There are lots of studies where people are fed some sort of single dish meal, say a stew, in a high and a low fat version, with attempts to make the stews indistinguishable. Because, double blind, right? Generally, people will eat similar amounts of the two stews, this is described as "failing to compensate for the fat calories." See the problem with that? Set things up, so that people can't distinguish between two foods which vary in calories, and then find it remarkable that they eat more calories with the higher calorie version....

There's a rat study where potato ships, high or low fat, are fed to the rats. Always feed them high fat, or always feed them low fat potato chips, and there's not much difference. But if you feed the animals low fat potato chips one day, high fat the next, and so on, they'll overeat. They'll fail to compensate for the fat, yes--because the low fat potato chips are mocked up to resemble the high fat ones. The low fat days set the animals up to over consume on the high fat days. A similar study was done with sugar and artificial sweetener. Sugar one day, artificial sweetener plus some sort of bulking agent to match the chows for calorie density the next--the animals got fatter this way than with just sugar, or just artificial sweetener every day. And again, the researchers explained this as the animals failing to compensate for the sugar calories.

It sort of makes sense that the animals would learn to eat enough of the low calorie chow to avoid discomfort of hunger. If you take away predictability--it makes sense that the lesson they'd learn is how much food predictably protects them from hunger--that hunger would be a stronger motivating force than the threat of feeling a little bit overfed sometimes.