring at printed p. 611, then Rosser, Bondi, Brill and Cohen, Moon and Spencer, Møller’s rotating ring, Brown, Nightingale, Lynden-Bell, Barbour and Bertotti, each under the heading “X’s Geocentrism”. That distinction matters, because a shell does single out a point, and the chapter knows it. At p. 623 it quotes Brill and Cohen — in a model where the shell sits at its own Schwarzschild radius, “there cannot be a rotation of the local inertial frame in the center relative to the large masses in the universe” — and draws the conclusion in its own voice: “either the shell of ‘fixed stars’ must be fixed around a rotating center, or the center must be the fixed point for a revolving shell.” The summary chapter says it again at ch. 12, printed p. 711, as the “all-important fact” about Barbour and Bertotti’s work: “An object at the center of the hollow sphere will not be affected by the inertial forces. The space around the Earth will exhibit the inertial effects of the distant sphere, but not the Earth itself, if it is centrally located.” (That sentence is the book’s own summary, not a quotation from Barbour and Bertotti.) This, not the rotating-cosmology literature, is where the cluster’s centre comes from, and it is what has to be answered.

Three things are wrong with reading a home for the Earth out of it. The centre belongs to the shell. What the geometry distinguishes is the point defined by the shell’s own symmetry; nothing in the relation between shell and interior says which body, if any, occupies it, or forbids that point from falling in empty space. Putting the Earth there is an assumption brought in from outside and then recovered — which is why ch. 11 has to settle the question with Hildegard of Bingen rather than with a metric. The vanishing is not pointlike. In Thirring’s interior solution the centrifugal-type term grows with distance from the rotation axis, so it vanishes along that entire axis — a line through the interior, not a point — while the Coriolis-type term is uniform inside and acts only on bodies already moving. “No inertial forces at the centre” is a statement about a locus and about which forces, not a proof that one body is privileged. And in the limit that makes the Brill–Cohen sentence true, it says the opposite of what it is being used for. “There cannot be a rotation of the local inertial frame in the center relative to the large masses” holds because the dragging is complete: the interior compass co-rotates with the shell. Brill and Cohen’s own gloss, in the passage the chapter quotes, is that this “explains why the ‘fixed stars’ are indeed fixed in our inertial frame”. A co-rotating interior is precisely the relabelling section 3 concedes; it is not a stationary Earth.

Where the chapter does reach for a rotating cosmology, it gets a direction and no place. Its one such citation is a long footnote at printed p. 608 quoting Obukhov’s “Rotation in Cosmology” (Gen. Rel. Grav. 24:121, 1992) for the claim that cosmic rotation carries no fatal observational consequences. Grant it, and see what it buys. Gödel’s rotating universe is spatially homogeneous: the isometry group carries any point to any other, so every dust grain in it may equally be described as sitting on the axis. Obukhov’s own 2000 review says the same of his Gödel-type expanding class in as many words — “models (3.1) are spatially homogeneous”, and “Coordinate z gives the direction of the global rotation.” The rotating solutions that review catalogues which are not homogeneous — Maitra’s generalisation of van Stockum’s metric, Wright’s dust-plus-Λ solution — are cylindrically symmetric: they have an axis, which is a line, and still not a point. Those two are our citations and not the chapter’s; they are here because they are the closest thing in the literature to what the argument needs.

So the argument gets its rotating universe from one footnote, gets no centre with it, and gets its centre instead from a shell whose centre is the shell’s. A metric solving the field equations for a matter distribution resembling ours and picking out a central body is not written down anywhere in ch. 10 of the 2006 scan, and is not in the Obukhov review the chapter cites. This is the gap the whole cluster steps over: “the equations permit rotation” and “the equations permit us” are different sentences, and only the first has been demonstrated.

2. The rate is wrong by thirteen orders of magnitude, and the honest bounds say so carefully

A geocentric cosmos has to turn once per sidereal day: ω = 7.292115 × 10⁻⁵ rad/s. Gödel’s solution fixes its own rate from its density, ω² = 4πGρ = −Λ. Recomputed here for ρ = 10⁻²⁷ kg/m³, within a factor of a few of the observed mean matter density: ω = 9.16 × 10⁻¹⁹ rad/s, one rotation per about 2.2 × 10¹¹ years. The rate scales only as the square root of the density, so no plausible revision of that figure moves the answer by an order of magnitude. That solution also requires a negative cosmological constant and does not expand, so it is doubly not our universe — a point Gödel himself pursued, asking astronomers for years whether any cosmic rotation had been detected.

Put the requirement in Hubble units and it is 3.3 × 10¹³ H₀. Now the bounds, in ascending order of how much a defender can argue with them. Saadeh, Feeney, Pontzen, Peiris and McEwen (PRL 117:131302, 2016) tested all the Bianchi degrees of freedom against Planck temperature and polarisation and report (ω/H)₀ < 5.2 × 10⁻¹¹ at 95% confidence, with anisotropic expansion disfavoured at odds of 121,000:1. That is 24 orders of magnitude below the requirement — and it is a bound inside a model class, so a defender is entitled to say his cosmos is not a Bianchi cosmos, which is why the weight of this section sits on section 1 and not here. The live proposal is more instructive than the bound anyway: Szigeti and colleagues’ rotating dark-fluid model, offered in 2025 as a resolution of the Hubble tension and explicitly argued to be “consistent with present observations”, runs at ω₀ ≈ 2 × 10⁻³ Gyr⁻¹, about 0.03 H₀. The most permissive rotating cosmology anyone is currently defending in the literature is around 10¹⁵ times too slow for this argument. Cosmic rotation being an open question is not the same as this being an open question.

And there is the light cylinder. At one turn per day, c/ω = 4.11 × 10¹² m = 27.5 AU, so everything from Neptune outwards would have to hold station in the Earth’s frame while moving faster than light. This objection is met head-on, and it is worth being exact about by whom. The reply usually quoted — if space can stretch superluminally, why can it not rotate superluminally; “Sauce for the general relativity goose is sauce for the geocentric gander” — is Martin Selbrede’s, from his 1994 paper answering Michael Martin Nieto, reprinted at printed p. 647 of the 2013 Vol. II as Appendix 1; it is not Sungenis and Bennett’s own text, and Selbrede does not appear on this review’s People tab. The book answers in its own voice elsewhere and differently: two pages past the Thirring section, at printed p. 616 of the 2006 Vol. I, “Rosser’s Geocentrism” quotes Rosser’s textbook that where gravitational fields are present “the velocities of either material bodies or of light can assume any numerical value”, so a co-rotating frame’s enormous centrifugal field licenses distant bodies exceeding c. Neither comparison holds, and the reason is worth stating rather than waving at. Superluminal recession in an expanding cosmos is a statement about comoving coordinates: every galaxy is at rest in a local inertial frame, no body overtakes light past any local observer, and the metric is well behaved everywhere. Rigid rotation is different in kind: beyond r = c/ω the co-rotating Killing vector turns spacelike, so “at rest with respect to the Earth” stops being a state of motion available to matter at all. That is also the answer to Rosser, who is quoted correctly and says something true: coordinate velocities in a gravitational field are unbounded, and nobody disputes it. The bound that bites is not on a coordinate velocity but on the causal character of the worldlines — matter must move on timelike curves, and past the light cylinder the curves that hold station in the Earth’s frame are not timelike. A defender may answer that the true geocentric spacetime is not flat space in rotating coordinates. Granted — and that answer is a promissory note for the metric section 1 says has not been produced.

3. The frame-dragging measurement is a measurement of the Earth’s spin

This is where the argument is convicted by its own evidence. The Lense–Thirring precession of a satellite’s node is

Ω̇_LT = 2GJ / [c²a³(1 − e²)^3/2]

where J is the angular momentum of the Earth. Take the standard polar moment of inertia C = 8.0340 × 10³⁷ kg m² and the sidereal rate Ω = 7.292115 × 10⁻⁵ rad/s, giving J = 5.86 × 10³³ kg m² s⁻¹. Recomputed here: a = 12,270 km, e = 0.0045 returns 30.7 mas/yr and a = 12,163 km, e = 0.0135 returns 31.5 mas/yr, against the 31 and 31.5 mas/yr published for LAGEOS and LAGEOS II. Set the Earth’s rotation to zero and both predictions are exactly zero. The quantity Ciufolini and Pavlis measured, and which the book presents as not proving a moving Earth, is a function whose only free input is how fast the Earth turns.

Be precise about the error bar, because a defender will check it. The 2004 paper reported 0.99 ± 0.05 of the general-relativistic value. Iorio disputed that budget for years: once the drift of the Earth’s even zonal harmonics is folded in, he argued, the accuracy “might be 15-45% (1-3σ)”, and comparing the GRACE-based gravity models against each other rather than trusting one model’s covariance gives a geopotential bias “as large as about 25-30%”. It does not matter here: at any of those error bars the result is non-zero and consistent with the J ≠ 0 prediction; the later three-satellite analysis returns μ = 0.994 ± 0.002 ± 0.05; and Gravity Probe B reached the same conclusion by a completely different route, four cryogenic gyroscopes in polar orbit returning a frame-dragging drift of −37.2 ± 7.2 mas/yr against −39.2 predicted — a prediction likewise computed from the Earth’s angular momentum.

Now the defence, at its strongest, and what it costs. A well-read geocentrist answers that J is angular momentum relative to the local compass of inertia; that in his model the compass is dragged round by the sky; and that the Earth therefore still has the same J relative to it, so the prediction is unchanged. That is correct, and it should be conceded in plain sight — and it concedes that the Earth rotates at 7.292 × 10⁻⁵ rad/s with respect to the local inertial compass, that being the rate fed into the prediction which the satellite nodes then confirm. The dispute that survives is over which body deserves the word “really”, and that is ARG-R01’s question, not an observation.

What does not survive is the 2013 edition’s asymmetry claim — that the effect “leans toward a rotating universe” because a great mass drags more easily than a small one. Two things answer it, and it is worth being exact about what the measurement is, because the tempting answer is wrong. The Ciufolini–Pavlis observable is not a difference between the two satellites. It is a positively weighted sum of their node residuals, δΩ̇(LAGEOS) + 0.546 × δΩ̇(LAGEOS II), and it was built, in Iorio’s words, to “cancel out the first even zonal harmonic J₂” — two node equations solved for two unknowns, the relativistic scaling μ and the Earth’s quadrupole. A spatially uniform dragging of the interior compass at some rate would enter that sum with weight 1.546 and be swallowed whole into μ. Within this combination there is no degree of freedom left with which to tell a constant from a gradient, and any claim otherwise should be struck.

What does the work instead is arithmetic of a different kind. General relativity predicts the number; the rival account fits it. The slope of that combination, 48.2 mas/yr, follows from the Earth’s independently measured polar moment and sidereal rate with nothing adjustable in it at all — our own recomputation of the two rates gives 47.9. A rotating cosmic shell has a mass, a radius and a rate to choose, and must choose them so as to land on a figure relativity computed in advance. And the distance dependence is testable — by adding a node at another altitude, not by combining these two. Uniform dragging displaces every node at the same rate whatever the orbit; the Lense–Thirring prediction falls as 1/a³. LARES, launched in 2012 at a = 7,821 km against LAGEOS’s 12,270 km, has a predicted rate of 118.4 mas/yr — recomputed here to the same figure — nearly four times LAGEOS’s 30.7, out of the same single J, with nothing tuned between them; and the three-node analysis returns μ = 0.994 ± 0.002 ± 0.05 against exactly that spread of predictions. Be exact about what that settles: any combination built to cancel the Earth’s zonal harmonics is fitted for a single scaling μ, there as here, so a constant offset is absorbed and not separated. What discriminates is not a residual shape but the fact that one account computes 30.7, 31.5 and 118.4 mas/yr before looking, from a moment of inertia and a spin rate measured by other means, and the other has a free constant and has to be told the answer.

4. The mechanism the book states in its own voice does not balance

The chapter is not content with equivalence, and this is where answering the source rather than the fragment costs the source most. Its stated physics at p. 615: the centrifugal force generated by the rotating universe is weakest near the Earth and strongest at the rim, an object that reaches escape velocity “joins the centrifugal force”, and the Sun and planets sit at the radii where that centrifugal force balances gravity. All of that is checkable with a calculator.

The cosmic centrifugal acceleration at radius r is ω²r with ω² = 5.3175 × 10⁻⁹ s⁻². At the Moon that is 2.04 m/s², while the Earth’s gravitational pull there is GM_⊕/r² = 2.70 × 10⁻³ m/s². The claimed balance is out by a factor of about 760. The mismatch is ω²r³/GM_⊕, so it grows as the cube of the distance: at the Sun’s distance it is 795 m/s² against 1.8 × 10⁻⁸ m/s², a factor of about 4 × 10¹⁰. The one radius where the two do balance is (GM_⊕/ω²)^1/3 = 42,164 km — geostationary orbit, which is what that number is — where the speed is 3.07 km/s and not the escape velocity the passage names. So the model as stated puts every body in the sky at the altitude of a television satellite.

The repair is available and it is the same one as before: hand the balancing back to Thirring’s induced Coriolis and centrifugal fields, which do the bookkeeping automatically for bodies carried round with the shell. Take it, and the arithmetic works — because the resulting model is the ordinary one written in rotating coordinates, with no ether, no rim, no residual drift, and no observation to its name. The book cannot have the physical rim and the relativistic equivalence at once. Its own p. 615 concession about the physicists it cites is the honest description of what is on offer: none of them, it says, present their findings as accepting geocentrism as a scientific fact; what they accept is that the same forces can be applied equally well to a geocentric universe. Equally well is the whole result, and the list turned it into evidence.

The list says

Indistinguishable rotating universe equivalence.

The source says

“geometrically speaking, there is no relative difference between a rotating universe around a fixed Earth and a rotating Earth in a fixed universe. They are, indeed, mirror images of one another. But there is only one true reality. As such, only one cosmology can be correct.”

A concession was re-used as a proof. The wording barely moves and the speech act moves completely, which is the ARG-R01 pattern exactly. Indistinguishable is the source’s own word for it — “mirror images” — and it is a statement that the geometry decides nothing. On the source’s page (Vol. I, printed p. 661, ch. 11) the very next sentences say so: only one cosmology can be correct, and the choice is then made, in that chapter, on the authority of Hildegard of Bingen’s visions. On the list the same proposition is item 52 in a sequence of numbered proofs that the Earth is not a spinning ball. A concession of non-discrimination is republished as discriminating evidence, with the rider that carried it dropped in transit.

The other two items drift the same way and one of them drifts across editions as well. Item 273, “GR allows rotating universe”, is a fair rendering of what the chapter’s footnoted Obukhov material supports — allows is the right verb — and it is offered as a proof. Item 274, “Lense–Thirring frame dragging”, is a bare noun phrase standing in for a paragraph whose 2006 wording is negative: the LAGEOS result “do[es] not prove either General Relativity or heliocentrism” (Vol. I, printed p. 613). The corresponding paragraph in the 2013 Vol. II, at printed p. 133, replaces “heliocentrism” with “a rotating Earth” and adds a sentence with no counterpart at p. 613 of the 2006 scan: the effect “leans toward a rotating universe since it is easier to see how the large mass of the universe would drag the matter inside of it as opposed to the small Earth trying to accomplish a similar task.” So on this argument the drift begins in the source’s own revision history and the list inherits the later, firmer form.

What the compression also throws away is the part that could be checked. The chapter does not stop at permission: at printed pp. 614–615 it commits to a physical rotating rim, an ether carrying its effect inward, a residual drift of about 4 km/s at the Earth, and planets sited where cosmic centrifugal force balances gravity. Those are falsifiable and the refutation above tests them — the balance condition misses by a factor of about 760 at the Moon. The three-word items keep only the unfalsifiable half. That is the usual direction of travel for this list, with an unusual consequence: the compressed version is safer than the book, and reads as a modest citation of mainstream relativity while quietly carrying the conclusion the book had to buy with a mechanism.

The refutation answers the source, not the fragment: it grants the equivalence in full, grants Thirring, Pfister–Braun and Brill–Cohen, grants that cosmic rotation is a live question in the current literature, and then puts the weight on the two things the source’s own citations cannot supply — a centre, since the shell results the chapter runs on distinguish the shell’s centre and not the body that sits in it, while the one rotating cosmology it cites is spatially homogeneous; and a rate, since Gödel’s relation gives one turn per about 2 × 10¹¹ years where a stationary Earth needs one per day.

the physics of the geocentric universe accounts perfectly for what we see and measure of the daily rotation whether that rotation is of the earth within the universe or the universe around the earth. In the final analysis, proofs based on dynamical equations are not proofs of anything; nor are they proofs against the geocentric universe.

— Geocentricity (1992), Appendix E, “Derivation of the Geocentric Equations for a Daily-Rotating Universe”, closing paragraph, printed p. 747. Quoted from the Internet Archive OCR of the 2013 edition — GEOCENTRICITY: CHRISTIANITY IN THE WOODSHED, Association for Biblical Astronomy, ISBN 9781890120900, archive item geocentricity-christianity-in-the-woodshed — and not from the 1992 first edition our work record names; not checked against a print copy. Three plain OCR artefacts are corrected: a comma for the full stop after “earth”, “fina!” for “final”, and a comma for the closing full stop.

Read the second clause. “Nor are they proofs against the geocentric universe.” The sentence is symmetric on its face, and Bouw means it symmetrically: two chapters earlier he applies the same rule to the Earth’s equatorial bulge and concludes that it “offers no proof for either heliocentric or geocentric theories” (ch. 28, p. 425). This is a claim that a whole class of argument settles nothing for anybody. On the list it becomes item 293 in a document whose title and structure present every line as a proof.

The chapter-1 version carries two more hedges, and they are the ones that vanish. At p. 4 the proposition is other people’s and is granted only in part: “The more subtle physicists, many of whom know well that the geocentric evidence is overwhelming, will claim, with some justification, that we can neither prove nor disprove the geocentric universe; but that we likewise can neither prove nor disprove the non-geocentric universe either.” Attributed, part-granted, and explicitly two-way. The middle clause is quoted here in full because it cuts against us and an earlier draft of this entry elided it: it is not a remark elsewhere on the page but part of the same sentence, and it is Bouw asserting in his own voice that the geocentric evidence is overwhelming. Cut it and he reads as more tentative than he is. That is the tension this entry is about, not a hedge we can hide behind.

Where item 310 comes from, and it is a real citation. Bouw’s chapter 36 note 2 reads “Ellis, G. F. R., 1978. General Relativity and Gravitation, 9:87. Quote is from page 92”, and the paper is exactly what he says it is: George Ellis, co-author of The Large Scale Structure of Space-Time, showing that spherically symmetric static spacetimes “can reproduce the same cosmological observations as the currently favored Friedmann-Robertson-Walker universes … provided that the universe is inhomogeneous and our galaxy is situated close to one of its centers.” Nothing in that chain is invented. Two things about the note need saying, though, and neither is in the item’s favour. The note is not attached to the equivalence result: the sentence it footnotes, at p. 534, is Ellis on the cosmological principle being “assumed for a priori reasons and not tested by observation”. And where Bouw handles the construction itself, at pp. 538–539, he does not transmit it as one of several models that fit — he calls it “an oddity among cosmological models” which “is not without its problems”, and reads it as a model that “the preponderance of geocentric evidence in cosmology has finally forced”. That last reading is the tension in a sentence: a claim that the evidence decides, in the same book as the claim that nothing can. What the chain drops is Ellis’s next sentence, which lists the routes out — and which the field then took.

What is established here is that the doctrine is Bouw’s in print, not that it is Bouw’s in origin. He attributes the proposition on p. 4 to “the more subtle physicists” — that is, to other people; item 318 could not be located in this text at all; and no earlier claimant was tested in this pass. This project records origin in three states rather than two, and the third is available here: the origination credit in our record is provisional, and is published as such.

On the edition. Our work record names Geocentricity (1992). No copy of that first edition was reachable from here, so every quotation in this treatment is from the 2013 edition, retitled Geocentricity: Christianity in the Woodshed, which is complete and searchable on the Internet Archive. That the same doctrine was in the 1992 book is corroborated twice from outside it: Danny Faulkner’s review of the 1992 edition (Journal of Creation 15(2), 2001) reports the model as “a co-ordinate change from the Sun to the Earth”, and Bouw’s published reply to that review restates the position while discussing the 1992 book by name.

SURFACE (weak — do not use). “Of course there is a falsifier: the Foucault pendulum / stellar parallax / a gyroscope.” Every one of those is answered inside the model by a rotating firmament that drags the local inertial frame, and a defender who has read Bouw will say so in one line. Equally weak: “unfalsifiable claims are worthless.” The claim on the table is that nothing discriminates, which is a claim about the evidence, and it has to be met with evidence rather than with a slogan about Popper — particularly against an author who devotes a page to rejecting Popper by name (ch. 26, pp. 407–408).

DEEPER. Underdetermination is real, and the specific logical point Bouw makes is simply valid. He quotes Charles Lane Poor on Einstein’s use of aberration — “How can an experiment, equally well explained by several different theories, be a ‘crucial test’ in favor of one of them?” (ch. 33, p. 473) — and that is correct as stated. A phenomenon derivable from two theories does not adjudicate between them. Anyone who wants to answer this cluster by denying that principle will lose, because the principle is ours too.

KERNEL. The strongest form is not Bouw’s and is not theology. It is that a named, credentialed relativist published the construction. Ellis, GRG 9:87 (1978): spherically symmetric static general-relativistic cosmologies reproduce the same observations as FLRW “provided that the universe is inhomogeneous and our galaxy is situated close to one of its centers.” No retraction, erratum or correction is attached to that paper on the publisher’s record page, which is where we looked; it seeded a live literature on Lemaître–Tolman–Bondi models, and Ellis went on to argue in general terms that the Copernican principle is an assumption rather than a measurement (“Issues in the Philosophy of Cosmology”, 2006). Add the honest state of the standard alternative: the horizon problem is a real problem, inflation is its standard cure, and whether inflation is falsifiable at all is a fight being conducted in Physics Letters B and in the pages of Scientific American by the people who built the subject. A defender who says “your own field cannot agree on what would refute its account of the initial data” is not making that up. Concede all of it.

Why it doesn’t save the claim.

Because Ellis wrote the way out in the same abstract, and then helped take it. The 1978 paper ends its own claim with a scope: only (i) unverifiable a priori assumptions, (ii) detailed physical and astrophysical arguments, or (iii) observation of the time variation of cosmological quantities could tell us we do not live in such a spacetime. That is not a declaration of permanent undecidability; it is a specification of what work would have to be done. Route (iii) became Uzan, Clarkson & Ellis, PRL 100:191303 (2008), redshift drift as a test of the Copernican principle — Ellis co-authoring a test of his own construction. Route (ii) became Caldwell & Stebbins, PRL 100:191302 (2008), Clarkson, Bassett & Lu, PRL 101:011301 (2008), and Zhang & Stebbins, PRL 107:041301 (2011). The underdetermination was a research programme with an address, and somebody went there.

And it does not save this cluster in particular, because the Ellis construction and the Tychonic model are not the same object. Ellis’s is a spacetime with a stress-energy tensor, a metric and a Hubble diagram to fit; it makes the Earth’s galaxy near a centre and says nothing whatever about the Earth. Bouw knows this and says so in his own book: the Milky Way “could just as well be viewed as located at the center of the cosmic shells … and the earth is not exactly at the center of the F-stars and G-stars, either. But that is where Scripture comes into play” (ch. 36, p. 556). The published construction reaches a galaxy; the remaining distance to a planet is covered, in the source’s own words, by Scripture. That is a candid sentence and it is fatal to item 293, which claims the gap is closed by the absence of evidence rather than left open by it.

First, the concession, and it is larger than the list probably expects. Two of the three things this cluster asserts are true. A coordinate change to an Earth-centred, Earth-fixed frame is exact, legal and performed several million times a day by every satellite receiver on the planet; nothing observable distinguishes a description from its relabelling, and no experiment ever will. And “multiple cosmologies fit the data” is not a crank line: George Ellis published one in General Relativity and Gravitation 9:87 (1978), showing that a static, spherically symmetric, inhomogeneous spacetime reproduces the same cosmological observations as the expanding models “provided that … our galaxy is situated close to one of its centers.” Anyone answering this cluster by asserting that no reputable cosmologist has ever written down an us-near-the-centre universe is wrong on the record and deserves to lose the exchange.

Second, the fork, which is where the cluster comes apart. Those two concessions are not the same concession, and they cannot both do the job the list needs.

A change of frame is undecidable because it has no content. There is nothing to test, so nothing tests it, so it can never be evidence for anybody — and the same sentence protects “the Earth orbits the Sun”, which is equally a statement about coordinates until you say what is doing the accelerating. Say this plainly rather than let it be extracted: the physical content on both sides is the relative motion of matter plus the local inertial structure, and that is measured, not chosen. A change of cosmology — a universe genuinely inhomogeneous about us — is the opposite kind of thing. It is a different spacetime with a different stress-energy content, so it makes different predictions, so it is testable. Item 293 is safe because it is empty; item 310 is interesting because it is not. Welding them together buys the emptiness of the first and the interest of the second and is entitled to neither.

Third, and this is the part that answers the source rather than the fragment: Bouw’s own model is on the testable side of that fork, and he puts it there himself. His cosmos is not a relabelling. The firmament, he writes at p. 5, “is a superdense medium that pervades all of space. It is the firmament that dictates the laws of physics, and it is the firmament that physically controls all motion.” In Appendix E the centrifugal and Coriolis terms of the rotating frame are not bookkeeping: they “keep the star in its place in the inertial field of the universe which is the gravitational field of the firmament” (p. 746). He has an answer ready to the charge that all this is relabelling, and it deserves one back: Appendix E opens (pp. 740–741) by calling the step from a kinematic to a dynamic equation a “sleight of hand” — multiplying one side by m/m and calling the product physical — and he is right that a factor of m confers nothing. But that factor is not what puts his cosmos on the testable side of the fork. The medium is. A firmament that “physically controls all motion” and whose gravitational field is what holds a star in its place is a substance with properties, not a convention, whichever side of the equation the mass sits on. At p. 523 the model is stated as making predictions in so many words — “Geocentricity predicts that earth’s lack of motion is absolute … Geocentricity further predicts that rotation is relative” — and, a page earlier, Sagnac interferometry is said to have “been performed accurately enough to discern the period of absolute rotation of the firmament is the sidereal day” (p. 522). That is a quantitative measurement of a physical medium, claimed as a result. A theory that predicts can fail; a theory whose medium has a gravitational field has a stress-energy tensor and consequences. So the source asserts a falsifiable cosmology in the body of the book and, in Appendix E, the proposition that no dynamical argument can settle anything either way. Both cannot be load-bearing. The list bought the second and kept spending the first — items 1, 2, 3 and 10 of the same document are Michelson–Morley, Michelson–Gale, Sagnac and Michelson–Pease, offered as experimental proof.

Fourth, what happened to the interesting half after 1978. Ellis did not claim the question was closed; he specified what would open it — detailed physical arguments, or observation of the time variation of cosmological quantities. Both routes were built. Redshift drift as a Copernican-principle test: Uzan, Clarkson & Ellis, PRL 100:191303 (2008), with Ellis co-authoring a test of his own 1978 construction. A general consistency test on H(z) and distance: Clarkson, Bassett & Lu, PRL 101:011301 (2008). The reionised universe used as a mirror, so that the blackbody spectrum of the microwave background reports on our own gravitational potential: Caldwell & Stebbins, PRL 100:191302 (2008), whose limits “exclude the largest voids which mimic cosmic acceleration”. The kinematic Sunyaev–Zel’dovich power spectrum: Zhang & Stebbins, PRL 107:041301 (2011), which “rules out the adiabatic void model as a viable alternative to dark energy”. And the combined-data fit: Moss, Zibin & Scott, Phys. Rev. D 83:103515 (2011), using supernovae, the full CMB spectrum, radial baryon acoustic oscillations, the local Hubble rate, ages, big-bang nucleosynthesis, the Compton y-distortion and σ₈, and finding voids “in severe tension with the data”.

State the limit of that as carefully as the papers do. This is not a clean kill and must not be sold as one. Zibin & Moss, Class. Quantum Grav. 28:164005 (2011), revisited the kSZ bound under a fully relativistic treatment and found its constraining power “considerably weakened (though still impressive)”, flagging “several theoretical ambiguities and observational shortcomings which further qualify the results” — while still concluding that “a very large class of void models is ruled out by the combination of kSZ and other methods”. What is established is narrower than “we are not near a centre”: it is that the specific Gpc-scale inhomogeneous models proposed as a replacement for dark energy do badly against the data. That is enough to settle the point at issue here, which is not whether such a universe is conceivable but whether no falsifier exists. Falsifiers were designed, published in Physical Review Letters, and applied.

Fifth, item 318 and the horizon problem, which deserves a straight answer. The problem is real: in an expanding universe without inflation, patches of the last-scattering surface more than a degree or two apart were never in causal contact, yet their temperatures agree to about one part in 10⁵. It is also true that the standard cure is contested by serious people — Ijjas, Steinhardt & Loeb, Phys. Lett. B 723:261 (2013), and again in Scientific American in February 2017, argued that inflation as practised had stopped being falsifiable; thirty-three physicists including Guth, Linde, Kaiser and Nomura replied in the same magazine in May 2017, and the authors replied to the reply. Pointing at that argument is not a foolish thing to do.

But does putting the observer at a centre solve it? Partly, and at a price that is itself measurable. A spherically symmetric model about us makes angular isotropy automatic, which is the appeal; it does not explain radial uniformity, which becomes a choice of initial data rather than a consequence of anything, so the tuning is moved rather than removed. And it converts a tuned initial condition into a tuned position — which, unlike the initial condition, can be measured. Alnes & Amarzguioui, Phys. Rev. D 74:103520 (2006), computed how far off-centre an observer in such a model may sit before the induced dipole exceeds the observed one: “within a radius of 15 Mpc from the center”. Fifteen megaparsecs is roughly fifty million light-years. It is a bound on the position of a galaxy. None of the results cited above — Alnes & Amarzguioui, Caldwell & Stebbins, Clarkson, Bassett & Lu, Zhang & Stebbins, Moss, Zibin & Scott — constrains an observer’s position at better than megaparsec resolution, and at that resolution the Earth and the Milky Way are the same point. The source agrees: Bouw’s own chapter 36 grants that the Milky Way would serve as well as the Earth and that the Earth is not at the centre of the nearby stellar distribution, then names Scripture as what closes the gap. One further note on provenance: a search of the full 2013 OCR text located no occurrence of the phrase “horizon problem”, and Bouw invokes inflation approvingly at p. 104, as the mechanism by which starlight reaches a young Earth. Item 318 is answered here on its merits; it is not laid at his door.

Sixth, the cost of spending this argument. R11 is a universal solvent and it dissolves the hand that holds it. If dynamical arguments prove nothing either way, then Michelson–Gale is not evidence, Sagnac is not evidence, “Airy’s failure” is not evidence and Miller’s drift is not evidence — and every one of those appears elsewhere on this same list as an experiment that came out the geocentrist’s way. ARG-R03 makes the same observation about the relativity principle; R11 is the sharper case, because R03 at least leaves the geocentrist his experiments while R11 explicitly retires them. Bouw took the consistent route and said what does the deciding: for a Bible believer, he writes at the conclusion of chapter 2, the scriptural point “should be enough” (p. 15). That is a coherent position. It is not, however, a proof, and it cannot be counted three times in a list of 461 of them.

Verdict: self-contradicted. Not because the underlying physics is wrong — the frame half is trivially right and the cosmology half was a legitimate open question in 1978 — but because the three items cannot be held together with the rest of the document, or with each other, or with their own source. A claim that no observation discriminates, filed as an observation-based proof; a claim that the equations are identical, filed beside forty items claiming that the experiments came out one way; and a construction whose author specified the tests, ran one of them, and reported that the models it supports are in trouble.

The list says

No falsifier distinguishing frames. / Multiple cosmologies fit data. / Horizon problem Earth initial data.

The source says

“the physics of the geocentric universe accounts perfectly for what we see and measure of the daily rotation … In the final analysis, proofs based on dynamical equations are not proofs of anything; nor are they proofs against the geocentric universe.” (Appendix E, p. 747) — and, at p. 4, “The more subtle physicists, many of whom know well that the geocentric evidence is overwhelming, will claim, with some justification, that we can neither prove nor disprove the geocentric universe; but that we likewise can neither prove nor disprove the non-geocentric universe either.”

A concession was re-used as a proof. The wording barely moves and the speech act moves completely, which is the R01 pattern in a different book. In Geocentricity the proposition is a symmetric disclaimer: dynamical arguments prove nothing for either side, and Bouw applies it against his own side too — the Earth’s oblateness “offers no proof for either heliocentric or geocentric theories” (p. 425). On the list it appears three times in a document that presents every line as a proof, so a statement that the evidence is silent is filed as evidence. Two further qualifications are dropped on the way. At p. 4 the proposition belongs to “the more subtle physicists” and is granted only “with some justification”; the list states it flat and in the source’s own name. And item 310’s pedigree runs through a citation Bouw gets right — Ellis, GRG 9:87 (1978) — whose abstract attaches an explicit escape clause naming the arguments and observations that would decide the matter; the clause does not survive into the item, and the field spent 2008–2011 acting on it. Bouw’s own handling of Ellis drifts the other way, and it is worth publishing against us as well as against the item: at p. 539 he calls the model “an oddity among cosmological models” that “is not without its problems”, and reads it not as underdetermination but as a model the “preponderance of geocentric evidence in cosmology has finally forced” — a claim that the evidence decides, which is nearer the opposite of item 310 than a restatement of it. The refutation above answers the source, not the fragment: it concedes the frame point outright, concedes that a reputable inhomogeneous cosmology was published in a mainstream journal and carries no retraction on the publisher’s record, quotes the caveat that the void-model constraints are weaker than the headline results suggest, and puts the weight on the fact that Bouw’s own firmament is a physical medium with a gravitational field that makes predictions he reports as measured. A third item, 318, invokes the horizon problem; a search of the full 2013 OCR text located no occurrence of that phrase, and Bouw invokes inflation approvingly at p. 104, so that item is answered on its merits rather than attributed to him.

This clearly shows that the Copernican Principle from which modern science creates its interpretations of the cosmological data is not scientific but philosophical. In other words, even if the empirical evidence shows Earth is not moving, the ever-present Copernican Principle requires that every piece of scientific data must be interpreted by assuming the earth is moving…

— Galileo Was Wrong: The Church Was Right (2006), Vol. I, chapter 1, “The New Galileo and the Truth about Copernicanism”, printed p. 92 of the seventh edition (2013); read in the Internet Archive three-volume scan (item galileo-was-wrong-the-church-was-right-sungenis-vol-1-3-complete), djvu text line 5,917. Not checked against a print copy.

Read what is being claimed, because it is bigger than the item. The four-word list entry says only that the Copernican principle is an assumption. The book says that it is a philosophical assumption which overrides evidence — that a datum showing a stationary Earth would still be read the other way, because the principle decides in advance how every datum is read. That is a checkable charge about scientific practice, and it is the one answered below. The bare version is not worth answering: it is true, and it is stated in the same words by the paper this page cites against it.

The dedicated treatment is in chapter 3, “Evidence Earth is in the Center of the Universe”, at printed pp. 308–311 of the seventh edition, where the principle is called something “taken as an a-priori truth to which the rest of cosmology must conform.” The same paragraph, word for word, is in the earlier CD edition of Volume I (ISBN 0-9779640-0-0) at printed p. 145, so the argument is the book’s and predates the 2014 film by the better part of a decade. This treatment therefore quotes and cites the book, whose byline — Sungenis and Bennett — is the one our record carries.

The chapter is built almost entirely out of working cosmologists’ own words, and quoted accurately: Bondi’s Cosmology, Hawking and Ellis’s The Large Scale Structure of Space-Time, George Gale in Scientific American (1981), Ellis’s 1995 Scientific American profile, Clifton, Ferreira and Land’s 2008 supernova paper, and Ellis’s 2011 review of inhomogeneity effects. Nothing here turns on a misquotation. It turns on where each quotation is cut — which is a harder problem, and the subject of section 3 below.

SURFACE (weak — do not use). “The Copernican principle is not an assumption, it is proven.” This loses immediately. The paper that closed the loophole opens: “The Copernican principle, a cornerstone of modern cosmology, remains largely unproven at Gpc radial scale and above” (Zhang & Stebbins, PRL 107:041301). Anyone who denies the premise is contradicted by the citation they are about to make.

DEEPER. “It is an assumption, but a testable one, and it has been tested.” True, and still incomplete, because it invites the obvious reply: the tests were run inside a framework that already assumes what they check, using structure-growth and foreground models built on ΛCDM.

