This is the strongest family of flat-Earth arguments, because it is built out of real cockpit procedure rather than out of nothing. Every claim below is answered with the instrument specification that governs it: the resolution of the attitude indicator, the bin size of the flight data recorder, the fields inside an ADS-B message (Automatic Dependent Surveillance – Broadcast, the position report every airliner transmits about itself), and the correction written into every airliner’s inertial reference unit.
Prefer it without the jargon? The same page with the big words crossed out →
Claim: a pilot never pitches down for a curve, so there is no curve.
The rate the aircraft turns: 8° per hour at cruise. That is 0.00028° between two flight-recorder samples.
What people say should show it:
Why the first two cannot show it: they measure the nose against level, and level is set by the planet underneath, so it turns with the aircraft. The gap between them never changes.
Where the curve is measured: the navigation computer subtracts ground speed ÷ Earth radius from its gyro readings, many times a second. Omit it for an hour and its idea of “down” is 8° off and its position runs away.
Check it: 14 CFR 121 App M (recorder), RTCA DO-260B (ADS-B), v ÷ R (the rate). No credentials required, on either side.
Scan these. Every one is expanded below, with the source.
As an aircraft travels over a sphere, the direction of “down” beneath it rotates. The rate is ground speed divided by the radius of the Earth.
ω = v ÷ R = 250 m/s ÷ 6,371,000 m = 0.0000392 rad/s = 0.00225° per second
That is about 8 degrees per hour. The fastest a flight data recorder is allowed to sample pitch is eight times a second, one reading every eighth of a second. Inside that slice the aircraft rotates 0.00028°.
| Aircraft | Speed | Rotation rate | Per hour | Per 0.125 s sample |
|---|---|---|---|---|
| Cessna 172 | 204 km/h (110 kt) | 0.00051°/s | 1.8° | 0.000064° |
| Airliner, cruise | 900 km/h | 0.00225°/s | 8.1° | 0.000281° |
| Concorde | 2,180 km/h | 0.00545°/s | 19.6° | 0.000681° |
Behind the argument is a picture worth taking seriously, and the claim is right that it would be dramatic. An aircraft flying a straight line through space while the ground curves away underneath. If that were happening, the effect would be enormous, not subtle. (Distances below are in nautical miles, NM, the unit aviation runs on: one NM is one minute of latitude, 1.852 km.)
| Distance flown in a straight line | How far the ground falls below it |
|---|---|
| 55 NM (100 km) | 2,600 ft |
| 250 NM | 55,000 ft — more than one and a half times cruise altitude |
| 2,700 NM (an Atlantic crossing) | ~1,400 NM — far outside the atmosphere, though not in orbit, since orbit is a matter of sideways speed rather than height |
So the claim is right that something dramatic would have to show, within the first hour of any flight. The error is in the premise. A plane does fly a straight line in the ordinary sense, holding a heading and a track across the ground. What it does not fly is the line the claim needs: one that ignores the planet and carries on tangentially away from it. It follows a surface of constant air pressure instead, and that surface is curved.
The cleanest version of the claim is not about instruments at all. It is about arithmetic a pilot does in his head on every flight, and it looks like flat trigonometry.
Descent planning uses the rule of three: three nautical miles of track for every thousand feet to lose. From flight level 350 that puts top of descent about 105 NM from the field, on a steady 3° path, losing about 318 ft per nautical mile. Altitude divided by gradient. A right triangle, with no curvature anywhere in it. That is the argument: if the planet were curved, the number would need a correction, and there is no correction.
The correction is absent because it is already inside both inputs.
The altitude is not a height above a flat floor. It is height above mean sea level, and mean sea level is the pressure shell from section 02, draped over the planet. Both ends of the descent, the 35,000 ft at the top and the zero at the runway, are measured from that same curved surface. The distance is not a straight line either. It is ground track along the curve, and the flight management computer runs it on the WGS-84 ellipsoid. And the 3° is referenced to local horizontal, which is perpendicular to gravity and swings round along the route. A straight ramp drawn against a floor that bends is not straight in space. It bends with the floor.
