Put two equal cups of water in a freezer, one hot and one cold. Surely the cold cup freezes first: it has less heat to lose. Usually, that is the sensible expectation. Yet reports of hot water apparently freezing sooner are old and persistent enough to have a name, the Mpemba effect. The puzzle is real as an experimental question, but the slogan “hot water freezes faster” is far too broad.¹
Before explaining anything, we must decide what “freezes” means. Does a cup win when its water first reaches 0°C, when the first ice crystal appears, or when the entire cup is solid? Those moments can be far apart. Water can cool below 0°C without forming ice, a state called supercooling, and then freeze suddenly when a crystal finally starts to grow. Comparing different endpoints can reverse the apparent winner.¹
Two cups that are not truly identical
Heating water changes more than its temperature. A hot cup may lose some water through evaporation, leaving a smaller volume to freeze. Its strong temperature difference can drive convection currents that move heat towards the surface. A hot container may also melt frost beneath it and make better thermal contact with a freezer shelf. Those are plausible ways for a particular setup to change the race, even though the hot water began with more heat to shed.²
Container shape, lid, placement, dissolved substances, freezer airflow and the water's previous heating history can matter too. If one cup sits closer to a cooling element, the experiment tells us more about the freezer than about water. Even using “the same freezer” is not enough if opening the door or adding hot cups changes its conditions.
The most subtle variable is often nucleation: the process that starts ice formation. Two otherwise similar samples may supercool by different amounts. A sample that stays liquid at a lower temperature can lose a race to another sample that begins crystallising earlier, despite having cooled faster up to that point.¹
Does it happen under careful control?
Researchers do not agree that there is one universal Mpemba effect with one mechanism. Some experiments report a hot sample reaching a chosen freezing milestone first. Others, under tighter or different controls, do not reproduce that outcome. A recent controlled investigation stresses how heavily results depend on definitions and experimental design.¹ The observation in one freezer does not become a law of water.
This also explains why apparently contradictory accounts can each contain a piece of truth. An experiment measuring first ice may find one result; a study timing complete solidification may find another. A cup with no lid can evaporate, whereas a sealed one cannot. The claim must include conditions before it can be tested or explained.
A better home experiment
If you want to investigate, use identical containers, equal measured masses of the same water, and temperature probes. Record both the cooling curves and the moment ice first appears. Repeat the trial several times and swap the positions of the containers. Decide beforehand whether you care about first ice or complete freezing. A single dramatic race is a starting observation, not a conclusion.³
Heat loss is not the only clock
The warmer sample must, in broad terms, release more thermal energy to reach the same final state if mass and other conditions truly match. That simple energy accounting remains useful. But the rate at which heat leaves is not constant. At first, the hot sample has a larger temperature difference from its surroundings, so it can shed energy more rapidly. The race involves both how much heat must be lost and how quickly it is lost at each stage.²
That does not by itself prove that the hot sample will overtake the cold one. In an idealised comparison with identical cooling behaviour, the initially warmer sample cannot simply pass through the colder sample's temperature history and finish first. A reversal needs some difference in conditions or in the way the freezing milestone occurs. This is why evaporation and nucleation are so important: they change the race rather than overturning basic thermodynamics.
Why replication is difficult
A freezer is not a perfectly uniform laboratory bath. Its cooling cycle switches on and off, air moves around shelves, and ice crystals or scratches on a container can seed freezing. Small differences may dominate a result when the two cups finish close together. An experimenter who reports only the winning cup may miss the decisive cooling curves.
Repeating an experiment matters because spontaneous ice nucleation can vary between runs. A result that appears once but reverses when the containers swap places is weak evidence for a general effect. The most informative reports describe their apparatus, volumes, starting temperatures and endpoint clearly enough that others can try the same comparison.¹ ³
The most interesting lesson is not that hot defeats cold. It is that freezing has stages, and that an apparently simple everyday question hides several variables. Under ordinary matched conditions, cold water has the head start. Under particular conditions, hot water can appear to win a carefully defined race. The science lies in finding exactly which race was run.