KERNEL. The strongest form is a theorem, not a grievance, and it is George Ellis’s. We observe the universe from essentially one point on one worldline, and the finite speed of light mixes distance with time, so isotropy is directly observable and homogeneity is not. The standard argument bridges the gap with the Ehlers–Geren–Sachs theorem — if every observer sees an isotropic background, the geometry is Friedmann–Lemaître–Robertson–Walker — and the phrase “every observer” is doing work no telescope can do for us. Worse for the standard picture, and this is a published theorem rather than a worry: Mustapha, Hellaby and Ellis (MNRAS 292:817, 1997) proved that “homogeneity cannot be proven without either a fully determinate theory of source evolution, or availability of distance measures that are independent of source evolution” — a spherically symmetric Lemaître–Tolman–Bondi model carries two free radial functions, and the freedom is enough to fit isotropic observations. So the underdetermination is exact, it is proved, and its author says out loud that cosmology chooses among the survivors on philosophical criteria: “I can construct you a spherically symmetrical universe with Earth at its center, and you cannot disprove it based on observations. You can only exclude it on philosophical grounds … A lot of cosmology tries to hide that.” That is Ellis in Scientific American, October 1995, read here as quoted at printed p. 310 of the seventh edition of Galileo Was Wrong; the magazine text itself was reachable only behind a paywall from here, so the wording is reported at one remove. Concede every word of it.

Why it doesn’t save the claim.

Because the man who proved the underdetermination also spent thirty years building the instruments that break it, and the book quotes the complaint without the programme. Ellis’s point was never “therefore we cannot know”; it was “precisely because of the foundational nature of the Copernican Principle for standard cosmology, we need to fully check this foundation.” That is a work order. It was filled.

Goodman proposed using scattered CMB photons to look at the sky from somewhere else in 1995. Caldwell and Stebbins turned it into a bound in 2008, treating the reionized universe as a mirror. Clarkson, Bassett and Lu produced a consistency test the same year that depends only on the geometry being Robertson–Walker — independent of curvature, of dark energy, of what the matter is and of the theory of gravity — which is exactly the model-independence the circularity objection demands. Clifton, Ferreira and Land showed the void and dark-energy readings separate in the redshift dependence of luminosity distance around z ~ 0.1–0.4. Uzan, Clarkson and Ellis proposed watching redshifts drift. Zhang and Stebbins then closed the largest gap with the kinetic Sunyaev–Zel’dovich power spectrum.

So the kernel points the other way at the exact step it was supposed to defend. An unexamined assumption that gets examined stops being an unexamined assumption; that is what happened here, in refereed journals, mostly before the seventh edition went to press. And the theorem never had the reach the argument needs anyway: the freedom Mustapha, Hellaby and Ellis established is the freedom to fit isotropic observations with a radially inhomogeneous model. It says nothing whatever about whether the Earth goes round the Sun.

First, the concession, and it is not grudging. The Copernican principle is an assumption. It was named after Copernicus rather than by him, its modern cosmological form dates to Bondi around 1948–1952, and it is a postulate about the large-scale distribution of matter that was adopted because it makes the equations tractable and the models predictive. The source’s own quoted authority says so plainly — George Gale in Scientific American, quoted at printed p. 309 of the seventh edition, dates the extended form to Bondi in 1948. And the paper this page cites in answer says so in its first sentence: “The Copernican principle, a cornerstone of modern cosmology, remains largely unproven at Gpc radial scale and above.” A rebuttal that begins by denying the premise has already lost.

Second, why it was an assumption, which is a better story than either side usually tells. We observe from what is effectively a single point, and every cosmological observation looks down a light cone that mixes how far away a thing is with how long ago it was. Isotropy — the sky looking the same in all directions — we can measure directly. Homogeneity — the universe looking the same from other places — we cannot, not by looking. The classical bridge is the Ehlers–Geren–Sachs theorem: if the background radiation is isotropic for every observer in an expanding universe, the geometry has to be Friedmann–Lemaître–Robertson–Walker. The words “every observer” are the assumption, and they carry the whole standard model. Sungenis and Bennett have located a real load-bearing beam.

Third, the sentence they quote, and the five pages in front of it. The book’s strongest single citation is Ellis’s 2011 review Inhomogeneity effects in Cosmology (arXiv:1103.2335; Class. Quantum Grav. 28:164001), footnote 427, cited to “pp. 19, 5”. The quoted line is real and is on p. 19: “Precisely because of the foundational nature of the Copernican Principle for standard cosmology, we need to fully check this foundation. And one must emphasize here that standard CMB anisotropy studies do not prove the Copernican principle: they assume it at the start.” What the same paper carries on p. 14, under the heading The argument for homogeneity, is this: “However it is now known that this assumption is indeed at least partly testable via measurements of CMB spectrum distortions, as will be discussed below. There are a number of other observational tests of the Copernican principle that are now possible, because of observational improvements in the past decade.” Section 4.4 is headed Observational tests of spatial homogeneity and sets out four kinds. Reference [128] of that paper is Zhang and Stebbins. The demand and its answer are in the same document, five pages apart, and the book carries the demand.

The pattern repeats with the other quotation the chapter leans on. At printed p. 320 it quotes Robert Caldwell wishing for an outside view — “It would be great if there were someone out there who could look back at us and tell us if we’re in a void” (Marcus Chown, New Scientist, 12 November 2008, p. 33). Caldwell had published that experiment with Stebbins six months earlier, in May 2008: use the reionized universe as a mirror, let it scatter background photons back to us, and read off what the sky looks like from out there. The wish in the quotation had already been granted by the man being quoted.

Fourth, the measurement, with the numbers. A universe that violates the Copernican principle radially — a deep void with us near the middle, big enough to fake cosmic acceleration — makes distant matter move relative to the background radiation, and that relative motion prints a kinetic Sunyaev–Zel’dovich signal on the small-angle microwave sky. Zhang and Stebbins computed it. The observed ceiling comes from two experiments at multipole ℓ = 3000: the South Pole Telescope’s 95% upper limit of ΔT² < 6.5 µK², the tighter of the two, and the Atacama Cosmology Telescope’s ΔT² < 8 µK² at the same angular scale. Void models within the 3σ contour of the UNION2 supernova fit predict ΔT² above 10³ µK² — in the paper’s words, “two orders of magnitude larger than the SPT upper limit” — and the models retuned to slip under that limit then miss the supernovae by Δχ² > 209. Stated model-independently, the deviation of any ~1 h⁻¹ Gpc shell from the overall expansion is held to about 1%. The paper is equally explicit about where the kSZ leg of the test stops, and so is this page: voids shallow enough (Ω₀ above about 0.8) or small enough that the underdensity ends below z ≈ 0.2 — radius under about 0.6 h⁻¹ Gpc — survive the kSZ bound, and are caught instead by the supernovae. That is why the two tests are always quoted together, and it is worth holding on to for section six. What the pair of them did is turn the principle from a postulate into a number with an error bar, and that is the specific measurement the verdict rests on.

Fifth, be honest about what is still open, because two things are. Ellis in 2011 recorded the kSZ claims as contested: Clarkson and Regis had argued the void models were only weakly constrained because the analyses fixed the Lemaître–Tolman–Bondi “bang time” function and studied structure formation with the wrong perturbation theory. That objection was answered on its own terms in 2012 by Bull, Clifton and Ferreira (Phys. Rev. D 85:024002), who let the bang time vary and found the extra freedom sufficient for individual observables but not for the supernovae, the small-angle CMB, the local Hubble rate and the kSZ effect together — and Clifton and Ferreira are the authors of “Living in a Void”, so the proposal was closed by its own proponents.

The second is not closed at all. The number-count dipole in radio and infrared source catalogues points roughly where the microwave-background dipole points but is about twice as large, which conflicts with reading the CMB dipole as pure motion. Secrest and colleagues put it at 4.9σ with 1.36 million CatWISE quasars in 2021 and at a 5.1σ joint significance across radio galaxies and quasars in 2022. A November 2025 reassessment of the same catalogue, with lognormal simulations of clustering, bias, selection and the survey mask, brings it down to 3.3–3.6σ and concludes that the anomaly is reduced but not accounted for by clustering or mask coupling. So the cosmological principle is under live empirical strain right now, and this page says so. But look at what the strain is: the dipole is in roughly the right direction and the wrong amplitude. It is evidence that our inferred velocity is not the whole story. It is not a compass pointing at the Earth, and it is worth noting that the same list treats the CMB dipole as fictitious elsewhere while this anomaly is defined by comparison with it.

Sixth, and decisively: suppose the principle failed tomorrow. What has actually been built and fitted is a Lemaître–Tolman–Bondi void. Ellis’s own specimen of an observationally viable one has the observer “within 10% of the central position” of an underdense region 160–250 h⁻¹ Mpc across — which puts it inside the window flagged in section four: small enough to slip under the kSZ bound, and left to the supernovae to exclude. And the LTB family is not the whole of the inhomogeneous literature, as the same review says in the next breath: “Actually you don’t need a void to explain the observations; more general models can do the job”, and the same section reports Bolejko and Sussman arguing that our improbable position is less improbable in a Szekeres solution than in an LTB one. That is a real caveat and it is conceded. But the extra freedom is freedom in the shape of the inhomogeneity, not a different prize at the end of it, and the tolerance below is what the prize looks like. The centring tolerance is set by the microwave dipole an off-centre observer would see, and it works out at tens of megaparsecs — the arithmetic is done at ARG-E09, where the same models are worked in the Hubble-tension vocabulary. Tens of megaparsecs is tens of millions of light-years. It contains the Local Group, the Virgo cluster’s outskirts and millions of galaxies, and it cannot tell the Earth apart from the Sun, from Alpha Centauri or from Andromeda. Inside that void the spacetime is still expanding, the age is still billions of years, and the Earth still orbits the Sun once a year. The maximum available prize is “our galaxy sits near the middle of a large hole”, and it is not the claim on the list.

Seventh, the two Copernican principles are not the same principle. The one Copernicus argued for is that the Earth orbits the Sun. The one Bondi named is that no place in the universe is privileged for the purpose of cosmological modelling. The second was retro-named after the first and the argument trades on the overlap. Heliocentrism is settled by things that have nothing to do with cosmological homogeneity: stellar parallax measured for about 1.47 billion sources by Gaia, annual aberration, radar ranging to Venus, and half a century of spacecraft navigating to targets on solutions that assume the Earth moves. Delete the cosmological principle in full — hand the geocentrist every void model in the literature — and every one of those measurements returns the same answer it returns today. The argument is aimed at a postulate whose failure the target does not depend on.

Eighth, the charge in the passage, answered on its own terms. The book’s claim is not merely that the principle is assumed; it is that the principle overrides evidence — that even a datum showing a stationary Earth would be re-read to preserve it. Test that against what the field did with the strongest anti-Copernican proposal ever seriously floated. When it turned out that a big enough local void could remove dark energy entirely, cosmologists did not suppress it: they wrote it up in Physical Review Letters, gave it the title “Living in a Void: Testing the Copernican Principle with Distant Supernovae”, specified the redshift range where it separates from ΛCDM, and then spent four years measuring it. Clarkson’s consistency test was built to be independent of curvature, dark energy, matter content and the theory of gravity precisely so that the answer could not be smuggled in through the assumptions. The evidence closed the option. The philosophy did not.

Verdict: refuted, on the source’s claim rather than the item’s. “It is an assumption” was true when written and is conceded here without reservation. “It is philosophy dressed as science, and it decides the data in advance” is the claim the book actually makes, and it is contradicted by a specific measurement: the microwave background’s small-angle power spectrum, which bounds radial inhomogeneity at the percent level per gigaparsec and excludes by two orders of magnitude the spherically symmetric Lemaître–Tolman–Bondi voids deep and wide enough to replace dark energy — the class the kSZ test was aimed at, and the class the argument’s own sources put forward. The foundation was checked, as the author they quote asked. It held.

The list says

Copernican principle assumption.

The source says

“This clearly shows that the Copernican Principle from which modern science creates its interpretations of the cosmological data is not scientific but philosophical. In other words, even if the empirical evidence shows Earth is not moving, the ever-present Copernican Principle requires that every piece of scientific data must be interpreted by assuming the earth is moving…”

A concession was re-used as a proof. This drift runs the opposite way to the usual one, and it is the ARG-R01 pattern rather than the ARG-A03 one: almost no wording moves, and the speech act does. In Galileo Was Wrong the proposition is a negative claim about somebody else’s foundation — you assumed this, you did not establish it, and you are using it to pre-read the data. That is ground-clearing. It is offered to remove an objection, and the chapter goes on to argue for a central Earth from other things entirely: microwave background alignments, quasar anisotropies, redshift structure. On the list the same four words stand alone as numbered proof 354, in a document whose organising claim is that each numbered line is a reason to believe the Earth does not move. A parity claim has become a proof, which is a promotion nothing in the source authorises.

The refutation above answers the source, not the item. The item’s bare proposition is true — the Copernican principle was an assumption, and the paper cited against it here opens by saying so — and beating it would beat nobody. What the book claims, and what gets answered, is the stronger thing: that the principle is philosophical rather than scientific and overrides evidence. One further note for anyone auditing this later. Rick DeLano, who produced the 2014 film with Sungenis, states the argument on his blog at a strength neither the book nor the list reaches — “Copernican Principle, 1532-2013, RIP” and “it has been scientifically falsified”. So the chain does not degrade monotonically. It hardens from book to film and then softens again into the four-word list entry, which is a useful corrective to any picture of compression as a one-way ratchet.

3. Surveyors' operations in the construction of railroads, tunnels, or canals are conducted without the slightest "allowance" being made for "curvature," although it is taught that this so-called allowance is absolutely necessary! This is a cutting proof that Earth is not a globe. 40. The Suez Canal, which joins the Red Sea with the Mediterranean, is about one hundred miles long; it forms a straight and level surface of water from one end to the other; and no "allowance" for any supposed "curvature" was made in its construction. It is a clear proof that the Earth is not a globe. 41. When astronomers assert that it is "necessary" to make "allowance for curvature" in canal construction, it is, of course, in order that, in their idea, a level cutting may be had for the water. How flagrantly, then, do they contradict themselves when they say that the curved surface of the Earth is a "true level!" What more can they want for a canal than a true level? Since they contradict themselves in such an elementary point as this, it is an evidence that the whole thing is a delusion, and we have a proof that the Earth is not a globe. 51. A "Standing Order" exists in the English Houses of Parliament that, in the cutting of canals, &c., the datum line employed shall be a "horizontal line, which shall be the same throughout the whole length of the work." Now, if the Earth were a globe, this "Order" could not be carried out: but, it is carried out: therefore, it is a proof that the Earth is not a globe.

— One Hundred Proofs that the Earth Is Not a Globe (1885), proofs 3, 40, 41 and 51 (index headings: “Surveyors' ‘allowance’”, “Suez Canal—100 miles—level”, “The ‘true level.’—a curve”, “The ‘Standing Order’”)

William Carpenter (1830–1896) emigrated from London to Baltimore and published One Hundred Proofs that the Earth Is Not a Globe there in 1885, dedicating it to the astronomer Richard Proctor as “A Challenge to the Johns Hopkins University.” It is the earliest numbered proof-list we have identified, and it works by condensing Samuel Rowbotham's sprawling Earth Not a Globe into a hundred one-line items, each closing with the same formula. Four of the hundred are about engineering practice, and they are the ancestor of the whole modern cluster.

Proof 41 is the one worth slowing down for. Carpenter has read the technical definition correctly — astronomers and surveyors do say the curved surface is the “true level” — and he treats that as a self-contradiction rather than as the answer to his own question. Rowbotham had quoted the source directly: in Zetetic Astronomy (1865), immediately before describing the Bedford Level experiment, he reproduces the Encyclopædia Britannica article “Levelling,” including the sentence “the curve of the Earth being the true level, and the tangent to it the apparent level,” and takes the article's “eight inches very nearly … at one mile's distance” as the arithmetic for the rest of the movement. The “8 inches per mile squared” figure enters flat-earth literature as a quotation from the encyclopaedia entry that defines level as curved.

The line downstream is documented. Wilbur Glenn Voliva reprinted Carpenter's hundred proofs under his own Zion, Illinois imprint in 1929, carrying the Victorian list into the twentieth century. Eric Dubay's 200 Proofs (2015) reproduces Carpenter's #3 as its #7 and Carpenter's #40 as its #8, with the Victorian phrasing modernised. Dubay names Carpenter elsewhere in the same list; this is descent and adaptation of a shared inheritance. The twelve modern restatements — pipelines, wind farms, Fresnel zones, aerial refuelling — are new examples fitted to the same four Victorian frames.

Surface (weak): “Engineers obviously do allow for curvature — the claim is just false.” This is the reflex answer and it is wrong. Walk onto a highway job, a pipeline right-of-way, a bridge deck or a wind farm and you will find no curvature term anywhere in the setting-out. The reflex answer loses the argument on contact with a working survey crew.

Deeper (true but incomplete): “The works are too short for it to matter.” True for most of the twelve modern examples — over a 400 m transmission-line span the departure from a tangent plane is about 3 mm against a catenary sag of several metres — but it does not cover the Suez Canal at 164 km, or a 57 km tunnel, or a 50 km microwave hop. And it concedes the wrong thing: it implies curvature is a small nuisance that gets rounded away, when in the cases that matter it is neither small nor rounded away.

Kernel: Rowbotham identified a real and specific technical fact, and stated it accurately. Early surveying did use “forward levelling” with an explicit allowance for convexity; the profession abandoned it. His own words: “the most eminent engineers abandoned the practice of the old ‘forward levelling’ and allowing for convexity; and adopted what is now called the ‘double sight’ or ‘back-and-fore sight’ method … for whatever degree of convexity existed, one ‘sight’ would compensate for the other.” That is not a caricature of surveying. It is what surveying does, and still does — the current NOAA/National Geodetic Survey levelling manual says curvature error “can be reduced by making the backsight and foresight distances in each setup nearly equal. If setups are thus balanced, within a tolerance, correction for curvature need not normally be made while leveling.” Carpenter and Rowbotham found the right invariant: there exists a single reproducible surface that every work in the world is built to, transferable over any distance by a procedure that never once mentions the shape of the Earth. That surface is real. Their inference — that a surface needing no curvature term must not be curved — is the strongest version of this argument, and the version worth answering.

Why it doesn’t save the claim.

The kernel points the other way, for the same reason Airy's null does. Cancellation is not absence. The back-and-fore-sight method cancels curvature because curvature error grows as the square of the sighting distance and is therefore equal on a balanced backsight and foresight, so it subtracts out. On a plane there would be no error to cancel, no reason to balance sights, and no tolerance on the imbalance. The NGS manual specifies a maximum permitted imbalance and gives tables of the combined refraction-and-curvature term to apply when a setup cannot be balanced — numbers computed from the Earth's radius. The technique cited as evidence for a plane exists only because the surface is curved.

The same structure holds for the surface itself. What the industry is built to is not a plane but an equipotential surface of the gravity field — the surface along which, in NGS's phrasing, “there is no change in potential energy.” Water finds it without being told to. A spirit bubble finds it. A plumb bob hangs perpendicular to it. It is closed, it wraps the planet, and on a rotating spheroid it is curved. “We built it level and it worked” is precisely what a curved equipotential predicts; it does not discriminate between the two hypotheses at all. The discriminating cases are the ones where the local trick stops working — and in every one, the radius reappears as an explicit number.

Start with the word doing all the work. “Level” in engineering does not mean “flat.” It means along a surface of constant gravitational potential — a surface you can slide along without gaining or losing energy. NGS defines the horizontal exactly that way, with the geoid being the equipotential surface that best fits the Earth. This is not a redefinition invented to rescue a theory; it is the operational definition every levelling instrument physically implements, because a bubble and a plumb line respond to gravity and nothing else. Water in a canal settles onto that surface by itself. So when Carpenter says the Suez Canal “forms a straight and level surface of water from one end to the other,” he is right about the observation, and the observation is neutral: on a spheroid a free water surface is necessarily an equipotential, and an equipotential is curved. Proof 41 sees the definition, calls it a contradiction, and stops one step short of the resolution.

Why no allowance is made — the honest answer, and it is in Rowbotham's own book. Modern differential levelling carries height forward by setting the instrument roughly midway between two rods and reading back then forward. Because curvature error scales as the square of sighting distance, equal distances produce equal errors that cancel on subtraction. NOAA Manual NOS NGS 3 states it plainly: the line of sight “would be parallel to the equipotential surface if the gravity field, at each setup, also defined a plane. This is not the case, however, since the gravity field defines a curved surface,” and therefore curvature error “can be reduced by making the backsight and foresight distances in each setup nearly equal.” The correction is not absent from the theory; it is engineered out of the procedure. Where it cannot be — unbalanced setups on steep ground, reciprocal setups across a river — the manual supplies numerical tables for it. Rowbotham described this changeover accurately in 1865 and read it as a retreat. It was a refinement.

Suez, specifically. The canal has no locks: it is a sea-level cut with seawater flowing freely through it, and as opened in 1869 it ran 164 km long and 8 m deep. Now ask what “making an allowance for curvature” would even mean there. The only geometric alternative to following the water surface is cutting a straight chord between the two ends — and a 164 km chord dips about 528 m below the water surface at its midpoint. By Carpenter's own arithmetic it is the same number: 8 inches × 102² = 6,936 ft of drop from the end tangent, of which the chord's mid-sagitta is a quarter, 1,734 ft ≈ 528 m. So the “allowance” being demanded would be an allowance to dig a half-kilometre-deep trench in the middle of a canal that is eight metres deep. Nobody proposed it, in 1869 or since, because the water would simply fill it. Not making that allowance is not a finding about the Earth's shape.

Where the correction is explicit, large and load-bearing. Four places, all checkable.

Radio and microwave links. This is the modern list item that fails hardest, because earth curvature is not merely modelled in RF path design, it is the dominant term. ITU-R Recommendation P.530 instructs designers to compute clearance over an effective Earth of radius k×R with median k = 4/3, then re-check against the k value exceeded 99.99% of the time, requiring roughly one full first-Fresnel-zone clearance over the highest obstacle. The ITU’s own handbook says why the trick exists: “To avoid having to calculate the path curvature when designing a link, the effective Earth radius can be used instead of the actual Earth radius. In this modified geometry … the path of the direct wave can be plotted as a straight line.” (R-HDB-54, §3.3, “Diffraction fading and path clearance”.) That is the same move as balanced backsights — curvature folded into the coordinate frame so the drawing looks straight. The magnitude: on a 50 km hop the mid-path earth bulge is about 37 m at k = 4/3, while the first Fresnel zone radius at 6 GHz is about 25 m. Design that link on a flat Earth and the bulge swallows the Fresnel zone and the link does not close.

Long tunnels. The Gotthard Base Tunnel is 57 km driven from multiple headings; final breakthrough came in with 8 cm transverse and 1 cm vertical deviation. That is a two-ended prediction test, not an assumption: independent surface networks are propagated down separate shafts and the two headings must meet. Agreement at the centimetre level over 57 km requires the surface control to be reduced on an ellipsoid, with deflection of the vertical accounted for, because a tunnel bored “level” in the plumb-line sense would be an arc, and one bored straight cannot also be plumb at both ends.

Geodesy and GNSS. Every GPS fix ever taken was computed in an Earth-centred, Earth-fixed frame on the WGS 84 ellipsoid, semi-major axis 6 378 137 m, flattening 1/298.257223563. The receiver reports a height above that ellipsoid and then applies a geoid model to give the orthometric height — the elevation a spirit level would agree with. The gap between those two numbers reaches tens of metres, and it exists precisely because “level” and “geometrically straight” are different surfaces.

The bridges — stated carefully. The Metropolitan Transportation Authority says of the Verrazzano-Narrows Bridge that its “693 foot high towers are 1 5/8 inches farther apart at their tops than at their bases because the 4,260 foot distance between them made it necessary to compensate for the earth's curvature.” The Humber Bridge's towers, over a 1,410 m span, “although vertical, are 1.4 inches (36 mm) farther apart at the top than the bottom due to the curvature of the Earth.” A defender is entitled to point out that these are design figures. They are — but note what generates them. Each tower is built plumb, aligned to its own local gravity vector, because a suspension tower must carry its load axially. Plumb lines 1.3 km apart are not parallel. The tower geometry is a consequence of plumbing, which is a direct physical operation, not an imported constant. (Akashi Kaikyō is often cited alongside these two; we found no operator or primary documentation of a curvature figure for it and have left it out.)

The rest of the twelve. They divide cleanly. Pipelines, highways, wind-farm layouts, power-line sag, short tunnels and aerial refuelling are all short-baseline: over a 400 m conductor span the departure from a tangent plane is 3 mm against metres of catenary sag; two 100 m turbine towers 1 km apart, each plumb, have tops about 16 mm further apart than their bases, well inside foundation tolerance; two tanker aircraft 30 m apart are on the same tangent plane to any precision that matters. Undersea cables are the opposite case: routes are planned and lengths costed along great-circle geodesics with slack allowance, and appear “straight” only on projections built to make geodesics look straight. Which brings the argument back to Carpenter's proof 51 and Parliament's “datum horizontal line.” A survey section is a plot of height against distance on a rectangular grid with enormous vertical exaggeration; the flat baseline is the datum, and the datum is the water surface. The flat line across the top of a cable-soundings chart is the geoid. It is drawn flat because it is the reference, not because it was measured to be a plane.

The list says

Engineering curvature corrections unnecessary. / Bridges no curvature allowance. / Microwave links flat. / Undersea cables straight. / Pipelines straight. / Railroads negligible curvature. / Highways horizon. / Wind farm layout flat. / Tunnel breakthroughs align. / Power line sag flat. / Microwave Fresnel zone flat. / Aerial refueling flat line.

The source says

“Surveyors’ operations in the construction of railroads, tunnels, or canals are conducted without the slightest ‘allowance’ being made for ‘curvature,’ although it is taught that this so-called allowance is absolutely necessary! This is a cutting proof that Earth is not a globe.” (proof 3) — and at proof 26 Carpenter does the arithmetic himself: the Atlantic “would … curvate downwards, in the 1665 miles across the Atlantic to Newfoundland, according to the astronomers’ own tables, more than three-hundred miles.”

A narrow case was generalised.

On strength there is no drift, and it should be said plainly: Carpenter is blunter than the list. He asserts flatly that no allowance is made and closes each item with it is a proof that the Earth is not a globe; the list’s “Railroads negligible curvature” is the milder of the two, conceding a curvature Carpenter denied outright. What has moved is scope. His engineering proofs cover surveyors’ cuttings, canals, railways, tunnels and the Atlantic cable soundings — the works of 1885 — and within them the observation is correct: the correction genuinely is not applied. The list carries the same frame into microwave path design, Fresnel-zone clearance, wind-farm layout, power-line sag, aerial refuelling and 57 km tunnel breakthroughs. None of those appear in Carpenter, and none in Dubay’s intermediate 200 Proofs, whose proof 7 adds only bridges. They are the point at which the claim stops being a mis-inference from a true observation and becomes simply false, because in an RF link budget the Earth’s radius is the dominant term rather than a neglected one. Our refutation is aimed correctly: it answers Carpenter’s own case first — the definition of level, Suez, the balanced back-and-fore sight, Parliament’s datum line — before it reaches the modern items.

Sigma Octantis is claimed to be a Southern central pole star similar to Polaris, around which the Southern hemisphere stars all rotate … Unlike Polaris, however, Sigma Octantis can NOT be seen simultaneously from every point along the same latitude, it is NOT central but allegedly 1 degree off-center, it is NOT motionless.

— 200 Proofs Earth Is Not a Spinning Ball (2015), proof 101 of 200. The 2015 free PDF is numbered by proof, not by page; read in the full-text reproduction posted 31 July 2018 at rexynotes.wordpress.com and cross-checked against the SlideShare scan of the PDF. Not checked against the 2018 print edition.

The sentence is trimmed for length; the clause that follows in the original adds that the star cannot be seen at all using publicly available telescopes. So there are four claims here, not one, and they are of two kinds. Two are checkable statements about a star’s geometry — it is not central, it is not motionless — and both are correct. Two are claims about what can be observed and by whom, and those are the ones this treatment tests.

Proof 101 matters more than its neighbours because of what proof 107 commits the model to. There Dubay reaches for the ring magnet found in loudspeakers, with a central north pole and the opposite south pole being all points along the outer circumference. On that model “south” is not a place; it is a direction that fans outward from the rim in every azimuth at once. A single southern point with the sky wheeling around it is therefore not an awkward detail for the model to absorb, it is the thing the model forbids — which is why proof 101 goes after whether the star can be observed at all rather than trying to fit it in.

Two of the neighbouring proofs are older than 2015. Proof 99, that Polaris is visible more than 20 degrees south of the equator, restates William Carpenter’s One Hundred Proofs that the Earth is Not a Globe (5th ed., 1885), proof 71: “The astronomers’ theory of a globular Earth necessitates the conclusion that, if we travel south of the equator, to see the North Star is an impossibility. Yet it is well known this star has been seen by navigators when they have been more than 20 degrees south of the equator.” The Polaris-is-fixed strand has an 1885 ancestor too, in Carpenter’s proof 80, which reasons from seeing the North Star through the very same corner of the very same pane of glass all the year round. What is new in 2015 is the naming of Sigma Octantis. The argument around it is not. Rowbotham’s Zetetic Astronomy: Earth Not a Globe (3rd ed., 1881) gives the southern sky a section of its own, “Motion of Stars North and South”, which opens on this exact claim — “IT has often been urged that the earth must be a globe, because the stars in the southern ‘hemisphere’ move round a south polar star” — and answers it with a denial: the southern region “is not central, but circumferential; and therefore there is no southern pole, no south pole star, and no southern circumpolar constellations.” Proof 101’s simultaneity objection is there in the same comparative form — “The Southern Cross is not at all times visible from every point of the southern hemisphere, as the ‘Great Bear’ is from every point in the northern” — and so is the perspective reply, which Rowbotham stages by asking the reader to stand with his back to the north on Arthur’s Seat, near Edinburgh, and watch the stars in his zenith. Quoted from the 1881 third edition; that section is not located in the 1865 first book edition as posted at Project Gutenberg (#69892), which we searched for circumpolar, pole star, south polar and Southern Cross. The 1865 text does carry the other strand, though — the north polar star seen “as far even as the tropic of Capricorn”, on a report of Captain Wilkins in the Times of 13 May 1862 — twenty years before Carpenter’s proof 71. Both halves of this cluster are inherited, and of the texts traced here the southern half is the earlier: 1881 against Carpenter’s fifth edition of 1885, whose own earlier editions we could not obtain.

The positive claim is stated most plainly outside the book. Dubay’s video Flat Earth Star Trails Explained, published 30 November 2018 and archived at the Internet Archive, holds that star trails in both hemispheres turn about Polaris in the same direction, east to west, rather than in opposite directions. That is a falsifiable statement, it is the one the geometry below answers, and none of it survives into the seven items the list carries.

SURFACE (weak — do not use). “You cannot see northern constellations from the southern hemisphere, so his numbers are nonsense.” Wrong, and it loses the exchange in one move. Proofs 103 to 105 list constellations with the latitude bands they can be seen from — Vulpecula from 90 degrees north to 55 degrees south, Virgo from 80 north to 80 south, Orion from about 85 north to about 75 south — and those bands are arithmetically right. They are also astronomy’s own numbers, printed in ordinary constellation guides, and they follow from one rule: a star of declination δ stays below the horizon for a northern observer at latitude φ only when δ < φ − 90°. A constellation straddling the celestial equator is therefore visible from almost pole to pole, on a globe, exactly as he says.

DEEPER. The northern observation at the heart of the cluster is correct and it is a real constraint. From any site in the northern hemisphere the stars trace concentric circles about one point, the circles close, and the constellations show no measurable distortion from one site to the next. That tells you two things before any model is chosen: the sky’s daily motion is a single rigid rotation about one axis, and the stars are so far away that no baseline on the ground changes their arrangement. Proof 102 adds a fair point of its own — that an object sinking towards the horizon as you walk away from it is generic, and not by itself a demonstration of curvature.

KERNEL. Proof 101 is the best thing in the cluster, because it identifies the decisive datum instead of walking past it. Of 200 proofs it is the one that names the observation the model has to defeat, and it goes at it head-on. And two of its four objections are simply true. Sigma Octantis is not on the pole: it sits slightly more than one degree from it, and precession is carrying the pole away from it. It is not motionless: it circles the pole once a sidereal day, and it is a Delta Scuti variable whose brightness shifts by about 0.03 magnitudes every 2.33 hours. Anyone who answers proof 101 by calling Sigma Octantis a fixed southern Polaris has conceded the point to him. Concede both facts first.

Why it doesn’t save the claim.

Because both true facts are true of Polaris, in the same way and to within a factor of two. Polaris was 0.66 degrees (39.6 arcminutes) from the north celestial pole in 2018, so it traces a circle about 1.3 degrees across every day; Sigma Octantis is about 1.06 degrees out and traces a circle about 2.1 degrees across. Polaris is a classical Cepheid and varies in brightness too. Precession is moving the north pole as well — towards Polaris until soon after 2100, when it comes within about 0.45 degrees, and away from it afterwards. So the test proof 101 applies to the southern pole star — off centre, and moving — retires the northern one on which proof 98 rests. A criterion that deletes your own benchmark is not a criterion.

And the southern rotation centre is not located by Sigma Octantis in the first place. It is fixed by the trails of every star in the southern cap, thousands of them, whose common centre can be measured on a photograph in which no star sits at the centre at all. Sigma Octantis is a convenience for a navigator, not the evidence. Removing it would leave the second pole exactly where it is.

Start by giving away the part that is right. The latitude bands in proofs 103 to 105 are correct, and the rule that generates them is the globe’s. Circumpolarity and invisibility are set by declination against latitude: a star never sets when φ + δ > +90° in the north or φ + δ < −90° in the south, and never rises when the same sums run the other way. Anyone who answers this cluster by insisting that each hemisphere has its own private sky is wrong, and will be shown to be wrong with the objector’s own star charts.

1. What the southern sky does. The measurements are these. The whole sky turns once per sidereal day, 23 h 56 m 04 s, at 15.04 degrees per hour, in both hemispheres. From a northern site the turn is anticlockwise about a point due north at an altitude equal to the observer’s latitude. From a southern site it is clockwise about a point due south, again at an altitude equal to the observer’s latitude. At Cape Town, Santiago and Sydney — all near 34 degrees south — that centre stands about 34 degrees above the southern horizon, and a cap of sky 34 degrees in radius around it never sets: 8.6 per cent of the celestial sphere, permanently up, with an equal 8.6 per cent around the north celestial pole permanently out of reach. The Southern Cross lies inside that cap; it is circumpolar south of 34 degrees south, and its brightest star, Acrux, sits at declination about −63 degrees. Sigma Octantis, about 1.06 degrees from the southern pole, is circumpolar for any observer more than about 1.1 degrees south of the equator — that is, for essentially the whole southern hemisphere.

2. Proof 101, clause by clause.

(a) “can NOT be seen simultaneously from every point along the same latitude.” Being circumpolar from about 1.1 degrees south onward, Sigma Octantis is above the horizon at every longitude of a southern parallel at the same instant. One honest qualification, which the list’s defenders are entitled to press: half those longitudes are in daylight at any given moment. Santiago (70.7°W), Cape Town (18.4°E) and Sydney (151.2°E) span 222 degrees of longitude and cannot all be dark together. But any two of them that are dark together find the star in the same place, due south, at an altitude matching their latitude, with the same stars wheeling around it — and the daylight limit applies identically to Polaris on the northern parallel, which proof 101 treats as the standard. Sunlight is not a fact about geometry.

(b) “NOT central but allegedly 1 degree off-center.” Correct, and conceded: slightly more than one degree. Polaris was 0.66 degrees off in 2018. Nobody has claimed either star sits on its pole; the claim is that each is the nearest star of its kind to one.

(c) “NOT motionless.” Correct, and conceded: it circles the pole once a sidereal day, in a circle about 2.1 degrees across, and its brightness varies by 0.03 magnitudes every 2.33 hours. Polaris circles in 1.3 degrees and varies too.

(d) cannot be seen at all using publicly available telescopes. This is the one that decides the exchange, and it is checkable by the reader. Sigma Octantis has apparent magnitude 5.42, about 294 light years away, spectral class F0 IV. The naked-eye limit under a dark sky is around magnitude 6.5, so the star is a naked-eye object for a southern observer away from town, and an easy one in any binocular. It is also on the flag of Brazil, adopted 19 November 1889, where it stands for the Federal District — and the official rationale for putting it there is that it is small, but all the other stars turn around it. That is a documented southern-hemisphere statement of the second pole, published 126 years before the 2015 PDF, by a government with no stake in this dispute.

3. What a single-plane model predicts instead. Take the standard flat map: the azimuthal-equidistant projection centred on the north pole, on which ground distance from the centre runs about 111 km per degree of colatitude, with Polaris hanging at some height h over the centre and the stars carried round on a dome that turns about the vertical axis through it. Five consequences follow, and each one is measured to be false.

(i) One centre of rotation, for everybody. A single axis produces exactly one apparent centre per observer, and it lies in the direction of that axis — northward, everywhere on the disc. Southern observers measure a second centre, in the opposite direction, with its own circumpolar cap.

(ii) One sense of rotation. Every observer on the disc sees the sky turn the same way about that one centre. Southern trails close in the opposite handedness from northern ones, at the same rate, on the same nights.

