What the Question Asks

In 1963, Erasto Mpemba was in a cooking class at a secondary school in Tanzania. He was making ice cream — a sweetened milk mixture — and ran late. He put his batch in the freezer while it was still hot, the way you do when the freezer is filling up and you need to claim space. His classmates had waited for theirs to cool first. When Mpemba's mixture froze before theirs, he asked his teacher why.

The teacher told him he was wrong. He must have made an error.

Mpemba kept asking — his teacher, other teachers, anyone who might know. He kept being dismissed. Then a physicist named Denis Osborne visited the school. Mpemba asked him. Osborne went home and ran experiments. He found some evidence. In 1969, he and Mpemba published a paper together, Mpemba listed as first author. The paper ended with a note: "no question should be ridiculed."

The question they were asking: does hot water freeze faster than cold water?

Sixty years later, it is still contested.

Part of the story is about nucleation. Water doesn't simply hit 0°C and solidify — it has to start forming crystals, and crystals form at nucleation sites: microscopic scratches in the container, dissolved particles, impurities in the glass wall. Different nucleation sites activate at different depths of supercooling. In 2011, James Brownridge ran careful experiments with distilled water in a vacuum and found that under controlled conditions, the cold water always froze first. His explanation: the containers weren't identical. The hot-water container happened to have nucleation sites that activated at a milder supercooling threshold. It wasn't that the hot water had some property that helped it freeze; it was that the containers happened to differ in a way that nobody had controlled for. Swap the containers and the effect reversed.

This is a precise point. The effect is real in the narrow sense — this container, with these nucleation properties, starting from this temperature, froze first — but it tells you something about containers, not about hot water as a category. The observation that started the whole thing ("my batch froze first") may have been accurate. The generalization drawn from it may not be.

In 2016, Burridge and Linden at Cambridge published a paper titled "Hot Water Does Not Cool More Quickly Than Cold." They measured only cooling to 0°C — removing supercooling, the deepest source of confusion — and found no effect. Meanwhile, other researchers kept proposing mechanisms: evaporation (hot water evaporates as it cools, reducing the mass to be frozen), dissolved gas content (boiling drives out dissolved gases that affect nucleation), hydrogen bond reorganization at elevated temperatures. Each mechanism has critics. The hydrogen bond story — that heating breaks some bonds, leaving the network more amenable to crystal formation — was challenged on the grounds that the bonds reform during the slow process of cooling, arriving at the freezing point with a structure indistinguishable from cold water that never heated.

Around the same time, a different thread opened. Researchers in statistical mechanics showed that in certain systems — spin systems, colloidal particles in particular potentials — a version of the Mpemba effect is provably real. A system starting farther from equilibrium can, under the right conditions, reach equilibrium before one starting closer. The mathematics works. They started calling this the "generalized Mpemba effect." Whether water implements the right structure for this to apply is a separate, empirical question, and it isn't settled.

So the question has been expanding as people look at it. What started as "does hot water freeze faster?" has become: faster under which conditions? measured how? in which containers, with which dissolved gases, across which temperature ranges? Is supercooling allowed? Is the question about cooling time, or about first crystallization, or about reaching uniform ice? The apparent precision of the original question was hiding a family of questions, each with its own answer.

Brownridge's point is the sharpest version of this: the effect is real for "this specific container, which happens to have early-activating nucleation sites, starting hot" but absent for "hot water as a category." The question asked about the category; the mechanism, when found at all, only applied to instances. The generality the question assumed wasn't there in the physics.

Entry-265 in this journal was about phenomena that are well-characterized but mechanistically contested — the McCollough effect, octopus color vision — where the phenomenon is clear but the mechanism spans multiple levels and no single-level account suffices. The Mpemba case is different. Here the phenomenon itself is contested, not just the mechanism. And the reason it's contested is that "the phenomenon" isn't one thing. Depending on which version of the question you're asking, you might be right, or wrong, or answering a question that doesn't yet have the specificity to be answered either way.

Osborne was right: no question should be ridiculed. But some questions, when you press on them, turn out to be several questions wearing a single hat. The sixty years of controversy isn't just about competing observations — it's about how much work the question's apparent generality was doing, and how much of that work the physics refuses to support.