Sorry to harp on this topic, but I wanted to produce one more bit of yield from my investment in trolling through the blogosphere.

PZ Myers offers a treasure trove of non-refutation methods here.

1. Argue that where there are many causal factors, it is wrong to think that any particular factor is important:

these statistical games [citing the different distributions of male and female math test scores] may be correct, but if and only if the property of success in science and math is a simple one, with one quantifiable attribute that is an indicator of this mysterious parameter called “math ability”

2. Relatedly, argue that to view any one factor as important is to deny all others:

people like Summers are trying to impose a single simplistic standard on scientists

3. Use the ad hominem freely -- Summers’ supporters are “chauvinists”, and Steven Pinker is a bunko artist:

One extremely popular source among the defenders of chauvinism is Steven Pinker. […] If people started walking out on presentations of fact-free, unsupported hypotheses, Pinker wouldn’t have a career.

4. Deny that there is any relevant evidence on male-female differences from math test results, or from anywhere else:

Summers ... presented a badly formed hypothesis with no evidence to support it

5. Compare the Summers Hypothesis to creationism:

This is exactly what we see from creationists, too.

6. Accuse Summers of bad faith:

We consider hypotheses of innate differences all the time in science; that’s very different from an administrator using half-baked ideas to rationalize away cultural stereotypes and prejudicial policies.

7. Turn "prejudice explains the observed pattern" from a hypothesis to a fact, and claim that its explanatory importance renders genetic factors of no possible relevance:

It just seems to me that the fact that women are subject to widespread, long-term bias against their scientific abilities, yet some still persevere and manage to make it, is convincing evidence that the “hypothesis” that they are innately inferior in these fields is bogus.

8. Insulate your position from refutation by refusing to look at any statistical evidence gathered in the world as it currently exists:

You can come back and tell me about “distribution curves” and “long tails” when the playing field is level and you can actually legitimately provide appropriate data.

To be fair, buried in Myers' rhetoric is one valid point:

you can’t use these kinds of distributions to argue for innate differences—they can be equally well (better, to my mind) explained by environmental factors.

If it were true that the different distributions of male and female teenage math test scores can be largely explained by environmental factors, then increasing the representation of women in science would still remain an uphill battle – only it would be a battle against deep-seated (child-rearing) culture rather than against our genes. The relatively small number of females among the top scientists would still be due to a relatively small pool of females choosing to enter those fields, rather than to "prejudicial policies" by universities.