Not that Bowles is against teaching negotiation to women; her book Women Don't Ask (2003) with Sara Laschever released a new edition earlier this year. Bowles, Babcock, and McGinn (2005) suggest some mediating and moderating factors for the gender effect on salary negotiation such as "the degree of uncertainty in parties’ understanding of the economic structure of the negotiation" makes a huge difference. Where uncertainty is low, women do nearly as well as men in negotiating. Training may be able to reduce the ambiguity for women.
So. How do we fix it? Funny that we were just talking about self-perpetuating differences yesterday in the comments. Women don't ask for more money, so it stands out more when they do. A woman asking for more money is rare; a man asking for more money is normal. So when a woman asks for more money, she's "unpleasant", and she doesn't get the job, or deals with resentment. Next time she doesn't ask. The man asks for more money, gets more money, and next time he asks again. How does this tie in with structural ambiguity? Perhaps it is just that when negotiations are understood to be part of the culture, women do it more often, and the rarity factor declines.