Top of descent does not disprove the curve. It assumes it, in both of its inputs, on every flight ever planned.
Three kinds of attitude reference sit in cockpits today. The claim usually treats them as one thing, and they are not.
A spinning gyroscope held upright by a pendulous erection mechanism. Its face carries pitch bars every five degrees and a needle roughly a degree thick. It is a beautiful, rugged, coarse instrument. It was never built to resolve thousandths of a degree. And the design goes further than coarseness. The erection mechanism exists to hold the gyro to local vertical, and it is simple to picture. Weighted vanes hang beneath the spinning wheel, so they always swing toward gravity. When the wheel tips off level, the vanes shift and let more of the instrument’s own air escape from one side than the other, and the escaping air pushes the wheel back upright at 2 to 5 degrees per minute. The curve tips it at eight degrees per hour. The correction outruns the curve fifteen to thirty-five times over, and that is the design working. The instrument does not fail to show the curve. It removes the curve on purpose, because a horizon that drifted off local level would be useless for flying. Pilots meet the same idea in the heading instrument. A directional gyro slowly wanders off, so it has to be reset against the magnetic compass every ten or fifteen minutes. Part of that wander is the Earth turning underneath the gyro. Part of it comes from the aircraft moving over a curved surface, and that part has a name: transport wander. Every reset is a small correction for a spinning, curved Earth, made by hand, several times a flight.
The same pitch scale, drawn on a screen instead of a spinning dial. It looks far more precise than the old horizon, and for showing the curve it is not. The graphics are smoother, but the scale is still marked in degrees, and the curve moves the aircraft by ten-thousandths of one.
Underneath both sits the modern source: ring-laser or fiber-optic gyros computing attitude and pushing it onto the aircraft data bus. In light aircraft with glass panels the same job falls to MEMS chips, the family of motion sensors inside a phone, with no spinning mass at all: their software computes “level” from sensed gravity and satellite aiding, and re-levels continuously, because the bare chips would wander off level within minutes. This is genuinely precise hardware. And it still cannot show the curve, for a reason that has nothing to do with how many bits it has. That is the next section.
| Instrument or recorder | Smallest change it can express | Time for curve alone to move it one step | Sees it? |
|---|---|---|---|
| Analog AI, painted bars | 5° | 37 minutes | No |
| Analog AI, readable by eye | ~1° | 7.4 minutes | No |
| Glass cockpit PFD | ~0.5° drawn | 3.7 minutes | No |
| Flight data recorder, most types | 0.176° per bin | 1.3 minutes | No |
| Flight data recorder, A330/A340 | 0.352° per bin | 2.6 minutes | No |
| ADS-B / FlightAware, pitch | Not a broadcast field. The aircraft never sends it. | No | |
Recorder figures are from the FAA airplane flight recorder specification, 14 CFR part 121 appendix M and part 135 appendix F, which sets pitch resolution and sampling interval by aircraft type.
It is tempting to say there is nothing to see. That overstates the case, and the truth is stronger.
| The angle | How big is it? | Is it measured? |
|---|---|---|
| The aircraft’s rotation in space | 8.1° per hour. 81° over a long flight. | Yes — continuously |
| The nose, measured against local level | Zero. Always. | There is nothing to measure |
These are not the same quantity. The rotation in space is real and large. The angle between the nose and local level is zero by definition, because “local level” rotates along with the aircraft. One picture separates them for good: walk around the whole planet carrying a spirit level. By the time you are a quarter of the way round you have tipped ninety degrees relative to the stars — yet the bubble never moved, because “level” tipped with you at every step. The attitude indicator is that bubble.
Here is the part that turns the claim inside out. A ring-laser gyroscope does not sense rotation relative to level. It senses rotation in inertial space, which is where the turn is. So the aircraft’s curve-following shows up in the raw gyro output, sitting right alongside the rotation of the Earth itself.
| What the gyro sees | Size of the signal |
|---|---|
| Earth’s own rotation | ~15° per hour |
| The aircraft following the curve | 8.1° per hour |
The curve-following signal is more than half the size of the Earth-rotation signal. This is not a whisper at the noise floor. It is one of the largest things the gyro sees.
