You sit men and women down and give them a maths test, and they will do fairly equally. Then you set up the same test, but with different people, and make them tick a box to say whether they are a man or a woman, and the women do significantly worse in the maths test than they did previously in a group set.