They measured, for instance, how often each student responded to questions posed by professors to the classroom as a whole. At the start of the semester, 11 percent of the female students attempted to answer questions posed to the entire class when the professor was male, and 7 percent of the female students attempted to answer questions posed to the entire class when the professor was female. By the end of the semester, the number of female students who attempted to answer questions posed by a male professor had not changed significantly: Only 7 percent of the women tried to answer such questions. But when classes were taught by a woman, the percentage of female students who attempted to answer questions by the semester's end rose to 46.
The researchers also measured how often students approached professors for help after class. Around 12 percent of the female students approached both male and female professors for help at the start of the semester. The number of female students approaching female professors was 14 percent at the end of the semester. But the number of female students asking for help from a male professor dropped to zero.
Finally, when Stout and Dasgupta evaluated how much the students identified with mathematics, they found that women ended up with less confidence in their mathematical abilities when their teachers were men rather than women. This happened even when women outperformed men on actual tests of math performance.
Full story from Slate.