I'll give a little example here. A lot of programmers think that good programmers are the kind of people who program in their spare time, for fun, with no particular goal in mind other than having fun with programming. It seems like a reasonable conclusion because they are that way, and perhaps all of their colleagues are that way. I read this book
a long time ago, and I don't own it so I can't look in it again, but I think I can remember enough to make this point… generally speaking, women are more likely to be into programming because of an end goal (i.e. "the program I'm working on will accomplish something good, therefore I am motivated to do it", rather than for the sake of programming itself. They may be just as good at it, but they don't sit around dreaming up things to program in their spare time. If you give them a task, or if they happen upon an open source project they care about, then they rock it. But if you ask them in an interview if they spend a lot of spare time programming for fun (a question that I myself have been asked in interviews both for graduate school and for jobs), they're more likely to say no.
The people asking this question aren't intentionally trying to weed out women, and the women answering it aren't intentionally making a choice about what job to take or what graduate school to go to, but little things like this could add up and mean they aren't, in the end, getting the same choices.
I don't have any statistics to show that women who don't program in their spare time are suffering lower pay because of it, or anything like that, but I'm just wondering about all these little subtleties and how they can add up without anyone realizing it.