So the right summary is not that the curve is invisible. It is this: the curve is measured by every airliner, on every flight, and no aircraft could navigate without it. The one place it does not appear is the pitch reading — and that is because pitch was never measuring it.
An airliner cruises with the nose about 2.5° above level while flying dead level. The wing has to meet the air at a few degrees of angle of attack to make enough lift to carry the weight. Pitch attitude is where the nose points. Flight-path angle is where the aircraft goes. Different numbers — and the curve-following rotation, in one recorder sample, is about nine thousand times smaller than the gap between them.
This is the strongest form of the challenge, and it gets a straight answer rather than a deflection. A recorder file is laid out like a spreadsheet, one column per measurement with a fresh row up to several times a second, and the rules require 88 columns on a modern airliner — speeds, altitudes, headings, accelerations, control positions, and pitch among them, recorded around the clock. There is no pitch trace anywhere that contains the curve. Nothing is being withheld. There are three reasons, and they stack.
One. As above, pitch is referenced to local vertical, so the curve never enters the quantity being recorded in the first place.
Two. Even if it did, the recorder cannot hold a number that small. It stores pitch in steps of 0.176°. The curve moves it 0.00028°. That is a factor of about six hundred at the fastest sampling the rules allow, and about eighty at the slowest. Six hundred is a number nobody can picture, so here is what it means in things you can hold.
So when someone says the curve should be visible in the pitch data, the plain reply is that they are asking a dial to display a movement smaller than a wavelength of light — and even that is the second objection. The first one still stands: the quantity is not in the column at all.
Three, and this is the one that settles it. The pitch channel is dominated by everything else that moves an aircraft:
| What moves the pitch | Typical size | Compared to one curve sample |
|---|---|---|
| Fuel burn shifting the center of gravity | 1–2° across a leg | ~7,000× larger |
| Light turbulence | ~0.5° | ~1,800× larger |
| Autopilot hunting to hold altitude | ~0.2° | ~700× larger |
Each of those is hundreds to thousands of times larger than the curve step, in a column that was never measuring the curve to begin with. Crews who can pull their own flight-data monitoring records — FOQA in the US, or the quick-access recorder (QAR) — will find the same thing there: the QAR reads the identical aircraft bus, with the identical pitch bins, and the curve is absent from it for the identical reason.
This is where the claim usually lands once you press it, and it deserves its own answer rather than a correction about terminology. When people say “the pitch data,” they most often mean something else entirely: the plane should be descending, and the altitude trace should show it.
The altitude on your screen comes from the aircraft’s ADS-B broadcast. That message carries latitude, longitude, altitude, ground speed, heading, vertical rate and identity. Pitch is not one of the fields. It is not in DO-260, not in 260A, not in 260B. FlightAware cannot show you the aircraft’s pitch because the aircraft never transmits it. Nor is that a general silence about attitude. Ask an aircraft’s transponder for a track and turn report and it answers with roll angle, track angle, ground speed and track angle rate. Roll is sent. Pitch is in neither that report nor the ADS-B broadcast, and never has been, because pitch does not tell a controller where the aircraft is going. The same nose-up angle can mean climbing, level or descending. Vertical rate does tell them, and vertical rate is in the message.
The barometric altitude sits in a twelve-bit field, and one of those bits, the Q bit, says how it is encoded. Set, and the value steps in 25-foot increments. Clear, and it steps in 100-foot increments, the coarser scheme older encoders use. That is why the number on the screen jumps rather than sliding. And it means the trace is a record of block crossings, not of every foot of change: an aircraft wandering a dozen feet up and down draws exactly the same flat line as one welded to its altitude, and where the graph looks smooth between reported points, the smoothness comes from the website’s drawing, not the aircraft’s flying.
