An interesting study is being widely commented on, but, as is often the case, the press glosses about "harassment" are a bit askew. For instance, the Washington Post (cleverly) reports that "Men Who Harass Women Online are Quite Literally Losers." The actual study is entitled "Male Status and Performance Moderates Female-directed Hostile and Amicable Behaviour". In the study, Halo 3 games were recorded by way of three experimental player accounts: control (no voice channel), male, and female voices. Interactions with these accounts and other (focal) players were transcribed and coded (N=126) as positive, negative, or neutral. Skill level of focal players correlated with the valence of their comments; that is, higher skill male players correlated with more positive and less negative statements towards women.
The authors don't mention "harassment." Also, because of the small sample size and that only 13% of that (11 individuals) uttered hostile sexist statements, "We found that the presence of sexist statements was not determined by differences in maximum skill achieved." The paper is really about the extent to which lower-status male players are bigger jerks to women players. They did find this with respect to negative and positive statements -- and we could (rightfully) call this sexist itself -- but they didn't have the statistical power to conclude anything about hostile sexist statements.
What I found interesting, methodologically, is that for the analysis they had to exclude two mega-jerks as outliers. "For the examination of negative statements, there were two focal players in the female-voiced manipulation that made 10 more negative statements than the next highest individuals (greater than 5 standard deviations from the mean). As a result, we removed them from our analysis to ensure they did not skew our results towards significance." Given the "rotten apple" thesis (a minority of jerks can spoil the barrel), what they did for the purposes of statistical analysis doesn't correspond to the experience women players may have. A minority of awful people can send the majority of awfulness. That is, I believe, if we excluded 5% of the most awful people online as outliers, the Net would be a lovely place!