(iii) Polaris at the wrong altitude nearly everywhere. On the disc the altitude of Polaris is arctan(h/r), with r the ground distance from the centre. Fit h so that the altitude comes out right at 45 degrees north and you need h = 5,000 km. That same height then puts Polaris 36.9 degrees up at 30 degrees north, where it is measured at 30; 29.4 degrees up at 10 degrees north, where it is measured at 10; and 26.6 degrees up at the equator, where it is on the horizon. The 10-degree error is 19.4 degrees, about thirty-nine full-moon widths, against a sextant that reads to an arcminute. Fitting each latitude separately gives a different height every time — about 5,800 km at 60 north, 3,850 km at 30 north, 1,570 km at 10 north, and zero at the equator. There is no single height, which is the arithmetic statement of the fact that altitude equals latitude on a sphere and cannot on a plane.

(iv) The southern pole would have to be in three directions at once. On that map Santiago, Cape Town and Sydney all sit about 13,800 km from the centre, on bearings 89, 133 and 138 degrees apart. Each observer’s due south is the direction straight away from the centre, so their three southern horizons face outward along three widely divergent lines. Sigma Octantis is 294 light years away; the three sight-lines to it are parallel to about a millionth of an arcsecond. On the disc they diverge by up to 138 degrees. One object cannot occupy three directions at once, which is the whole of the argument, stated without any appeal to authority.

(v) The rim is not a point. Dubay’s own proof 107 makes the south the entire outer circumference. On that map the rim is a circle roughly 126,000 km around. The southern sky says the southern axis is a point: from every southern longitude the trails share one centre, and its altitude tracks the observer’s latitude one for one from the tropics to the Antarctic coast.

4. The vanishing-point reply, answered before it is made. The standard response at this stage is proof 102’s: perspective on a plane. Distant things converge; a southern observer looking outward across the disc is seeing the far stars foreshortened into an apparent centre, the way rails appear to meet. Three measurements close this off. First, a vanishing point on a plane lies on the horizon, at altitude zero, by construction; the southern centre stands at an altitude equal to the observer’s latitude — about 12 degrees at Darwin, 34 at Cape Town, 53 at Punta Arenas — and moves up as you travel south. Second, a vanishing point is fixed with respect to the observer and does not turn; the southern centre has stars circling it at 15.04 degrees per hour, closing into complete circles in one sidereal day, which is a rotation and not a projection. Third, perspective foreshortens: an off-centre observer under a dome of finite height would see the constellations near the far rim compressed, and their angular sizes would change with his distance from the centre. The Southern Cross subtends the same angle from Chile as from Australia.

5. The two items that make a claim about time. “Polaris constant alignment over millennia” and “Zodiacal alignment constant” are the only items in the cluster that assert something about history, and precession settles both. The axis completes a circuit in roughly 25,800 years, moving the pole about one degree every 72 years. In classical antiquity the north celestial pole stood about 10 degrees from Polaris and about as far from Kochab; Pytheas, around 320 BC, described the pole as devoid of stars. Polaris will be closest to it soon after 2100, at about 0.45 degrees, and will then recede for the next thirteen millennia. The equinox has slid about a full zodiacal sign since Hipparchus measured the drift in the second century BC. These are not fine corrections rescuing a model; the measurement of this drift is where the study of the sky’s long-term motion began.

What is actually in dispute. Three of the seven items — perfect circular star trails around Polaris, perfect star circles, Polaris fixed — report things that are true and that nobody contests. Northern trails do close into near-perfect circles about a point 0.66 degrees from Polaris. Refuting an observation is not on offer and is not needed. What fails is the inference, and it fails on a second observation the same instrument makes from the other half of the world: point the camera south from Paranal and the trails close about a different centre, turning the other way. Two poles, one sky, one rotation. A disc with Polaris over its middle has one axis and can deliver one centre, which is why the argument arrives at the southern pole star with a denial rather than an explanation. Verdict: refuted — and refuted by an observation any reader with a camera, a tripod and a flight to the southern hemisphere can make without trusting anyone.

The list says

Constant constellations. / Constellation stability.

The source says

“There are several constellations which can be seen from far greater distances over the face of the Earth than should be possible if the world were a rotating, revolving, wobbling ball.” … “The constellation Vulpecula can be seen from 90 degrees North latitude, all the way to 55 degrees South latitude.” (proofs 103–104)

A narrow case was generalised. Dubay’s claim is spatial and specific: named constellations, each with the band of latitudes it can be seen from, offered as bands too wide for a globe. The list renders this as “Constant constellations” and “Constellation stability” — two-word fragments that a reader will take as a claim about time, that the constellations do not change. Read that way they are false for reasons Dubay never argued (precession, proper motion); read the source’s way they are a claim about visibility geometry with checkable numbers, and the numbers are right. The refutation above answers the source’s version — which is why it opens by conceding the latitude bands rather than disputing them. Item 145, “Zodiacal alignment constant”, is the one this pass could place least confidently against any proof in the text read.

A second movement, and it runs the other way from every drift on this page. The seven items report only the northern sky: perfect circular star trails, Polaris fixed, constellations constant. Nothing south of the equator survives into them. But the southern sky is where the source stakes its case — proofs 99 to 102 on Polaris below the equator, the Southern Cross, and Sigma Octantis, and the 2018 video Flat Earth Star Trails Explained, which holds that trails in both hemispheres turn about Polaris in the same direction. Those claims are bold and checkable. The list kept the half that is uncontested — northern trails really do close into circles about Polaris — and dropped the half that fails. So the fragment is not overstated here; it is disarmed, and a reader meeting item 16 as a proof meets something no astronomer disputes. The seven drift types have no value for a weakening, so it is recorded in words: on this cluster the chain degraded towards safety rather than towards certainty, which is a counter-example to the pattern the rest of this page has found, and worth more than the label.

The square of 24, multiplied by 8 inches, shows a declination of 384 feet. The altitude of the lights in Poolbeg Lighthouse is 68 feet; and of the red light on Holyhead Pier, 44 feet. Hence, if the earth were a globe, the former would always be 316 feet and the latter 340 feet below the horizon … The line of sight H, S, would be a tangent touching the horizon at H, and passing more than 300 feet over the top of each lighthouse. [the table] The Egerö Light, on west point of Island, south coast of Norway, is fitted up with the first order of the dioptric lights, is visible 28 statute miles, and the altitude above high water is 154 feet. The Dunkerque Light, on the south coast of France, is 194 feet high, and is visible 28 statute miles. The Cordonan Light, on the River Gironde, west coast of France, is visible 31 statute miles, and its altitude is 207 feet. The Light at Madras, on the Esplanade, is 132 feet high, and is visible 28 statute miles. The Port Nicholson Light, in New Zealand (erected in 1859), is visible 35 statute miles, the altitude being 420 feet above high water. The Light on Cape Bonavista, Newfoundland, is 150 feet above high water, and is visible 35 statute miles. Many instances could be given of lights being visible at sea for distances which would be utterly impossible upon a globular surface of 25,000 miles in circumference … The only modification which can be made in the above calculations is the allowance for refraction, which is generally considered by surveyors to amount to one-twelfth the altitude of the object observed. If we make this allowance, it will reduce the various quotients so little that the whole will be substantially the same. [and, quoted by Rowbotham himself, from the Encyclopædia Britannica, article “Levelling”] We suppose the visual ray to be a straight line, whereas on account of the unequal densities of the air at different distances from the earth, the rays of light are incurvated by refraction. The effect of this is to lessen the difference between the true and apparent levels, but in such an extremely variable and uncertain manner that if any constant or fixed allowance is made for it in formula or tables, it will often lead to a greater error than what it was intended to obviate. For though the refraction may at a mean compensate for about one-seventh of the curvature of the earth, it sometimes exceeds one-fifth, and at other times does not amount to one-fifteenth. We have, therefore, made no allowance for refraction in the foregone formulæ.

— Earth Not a Globe (1865), 3rd ed., rev. and enl. (London: Day, 1881), ch. II, pp. 28–35 — the lighthouse table, the refraction allowance, and the Encyclopædia Britannica “Levelling” extract. Not the 1849 pamphlet: see the gloss

Our own attribution is wrong, and the text dates itself. This cluster is credited to Rowbotham's 1849 Zetetic Astronomy — a sixteen-page pamphlet. The passage above names “The Port Nicholson Light, in New Zealand (erected in 1859).” A text that cites an 1859 lighthouse cannot appear in an 1849 publication. Nor is it in the 1865 first book edition so far as we can search it: none of Egerö, Dunkerque, Cordonan, Madras, Port Nicholson, Bonavista or Poolbeg occurs in the Gutenberg text of 1865 through p. 83, which is past the point where the equivalent material sits in the later edition. The earliest text we can document is the third edition of 1881. The lighthouse argument is therefore a mature-Rowbotham addition of the 1860s–70s, not a founding claim, and our record has it about thirty years early. This is a correction to us, not only to them.

What Rowbotham actually does, which is more careful than the item that descends from it. He does not simply multiply eight inches by the square of the whole distance. He allows for the observer's own height and subtracts the resulting horizon distance first — “Allowing 16 feet for the altitude of the observer … 5 miles must be taken from the 30 miles, as the distance of the horizon” — and only then squares the remainder. That is the correct geometric procedure, and 16 ft does give a horizon at about 4.9 statute miles. That worked example is the Spurn Point light; in the lighthouse table itself he uses the ten-foot eye he calls the standard and takes off four miles. He converts the light lists' nautical miles into statute miles and says so each time. And he concedes refraction outright, quantifies it, applies it, and shows his working on Cape Bonavista: 150 ft divided by 12 gives “13 feet as the amount to be deducted from 491 feet, making instead 478 feet.” Every one of those qualifications is gone from the list item, which reads in full: “Lighthouse visibility.”

The Britannica extract is the most important thing on the page, and it is his. Rowbotham reprints it, in both editions, immediately around this material. It tells him the mechanism (“the unequal densities of the air at different distances from the earth”), the coefficient (“at a mean … about one-seventh of the curvature”, which is k ≈ 0.14, the 7/6 effective-Earth-radius rule still in the surveying textbooks), and the variability (“sometimes exceeds one-fifth, and at other times does not amount to one-fifteenth”). It also explains why the encyclopaedia's own tables omit refraction — not because it is absent but because a fixed allowance would mislead. Rowbotham reads that as licence to set refraction aside; the passage says the opposite. Elsewhere he goes further and denies the mechanism outright: “Refraction can only exist when the line of sight passes from one medium into another of different density.” The article he had just typeset says the line of sight does pass continuously through air of different densities. Compare ARG-B06, where the same book supplies the “8 inches per mile” figure out of the same encyclopaedia entry.

Carpenter's contribution is American and it is first-person. William Carpenter emigrated to Baltimore and localised the argument. Four of his hundred proofs are in this cluster, and three of the four are his own: the Cape Hatteras figure and the two Chesapeake cases are not located in the third edition of 1881. The fourth he had from Rowbotham. Proof 5: “The lights which are exhibited in lighthouses are seen by navigators at distances at which, according to the scale of the supposed ‘curvature’ given by astronomers, they ought to be many hundreds of feet, in some cases, down below the line of sight! For instance: the light at Cape Hatteras is seen at such a distance (40 miles) that, according to theory, it ought to be nine-hundred feet higher above the level of the sea than it absolutely is, in order to be visible!” Proof 36: “If we take a journey down the Chesapeake Bay, by night, we shall see the ‘light’ exhibited at Sharpe's Island for an hour before the steamer gets to it. We may take up a position on the deck so that the rail of the vessel's side will be in a line with the ‘light’ … and we shall find that in the whole journey the light will not vary in the slightest degree in its apparent elevation. But, say that a distance of thirteen miles has been traversed, the astronomers' theory of ‘curvature’ demands a difference (one way or the other!) … of 112 feet 8 inches!” Proof 7 is the daytime version, with the far shore's “tall trees, towering up, in perspective, over the heads of the ‘hull-down’ ships”. Proof 63 is the telescope claim: “a good telescope will restore to our view this portion of the vessel. Now, since telescopes are not made to enable people to see through a ‘hill of water’ …” That one is not Carpenter's invention, and the difference between the two versions is the whole of it. Rowbotham had written it in ch. XIV with a condition attached — “even on the sea, when the water is very calm, if a vessel is observed until it is just ‘hull down,’ a powerful telescope turned upon it will restore the hull to sight” — and he states the contrary case just as plainly, because on a moving sea the waves are “magnified and rendered more obstructive by the very instrument” and “a telescope fails to restore it, however powerful it may be.” Carpenter's proof 63 asserts the effect with the condition struck out, and the Flat Earth Society wiki still repeats it in that form. The hedge dropped at that step is the one that had made the claim survivable: Rowbotham's version is hard to test, because it holds only on a glassy sea and only at the moment the hull is just lost; Carpenter's is testable on any afternoon, and is tested in section 8 below.

The descent, corrected. Eric Dubay's 200 Proofs (2015) reprints Rowbotham's table almost intact as proofs 82–86 (Port Nicholson, Egerö, Madras, Cordonan, Cape Bonavista) and the Poolbeg/Holyhead case as proof 93; Carpenter's Chesapeake night journey is proof 96, quoted with attribution. But Cape Hatteras is not in 200 Proofs at all, and proof 89 — which our basis line identifies with it — is the Cape L'Agulhas light, a case with no Victorian ancestor. Carpenter's proof 5 survives instead in Dubay's 2018 post “Enlightenment from Lighthouses”, quoted verbatim and credited to Carpenter. So the lighthouse spine of this cluster is Rowbotham's, not Carpenter's; the two Chesapeake cases are Carpenter's, not Rowbotham's; and one of the two links our own record asserts does not exist.

SURFACE (weak — and a knowledgeable defender will punish it). “Refraction explains it, next.” Said bare, this is the single worst move available, for three reasons. It sounds like a parameter tuned after the fact to whatever the observation needs; it is not even the right answer for most of the lighthouse items, which fail for a different reason entirely; and it concedes by implication that the naive calculation was sound and only needed a nudge. It was not sound. Anyone who opens with “refraction” has agreed to defend the weakest version of their own case.

DEEPER (true, incomplete). “The observations are real; your calculator is wrong.” Correct, and it has to be said out loud: the “8 inches per mile squared” figure that dominates this debate is wrong as it is almost always used. It computes the drop of the sphere below the plane tangent at the observer — “the drop x is not what is hidden by the curvature of the earth!” — and it is only a hidden-height figure when the observer's eye is exactly at the surface. Raise the eye and the correct quantity is the drop over the distance beyond your own horizon, which is much smaller. Popular online curvature calculators routinely report the first number, default to zero refraction, and consequently predict that targets are hidden which are in fact seen every clear day. Flat-earthers who run those calculators are getting an honest answer to the wrong sum. But this stated alone still invites the reply that we are just moving the goalposts to wherever the target happens to be.

KERNEL — refraction is not a rescue, it is a standing correction with a published coefficient, and the source printed the coefficient himself. The mean value of terrestrial refraction is not a free parameter. Andrew Young states the standard case flatly: the radius of curvature of a near-horizontal ray is about seven times the Earth's, so the geometry is handled by an effective radius R′ = R × 7/6. Surveyors use k ≈ 0.13–0.14 and radio-path engineering uses k = 4/3 for the same reason and by the same construction. It is written into navigation: Bowditch's geographic-range figures are 1.17√h nautical miles rather than the geometric 1.06√h, so about ten per cent of refraction is inside the table before a mariner touches it, and a 150-foot light is tabulated at 14.3 nm rather than 13.0. And the Encyclopædia Britannica article Rowbotham reprints in his own chapter gives the same number in 1860s language — refraction “at a mean compensate[s] for about one-seventh of the curvature of the earth.” One-seventh is the 7/6 rule. The movement's founding text contains, verbatim, the correction its own argument omits, together with the mechanism (“the unequal densities of the air at different distances from the earth”). So the honest position, which we should state before we say anything else: many of these sightings are real, are correctly reported, and are exactly what a globe with an atmosphere predicts. That is why the verdict is MISLEADING and not REFUTED. The observations are usually sound. It is the inference that fails.

And one more piece of ground that belongs to them. Rowbotham's account of the ship's hull is not “perspective” hand-waving; he proposes a physical mechanism — “the natural result of the law of perspective operating on a plane surface, but modified by the mobility of the water” — and at the eye heights he used, two or three feet above a choppy surface, waves really do occlude low targets. He is right that a rough sea hides things a calm sea shows. He simply never asks what happens to that mechanism when the eye is 30 metres up and the sea is glass.

Why it doesn’t save the claim.

Because the same article that supplies the coefficient supplies the discriminator, in the same sentence. Refraction “sometimes exceeds one-fifth, and at other times does not amount to one-fifteenth” — a factor of three either side of the mean, stated in the 1860s, and confirmed by modern measurement: over Rainy Lake the refraction coefficient was measured running from about k = 0.19 to k = 0.41 across observing sessions, lifting a target 5.4 km away by about half a metre and one 9.5 km away by about a metre. A flat plane predicts visibility limited only by luminous intensity and atmospheric extinction: monotonic, and independent of the temperature profile. What is actually observed is a target lifted into view one morning and hidden the next, over the same water, at the same distance, from the same eye. Variability is not a defect in the globe answer. It is the observable the flat model cannot generate at all.

Because half these numbers were never horizon distances. The lighthouse tables are catalogue figures, not sightings. A light list's stated range is a nominal or luminous range — how far the lamp is bright enough to be seen in stated visibility — and the mariner is instructed to compute the geographic range separately and take the lesser. Cape Agulhas, Dubay's proof 89: the tower is 27 m (89 ft) and the focal plane 31 m (102 ft) above high water, not the “33 feet high, 238 feet above sea level” the proof asserts, and its charted 30 nautical miles is a luminous range against a geographic range of about 16 nm from a fifteen-foot eye. Nobody has ever stood on a deck and seen that light at fifty miles. The argument is not fighting refraction here; it is comparing a candela figure with a geometry figure and reporting the difference as curvature.

Because the tables are height-dependent in exactly the way a sphere requires and a plane forbids. Every geographic-range table in every light list is a function of two heights — the light's and the observer's eye — added together as square roots. On a plane the observer's eye height would have no bearing whatever on whether a distant light is visible; only its brightness and the air's clarity would. Eye-height dependence is the fingerprint, it is tabulated, and mariners use it every night.

Because waves do not explain the observation Rowbotham himself reported. He records the Eddystone light seen from five feet above the water in calm weather, while in rough weather at the same place the vane one hundred feet above the foundation was “entirely out of sight.” No sea state occludes a hundred feet. And the Nab light-ship from Victoria Pier: from 32 inches up, “when it was very calm, the greater part of the hull … was, through a good telescope, plainly visible. But on other occasions, when the water was much disturbed, no portion of the hull could be seen.” He needs long-range visibility to be a stable property of a flat surface. His own field notes record it changing with the day.

1. Concede the arithmetic first, because the arithmetic really is wrong. The figure that runs this whole debate — 8 inches per mile squared — is a correct number for the wrong quantity. Taking R = 6,371 km, one statute mile of horizontal run puts the surface 0.203 m, or 8.00 inches, below the plane tangent at the observer. That is drop from a tangent plane. It is not how much of a distant object is hidden, and it becomes that only if the observer's eye is on the surface. Lift the eye and the hidden height is the drop measured over the distance past your own horizon: with the eye at height h feet the horizon lies about 1.22√h statute miles away, and only the excess beyond that is squared. Rowbotham knew this and did it correctly — four miles off each distance in the lighthouse table, five in the Spurn Point example where he allows sixteen feet of eye height, and only then square. The modern meme usually does not, and neither do most of the calculators circulating with it. Before refraction is mentioned at all, the popular version of the globe prediction is an overstatement, and the people running it against real sightings are getting a true answer to a false sum.

2. Then add the correction that is not optional and not ad hoc. Air is denser near the ground, so a near-horizontal ray bends downward continuously; its radius of curvature is roughly seven times the Earth's, and the standard device is to keep the ray straight and inflate the planet: an effective radius R′ = R/(1 − k), with the geodetic standard k ≈ 0.13–0.14 giving the familiar 7/6 rule. Radio-path engineering uses the same construction with k = 4/3, which is why ITU-R P.530 has designers draw straight lines over a 4/3-radius Earth. Under mean refraction the 8-inch figure becomes about 6.9 inches per mile squared. This is not a correction invented for flat-earthers; it predates the dispute, it is used by people with no stake in it, and it is derived from a measurable temperature profile rather than fitted to the observation it is meant to save. See ARG-B07, which is the cluster asserting that refraction is invoked ad hoc.

3. The source contains the correction. This is the finding. Rowbotham reprints the Encyclopædia Britannica article “Levelling” in the same chapter as his lighthouse table. It states the mechanism — “on account of the unequal densities of the air at different distances from the earth, the rays of light are incurvated by refraction” — and then the number: “the refraction may at a mean compensate for about one-seventh of the curvature of the earth.” One-seventh is k = 0.143, which is the 7/6 rule to three digits. The Victorian encyclopaedia and the modern surveying textbook agree, and the agreement is printed inside the book being cited against them. Rowbotham's own substitute — “one-twelfth the altitude of the object observed” — is the wrong quantity in the wrong place: refraction scales with the square of the distance, not with the height of the target, so on Cape Bonavista he deducts 13 feet from a curvature term of about 641 feet. The 491 he prints there is not the curvature term itself; it is that term less the light's own 150 feet, which is how every row of his table is built. The Poolbeg case quoted above shows the construction in his own words: a declination of 384 feet, less lights of 68 and 44 feet, gives the 316 and 340 he reports. Egerö is the check on the table proper — 28 statute miles less the four-mile horizon of the ten-foot eye he calls the standard is 24; 24² × 8 inches is 384 feet; 384 less the light's 154 feet is the 230 feet of depression that comes down intact to Dubay's proof 83. Bonavista runs the same way: 35 miles less four is 31, and 31² × 8 inches is 641 feet. So the encyclopaedia's own mean would take about a seventh of 641, some 90 feet, and a strong-refraction day nearer 130. That correction does not rescue the sighting either — we will come to why — but the allowance he made was about seven times too small and applied to the wrong variable.

4. Navigation has all of this built in, and has for two centuries. Bowditch tabulates a light's geographic range as the sum of two horizon distances, the observer's and the light's, at 1.17√h nautical miles for h in feet: 150 ft of light gives 14.3 nm, a 5-foot eye adds 2.6, a 30-foot eye 6.4, a 70-foot eye 9.8. The purely geometric coefficient is 1.06, so roughly a tenth of the tabulated distance is refraction, standing, unremarked, in a table printed for people who need to make landfall. Bowditch also states the limit of the method in the same breath: “Abnormal refraction patterns might change this range; therefore, one cannot exactly predict the range at which a light will be seen.” That is a sphere-plus-atmosphere model telling you honestly which part of its prediction is soft.

5. But most of the lighthouse items never reach the refraction question, because the numbers are not sightings. This is the part of the cluster that fails first, and it fails on definitions. A light list quotes a nominal range — the distance at which the lamp would be seen if meteorological visibility were 10 nautical miles — or a luminous range, its equivalent in the actual weather. Both are functions of candela and air clarity and have nothing to do with the horizon. The geographic range is the horizon-limited figure. The mariner computes both and uses the smaller, and the US Coast Guard Light List adds the loose end explicitly: “The ‘loom’ (glow) of a powerful light is often seen beyond the limit of visibility of the actual rays of the light.” Now take the two named cases.

• Cape Agulhas — Dubay's proof 89, “33 feet high, 238 feet above sea level, and can be seen for over 50 miles.” The tower is 27 m (89 ft); the focal plane is 31 m (102 ft) above high water; the charted range is 30 nautical miles, which is 56 km, which is 35 statute miles. All three of the proof's numbers are wrong, and the geographic range from a fifteen-foot eye is about 16 nm, so the charted 30 is plainly a luminous figure. It is worth noting how the error may have arisen: “56 km” read as miles is “over 50 miles.” We cannot prove that is what happened, but the claim is fifty per cent adrift from any published figure.
• Cape Hatteras — Carpenter's proof 5, “seen at such a distance (40 miles).” The National Park Service gives the focal height as 192.2 ft, an official range of 24 nautical miles, and a practical figure of “up to 20 nautical miles at sea” for most vessels in clear weather. Twenty nautical miles is 23 statute miles, and the geographic range for a 192-foot light and a fifteen-foot eye works out at 20.7 nm. In other words the range mariners actually get from that light is the range a sphere with a refracting atmosphere predicts, to within a few per cent. Carpenter's 40 miles is not a published figure then or now; we could not locate an 1880s Light List to trace where he got it.

His “nine-hundred feet” then follows mechanically: 8 in × 40² is 1,067 feet, less the 191-foot light, gives 876. It is the naive drop-from-tangent sum, computed from a zero-height eye, on a distance twice the real one.

6. What the light lists prove, if you read them the other way. Every geographic-range table is indexed by two heights and adds their square roots. On a plane, the height of the observer's eye could not matter: a light would be visible until it was too dim or the air too thick, and standing on a box would do nothing. The tables are built the way they are because raising the eye moves the horizon, and moving the horizon uncovers the target. That is the fingerprint, and it is not subtle — it is the difference between a 2.6 nm and a 9.8 nm contribution for the same light seen from a dinghy and a bridge wing.

7. Refraction's variability is the discriminating observation, and it is enormous. The atmosphere does not merely add a fixed 15 per cent to the horizon. Depending on the vertical temperature gradient it produces a whole taxonomy, all of it standard optics and none of it available to a plane:

• Looming — a thermal inversion steepens the density gradient, the ray bends harder, and objects normally below the horizon are lifted into view. This is the flat-earther's favourite photograph, and it is a named, modelled, predicted phenomenon.
• Sinking — a steep lapse rate does the reverse: targets ordinarily visible drop out of sight. A plane has no mechanism for this at all. If the surface were flat and visibility were limited by extinction, nothing that was visible yesterday could be hidden today in clearer air.
• Towering and stooping — vertical stretching and compression of the image, from curvature in the temperature profile. The Chicago skyline photographed across Lake Michigan shows both: the narrow section of the Willis Tower below the spires is visibly stretched, which is why the picture looks wrong rather than merely distant.
• Ducting — when the gradient reaches k = 1 the ray curves with the Earth and the surface behaves optically as though it were flat; stronger still and rays are trapped and guided round the curve. This is the regime in which the record-breaking sightings occur, and it is transient, local and measurable.
• Inferior mirage and the Fata Morgana — inverted images below or above the erect one, and the multiple alternately erect and inverted strips of the Fata Morgana. Under inferior-mirage conditions the apparent horizon can pull in to about 4 km against a standard 13–14 km. The horizon itself moves.

The numbers are real and they are measured, not assumed. Across Rainy Lake, targets 0.37–1.6 m above the water at 1.1–9.5 km were surveyed against GNSS-fixed positions and returned refraction coefficients from about 0.19 to 0.41 between sessions, with apparent lifts of roughly 0.5 m at 5.4 km and 1 m at 9.5 km. And even generous refraction does not flatten the world: from Grand Mere State Park, 56.5 statute miles from the Willis Tower, the hidden height is 453 m with no refraction, 404 m at 10 per cent and 364 m at 20 per cent — so on the best day most of a 442-metre building is still gone, which is precisely what the photographs show. The claim needs the sightings to be routine and stable. They are episodic and correlated with the weather, and the correlation is the thing being explained.

8. Hull-down is the discriminator on ships, and Carpenter made it testable. A ship going away does not shrink uniformly to a point: it is occluded from the waterline upward, the hull first, then the deck, then the funnel, with a hard edge at the horizon and full angular size retained in what remains. That is occlusion, not resolution. Carpenter's proof 63 makes the opposing prediction explicit — “a good telescope will restore to our view this portion of the vessel” — and the Flat Earth Society wiki still asserts it today: “a good telescope with sufficient zoom will change the observer's perspective and bring the ship's hull back in full view.” It is a falsifiable claim and it is false. Zooming is cropping and magnifying; it cannot recover an object that is geometrically behind the surface. The controlled version of the test is trivial: photograph a distant vessel, zoom by a factor of two, photograph again, then enlarge the first frame by two and compare the gap between the waterline and the horizon. The gap is unchanged. What does restore the hull is raising the camera, and the amount restored is the amount the horizon-shift predicts. Rowbotham's own alternative — that a rough sea occludes the hull — is a real effect at his 32-inch eye height and useless at 30 metres, where the hulls still vanish on a glassy sea.

9. The two Chesapeake proofs, on their own terms. Proof 7's daytime observation — tall trees on the far shore visible over the heads of hull-down ships — is real and is exactly what a sphere gives you, because hidden height depends on the target's distance and the observer's eye, and tall things far away can clear what short things nearer cannot. Numerically: from a 2-metre deck the horizon is about 5.5 km; a ship at 12 km loses about 2.9 m at the waterline, enough for a laden hull, while a treeline at 20 km loses about 14 m, so a 30-metre tree still shows fifteen metres of itself. The observation discriminates nothing; it merely feels paradoxical. Proof 36 is weaker still. Carpenter sights the light along “the rail of the vessel's side” — a bulwark on a hull that pitches, rolls, trims and settles as the coal burns off, which is not a levelling instrument by any definition — and demands a change in apparent elevation “one way or the other!”, which concedes that he does not know what the theory he is refuting predicts. It predicts one direction: approaching a light, it rises relative to the horizon. And his opening sentence gives the case away. “We shall see the light … for an hour before the steamer gets to it” describes a light that comes into view at a finite range. On a plane, why would it ever not have been in view?

10. Why this is MISLEADING and not REFUTED, which matters. Most of the sightings in this cluster happened. Oil platforms are seen from beaches that a bad calculator says should hide them; skylines do appear across lakes; ships are visible past the naive horizon. We are not disputing the reports, and a page that did would deserve to lose. What fails is the step from the report to the conclusion, and it fails at three separate joints: the arithmetic computes the wrong quantity, the atmosphere is omitted from a calculation that navigation and surveying have included since before the dispute began, and half the numbers cited are luminous ranges out of a catalogue rather than observations of anything. A true observation, correctly reported, married to an inference the observation does not license — that is what MISLEADING names, and it is the right verdict here.

11. Scope. This cluster is the optical claim. The hydrostatic premise — what “level” means and what shape a free water surface takes — is ARG-B01, and it deliberately rests on gravimetry and tide gauges rather than sight lines for exactly this reason. The Old Bedford Level trials, including Lady Blount's 1904 repeat, belong to ARG-B03; we note only that Blount's own photographer, Edgar Clifton of Dallmeyer's, working from two feet above the water, recorded “an aqueous shimmering vapour” floating “unevenly on the surface of the canal and adjoining fields”, and that he and Blount reported seeing the target and its reflection — two images, which is the signature of a mirage and not of a plane. The engineering “no curvature allowance” form is ARG-B05; the horizon's flatness and dip is ARG-B02; the charge that refraction is an ad hoc rescue is ARG-B07. And ARG-B13 is where the list contradicts itself: it carries “Harbor mirage charts.” as its own item while carrying “Harbor lights flat lines.” here. Conceding an atmospheric optical mechanism in one item and denying it in the next is not two arguments. It is one argument and its refutation, filed under different numbers.

The list says

Repeated long-range visibility. (222) / Tower sight lines beyond curve. (260) / Harbor lights flat lines. (378) / Lighthouse visibility. (379) / Oil rig sightlines. (381) / Ship optical range. (401)

The source says

“Allowing 16 feet for the altitude of the observer … 5 miles must be taken from the 30 miles, as the distance of the horizon” … “The only modification which can be made in the above calculations is the allowance for refraction, which is generally considered by surveyors to amount to one-twelfth the altitude of the object observed” … and, quoted by Rowbotham himself from the Encyclopædia Britannica: “the rays of light are incurvated by refraction … though the refraction may at a mean compensate for about one-seventh of the curvature of the earth, it sometimes exceeds one-fifth, and at other times does not amount to one-fifteenth.”

The qualifier was dropped.

The dominant drift is hedge_dropped, and this is one of the starkest cases on the page: the source shows its working and the list keeps two words of it. Item 379 reads, in full, “Lighthouse visibility.” The passage it descends from allows for the observer's own height and subtracts the resulting horizon before squaring — four miles in the table, five in the Spurn Point example where he allows sixteen feet of eye — converts the light lists' nautical miles into statute miles and says so each time, concedes an allowance for refraction, quantifies it, applies it, and prints the worked deduction on Cape Bonavista. Every qualification Rowbotham made survives into none of the six items. This matters more than usual because the qualifications are where he can be answered: his refraction allowance is the wrong quantity applied to the wrong variable — a twelfth of the target's height rather than about a seventh of the curvature term — and you cannot show that against a fragment reading “Lighthouse visibility.” The refutation above is therefore aimed at his arithmetic, not at the fragment.

The hedge that was dropped is the one that answers the argument. The strongest sentence in this whole cluster is not ours; it is the Encyclopædia Britannica extract Rowbotham typeset into his own chapter, which supplies the mechanism, the mean coefficient (one seventh of the curvature, i.e. the modern 7/6 rule) and the variability (“sometimes exceeds one-fifth … at other times does not amount to one-fifteenth”). The list carries none of it, and neither does any modern restatement we found. The tradition dropped a hedge that was already printed on its own founding page.

A second hedge went the same way, and it is the cleanest specimen in the cluster. Rowbotham's telescope claim is conditional: a powerful glass restores the hull “when the water is very calm” and the vessel is just hull down, while on a running sea “a telescope fails to restore it, however powerful it may be”. Carpenter's proof 63 states the effect with no condition at all, and the Flat Earth Society wiki carries it forward in that unconditional form. The condition was the part that made the claim hard to check; removing it is what turned a hedged observation into a falsifiable one, and it is now false.

Item 381 is a different drift and a harder finding: unsourced_addition, and specifically an anachronism. “Oil rig sightlines.” cannot descend from the credited source. Rowbotham died in 1884; Carpenter published in 1885; the first offshore platform standing out of sight of land was Kerr-McGee's Kermac Rig No. 16, ten miles off Louisiana, in production from 14 November 1947. It is a twentieth-century example fitted to a Victorian frame, exactly as the microwave and wind-farm items are in ARG-B05. Items 222, 260, 378 and 401 sit between the two cases: they are bare generalisations (“Repeated long-range visibility.”) of arguments the sources make about named, dated, checkable instances, which is scope_widened in effect — a claim about the Egerö light and the Chesapeake steamer restated as a claim about long-range visibility in general, with the instances that could be falsified removed.

And our own record drifted too. The cluster credits the whole cluster to Rowbotham's 1849 pamphlet; the ancestral passage names a lighthouse built in 1859. Our basis line says Carpenter's Cape Hatteras proof reappears as Dubay's proof 89; Dubay's 89 is the Cape L'Agulhas light and “Cape Hatteras” does not occur in 200 Proofs at all. Both corrections are set out in the gloss. On a page about provenance, an error in our own genealogy is the same class of failure we are documenting, and it is published rather than quietly fixed.

[1865, p. 54] It is commonly believed that surveyors when laying out railways and canals, are obliged to allow 8 inches per mile for the Earth's curvature … [1865, p. 55, quoting the standing order] That the section be drawn to the same horizontal scale as the plan; and to a vertical scale of not less than one inch to every one hundred feet; and shall show the surface of the ground marked on the plan, the intended level of the proposed work, the height of every embankment, and the depth of every cutting; and a datum HORIZONTAL LINE, which shall be the same throughout the whole length of the work, or any branch thereof respectively; and shall be referred to some fixed point stated in writing on the section, near some portion of such work; and in the case of a canal, cut, navigation, turnpike, or other carriage road, or railway, near either of the termini. [1865, p. 56] Every survey of this and other countries, whether ordnance or otherwise, is now carried out in connection with a horizontal datum, and therefore, as no other method proves satisfactory, it is virtually an admission by all the most practical scientific men of the day that the Earth cannot be other than a plane! [1865, p. 56] … thus it is evident that the doctrine of the Earth's rotundity cannot be mixed up with the practical operations of civil engineers and surveyors, and to prevent the waste of time and the destruction of property which necessarily followed the doings of some who were determined to involve the convexity of the Earth's surface in their calculations, the very Government of the country has been obliged to interfere! [1881 3rd ed., p. 57] In all these extensive surveys the doctrine of rotundity is, of necessity, entirely ignored; and the principle that the earth is a plane is practically adopted, and found to be the only one consistent with the results, and agreeing with the plans of the great surveyors and engineers of the day.

— Earth Not a Globe (1865), 1st book ed. (London: Simpkin, Marshall; Bath: S. Hayward, 1865), Section I, pp. 54-56 — Project Gutenberg #69892, sentences verified on two independent mirrors. The same argument, revised and extended, at 3rd ed. rev. and enl. (London: Day, 1881), ch. II, Experiment 13, pp. 47-57. Not the 1849 pamphlet: see the gloss

The text dates itself, and it is not 1849. The argument rests on a quoted Standing Order which the third edition introduces as “the Standing Orders of the Houses of Lords and Commons on Railway Operations, for the Session of 1862”. A text quoting an 1862 Standing Order cannot stand in a pamphlet published in 1849. The earliest text we located is the 1865 first book edition, where the passage sits at pp. 54–56; no copy or transcription of the sixteen-page 1849 pamphlet could be reached at all, and the same was recorded when ARG-B01 went looking for it. The 1881 revision extends the same argument with the Mont Fréjus tunnel (completed 1871), the Suez Canal (opened 1869) and the Atlantic cable soundings; none of those was found in the portion of the 1865 text returned by our searches, and the first two postdate it. So this is a claim of the 1860s, sharpened in the 1880s.