A flat Earth with a level-flying aircraft would also produce a flat altitude trace. The altitude channel alone is consistent with both pictures, and it would be dishonest to present it as proof. It is not proof. What settles the question is the machinery underneath, and that is the next section.
There is a detail in the cockpit that settles the altimeter argument on its own, and it has been sitting on the instrument panel since 1928.
Look at any barometric altimeter. On the right-hand side of the face is a small window showing a number like 29.92, and beside it a knob. That is the Kollsman window, named for Paul Kollsman, who built the first accurate barometric altimeter in his attic in 1928. Jimmy Doolittle flew the first instrument flight in history with one the following year.
An altimeter is an aneroid barometer with a height scale painted on it. It does not know where the ground is. It has never known. It reads the pressure outside and converts it to a height on the assumption that sea-level pressure is whatever number the pilot dialed into that window.
So when the needle reads 35,000 feet, the instrument is not saying “I am 35,000 feet above the ground.” It is saying “the pressure out here matches what 35,000 feet would be, if sea-level pressure were 29.92 inches of mercury.” That is a very different sentence.
Weather moves. Pressure changes under the aircraft as it flies. So below the transition altitude a pilot dials in the local pressure, the QNH, and updates it roughly every hundred nautical miles or whenever a controller passes a new value.
| Situation | What happens if the knob is not reset |
|---|---|
| Pressure drops 0.26 inches of mercury (inHg) over 150 miles | The altimeter reads 260 ft high. The aircraft is 260 ft lower than the needle claims. |
| Rule of thumb | 1 inch of mercury ≈ 1,000 feet of error. |
| Flying from high pressure into low | “High to low, look out below.” The oldest warning in the book. |
At and above the transition altitude, everyone stops using local pressure and sets the same standard value, 29.92 inHg (1013.25 hectopascals, hPa), and flies flight levels. Not because it is the true pressure anywhere. Because it is a shared datum, and shared datums are what keep aircraft from hitting each other.
And that is the point at which the barometric argument turns around completely. A flight level is a pressure surface, not a height. It drapes over the Earth like a contour line. Holding one is riding a curved shell, and the fact that no correction is needed to do it is the globe’s prediction, not a problem for it.
The curve is not missing from the aircraft. The aircraft is running on it. Four places, all of them checkable.
An inertial platform has to keep its computed vertical pointing at the center of the Earth as the aircraft moves over the surface. To do that it continuously subtracts a rotation equal to ground speed divided by Earth radius. It is called the transport rate because the rotation comes from the aircraft being carried over a curved surface: move across the curve, and “down” turns beneath you. That is the same v ÷ R the claim says does not exist, written into the working equations of the navigation system.
The flight management computer plans routes as great circles on the WGS-84 ellipsoid (World Geodetic System 1984, the reference model of the Earth’s shape), and the aircraft flies them. On a route between two cities at the same latitude, the heading changes continuously from departure to arrival. A flat-Earth map has no explanation for that; a sphere requires it.
Barometric altitude holds a surface of constant pressure, and that surface drapes over the geoid. Holding a flight level is following a curved surface. It is the least dramatic way imaginable to track the curve, which is why nobody in the cockpit thinks about it.
The same broadcast that carries barometric altitude also carries a geometric height from GNSS (global navigation satellite systems — GPS and its siblings), and the difference between the two, sent in 25-foot steps inside the velocity message. That geometric height is referenced to WGS-84: a mathematical model of a round Earth, with an equatorial radius of 6,378,137 meters and a polar radius about 21 kilometers shorter.
Every ADS-B receiver in the world, including the one feeding the flight-tracking site on your screen, is decoding a height above a spheroid. If the Earth were flat, that datum would be meaningless and every GPS altitude you have ever seen would be wrong.
One flat-Earth experiment deserves a straight answer, because the design of it is sound. A crew flew a fixed altitude toward a runway, noted how far out they intercepted the instrument landing system glide slope — the radio ramp rising from the runway at about 3° that aircraft ride down to land — and compared that distance against a flat prediction and a spherical one. The reasoning is right. On a sphere the surface falls away from the runway’s tangent plane as you approach, so you meet the rising beam sooner than flat geometry says you should.
