A question the internet answers badly
Half a Turn to the Floor
There are two confident wrong answers online. One says toast is out to get you — Murphy's Law, cosmic spite. The other, the clever debunk, says it's all in your head: selective memory, really a coin flip. The truth is sturdier and stranger than either. Toast slides off a table edge buttered-side up, and from table height it has time for only about half a turn before it lands — so it arrives inverted. It's geometry, not luck and not the weight of the butter. Drag the table up toward 3 m and watch it finally come back butter-up — then meet the reason no kitchen is that tall.
1 · Drop it and count the turns
A slice of toast resting with its centre of mass just past the table edge tips over, grips the edge for a moment, and lets go spinning. After that it is in free fall, turning at a steady rate until the floor. Lands in the window between a quarter and three-quarter turn (90°–270°) and the buttered face is pointing down. Raise the table and you buy more spin; pull the toast right to the brink (less grip) and you buy more too.
The total turn is the real recomputed figure; the tumbling slab is a schematic of that turn. Butter is the yellow edge — it starts up. Notice it crosses 270° (butter-up again) only when the table is taller than a tall human.
2 · Why half a turn — and why the butter doesn't matter
The spin is set the instant the toast leaves the edge. Tipping over the corner under gravity, it departs with an angular velocity Matthews wrote as ω₀² = (6g/a)·(η₀/(1+3η₀²))·sin φ, where a is the slice's half-length and η₀ is how far its centre of mass overhung the edge. Then it free-falls for τ = √(2(h−2a)/g) seconds, turning all the while. From a normal table that product comes to a touch under 180° — half a turn. Butter-up to butter-down.
Two things fall out that surprise people. First, the butter is almost irrelevant: its mass is tiny and is soaked into the crumb, so it barely shifts the centre of gravity. The toast lands butter-down because butter is on the side that started up — not because that side is heavier. Second, g nearly cancels: a faster fall spins the toast faster but gives it less time, so the number of turns depends on the slice's size against the drop height, not on how strong gravity is. The same half-turn would betray your breakfast on Mars.
The one honest escape, short of a taller table, is speed: flick the toast off briskly — past about 1.6 m/s — and it sails clear before it can wind up a half-turn, landing butter-up. Slow nudges, the usual way toast leaves a plate, are the trap.
3 · The debunk tested the wrong toast
So why do careful people insist it's a myth? Because they dropped it wrong. Throw toast in the air and it lands butter-down half the time — pure chance, no edge to wind it up (the BBC's QED showed exactly this in 1991). Drop it flat from a machine and you get 50-50 again. Both are real results; neither is the situation Murphy described, which is toast sliding off an edge.
MythBusters ran straight into this in 2005 and, to their credit, recorded both answers in one episode. Their automatic dropper: an even 50-50 split — the myth, they declared, busted. But when they "simply pushed toast off the side of a table," it "showed preference to flip once and land buttered-side down." They had quietly confirmed Matthews while busting a claim he never made. The drop method is the whole story.
And the edge case has been measured at scale. In a 2001 schools experiment, about a thousand children let toast fall off tables: 6,101 of 9,821 slides landed butter-down — 62%, a margin no coin flip survives (that's 24 standard deviations from 50-50). The bias is real. It just isn't supernatural, and it isn't the butter. It's the edge and the half-turn.
4 · Why our tables are too short — ask the fine-structure constant
To land butter-up from a single fall, the toast needs to reach 270°, which means a table about 2.8 m tall. We don't build them — and Matthews's playful last move is to argue we can't. Table height tracks human height; human height is capped because a fall onto the skull mustn't be lethal; bone strength bottoms out in the binding energy of atoms; and that traces to the fundamental constants. Following a 1980 argument of William Press, the tallest a bipedal animal can be is roughly L_H ≈ 50·(α/α_G)¼·a₀ ≈ 2.8 m — set by the fine-structure constant α, the gravitational coupling α_G, and the Bohr radius a₀. The tallest human who ever lived, Robert Wadlow, reached 2.72 m: right against the ceiling. So our tables top out around half that — and the toast never gets its full turn.