What he actually claims is stronger than the list fragment, not weaker. The item that descends from this reads, in full, “Surveying plane assumption.” Rowbotham's sentence is an assertion about the entire profession and about Parliament: the horizontal datum is “virtually an admission by all the most practical scientific men of the day that the Earth cannot be other than a plane”, and the Standing Order exists, he says, because “the very Government of the country has been obliged to interfere” to stop engineers ruining works by putting convexity into their calculations. That is the version answered below. No purpose of that kind appears in the words of the order as he himself prints them — the clauses he quotes specify scales, an embankment and cutting schedule, a datum line constant along the work, and a requirement that the datum “be referred to some fixed point stated in writing on the section”.

That last clause is the whole answer, and it is inside his own quotation. A datum line on a railway section is not a geometrical plane hung in space; it is a stated height above a physical benchmark. British ordnance heights of this period ran from Ordnance Datum Liverpool, redefined from tidal observations taken at Victoria Dock in 1844 — mean sea level read off a tide gauge. In his own Experiment 14 he prints a section diagram labelled “D, D, the datum line—the Trinity high water mark”, which is a tidal datum too. The surface every one of these works is referred to is the surface of the sea, and ARG-B01 establishes what that surface is.

He calls the Ordnance Survey as his witness. “Every survey of this and other countries, whether ordnance or otherwise” — and seven years before the edition quoted here, the Ordnance Survey had published Captain A. R. Clarke's official account of the Principal Triangulation of Great Britain and Ireland under the title Account of the observations and calculations of the Principal Triangulation; and of the figure, dimensions and mean specific gravity of the Earth as derived therefrom (London, 1858). It adjusted 289 stations by least squares, carried spherical excess through the triangles, computed each station's latitude and longitude on the Airy ellipsoid, and published a figure of the earth from the result. The witness he names had, in print, done the opposite of what he says every survey does.

And in 1881 he quotes that very survey back at us. Experiment 14 of the third edition observes that “the length of some of the sides of the great triangles (in the English survey) is upwards of 100 miles”, and cites “Lieutenant-Colonel Portlock, R.E., who observed the station on Precelly, a mountain in South Wales, from the station on Kippure, a mountain about 10 miles south-west of Dublin—the distance between the stations being 108 miles”. Those are trigonometrical stations of the Principal Triangulation, and both are mountain tops — which is the only way to get a 108-mile sight, and the reason the survey put its stations there.

The encyclopaedia article he reprints is a surveying article. Twenty-five pages earlier in the same chapter he typesets an extract from the Encyclopædia Britannica, article “Levelling”. ARG-B04 takes what it gives him about refraction — the mechanism, the one-seventh mean, the variability. The half that belongs here is what the article is about: the difference between the true and the apparent level, which is the curvature term in levelling. Its own sentence says refraction “may at a mean compensate for about one-seventh of the curvature of the earth”, and it closes “we have, therefore, made no allowance for refraction in the foregone formulæ” — the foregone formulæ being the ones that compute that curvature for a leveller to use. Rowbotham's comment on it is “It will be seen from the above that, in practice, refraction need not be allowed for”. He read a levelling manual's instruction not to apply a fixed refraction correction as permission to discard the curvature correction the manual was supplying.

SURFACE (weak — and a working surveyor will take it apart). “Surveyors do allow for curvature.” Said bare, this loses, because for the overwhelming majority of jobs they demonstrably do not. A crew setting out a subdivision, running a level circuit round a site, or staking a road never adds eight inches per mile to anything, and the textbooks say so in as many words: plane surveying is defined as the surveying in which the curvature of the earth is neglected. Anyone who opens by denying that has picked a fight with the manuals and will lose it.

DEEPER (true, incomplete). “Plane surveying is a bounded approximation with a published error.” Correct. The standard texts put the limit at roughly 260 km² (about 100 square miles) and the reason is arithmetic anyone can check: the spherical excess of a triangle on the earth is one arcsecond per R² sin 1″ of area, and with R = 6371 km that is one arcsecond per about 197 km². A 260 km² figure therefore carries about 1.3″ of excess — below what most instruments and most jobs can see, and above zero. The flat-earth site that runs this argument does the sum honestly and gets 20 m of “ignored curvature” across a 100-square-mile block; the drop from a tangent plane over 16.1 km is indeed 16,100²/(2 × 6,371,000) = 20.3 m. But stated alone this invites the obvious reply: then when do they start allowing for it, and why does nothing ever go wrong if they don't?

KERNEL — the true thing is that a leveller really does not add eight inches a mile, and the reason is a procedure built around the curvature. This is the specific correct observation at the bottom of the cluster, and it deserves to be stated in its strongest form. In differential levelling the instrument is set between two staves and the backsight and foresight are made equal in length. Both readings are then displaced by exactly the same curvature-and-refraction term, and the term cancels identically in the difference. The manual puts it plainly: “For most work it is sufficient to keep the foresight and backsight distances approximately equal so that the refraction and curvature effects cancel out.” So the surveyor makes no allowance — and this is not sloppiness or convention, it is the correct handling of a term that is really there. The instruction to balance the sights exists because the earth is curved. On a plane there would be nothing to cancel and the rule would not be in the book.

Second half of the kernel: plane surveying's plane is horizontal, never vertical. The 20 m figure is a drop from a tangent plane, and no survey on earth measures heights from a tangent plane. Elevations are referred to a level surface — the USGS glossary defines horizontal as “a direction perpendicular to the force of gravity” and the geodetic levelling manual calls the reference surfaces “level” or equipotential surfaces. What plane surveying approximates is the horizontal position computation over a small block. Item 382 of this cluster, “Leveling equipotential planes”, has already conceded the noun and is arguing about the adjective.

Why it doesn’t save the claim.

Because every procedure that lets a surveyor ignore curvature is engineered around it, and each one fails in the direction a sphere predicts the moment you stop obeying it. Four cases, all from ordinary practice, none of them exotic.

Unbalanced sights. The cancellation is exact only while the two sight lengths match. Let them differ and the combined curvature-and-refraction term enters at 0.0675D² metres with D in kilometres — 0.7 mm at a 100 m sight, 6.8 cm at a kilometre. That is why the tolerance tables exist and why precise-levelling specifications cap the imbalance.

Trigonometric heighting. You cannot balance anything when you shoot a single long ray with a total station, so the correction goes back in explicitly — and it is a switch on the instrument. Trimble's own documentation gives the earth-curvature correction as “approximately 16″ per km measured distance”, which is D/2R to three figures, and puts refraction at “approximately one-seventh of the earth curvature correction”. That is the same one-seventh Rowbotham printed out of the Britannica.

Level networks. Run a long line north and the level surfaces themselves converge; the orthometric correction that a national levelling network applies is zero for a strictly east–west run and grows with the north–south component, because “level surfaces are irregular and converge toward the poles”. And when Britain ran its Second Geodetic Levelling it found mean sea level differing by 0.81 ft between the Dunbar and Newlyn gauges — which is why Newlyn alone became the datum. On a plane surface of still water there is no such difference to find.

Cadastral survey. The most aggressively planar survey system ever built, the American township grid, is interrupted on purpose. The BLM Manual: “The convergency is taken up at intervals by the running of standard parallels, on which the measurements are again made full … The usual interval between the standard parallels is 24 miles.” A surveyor running range lines north hits a correction line every 24 miles and starts again, because the meridians he is following are converging.

So the kernel is real and it points the other way. “No allowance is made” is true of a procedure whose entire design is an allowance.

Concede the observations first, because they are correct. Plane surveying exists, it is taught under that name, and it treats the survey area as a plane. Railway sections really are drawn to a single datum line running the length of the work, because the Standing Order Rowbotham quotes really does require it. A crew running levels really does not add eight inches per mile. None of that is in dispute, and a rebuttal that starts by denying it has lost the argument to any reader who has held a staff.

1. “Horizontal” and “datum” are gravity words, and the Standing Order says so itself. The argument needs horizontal datum line to mean plane. In surveying it does not. The USGS glossary defines horizontal as “a direction perpendicular to the force of gravity” and level as “a line or surface whose segments are all horizontal”; the geodetic levelling manual states that the gravity field “can be represented as a series of surfaces of equal potential, termed ‘level’ or equipotential surfaces”. A spirit level is a gravity instrument: the bubble finds the perpendicular to the local plumb line and nothing else. And the order Rowbotham prints does not stop at requiring a datum — it requires that the datum “be referred to some fixed point stated in writing on the section”. That fixed point is a benchmark, and British ordnance benchmarks of the 1860s carried heights above Ordnance Datum Liverpool, redefined from tidal observations at Victoria Dock in 1844. His own Experiment 14 labels a section diagram “the datum line—the Trinity high water mark”. The datum is the sea. Reading a sea-surface datum as proof of a plane is the equivocation ARG-B01 takes apart, arriving here by a different road.

2. What plane surveying approximates, and what it never approximates. The plane in plane surveying is the horizontal-position plane over a limited block. Heights are never referred to it. The published limit is about 260 km², and the reason is checkable arithmetic rather than convention: spherical excess accumulates at one arcsecond per R² sin 1″ ≈ 197 km² of triangle area, so a block of that size carries roughly 1.3″ — small enough to drop for a boundary survey, and a quantity somebody has computed rather than assumed. The tangent-plane drop across the same block, about 20 m over 16 km, is not an error anybody is carrying, because no survey measures a height from a tangent plane. A stated threshold with a computed error term is the opposite of an assumption. The honest form of the flat-earth question — at what point are they trained to factor it in? — has a numerical answer in every textbook, and the answers below are the specific ones.

3. Levelling: the allowance is a procedure, not a number. Set the level midway and make the backsight and foresight equal, and the curvature-and-refraction displacement is identical in both readings and cancels in the difference — “keep the foresight and backsight distances approximately equal so that the refraction and curvature effects cancel out”. Break the balance and the term reappears at 0.0675D² metres for D in kilometres: sub-millimetre at a hundred metres, 6.8 cm at a kilometre. Reciprocal levelling across a river — the case where you cannot put the instrument in the middle — is done by observing both ways and meaning the results, which cancels the same term by a different trick. Every one of these is a technique for handling a real quantity. The eight inches Rowbotham says nobody allows for is the first-mile value of that quantity: the drop of a sphere below its tangent is D²/2R, and for one statute mile 0.0785 × 1.609² = 0.2032 m, which is 8.00 inches. It is the figure the nineteenth-century levelling literature carries — Simms's Treatise on the Principles and Practice of Levelling uses 8 in per mile squared — and it is the subject of the encyclopaedia article Rowbotham reprints in the same chapter, whose whole business is the difference between the true and the apparent level.

4. Where you cannot cancel it, it is a setting on the instrument. Trigonometric heights from a total station are single long rays with nothing to balance against, so the correction is applied outright, and the manufacturer's documentation states its size: the earth-curvature correction is “approximately 16″ per km measured distance” — D/2R in arcseconds is 16.2 — and the refraction correction runs the other way at “approximately one-seventh of the earth curvature correction”, with coefficients of 0.13, 0.142 or 0.2 selectable. The one-seventh in a 2020s field-software manual is the one-seventh Rowbotham typeset out of the Encyclopædia Britannica and dismissed. The refraction half of that inheritance is ARG-B04's and ARG-B07's; what matters here is that both halves come from an article about how to run a level.

5. The extensive surveys, which is where the claim breaks outright. “In all these extensive surveys the doctrine of rotundity is, of necessity, entirely ignored”, and “every survey of this and other countries, whether ordnance or otherwise”. Take him at his word and go to the ordnance survey he names. Its Principal Triangulation, observed from 1791 and adjusted by Captain A. R. Clarke, was published in 1858 as Account of the observations and calculations of the Principal Triangulation; and of the figure, dimensions and mean specific gravity of the Earth as derived therefrom — 289 stations, adjusted by least squares, with the spherical excess of the triangles carried through the computation, positions reduced on the Airy ellipsoid, and a figure of the earth published out of the far end (from British data alone, a = 20,927,005 ft with flattening 1/299.33). The claim is not merely that these surveys used a curved earth; it is that deriving the shape of the earth was one of the stated products, on the title page, seven years before the edition quoted here. And the third edition cites that same survey approvingly, invoking Portlock's 108-mile sight from Kippure to Precelly. Those are two of its trig stations, and both are mountain summits — the elevation being exactly what buys the sight line.

6. Cadastral survey: the flattest grid ever drawn is cut every 24 miles. The United States Public Land Survey lays townships out in nominal six-mile squares, which is as close to assuming a plane as surveying gets. It cannot close, because meridians converge, and the Manual of Surveying Instructions says what is done about it: “The convergency is taken up at intervals by the running of standard parallels, on which the measurements are again made full. On the standard parallels (first named ‘correction lines’) there are offsets in the range lines and two sets of corners … The usual interval between the standard parallels is 24 miles.” Every jog in the American township grid — visible from the air, and in the kink of any road that follows a range line — is a curvature allowance carried out with a chain by a surveyor who was, in every other respect, treating the ground as flat.

7. Item 47, which is a different claim and answers itself underground. “Surveying assumes stationary ground” is about motion, not shape, and we did not locate any argument of that kind in the passages searched. Answered on its merits: the standard instrument for orienting a survey where the sky is not visible is the gyrotheodolite, and what it senses is the earth's rotation. A wheel spun at 20,000 rpm and released near the meridian precesses into alignment with it, because “the gyroscopic reaction of spin and Earth's rotation results in precession of the spin axis in the direction of alignment with the plane of the meridian”. It is “the main instrument for orientation in mine surveying and in tunnel engineering”; it steered the Channel Tunnel; it gives the meridian to about 10 arcseconds; and it stops working within roughly 15° of the pole, “where the angle between the earth's rotation and the direction of gravity is too small for it to work reliably”. That latitude dependence is a rotating sphere's signature, and it is the reason mine surveyors buy a different instrument at high latitudes. Above ground the same point is made by the geodetic datums themselves, which no longer pretend the ground is fixed: modern reference frames carry station velocities and epoch-dependent coordinates precisely because the crust moves.

8. Item 395, mine surveys, and item 383. Mine surveying is where the argument is weakest rather than strongest: an underground survey is tied to the surface network through the shaft, oriented by an instrument that works only because the earth turns, and computed on the same national grid as everything above it. It is also, historically, where the mass of the earth was weighed — Airy's 1854 pendulum experiment in the Harton Colliery was published as an Account of pendulum experiments undertaken in the Harton Colliery, for the purpose of determining the mean density of the Earth, and “mean specific gravity of the Earth” is on the title page of the Ordnance Survey's triangulation report for the same reason. Item 383 is two words, “Hydrology planar”, and we read it as the river-gradient claim — an inference from those two words rather than anything the item states. On that reading the earliest text ARG-B01 could document for it is Carpenter's 1885 proof 4 rather than anything of Rowbotham's. The engineering form of the no-allowance argument — canals, rail, pipelines, bridges — is ARG-B05's and is not re-argued here.

Verdict: self-contradicted, and by a witness of his own choosing. The argument's evidence is (i) an encyclopaedia article on levelling, whose subject is the difference between the true and the apparent level and which states that refraction offsets about one-seventh of the curvature of the earth; (ii) a Standing Order requiring a datum line referred to a fixed benchmark, on a network whose zero was a tide gauge; and (iii) the Ordnance Survey, whose principal publication of the period derived the figure and dimensions of the earth from the survey in question. All three were in the book. Two of them he printed himself.

The list says

Surveying assumes stationary ground.

The source says

“Every survey of this and other countries, whether ordnance or otherwise, is now carried out in connection with a horizontal datum, and therefore, as no other method proves satisfactory, it is virtually an admission by all the most practical scientific men of the day that the Earth cannot be other than a plane!” (1865, p. 56)

The source does not contain this.

The drift here is not softening — the source is more emphatic than anything on the list, claiming an admission by the whole profession and an intervention by Parliament. The drift is that four of the cluster's five items are not the source's argument.

Item 223, “Surveying plane assumption”, is faithful and is what the refutation above answers, at the strength quoted. Item 47, “Surveying assumes stationary ground”, is a claim about motion, not shape; we did not locate any argument of that kind in the texts searched (1865 ed., Project Gutenberg #69892, the datum and survey passage at pp. 54–56; 1881 3rd ed., ch. II Experiments 13–14 via the Internet Sacred Text Archive), and we did not run a full-text search of either edition for it. Items 382, 383 and 395 introduce vocabulary — equipotential surfaces, hydrology, mine surveying — for which we located no ancestor in those same passages; item 383, “Hydrology planar”, we read as the river-gradient claim, which is an inference from two words rather than anything the item states, and on that reading the earliest text ARG-B01 could document for it is Carpenter's 1885 proof 4 rather than anything of Rowbotham's. Unreachable is not absent, and this is a statement about the passages we read, not about the corpus.

That pattern — one cluster attribution applied to every item in the cluster — is the same failure recorded against ARG-R06 in the corrections log, and the same test applied to ARG-E13 moved the published totals in the other direction. It is worth stating plainly which way it cuts here: the surveying argument really is Rowbotham's, and it is his in a stronger form than the list states. Four items riding on it are of unestablished authorship.

There is no original to quote, and as at ARG-C08 and ARG-C09 that is a conclusion rather than a shrug. The specimen carries no citation for these four items — it carries none for any of the 461 — so the search ran outward through the literature the rest of the list demonstrably draws on. Here is the route, and where it stopped.

The movement’s largest work reaches none of these four instruments. The full text of Galileo Was Wrong Vol. I (Internet Archive item GallileoWasWrong, 462,032 words) and Vol. II (item …Bennett4276, seventh edition, 2013, chapters 7–13, 284,044 words) were downloaded and searched offline on 2026-08-10. “Photogrammetry”, “sonar”, “bathymetry”, “LiDAR”, “synthetic aperture”, “echo sounding” and “remote sensing” are not located in the text of either volume as searched. “Radar” is there — 24 occurrences in Vol. I and 22 in Vol. II — and every one was read. They fall into four groups. Most are the Shapiro Venus time-delay experiment and Bryan Wallace’s challenge to it. Then Van Flandern’s planetary radar-ranging item; a single sentence about the Department of Defense using radar to map GPS satellites to ground reference points; and contents-page and bibliography lines pointing back at the first group. One occurrence in Vol. I is neither: it is the figure of speech “not merely a blip on the radar screen”. So: planetary radar, satellite tracking, and a metaphor. A terrestrial radar horizon is not among them. “Interferometry” is there too, 23 and 9 times, and every occurrence located is Michelson–Morley, Miller, Allais, Sagnac or very-long-baseline — none of it synthetic-aperture.

The largest modern flat-earth proof list reaches them even less. The full text of Dubay’s 200 Proofs Earth Is Not a Spinning Ball was searched the same way on the same day, for radar, sonar, LiDAR, photogrammetry, bathymetry, interferometry, synthetic aperture and remote sensing. The count for every one of those terms in that text is zero.

Two more places a claim of this shape would live. The Flat Earth Society wiki’s Frequently Asked Questions was read end to end on 2026-08-10; none of these terms is located in it. And FlatEarth.ws — a debunking site, but a useful census of what actually circulates, because it writes up claims in order to answer them — carries a radar topic index of three articles. Exactly one flat-earth radar claim is described there, and it is a different claim from ours: that marine radar sets are marketed with maximum ranges far exceeding the distance to the horizon (“Marine Radar”, 9 April 2020). No originator is named for it. It is answered in section 1 below anyway, because it is the most plausible thing item 217 could mean.

Why nothing older was searched. Rowbotham and Carpenter are the ancestors for most of lane B, and they are not candidates here: radar dates from the 1930s, sonar from 1912, photogrammetric aerotriangulation from the 1920s, synthetic-aperture radar from 1951 and interferometric SAR from 1974. Hunting a Victorian ancestor for these four lines would be theatre. What is worth recording instead is the shape of the passage they sit in. Items 385 to 402 run: pipelines, railroads, highways, wind farms, drone RTK, photogrammetry, LiDAR, SAR, seismology, volcano infrasound, mining surveys, architecture, skyscrapers, pendulum clocks, GNSS pseudorange, bathymetry, ship optical range, harbour mirage charts. Eighteen consecutive lines, one per profession, most of them without a verb. That is a list grown by enumerating disciplines rather than by citing anyone, and it is the clearest specimen of the mechanism on the whole page.

An honest note on our limits. No author found means we did not find one, not that none exists. Four noun phrases can enter circulation through a livestream, a comment thread, a slide or an image caption that leaves nothing searchable. A reader who can point us at someone who actually argued from radar, photogrammetry, SAR or sonar to a flat Earth — in print, on air, anywhere datable — will improve this entry, and we will publish the correction.

SURFACE (weak — do not use). “Obviously these instruments account for curvature; the claim is silly.” This loses, and it loses to citations, because a flat Earth is a named, documented, daily-used object in all four of these fields. Anyone who opens with the easy bust will be handed four references and will deserve it.

DEEPER. Working in a local tangent plane is not a shortcut, it is the professional standard over short baselines, and it has a name: plane surveying, as distinct from geodetic surveying. Over the extent of one job the plane is not merely convenient, it is accurate. A defender who says this has said something every practitioner agrees with.

KERNEL. The strongest form names the four objects individually, and every one of them is real. In radio propagation there is a standard model called plane earth loss — the two-ray ground-reflection model, in every RF textbook, which treats the ground as an infinite plane. In photogrammetry the collinearity equations that underpin every bundle adjustment are written for a three-dimensional Cartesian space, with the Earth’s curvature entering afterwards, as a correction applied to a model that did not contain it. In SAR interferometry there is a processing step whose name is flat-earth phase removal. In shallow-water acoustics there is the isovelocity assumption, under which sound is taken to travel in straight lines. So the honest version of this cluster is not “these instruments secretly know the Earth is round” versus “they assume a plane”. It is: the flat Earth is a working object in all four disciplines, it is in the textbooks under that description, and the list did not invent it.

Why it doesn’t save the claim.

Because each discipline publishes the limit of its own flat approximation, and the limit is computed from the Earth’s radius. A truncated series is not a rival to the function it was truncated from.

The cleanest statement of this is in the radio literature, where both models exist side by side and the second was published because the first fails. Yang, Molisch, Ekman, Røste and Berbineau, proposing a Round Earth Loss model for open-sea propagation, state the reason directly: “the PEL assumption of a plane earth surface inherent in this model is not fulfilled for maritime radio links at larger distances … This is due to the fact that the diffraction loss caused by the earth curvature and sea roughness can not be ignored.” Plane earth loss is real, taught, and bounded — and what bounds it is the curvature.

The same shape holds in each of the other three, and section by section the refutation gives the numbers: the photogrammetric planar model is corrected by a term of the form r³(H − Z)/2c²R, which has the Earth’s radius in the denominator and is worth about 40 micrometres at the corner of an ordinary aerial frame; the flat-earth phase in InSAR is defined as the phase due to the curvature of the reference surface and is computed on WGS 84; and the straight-ray assumption in sonar is what produces the artefacts hydrographers call smiles and frowns, which is why they measure the profile instead. In every case the flat model is derived from the round one, carries an error term proportional to 1/R, and is used inside the range where that term is smaller than the tolerance. You cannot know where a plane stops working without knowing the radius of the thing it is approximating.

What this cluster covers, stated first, because two of its four items are not the same kind of claim as the other two. The four are item 217 “Flat radar horizon.”, item 390 “Photogrammetry planar.”, item 392 “SAR interferometry.” and item 400 “Bathymetry sonar straight.” The datum answer — that these systems run on an ellipsoidal reference — is the right answer to 390 and 392, and it is scored in its general form at ARG-R08. It is not the answer to 217 or 400, because a radar horizon is not a datum question and a sonar ray path is not a datum question. Those two need different answers and get them below.

And one concession before anything else, because it governs how the rest should be read. Items 390 and 392 are noun phrases with no predicate. “SAR interferometry.” asserts nothing at all on its own; the claim is supplied by the page heading these lines sit under, “435 Pieces of Evidence The Earth is Not A Spinning Ball”. So what follows answers the strongest reading available — that these techniques model the ground as a plane and thereby show it is one — and says so out loud rather than pretending the items argued it.

1. The radar horizon has a formula, and the Earth’s radius is in it

Take reading (a) first: that radar sees a plane. The distance at which a radar loses a low target is standard engineering, not a controversy. With the Earth’s radius R and antenna height h, the geometric horizon is d = √(2Rh); the working version folds in atmospheric refraction by replacing R with an effective ⁴⁄₃R = 8,495 km, giving the number every radar text carries: d(km) ≈ 4.12 √h(m). Recomputed here 2026-08-10: √(2 × ⁴⁄₃ × 6.371 × 10⁶) = 4,122, so 4.12 with h in metres and d in kilometres; the unrefracted constant is 3.57. A 10 m mast reaches 13 km, a 23 m one 19.8 km, an aircraft at 10 km reaches 412 km. The whole quantity is a curvature; set the curvature to zero and the formula returns an infinite horizon, which is not what anybody measures.

This is not a textbook curiosity, it is a daily operational constraint. The United States weather-radar network is limited by it every hour: as the National Weather Service puts it, “Earth curvature and standard refraction dictate that the beam becomes more elevated above the surface with increasing range”, so the radar overshoots shallow precipitation at distance, and “overshooting of precipitation by the radar beam often produces the largest errors, usually causing precipitation underestimation.” The 4/3-earth beam-height relation puts a 0.5° beam at 0.58 km above ground at 50 km, 1.46 km at 100 km and 5.1 km at 230 km — recomputed here. The forecaster in the next county cannot see the rain falling under that beam, and the reason has a radius in it.

Two entire equipment categories exist because of this limit. Airborne early warning aircraft carry a radar up a tower fifty times higher than any mast can be built, for no other reason. And over-the-horizon radar — the Australian Jindalee network, with an official range of 3,000 km — abandons microwaves altogether and bounces high frequencies off the ionosphere, because, in the reference literature’s own words, microwave radar propagation “generally limits the detection range of radar systems to objects on their horizon … due to the curvature of the Earth.” Nations do not fund a second physics for a limit that does not exist.

Now reading (b), which is the version documented as actually circulating: that marine radars are advertised with ranges far beyond the horizon, so the horizon cannot be real. The advertised figure is a signal range — how far the transmitter can put usable power and still resolve a return — not a detection range for a surface target. Recomputed: a set with its scanner 20 m up, looking at a ship whose superstructure reaches 30 m, has a radar horizon of 4.12(√20 + √30) = 41 km, about 22 nautical miles. That set may be sold as a 96-mile radar, and it will indeed paint a mountain or a rain cell at 96 miles, because those are tall. It will not paint a hull. Every navigation course teaches the distinction, and the ranges at which it fails are the ranges the formula predicts.

The honest qualification, which a defender will otherwise supply. Radar does routinely detect targets beyond the 4/3 horizon, through ducting and other anomalous propagation, and sometimes by hundreds of kilometres. That is real. It also does not help, because every stage of the standard engineering treatment of it is parameterised by the Earth’s radius. The effective-radius factor is k = 1/(1 + R dn/dh), with n the refractive index; the 4/3 value is the standard atmosphere, a sub-refractive layer is a smaller k and superrefraction a larger one. A duct is not simply a larger k still: it is the limit at which that model stops working. At a refractivity gradient of −157 N-units per kilometre the denominator vanishes and k passes through infinity — the case in which, as the propagation notes put it, “the curvature of the ray matches the curvature of the earth and the ray maintains the same altitude as it travels” — and past that the effective radius is negative. What engineering practice switches to there is modified refractivity, M = N + (h/R_e) × 10⁶ ≈ N + 0.157h with h in metres, introduced in the over-the-horizon literature because “the influence of the curvature of the Earth can be achieved by modifying the refractivity of the troposphere”. Recomputed here: 0.157 per metre is 10⁶/R at R = 6,371 km. So the term added to the refractivity in order to compute ducted propagation on a flat grid is the height divided by the Earth’s radius — section 5’s structure arriving from the least convenient direction available. Anomalous propagation is parameterised by the curvature it bends around at every level of the treatment, never offered as an alternative to it.

2. Photogrammetry: the planar model is real, and it ships with a curvature term

Item 390 is true as far as it goes, and that is the interesting part. Analytical photogrammetry does work in Cartesian space: the collinearity condition that ties an image point to a ground point is written for a three-dimensional Cartesian frame at both ends, and over one frame that frame is effectively flat. The discipline then applies a correction, and the correction is published in the same chapter as the model. The radial displacement of an image point caused by the Earth’s curvature is

d_earth = r³(H − Z_P) / (2c²R)

with r the radial distance in the image, c the principal distance, H − Z_P the height above the ground point, and R the Earth’s radius, taken in the standard treatment as a sphere of 6,372.2 km. Note where R sits: in the denominator, so the correction vanishes only as the radius goes to infinity.

Put ordinary numbers in it. A survey flown 1,500 m above the ground with a 100 mm lens, at a point 150 mm out from the principal point: d = 39.7 micrometres in the image — about eight pixels on a 5 µm sensor — which at 1:15,000 is 0.60 m on the ground. At the true corner of a 230 mm frame it is 50.6 µm and 0.76 m. That was derived here a second way as a check, and the two agree exactly: the ground radius is 2,250 m, the curvature drop over 2,250 m is 2250²/2R = 0.397 m, and the relief displacement of a 0.397 m height change at that position is 39.7 µm. The curvature is not a rounding error hidden in the software; it is a systematic tilt of the whole frame outward, larger — as the standard treatment notes in as many words — than the atmospheric-refraction correction sitting next to it. A photogrammetrist who left it out would deliver a block that fails its own check points.

3. SAR interferometry: “flat earth” is the name of the thing you subtract

Item 392 names a technique in which the phrase “flat earth” is genuinely standard vocabulary — and it is the name of an artefact, computed from an ellipsoid, and removed. The argument here is not the pun; it is the definition. ESA’s own processing documentation for the interferogram operator states it: the flat-earth phase is “the phase present in the interferometric signal due to the curvature of the reference surface”; the reference system is “for now only WGS84 supported, which the reference system used by all space-borne SAR systems”; and if you skip the step, “the formed interferogram will still have the fringes caused by the earth curvature, and could hamper further interferometric processing and analysis.” The reference summary of InSAR says the same in one line: “the interferometric phase due to the curvature of the Earth is removed, a process referred to as flattening.” Those fringes are observed. They are counted. They are then modelled away using a rotational ellipsoid with a specified flattening, and what is left over is the topography and the ground motion the mission was flown to measure.

Two further things are true of every spaceborne SAR image, and neither is optional.

The ellipsoid is one of the three equations. Locating a SAR pixel on the ground is a simultaneous solution of a slant-range equation, a Doppler equation, and an Earth-model equation — and the third one is the ellipsoid, printed in the geolocation literature as x²/R_e² + y²/R_e² + z²/R_p² = 1 with an equatorial radius and a distinct polar radius. Remove the third equation and the system is underdetermined: the range and Doppler conditions alone place the target on a circle, and it is the Earth model that picks the point. There is no spaceborne SAR image whose pixels have ground coordinates without an oblate spheroid in the solver.

And the spacecraft physically turns to cancel the Earth’s rotation. A side-looking radar in a polar orbit sees a Doppler shift from the ground it is illuminating, and part of that shift is the ground moving underneath: as the SAR attitude-control literature puts it, “the Doppler centroid is not zero in the conventional broadside mode … due to the earth rotation and eccentricity of the orbit.” The standard fix, developed for TerraSAR-X and known as total zero-Doppler steering, is a yaw manoeuvre computed to null it. The satellite is aimed slightly off square, continuously, at an angle derived from ω_⊕. Set the Earth’s rotation to zero and the steering law has nothing to correct.

4. Bathymetry: the rays are the problem, and the curvature is not

Start with the part that goes against us, because it does. Over the width of one multibeam swath the Earth’s curvature is negligible, and no amount of rhetoric changes that. Recomputed here: the curvature drop across a 60 m half-swath — a typical coastal survey in 30 m of water — is three tenths of a millimetre; at 500 m it is 2 cm; it only reaches 7.8 m at a 10 km half-swath in the deep ocean, where the IHO’s Order 2 vertical allowance at 5,000 m depth is 115 m. So anyone answering item 400 by saying “but the sea floor curves” is answering with a term far below the noise. That is not the answer.

The answer is that the word in the item is “straight”, and sonar rays are not. Sound speed in the sea is not a constant: the Mackenzie equation’s linear term is 4.591T, so roughly 4.6 m/s per degree Celsius, with further dependence on salinity and pressure. A ray crossing a layered water column therefore refracts, by Snell’s law, exactly as light does entering glass. This is not a subtlety at the edge of the error budget; it is the discipline’s principal systematic, and it has names. Get the profile wrong and the swath develops what hydrographers call smiles and frowns — “two classic error patterns in multibeam data, where the seafloor appears to curve up or down at the edges of the swath”. Note what that means: assuming straight rays does not flatten the sea floor, it bends it, into a curvature that is not there.

The size of it, recomputed here, with each number attached to the mechanism it comes from, because two different ones are at work. A 1% error in the sound-speed profile is about 15 m/s, which is roughly three degrees of temperature. The first mechanism is plain range scaling: a travel time converted with a speed 1% high yields a range 1% long, which at 100 m depth is 1.0 m of vertical error before any ray has been bent at all — 72% of the 1.39 m an IHO Order 1a survey allows at that depth, from the bias alone. The second is the bending. By Snell’s law the same 1% turns a beam launched at 60° into one travelling at 61.0°, and that one degree moves the point where the beam meets a 100 m seafloor 7.3 m across-track. Carry both through the same geolocation — range 1% long, angle 1° off — and the outer sounding lands about 3.5 m out horizontally and about 2 m out in depth, in the shallow direction; a 1% bias the other way puts it about 2 m deep. So the vertical error at the swath edge is not 72% of the allowance but roughly half again the whole of it, from one un-measured profile. This is why survey vessels stop, lower a sound-velocity probe, and repeat it as conditions change, and why permanent sensors are fitted at the transducer head.

It has a documented history, too. The refraction of sound by a warm surface layer was identified during the Second World War as the afternoon effect, in which sonar detection ranges collapsed on calm sunny days: Iselin and Woodcock reported for Woods Hole in 1942 that “with moderate or light winds and a clear sky the diurnal heating which occurs near the sea surface can cause a serious reduction in the range of submarine detection.” The response was the bathythermograph — an instrument invented for no other purpose than to measure the profile that bends the rays. Straight-line sonar is not the assumption of the field. It is the failure mode the field was built around.

And the vertical reference is not a plane either. Modern hydrography is positioned by GNSS in a three-dimensional geodetic frame and reduced through a separation model: NOAA’s ellipsoidally referenced surveys describe geoid undulation as “the outward-normal distance from the reference ellipsoid”, and the chart datum a sounding is finally referred to is derived from tidal and gravimetric surfaces, not from a level plane. That half of the answer is the datum argument, and it belongs with 390 and 392.

5. Why the plane is allowed, and what allowing it costs the argument

Nothing above says a practitioner is wrong to work in a plane. The opposite: the plane is correct, within a tolerance, and the discipline states the tolerance. Two numbers define it, both recomputed here. The spherical excess of a triangle — the amount by which its angles exceed 180° — is its area divided by R², so a triangle of 197 km² has an excess of one arcsecond. And a line on the Earth’s surface exceeds its own chord by roughly s³/24R², which is 8 mm at 20 km and 6 mm at 18 km. Those are the numbers that separate plane surveying from geodetic surveying, and both of them have R in them.

That is the whole structure of the answer. A flat-Earth approximation is a truncated series. It is the first term of an expansion whose next term is proportional to 1/R, and you use it precisely where that next term is smaller than what you are trying to measure. Every one of the four disciplines named here does that, and every one of them publishes the crossover. The claim needs the plane to be the model. What the textbooks contain is the plane as an approximation with a stated radius of validity — and you cannot state the radius of validity without the radius. The surveying form of this argument is scored at ARG-B06, and the engineering form — canals, rail, pipelines — at ARG-B05.

6. The list cancels itself, two lines apart

Read the specimen’s own sequence. Item 390 is “Photogrammetry planar.” Item 391, the next line, is “LiDAR ECEF.” Item 389, the line before, is “Drone RTK Earth grid.” ECEF is Earth-Centred, Earth-Fixed: a frame whose datum is the WGS 84 ellipsoid and whose transformation to an inertial frame contains the Earth’s rotation rate as an explicit term — the point scored at ARG-R08. Airborne LiDAR and aerial photogrammetry are commonly flown on the same aircraft, on the same GNSS/inertial trajectory, adjusted against the same control and delivered in the same coordinate reference system. The list offers the planar character of one and the ellipsoidal frame of the other as two separate proofs of the same conclusion, on consecutive lines.

And the cancellation does not need two items. It is available inside item 392 alone: the technique it names contains a planar term that exists only as something computed from the WGS 84 ellipsoid and then subtracted, so “SAR interferometry” is simultaneously the list’s evidence for a plane and the clearest instrument on the list for measuring the curvature it removes. Which is the point of counting arguments rather than items. Four lines here, and the two halves of the same processing chain are being spent on both sides of the same ledger.

