Scientific American has an article about a study that shows that people can (with better-than-chance accuracy) identify which men are gay and which are straight solely by looking at faces.
There are all sorts of nitpicks and questions to be addressed there--like, hey, what about bisexuals? And does "better than chance" mean 60% (in which case it's scientifically interesting but not all that practical) or 90%? And did the study gather any information about the study participants' backgrounds (like whether they are, or know anyone who is, gay)? And did they try showing photos of guys who were gay but closeted? And how do they know that all the supposedly straight guys whose pictures they used were really straight? And so on.
The studies themselves probably answer most of those questions, but all I've seen is the abstract of one of the studies, plus this Sci Am article.
But it's still a pretty interesting study, and it goes further than I would have expected in controlling for various factors.
Note: When I link to articles about studies, people often post comments claiming that the study is flawed based on my description of the study. I strongly recommend taking a look at the article before suggesting ways that the study is flawed; as is often the case, some of the obvious ways that it could be flawed were actually addressed by the study itself. And, of course, the article is not the study; news articles often get things wrong or leave things out when describing scientific studies.
(Wrote this back in late February, but didn't post it for some reason.)