Fun With Science  /  Globe Deconstruction  /  Self-Test Protocol

Self-Test Protocol

Thirteen falsifiable tests from Levi Miller's own claims and his own stated methodology (falsification-first, "no assumptions," Team A vs Team B): seven rescale the book's proposals, six are the checks the claim pages carry. Each is runnable with cheap equipment, and each states its globe prediction and its flat-plane prediction before the result.

The book frames its project as falsification-driven — your own controlled tests, not appeals to authority. This page takes that at face value: a setup, then both predictions, as numbers wherever the linked page has them. The book's own headline test needs rescaling first: at the heights it specifies, curvature cannot hide the light until about 4,271 feet; the baseline it proposes is 410 feet. 410 ft cannot discriminate.

Conceded once: this review has run none of the thirteen as written. Where a linked page reports a measurement the entry says so; the Io date is open until mid-November 2026; the rest are proposals. Tests 8–13 copy their numbers from the pages they link.

"The decisive question isn't whether wall-distance-dependence disappears; it's whether the vehicle moves at all in the condition built to minimize recirculation."

TEST 1

The bottom-up disappearance test — properly scaled

Targets: Geocentric flat-Earth claim #1 (p.26) — objects disappear bottom-up due to angular resolution, not curvature. Answered on Bottom Up Observations.

His proposed setup (p.127): flashlight 2.5" above calm water, camera 0.5" above water, 410 feet apart. The horizon-distance formula (d = √(2Rh), R = 3,959 mi) gives a 0.5" camera a horizon of about 1,320 ft and a 2.5" light about 2,951 ft; the light stays geometrically visible to roughly their sum, 4,271 feet (~0.81 miles) — over 10× his baseline. The curvature bulge at 410 ft is about 0.05 inches, smaller than either height. Neither model predicts disappearance at 410 ft, so a result there cannot discriminate.

The same setup has a 150-year history

Structurally this is Samuel Rowbotham's 1838 Old Bedford River test: telescope 8 inches above the water, a boat flag 3 feet above it, 6 miles off; the flag stayed visible where curvature predicted an ~11-foot drop. In 1870 Alfred Russel Wallace raised the sightline to 13 feet — above the layer where warm air over water bends grazing light unpredictably — added a midpoint marker, and the curvature appeared cleanly. The inches-above-water setup is the 1838 geometry, not the 1870 correction.

Rescaled protocol

Four distances (500ft, 1,500ft, 3,000ft, 5,000ft), heights fixed, light and camera a few feet above the water rather than inches, a fixed marker at the midpoint. Then raise the camera to four times its height and repeat: the cutoff should move out by a factor of two. Refraction shifts a cutoff by a modest fraction; it neither removes it nor changes the square-root scaling, and that scaling is the discriminator — "perspective" predicts no cutoff to shift.

Globe predicts
Sharp disappearance near the calculated horizon-distance sum for whatever heights are used, shifting as the square root of height (√2 per doubling).
Flat plane predicts
Visibility at all four distances, limited only by brightness/optics — no hard geometric cutoff at any of them.
Cost: tripod + flashlight + a truck, ~$0–50
TEST 2

Full-moon "no highlight" claim, at his desk

Targets: Q7 (p.50) — "If the Moon is a spherical rock in a vacuum, how does the light of the Sun perfectly distribute across the entire surface during a full Moon without any highlight points?" Answered on Full-Moon Lighting.

Setup: a rough, matte sphere (a golf ball, a stucco-coated craft ball, a scored orange) and, as a control, a glossy one (a billiard ball), lit by a single point source across the room. Photograph from an angle offset from the lamp, then with the camera beside the lamp — near-zero phase angle (the angle at the ball between light and camera): full-Moon geometry.

Standard photometrics predict
The matte ball at near-zero phase shows a flat, evenly lit disc with no highlight point — the lunar "opposition effect," every grain's shadow hidden behind it. Only the glossy ball shows a highlight, in both frames: a highlight reports surface finish, not shape.
Q7 as stated predicts
A sphere lit by a single source shows a highlight point; if that is what lit spheres do, the matte ball must show one at zero phase as well.
Cost: a $2 ball and a desk lamp he already owns
TEST 3

Moon-tilt illusion, with a lamp and a ball on a string

Targets: Claim #3 sub-claim — moon-tilt illusion "explanation is not satisfactory" (p.165, and the worked example at pp.167–168). Answered on The Moon-Tilt Illusion.

Setup: a lamp as the "Sun," a ball hung as the "Moon" at several angles and heights against a fixed horizon reference (a table edge). Photograph the terminator (light/shadow boundary) at each position.