There is no original to hold these four lines against, and the search that established it is set out in full under “No original to quote” above: the specimen cites nothing; the full texts of Galileo Was Wrong Vol. I and Vol. II were searched offline and photogrammetry, sonar, bathymetry, LiDAR and synthetic aperture are not located in either as searched, while every one of the 46 occurrences of “radar” across the two volumes is planetary radar, satellite tracking, a contents or bibliography line, or — once — a figure of speech; Dubay’s 200 Proofs returns zero for every one of those terms; and the one flat-earth radar claim the claim-cataloguing literature does record is a different claim about marine radar advertising, with no originator named. The hedge rule has nothing to bite on here, and that is the result. Where an argument has an author we can show the list hardening a hedge; here there is no hedge, because there is no sentence upstream of the fragment.

Two things are worth publishing anyway, and neither is a drift. The first is that the items have no predicates. “Photogrammetry planar.” and “SAR interferometry.” assert nothing on their own; the argument is supplied entirely by the heading the lines sit under, “435 Pieces of Evidence The Earth is Not A Spinning Ball”. A reader meets them already told what they are for. That is compression with the source removed rather than compression of a source, and the seven drift types have no word for it because all seven presuppose an upstream sentence.

The second is that the items borrow technical vocabulary, and vocabulary can be checked even when authorship cannot. Item 392 is the clean case: “flat earth” is standard terminology in SAR interferometry, and in ESA’s own processing documentation it denotes “the phase present in the interferometric signal due to the curvature of the reference surface”, computed on WGS 84 and subtracted. The list has picked up a phrase whose published definition is the curvature of the Earth. Nothing was misquoted, because nothing was quoted; the reversal happened between a discipline and a noun phrase, with no author in between.

Recorded as no-source rather than unassessed. The comparison was attempted and it terminated in an answer, which is a finding about how this part of the list was assembled — eighteen consecutive lines, one per profession, none of them citing anyone — and not a backlog item.

Proposed instead: SELF-CONTRADICTED

MISLEADING is defined on this page as real data with a wrong conclusion made to look supported. That fits items 390 and 392, where the underlying practice is real: photogrammetry's collinearity model genuinely is Cartesian, and SAR interferometry genuinely does contain a step called flat-earth phase removal. It does not fit items 217 and 400, which are not real data with a wrong conclusion attached but false descriptions of the instruments they name. Radar horizons are not flat — the range relation d = 4.12 sqrt(h) is a curvature, the National Weather Service states that Earth curvature and standard refraction raise the beam above the surface with increasing range (the 5.1 km figure at 230 km given in the refutation is our own computation from the 4/3-earth relation, not the Weather Service's number), and airborne early warning and over-the-horizon radar are two equipment categories built to defeat the limit. Sonar rays are not straight — sound speed varies about 4.6 m/s per degree Celsius, rays refract by Snell's law, and the straight-ray assumption is what produces the artefacts hydrographers call smiles and frowns. SELF-CONTRADICTED is defined in the legend at the foot of this page as the claim's own source, or another item on the same list, pointing the other way, and both halves of that definition are already in use: at C05 and C10 the list runs two proofs that cancel (C10's counterpart sits in a different cluster, C03), and at A19 the cluster has no source at all and the verdict rests on the named instrument — a gyrocompass finds north by sensing the rotation the item denies. Both halves apply here. Across items: item 390 'Photogrammetry planar.' and item 391 'LiDAR ECEF.' are consecutive lines in the specimen, and ECEF is the WGS 84 ellipsoid in a rotating frame — the two instruments are routinely flown on the same aircraft, adjusted in the same block and delivered in the same reference system. Within a single item: the flat-earth phase item 392 invokes is defined in ESA's own processing documentation as the phase due to the curvature of the reference surface, computed on WGS 84 and subtracted, so item 392 offers as evidence for a plane a technique whose planar term exists only as a measured consequence of curvature. The fallback, if a reviewer reads SELF-CONTRADICTED as reserved for the fact-denial tier rather than for cancelling proofs, is REFUTED, on the strength of the two items that are contradicted by specific measurements. What should not survive either way is treating all four items as one kind of claim: the datum answer covers two of them and is not the answer to the other two, and the refutation above says which is which in its opening paragraph.

The verdict on the scorecard is unchanged until the question is settled. Publishing the disagreement is the point: a rubric that never produces dissent is not being applied.

[1865, Section I] Ascending or descending, the distant horizon does the same. It rises and falls with the observer, and is always on a level with his eye. … [enlarged ed., Experiment 11, p. 41] It was affirmed that, through this instrument, when “levelled,” the horizon always appeared below the cross-hair … In every instance when the experiment was tried, this appearance was found to exist … He ascertained that in those of the very best construction, and of the most perfect adjustment, there existed a certain degree of refraction, or, as it is called technically, “collimation,” or a slight divergence of the rays of light from the axis of the eye, on passing through the several glasses of the theodolite. He therefore obtained an iron tube, about 18 inches in length; one end was closed, except a very small aperture in the centre; and at the other end cross-hairs were fixed. A spirit level was then attached, and the whole carefully adjusted. On directing it, from a considerable elevation, towards the sea … the cross-hair at the opposite end was seen to cut or to fall close to the horizon. … [enlarged ed., TANGENTIAL HORIZON, p. 203] The infinitesimal or mathematical point actually in the centre is, of course, not visibly influenced, being in the very centre or on the true axis of the eye, but any part in the minutest degree out of that abstract centre is dilated, or diverged, or thrown further away from it than it would be to the naked eye.

— Earth Not a Globe (1865), Two texts, quoted separately, with two paginations that must not be mixed. (1) Earth Not a Globe, 1865, Section I — the eye-level sentence and the lens argument, from the Project Gutenberg plain text (ebook 69892, “Original publication: United Kingdom: Simpkin, Marshall, and co., 1865”). (2) The enlarged edition — Experiment 11 and TANGENTIAL HORIZON — from the archive.org item zeteticastronomy-earthnotaglobe, file ZeteticAstronomy-EarthNotaGlobe-3e-format2_djvu.txt, whose title page is dated 1881 and which carries every section absent from 1865. Which enlarged edition supplied its copy-text was not established here: the only preface located in that file is “Preface to the second edition”. Page numbers below are that file's own (Experiment 11 at 40–41, TANGENTIAL HORIZON at 202–203); a different set for the same sections circulates from another reproduction and this entry adjudicates nothing between them. Obvious OCR artefacts are normalised; nothing else is altered.

The first sentence is the item; the rest is the argument. Read on its own, “the horizon … is always on a level with his eye” invites the easy answer — level a theodolite at the sea and read the depression off the vertical circle. Rowbotham got there first, in print, in both editions. He reports that surveyors challenged him to do exactly that, that he did it, and that “in every instance when the experiment was tried, this appearance was found to exist.” He is not disputing the reading. He is disputing its cause.

And he did not stop at a hypothesis: he built the null instrument. This is the part of the source that has to be answered. Having named the cause — collimation, “a slight divergence of the rays of light from the axis of the eye, on passing through the several glasses of the theodolite” — he made an 18-inch iron tube with a pinhole at one end and cross-hairs at the other, levelled it, pointed it at the sea from “a considerable elevation”, and reports that the cross-hair “was seen to cut or to fall close to the horizon”, “always with the same result”. A lens-free instrument, a stated null, and a mechanism for the discrepancy. Whatever else this is, it is not an argument from having failed to look.

The scope clause, which is what the list loses. In THE “DIP SECTOR”, answering Herschel, the claim is stated with its restriction attached: “unless some telescopic instrument is used no dip whatever is required to meet the sea horizon.” That is a claim about unaided sight. The instrumental case is handled separately and conceded outright under THEODOLITE TANGENT: “The author has made experiments similar to the above, and found it to be as stated; but the cause is not that the line of sight is a tangent, but the same ‘collimation’.”

Where it strains is on his own page, and it is worth following slowly, because he saw the strain too. The defect he describes is an off-axis effect. His demonstration of it is a magnifying glass held over a straight line: the part seen through the lens is displaced, and — his words — “The infinitesimal or mathematical point actually in the centre is, of course, not visibly influenced, being in the very centre or on the true axis of the eye.” Now read his own description of the measurement, in the same section: “let the instrument be inclined downwards until the cross-hair touches the horizon”, and take the dip as the angle turned through. That procedure brings the horizon onto the axis and then reads a mechanical angle off a graduated circle. On the optics he himself sets out, that reading looks exempt from the defect he is invoking against it. He saw the objection, and his answer is in both editions, not filed away somewhere else. 1865 puts it as “the axis or actual centre is always occupied by the cross-hair”. The enlarged edition puts it inside this same section: the dip taken this way is the geometric angle “to which must be added the collimation”, because “the spider’s web of which the cross hair is made is placed in the actual centre; hence, in an observation, the point absolutely opposite to it is not seen, but only some other point minutely distant from it” (pp. 203–204 of the file read here). So on his account nothing is ever quite axial. But whatever residual that leaves is a fixed property of a particular tube, and the refutation below turns on three things a fixed property of a tube cannot do.

The witnesses he calls do not testify for a plane either. The four balloon accounts in the 1865 text describe the Earth as looking concave — “a huge dark bowl”, “an immense terrestrial basin”, two watch-glasses edge to edge. He has a theory of why a plane would look like a bowl, and it is on the same pages: two paragraphs earlier the sea “appears to rise up on all sides equally and to surround him like the walls of an immense amphitheatre. He seems to be in the centre of a large concavity”; immediately after the balloon accounts he explains it, the surface rising to the line of sight “in the same manner that the ceiling and the floor of a long room, or the top and bottom of a tunnel appear to approach each other”. So these are not neutral witnesses he happened to collect; they are reports his perspective model was built to accommodate. What none of them reports is a plane. What they report is an appearance, and the appearance is the quantity in dispute. Wise, the first of them, calls the appearance “an optical illusion which increases as you recede from it”; Elliott, the third, says that “the aëronaut may well be the most sceptical man about the rotundity of the Earth” — and then gives the point away in the next breath: “Philosophy imposes the truth upon us; but the view of the Earth from the elevation of a balloon is that of an immense terrestrial basin.” Rowbotham prints both sentences without comment.

What this passage is being cited as. The earliest located text carrying the argument in the form the list uses. Note which edition carries what: the eye-level sentence, the balloon quotations and the lens explanation are in 1865; the levelled board, the pier quadrant, the balloon arithmetic, the iron tube and the Grand Hotel clinometer are additions in the enlarged editions. The strongest version of this argument is not located in either text read here. The 16-page 1849 pamphlet — whose own title advertises “several experiments which prove that the surface of the sea is a perfect plane” — was not reached for this review and has not been checked for it, though Experiment 15 is sited at Brighton buildings that postdate 1849.

SURFACE (weak — do not use). “Look out of an aeroplane window, or at any high-altitude photograph — the curve is obvious.” This loses, and it loses to a refereed paper in the right journal. Lynch, Applied Optics 47(34):H39 (2008), concludes that the minimum altitude at which the curvature of the horizon can be detected is “at or slightly below 35,000 ft”, and then only with a 60° field of view and nearly cloud-free air; he adds that “virtually all camera lenses project an image that suffers from barrel distortion”, which disposes of most photographs offered in either direction, and that for a fair test the horizon must be “placed precisely in the center of the image, i.e., on the optical axis.”

DEEPER. The dip at any height a person stands at is below the threshold of unaided judgement. On a beach it is 2.5 arcminutes; on a hundred-foot cliff, 10.6; at the top of Rowbotham’s Brighton pier, 6.2. The eye carries no internal reference to the horizontal — “eye level” is not a quantity the eye measures — so “the horizon is level with my eye” at these heights is a report about the precision of the method. A defender who says only this has said nothing an optician would contest.

KERNEL. The strongest form is not about eyesight at all, and it belongs to Rowbotham rather than to any modern list. It runs: I accept the theodolite result. I reproduced it. A levelled telescope sighted at the sea does show the horizon below the cross-hair, and the angles of a large survey triangle do sum to more than 180°. But a telescope is a stack of lenses, and a divergence of the ray from the axis of vision through those glasses is a real, named, universally acknowledged instrumental error — so well known that the Ordnance Survey repeated its observations upwards of a hundred times to average errors down. I did not merely assert this. I built an instrument with no lenses in it, levelled it, pointed it at the sea from height, and the horizon came up on the cross-hair. You are inferring the shape of the world from a residual of a few arcminutes that disappears when the glass is taken out of the path. Show me that it is not the glass. That is the demand a careful experimentalist would make, it is answerable, and it is the only version of this argument worth answering.

Why it doesn’t save the claim.

Two replies, and the second does not depend on the first.

First: the mechanism is exempted by its own author. The defect he describes acts off the axis — “The infinitesimal or mathematical point actually in the centre is, of course, not visibly influenced” — and the measurement he is explaining away is taken by tilting the telescope until the horizon sits at the centre and reading the angle turned through off a graduated circle. The quantity in dispute is a mechanical rotation, not an image position. He anticipated this and answered it in the same section — the dip is the geometric angle “to which must be added the collimation”, because the cross-hair itself sits at the centre, so “the point absolutely opposite to it is not seen, but only some other point minutely distant from it” — which is why the second reply below is written not to need this one.

Second, and this is where the weight sits: it cannot be the glass, because the residual does three things no property of a tube can do.

It reverses. First, be exact about whose error is being tested. What Parallax calls “collimation” is not the surveyor’s fixed offset between the line of sight and the instrument’s mechanical axis; it is “a slight divergence of the rays of light from the axis of the eye, on passing through the several glasses of the theodolite”, demonstrated with a magnifying glass and radial — spreading outward from the centre, and by his own statement zero at the centre. That gives him a dilemma rather than a defence. If the divergence is symmetric about the axis it displaces nothing on the axis, which is the first reply above. If instead the residual is a small offset in some definite direction — which is what it must be to push the horizon off the cross-hair at all, and what his decentred-lens diagram describes — then it is a direction fixed in the instrument, and a direction fixed in the instrument is exactly what the two-face procedure kills. Transit the telescope, rotate 180° in azimuth, and the offset changes sign; the mean of the two faces is free of it. Wikipedia’s summary of standard theodolite practice states the rule plainly — “By measuring the same horizontal and vertical angles in these two modes and then averaging the results, centering and collimating errors in the instrument can be eliminated.” The dip does not reverse, because it is fixed by gravity and geometry. Face-left/face-right observing is not a modern refinement invented to answer Rowbotham; it is why collimation has a name.

It cannot grow as the square root of height. Dip goes as sqrt(h): the same instrument carried from 2 m to 100 m must read seven times more, from a beach to 35,000 ft must read from 2.5 arcminutes to 3.3 degrees — a factor of eighty. Nothing about a tube, a lens or a pinhole knows how high it has been carried. Rowbotham himself set up this exact test, at Experiment 15, and ran it over three floors of one hotel, where the predicted change is a few arcminutes.

It cannot grow with a triangle’s area. He offers one cause for two effects, and the second convicts it. Spherical excess is A/R² — recomputed here, equilateral triangles of 30, 60 and 100 km side give 1.98″, 7.92″ and 22.0″. He quotes the rule himself, from Castle’s Treatise on Levelling (“divide the area of the triangle in feet by the radius of the earth in seconds”), and quotes the behaviour himself (“This excess is inappreciable in common cases, but in the larger triangles it becomes necessary to allow for it”), immediately before proposing a fixed optical defect as its cause. A defect in the glass does not switch on when the triangle gets bigger. And his supporting detail runs the wrong way for him: he notes that the Survey repeated angles “upwards of a hundred times, in order to diminish the personal and instrumental errors” — repetition beats down random error, so a residual that survives a hundred repetitions and scales with enclosed area is behaving like a geometric quantity, which is precisely what he needs it not to be.

Start with the concessions, because there are two and both are permanent.

First: the horizon looks flat, and it is supposed to. The authority here is not a debunking site but Lynch in Applied Optics, whose finding is that the minimum altitude at which the curvature of the horizon can be detected is “at or slightly below 35,000 ft”, with a wide field of view and clear air — and who warns in the same paper that “virtually all camera lenses project an image that suffers from barrel distortion”, which bows a horizon down when it sits above frame centre and up when it sits below, in everyone’s footage, in both directions. Item 43, taken by itself, reports something true.

Second: the dip is not a constant. The geometric value is 1.93′ × sqrt(h in metres); the value used in practice is Young’s 1.75′, the difference being ordinary refraction at a coefficient of about one sixth. But Young also records that the effective k “varies from moderate negative values (on sunny days) to values considerably larger than 1 (at night, or for sight-lines over cold water)”, and that when the observer is inside a duct “a pseudo-horizon appears above the astronomical one, so the dip of this apparent horizon is negative — a remarkable phenomenon that really is observed, occasionally.” So a single low-altitude naked-eye sighting of the horizon settles nothing, and that cuts against demonstrations offered in our favour exactly as hard as it cuts against Rowbotham’s.

1. What the verdict ranges over

Not “the horizon looks curved.” The proposition is the conjunction the cluster name states, and its operative half is the second: that the horizon rises to eye level — that the dip is zero. That is a claim about a measured quantity, the quantity is not zero, and the man the claim is attributed to measured it himself before explaining it away.

2. The claim actually on the table is about lenses, not about looking

By the enlarged edition Rowbotham has a section answering Herschel’s dip-sector argument, and its key sentence carries a restriction the list never inherits: “unless some telescopic instrument is used no dip whatever is required to meet the sea horizon.” Under THEODOLITE TANGENT he concedes the instrumental result in as many words: “The author has made experiments similar to the above, and found it to be as stated; but the cause is not that the line of sight is a tangent, but the same ‘collimation’.” And at Experiment 11 he goes further than a hypothesis — he builds an 18-inch iron tube with a pinhole and cross-hairs, attaches a spirit level, points it at the sea from height, and reports that the cross-hair “was seen to cut or to fall close to the horizon”.

So the argument is: the dip is an artefact of the glasses, and here is the instrument without glasses. Three answers, in increasing order of how little they depend on him.

(a) His own optics appear to exempt the measurement — and he knew it. The defect he names acts away from the axis. His demonstration is a magnifying glass held over a drawn line, and his conclusion about it is explicit: “The infinitesimal or mathematical point actually in the centre is, of course, not visibly influenced, being in the very centre or on the true axis of the eye.” But the dip is not read as an image displacement. It is read, in his own description, by inclining the instrument “until the cross-hair touches the horizon” and taking the angle turned through. That puts the horizon on the axis and converts the question into a mechanical rotation on a graduated circle.

He saw the objection. In the same section he writes that the dip required to bring cross-hair and horizon together is the geometric angle “to which must be added the collimation”, and repeats it — the angle observed is “always A, T, S, plus collimation”. His reason arrives two sentences after the exemption just quoted: “the spider’s web of which the cross hair is made is placed in the actual centre; hence, in an observation, the point absolutely opposite to it is not seen, but only some other point minutely distant from it” (enlarged ed., pp. 203–204). That answer is available to him, so (a) is a pressure point and not a kill. It is also why (c) below is written not to depend on it: whatever residual the spider-line leaves is a fixed property of one instrument.

(b) The null instrument is null at the wrong precision. Take the tube entirely at face value and ask what it could have resolved. It is 18 inches — 457 mm — long, and the eye looks through “a very small aperture”. Eye-position freedom inside an aperture of 1 mm is already ±3.8 arcminutes of pointing; inside half a millimetre, ±1.9. Against that, the dip at his Brighton pier is 6.2 arcminutes and at a hundred feet 10.6. The reference is a spirit level “carefully adjusted”, and the passage as printed in the reproduction read here (Experiment 11, pp. 40–41) states no height of eye, no angular value and no repeat count — the result is reported as the cross-hair cutting or falling close to the horizon. A null with that phrasing, from that apparatus, at those heights, is consistent with a dip of several arcminutes. It is not evidence that the dip is zero; it is an instrument whose resolution is the same size as the thing it was built to detect.

(c) And none of that is needed, because of what the residual does. It reverses when the telescope is transited and the instrument turned through 180° — the standard two-face procedure that exists precisely to cancel collimation — and the dip does not. His version of the defect is a radial spreading through the glasses rather than the surveyor’s line-of-sight offset, but that does not exempt it: a spreading that is symmetric about the axis moves nothing that sits on the axis, and any residual that does displace the horizon while the cross-hair is on it points some definite way in the instrument, which is the thing two-face observing averages out. It grows as sqrt(h), from 2.5 arcminutes at a standing observer’s eye to 3.31° at 35,000 ft and 5.75° at the “20+ miles” the modern form of this claim ranges over: a factor of eighty across a range no property of a tube can track. And the same single cause is offered for spherical excess, which grows with the area of a triangle — about 2″ on a 30 km triangle, 22″ on a 100 km one — while the authority Rowbotham quotes on the same page puts the survey’s total angular error at “never exceeded three seconds”.

3. The angle that was never taken

Here is the argument’s load-bearing member, and it is arithmetic anyone can check. Discussing the balloon two miles up, Rowbotham writes that on a globe “the horizon would have been 127 miles away” and “over 20,000 feet below the line of sight”, and concludes that the dip “would be so great that the aëronaut could not fail to observe it.” Both of his figures are right — a sphere of radius 6,371 km puts the horizon at 125.85 miles and the drop below the eye’s horizontal plane at 2h = 21,120 ft. The conclusion does not follow from them:

arctan(21,120 ft ÷ 125.85 miles) = 1.82°

Under two degrees, judged from a swinging basket with no levelled reference, no artificial horizon and no instrument he would have trusted. Searching the Experiment 10–11 region of the text read here returns no angular value for the dip at all: the two legs are printed and the ratio is left untaken. The same gap governs the ground-level work. At the Brighton pier his quadrant stood 34 feet up and he computes, correctly, that the horizon would lie 68 feet below the line of sight at rather more than seven miles; as an angle that is 6.2 arcminutes, a fifth of the Moon’s width, to be caught by eye over the top edge of a wooden quadrant.

4. The levelled board does not test what it is said to test

Experiment 7 — a 6-to-12-foot board, levelled on tripods, sighted over towards the sea — survives into Dubay’s proof 60 down to the same 66-foot figure. Both state the globe’s prediction the same way: the horizon would “gradually decline from the centre” of the board, sitting over 66 feet lower at ten miles to each side.

That is a mis-derivation. 66.7 feet is the drop of the surface at ten miles measured sideways from the tangent point directly ahead — and that place is past the horizon, which is why it is not visible. The horizon of a sphere is the locus of tangent points: a circle at the same angular depression in every azimuth, to the precision at issue here over open sea. A levelled straight edge at eye height likewise subtends zero elevation at every azimuth along its length. So compared azimuth by azimuth — which is how both texts describe the procedure, Rowbotham directing the eye “in an angular direction to the left and to the right” — the sphere predicts the horizon parallel to the upper edge and uniformly below it.

Now be precise about the residual, because there is one and it is small. Take the whole span in at one fixed gaze, or photograph it, and the flat image plane adds an extra sag of d(sec φ − 1) towards the ends — the geometric root of the apparent curvature Lynch analyses. For a 20-mile trace from a 200-foot eye that is about 2.9 arcminutes at each end. And Rowbotham’s own 66 feet, converted across the ten-to-twenty-mile traces both texts name, is between about 2.2 and 4.3 arcminutes. The two predictions are an arcminute or two apart. A wooden plank and an unaided eye cannot separate them, which is the real verdict on the experiment: not that it refutes a caricature, but that it cannot decide the question it is offered to decide, while being reported as though it were conclusive — “This would be impossible if the earth were a globe.”

And notice what both texts actually report. Rowbotham: the horizon “will be observed to run perfectly parallel with its upper edge.” Dubay: it “will always align perfectly parallel with the upper edge of the board.” Parallel, not coincident. The quantity that would settle this is the offset between the edge and the horizon — the dip — and the apparatus as described is used to check parallelism and not to measure the offset. Done as a dip measurement instead, the same rig becomes decisive at altitude: at Rowbotham’s 34 feet the offset is 6.2 arcminutes, about five millimetres of eye position at three metres behind the board and hopeless by hand; at 10,000 feet it is 1.77°, which is nine centimetres at the same three metres; at the “20+ miles” the modern claim ranges over it is 5.75°.

5. Two horizon distances of their own, and only the larger one decides anything

Carpenter supplies them sixty-six-fold apart. The small one has been offered as a clean check on this page before; it is not one, and the reason is worth setting out before the one that is.

Proof 32 does not discriminate, and it is worth saying so before anyone leans on it. On level ground “the horizon is formed at about three miles all around us”, and in the same sentence objects “no higher than we stand—say, six feet—and which are at that distance (three miles)” have reached the vanishing point. On a sphere, sqrt(2Rh) at h = 6 ft is 3.00 miles. But it is equally the zetetic tradition’s own rule, which Parallax states quantitatively and endorses — quoting Mayhew’s Wonders of Science, p. 357 — that “when an object is removed from the eye 3000 times its own diameter, it will only just be distinguishable”, and then adopting it in his own voice: “The above may be called the law of perspective.” (enlarged ed., p. 160). That is the rule he uses to place the sea horizon: the eye-line and the water are the eye’s height apart, and they meet where that separation has shrunk to a minute of arc. Three thousand times six feet is 3.4 miles. Against Carpenter’s “about three miles” the two laws are indistinguishable, and they are indistinguishable for a structural reason: sqrt(2Rh) and 3000h cross at an eye height of about 1.4 m and are still within half a mile of each other at the six feet he names. The check is being made at the one height where the rival predictions agree.

Proof 6 is where they part. Their law is linear in the height of the eye; the sphere’s goes as its square root. Carpenter reports that from Coxwell’s balloon the horizon is at eye level “though it be two-hundred miles away, as seen by Mr. J. Glaisher”. A 200-mile horizon on a sphere requires an altitude of 8.13 km, 26,675 feet — inside the range of the 5 September 1862 ascent he is citing (Glaisher lost consciousness around 28,900 ft; estimates for the peak run from 31,200 to 35,800 ft). The same three-thousand rule, given an eye 26,675 feet above the plane, puts the horizon some fifteen thousand miles away. Carpenter reports two hundred. Run their rule the other way and it is worse: 200 miles of horizon needs an eye only 352 feet up, so on their own law Glaisher should have reached that view from a tall building and then watched it run out to thousands of miles as he climbed.

And the height-dependence is theirs, not something imputed to them. Carpenter, in the same proof: “no matter how high we ascend above the level of the sea, the horizon rises on and still on as we rise”, and from a hill “such as Federal Hill, Baltimore, we may see twenty-five miles away”. Parallax, in his stated second proposition, makes the vanishing distance grow with the separation of the two things converging — “the further any two bodies, or any two parts of the same body, are asunder, the further must they recede before they appear to converge to the same point” — and says of the sea horizon that it forms “at a certain distance, less or greater, according to the elevation of the observer”. So the range does grow with height on their account. It grows in direct proportion to the height, not as its square root, and that is the difference their own reported numbers decide against them.

6. The experiment the argument needed, and the one place it was nearly done

Dip is the discriminator, and dip is a routine correction rather than an exotic measurement: the altitude of a body taken with a sextant against the visible sea horizon has the dip for the observer’s height of eye taken out of it before the sight is worked, and the value used in practice is Young’s 1.75′ × sqrt(h in metres). From a bridge 15 m above the water that is about 7′, and 7′ of latitude is seven nautical miles of position. If the horizon stood at eye level the correction would be zero and every fix worked from a height of eye would be wrong by a distance the navigator would meet on approach.

Rowbotham designed the right test and under-ran it. Experiment 15 takes a levelled clinometer to successive floors of the Grand Hotel at Brighton — exactly the arrangement that exposes the square-root law, since a lens defect cannot know which floor it is on. Over the storeys of one hotel the predicted change is a few arcminutes. Carried instead from the beach to a high cliff, or to a balloon, or to the 35,000 ft where the modern version of the claim is made, the same instrument must read 3.3 degrees. That is the measurement the argument has needed since 1865, and the altitudes at which it is now asserted are exactly the altitudes at which it would be easy.

7. What is left, stated without decoration

Two of the three items report true observations. The horizon does look flat below about 35,000 feet, and it does sit within a few arcminutes of eye level at any height a person can stand on; both are predictions of a sphere 6,371 km in radius and neither discriminates. The third claim — that no dip exists — is false, was known by the author of the argument to be false of instrumental observation, and was assigned by him to a lens defect that his own account of it exempts from his own method, that reverses when the instrument is reversed, that cannot grow as the square root of the height, and that cannot grow with the area of a triangle. What the unaided eye cannot resolve, a graduated circle can; the reading grows with height in the way the geometry says and in no way an instrument could imitate; and the world’s navigators subtract it before breakfast.

The list says

Horizon rises to eye level.

The source says

“At every altitude where special observations have been made, the sea surface has been found to ascend to a line of sight at right angles to a plumb line; and that unless some telescopic instrument is used no dip whatever is required to meet the sea horizon.” (Zetetic Astronomy, enlarged ed., THE “DIP SECTOR”)

A narrow case was generalised. The clause that goes missing is “unless some telescopic instrument is used”. In the source the no-dip claim is about unaided sight. The instrumental case is handled in its own sections, conceded — “The author has made experiments similar to the above, and found it to be as stated” — and blamed on collimation in the telescope. Item 44 carries the assertion without the restriction, so a reader meets as a universal claim what the source stated for one class of observation and argued for in the other.

What else is lost is the apparatus. The source’s evidence for “no dip” is not the bare sentence: it is an 18-inch iron tube with a pinhole, cross-hairs and a spirit level, reported at Experiment 11 as showing the cross-hair cutting “or falling close to” the horizon. The compressed item keeps the conclusion and drops the experiment, which is the opposite of the usual complaint about this literature and worth saying plainly: the list is thinner than its own source, not merely firmer.

The widening is not idle, because the modern evidence base is the excluded class. The nearest located antecedent of item 43 is Dubay’s proof 1 — the horizon “always appears perfectly flat 360 degrees around the observer regardless of altitude” — whose support is amateur balloon, rocket, plane and drone footage. That is lens-mediated observation, and lens-mediated observation is precisely what the founding text says manufactures a false horizon. The position has travelled with its escape clause deleted, and the modern version now rests its case on instruments its own founder needed to be untrustworthy.

Where the specimen is the more cautious document, say so. Item 43 says “in long-distance views” where Dubay says “regardless of altitude”, and item 44 drops Dubay’s “so you never have to look down to see it”. On this cluster the sharpest drift happened earlier in the chain, between the enlarged Rowbotham and the 2015 list, not at the specimen. Item 111 — “Horizon tests flat” — is four words naming no apparatus; its nearest located antecedent is Dubay’s proof 60, which is Rowbotham’s Experiment 7 reworded down to the same 66-foot figure. That identification rests on the match between those two texts and on nothing in the item’s own wording.

scope_widened is recorded because the operative change is that a claim stated for the unaided eye is applied to instrument and camera observation. hedge_dropped fits the same sentence — “unless X” is a qualification and it is gone — and a reader who prefers that reading is not wrong; both texts are printed above so the choice can be checked rather than taken on trust.

The refutation answers the source, not the fragment: it grants the flat-looking horizon at full strength on Lynch’s authority, grants that the dip is beneath unaided judgement at every height Rowbotham worked at, takes the iron tube seriously enough to compute what it could have resolved, and then puts the weight where no property of any tube can follow — the sign reversal under two-face observing, the square-root growth with height, and the area scaling of spherical excess.

The fallacy involved in all the attempts to prove the oblate spheroidal form of the earth, is, that the earth is first assumed to be a globe, the celestial surface above it to be concave, and the plumb-lines to be radii. If this were the true condition of things, then all the degrees of latitude would be the same in length; and if the earth were really “flattened at the poles,” the degrees would certainly shorten in going from the equator towards the north. If, however, the celestial surface is not concave, but horizontal, two plumb-lines suspended north and south of each other would be parallel, and would indicate equal length in all the degrees of latitude, thereby shewing the earth to be parallel with the celestial surface, and therefore a plane. The differences required by a globe are not found in practice, but such as a plane would produce are invariably found. Hence the failure of geodesy becomes evidence against rotundity, but demonstrating that the earth is parallel to the horizontal heavens, and therefore of mathematical and logical necessity A plane.

— Earth Not a Globe (1865), 3rd ed., rev. and enl. (London: Day, 1881), ch. XIV “Examination of the So-Called ‘Proofs’ of the Earth's Rotundity”, section “Arcs of the Meridian”, printed p. 245 as paginated at sacred-texts za40.htm, whose “[p. N]” markers stand at the head of page N. The section opens at p. 241; the von Gumpach extract quoted in the straw-man note is at pp. 242–243 and the degree table at p. 244. Not the 1865 first book edition — see the gloss. A reformatted archive.org scan of the third edition, which numbers pages at the foot, carries the same section at pp. 187–189; the two paginations are not interchangeable, and neither transcription was checked against print

Read the last three sentences before deciding this is a hypothetical. The argument opens on a condition — if the celestial surface is horizontal — and it would be convenient for us to stop there and say the source never asserted anything. It does. The condition is followed immediately by a flat empirical claim (“The differences required by a globe are not found in practice”) and a flat conclusion (“of mathematical and logical necessity A plane”). Rowbotham is not hedging and this page does not pretend he is.

What he is arguing from is a table, not a plumb line. The section is about arcs of the meridian, and at its centre is a table of the British Ordnance Survey’s degree lengths, eight arcs paired from five stations — Clifton, Arbury Hill, Blenheim, Greenwich and Dunnose — quoted from Hugh Murray’s Encyclopaedia of Geography. Two remarks of Murray’s travel with it and they are about two different failures. Before the table, about a degree measured perpendicular to the meridian between Beachy Head and Dunnose: the polar-to-equatorial axis ratio computed from it “differed considerably from that obtained by meridional degrees. It has been found impossible to explain the want of agreement in a satisfactory way.” After the table, about the tabulated meridional degrees themselves: “instead of the degrees increasing as we proceed from north to south, they appear to decrease, as if the earth were an oblong instead of an oblate spheroid.” It is the second that Rowbotham builds on; the first is a separate concession about a separate quantity, and this page does not pool them. The parallel plumb lines are an inference from the tabulated discrepancy. Neither this passage nor Carpenter’s reports hanging a wire and measuring one, and the refutation below is aimed at the inference rather than at a measurement they did not make.

The edition matters. This passage is in the third edition of 1881. It is not located in the 1865 first book edition: the Project Gutenberg text of that edition (#69892) was searched for plumb, perpendicular, wall, building, spire and tower, and its three occurrences of “plumb” are all instrumental — the Encyclopædia Britannica “Levelling” extract that defines the level as a line crossing the plumb-line at right angles, an air-gun “carefully adjusted by a plumb-line”, and a mirror at Plymouth Hoe “fixed, by the aid of a plumb-line, in a true vertical position”. The 1865 text carries the arcs material and the “oblong” quotation; what 1881 adds is the von Gumpach extract and the plumb-lines-are-parallel inference drawn from them. Both online transcriptions of the third edition read “spewing” in the sentence above, an OCR slip for “shewing”.

The other half of the cluster is Carpenter’s, and it is a different claim. Items 396 and 397 — architecture and skyscrapers — match One Hundred Proofs that the Earth Is Not a Globe (Baltimore, 1885), proof 72, indexed by its author as “Walls not parallel!”: “Astronomers tell us that, in consequence of the Earth’s ‘rotundity,’ the perpendicular walls of buildings are, nowhere, parallel, and that even the walls of houses on opposite sides of a street are not strictly so! But, since all observation fails to find any evidence of this want of parallelism which theory demands, the idea must be renounced as being absurd and in opposition to all well-known facts.” Note what that is: an accurate statement of the globe’s prediction, followed by a claim that the prediction is not observable. At his scale it was not. Across a twenty-metre street the two walls lean apart by 0.65 of an arcsecond.

What these passages are being cited as. The earliest texts located carrying the two halves of this cluster in the form the list uses. They are ancestors and that is all — no earlier flat-earth or geocentric text was traced that puts either claim first, and neither man is credited here with originating it.

SURFACE (weak — do not use). “Plumb lines obviously converge on the centre; this is trivially false.” It walks straight into Carpenter’s actual point. He never denied what the theory predicts; he denied that anybody had seen it. Across the width of a street the predicted lean is 0.65 arcsec, far below what a mason’s plumb bob, a builder’s square or a spirit level resolves. Answering “but the theory says so” to a man complaining that the theory’s prediction has never been observed is answering nothing.

DEEPER. The argument is non-discriminating: “plumb” means along local gravity and “level” means perpendicular to it, so a builder who works plumb and level has measured the direction of gravity at one site and nothing else. True, and it disposes of items 396 and 397 — but it does not touch Rowbotham, whose claim is not about definitions at all. His is an inference from published survey data, and the data really were discordant.

