Despite sounding like the most egregious contradiction in physics, hot water appears to freeze faster than cold water under certain circumstances. The phenomenon can be traced back to Aristotle himself, but after centuries of experiments demonstrating this phenomenon, no one's been able to explain it.
Now physicists are pointing to strange properties of hydrogen bonds as the solution to one of the oldest mysteries in physics - but others are claiming the so-called Mpemba effect doesn't even exist at all.
For a bit of background into the Mpemba effect, this phenomenon has been confounding physicists since Aristotle first noticed it more than 2,000 years ago.
After similar accounts from the likes of Francis Bacon and René Descartes, the possibility of hot water freezing faster than cold water finally gained widespread acceptance in the 1960s, thanks to a Tanzanian schoolboy who noticed the effect when making ice cream.
Erasto Mpemba and his schoolmates often made ice cream by boiling milk and mixing it with sugar, and letting it cool before placing it in the freezer.
One day, Mpemba grew impatient, and instead of letting his mix cool before placing it in the freezer, he put his still-boiling milk in anyway, and hoped for the best.
To everyone's surprise, his ice cream set quicker than his peers', and in 1969, Mpemba teamed up with a physics professor to publish a paper describing the apparent phenomenon.
But there's a big problem with the Mpemba effect. While it's been more or less accepted as fact, physicists can't agree on how exactly it works, because how can hot water hit freezing point faster than cold water, when cold water already has a massive head-start?
There's also the lingering problem of replication.
Attempts to replicate the Mpemba effect in a foolproof, consistent manner have failed, but there's enough inconsistent evidence out there to prevent it from being debunked altogether.
Back in 2012, the Royal Society of Chemistry ran a competition asking scientists to explain the phenomenon, and despite receiving some 22,000 papers from all over the world, none of the explanations were convincing enough on their own to draw widespread consensus.
As Signe Dean reported for us last year:
"The most commonly proposed hypothesis ... is that hot water evaporates more quickly, losing mass and therefore needing to lose less heat in order to freeze. However, scientists have also demonstrated the Mpemba effect with closed containers where evaporation doesn't take place.
Another theoretical speculation is that water develops convection currents and temperature gradients as it cools.
A rapidly cooling glass of hot water will have greater temperature differences throughout, and lose heat more quickly from the surface, whereas a uniformly cool glass of water has less of a temperature difference, and there's less convection to accelerate the process.
But this idea has not been entirely verified either."
So after centuries of experiments, we're still looking for answers.
Now researchers from the Southern Methodist University in Dallas and Nanjing University in China think they might have a solution - strange properties of bonds formed between hydrogen and oxygen atoms in water molecules could be the key to explaining the elusive Mpemba effect.
Simulations of water molecule clusters revealed that the strength of hydrogen bonds (H-bonds) in a given water molecule depends on the arrangements of neighbouring water molecules.
"As water is heated, weaker bonds break, and groups of molecules form into fragments that can realign to form the crystalline structure of ice, serving as a starting point for the freezing process," Emily Conover reports for Science News.
"For cold water to rearrange in this way, weak hydrogen bonds first have to be broken."
In other words, we find a higher percentage of strong hydrogen bonds in warm water than cold, because the weaker ones were broken as the temperatures increased.
"The analysis ... leads us to propose a molecular explanation for the Mpemba effect. In warm water, the weaker H-bonds with predominantly electrostatic contributions are broken, and smaller water clusters with ... strong H-bonding arrangements exist that accelerate the nucleation process that leads to the hexagonal lattice of solid ice.
Therefore, water freezes faster than cold water in which the transformation from a randomly-arranged water clusters costs time and energy."
But as with all the explanations that have come before this one, we're going to need to see more proof before we can know for sure that this - or a combination of factors - is truly at play in the Mpemba effect.
While some put the replication problem down to several factors coming together in different ways to achieve the phenomenon - including convection, evaporation, and supercooling - and the fact that freezing is a gradual, not instantaneous, process, others say the Mpemba effect is nothing more than an incredibly persistent myth.
Another recent paper by a team from Imperial College London monitored the time it took for hot and cold water samples to drop to freezing point (0 degrees Celsius).
"No matter what we did, we could not observe anything akin to the Mpemba effect," one of the researchers, Henry Burridge, told Science News.
So what's actually going on here? We'll have to wait and see which conclusion - if any - bears out with further research, but one thing's for sure when it comes to water - it's still surprising us, even after all these years.