Saw this in a comment somewhere.

There have been a huge number of studies using rats and mice negotiating mazes. They purport to do things such as testing whether drugs enhance or impare functioning.

So some people, for whatever reason, decided to test the tests. To see if the maze tests actually tested what the people thought they tested. In essence: they took a rat’s eye view of the mazes, “if I were a rat in the maze how would I “see” things.” They started, and went on an on, eliminating clues that rats would use, urine trails, hair drops, on and on, vibrations from outside traffic, all kinds of things I don’t even remember. Several times they thought they got it right, then they discovered there was some other clue the rats were using.

Finally, they did get it right. No clues for the rats, all of the clues finally eliminated. Then, testing the tests. None of the tests of drugs etc. could be reproduced. None of them. The rats had been negotiating the mazes based on clues in the mazes and surroundings.

The point of all that is: the researchers may actually have got what they say they got, but it isn’t reproducible because they don’t understand what was “exactly right” for that test.

Even when it works they don’t know why.