31C is 88F. It's not absurdly hot, but it's hot enough that it's surprisingly hot to me.
Controlling a variable means setting up the experiments such that the variable doesn't influence the results. This is typically done by making sure it is the same in both the experimental and control group. Experiments typically control a large number variables, and only vary a small number (ideally, one).
On the other hand, prior to this experiment, there was no reason to suspect that mice at 31C would be significantly different than mice at 25C. So why waste time, money, and mice by re-doing everything at 31C? Remember, things don't end well for the control group just because they're the no treatment group: they've got to be killed and autopsied at the same rate as the experimental group. When the study says "chemical X resulted in a 35% increase in tumors", they know that because they cut open both the experimental and control group, counted their tumors, weighed them, compared them, and came up with that number of 35%.
Prior to this experiment, there was no reason to think that mice at 31C would be any different to mice at 25C, any more than you would expect mice wearing funny hats to be any different than mice without. Would you expect scientists to run a second control group of mice wearing funny hats, and slice them open and count their tumors, just in case it ended up being different?
I mean, England barely ever has a high of 31C, and there are plenty of mice there. I wouldn't suspect that every mouse in England is immunocompromised.
Nests in cages mean mice can't be observed, so are usually not provided.
Mice are kept lean, so have limited body fat for warmth.
Mice are kept in overly-well ventilated cages, which adds to the cold felt.
Many mice are bred for limited or low fur, which again means they need a higher mean temperature to stay warm.
Lastly the cages encourage non wild behaviour, and are rarely enriched with toys, or have long spaces in which to run, which the animals also use to mediate their body temperature.
The moden scientific laboratory design puts so much emphasis on precision (cages, diets, etc, for consistent experimental outcomes), that it fails miserably at accuracy (similarity to what non-experimental individuals)
Mice have been valuable tools not because of how good a model they are but because they're cheap and fast. Apes are a much better animal model for most purposes, but they're expensive and slow, and raise way more moral questions. And people as test subjects are all that and a bag of chips.
Instead, researchers using animals work to understand how an animal model compares to humans. They pick their animal test subjects based on what they're researching; for instance, rabbits might be better for researching a particular disease, and cats might be better for researching a particular cognitive disorder.
So -- assuming this new information holds up to replication -- what this means is that we potentially have a more accurate, cheaper animal model on hand than we previously thought. If mice at 31C behave significantly differently than mice at 25C, then researchers will be able to use mice in cases they previously weren't, eliminating the need for other animal models or even some human testing in some cases (eg, the experiments will fail at mice and never make it to human testing). This will save money, and speed up research.
http://en.wikipedia.org/wiki/Gravity_of_Earth#Variation_in_g...