Biology is complicated; it's hard to say there couldn't be a fundamental issue that meant the average man was better/worse at something than the average woman.
BUT: let's talk about that after all the frank discrimination has gone, at which point we might have a hope of understanding something. We need to get rid of the discrimination anyway, since we're only ever going to be talking about overlapping distributions.
There is a lovely research methodology that involves taking paperwork about an individual, e.g. a CV, and manipulating it so that it shows the individual to be male/female/white/black/a parent/whatever you want to investigate, while not changing any aspect of the paperwork that ought to make a difference. Then you give contrasting versions of the paperwork to a group of evaluators. You find - robustly, and replicated in many fields now - that a candidate perceived as female will be judged worse than one perceived male, etc., even when there simply is no difference besides the name.
Moreover, humans are good at rationalising patterns, and it behooves us to remember that we are often confabulating. E.g. proportions of women in different subjects vary dramatically from country to country, casting doubt on the idea of innate biological influence. Another example: the proportion of women in mathematical computer science is much lower than that in less mathematical areas of computer science, which sometimes leads people in CS to claim that it's just that women don't do maths - except that the proportion of women in maths is much higher than the comparable proportion in pretty much any part of CS! (While elsewhere, people like to claim women are drawn to interdisciplinary areas ;-) I think a lot of it will be historical accident causing cultural differences - stuff like role models really matters, and those vicious or virtuous circles have a *huge* influence.
