From a news follow-up article, “Do genes play a role in science gender gap?”, by Carolyn Y. Johnson of the Boston Globe:

Summers was faulted by some conference participants and others for drawing a link between test scores and career success -- a link that they said has been disproved by research.

Really? Disproved? Research has shown that there’s no link? This is research I’d like to see.

Yu Xie, a sociology professor at the University of Michigan who spoke at the conference, has shown that men outnumber women 2-to-1 in the top 5 percent of scorers on math tests -- research that Summers cited in his talk. But Xie said in an interview that his studies show that not all the best scorers succeed professionally, and not all men with successful science careers received the top scores.

Sorry, those aren't the research results we need to show that there’s no link. Of course not 100% of the top-5% scorers end up in the top (say) 10% of their fields. But surely more than 10% do? What is the percentage? Of course not 100% of all men with the most successful (top 5%) careers are among the top 5% of test scorers. But surely more than 5% of the most successful scientists are? What is the percentage? I would find it hard to believe that there’s no link, that the top 5% of test scorers are not more heavily represented than others (are not more than 5% of the population) among successful career scientists. I await the evidence to the contrary.

Let me borrow an analogy from David Bernstein of The Volokh Conspiracy. Suppose a hypothetical Larry Winters says that perhaps, in part, goyishe men dominate the National Basketball Association because there is a link between height and basketball success, and goyishe men are more heavily represented among the very tallest. The analog to Xie’s comments would be: “But wait! Not all the tallest people have successful NBA careers! And not all successful NBA players are above the 95th percentile in height!” True and true, but beside the point. Less-than-perfect correlation doesn’t show that height doesn’t matter, or that there’s no link between height and professional basketball success.

Xie and Kimberlee Shauman of the University of California at Davis, who coauthored ''Women in Science: Career Processes and Outcomes," concluded that no single factor could explain why women did not make it into the most elite science positions, and that a combination of social, cultural, and psychological factors were probably at work.

Again, not the evidence we need. Lots of other factors also matter, sure. Test scores are not a perfect predictor (do not have an r-squared of 1.00), sure. That doesn’t show that test scores have no predictive value, other things equal. Or even that test scores aren’t a better predictor than any other single factor or combination of factors.

To be fair to Prof. Xie, please notice that he isn't quoted as saying that the existence of a link has been disproved. That claim the reporter only attributes to an unnamed "they".