Standard geometry predicts
The terminator angle visibly changes with viewing geometry, reproducing the "illusion" in his Santa Fe/Los Alamos photo without needing to trust that photo.
His claim predicts
No consistent geometric/perceptual explanation should reproduce the effect on demand.
Cost: string + a ball he already has
TEST 4

"Local light source" cloud-brightness claim — one setting change

Targets: pp.170–172 — two cloud photos showing very different brightness "with the Sun still in full view." See The Sun Does Not Shrink.

Setup: redo the above-the-clouds sequence with the camera locked to manual exposure/ISO/shutter (every phone has a "Pro" mode) instead of auto-exposure.

Camera-artifact hypothesis predicts
The brightness swing disappears once exposure is held constant — the original result was auto-exposure metering, not a change in the light source.
His claim predicts
The brightness difference persists with exposure locked, indicating a genuine local/variable light source.
Cost: free — a phone setting
TEST 5

Jupiter's moon-shadow "misalignment," as a dated blind prediction

Targets: Q6 (p.49) — Jupiter's moon shadows "not aligned with the singular light source of the Sun" (pp.75, 160–161, 186–187). Answered on Jupiter's Shadows Point at One Sun, which published this prediction before the fact.

Setup: over 15–17 November 2026, at Jupiter's western quadrature (Jupiter ninety degrees from the Sun as seen from Earth), time one Io shadow transit with a modest telescope and a stopwatch: start when the shadow touches the disc, stop when Io does. The only outside input is a published transit prediction for Io itself (Sky & Telescope prints these in advance).

Standard model predicts
Io's shadow touches the disc 74.9 to 76.0 minutes before Io (the one-line law gives 75.7). In April 2026 the same shadow arrived 76.7 to 78.3 minutes after Io. Europa's offset about twice Io's, Ganymede's about four times.
His claim predicts
The shadow should not match an advance prediction from heliocentric geometry: it does not lead Io in November, or the interval is nowhere near 75 minutes.

An advance prediction with a sign: the flip from "after" in April to "before" in November is a direction no fudge factor supplies afterwards. To the standing reply — heliocentric software made the prediction, so a match is circular — the answer is that the book's model must state its own advance interval; with one prediction on the table, only that one is tested.

Cost: a modest telescope, or borrowed time on one
TEST 6

Kampf's Law, properly instrumented

Targets: "Kampf's Law" (pp.79–84) — gas-propelled thrust requires an external medium to push against. Answered on Rockets Don't Push Against Air, Action Lab Footage and Thrust, Measured in Flight.

Setup: a real vacuum gauge (~$30–50), reporting the pressure reached in torr or mbar (a torr is 1/760 of an atmosphere), not "vacuum on/off." Run the far-wall trial with the wall backed in absorptive material (foam, cloth) to reduce exhaust bounce-back, at the best achievable vacuum.

Why "does wall-distance stop mattering" is the wrong test

Even at a good vacuum, a nearby wall lets the vehicle's own exhaust bounce back against it — a distance-dependent push unrelated to ambient pressure. Propulsion labs call these facility effects and size chambers to keep them out of thrust measurements (AIP Physics of Plasmas, "A review of the impact of ground test-related facility effects on gridded ion thruster operation and performance," 2024). So "closer wall → bigger reaction" (the book's p.80 finding) doesn't discriminate — both hypotheses predict it. The decisive question isn't whether wall-distance-dependence disappears; it's whether the vehicle moves at all in the condition built to minimize recirculation.

Newton's third law predicts
Nonzero motion in the farthest-wall / best-vacuum / absorptive-backing trial — any motion there falsifies "requires an external substrate to work at all," even if closer-wall trials show a bigger effect (the recirculation confound), which should decay toward the far-wall baseline with distance.
Kampf's Law predicts
Zero motion in that same far-wall/absorptive/best-vacuum trial — no reaction at all without a nearby surface to push against.
Cost: ~$30–50 vacuum gauge + foam/cloth backing
TEST 6B

Kampf's Law — the sharper claim: timing, not magnitude

Targets: his refined mechanism — the push originates when the gas contacts the wall and is transmitted back along the gas column, not at the moment of ejection

This is what he pointed to in the Action Lab footage: an apparent delay between gas leaving the nozzle and the vehicle visibly moving, matching the cloud reaching the wall. On rewatch, the syringe's substantial motion is timed with wall contact, but a smaller motion precedes it. And a still-vertical filament just after gas exit is kinematics: an impulse changes velocity at once, but a filament's angle shows displacement — for a small puff, a fraction of a millimeter in the first tens of milliseconds, below what normal-speed video shows.

Instrumented, not just visual

Keep the horizontal-tube/syringe geometry; add force instrumentation to the same rig. An in-line load cell (strain gauge or piezo, ~$20–100, the part hobby-rocketry thrust stands use) between syringe and support, replacing the filament, gives a continuous force-vs-time trace at the source; a pressure-sensitive plate as the target wall gives a second trace of when the gas arrives. Run both together across a few wall distances.

