Why it isn't a coin toss
The slice is already spinning when it leaves the table
01 / slipThe slice topples over the edge, it doesn't just fall. Gravity creates a toppling torque around the table edge, and the slice gains angular velocity before it leaves the table.
02 / fixed rateThe angular velocity, ω, is set by the slice's geometry. Matthews' formula ties ω to the slice's half-length, its overhang at slip, and its tilt angle, not to table height.
03 / fall timeOnly height decides how long the slice keeps spinning. Fall time grows with the square root of height, so the total rotation, φ ≈ ω × t, grows with the square root of height too.
≈8.04 rad/sthe angular velocity ω from Matthews' formula for the illustrative parameters chosen for this model, about one and a quarter turns a second
The coincidence that isn't one
An ordinary table gives exactly half a turn
At 0.75 metres, fall time is about 0.39 seconds. Multiplied by ω, that gives a rotation of about 180°, half a turn: the slice leaves butter-up and reaches the floor butter-down. For the landing to come back butter-up, it would need close to a full turn, ~360°, which requires four times the height, because fall time grows only with the square root of height, not with height itself.
No kitchen or living-room table comes close to 3 metres. That's why the result has nothing to do with a particular bread, butter, or plate: it comes from the fact that ordinary tables are, structurally, about a quarter of the height a nearly full rotation would need.
Confirmation at large scale
In 62.1% of drops, the slices landed butter-side down
Matthews ran a practical trial with nearly 1,000 UK schoolchildren, who dropped buttered slices from ordinary tables. The result confirms, on a large sample, exactly what the geometry predicts: butter-down wins more often than a coin toss would.
| Quantity | Value |
|---|---|
| Drops recorded | 9,821 drops |
| Butter-down landings | 6,101 landings |
| Butter-down percentage | 62.1% |
| A fair coin's percentage | 50.0% |
Method and limits
A simple model explains an old observation
R.A.J. Matthews published, in 1995, a torque analysis of a slice toppling off a table edge. Written in dimensionless form — with η = δ/a, the ratio of the overhang at slip to the slice's half-length —, his formula is ω² = (6g/a) × [η/(1+3η²)] × sinθ, where a is the slice's half-length, δ the overhang at the moment of slip, θ the tilt angle, and g gravitational acceleration. Writing it with η keeps every term inside the brackets dimensionless, so the units of ω² (1/s²) come solely from the g/a factor. For the illustrative parameters chosen to reproduce the two published landmarks (a = 5 cm, δ ≈ 0.34 cm, θ ≈ 55°), the formula gives ω ≈ 8.04 rad/s; these parameters are not typical measurements of a real slice. The dial uses this value of ω and approximates the in-flight rotation linearly, φ ≈ ω × t, where t = √(2h/g) is the fall time from height h. This simple model reproduces the two published landmarks: about 180° at 0.75 metres and about 360° at 3 metres.
Matthews' work earned him the 1996 Ig Nobel Prize in Physics, a distinction awarded to research that first makes people laugh, then makes them think. A later refinement, published by Bacon, Heald and James in 2001, shows that Matthews' frictionless-pivot model underpredicts angular velocity unless slipping at the table edge is included explicitly; the authors reanalysed the problem through video analysis and numerical solving. The refinement changes the fine values of ω without changing the geometric conclusion: at ordinary table height, the rotation stays close to half a turn.
The limits stay clear: 180° and 360° are landmarks of an illustrative model, not measurements of a particular slice; ω depends on the slice's geometry at slip, not on table height; and the 62.1% figure comes from a large field trial, not a laboratory measurement. Together, the torque mechanics and the large sample point at the same thing from different angles.