John Tierney argues that Larry Summers' take on the role of gender and science, which got him fired as president of Harvard, was right:

The Duke researchers — Jonathan Wai, Megan Cacchio, Martha Putallaz and Matthew C. Makel — focused on the extreme right tail of the distribution curve: people ranking in the top 0.01 percent of the general population, which for a seventh grader means scoring above 700 on the SAT math test. In the early 1980s, there were 13 boys for every girl in that group, but by 1991 the gender gap had narrowed to four to one, presumably because of sociocultural factors like encouragement and instruction in math offered to girls.
Since then, however, the math gender gap hasn’t narrowed, despite the continuing programs to encourage girls. The Duke researchers report that there are still four boys for every girl at the extreme right tail of the scores for the SAT math test. The boy-girl ratio has also remained fairly constant, at about three to one, at the right tail of the ACT tests of both math and science reasoning. Among the 19 students who got a perfect score on the ACT science test in the past two decades, 18 were boys.
Meanwhile, the seventh-grade girls outnumbered the boys at the right tail of tests measuring verbal reasoning and writing ability. The Duke researchers report in Intelligence, “Our data clearly show that there are sex differences in cognitive abilities in the extreme right tail, with some favoring males and some favoring females."

Mark Perry adds some related charts of SAT math scores by gender:

A couple comments. It's very difficult to disentangle nature from nurture. There's nothing like having a daughter to make you understand the ways that culture can effect a person's development from a very young age. But I think the point is that, from the perspective of how a university should treat the problem, it fundamentally doesn't matter. From a societal perspective, it's crucial to identify sexism and differential socialization that may contribute to the disparity in male-female math performance at high levels. But Summers was addressing how universities should treat the problem, and it's clear that by the time you're dealing with university-level students, a significant gender disparity is already baked in the cake. This is not to say that raising awareness of gender discrimination has no role in addressing the problem, merely that we need a realistic baseline of what non-discriminatory policies at the university level can achieve.

Part of the argument on this topic, of course, centers around the permissibility of discussing it. Liberals have a strong taboo against discussing the role natural ability, as opposed to discrimination or societal pressure, plays in the math and science gender disparity. The best justification for the taboo is that public discussion of gender differences in cognitive ability in math and science will become distilled into an oversimplified message that "girls can't do math," and thus discourage girls from developing their talents.

That's certainly a valid fear. But it seems to argue not for a taboo but for taking care to ensure that the issue is broached in a careful way, acknowledging that the greater number of boys with very high-level math ability does not mean that girls can't also perform at a high level. Indeed, it seems to be the very people creating the taboo who are distilling this claim into an oversimplified "girls can't do math" message, which is quite the opposite of what Summers argued.