№ 00816 NOVEMBER 2026PHENOMENONTANZANIA; PHYSICS LABS WORLDWIDE
The Mpemba Effect
Mpemba's Physics, Not the Universal Kind
Unsolved since 1969In 1963, a Tanzanian schoolboy noticed his hot ice-cream mix froze before his classmate's cooled batch — and was told he was confused. Sixty years, two independent 1969 papers, and one blunt 2016 takedown later, physicists still can't agree whether hot water ever really beats cold to the freezer.
What We Know
Erasto Mpemba was a form-three student racing to save his turn at the one refrigerator his Tanzanian school owned. It was 1963, and making ice cream meant boiling milk with sugar, cooling it, then freezing it — with a lot of boys doing it the same way, there was a rush for freezer space. Rather than wait, Mpemba poured his batch in hot. Ninety minutes later, it had frozen solid. A classmate’s milk, cooled properly first, was still liquid.
He asked his physics teacher why. “You were confused,” the teacher told him. “That cannot happen.” Mpemba believed him, until he raised the same question again at a different school, during a lesson on Newton’s law of cooling. This teacher didn’t believe him either, and delivered a line that stuck: “That is Mpemba’s physics, and not the universal physics.” His classmates adopted it as a running joke, invoking “Mpemba’s physics” whenever he got anything wrong.
Then a visiting physicist, Denis Osborne, gave a talk at Mpemba’s school and took questions. Mpemba asked his. Osborne admitted he thought the boy was probably mistaken — but he’d told his students no question deserved to be ridiculed, so back in his own lab he had a technician test it. The technician got the same result and, delighted, vowed to keep repeating the experiment until they got a different one. They never did.
In 1969, Osborne and Mpemba published their findings together, under both names, in the journal Physics Education. Their own measurements ruled out evaporation as the main cause and pointed instead toward convection, though the mechanism they offered was an explicit guess, not a proof. That same year, unaware of either man, a chemist in Canada published an entirely separate paper chasing the identical claim, down an old folk belief about pails of hot water left outside in winter.
The Roadblock
Freezing doesn’t have one honest finish line. “Frozen” can mean the water first touches 0°C, the moment the first ice crystal appears, or the point the whole sample turns solid, and each definition can rank the same two beakers differently. Ice doesn’t form the instant water reaches freezing, either — it needs a nucleation site, and nucleation is a matter of chance, so two identical samples cooled identically can still solidify at different times for no deeper reason than luck.
Stack onto that a long list of variables that are brutal to hold constant between a kitchen freezer, a university cooling plate, and a Tanzanian icebox in 1963: container shape, thermometer depth, whether the water was boiled first, its mineral or dissolved-gas content, whether built-up frost changes contact with the freezer wall, how much surface is exposed to evaporate from. Shift any one and the result can flip. When Cambridge researchers tried in 2016 to line up data from eight independent studies on common axes, the studies hadn’t just reached different conclusions; several had measured different things and called it the same effect.
Best Guesses
“Where do you stand” asks for a side, and picking one means learning that “hot water freezes faster” was never a single claim. It’s four separate physics arguments wearing one headline. Two try to explain a real effect through ordinary heat transfer. One tries to explain it through water chemistry. One argues there’s nothing left to explain, and that decades of an effect that “sometimes replicates” is exactly what noise looks like when nobody agrees on how to measure it. None of the four has closed the case.
Evaporative mass loss
Mpemba and Osborne’s own paper undercuts this one first: they measured evaporation in their setup and found it could account for at most 30% of the heat lost, not enough by their own numbers to explain the rapid freezing alone. What still argues for the theory is G.S. Kell’s separate 1969 study, sparked independently by Canadian folklore about pails of hot water freezing first outdoors: an 80°C sample lost about 6% of its mass to evaporation in sub-zero air, enough alone, by his calculation, to overtake a heavier, cooler sample — a conclusion a 2010 refinement of his model backed up.
