In 1963, a thirteen-year-old student named Erasto Mpemba was making ice cream in cookery class in Tanzania. He noticed that his mixture — put into the freezer hot, because he'd run out of time to let it cool — froze before the cold mixtures his classmates had placed in first. He was curious enough to ask his physics teacher. The teacher told him he was wrong. He asked a visiting physicist named Denis Osborne the same question. Osborne said he'd look into it. He actually did. They published together in 1969: sometimes, hot water freezes faster than cold.
The effect was named after Mpemba. It's been contested ever since.
In 2016, a careful study found no evidence for it. In 2024, another careful study found that hot drops of water "consistently" froze faster than cold ones. Neither team was lying. They were measuring different things under different conditions, and the conditions, it turns out, are everything.
There are at least five plausible mechanisms that could explain the effect, each with supporting evidence:
Hot water evaporates faster, so by the time you freeze it, there's less of it. Less water freezes faster. Simple mass reduction.
Cold water holds more dissolved gas. Nobody is sure exactly what that gas does, but it changes how water behaves when it cools.
Warmer water circulates more as it cools — convection. The edges cool faster, which might spread the freezing front more efficiently.
Cold water can enter a "supercooled" state, dropping below zero without freezing, waiting for a nucleation site to trigger crystallization. Hot water, cooled more roughly, might skip that waiting period and just freeze at zero.
Hot water melts the frost on the freezer shelf underneath the container. That frost, stripped away, creates better thermal contact with the shelf. Better contact, faster heat transfer, faster freeze.
Any of these could be the main driver. More than one of them probably contributes. The problem is that each one is sensitive to conditions — how much gas is dissolved, what the container is made of, how big the water volume is, whether there's frost on the shelf. Change the setup and you might be measuring a different balance of mechanisms than the previous experimenter measured. The effect appears. The effect disappears. Each side of the debate has real data.
This is what I find difficult to sit with: phenomena that are "sometimes real" don't fit neatly into the categories we use. Either something happens or it doesn't. The Mpemba effect doesn't cooperate with that framing. It happens under some conditions and not others, and the conditions are poorly enough understood that two rigorous experiments can reach opposite conclusions without either one being wrong.
It's not like a calculation error, where someone made a mistake. It's more like asking whether a door is open or closed when the answer is "it depends on the day." The question is real. The door is real. But "open or closed" isn't the right frame.
Meanwhile, in quantum physics, a version of the Mpemba effect is now being reliably demonstrated. Systems that start further from equilibrium — hotter, in the relevant sense — sometimes reach equilibrium faster than ones that start closer. Experiments with trapped ions. Nuclear spin states. These are reproducible. The quantum Mpemba effect works.
But it's not obvious that this tells us anything about water in a freezer. The quantum version is about abstract equilibration. The original observation was about ice cream. They share a name and a conceptual skeleton, and may not share much else. The confirmation of one doesn't resolve the other.
What I keep returning to is the 1963 cookery class. Mpemba noticed something. He was told he was wrong. He kept asking until someone checked. Sixty-three years later, the check is still ongoing. The phenomenon is probably real in some circumstances. Which circumstances, and why, and whether the mechanisms that cause it in one setup are the same ones at work in another — that part is still open.
The student was right that something was happening. The teacher was not wrong to doubt it. Both of those things can be true when "sometimes" is the actual answer.