I've heard that girls learn better in an all-girl classroom setting, but that boys learn better in a mixed boy/girl classroom composition. Is there any statistical evidence that this lends merit to this argument? What factors were measured, and what is the psychological explanation behind the findings?
The answer depends a lot on the culture the children are raised in. For example, when working with children raised in a culture where strict gender roles are the norm, and boys and girls aren't normally allowed to socialize with one another, I found that children performed considerably better when segregated by sex. Boys and girls just weren't sure how to act one another, and they were wasting processor cycles on that instead of their schoolwork.
On the other hand, when working with children raised in a culture where sex roles were not so divided (boys and girls might favor different toys, but they socialized freely and didn't see one another as foreign and confusing), I found no difference in the performance of same-sex and mixed classroom in aggregate.
However this isn't great anecdotal evidence, because I never experienced a case where classrooms segregated by sex were taught the same material the same way. In every case, girls' classes centered around language skills and promoted conservative strategies, while boys' classes emphasized math and science and promoted risk-taking. This could be interpreted to mean that girls and boys raised in strict sex roles have different educational needs, or it could be interpreted to mean that girls and boys raised that way appear to perform better when the subjects and strategies stereotypically associated with the other gender (and bearing a stigma for their own gender) are de-emphasized.
As a side note, though the question only addressed educational performance, I felt the need to address the impact on socialization between the sexes as well. In both settings, children educated in same-sex settings had trouble working with the opposite sex in a professional setting as adults (some got over it with experience, some didn't, and it was far more pronounced in cultures with strict sex roles). Additionally, teenagers who spent their entire childhoods in same-sex education tended to have more stereotypical views of the opposite sex: this was adaptive in a society where those stereotypes were the only permitted behavior, it was like learning a script you would have to perform later, but was maladaptive in a more varied society where both boys' and girls' real behavior was more individual.
Most of my queries on Google Scholar returned results for a very specific age group (grade school, college, etc), subject (science, computer skills, etc), or learning approaches/settings (in-class instruction, small group projects, etc). Finding a more broadly-scoped experiment or observational study on this topic would require quite a bit of digging, and perhaps some help from a subject librarian at your local library or college/university.
HedgeMage mentioned the distinction between academic learning and social learning, and I would like to underscore this distinction. The answer to your question will likely be different if you only care about academic performance vs if you only care about social development vs if you care about overall performance/development. Complicating the matter further, both academic performance and social development can be broken down into multiple categories (exposure to facts/ideas, development of important technical skills, ability to work with others, respect for other people, respect for other cultures, etc). And there are probably other aspects of the educational process that I'm not even considering now!
This is probably why it will be difficult to find any statistical evidence for a broadly-scoped study of this topic, and why so many of the studies focus on specifics. It might be worth looking at a few different studies that approach the topic from different angles.