KERNEL. Two genuinely true things sit underneath this cluster. The first: the British arc measurements came out wrong, and a standard reference work of the day said so. The tabulated degrees decreased going north, which is the signature of a prolate Earth and the opposite of what flattening at the poles requires; and the same source records a second, separate failure a page earlier — that the axis ratio got from a degree measured perpendicular to the meridian “differed considerably” from the ratio got from meridional degrees, and that “It has been found impossible to explain the want of agreement in a satisfactory way.” Rowbotham invented neither; he found both in print and quoted them accurately. The second, which is sharper: the plumb line genuinely does not point at the centre of the Earth, and geodesy genuinely does substitute a computed ellipsoid normal for the observed vertical. Rowbotham prints von Gumpach making exactly that charge — that astronomy “gives to the plumb-lines such imaginary directions as are needed in order to adapt the empirical results of geodetic measurements to the earth’s imagined form.” The distinction between the direction a plumb bob actually hangs and the direction the reference figure says it should is real, has a name, and is the subject of a whole sub-discipline.

Why it doesn’t save the claim.

Because both true things run on the same quantity, and that quantity exists only if the plumb line moves.

An astronomical latitude is the angle between the plumb line at a station and the celestial pole; a table of degree lengths is built out of differences between such angles. So every number Rowbotham argues from is a measurement of how much the local vertical has turned between two stations — and between the ends of his own table, Dunnose on the Isle of Wight and Clifton in Yorkshire, it has turned by 2° 50′. As for the scatter that he reads as the failure of geodesy: it is 124 fathoms in 60,800, or 0.20%, and the shortest arc in the table is Arbury Hill to Greenwich, 44′ 48″ or 2,688 arcseconds, of which 0.20% is about 5.4 arcseconds of latitude. That is the difference between two ordinary station deflections — the 1911 Encyclopædia Britannica computed them for this same English survey and found six of sixteen stations under 2″, six between 2″ and 4″, four above 4″. We located no published reanalysis of Rowbotham’s particular table, so that is a statement about magnitudes and not an attribution of cause.

So the argument runs on a quantity that exists only because plumb lines are not parallel. Take the non-parallelism away and there is no station error, no discordance between arcs, and nothing for the passage to be about. The strongest thing in the cluster — von Gumpach’s complaint that the plumb line does not point where the ellipsoid says — is a complaint that the local vertical wanders from place to place. That is the claim being denied.

The concession first, because it is real and because the easy version of this rebuttal loses to it. At the scale Carpenter names, he was right that nothing was observable. Two plumb lines a distance d apart converge by d/R radians. Across a twenty-metre street that is 3.1 × 10⁻⁶ rad, or 0.65 arcseconds; across a hundred-metre city block, 3.2 arcseconds. No mason, spirit level or Victorian theodolite was going to find that in a party wall, and saying “the theory predicts it” to a man whose objection is that the prediction has never been seen concedes his point rather than answering it. (Recomputed here 2026-08-10 with R = 6,371 km.)

What the verdict ranges over. Not “a plumb line hangs perpendicular to the local level surface” — that is true, on a globe as much as on a plane, and it is what a builder means by plumb. The cluster’s claim is that plumb lines are parallel to each other, everywhere, and that this is evidence for a plane. Both halves fail, and they fail differently: the parallelism is contradicted by direct measurement at scales where it is big enough to matter, and the inference Rowbotham draws it from is self-defeating on his own optics.

1. Where the effect is large enough to matter, it is measured and built for

Go from a street to a strait. The Verrazzano-Narrows Bridge carries its towers 4,260 ft (1,298 m) apart and 693 ft (211 m) high. The two verticals at those foundations differ by 42 arcseconds, which over the tower height is 43 mm. The Metropolitan Transportation Authority, which owns and operates the bridge, puts it in its own public description: the towers are “1 5/8 inches farther apart at their tops than at their bases because the 4,260 foot distance between them made it necessary to compensate for the earth’s curvature”. The standard description of the structure states the conclusion in Carpenter’s own words and the opposite sense: “The towers are not parallel to each other.” The Humber Bridge, 1,410 m between towers and 155.5 m high, is described the same way — “although vertical, [the towers] are 1.4 inches (36 mm) farther apart at the top than the bottom due to the curvature of the Earth.” Our arithmetic gives 43.0 mm and 34.4 mm against the published 41.3 mm and 36 mm: agreement to within about 5%, the residue being the local radius of curvature and where the base is reckoned from. This is not a debating point produced for a website. It is a dimension on a drawing, and the two structures were built to it.

The same correction turns up wherever a straight line is long. LIGO’s public description of its own construction: “Over the 4km length of each arm, the Earth curves away by nearly a meter!” — and the concrete slab under the beam tube had to be precision-levelled so that a laser leaving the corner station in a straight line “strikes the test mass/mirror at the end of each arm, and not a meter above it.” The sagitta of a 4 km chord is 1.26 m. An instrument built to detect a strain of one part in 10²¹ is aligned around the fact that the local vertical rotates by two arcminutes over 4 km.

2. The direction of a plumb line against the stars is what latitude is

Hang a plumb bob, or level a bubble, which is the same measurement made flat: you have fixed the local vertical. Now measure the angle from it to a star. Measured against the celestial pole, that angle is your astronomical latitude. Travel and it changes: the local vertical swings by one degree for every 111 km of ground you cross, in any direction on the surface, and by one degree of latitude for every 111 km of northing. The unit sailors use encodes it — the nautical mile is one minute of latitude, 1,853 m, the distance over which the plumb line turns by one arcminute relative to the sky. If plumb lines were parallel worldwide and the sky were far off, that angle would be the same everywhere, latitude would not exist as a measurable quantity, and there would be no degrees of the meridian to tabulate.

Which is Rowbotham’s problem, because the tabulated degrees are his evidence. His premise is that the celestial surface is “not concave, but horizontal” — a sky parallel to the plane. Take that premise seriously in either of its two available forms:

• If the sky is far enough away that its rays arrive parallel, the angle between a plumb line and a star is identical at every station. Polaris stands at the same altitude in London and in Ceylon. There is no latitude, no arc, and no table — and the quantity Rowbotham spends the whole section arguing about cannot be measured at all.
• If the sky is at a finite height h — which is his actual position, set out earlier in the same chapter under the heading “Declination of the Pole Star” (pp. 230–232), where Polaris sinking towards the equator is “an ordinary effect of perspective” like a receding row of lamp-posts — then a station seeing Polaris at altitude α stands at ground distance h cot α from beneath it, and one degree of “latitude” occupies a ground length proportional to 1/sin²α. At α = 51° that is 0.029h; at α = 10° it is 0.579h. A degree near the equator would be twenty times longer than a degree in England, and lengthening fast as you go south. The measured degree runs from 110.574 km at the equator to 111.694 km at the pole — about one per cent, and in the other direction.

So the sentence “two plumb-lines suspended north and south of each other would be parallel, and would indicate equal length in all the degrees of latitude” does not follow from either version of his own sky. On the first there are no degrees; on the second they are grossly unequal. The globe predicts a one-per-cent lengthening polewards and that is what the surveys find, to the point where the residual disagreements are small enough to be diagnostic of something else.

3. That something else is the size of the plumb line’s own wandering — which was not invented to save the ellipsoid

Rowbotham’s anomaly is genuine: the tabulated British degrees shorten going north, an oblong figure, and the reference work he quotes prints that observation — and, of a separate result on the page before, that the axis ratio from a degree measured perpendicular to the meridian “differed considerably from that obtained by meridional degrees. It has been found impossible to explain the want of agreement in a satisfactory way.” What his source does not do is say what produced the northward decrease, and we should be equally careful: we found no published reanalysis of this particular table, so what follows is an argument about magnitudes, not an attribution of cause.

The candidate is the one geodesy already had. The astronomical latitude of a station is read against a plumb line that the mass around it pulls sideways, and the 1911 Encyclopædia Britannica gives the computed numbers for the English survey specifically: sixteen stations, six deflected by under 2 arcseconds by the form of the ground, six between 2″ and 4″, four above 4″, with one exceptional station on the Banffshire coast reaching 10″. It remarks that failing to recognise this “often led to much perplexity in the early history of geodesy”, and it states the sensitivity outright: comparing two degree-arcs about five degrees of latitude apart, “if ε be so great as 2″ the probable error of the resulting ellipticity will be greater than the ellipticity itself”, ε being the probable error of an observed latitude. Now measure Rowbotham’s table against that. Its spread is 124 fathoms in 60,800, or 0.20%; its shortest arc is Arbury Hill to Greenwich, 44′ 48″ or 2,688 arcseconds; and 0.20% of that arc is about 5.4 arcseconds of latitude — the difference between two ordinary station deflections. The whole failure of geodesy he is reporting is the size of the effect he is arguing does not exist.

Von Gumpach’s reply, which Rowbotham prints, is that this is circular: astronomers assign the plumb lines “such imaginary directions as are needed” to fit the shape they have assumed. It is the best objection in the cluster and it has three answers.

First, the theory made a prediction and the prediction failed. Everest’s Great Trigonometrical Survey found the latitude difference between Kaliana and Kalianpur 5.24 arcseconds smaller by triangulation than by astronomy. J. H. Pratt, in a paper read on 7 December 1854 and published in 1855, computed forward from the visible mass of the Himalaya what deflection it ought to produce, and got 15.885 arcseconds — more than three times the observed value. A quantity invented to absorb discrepancies does not overshoot by a factor of three. What followed was Airy’s three-page reply of 1855 proposing compensating light material at depth, which is the origin of isostasy, and whose crustal roots were later seen independently by seismology. (The same G. B. Airy as ARG-A03, and the same year.)

Second, the direction of the plumb line is now measured, not assigned. The US National Geodetic Survey’s deflection-of-the-vertical surveys set a camera on a benchmark, align it precisely to the local plumb line, and compare the observed star field with the positions expected from a GNSS-derived location and time. Accuracy: 0.1 arcsecond. Observed deflections: “many 10s of arc-seconds”. The agency’s own description of the method notes that while GNSS supplies position and time, “the slope is determined solely by the camera system”. That is the plumb bob and the sky, and nothing else.

Third, independent instruments agree. A 2016 comparison in Sensors ran three unrelated methods at the same sites — an astro-geodetic zenith camera, GNSS heights differenced against spirit levelling, and a gravimetric geoid computed from measured gravity — and reports: “The three methods are in excellent agreement, with an operational supremacy of the astro-geodetic method, being faster and more precise than the others.” The two numbers usually quoted from it are each method’s own precision, not a disagreement between methods: the GNSS-plus-levelling combination “has slightly larger standard deviations; although well within the 1 arcsec level, which was assumed as threshold”, while “the geoid model based method, whose 2.5 arcsec standard deviations exceed this threshold, is also statistically consistent with the others”. A star camera, a levelling staff and a gravimeter are three different physical measurements — starlight, height differences, and the pull of gravity itself. They return the same tilt.

4. Architecture and skyscrapers: a measurement of one point

Items 396 and 397 need less. A builder’s plumb is defined as the direction a weight hangs at that site and level as the surface perpendicular to it; a skyscraper is erected to a single vertical established at a single foundation. Nothing about that procedure compares the vertical at one site with the vertical at another, so nothing about it can show they are parallel — it is the convenience-frame move of ARG-R08 in mortar rather than in software. The place where two verticals do have to be reconciled is a structure long enough to stand on both, and there, as section 1 shows, the answer is on the drawings. And the stability of a tall building follows from each floor being level with respect to gravity where it is, which is a curved family of surfaces, not a stack of parallel planes — the same point ARG-B06 makes about the surveyor’s word level, which the Encyclopædia Britannica article Rowbotham himself reprints defines as “the curve of the Earth being the true level, and the tangent to it the apparent level.”

5. What is left

Carpenter’s observation was correct for his instruments and is now false for ours: the non-parallelism he could not find is 41 millimetres of designed spread between two bridge towers, and a tenth of an arcsecond is the routine precision with which the local vertical is now measured against the stars. Rowbotham’s anomaly was real, and it is the size of the effect he was arguing against. Neither text observed a plumb line; both argued from what could not be detected in 1881 and 1885. The claim is not merely unsupported by the modern measurement — it is the quantity the modern measurement reports, with a value, at every benchmark.

Somebody did hang the wires, and the answer is arithmetic rather than dismissal. At the Tamarack mine at Calumet, Michigan in 1901–02, plumb lines 4,250 ft long and 15 to 16 ft apart were suspended down the shafts and came out farther apart at the bottom; F. W. McNair repeated the work across three shafts and published it in Science XV on 20 June 1902, recording deviations from −0.028 to +0.141 ft with a stated measurement error “not greater than 0.003 feet”, and traced the divergence to convection currents in the shafts. Put the expected signal beside those numbers: two lines 15 ft apart converging over 4,250 ft of depth should close by 15 × 4,250/R = 0.0031 ft, about 0.9 millimetres. That is the same size as his own error bar and between a ninth and a forty-fifth of the deviations he actually measured. A shaft with air moving in it could not have resolved the effect in either direction — which is a statement about the apparatus, not about the men, whose measurements were careful and honestly reported. The place the effect is resolvable is the one section 1 describes: a structure long enough to stand on two verticals at once.

The list says

Plumb lines perpendicular worldwide.

The source says

“If, however, the celestial surface is not concave, but horizontal, two plumb-lines suspended north and south of each other would be parallel, and would indicate equal length in all the degrees of latitude … The differences required by a globe are not found in practice, but such as a plane would produce are invariably found.”

A narrow case was generalised. The source’s claim is about two plumb lines, north and south of each other, inferred from one printed table of British meridian arcs running between Dunnose on the Isle of Wight and Clifton in Yorkshire. The item says worldwide. That is the drift — though not in the direction it first appears. Within the survey the two verticals are not nearly parallel either: Dunnose and Clifton lean apart by 2° 50′, some 10,200 arcseconds, and that lean is the very quantity his table records as latitude, so his own evidence presupposes it. What “worldwide” adds is the scale at which two verticals differ by up to 180° and at which bridge towers are built non-parallel on purpose. The list restates the claim at the scale where it is most easily checked, and it fails there as it already failed at his.

Two more things travel with the argument and neither survives into the items. The condition: the sentence is the consequent of “if the celestial surface is not concave, but horizontal” — a premise about the sky, which the item does not carry and which, taken either way, defeats the conclusion it was introduced to support. The evidence: Rowbotham is not reporting an observation of plumb lines but an inference from degree lengths he quotes from Hugh Murray’s Encyclopaedia of Geography, including Murray’s own admission that the disagreement between arcs could not be satisfactorily explained. The item presents as a fact about plumb lines what the source presents as a deduction from a surveying discrepancy. That second gap is a kind the seven-value list has no exact slot for — an inference republished as an observation — and scope_widened is recorded because it is the plainest and most checkable of the three, with both texts side by side above.

Items 396 and 397 drift the same way from a different author. Carpenter’s proof 72 is scoped to walls of buildings and to houses across a street, where the effect is 0.65 arcseconds and his claim that it was not to be found by observation held for the instruments and the building practice of 1885. “Architecture plumb/level universal” and “Skyscraper vertical stable” keep his conclusion and drop the scale it depended on. There is a second, quieter slippage in the item wording itself: Carpenter’s “perpendicular walls” means walls that are each plumb at their own site, which is true on a globe; the item’s “perpendicular…worldwide” is read as all of them being parallel to one another, which is not. The one word carries both claims and only the second is at issue.

The refutation answers the sources, not the fragments: it concedes Carpenter’s observation at Carpenter’s scale in its own voice, quotes Rowbotham through to his flat conclusion rather than stopping at his conditional, and puts the weight on his own inference — that a table of latitude differences cannot exist if the local vertical does not vary, and that the discrepancy in the table is the size a few arcseconds of plumb-line deflection produces.

Size, except in the case of very small stars, may be as easily determined. Let the instrument with which the angular distance was taken be graduated to degrees, minutes and seconds, the minutes and seconds corresponding to miles and sixtieths of miles on the earth's surface. … Instead of the diameter of the moon being 2,160 miles, as we are informed by the men of science of to-day, it is, by the above process, found to be about 32 nautical miles in diameter. [Here the quotation jumps forty-nine pages, from the Moon on printed p. 71 to the section “Sun's Diameter” at printed p. 120:] If the navigator neglects to apply the sun's semi-diameter to his observation at sea, he is 16 nautical miles (nearly) out in calculating the position his ship is in. A minute of arc on the sextant represents a nautical mile, and if the semi-diameter be 16 miles, the diameter is of course 32 miles. And as measured by the sextant, the sun's diameter is 32 minutes of arc, that is 32 nautical miles in diameter. Let him disprove this who can.

— Zetetic Cosmogony; or, Conclusive evidence that the world is not a rotating-revolving-globe, but a stationary plane circle (1899), 2nd ed., enl. (Durban: T. L. Cullingworth, 1899). The method and the Moon at printed p. 71; the Sun at printed pp. 119–120. Pagination follows the printed-page markers in the Google/archive.org scan (item ZeteticCosmogony) and was not checked against a print copy. The p. 120 sentence is also quoted, with that page attribution, by David Wardlaw Scott, Terra Firma (1901), printed pp. 173–174

This is where “equal apparent size” gets its teeth, and it is worth seeing what the equality is made of. Winship runs one procedure twice — sextant to the lower limb, sextant to the upper limb, difference read off as the diameter — and gets 32 nautical miles for the Moon on printed p. 71 and 32 for the Sun on p. 120. The equality is not observed at the end of the process; it is guaranteed at the start of it, because the rule he converts by assigns the same range to everything it is applied to. His own page has the hedge and then removes it: “the moon is next in importance to the sun, if, indeed, she is not quite as large”, and three paragraphs on, the flat figure.

The two items about constancy come from somewhere else. Carpenter, One Hundred Proofs (1885), proof 89, indexed by him as “Luminous objects”: “It is well known that the law regulating the apparent decrease in the size of objects as we leave them in the distance … is very different with luminous bodies from what it is in the case of those which are non-luminous. … Proctor says, in speaking of the Sun: ‘his apparent size does not change,’ —far off or near.” The pamphlet is dedicated to Richard A. Proctor, “The Greatest Astronomer of the Age”, and the sentence about constancy is Proctor’s, quoted inside a rebuttal of Proctor’s objection that a traveller going far enough south would, on a plane, find that the Sun “should therefore look much larger”. Proctor’s own text was not reached for this entry: neither that sentence nor the English Mechanic letter of 20 October 1871 that Carpenter cites was located in any text searched here, so both are reported as Carpenter’s report of Proctor.

Where the modern statement of item 31 lives. Dubay, 200 Proofs (2015), #147: the ball model, he writes, asks us to accept as coincidence what “cannot be explained other than by natural design”, the two bodies having “been measured with sextants to be of equal size and equal distance”. The procedure is Winship’s, and #123 carries the same 32-mile figure for both bodies. Neither proof says so: #123 credits the figure to “Flat-Earthers throughout the ages” and #147 cites nobody, and the only naming of Zetetic Cosmogony located in the archive.org text of 200 Proofs is #31, which is about kite-flying and wind. So the citation trail runs only as far as Scott, who in 1901 quotes the sun passage and gives the page; from there the figure and the sextant appeal travel unattributed, which is how a derivation is usually visible in this literature. The premise doing the work travels the whole way down whether or not the citation does.

Three real things sit underneath these items, and the easy bust misses all of them. First, Winship is describing a genuine nautical procedure correctly. A navigator taking a sun sight really does shoot the lower limb and add the semi-diameter, about 16′, to reach the centre; the tables really do print that correction; and the sextant really does measure the Sun’s angular diameter to better than an arcminute. He is not making up an instrument. Second, the coincidence in item 31 is real and astronomers say so in their own textbooks: two bodies whose distances differ by a factor of about 390 subtend almost the same angle, which is why total solar eclipses look the way they do and why the corona can be seen with the naked eye at all. Treating that as unremarkable would be the strawman. Third, Carpenter’s observation about luminous objects is a real perceptual effect. A street lamp a mile off does not look proportionately smaller than one at fifty yards, because an unresolved bright source is seen as a glare disc whose size is set by scattering in the air and in the eye rather than by geometry. Anyone who has driven at night knows the phenomenon he is pointing at.

Why it doesn’t save the claim.

Each of the three is true, and each points the other way.

The sextant measures an angle, and the conversion to miles is the whole argument. “A minute of arc represents a nautical mile” is a definition about arcs on the Earth’s surface — the nautical mile was fixed as one minute of latitude. Applied to an object at unknown range it is not a measurement but an assumption, and the assumption is recoverable: 1852 m divided by one arcminute in radians (2.90888×10⁻⁴) is 6,367 km, and the Earth’s mean radius is 6,371 km. The rule places every object it touches at one Earth radius. That is why the Sun and the Moon come out the same size and the same distance in his hands: the answer was in the unit. His own numbers show the seam — he puts the Sun 2,700 nautical miles up, and 2,700 multiplied by 32′ in radians is 25.1 nautical miles, not 32.

The coincidence is real, inexact and temporary, and it is silent about the Earth. The Moon’s disc runs 29′26″ to 33′30″ and the Sun’s 31′36″ to 32′42″: the ranges overlap, which is the coincidence, and the Moon’s swing is nearly four times the Sun’s, which is not what two objects at one distance do. The Moon recedes 38 mm a year, so the match is a phase we are living through rather than a fact about the world. And whichever way it is read, it is a claim about the ratio of two objects in the sky, from which nothing about the shape or the motion of the ground follows.

The glare effect is a reason to change instrument, not a law of nature. It is exactly because a bright source resists visual shrinking that the naked eye is the wrong tool and a filtered disc measured against a graticule is the right one — a point the tradition’s founding text concedes, since Rowbotham’s 1865 Section 8 settles the horizon question by appeal to the micrometer. Once the glare is removed there is a number, and the number is the one reported below.

Two claims travel in this cluster and they need separating, because they come from different books and fail in different ways.

1. The two discs really are close to equal. That is the observation, and it is not in dispute.

The Sun’s apparent diameter runs from 31′36″ near 4 July to 32′42″ near 3 January; the Moon’s from 29′26″ at apogee to 33′30″ at perigee. Mean values, computed from the standard sizes and distances — 2×695,700 km over 149,598,000 km, and 2×1,737.4 km over 384,400 km — are 31.97′ for the Sun and 31.07′ for the Moon. So the Moon’s disc is on average the smaller of the two, by about 3%. That is not a quibble: it is why a central eclipse sometimes leaves a ring of photosphere all the way round, which is a direct visual demonstration that the two discs are not the same size. The equality is an overlap of two ranges, not an identity.

It is also passing. The Moon recedes about 38 mm a year, and published estimates of when it will stop being able to cover the Sun range between 650 million and 1.4 billion years from now. A coincidence with an expiry date is a poor foundation for a cosmology.

2. The step the item drops is where the argument lives, and it is a units error.

Nobody disagrees that the two look about the same size, so “equal apparent size” on its own settles nothing about the shape or the motion of the Earth. The source knows this, which is why the source says more: Winship’s sextant procedure turns the angle into miles, and Dubay’s #147 reports the output as the two bodies having been “measured with sextants to be of equal size and equal distance”. That conversion is the argument, and it does not work. One minute of arc corresponds to one nautical mile because the nautical mile was defined as a minute of latitude on the Earth’s surface; the correspondence holds for arcs subtended at the Earth’s centre and for nothing else. Applied to a body at unknown range it fixes that range: 1852 m divided by 2.90888×10⁻⁴ rad is 6,367 km, against an Earth radius of 6,371 km. Every object measured this way comes back the same size and the same distance because the unit put it there. The “equal distance” is not a finding; it is the arithmetic reporting its own assumption.

The two distances are, in fact, separately measured, and they are not equal. Take the cheapest observation: the Moon shows a large daily parallax and the Sun does not. From the same distances above, the horizontal parallax is arcsin(6371/384,400) = 56.97′ for the Moon and arcsin(6371/149,598,000) = 8.8″ for the Sun — a factor of about 390, the same factor as the distance ratio. Widely separated observers see the Moon displaced against the star field, at the same instant, by tens of arcminutes — a shift of the order of its own apparent width — where the Sun’s displacement is under nine arcseconds. That is a measurement of the difference in distance, made with the same class of instrument Winship trusts.

3. “Constant solar diameter” is a perceptual law, propped up by an opponent’s admission.

Carpenter’s proof 89 does not report a measurement of the Sun. It opens by asserting a perceptual law in his own voice and with his own example — “Sail past the light of a small lamp in a row-boat on a dark night, and it will seem to be no smaller when a mile off than it was when close to it” — and only then reaches for Proctor, whose line about the Sun’s apparent size not changing he takes as confirmation: “And then he forgets the fact!” So the constancy is his as much as Proctor’s, and it would be too generous to us to call it borrowed. What the paragraph is for, though, is defence. Proctor’s objection — that a traveller far enough south should, on a plane, find the Sun much larger — is dismissed as resting on “the non-appearance of a thing which has never been known to appear at all”, and therefore as “a counterfeit—a fraud—no valid objection at all”. That is an argument that constancy cannot be used against a plane, not a datum in favour of one. The list keeps four words about the Sun and drops the law, the lamp and the argument they were serving.

And the perceptual law is a reason to reach for an instrument, which is precisely what the tradition’s founding text does. Rowbotham’s 1865 Section 8 disposes of the enlarged horizon Sun by quoting Sir Richard Philips: take the angle “either with a tube or micrometer” and “the measure is identical” at the horizon and at the meridian. So the movement’s first book concedes both the instrument and the result. Filter the glare, measure the disc, and there is a number.

4. The number destroys the near Sun, on the sources’ own figures.

Angular size falls as one over distance. When the Sun sets for an observer it is vertically over a point a quarter of the way around the world; on Winship’s own scale — 45° of latitude equals 2,700 nautical miles — that is a ground distance of at least about 5,400 nautical miles. Put his Sun 2,700 nautical miles up and its setting distance is √(2700²+5400²) = 6,037, so the disc should set at 0.45 of its noon width, near 14′ instead of 32′. Put Rowbotham’s 1881 Sun “less than 700 statute miles” up (ch. V, printed p. 104) and it is worse: the ratio is 0.11 — a setting disc some 3.6′ across, roughly a ninth of the noon Sun and a ninth of the full Moon. Nothing of the kind is seen. Turn that round and it prices the atmospheric rescue: to hold Rowbotham’s 700-mile Sun at its noon width all the way down to the horizon, the damp low air would have to magnify the disc about ninefold. Chapter X of 1881 claims an enlargement and names a mechanism for it; what those two printed pages never supply is a magnitude. The one measurement Rowbotham does print, the Philips micrometer reading in Section 8 of 1865, has the angle identical at horizon and meridian — a magnification of one. The setting Sun measures the same width as the noon Sun to well within a percent, which is what a body 150 million kilometres away predicts: the observer’s own displacement of one Earth radius changes the distance by 0.004%.

Refraction does alter the low Sun, and in the wrong direction for the argument. It lifts the lower limb more than the upper, squashing the disc vertically into the familiar oval, while leaving the horizontal width essentially untouched. So the horizontal measurement is the clean one, and it is flat all day.

5. Through a year the diameter is not constant, and the variation matches the orbit.

31′36″ to 32′42″ is a 3.4% swing, in step with an Earth–Sun distance that varies by about the same fraction: the distance is least in early January and greatest in early July, and the disc is largest and smallest on exactly those dates. That is what an elliptical orbit requires, and it is a measurement a reader can make — a solar filter, a fixed focal length, two photographs six months apart. So the fragment is false at the one timescale where it would be doing work against the globe, and true at the timescale where its truth is a problem for the flat model instead. Both halves of the measurement point the same way.

6. The tradition carries both claims, which is worth knowing before answering either.

What changed between Rowbotham’s editions is narrower than a reversal, and sharper. Both carry the same mechanism. Section 8 of 1865 already has the damp air doing the work — “as the air near the Earth is both more dense and more damp … the light of the Sun must be dilated or enlarged as well as modified in colour” — and then subordinates it to an instrument: “But the enlarged appearance of the Sun when rising and setting is only an optical impression, as proved by actual measurement”, followed by Philips on the tube and the micrometer. The 1881 third edition keeps the mechanism and drops the concession. Chapter X (printed pp. 128–129) is built on “the disc of light projected upon the atmosphere”, drops both the micrometer sentence and the Philips citation, and promotes what is left to “a striking argument against the rotundity of the earth”, on the ground that “the atmosphere surrounding a globe would not permit of anything like the same degree of enlargement of the sun when rising and setting, as we daily see in nature”. Carpenter, four years later, needs no change at all. So the defence that he only ever meant the glare and never the instrumental reading is available in 1865 and not in 1881 — and it is the 1865 sentence, the one the rewrite removed, that measurement bears out: the horizontal disc neither grows nor shrinks through a day. Chapter IX supplies the reason the enlargement reading fails on its own terms: Rowbotham’s illustration of perspective there is the Mont Cenis tunnel, whose far opening “seemed like a bright star” and whose “volume increased” as the observer approached. A luminous source diminishing with distance, printed one chapter before the Sun is required not to.

7. One thing here is genuinely open, and it is not this.

Whether the Sun’s physical diameter changes over the activity cycle is an unresolved question in the current literature, and this page does not claim otherwise: Rozelot, Kosovichev & Kilcik (2018) write that despite dedicated space instruments “no consensus was reached on this issue”, and helioseismic estimates of the kind reported by Antia and colleagues put any change at “about 5 km from minimum to maximum activity” — roughly ten milliarcseconds on a disc of about 1,920″. That debate is four orders of magnitude below the annual orbital signal and five below the change a Sun a few thousand miles up would have to show between noon and sunset. It settles nothing here in either direction, which is exactly why it is stated rather than borrowed.

The list says

Equal apparent size of Sun and Moon. / Constant solar diameter. / Constant solar diameter.

The source says

“A minute of arc on the sextant represents a nautical mile, and if the semi-diameter be 16 miles, the diameter is of course 32 miles.” (Winship, 1899, p. 120) · “Sail past the light of a small lamp in a row-boat on a dark night, and it will seem to be no smaller when a mile off than it was when close to it. Proctor says, in speaking of the Sun: ‘his apparent size does not change,’—far off or near. And then he forgets the fact!” (Carpenter, 1885, proof 89)

The qualifier was dropped.

The two constancy items (97, 220) compress a law about seeing into a datum about the Sun. It is tempting to say the words are Proctor’s and the list has borrowed an opponent’s concession, and that would be too flattering to us. Carpenter states the luminous-object law first, in his own voice, with his own example — the lamp in a row-boat that “will seem to be no smaller when a mile off” — and then endorses Proctor’s line as true: “And then he forgets the fact!” The constancy is already a positive claim in the movement’s voice inside the source, and proof 89 is already a numbered proof, so nothing is upgraded from concession to proof between Carpenter and the list. What the fragment removes is everything that scoped the claim: that the law is about how luminous bodies appear at a distance, that it is asserted in order to deny that Proctor’s objection follows, and that the paragraph ends by calling that objection a counterfeit rather than by reporting a measurement. “Constant solar diameter” reads as an astronomical fact about the Sun, which is a different and flatter thing than proof 89 asserts. No enum value fits that exactly; it is filed under the dropped qualifier, which is the closest of the seven.

Item 31 drifts the other way, and the enum has no value for it. Winship’s claim, and Dubay’s after him, is not that the two discs look alike — it is that a sextant shows them to be “of equal size and equal distance”, 32 nautical miles across apiece. The item keeps only the part nobody disputes and drops the conversion that turns it into an argument. That is a compression which makes the claim weaker and simultaneously makes it look unanswerable, because what is left is a true statement about the sky. The published gap is therefore not that the list overstated its source but that it removed the source’s only checkable step: a reader meeting “equal apparent size of Sun and Moon” is never shown the minute-of-arc-equals-a-nautical-mile rule and so never gets to notice that it hides an Earth radius. The refutation above answers the sources at full strength — the sextant derivation and Carpenter’s luminous-object law — and not the fragments.

A third gap, in the fragment itself. “Constant solar diameter” names no interval, and the two readings have opposite truth values: constant through a day, not constant through a year. The annual variation is nowhere at issue in proof 89 as quoted above, so the annual reading is one the list makes available and neither of the texts quoted here supplies.

“Parallels of latitude” only--of all imaginary lines on the surface of the Earth--are circles, which increase, progressively, from the northern centre to the southern circumference. The mariner's course in the direction of any one of these concentric circles is his longitude, the degrees of which INCREASE to such an extent beyond the equator (going southwards) that hundreds of vessels have been wrecked because of the false idea created by the untruthfulness of the charts and the globular theory together, causing the sailor to be continually getting out of his reckoning. With a map of the Earth in its true form all difficulty is done away with, and ships may be conducted anywhere with perfect safety. This, then, is a very important practical proof that the Earth is not a globe.

— One Hundred Proofs that the Earth Is Not a Globe (1885), Proof 14 of One Hundred Proofs that the Earth Is Not a Globe, Baltimore 1885, as printed in the Project Gutenberg text #55387 (5th edition). Proofs 8–16 are the pamphlet's navigation run; 14 is the one that states a measurable claim. Not checked against a print copy.

Read what this proof is willing to be wrong about. The risk it takes is not the observation that navigators use pole-centred charts, which is true and settles nothing. It is the assertion that parallels of latitude increase southward past the equator, that the increase is large enough to wreck ships, and that a map drawn the other way would end the wrecks. That is a quantity, a consequence and a remedy — a testable claim, offered as such, in 1885. The refutation below answers that claim and not the four-word items.

Where it comes from. Twenty years earlier Rowbotham had set out the same geometry at length in Earth Not a Globe (1865). Sailing due west, he writes, is steering at right angles to a north–south line fixed “more accurately by the meridian lines which converge to the northern centre of the Earth”, so that a mariner “practically circumnavigates a plane surface … because the earth is a plane, having a central region, towards which the compass and the meridian lines which guide him, converge.” Read him carefully: he is not confined to the modest point that a plane could be circumnavigated, though he makes that point too and builds a syllogism out of it; in the same sentence he states the plane as the cause. Carpenter then cut the material into numbered one-line proofs — 8 through 16 — which is the form the specimen list is written in, and proof 11 draws the conclusion that matters for item 128: “there is no south ‘point’ or ‘pole’ but that, while the centre is North, a vast circumference must be South in its whole extent.”

And it is still in circulation, quotation marks and all. Dubay’s 200 Proofs (2015) carries the run at numbers 34–39 and 108–111: “The North central Pole is the only proven fixed point on our flat Earth”; “every line of latitude south of the equator should measure a gradually larger and larger circumference the farther South travelled”. Proof 38 reproduces the Reverend Thomas Milner shipwreck passage that Rowbotham had printed in 1865 with a footnote naming Milner’s Tour through Creation — the same sentences, a century and a half later, about navigators to India who “fancied themselves east of the Cape when still west”. Very little in this cluster has moved since Palmerston was prime minister.

What this passage is being cited as. The earliest text located that states this cluster’s argument in numbered-proof form, with Rowbotham 1865 as the earlier prose statement of the same geometry and Dubay 2015 as the modern carrier. That is an ancestry claim and not a claim about who originated it: a search for an earlier numbered navigation proof was not run beyond these, and Rowbotham’s own 1849 pamphlet is not digitised anywhere this review could reach.

SURFACE (weak — do not use). “Navigators know the Earth is round, so the argument is silly.” This loses to a page of the navigator’s own manual. Bowditch, ch. 24, defines plane sailing as solving for course, distance, difference of latitude and departure “in which the Earth is regarded as a plane surface”, and defines dead reckoning as projecting course and distance run from a known position — a plane construction with a straightedge. Anyone who opens by denying that navigation treats the sea as flat is contradicted by the standard reference in one sentence.

DEEPER. The pole really is privileged, and not by convention alone. It is the one direction a magnetic needle finds unaided, the one point a chart can be built around without choosing an arbitrary origin, and the one place a star sits nearly still all night. Charts, compass roses, gyro repeaters and grid overlays are all organised around it. A defender who says this has said something true that no navigator would dispute.

KERNEL. The strongest form is historical and it is uncomfortable, because for about forty years the data really did misbehave. Ships working the Southern Ocean in the 1830s and 1840s came out of their reckoning by tens of miles a day and said so in print. Rowbotham quotes Ross’s own account of the voyage, at page 96: “We found ourselves every day from 12 to 16 miles by observation in advance of our reckoning”, and “… we found ourselves 58 miles to the eastward of our reckoning in two days”. Dubay’s proof 37 adds that Wilkes “in his journals also mentioned being consistently east of his reckoning, sometimes over 20 miles in less than 18 hours” — that wording is Dubay’s, not Wilkes’s, and Wilkes’s Narrative of the United States Exploring Expedition was not reached for this entry, so treat it as a modern report of a journal rather than as the journal. Nothing below depends on it. What Rowbotham then asks for is a measurement: a degree of longitude taken on the ground far south of the equator, by “proper geodetical operations”, because the point “has yet to be settled”. So the honest 1865 position is: the southern hemisphere is charted from computation rather than from survey, the ships that sail it keep coming out wrong, and the people insisting the charts are right have not been there with a chain. That is a legitimate demand for evidence, and it was made before the evidence existed.

Why it doesn’t save the claim.

Because the demand was met, and because the instruments the argument points at have the Earth’s size and spin written into them.

Rowbotham named his own test: measure a degree of longitude in the far south. He predicted 69.44 statute miles at the parallel of Port Jackson against 45 statute miles at 50°N, and wrote that the question “has yet to be settled”. It was settled. The measured figure at 45° is 48.99 statute miles — 8.9% above his northern value, not 54% above it. His northern number was right to within one per cent, which shows he was reading the tables correctly; his southern number is the prediction, and the prediction failed.