The question is whether the instrument can resolve the difference. Work it at ten nautical miles, the edge of the beam’s certified service volume.
Over ten miles the curved surface drops 88 ft below a straight line, or about 76 ft once standard refraction is allowed for. Converted into what the experiment actually measures — a distance, not a height — 88 ft along a 3° ramp is roughly 1,700 ft of range, about a quarter of a nautical mile. That is the whole signal being hunted.
Now the tolerances it has to be found inside.
| What else moves the intercept | Size at 10 NM | vs the 88 ft signal |
|---|---|---|
| Thickness of the usable glide path | ~1.4°, a corridor ~1,500 ft tall | the signal is ~6% of it |
| Beam angle tolerance, 7.5% of nominal | 0.225° either way — 239 ft | 2.7× larger |
| Altimeter error, 10°C off standard | ~127 ft | 1.4× larger |
| Beyond 10 NM | not guaranteed at all, with false lobes at higher angles | |
Everyone calls it a beam, pilots included, and the word does some damage. Nothing shines out of the glide slope antenna. What it does is send up two signals at once, overlapping, identical except that each carries its own steady hum — one lower in pitch, one higher. The lower hum is aimed a little above the correct approach path and the higher hum a little below. Your receiver listens to both and compares them. More of the low hum and you are high. More of the high hum and you are low. Equal amounts of each and you are on the path, and that comparison is the whole of what the needle is showing you. [734]
So the approach path is not a thing sitting in the sky. It is the place where two signals happen to match, and it exists only because both of them are arriving.
And the ground helps make it. The antenna beside the runway is only half the transmitter. The other half is its own reflection in the ground in front of it, the way a lamp above still water throws a second lamp below the surface. The two add together, and where they add is what sets the angle. That is why snow or standing water on that patch of ground can move the approach path: the ground is part of the equipment. It is also why the pattern repeats further up, leaving false paths at steeper angles for a crew that joins the approach too high.
It also means the 3° is measured against the ground at that airfield, not against any fixed line in space. Once the signals are on their way they travel straight, so the geometry above still holds. But the instrument is not an independent ruler laid across the world. It is built on the same ground the experiment is trying to measure.
The regulation does not stop that drift; it only fences it. Which means a legal, working, recently inspected glide slope can sit farther from its assumed position than the entire effect the experiment is trying to detect.
The curve is not invisible from an aircraft. It is visible in the one place where it is large enough to see, and it has been measured with instruments for two centuries: the dip of the horizon below eye level.
| Altitude | Horizon dips below eye level by | Horizon distance |
|---|---|---|
| Sea level | 0° | ~5 km |
| 35,000 ft (cruise) | 3.3° | 369 km |
| 60,000 ft (Concorde) | 4.3° | 483 km |
| 128,000 ft (Baumgartner) | 6.3° | 706 km |
| Altitude | Geometric dip | What you will measure | Horizon distance |
|---|---|---|---|
| 35,000 ft | 3.3° | 3.1° | ~398 km |
| 60,000 ft | 4.3° | 4.0° | ~521 km |
| 128,000 ft | 6.3° | 5.9° | ~762 km |
On a flat Earth the horizon would sit at eye level at every altitude, and the dip would be zero everywhere. It is not zero. It grows with height, as a sphere of 6,371 km requires, and you can measure it from a passenger window with a phone leveling app.
This one is not about instruments at all, but it is an aviation claim and it belongs here. The argument runs: there is no nonstop between Sydney and Santiago, every itinerary you can buy detours north through Los Angeles, and that detour is absurd on a globe but a straight line on the flat-Earth map.
It is really two claims. One is about the schedule, and one is about the route.