Newton's third law predicts
Force onset at t≈0 (gas exit) in the source trace, independent of wall distance; if the wall is close enough for recirculation (Test 6), a second bump in the same trace just after the plate registers impact — two peaks, not one delayed peak.
His mechanism predicts
Near-zero force in the source trace until a signal time-correlated with (not preceding) the plate's own contact registration — one peak, gated by wall contact.

Bonus check: total impulse (area under the source curve) should match the momentum of the ejected gas. Fallback without load cells: 240fps phone video tracking the filament's angle frame by frame, timestamping gas-exit onset, first deflection and wall contact across several distances, including one where gas transit takes a full second or more.

Primary cost: two load cells, ~$40–200 total. Fallback cost: a phone with slow-motion video (already owned) + free motion-tracking software
TEST 6C

Kampf's Law — turn the exhaust instead of lengthening the chamber

Targets: the same p.80 mechanism, by direction rather than by timing. The Action Lab frames on Action Lab Footage show the ordering the book expects and cannot, on their own, run the book's two-distance test; this version needs no second chamber and no instrument.

Why not a longer tube, or a bigger syringe: a tube sends the exhaust straight down the axis and returns whatever comes back along the same axis, so the push from the nozzle and any push from the wall arrive from the same direction and the two hypotheses only differ in when — a difference that sits inside a single video frame at any consumer rate (Test 6B). Changing the syringe does not help either: a larger volume changes how the propellant burns (the confined-burn law on the Action Lab page), so a change in onset with syringe size is the engine's own pressure build-up, which both hypotheses share. The discriminator that survives is where the recoil points.

Setup: a wide chamber rather than a tube — a bell jar or a pressure-cooker-sized vessel is enough — with the syringe hung on two threads so it is free to swing in any horizontal direction, and a phone filming from above. Run A: nozzle aimed square at the nearest wall, at some distance d. Run B: same syringe, same d, but the exhaust is turned through 90° by a smooth plate fixed to the chamber (not to the syringe) a few centimetres past the nozzle, so the gas ends its flight against a side wall and anything that returns comes from the side. Run C, optional: nozzle turned through some angle θ with no plate. Mark the nozzle direction on the floor before each run.

Newton's third law predicts
The syringe swings straight back along its own nozzle line in every run, by the same amount in A and B (to within the few percent of recirculation Test 6 allows for), and by the angle θ in C. Where the gas goes afterwards, and which wall it meets, is irrelevant to the recoil.
Kampf's Law predicts
The push comes from gas compressed against a surface and returned to the vehicle, so the recoil should follow the geometry of the return: in B the wall is beside the syringe, not behind the jet, and the recoil should be sideways, or a small fraction of A's (only what back-scatters from the plate); in C the swing should follow the wall the gas ends up against rather than the nozzle.

The point is that turning the nozzle by an angle turns the recoil by the same angle under Newton and by something else under the wall mechanism. A direction can be read off a phone video from above; a delay of a few milliseconds cannot. The standing reply — that diffuse scatter off the plate still returns some gas along the axis — is a magnitude claim, and A against B measures it: the Action Lab frames put returning material at 0.05–3% of the outgoing momentum.

Cost: the Test 6 rig, a second thread, a flat plate and a phone
TEST 7

Water-curvature "skinny warehouse" test, scaled to what's accessible

Targets: p.89–90 — a proposed 1000-meter controlled water-curvature test, not yet run. Answered on Bending the Laws of Hydrostatics.

Setup: a laser level plus a calm canal or long pool. At 100–200 meters the predicted drop is under a millimeter — likely below what basic gear can resolve, which is itself informative: it is why casual short-range observation never "sees" the curve. A mile-plus baseline with a surveying-grade level brings the predicted ~8 inches per mile² (a level surface's drop below a straight line, growing with the square of distance) within easy reach.

Globe predicts
A measurable drop once the baseline is long enough relative to instrument precision, matching the standard curvature formula.
Flat plane predicts
No drop at any baseline, regardless of instrument precision or distance.
Cost: laser level rental, ~$30–100/day
TEST 8

The tilted glass, and the straight kilometre

Targets: Q2 (p.45) — whether gravity can "bend the laws of hydrostatics," and the chapter's premise that water takes the shape of its container. From Bending the Laws of Hydrostatics.

Setup: half-fill a glass and tilt it. Then the trough the book proposes at p.89, with one specification: a kilometre of still water over a floor built straight (what light in a vacuum gives) rather than level (what a spirit level gives). Built level, nothing happens in either world; the floor, not the water, is the instrument.