Convective hot-top cooling
This was Mpemba and Osborne’s own guess, and their own thermocouple measurements backed it up: a hot sample’s surface stays warmer than its bulk average as convection keeps circulating warm water upward, so heat escapes mainly from that surface. A hot sample can be losing heat faster than a cooler one even after their averages match. That’s measured physics, not speculation. But a later analysis of the flow conditions argues convection at that scale should erase such differences, not sustain them. It’s the only explanation here its own discoverers watched happen, thermocouple traces and all — and the only one they’d call no more than a guess.
Dissolved-mineral depletion
Physicist Jonathan Katz proposed a sharper, testable version in 2009: boiling hard water drives off dissolved carbon dioxide and precipitates calcium and magnesium out of solution, while unheated hard water keeps solutes that concentrate ahead of an advancing ice front and drag its freezing point down. The theory predicts the effect only in mineral-rich water, and a 2026 study of 176 trials found close to that pattern — tap water produced the largest, most reliable gaps of any sample tested. What it can’t touch is Mpemba’s own tray: milk and sugar, not tap water, with no mineral content to work on.
No general effect exists
In 2016, two Cambridge researchers replotted cooling-time data from eight independent studies on common axes and found the trend ran backward: hotter water generally took longer to cool, as ordinary thermodynamics predicts. Only one dataset broke the pattern, the original 1969 paper, unmatched since, and their conclusion was blunt: no genuine effect, full stop. A 2025 analysis reinforced the reading, arguing that apparent cases in pure water are measurement scatter surfacing only near one particular freezer temperature. Unanswered since 2017, though, is a physicist’s objection that they measured the wrong finish line, and a 2026 study found the effect clustered non-randomly in exactly the chemistry his rival theory had predicted.
The Verdict
The two camps are still arguing from unequal ground, decades after the milk froze in that first beaker. One has a founding dataset nobody has reproduced at its original scale. The other has a definition of “frozen” its critics call too narrow to be the real question. Settling this needs both sides measuring the same finish line (full solidification, not just the moment a sample reaches 0°C) under a protocol that controls for random nucleation, repeated across enough labs and enough water chemistries that a tap-water signal like 2026’s either grows into something undeniable or dissolves into scatter like most of what came before it.
So far, only fragments of that experiment have been run. What’s left is a phenomenon two credentialed camps can’t agree even exists, traced back to two beakers a teenager set in an empty biology lab, one filled from the tap and one from a boiler, on an afternoon when no teacher happened to be watching.
Where do you stand?
That’s one. There’s another every morning at 6 a.m. local time.
Share where you stand
What Didn't Hold Up
Details that circulate with this case but failed verification against independent sources. Cut from the story above — listed so you know they weren't missed.
“Erasto Mpemba was the first person to ever notice that hot water can freeze faster than cold — his 1963 experiment is treated in most retellings as the discovery, full stop.”
Burridge & Linden's 2016 paper, the skeptical case's own source, quotes Aristotle's Meteorology, Francis Bacon, and a 1637 Descartes letter, describing warmed water freezing faster centuries before 1963. Confirmed: Mpemba reintroduced the observation to modern science and named it. He wasn't the first to notice it.
“Recent physics research has finally explained the Mpemba effect — 2025–2026 headlines about a "quantum Mpemba effect" read as settling why hot water freezes faster.”
Quantum-Mpemba research (spin chains, trapped ions, open quantum systems) is real and active, but it studies relaxation in quantum many-body systems, not water in a freezer. It says nothing about whether hot tap water out-freezes cold. The classical question this article covers is exactly as open as argued above.
The Rabbit Hole
- The original 1969 paper, in fullMpemba's own account, in his own words, followed by Osborne's reply
- Burridge & Linden's full 2016 takedownthe Cambridge dataset-by-dataset case that the effect isn't real
- Katz's unanswered 2017 rebuttalwhy a hard-water chemist thinks Cambridge measured the wrong finish line
- The 2026 tap-water study176 fresh trials, and the newest data either side can point to
- Skeptical Inquirer's readable history of the disputethe clearest plain-English walkthrough of the whole fight