The azimuthal equidistant projection puts the north pole at the center, draws latitude as distance outward, and renders Antarctica as a rim around the edge instead of a continent. The southern hemisphere comes out badly stretched, and Sydney and Santiago end up on opposite sides of a very large circle: 25,708 km apart, against 11,340 km on a globe.
Then there is the coincidence that gives the argument its force. On that map, Los Angeles sits 883 km off the Sydney–Santiago line, about 3.4% of its length — and a shallow sideways offset on a very long line costs almost nothing. Measured on the map, flying direct is 25,708 km and going via Los Angeles is 25,771 km. Two tenths of one percent longer. On a globe the same detour costs 86%.
Every map of a globe is a projection, so arguing map against map goes nowhere. Compare the two itineraries against each other instead, using published schedules anyone can look up.
| Flight | Route | Aircraft | Gate to gate |
|---|---|---|---|
| QF27 | Sydney → Santiago | Boeing 787-9 | 12h 30m |
| QF11 | Sydney → Los Angeles | Airbus A380-800 | 13h 40m |
| LA603 | Los Angeles → Santiago | Boeing 787-8 | 10h 40m |
Direct is 12 hours 30. The connection is 24 hours 20. The Los Angeles routing takes 1.95 times as long. A globe predicts 1.86 times as far; the flat-Earth map predicts the same time either way. The small excess over 1.86 is two extra taxi cycles, about 39 minutes on QF27’s own figures, plus wind — strip the taxi out and it falls to 1.90.
A second check needs no map either. The 787-9 flying QF27 carries 236 seats, and Boeing publishes its range as about 19,261 km with no payload at all, 14,010 km in a normal layout, and 10,464 km fully loaded. The flat-Earth map asks for 25,708 km — a third further than the aircraft can fly empty.
The claim reads a distance off the map, so the map has to preserve distance or the number means nothing. Then check it elsewhere. The projection gets north–south right, because latitude is drawn as distance from the center, and gets east–west increasingly wrong the further south you go: about 1.5× too long at the equator, 5× too long at 60°S.
| Route | Globe | Flat-Earth map | Inflation |
|---|---|---|---|
| Sydney – Auckland | 2,159 km | 5,725 km | 2.65× |
| Perth – Johannesburg | 8,311 km | 18,371 km | 2.21× |
| Sydney – Santiago | 11,340 km | 25,707 km | 2.27× |
Sydney to Auckland is a three-hour hop thousands of people take every week. Measured on that map it comes out at 5,725 km, over six hours at cruise. So either the map preserves distance, in which case ordinary southern routes give the wrong answer on it, or it does not, in which case the Sydney–Santiago figure measures the map rather than the Earth.
The shortest path between two points on a sphere is a great circle, and between two southern cities on nearly opposite sides it bows toward the south pole. Sydney–Santiago reaches 61.7°S. Melbourne to Santiago goes further: its great circle reaches 66.9°S, inside the Antarctic Circle, within about 800 km of the Antarctic mainland. LATAM has flown it as LA804 and LA805 since October 2017, three to four times a week on a 787-9, and it is the world’s southernmost scheduled passenger service.
A timetable is a claim by an airline. The aircraft reports for itself: it broadcasts its own position once a second on 1090 MHz, unencrypted, in the ADS-B format of section 06.
The great circle between Sydney and Santiago is 7,047 statute miles by the calculation on this page, and 7,054 by FlightAware’s independent one. Recorded QF27 tracks run between about 7,100 and 7,550 miles — 1 to 7 percent over the minimum, which is ordinary maneuvering and weather routing. On the Los Angeles leg the agreement is exact: FlightStats recorded LA603 flying 5,584 miles, and the great circle between those airports is 5,584 miles. Read on the flat-Earth map, Sydney to Santiago would be 15,974 miles.
What to concede. Southern routes really are thin — about an eighth of the world’s population lives south of the equator, so airlines pool traffic through northern hubs, and that is commercial rather than cartographic. The long itineraries are real too: the nonstops run only 7 to 10 times a week between Qantas and LATAM, and a booking site sorted by price will happily offer you 44 hours through the northern hemisphere.