Globe predicts
The surface stays put while the glass rotates under it — a free surface is perpendicular to the downward pull, not its container. Over a straight kilometre, level and straight part by 78.5 mm, four times that over two.
Flat plane predicts
The glass behaves identically, but over the kilometre straight and level coincide: the departure is zero at any length.
Cost: a glass of water; the kilometre is the book's own proposal
TEST 9

Two filtered photographs of the Sun, an hour apart

Targets: Q8 (p.51) — the Sun does not set but "shrinks due to linear perspective and disappears completely from a lack of visibility." From The Sun Does Not Shrink.

Setup: mid-afternoon, the Sun near 45° up, a solar filter on the lens; two frames sixty minutes apart at one focal length; measure the disc's width in pixels. (An arcminute is a sixtieth of a degree, an arcsecond a sixtieth of that; the Sun is about 32 arcminutes, or 1,920 arcseconds, across.) Total refraction at 45° is one arcminute, so the air cannot excuse a sixth.

Globe predicts
The same diameter in both frames, about 32′; the second merely darker or brighter.
Flat plane predicts
Recession costs the disc a sixth of its diameter within the hour at 45° — 16.7 per cent, 321″ per hour — for a lamp 5,000 km up moving at 1,670 km/h, the smallest defensible figures; the flat model's own map makes it larger.
Cost: a solar filter and a camera he already owns
TEST 10

The Moon against the stars, across one night

Targets: Q5 (p.48) — "Why is there no difference in visual speed between stars and planets as the Earth spins during a single night?" From The Sky Turns at One Rate.

Setup: at dusk, note which stars sit beside the Moon; look again before dawn. Naked eye, one night, exactly as the question asks.

Globe predicts
The Moon drifts eastward against the stars at 1,977″ per hour — 4.39° over an eight-hour night, about eight of its own diameters — while the stars keep their pattern. The same rule puts Mars at opposition (opposite the Sun in the sky) at 54″ per hour.
Flat plane predicts
As Q5 has it, no difference in visual speed: the Moon holds its place among the same stars all night.
Cost: nothing
TEST 11

Summit to reflected summit, at his lake

Targets: Q4 (p.47) — "Do we observe perfectly mirrored reflections over large bodies of water at rest as we increase observer height?" and the Lake Pukaki drone proposal at p.116. From The Mirror Settles Nothing. The Same Photograph Does.

Setup: do not measure the reflection's shape; measure the angle from the summit down to its reflection. It is twice the summit's apparent elevation, the camera's height cancels out (to better than 0.2′), and the water supplies the level. The inputs — the peak's height and distance, 64.6 km for Aoraki/Mount Cook from the Pukaki shore — go to both models alike, so an error in them moves both predictions together.

Globe predicts
311–317 arcminutes across every atmosphere the lake could plausibly have had. A frame taken there on 5 August 2020 measures 311 ± 3′.
Flat plane predicts
346–352 arcminutes for the same atmospheres — thirty-five arcminutes above the globe, about ten per cent of the angle measured.
Cost: a camera, a map and a still morning
TEST 12

Two phones, two hemispheres, one Jupiter

Targets: the "Unknown Luminaries" chapter (p.161) and the seven-planet photograph at p.163. From The Eight-Planet Photograph.

Setup: two observers, one in each hemisphere, agree a timestamp and each photograph Jupiter with a compass app open; neither needs software to find it. The arrangement of the planets is the same from everywhere on Earth, so every difference between the frames is orientation — where a flat plane and a globe part company.

Globe predicts
For 22 February 2025 at 18:45 UT: from London, Jupiter at altitude +60°, azimuth 187° (compass bearing: due south); from Cape Town, +30°, azimuth 336° (north-northwest) — 149° apart in azimuth, with the Jupiter–Mars–Uranus separations identical to within 21 arcseconds.
Flat plane predicts
Lights above a flat plane converge on the observers beneath them: Jupiter in the same compass direction from both sites, never high in opposite compass halves at one moment.
Cost: two phones and a friend abroad
TEST 13

Which side the bite is on, after totality

Targets: the fourth "red flag" of Claim #3 — the eclipse "came in from the wrong side," an argument reached through the book's QR that belongs to Jeremy McGarry and is tested on his own frames. From The Eclipse Came In From the Wrong Side.

Setup: in any partial-phase frames — his from Erie, Pennsylvania on 8 April 2024, or your own at the next eclipse — note which side of the Sun the bite sits on before totality, and after. The direction is a position angle: measured at the Sun from straight up, positive toward increasing azimuth.

Globe predicts
The bite rotates through about 160°: 147° to 154° (lower right, in a level frame) before totality, −11° to −17° (top) after. His three post-totality Nikon frames, stamped 15:23, 15:25 and 15:46, show it at the top.
The video's expectation predicts
The Moon approaches along the one daily arc the two bodies share, so the bite stays on the same side through the event — still at the lower right after totality.
Cost: nothing — the frames already exist