And the equipment convicts him twice over. The mariner’s compass he calls a proof of a central north dips into the ground at an angle that runs from 0° at the magnetic equator to 90° at the magnetic poles, which is why compass cards are balanced for magnetic zones. The gyrocompass that replaced it finds north by sensing the Earth’s rotation, so its directive force “is maximum at the equator and decreases to zero at the poles” — and its natural period is about 84 minutes, which Bowditch stops to explain is “the period of oscillation of a pendulum with an arm equal to the radius of the earth”. The instrument that finds the pole for him has the planet’s radius in its equation of motion.

Concede the description first, because it is accurate and it is in the standard reference. Charts really are built around a fixed pole. Dead reckoning really is a plane construction: Bowditch, The American Practical Navigator, art. 2400, defines it as “the determination of one’s present or future position by projecting the ship’s course and distance run from a known position”. Plane sailing really does treat the sea as flat — the same chapter says so in those words, solving for course, distance, difference of latitude and departure “in which the Earth is regarded as a plane surface”. Polar navigation really is a special case with its own apparatus. And celestial navigation really is done on a dome: art. 1524 has the navigator imagining the stars on the inside of a sphere. None of that is in dispute here, and an answer that starts by denying it deserves to lose.

What the verdict ranges over. Not whether navigation uses pole-centred, locally flat methods. It does. The claim is that those methods imply a disc with a fixed northern centre — and that claim, in the form its own authors gave it, names a measurement and gets it wrong.

1. The test Carpenter set, and it does not depend on which flat map you like

Proof 14 grants the whole geometry the items assert: the pole is the centre, meridians run outward from it, and parallels of latitude are “circles, which increase, progressively, from the northern centre to the southern circumference”. Take that at face value. On any such figure the east–west length of a parallel is its radius times the longitude interval, and the radius grows as you go south, without exception, because south is outward. So on every north-centred disc — Gleason’s, Rowbotham’s, or one not yet drawn — a degree of longitude must be longer at 45°S than at the equator, and longer still at 60°S.

It is shorter. Computed here on the GRS80 ellipsoid, 2026-08-10: a degree of longitude is 111.32 km at the equator, 78.85 km at 45°S and 71.70 km at 50°N — and the 45°S figure is identical to the one at 45°N, because an ellipsoid of revolution is symmetric about its equator. The southern value is 71% of the equatorial value where the disc requires it to be larger. That is the sphere’s answer, cos φ, and it is fatal to the class of models rather than to one member of it. Nothing in the paragraph rests on the azimuthal-equidistant chart, so the reply “that is not our map” does not reach it.

2. Rowbotham asked for the measurement in writing, and named the number

He is more specific than Carpenter and more honest about the state of the evidence. In 1865 he sets the problem out as a challenge: a degree of longitude at 50°N is 45 statute miles, and “if the Earth is a plane, and the distances above referred to as given by nautical men are correct, a degree of longitude on the parallel of Port Jackson will be 69·44 statute miles … This is the point which has yet to be settled.” He wanted geodetic operations in the south, and he was entitled to want them.

The answer at that parallel is 48.99 statute miles. His northern figure is right to within about one per cent; his southern prediction is 42% too high. Carpenter’s harder version of the same claim — proof 16, that the circuit at 45°S is “found by navigators to be twice the distance” of the circuit at 45°N — asks for a ratio of 2. The measured ratio is 1.00.

3. Where the Victorian numbers came from, and why they fit no map at all

Rowbotham’s southern circuit is assembled from sailing distances: Cape of Good Hope to Port Jackson 8,000 miles, Port Jackson to Cape Horn 8,000, Cape Horn to the Cape 6,000, total 22,000, which he compares with the parallel of 45°. The three great-circle distances between those places are 5,946, 5,065 and 3,612 nautical miles — 14,623 in total, comfortably inside the 15,282 nmi his own table gives for the 45° parallel. (His page prints 14,282 for that circuit; 360 × 42·45 is 15,282, and the slip inflates his claimed discrepancy by a thousand miles.)

The gap is not curvature. A sailing distance is what a ship covered, and square-riggers running the Southern Ocean did not follow parallels; they ran composite great-circle tracks far south of the rhumb line to hold the westerlies, and beat where they had to. But the figure does not rescue the disc either, and this is worth doing carefully because it is easy to get backwards. On a north-polar azimuthal-equidistant chart a parallel is drawn at a radius proportional to its co-latitude, so 45°S sits at 135° from the centre — a radius of 15,011 km — and measures 94,319 km, or 50,928 nautical miles. Rowbotham’s 22,000 nmi is 44% more than the globe’s 15,282 and 57% less than the disc’s 50,928. His data fit neither figure: too long for the globe and far too short for the disc, which is the signature of a category error rather than of a discovery. Note which way that cuts. To make 22,000 nmi come out right on the azimuthal-equidistant chart, the southern parallels would have to be less than half the length that chart gives them — the opposite of the stretching the argument needs.

4. The modern version of his experiment, which anybody can run this week

Sydney to Perth is 35° of longitude at about 33°S, flown nonstop about eleven times a day, published at 3,294 km and scheduled at roughly 5 hours 15 minutes. On the north-polar azimuthal-equidistant chart those two airports are 8,301 km apart in a straight line, so the flight would have to average about 1,580 km/h — supersonic at cruising altitude, in a subsonic airliner. Sydney to Santiago, Qantas QF27, is published at 11,333 km and scheduled at about 12 h 30; the same chart puts those cities 25,684 km apart, beyond the range of any airliner ever built. The identical calculation for London–New York gives a discrepancy of 7%, which is why northern routes never expose the problem and southern ones always do.

5. Dead reckoning: what it is, and what the Southern Ocean reports can and cannot settle

Dead reckoning is not a claim about the shape of the Earth. It is an estimate carried forward from the last fix by course and distance run, and it is wrong the moment anything moves the ship that the log cannot see — because a log measures speed through the water, not over the ground. The Southern Ocean is where that error is largest, and it took until the twentieth century to chart why: the Antarctic Circumpolar Current, whose transport through the Drake Passage was measured at 173.3 Sv by Donohue and colleagues in 2016, is the largest current on the planet. The easting reported in Dubay’s proof 37 — over 20 miles in less than 18 hours — is a set of about 1.1 knots, the order of magnitude of that current. That is offered as a scale and not as a diagnosis: this review did not establish the date or position of the report from Wilkes’s own record, and no mechanism at all is assigned to Milner’s Cape case.

The direction of these errors decides nothing, and it is worth saying why, because it looks as though it should. The two reports point opposite ways on the chart — Wilkes east of his reckoning, Milner’s India-bound ships west of theirs, “east of the Cape when still west” — and it is tempting to read the reversal as the signature of local currents rather than of a rescaled hemisphere. It is not. Under any uniform stretching of the southern parallels the navigator converts his distance run into difference of longitude using the globe’s figure, and therefore over-reckons his change of longitude on every heading: eastbound he plots himself too far east and the truth lies west of his reckoning; westbound he plots himself too far west and the truth lies east of it. Same error along the track, opposite signs on the compass, produced by heading alone. And the headings run the wrong way for this move. Milner’s ships were eastbound to India; Wilkes’s Antarctic coastal work of January and February 1840 ran westward, from 140°30′E to 112°16′E and on toward a goal of 105°E. Both reports are what Carpenter’s geometry predicts, and an argument built on their reversal would be an argument for the other side. (The eastward set of the Circumpolar Current would displace a ship east of its reckoning on any heading, which fits the one report and not the other; that too settles nothing on two anecdotes.)

What the stretched hemisphere does predict is not an anecdote but a constant. The over-reckoning is fixed by latitude alone — on Rowbotham’s own numbers, 69·44 statute miles to the degree of longitude where the charts give 49·52, a reckoned change of longitude 40% larger than the true one — and it applies to every ship, on every heading, on every day, at that parallel. A claim of that size is not settled by journals recording that a ship was tens of miles out after a day’s run, which is what dead reckoning does when it cannot see the water moving. It is settled by measuring the ground, which is sections 1 and 2, and the ground was measured.

The two wrecks named in that quotation are worth following, since they have travelled unedited for a hundred and sixty years. The Royal Navy Challenger whose loss is documented in the period is HMS Challenger (1826), wrecked off Mocha Island, Chile, on 19 May 1835 — and the account notes that overcast skies had prevented sightings since 17 May, so her longitude was being carried on dead reckoning with no celestial fix, which is the ordinary explanation for running onto a coast. Dubay prints a longer form of the passage than the 1865 text carries, adding “How came Her Majesty’s Ship ‘Conqueror’ to be lost?” HMS Conqueror, 100 guns, was lost on Rum Cay in the Bahamas on 29 December 1861, at 23°N.

6. The compass, taken seriously

The Victorian version of item 409 is Carpenter’s proofs 10–13: a needle points north and south at once, which he says is impossible on a globe, so meridians must be straight lines to a central north. Dubay modernises it — on a ball “the opposing ‘South’ needle would actually be pointing up and off into outer-space”.

It points into the ground, and by a measured amount. Bowditch art. 602: the angle of magnetic dip “increases from 0° at the magnetic equator to 90° at the magnetic poles”, with the horizontal component greatest at the magnetic equator and the vertical component greatest toward either pole. This is not a modern discovery invoked to save a theory; Robert Norman measured dip in London in 1581. Its practical consequence sits in every chandlery: compass cards are counterweighted for magnetic balancing zones, so a card balanced for northern service sits askew in southern latitudes. The same article records that the magnetic poles are “near, but not coincidental with, the Earth’s geographic poles”, which is why every chart carries a variation figure and why the World Magnetic Model is reissued on a five-year cycle. A needle that pointed at a fixed central north would need neither.

7. Polar navigation, and the dome

Polar navigation is genuinely a special case, and the reason is the one the argument cannot afford. A gyrocompass seeks the meridian by sensing the Earth’s rotation; Bowditch art. 620 states the consequence flatly — the directive force “is maximum at the equator and decreases to zero at the poles”, so “vessels operating in high latitudes must construct error curves”. That is a cos φ law, the horizontal component of a rotation vector tilting out of the local horizontal as you move over a sphere. On a stationary plane a gyrocompass would find nothing anywhere; on a spinning disc the rotation vector stands perpendicular to the surface at every point, so the horizontal component is zero everywhere and the instrument would never settle at any latitude. It settles everywhere except near the poles, exactly as the sphere requires.

Hence grid navigation, where meridians converge too fast to steer by — and here is the detail that decides item 408. Grid navigation is used at both ends. At the South Pole the local grid is defined so that “Local Grid North aligns with the Greenwich meridian”. Carpenter’s proof 11 says there is no south point at all, only a circumference; a circumference has no convergence, and no grid would be needed to cope with one. The southern practice exists because the southern pole does.

As for the dome: the celestial sphere is defined in the navigator’s own manual, art. 1524, as “an imaginary sphere of infinite radius with the Earth at its center”, and art. 1501 introduces it with the word imagine. It is a coordinate device, in the same way the Earth-fixed frame at ARG-R08 is a coordinate device. What is not imaginary is the geometry underneath it. The navigational triangle solved in every sight reduction is a spherical triangle, and art. 1529 notes that its terrestrial counterpart is one too: vertices at the elevated pole, the observer and the body’s geographic position, with sides that are co-latitude, polar distance and zenith distance. The intercept a navigator plots is the difference between two arcs of a great circle on the Earth’s surface, converted at sixty nautical miles to the degree — and the same chapter records that a degree of latitude runs “from about 59.7 nautical miles at the equator to about 60.3 nautical miles at the poles”, with the round 60 “correct at about latitude 45°”. That variation is the Earth’s oblateness, printed in the manual, in the unit the whole trade measures distance in. If the second reading of “star dome” is meant — the aircraft astrodome, the perspex blister through which a bubble sextant was shot on polar routes — the answer is the same one: what was done through it was spherical trigonometry with the pole as a vertex.

8. What is left

Every device in this cluster is pole-centred, and every one of them carries the globe inside it: the plane-sailing method that hands off to the spherical sailings after a few hundred miles and whose own mid-latitude formula is p = DLo cos L; the magnetic needle that dips to vertical over the poles; the gyrocompass that loses its north-seeking force as the cosine of the latitude and swings with the period of a pendulum as long as the Earth’s radius; the imaginary star dome resolved by spherical triangles in units of one minute of arc. The argument found something real — navigation is organised around the pole — and read it as a claim about shape. Read as a claim about shape it makes a prediction, in Carpenter’s own words, about how parallels behave south of the equator. They behave the other way.

The list says

Navigation charts fixed pole. / Polar navigation star dome. / Dead reckoning dome compass.

The source says

“‘Parallels of latitude’ only … are circles, which increase, progressively, from the northern centre to the southern circumference. … the degrees of which INCREASE to such an extent beyond the equator (going southwards) that hundreds of vessels have been wrecked … causing the sailor to be continually getting out of his reckoning.”

The source does not contain this. The recorded drift is the word “dome”, which two of the three items carry and the navigation argument's own texts do not. Searches run 2026-08-10: the Project Gutenberg text of Carpenter's One Hundred Proofs (#55387) returns no hit for it; the Project Gutenberg text of the 1865 Earth Not a Globe (#69892) returns one, “the dome of the Pantheon”, in the section on Foucault's pendulum; the archive.org OCR of Dubay's 200 Proofs that was searched returns none. All three argue for a plane with a northern centre and a southern circumference, and Carpenter's proof 11 denies a southern pole outright — “while the centre is North, a vast circumference must be South in its whole extent.” The enclosing dome is later furniture, from the 2015 enclosed-world material, fitted to a Victorian navigation proof that neither needs nor mentions it.

The larger change runs the other way from this project's usual finding, and the enum has no value for it. Almost every argument audited here is firmer in the list than in the book. This one is weaker. Carpenter's proof 14 commits to a quantity — parallels of latitude increase going south, enough to wreck ships — and proof 16 puts a number on it, the circuit at 45°S being “twice the distance” of the circuit at 45°N. Rowbotham in 1865 goes further and names the experiment that would settle it, predicting 69.44 statute miles to the degree of longitude at the parallel of Port Jackson and writing that the point “has yet to be settled”. What survives into the list is “Navigation charts fixed pole” — true of the globe as well, and committed to nothing. The compression dropped the falsifiable half. That is a drift toward unfalsifiability rather than toward overstatement, and it is worth naming because the usual diagnosis does not fit: on this argument the source is the party that took a risk, and the list is the party that stopped taking it. unsourced_addition is recorded because it is the plainest and most checkable of the two changes, following the precedent set at ARG-E08, and both texts are above so a reader can judge.

The refutation answers the source, not the fragment. It grants the description at full strength in its first paragraph, in the navigator's own manual's words, and then answers Carpenter's stated prediction about southern parallels and Rowbotham's stated number — not the four-word items, which assert too little to be wrong.

ZETETIC ASTRONOMY. A description of several experiments which prove that the surface of the sea is a perfect plane, and that the earth is not a globe! [title page, continued] Being the substance of a paper read before the Royal Astronomical Society on the evening of Dec. 8, 1848. [the sixth experiment] A small boat was sent out 6 miles from the Theodolite—as represented in fig. 9, but no convexity whatever could be detected! the surface of the water was perfectly level!!

— Zetetic Astronomy: A description of several experiments which prove that the surface of the sea is a perfect plane, and that the earth is not a globe (1849), title page, and the sixth experiment (fig. 9). No transcription or scan of the 16-page pamphlet could be reached; the body sentence is quoted from Schadewald, The Plane Truth, ch. 1, which reproduces it verbatim

The claim is in the title. Rowbotham's first publication — sixteen pages, Birmingham, 1849, under the pseudonym “Parallax” — does not argue from hydrostatic principle at all. It asserts a result: the surface of the sea is a perfect plane, established by six experiments with a surveyor's theodolite. The logical form is a modus tollens, and it is the same form in the mature book. Earth Not a Globe opens its water chapter: “If the earth is a globe, and is 25,000 English statute miles in circumference, the surface of all standing water must have a certain degree of convexity—every part must be an arc of a circle.” Then the observation, then the denial of the consequent. Rowbotham is not saying “water is level by nature, therefore the Earth is flat.” He is saying “a globe predicts convex standing water; I went and looked; it was not convex.”

We could not reach the 1849 text. Searched: Project Gutenberg, archive.org, sacred-texts, the Library of Congress flat-earth research guide, and Google Books. The pamphlet is catalogued but not digitised anywhere we can read. The quotation above is therefore the title, which is the claim, plus the one body sentence Robert Schadewald reproduces verbatim. WRK-ROWBOTHAM-1849 is public domain and we would quote it at length if we could. This is a coverage gap in the primary record, not a stylistic choice.

The title page contains a documented misrepresentation. “Being the substance of a paper read before the Royal Astronomical Society on the evening of Dec. 8, 1848” is not what happened. Augustus De Morgan, who was the Society's Secretary at the time and is therefore the best possible witness, recorded in A Budget of Paradoxes: “No account of such a paper appears in the Notice for that month … Dec. 8, 1848, the Secretary of the Astronomical Society (De Morgan by name) said, at the close of the proceedings,—‘Now, gentlemen, if you will promise not to tell the Council, I will read something for your amusement:’ and he then read a few of the arguments that had been transmitted by the lecturer.” The founding document of the tradition claims an institutional hearing it did not get, on its title page, in its first year.

What the circulating meme actually cites. The post that prompted this treatment — “we have SEA LEVEL! It isn't SEA CURVE!” — links Gutenberg #55387, William Carpenter's One Hundred Proofs that the Earth Is Not a Globe (Baltimore, 1885), not Rowbotham and not Dubay. Carpenter restates the claim as proof 2: “Whenever experiments have been tried on the surface of standing water, this surface has always been found to be level. If the Earth were a globe, the surface of all standing water would be convex. This is an experimental proof that Earth is not a globe.” Carpenter is explicit that he is a follower, not an author: his introduction opens “‘Parallax,’ the Founder of the Zetetic Philosophy, is dead,” calls Rowbotham “certainly, one of the most gifted of men,” and says he is “proud of having spent many a pleasant hour in the company of Samuel Birley Rowbotham.” So the meme cites the distributor as the authority, thirty-six years downstream of the claim, in a book that says on its first page whose claim it is.

And it cites the wrong proof inside the right book. Proof 2 is empirical — a claim about what canal experiments found. The meme argues from the word: we say “sea level,” not “sea curve,” therefore the sea is flat. That is a different argument, and it is not proof 2 — but it is not new either, and it is not the meme's invention. It is Carpenter's proof 18: “Every man in full command of his senses knows that a level surface is a flat or horizontal one; but astronomers tell us that the true level is the curved surface of a globe! They know that man requires a level surface on which to live, so they give him one in name which is not one in fact!” The argument from terminology has been in the tradition since 1885. It is simply filed under a different number than the one the meme links.

The second item in this cluster has a different father. Item 384, “River grades.”, is Carpenter's proof 4: “There are rivers that flow for hundreds of miles towards the level of the sea without falling more than a few feet—notably, the Nile, which, in a thousand miles, falls but a foot. A level expanse of this extent is quite incompatible with the idea of the Earth's ‘convexity.’” We looked for an earlier source and did not find one: the argument is absent from the 1849 pamphlet as quoted, the word “Nile” does not occur in the full text of the 1865 Earth Not a Globe, and the 1881 third edition's own General Index has no entry for “Nile” or “Rivers” (it has one for “Standing water, experiments demonstrating the true form of, 9–62”). Carpenter's sentence then passes into Eric Dubay's 200 Proofs (2015) almost unaltered as proof 5 — “One portion of the Nile River flows for a thousand miles with a fall of only one foot” — alongside proof 3, “The natural physics of water is to find and maintain its level,” which is proof 2 modernised. Our own cluster record attributes both items to Rowbotham 1849; on the river item that is an error, and it is recorded below.

SURFACE (weak — and it loses the argument on contact). “Water doesn't find its level.” Anyone who says this has conceded the exchange. Water in a connected vessel absolutely does come to rest on a single, reproducible, well-defined surface, it does so quickly, and it does so identically in a bathtub, a canal, a spirit level and an ocean. It is one of the most robust everyday facts there is. Do not attack it. A close cousin, equally bad: “the curvature is too small to see.” True over a bathtub and irrelevant over the Atlantic, and it silently concedes that the surface really is a plane to within measurement, which is not what we mean.

DEEPER (true but incomplete). “‘Level’ is a technical term meaning equipotential, and an equipotential on a rotating spheroid is curved.” This is correct, it is the operative definition in geodesy, and it is the answer. But stated bare it sounds exactly like what a defender says it is — a redefinition produced on demand to rescue a theory — and it invites the reply “then no observation of a water surface could ever count against you.” That reply has to be answered, not asserted past.

KERNEL — the premise is not merely defensible, it is the founding premise of the science being attacked. “A free water surface is an equilibrium surface” is precisely the assumption Isaac Newton used to derive the shape of the Earth in the Principia, by what became known as the principle of canals: imagine two water-filled channels bored from the surface to the centre, one from the equator and one from the pole, meeting there. If the water is in equilibrium the two columns must balance. The equatorial column is lightened by the rotation, so it has to be longer. Newton solved it and got a polar-to-equatorial axis ratio of 689 to 692 — a flattening of about 1 in 230 — from hydrostatics alone, with no telescope, no voyage and no survey. That is “water finds its level” carried through to a number. Rowbotham had the right premise. He simply stopped at the premise.

And the terminology really is misleading — this part of the complaint is fair. A spirit level does define a tangent plane. A surveyor setting out a slab does treat “level” as “flat.” In ordinary English a level line is a straight horizontal one, and the two senses are not distinguishable by the instrument at the scale the instrument is used: over a 30 m run the geoid departs from the tangent plane by 0.07 mm, and the angle between them is about 1 arcsecond — roughly a quarter of one division on the most sensitive machinist's bubble level made. Rowbotham was not being stupid and was not playing dumb. He was using the word the way his readers used it, in the one regime where the two meanings genuinely coincide.

Why it doesn’t save the claim.

Because the premise, solved rather than merely stated, returns a curved surface — and the number is right. Newton's hydrostatic argument predicted a flattening near 1/230; the measured value is 1/298.257223563, and the WGS 84 figure of the Earth stands 6,378,137.0 m from centre to equator against 6,356,752.3 m to the poles. The ocean's own surface is 21.4 km further from the centre at the equator than at the poles, and it is there because the planet turns. Newton was wrong in the third digit and right in the physics, which is the ordinary condition of a first-principles prediction. “Water finds its level” is not a fact the flat model owns and the round model must explain away. It is a fact the round model used, two centuries before Rowbotham, to compute its own shape.

Because the scale at which the two senses of “level” coincide is exactly the scale at which the claim carries no information. The convergence is real and it is quantitative: 1 arcsecond over 30 m, undetectable; 32 arcseconds over 1 km, about eight divisions on the same instrument, and unmissable. That is not a boundary anyone drew to be convenient. It is the reason precise levelling procedure balances backsights against foresights and tolerances the imbalance — see ARG-B05. The observation “my level says the water is flat” is true and empty; the observation “the water surface is planar over a kilometre” is a measurement, and it comes out the other way.

Because a level surface exists on both models, so its existence discriminates nothing. This is the question the argument never asks. On a flat Earth there is a level water surface. On a spheroidal Earth there is a level water surface. Every hydrostatic fact in the cluster — water settles, canals hold water, a bubble centres, a river runs downhill — is predicted by both. The discriminating content is not whether there is a level surface but what shape it has, and that has been surveyed, sounded, gravimetrically modelled and mapped from orbit. It is not a plane, it is not even an ellipsoid, and its departures from the ellipsoid reach 100 m in either direction and track the mass anomalies that produce them.

1. Grant the premise completely, because it is true. A fluid at rest, free to move under gravity alone, comes to rest with its surface everywhere perpendicular to the local gravity vector. That is hydrostatic equilibrium: were any part of the surface not perpendicular, there would be an unbalanced component along it and the water would still be moving. This is not disputed physics, it is not a recent formulation, and it is not something anyone answering this argument should want to weaken. Water finds its level. The entire content of the dispute is in the next sentence.

2. What “level” means, in the trade that owns the word. A level surface is an equipotential surface of the gravity field: a surface along which you can move a mass without doing work on it or extracting work from it. NOAA's National Geodetic Survey puts it in exactly those terms — “We define horizontal motion, when there is no change in potential energy”, as against “motions along the vertical that are associated with a gain or loss of energy” — and states the consequence for fluids directly: “Water will flow as a function of height difference and/or changes in the gravity field. This combination is defined as geopotential.” The particular equipotential that best fits global mean sea level is the geoid. It is closed, it wraps the planet, it is the reference for every orthometric height ever published, and it is curved. A plumb line hangs along its normal; a bubble level sits in its tangent plane; still water lies in it.

The word's own history says the same thing. “Level” comes into English in the mid fourteenth century as the name of a tool — Old French livel, from Latin libella, “a balance,” the diminutive of libra, the scales. It is named after the instrument that finds the direction of gravity, and the geometric senses are later growths from that root: the Online Etymology Dictionary dates the adjective “having an even surface” to the early fifteenth century, and “lying in the same horizontal plane” to the 1550s — two centuries after the tool. Both are genuine senses of the word, as the steelman above concedes, but neither is the primitive one; and “sea level” itself descends from the noun sense “position as marked by a horizontal line” (1530s), which is a height, not a shape. An argument that runs “we say sea level, not sea curve” leans on the derived sense while presenting it as the original one, and the original one is the name of a gravity instrument.

3. Where the equivocation happens, in numbers. The argument needs one word to carry two meanings: (a) level = perpendicular to gravity everywhere, and (b) level = lying in a Euclidean plane. Over short runs these are indistinguishable, and the honest thing is to say by how much. Taking R = 6,371 km, a tangent plane departs from the equipotential by d²/2R:

• over 30 m (a builder's straightedge): 0.07 mm; angle between the two surfaces, 0.97″. The most sensitive machinist's level in normal use reads about 0.02 mm/m, or 4.1″ per division. The curvature is a quarter of one division. The instrument physically cannot see it.
• over 1 km: 78 mm; angle 32″, close to eight divisions. Now it is the dominant error, which is why levelling procedure is built to cancel it.
• over 800 km (Aswan to Cairo): about 50 km.

Sense (a) is the one every instrument implements, because bubbles and plumb bobs respond to gravity and to nothing else. Sense (b) is an approximation to it that is superb below a hundred metres and worthless above a hundred kilometres. The argument takes an observation valid in the first regime and asserts a conclusion about the second.

4. So does anything here distinguish a flat Earth from a globe? Not the existence of a level surface, which both models have. The discriminating question is what shape the level surface has, and it has been measured four independent ways, none of them optical and none of them dependent on a photograph.

(i) The sea stands 21.4 km higher at the equator. The equatorial radius of the reference figure is 6,378,137.0 m and the polar radius 6,356,752.3 m — a flattening of 1/298.257223563. That bulge is not a land feature; the ocean surface participates in it, because the ocean surface is the equipotential. And its cause is rotation: a self-gravitating fluid that turns must be oblate, which is the whole content of Newton's canal calculation and of Clairaut's theorem after him. The observed value sits within a few tenths of Newton's 1687 estimate. Water found its level, and the level it found is the signature of a spinning planet.

(ii) The level surface is not even an ellipsoid, and the lumps are where the mass is. The geoid departs from the best-fitting ellipsoid by up to 100 m in either direction, and those departures correlate with density structure — which is what makes the geoid a working geophysical tool rather than a fitted curve. Satellite gravimetry maps it globally and repeatedly, and terrestrial levelling against tide gauges agrees with it. A model in which “level” means “planar” has nothing to say about any of this, and predicts none of it.

(iii) Two connected seas are not at the same height. Mean sea level at the Pacific end of the Panama Canal stands about 20 cm above mean sea level at the Atlantic end, from density and wind differences. The Permanent Service for Mean Sea Level states the general case flatly: mean sea level is not a “level surface”. This is worth dwelling on, because it cuts against the flat model in an unexpected direction: the sea is not even exactly equipotential, let alone planar. The argument's own premise, taken at the precision now available, is an approximation — and the departures from it are measured, explained and modelled.

(iv) The level surface moves, twice a day. Tides are the equipotential itself being deformed by the Moon and the Sun, and the water following it. A surface whose defining property is that it is perpendicular to the total gravity field must move when the field changes, and it does, by metres, on schedule, everywhere, in step with an astronomical calculation. On the “level means planar” reading there is no reason for any of this to happen.

5. River grades — item 384, and Carpenter's proof 4. The sub-argument is that rivers fall almost nothing over enormous distances, so the land they cross must be a plane; in its stronger form, that on a globe some rivers would have to flow uphill. Three answers, in order of increasing force.

First, the headline number is simply wrong. Carpenter wrote that the Nile “in a thousand miles, falls but a foot,” and Dubay reprints it. The Nile falls from 378 m above sea level at Khartoum to 91 m at Aswan — over 280 m in some 1,847 km, a gradient of about 15 cm/km. From Aswan to Cairo, roughly 500 miles, the modern river falls about 70 m, some 460 feet per thousand miles. Even the flattest reach in the whole system, the White Nile between Malakal and Khartoum through the Sudd, runs at about 1:101,000, under 1 cm/km — which is still 52 feet per thousand miles, fifty-two times Carpenter's figure. There is no reach of the Nile that falls a foot in a thousand miles, and the flattest large-river reach on Earth is not within an order of magnitude of it.

Second, gradient is measured against the geoid, not against a plane. A river's slope is a loss of potential, which is what makes water move; NGS again, “water will flow as a function of height difference and/or changes in the gravity field.” The heights in every river-gradient figure ever quoted — gauge datums, SRTM and satellite altimetry products, national levelling networks — are orthometric heights, referred to the geoid. So a river falling 70 m over 800 km is losing 70 m of height above the equipotential. Over that same 800 km the equipotential departs from a tangent plane by about 50 km. The two quantities are not competing; they are not even the same kind of quantity. The intuition that the river must “climb the curve” treats a 50 km geometric departure as though it were a hydraulic slope, and it is not one: the equipotential has, by construction, zero slope. Water is not being asked to run uphill, because the curve is not uphill. The curve is the definition of horizontal.

Third, this is not a second argument. It is the same equivocation applied to a moving fluid instead of a standing one. Grant “level” its technical sense and both items dissolve together; insist on the planar sense and both items still fail, because the numbers are wrong.

The companion claim, Dubay's proof 4, that rivers flow “North, South, East, West and all other intermediary directions” and that on a globe many would be flowing uphill, is answered by the same sentence. Compass direction has nothing to do with potential. The Nile flows north and downhill; the Amazon flows east and downhill; every river on the planet flows in whatever compass direction its terrain runs and in exactly one direction with respect to the geopotential, which is down.

6. Is this just a definition rigged to be unfalsifiable? The most serious objection to everything above is that “level” has been redefined so that no observation of water could ever contradict the round Earth. It has not, and the test is whether the definition generates numbers that could have come out otherwise. Three did.

(a) Newton's flattening was a prediction, and it was wrong — in the direction that could have killed it. He got 1/230 from hydrostatics; his contemporary rivals, the Cassinis, held from French meridian arcs that the Earth was prolate, lengthened at the poles, which is the opposite sign. That dispute was settled by going and measuring, on the Lapland and Peru expeditions of the 1730s, and it could have gone the other way. It went Newton's, and the modern value is 1/298.257 — a real disagreement with 1/230 that later theory (Clairaut, and the Earth's non-uniform density) had to earn its way out of.

(b) The geoid is measured before it is used, and it has a residual. If “level” simply meant “whatever the water does,” the Panama result — 20 cm of head between two connected oceans — could not be stated at all, because there would be nothing for it to be a departure from. It is stated, and quantified, precisely because the equipotential is defined independently of the sea surface. That is the mark of a real reference, not a rigged one.

(c) The 8-inches-per-mile-squared figure is the flat model's own arithmetic, and it is a prediction with a threshold. Below about a hundred metres it says curvature is unmeasurable with a bubble; above a kilometre it says curvature dominates. Both halves are checkable, and the second half is checked every working day by levelling crews who balance their sights for exactly this reason and tabulate the correction when they cannot. Rowbotham took that figure from the Encyclopædia Britannica article on Levelling — the same article that told him “the curve of the Earth being the true level, and the tangent to it the apparent level” — and reprinted the definition that answers him. See ARG-B06.

7. Scope, and where the neighbouring arguments live. This cluster is the hydrostatic claim: what a water surface is, and what “level” means. It is not the optical one. Whether ships, lighthouses and towers stay visible further than a spherical Earth allows is ARG-B04, and turns on atmospheric refraction. Rowbotham's Old Bedford Canal trials, which are where his 1849 water sentence comes from, are ARG-B03: that cluster covers the two-point near-water sightline that produces a false null, Wallace's three-point 1870 repeat and Oldham's 1901 replication. The engineering form — canals and railways built without a curvature allowance — is ARG-B05. Plumb-line convergence, which is the same equipotential fact stated as a direction rather than a surface, is ARG-B09. B01 answers the premise those four all borrow.

Verdict: refuted, and refuted by conceding the premise. Water finds its level; a level surface is an equipotential surface; an equipotential surface on a rotating, self-gravitating body is curved and specifically oblate; the oblateness was computed from this premise in 1687 and measured to nine significant figures since. The argument survives only on an equivocation between two senses of one word, in the regime where they happen to agree to within a fraction of an arcsecond — and it is stated in a vocabulary borrowed wholesale from the surveyors who use the technical sense.

The list says

Water finds level. (item 42) / River grades. (item 384)

The source says

“A description of several experiments which prove that the surface of the sea is a perfect plane, and that the earth is not a globe!” — and, in the mature statement, “If the earth is a globe … the surface of all standing water must have a certain degree of convexity—every part must be an arc of a circle.” On rivers, Rowbotham says nothing at all.

The source does not contain this.

The two items in this cluster drift differently, and one of them does not drift. The enum forces a single value, so it is set to the finding that actually changes the record; both are set out here.

Item 42 — no drift. “Water finds level.” is a fair label for Rowbotham's claim at Rowbotham's strength. If anything the list is weaker than its source: the item states a hydrostatic principle that mainstream physics also asserts, while the 1849 title asserts the far bolder and quite false proposition that the sea's surface is a perfect plane. There is no hedge to have dropped — Rowbotham does not hedge, in 1849 or in 1881 — and no scope to have widened. We looked for a drift here and did not find one, and say so rather than manufacture one.

Item 384 — unsourced addition. “River grades.” is credited by our own cluster record, and by the movement's own genealogy, to Rowbotham's 1849 Zetetic Astronomy. It is not there. The argument does not appear in the 1849 pamphlet as Schadewald quotes it; the word “Nile” does not occur in the full text of the 1865 Earth Not a Globe; and the 1881 third edition's own General Index carries no entry for “Nile”, “Rivers”, “Amazon” or “Mississippi”. The earliest text we can document is Carpenter, 1885, proof 4 — “the Nile, which, in a thousand miles, falls but a foot” — which passes near-verbatim into Dubay's proof 5 in 2015. So a claim invented by the distributor is attributed to the originator, and the modern list inherits the misattribution intact. We inherited it too: the correction is to our own record, not only to theirs.

The circulating meme drifts a third way, and it is the interesting one. “Why do people think the Earth is a globe when we have SEA LEVEL! It isn't SEA CURVE!” is an argument from terminology, and it is posted over a link to Carpenter's proof 2, which is an argument from experiment. Taken as a gloss on proof 2 that is a clean category shift — an empirical claim about canal trials restated as a claim about what a word implies. But the shift is not the meme's, and it is not modern. Carpenter's proof 18 makes exactly the terminological argument, in 1885, in the same book: “Every man in full command of his senses knows that a level surface is a flat or horizontal one; but astronomers tell us that the true level is the curved surface of a globe!” The meme is a faithful descendant of proof 18 that has lost track of its own citation and attached itself to proof 2. That is the characteristic failure mode of a tradition that circulates by restatement: the argument survives, the provenance does not, and the source it names is right about the book and wrong about the page. Both the refutation and the steelman above answer Rowbotham's and Carpenter's versions on the merits, not the fragments — which is why section 1 concedes the premise outright and section 6 answers the “you have redefined the word” reply directly.

There is no doubt, however, that it is possible for the atmosphere to have different temperature and density at two stations six miles apart; and some degree of refraction would thence result; but on several occasions the following steps were taken to ascertain whether any such differences existed. Two barometers, two thermometers, and two hygrometers, were obtained, each two being of the same make, and reading exactly alike. … One of each kind was then taken to the opposite station, and at three o'clock each instrument was carefully examined, and the readings recorded, and the observation to the flag, &c., then immediately taken. … the temperature, density, and moisture of the air did not differ at the two stations at the time the experiment with the telescope and flag-staff was made. Hence it was concluded that refraction had not played any part in the observation, and could not be allowed for, nor permitted to influence, in any way whatever, the general result. [p. 34, after quoting the Encyclopædia Britannica] It will be seen from the above that, in practice, refraction need not be allowed for. It can only exist when the line of sight passes from one medium into another of different density; or where the same medium differs at the point of observation and the point observed.