A standard table sits far below the flip line — the ~2.8 m a table would need for toast to come back butter-up. That flip line lands just above the tallest possible human (and the tallest who ever lived), because the same constants set both. The kitchen is too short by construction.
It is, of course, a "rough and ready" estimate — Matthews said so himself, and Press's first version missed real human height by a hundredfold before the polymer correction. Take it in the spirit it was offered: not a proof that the universe hates breakfast, but a real chain showing that even a buttered-toast joke, followed far enough down, ends at the constants of nature. That's the part that's true.
The check — recomputed live, 25/25 green
- From a 0.75 m table, toast turns ≈178° — half a turn, landing in the (90°,270°) butter-down window. (≈90° is gripped at the edge; ≈88° is added in free fall.)
- Critical table height for a butter-up landing: h/a = 2 + π²(1+3η₀²)/(12η₀) → ≈2.84 m for real toast (η₀≈0.015, a≈5 cm) — Matthews's "about 3 m."
- At a normal table you'd need the overhang η₀ > ≈0.064 to flip butter-up; real toast sits at 0.015, far under.
- Fundamental-constants ceiling: L_H = 50·(α/α_G)¼·a₀ = 2.79 m, from α_G = Gm_p²/(ħc) = 5.91×10⁻³⁹. Wadlow (2.72 m) sits inside it.
- Schools experiment: 6,101/9,821 = 62.1% butter-down → z ≈ 24 vs the 50-50 null (overwhelmingly significant).
- What it does not claim: that thrown or dropped-flat toast is biased (it's ~50-50), that the butter's weight causes it, or that the constants argument is more than an order-of-magnitude estimate.
Reproduce: node research/butter-side-down/verify.mjs
Sources & honest apparatus
The toast dynamics, the critical-height and overhang formulas, the ~3 m flip height and the fundamental-constants bound are all Matthews's. The rotation figures, the critical height, the overhang threshold, the constants chain and the schools-test statistics are recomputed by the verifier; the drop counts and the lab/TV results are cited, not computed.
Matthews, R. A. J. "Tumbling toast, Murphy's Law and the fundamental constants." European Journal of Physics 16(4), 172–176 (1995). DOI 10.1088/0143-0807/16/4/005. — the model and the constants argument; 1996 Ig Nobel Prize in Physics.
Bacon, M. E.; Heald, G.; James, M. "A closer look at tumbling toast." American Journal of Physics 69(1), 38–43 (2001). DOI 10.1119/1.1289213. — peer-reviewed refinement; confirms the <270°, butter-down result from a ~76 cm table.
Press, W. H. "Man's size in terms of fundamental constants." American Journal of Physics 48(8), 597–598 (1980). — the L_H ∝ α_G−¼·a₀ argument Matthews adapts.
Matthews, R. A. J. "Testing Murphy's Law: urban myths as a source of school science projects." School Science Review 83(302), 23–28 (2001). — the schools drop experiment.
MythBusters, episode 28 (2005): dropped → 50-50 ("busted"); pushed off a table edge → butter-down. Verbatim at mythresults.com/episode28.
Uncertainties named, so as not to trade one tidy story for another: (1) we could not machine-read the original 1995 scan (its PDF text layer is corrupted), so the equations come from the Oikofuge transcription cross-checked against the peer-reviewed Bacon et al. The detachment angle we draw (φ≈90°, i.e. sin φ→1) is the value that makes Matthews's two published thresholds exact and reproduces the half-turn on its own — a reconstruction, flagged as such. (2) The pair 9,821 / 6,101 traces through secondary sources, not a copy of the paper we could open; the popular "21,000 drops" is the total across all three of his experiments, with no published breakdown — we use the figures we could corroborate and say so. (3) The constants bound is explicitly order-of-magnitude; the prefactor ≈50 is a rough composite of biological factors, and Matthews himself calls the estimate "pretty rough and ready." It is offered as a true chain, not a precise number.