— Earth Not a Globe (1865), Zetetic Astronomy: Earth Not a Globe, 3rd ed. (London: Day, 1881), ch. II, the discussion following EXPERIMENT 9, at printed pp. 32–34 (sacred-texts transcription za14.htm; the refraction block runs pp. 31–35). The Encyclopædia Britannica extract quoted in the gloss is at p. 34 of the same chapter, and in a fuller form in the 1865 first book edition (Project Gutenberg #69892). No print copy of either edition was consulted. Two stray marks in the sacred-texts transcription — a full stop inside “one-twelfth the altitude. of the object observed” at p. 31, and an opening quotation mark inside “the line of "sight passes” at p. 34 — are dropped where those sentences are quoted here

Both halves are printed above on purpose. The concession — “There is no doubt, however, that it is possible … and some degree of refraction would thence result” — is where a hostile summary would stop, and stopping there would misrepresent him in our favour. He does not leave it as a possibility. He reports a test, and then states the conclusion flat: refraction “had not played any part … nor permitted to influence, in any way whatever, the general result”, and, a page later, that “in practice, refraction need not be allowed for.” That flat conclusion is what has to be answered, and it is what the refutation below answers.

This is a control experiment, and it deserves the name. Two barometers, two thermometers and two hygrometers, matched against each other at noon so that each pair read alike, one of each carried to the far station, all six read at three o'clock, the sighting taken at that instant, the notes compared midway along the bank afterwards. The 1851 repeat across Dublin Bay, at pp. 33–34, runs the same protocol with a theodolite on Kingstown Pier and a flag at the Hill of Howth. Whatever else is true of Rowbotham, he did not wave the objection away; he built an apparatus for it. ARG-D11 is where the method that produced it is discussed.

He is not against refraction. He allows for it, three separate times. In the same chapter, at p. 31, he takes an allowance from the surveyors and applies it to every row of his lighthouse table: “The only modification which can be made in the above calculations is the allowance for refraction, which is generally considered by surveyors to amount to one-twelfth the altitude of the object observed” — attributed in the 1881 edition to “the ordnance surveyors”, and in the 1865 first book edition, more loosely, to “surveyors generally”. In EXPERIMENT 3, in the same chapter, his own theodolite reading is described as falling on the target points “at an altitude, making allowance for refraction, equal to that of the observer” (the sacred-texts transcription carries no page marker on that page). And at p. 34 he reprints the Encyclopædia Britannica article “Levelling”, which supplies the mechanism — “on account of the unequal densities of the air at different distances from the earth, the rays of light are incurvated by refraction” — and the coefficient: refraction “may at a mean compensate for about one-seventh of the curvature of the earth, it sometimes exceeds one-fifth, and at other times does not amount to one-fifteenth.” One-seventh is k ≈ 0.14, which is the 7/6 effective-radius rule still in the surveying textbooks. The number the list says was invented to rescue the globe is printed inside the founding text of the tradition, in both editions, in the same chapter as the argument against it.

The sentence that does not survive contact with the sentence above it. “It can only exist when the line of sight passes from one medium into another of different density” is an interface rule. The encyclopaedia paragraph he has just typeset describes a continuous gradient — air that thins with height, bending the ray all along its length, with no interface anywhere. Those two statements are on facing pages of the same chapter and they cannot both be right. The refutation takes the encyclopaedia's version, because it is the one his critics were actually using and the one that is true.

SURFACE (weak — and it will lose). “They don't understand optics; refraction is real, next question.” Anyone who opens this way has not read the source and will be shown that within a sentence. Rowbotham reprints the refraction coefficient, cites the mechanism, applies an allowance to his own lighthouse table and to his own theodolite readings, and tested for the effect with matched instruments at both ends of the sightline. He is not ignorant of refraction. He has a position about it.

DEEPER (true, incomplete). “His interface rule is wrong — refraction happens in a gradient, not only at a boundary.” Correct, and it is a real self-contradiction inside his own chapter. But on its own it invites the obvious reply: fine, so there is some bending; you still have not told me how much there was over that canal on that morning, and you cannot, because nobody measured it.

KERNEL. The strongest version is not about optics at all. It is a demand about method, and it runs: you have a correction whose size you do not know, which you apply after the fact, in whatever amount the observation requires. When a distant target is visible that your model says should be hidden, it was refraction. When it is hidden, the atmosphere was standard. Name, in advance, the value for tomorrow. If you cannot, you are not correcting a measurement; you are absorbing a refutation. And the modern literature makes this worse rather than better: Hirt and colleagues, measuring the refraction coefficient at 1.8 m over grass with paired total stations at one-minute sampling, found it swinging between −4 and +16 on sunny summer days, and concluded that “the frequently used Gaussian refraction coefficient of +0.13 is not suited for describing refraction effects in the lower atmosphere.” A parameter with a working range of twenty, quoted from a geodesy journal rather than from a flat-earth pamphlet, is a serious thing to hand an opponent. The charge also lands on a great deal of actual debunking, which says “refraction” and stops, having measured nothing.

Why it doesn’t save the claim.

Because refraction is an observable, and observables can be measured before the thing they explain. That is the whole difference between a correction and an excuse, and it is not a debating point — it is a set of instruments. The coefficient is recovered from simultaneous reciprocal vertical angles at the two ends of a line, which is how Gauss got +0.13 near Hannover around 1826 and how Hirt's total stations got their time series in 2008. It is recovered independently from a temperature profile: k = 503(p/T²)(0.0343 + ∂T/∂z), which contains no free constant, only the lab-measured refractivity of air and hydrostatics. Put a thermistor mast or a radiosonde on the path and you have k without looking at the target at all. The excursions Hirt measured are not evidence that k is unknowable; they are the record of somebody knowing it, minute by minute, for thirty-three hours.

Because it is published in advance, at a fixed value, and used to predict. The US Naval Observatory defines sunrise and sunset as the moment the Sun's geometric zenith distance reaches 90.8333° — a depression of 50 arcminutes which the office obtains, in its own words, by adding “the average apparent radius of the Sun (16 arcminutes) to the average amount of atmospheric refraction at the horizon (34 arcminutes)”. That 34′ is not tuned per observation. It is stamped into every rise and set time the office issues, for every latitude, years ahead, and the tables work. A quantity fixed in advance and used to make daily quantitative predictions across the whole planet is not an ad hoc rescue, whatever else it is.

And because refusing it does not leave the argument with no adjustable parameter — it leaves it with a worse one. Having declined the atmospheric coefficient, Rowbotham needs something else to explain the theodolite results, and he supplies it: a divergence of the ray inside the telescope, which he calls “collimation, or refraction” and deploys against the horizon dip, against the 1870 Bedford Level result, and against the spherical excess of the Principal Triangulation of Great Britain. No measurement of it is located in the chapters read for this entry. The one magnitude he does give for it there, borrowed from Heather's Treatise on Mathematical Instruments, is about 240 times too small for the job — the arithmetic is in the refutation. The kernel's own standard, applied evenly, convicts the substitute long before it convicts the correction.

Concede the method complaint first, because it is a real one. A correction applied only when it is needed, in whatever amount is needed, would be exactly what this item says it is. Much popular debunking does behave that way: it names refraction, measures nothing, and moves on. If that were all there were, the charge would stick. So the question this entry has to answer is not does light bend — the source agrees that it does — but is the correction fixed independently of the observation it is used on. Three tests decide that, and refraction passes all three.

1. It was measured, tabulated and argued about for centuries before anyone disputed the Earth's shape

Ptolemy's Optics, Book V, already carries a detailed model of atmospheric refraction. Tycho Brahe “was the first to measure atmospheric refraction properly” and published his results in 1596; Kepler adopted Tycho's tables in his Optics of 1604. The extreme case was recorded before either was in print: in January 1597 Barentsz's party, wintering at Novaya Zemlya, saw the Sun when it was “5°26′ below the horizon” — what Lehn and van der Werf call “the very first report of a scientifically documented and recognized mirage”, and what is now understood as optical ducting in a temperature inversion. For the surveying coefficient specifically, Gauss obtained his average value of k ≈ +0.13 from reciprocal vertical angle measurements near Hannover, on the triangulation he ran through the 1820s. That is more than two decades before Rowbotham's 1849 pamphlet and nearly four before the first book edition. Nothing in that chronology was produced in response to this argument, because when it was produced this argument did not exist.

2. It is measured independently of the sighting it is used to explain

Two routes, neither of which looks at the disputed target. The first is simultaneous reciprocal vertical angles: two instruments at the two ends of a line, each measuring the other's zenith angle at the same instant, from which k falls out directly. That is Gauss's method and it is still the standard one — Hirt and colleagues ran two pairs of total stations along adjacent lines of sight over five days in 2008 and logged k at one-minute sampling for thirty-three hours. The second route bypasses angles entirely: k = 503(p/T²)(0.0343 + ∂T/∂z), with pressure in hectopascals, temperature in kelvin and the vertical temperature gradient in K/m. Every term is measurable with a barometer, a thermometer and a mast, and the constants come from the laboratory refractivity of air and from hydrostatics, not from any fit to a horizon photograph.

What those measurements show is that the atmosphere near the ground is far more variable than the textbook constant. Hirt et al. report k ranging from −4 to +16 at 1.8 m over grass on sunny summer days, and say so bluntly: the Gaussian +0.13 “is not suited for describing refraction effects in the lower atmosphere.” This page is not going to hide that number, because it is the best card the argument has. What has to be said alongside it is where the number came from: a paper whose entire method is measuring the thing minute by minute. Variability established by measurement is not the same as a free parameter, and it is the opposite of an unmeasured one.

3. It is fixed in advance and used to predict

The US Naval Observatory defines sunrise and sunset as the instant the Sun's geometric zenith distance reaches 90.8333°: a depression of 50 arcminutes which the office describes as “obtained by adding the average apparent radius of the Sun (16 arcminutes) to the average amount of atmospheric refraction at the horizon (34 arcminutes)” — what Andrew Young, cited below, calls “a nominal horizontal refraction of 34 minutes”. That 34′ is not adjusted per sighting; it is baked into every rise and set time the office publishes, computed years ahead, for every latitude. And the honest limit belongs in the same paragraph, stated by the same community rather than extracted from it: Andrew Young, whose SDSU pages on refraction are cited as the standard reference throughout this review, titles one of them Why We Can't Predict Sunset Times Exactly, because “the refraction at any instant may differ by several minutes of arc from the most accurate value that can be calculated.” The mean is fixed and published; the excursions are not predictable. That combination is what an honest physical correction looks like, and this entry claims nothing stronger.

4. The control experiment measured the wrong quantity

This is the paragraph that answers the source rather than the item, and it is the one that matters. Rowbotham's test compared the atmosphere at the two ends: matched barometers, thermometers and hygrometers, one set carried six miles, read at the same hour, notes compared afterwards. They agreed, and he concluded that refraction “had not played any part in the observation.”

But look again at the relation in section 2. Refraction along a near-horizontal ray is set by ∂T/∂z, the rate at which temperature changes with height. No term in it is a difference between two stations. A thermometer at eye level at one end and an identical thermometer at eye level at the other can agree perfectly while the air one foot above the water is several degrees colder than the air ten feet above it — and that vertical structure, not the horizontal comparison, is what bends the ray. His instruments were sampling a direction the effect does not live in.

And the test fails in a second, sharper way, which does not depend on any modern equipment. Horizontal uniformity is the condition under which a stratified atmosphere refracts most regularly, not the condition under which it stops refracting. A path whose two ends read alike is a path where the layers run parallel to the surface all the way along — the textbook case, the one the standard correction is derived for. The protocol could register only a horizontal anomaly, and it returned a null; he read that null as the absence of refraction, when what it indicates is the well-behaved case. Nothing about this requires hindsight about instruments he could not have owned: the geometry of the objection was available in 1881, and the encyclopaedia paragraph on the facing page — air of “unequal densities at different distances from the earth” — states the vertical variable explicitly.

How much bending is available over still water? Set k = 1 in the relation above with sea-level pressure and 288 K, and the temperature gradient required works out at about +0.13 K per metre — recomputed here on 2026-08-10. At k = 1 the ray curves with the same radius as the Earth and a curved surface photographs flat. Hirt et al. report measured gradients of 1 to 2 K/m shortly after sunset, an order of magnitude past that threshold, over grass. Nobody recorded the profile over the Old Bedford Canal on the mornings in question, and this entry does not pretend otherwise. The claim is the narrow one: for that geometry — eye eight inches up, sightline grazing still water for six miles — a flat null sits well inside the range a curved Earth produces once the profile is allowed the values that are routinely measured, so the null does not pick out a plane. Which is why the experiment that decides is the one that changes the geometry: raise the sightline out of the surface layer and put a marker at the midpoint as well as the ends, so that curvature and a uniform bend make different predictions. Wallace did that in 1870 and Oldham in 1901. ARG-B03 owns those trials.

5. What he put in refraction's place, and what it costs

Declining the atmospheric coefficient does not leave the argument free of adjustable quantities. It leaves it with one, and the thing worth stopping on is that the substitute is also called “refraction”. At p. 41, explaining why a levelled theodolite reads the sea horizon below the cross-hair, Rowbotham writes that in instruments of the best construction “there existed a certain degree of refraction, or, as it is called technically, ‘collimation,’ or a slight divergence of the rays of light from the axis of the eye, on passing through the several glasses of the theodolite.” He then uses it as a universal solvent: for the horizon dip, for the Ordnance Survey's spherical excess (pp. 263–264), and, under TANGENTIAL HORIZON at p. 272, for the 1870 Bedford Level result, where he calls “the well-known and admitted refraction inseparable from the instruments employed … fully sufficient” to explain what the experimenters saw.

The one magnitude he supplies for it, in the chapters read for this entry, is quoted from J. F. Heather's Treatise on Mathematical Instruments: after adjustment the instrument is in order for any distance up to ten chains, “the maximum error being only 1/1000 of a foot.” Take him at his word and scale it, which is what he does when he says the effect is “considerable in distances of several miles.” One thousandth of a foot in 660 feet is an angle of 1.5 × 10⁻⁶ radians, about 0.31 arcseconds. Carried over six statute miles — 31,680 feet — it comes to 0.048 feet, or about half an inch. For collimation to stand in for curvature at these ranges it has to be able to produce what curvature produces there, and the yardstick is his own arithmetic throughout: 11 feet 8 inches over six miles for the boat of EXPERIMENT 1 (the fig. 3 case, restated at p. 35), 6 feet at three miles for EXPERIMENT 3, and — at the place where he actually deploys the substitute against a reported canal result, the 1870 Bedford Level trial at p. 272 — the several feet he calls it “fully sufficient” to account for, writing that a hair's-breadth of dip, magnified, would “make it appear to be lifted up for several feet.” His substitute is short by a factor of roughly 240 against the first, roughly 250 against the second, and of order a hundred against the several feet of the third. (Recomputed here 2026-08-10 from the figures he prints.) Note also what Heather's sentence actually says — it is a statement of the residual after correct adjustment, a tolerance, which Rowbotham converts into a systematic divergence the lenses “necessarily produce”. The three structural reasons collimation cannot be the cause of the dip — it reverses under face-left/face-right, it cannot scale as √h, and spherical excess grows with the area of a triangle while an instrument bias does not — are set out at ARG-B02. This entry adds only the size, and the size is enough.

The same asymmetry shows in the allowance he does make. “One-twelfth the altitude of the object observed” is the right kind of number attached to the wrong variable: refraction scales with the square of the distance, not with the height of the target. On his own Cape Bonavista row the curvature term is 8 inches × (35 − 4)² = 641 feet, of which the 491 feet he prints is that figure less the light's own 150 feet. His rule deducts 150/12 = 13 feet. The Encyclopædia Britannica's own mean, one-seventh of the curvature, would take about 92 feet, and its strong-refraction figure, one-fifth, about 128. He is applying a seventh of the correction his own quoted authority specifies, and attaching it to a different quantity from the one that authority attaches it to. Where the one-twelfth rule came from has not been established here; it is not traced to a named surveying text in this entry, and no substitute origin is offered for it.

6. What would refute this, named in advance

The demand behind the item is fair and it deserves an answer in the form it was asked. Refraction lifts distant objects; it does not do so without leaving marks, and the marks are the falsifier.

• Strong lift comes with distortion. Large positive k compresses, stretches and, past the critical gradient across part of the beam, inverts — which is why the taxonomy has separate names for looming, towering, stooping and superior mirage. A distant skyline photographed far beyond the geometric horizon with its bases visible, undistorted, no vertical stretching, under a measured neutral temperature profile, would falsify the account given here. Nothing answering that description was located in the material surveyed for this entry or for ARG-B04, where the Lake Michigan photographs are examined; the frames looked at there have the lower floors missing and the remainder visibly stretched.
• It is episodic and correlated with the weather. The same target, from the same spot, at the same distance, appears one morning and is gone the next. A plane predicts a visibility limit set by luminous intensity and haze — monotonic, and indifferent to the temperature profile — unless it imports the very refraction this item objects to, which is what item 263 does. The observed correlation with the profile is the thing that needs explaining, and it is not a thing a plane with straight sightlines can generate.
• It is bounded, and the bound has a signature. At the coefficients ARG-B04 tabulates, 10 and 20 per cent, most of a 442-metre building is still hidden from 56.5 miles across Lake Michigan — 404 and 364 metres of it, against 453 with no refraction. Making the whole tower visible needs k up near 1, and k near 1 is a duct, which does not arrive quietly: it brings the compression, stretching and multiple imaging listed above. The escape hatch has a fingerprint on it.

7. The modern form of the item, and the concession inside it

The nearest modern form of the same charge — not traced here as item 221's source — is Dubay's proof 71, about Joshua Nowicki's 2015 photograph of the Chicago skyline from across Lake Michigan: news channels called it a superior mirage, and “while these certainly do occur, the skyline in question was facing right-side up and clearly seen unlike a hazy illusory mirage, and on a ball-Earth 25,000 miles in circumference should be 2,400 feet below the horizon.” Read that carefully, because he is half right and the half he is right about is the interesting one. In the standard taxonomy — Andrew Young's, not one invented for this reply — an erect, sharp, undistorted lift is looming, and looming is explicitly classified as a refraction anomaly without inverted or multiple images “and therefore without mirages.” The news channels reached for the wrong word and Dubay caught them at it. But the right word names an effect that raises distant objects into view while keeping them the right way up, which is precisely the thing his objection assumes cannot happen. The true observation points the other way.

8. Scope, stated plainly

This entry settles one question: whether the refraction correction is fixed independently of the observations it is applied to. It is. It does not claim that refraction accounts for the long-range sightings in the source's lighthouse table — correcting his arithmetic leaves Cape Bonavista unexplained by geometry alone, and the reason is that a light list's stated range is a luminous figure rather than a horizon figure. That is ARG-B04's argument, not this one's. The canal trials are ARG-B03, the dip and collimation are ARG-B02, the engineering “no allowance” form is ARG-B05. And the list's own position on atmospheric optics is not one position: it carries “Mirage optical ducts.” as proof item 263, where ducting is the mechanism being asserted, alongside item 221, where invoking it is the offence. ARG-B13 is where that pair is scored.

The list says

Refraction invoked for curvature.

The source says

“There is no doubt, however, that it is possible for the atmosphere to have different temperature and density at two stations six miles apart; and some degree of refraction would thence result; but on several occasions the following steps were taken to ascertain whether any such differences existed.” — Rowbotham, 3rd ed. 1881, ch. II, p. 32

“… several news channels quickly claimed his picture to be a ‘superior mirage,’ an atmospheric anomaly caused by temperature inversion. While these certainly do occur, the skyline in question was facing right-side up and clearly seen unlike a hazy illusory mirage, and on a ball-Earth 25,000 miles in circumference should be 2,400 feet below the horizon.” — Dubay, 200 Proofs, proof 71 (printed as a parallel, not as this item's traced source — see below)

The qualifier was dropped. The source hedges and the item does not. Rowbotham grants that the atmosphere may differ over six miles and that “some degree of refraction would thence result”, then reports a control experiment — six matched instruments, simultaneous readings at the two ends — and only afterwards states the conclusion flat. Four words on the list carry neither the concession nor the experiment.

The Dubay text above is a parallel, not a traced source. Proof 71 is the nearest modern form of the same charge, and it hedges the same way — “while these certainly do occur” — before finishing the sentence flat. But it argues from the appearance of one photograph, it does not use the word “refraction” anywhere in the text reproduced at flatearth.ws (searched 2026-08-11), and no line of descent from it to item 221 is established in this entry. It is printed here so a reader can see the same concession being made in the modern literature; the drift type recorded above is computed on the Rowbotham comparison alone.

Three things travel with the claim in the source and none of them survives. The coefficient: at p. 34 Rowbotham reprints the Encyclopædia Britannica article “Levelling”, which gives him the mechanism and the number — refraction “may at a mean compensate for about one-seventh of the curvature of the earth”, which is the modern 7/6 rule. The allowance: he applies a refraction correction to every row of his own lighthouse table and to his own theodolite reading in Experiment 3. The test: he did not assert that the correction was unavailable, he tried to measure whether it applied. A reader who meets only item 221 will think the tradition denies that light bends in air. Its founding text prints the coefficient for it, twice.

The item is also ambiguous, and the source supports both readings. “Refraction invoked for curvature” can mean refraction is the excuse used to save curvature — the cluster's reading, and the one the neighbouring items support — or refraction is what produces the appearance of curvature, which is the argument at pp. 41 and 266–273, where the dip and the spherical excess are blamed on a divergence inside the telescope that he also calls refraction. The compression is loose enough that both are available, and this note records that rather than picking one.

scope_widened was considered and rejected: it would describe the gap between Dubay's single photograph and the item's standing indictment of a method, and that is exactly the pairing this entry does not claim to have traced. hedge_dropped is recorded on the Rowbotham comparison, which is documented, and the reader has that text above.

The refutation answers the source, not the fragment: it grants that the correction is variable at the strength a geodesy journal states it, grants that the control experiment was honestly run, and puts the weight on what the control could sample — a comparison between two stations at one height contains no term for the vertical temperature gradient that bends the ray — and on the magnitude of the mechanism he substitutes.

Proposed instead: SELF-CONTRADICTED

MISLEADING is defined on this page as real data with a wrong conclusion made to look supported. Item 221 is not a data claim; it is a charge about the opponent's method, so the definition fits it awkwardly at best. SELF-CONTRADICTED is defined as the claim's own source, or another item on the same list, pointing the other way, and both limbs are satisfied on located text. The source: Zetetic Astronomy reprints the Encyclopaedia Britannica article "Levelling" at p. 34 of the 1881 third edition and at greater length in the 1865 first book edition, which supplies both the mechanism (rays "incurvated by refraction" by "the unequal densities of the air at different distances from the earth") and the coefficient ("about one-seventh of the curvature of the earth"); it applies a refraction allowance to every row of the lighthouse table at p. 31; and in EXPERIMENT 3 at p. 17 the author's own theodolite reading is described as taken "at an altitude, making allowance for refraction". The list: item 263, "Mirage optical ducts.", asserts atmospheric ducting as a mechanism in its own right, forty-two items further down the same list from the point where 221 charges that invoking it is illegitimate. Sibling precedent runs the same way and on the same book: B06, "Surveyors assume a plane and make no allowance", and B13, "Mirage and optical ducting explain far sightings", are both recorded SELF-CONTRADICTED. Against the change: the item does carry a genuine methodological complaint about corrections applied without measured magnitudes, and MISLEADING reads more naturally as a verdict on a complaint than on a contradiction. That is why this is filed as a challenge rather than treated as settled. The refutation above is written to read correctly under either label.

The verdict on the scorecard is unchanged until the question is settled. Publishing the disagreement is the point: a rubric that never produces dissent is not being applied.

Bellarmine cites Ecclesiastes 1:5 — “The sun also riseth, and the sun goeth down, and hasteth to his place where he arose” — as evidence that the sun really moves and the Earth stands still, and argues that unlike a passenger who sees the shore recede, here “no wise man has any need to correct the error.” That is this cluster's argument, in 1615, four centuries before the list. The modern compilers are distributing it, not inventing it — which is what this review exists to show.

Carried forward by:

• Robert Schadewald, 1987 — “The Flat-Earth Bible,” Bulletin of the Tychonian Society 44. surveyed and catalogued the corpus — from the debunking side, writing for geocentrist readers
• Gerardus Bouw, 1992/1999 — Geocentricity; A Geocentricity Primer. runs Joshua 10:12–13, Isaiah 38:8 and Psalm 19:4–6 as evidence the Sun's motion is real rather than apparent
• Rob Skiba, 2015 — biblical-cosmology teaching. the proximate route these sixteen items travelled into the modern flat-earth compilations; quotes Schadewald by name

Listed as distributors, not authors. Naming them as originators would be the same error this page documents elsewhere.

39. We have abundance of evidence that the Sun moves daily round and over the Earth in circles concentric with the northern region over which hangs the North Star; but, since the theory of the Earth being a globe is necessarily connected with the theory of its motion round the Sun in a yearly orbit, it falls to the ground when we bring forward the evidence of which we speak, and, in so doing, forms a proof that the Earth is not a globe. 50. We read in the inspired book, or collection of books, called The Bible, nothing at all about the Earth being a globe or a planet, from beginning to end, but hundreds of allusions there are in its pages which could not be made if the Earth were a globe … we have a store from which to take all the proofs we need, but we will just put down one proof—the Scriptural proof—that Earth is not a globe.

— One Hundred Proofs that the Earth Is Not a Globe (1885), Proofs 39 and 50

The numbered-proof format these sixteen items inherit begins with Carpenter. But the genealogy runs differently than the modern list assumes, and the primary text settles it.

Carpenter's book contains one hundred numbered proofs. Exactly one is scriptural: Proof 50. Carpenter says so himself, inside the proof — “we will just put down one proof—the Scriptural proof.” He treats his scriptural material as a single entry in a list of a hundred, not as a foundation. And the passages he gestures at there are the immovability and stretched-out-earth texts, not sun-motion texts at all.

The words Joshua, Habakkuk and Ecclesiastes are not located anywhere in the text we searched: the Project Gutenberg plain text of the 1885 pamphlet (ebook 55387), all 143,000 characters of it, queried in full. Carpenter's actual sun-motion proofs — 38 and 39 — argue from what an observer sees: the midnight sun skimming the horizon, the sun's daily circuit. Whatever one makes of that reasoning, it is an appeal to appearance, not to a proof-text.

So the sixteen items here are a later accretion. The numbered-proof form is Carpenter's; the scriptural sun-motion content was assembled afterwards, and reaches its present shape in twentieth- and twenty-first-century compilations sorted under headings like “Sun Moves, not the Earth.” Attributing that corpus to Carpenter overstates his role and misdescribes what he thought he was doing.

The strongest version of this argument is not about astronomy at all. It is about hermeneutic consistency, and it deserves to be stated at full strength.

Suppose you hold a text to be authoritative and to speak plainly about the world. You did not adopt that position casually; it is load-bearing for a great deal else you believe, and it has a long and serious intellectual pedigree. Now you notice that the same text describes the sun as running a circuit, as being halted, as reversing on a sundial. You have two options. You can read those passages plainly, the way you read the passages you build your life on. Or you can read them non-literally — but then you owe an account of why these and not others, and if the honest answer is “because modern astronomy says so,” you have made astronomy the arbiter of your text.

Someone who declines to do that is not being naive. They are refusing a move they regard as special pleading. Consistency is a real intellectual virtue, and applying one's hermeneutic without exception when it is inconvenient is a harder discipline than applying it only when it is comfortable. That is a defensible position, and treating it as ignorance is both wrong and rude.

Why it doesn’t save the claim.

The kernel is defensible right up to a specific point, and it is worth naming that point precisely.

A commitment about how to read a text is a commitment about reading. It stays entirely inside its own domain, and nothing this review does can touch it. But the argument here does not stop there. It continues: therefore the sun physically moves over a stationary Earth. That “therefore” carries the claim across a boundary. The moment a reading is asserted to entail that a gyroscope will precess a certain way, that a pendulum will rotate at a certain rate, that a ring interferometer will show a certain fringe shift — the claim has become a claim about instrument readings.

And instrument readings are answerable to instruments. Not to argument, not to authority, and not to the interpretive question, which remains exactly as open as it was. The dilemma the steelman poses is real, but its second horn was never “let astronomy arbitrate your text.” It is that a reading which generates a physical prediction has stepped into a domain with its own arbiter, and the prediction is now separable from the reading that produced it. The prediction can fail without the reading being adjudicated at all — which is precisely why this treatment can reach a verdict without touching the theology.

This section is about domain, not about the text. Nothing here argues that a scriptural passage is false, and nothing here takes a position on whether God exists. Those are not questions a review of measurements can answer, in either direction, and this review does not attempt them. What follows concerns only what kind of claim is being made and what can settle it.

1. Describing appearance is not asserting mechanism. This is an observation about language, not a claim about what any text means. “The sun rises” is how observers in every language and every era have described what they see from where they stand. It is also how working astronomy describes it today: the US Naval Observatory and HM Nautical Almanac Office both publish sunrise and sunset tables, every marine almanac and aviation chart uses the same vocabulary, and celestial navigation is computed in a frame where the observer is fixed and the bodies move. Nobody drawing a sight reduction from an almanac believes they have thereby asserted geocentrism. Noting this does not tell us what any given text intends by it — that is the interpretive question, addressed next.

2. The interpretive question is contested, and it is not ours. Many religious traditions, commentators and scholars have read these passages non-literally for a very long time. Others read them literally. Both are theological positions, held by serious people, argued on textual and traditional grounds. This review takes no side, and readers should not infer one from anything above. The dispute is internal to interpretation, and — this is the operative point — no instrument resolves it. There is no measurement whose outcome makes one reading correct. That is not a criticism of either reading. It is a statement about what measurement is for, and about the limits of this review's remit.

3. What is in our domain: the proof-text method conflicts with itself. The lists these sixteen items come from are not offered as devotional reading. They are offered as evidence, sorted into evidentiary categories, in a dispute about the shape of the Earth. Assessing whether a stated method delivers consistent results is fair comment on the argument being made, and it requires no theology.

It does not deliver consistent results, and the list under review demonstrates that without any help from us. Item 414 enters Isaiah 40:22, the circle of the Earth. Item 420 enters Revelation 7:1, the four corners. Items 410 and 415 enter the pillars of Psalm 75:3 and 1 Samuel 2:8. Item 412 enters Job 26:7, the Earth hung upon nothing. That is four different shapes and two incompatible supports, entered as five separate confirmations of a single physical model. A disc has no corners; a thing resting on pillars is not hanging upon nothing. A method returning all five as independent evidence is not filtering anything — it is counting. Every number above is checkable in the item table on this page, which is the point: the internal conflict is a property of the list, not of our reading of it.

Note carefully what this is and is not. It is not a claim that the texts contradict each other — every one of those passages has been read coherently within the tradition for centuries, and how they fit together is again the interpretive question we are not touching. It is a claim about the method: read this way, as a shape-extraction procedure, it yields mutually exclusive shapes.

4. The domain distinction comes from inside the tradition. The fairest available framing of this boundary was not written by a critic of religion. It was written in 1615 by Galileo, a believer, in his Letter to the Grand Duchess Christina, making a theological argument to a devout audience. Quoting Cardinal Baronius, he wrote: “That the intention of the Holy Ghost is to teach us how one goes to heaven, not how heaven goes.” He develops the point at length, observing that scripture speaks of physical things “but casually,” and concluding that “in discussions of physical problems we ought to begin not from the authority of scriptural passages but from sense-experiences and necessary demonstrations.”

Galileo's own view was that scripture and nature “proceed alike from the divine Word.” We cite him not as an authority on what the texts mean — he was making a contested theological argument and knew it — but because the domain distinction drawn here is not an outsider's imposition. It has been argued from within, by believers, for four hundred years.

5. Joshua 10, specifically. This passage has been read in several ways within the tradition, over a long period, by people who took it entirely seriously. We are not going to tell readers which reading is right; that is exactly the line this treatment does not cross. The only thing within our remit is this: astronomy does not settle it. There is no observation that arbitrates between the readings, and anyone claiming a telescope has decided the question has made the same category error the proof-texts are being asked to make. The passage is not evidence in a physics dispute — in either direction.

Verdict. UNFALSIFIABLE, in a precise and non-pejorative sense: the claim as stated cannot be confirmed or refuted by measurement, because a reading of a text is not the kind of thing measurement addresses. This is a statement about our remit and about the structure of the claim. It is not a statement about the value of the text, and it should not be read as one.

This comparison could have gone the other way, and it is worth saying why it did not. The concern was that the source might treat these verses as devotional or observer-relative language — how the sky looks from where a person stands — while the list re-presented them as physical claims about the sun’s motion. That would be a category shift, and it is the commonest failure in this family. Read against the named source’s own text, it is not what happened. Skiba states the conclusion in the flattest terms available: the earth is not rotating either, and is quite stationary in every respect. The appearance-from-the-observer reading is not conceded and then overrun; it is never entertained, and he adds that on a plain reading no one would think otherwise. The list items are bare verse labels and carry less assertion than the paragraphs they were drawn from, not more.

The treatment above is therefore aimed at the source’s own claim — a stationary Earth asserted as fact, not a reading offered as devotion. One caveat we record rather than resolve: two Psalm 19 items sit in this cluster, though the compilations file those verses under headings about the firmament and the “line” of creation rather than under sun-motion.

I don't make any claims to having the definitive answer to this controversy. Far from it. … If I were to be perfectly honest here, I'd have to say that I do not get a spherical globe, spinning around in an endless void of space from my reading of the above Scriptures.

— Testing the Globe: A Zetetic Investigation (2018), The Bible and the Still Flat Earth, the 2015–2016 teaching document issued through testingtheglobe.com. The work record WRK-SKIBA-2018 catalogues Testing the Globe, the later book carrying the same teaching; the cluster's date of 2015 belongs to the document, and the document is what is quoted.

Three provenance findings, in ascending order of seriousness. Note first a smaller one about our own record: the work we catalogue for this cluster is the 2018 book, while the cluster is dated 2015 and the teaching that actually carries these nine passages is the freely circulated PDF quoted above. Both are Skiba; they are not the same artefact, and the excerpt is from the dated one.

1. The corpus is older than the man our record names it after. The firmament proof-text set — Genesis 1, Job 37:18, Ezekiel 1:22, Isaiah 40:22, Psalm 19:1, Revelation 6:13–16, with 1 Enoch alongside the canonical texts — was assembled in print in July 1987, in an essay called “The Flat-Earth Bible”, Bulletin of the Tychonian Society 44. Its author was Robert Schadewald, later president of the National Center for Science Education, a career critic of creationism and flat-earthism who was once expelled from the International Flat Earth Research Society for his “spherical” tendencies. He wrote it for a geocentrist readership to argue that the biblical writers held an ancient cosmology, and added that “it is a grave error to reinterpret ancient documents to force their authors to speak with modern voices.” The same compilation now circulates as a flat-earth proof list. Skiba takes it from Schadewald, and says so. The teaching document quoted above introduces “an online article, titled ‘The Flat-Earth Bible,’ written by Robert J. Schadewald, © 1987, 1995”, block-quotes several paragraphs of it, and tells the reader “I recommend you actually read the entire article too”. Job 37:18, Isaiah 40:22, Psalm 19:1, Psalm 102:25, Job 9:8 and Ezekiel 1:22–26 all reach the page inside that quotation. This is not concealment and nothing here suggests it is: the credit is open, named and dated, and the reader is sent to the original. It is something rarer and more useful — a documented link in a transmission chain, in which a distributor's proof-text set is carried over, with attribution, from a critic who assembled it to argue the opposite. The compilation predates the teaching that now uses it by twenty-eight years, and its first modern compiler was a debunker.

2. Three of the nine items cannot be located in the named source. Tobit 13 dome heaven (item 424): Tobit 13 is a hymn of thanksgiving and a vision of Jerusalem rebuilt in sapphire and gold. It contains no dome, no vault, no firmament and no description of the sky's structure at all. Tobit is not cited in Skiba's 2015–2016 teaching document, in Schadewald's 1987 essay as reprinted online, or in the standard circulating compilation. The deuterocanonical text that does say what the item wants is Sirach 43, “the pride of the heavenly heights is the clear firmament”, with the sun and moon “in the firmament of heaven” — we note the likely candidate and do not assert the substitution, because we cannot demonstrate it. Psalm 104 firmament supporting lights (item 439): the word firmament is not located anywhere in Psalm 104 in the KJV, and no verse of it places any light in anything. What Skiba actually uses from that psalm is verse 2, the heavens stretched out “like a curtain”, and verse 3, the beams of God's chambers laid in the waters — and the standard compilation likewise files Psalm 104:2 under the heading “Where is the sun?”, quoting the curtain, not a firmament. 1 Enoch concentric heavens (item 425): 1 Enoch's astronomical cosmology is a portal cosmology — “I saw six portals in which the sun rises, and six portals in which the sun sets” — not a scheme of nested heavens. The layered-heavens schema belongs to 2 Enoch, whose Enoch ascends through ten of them. The item names the wrong book for the feature it describes. (1 Enoch does use “heaven of heavens” language elsewhere, which is why this is recorded as a mismatch and not a fabrication.)