The media aregleefullyreporting a recent paper showing that "gaydar is real" - we can tell who's gay just by looking: The Roles of Featural and Configural Face Processing in Snap Judgments of Sexual Orientation
While it's a fine paper, I'm afraid that the results really aren't that exciting.
American undergraduate students were able to classify people as gay or straight with better than chance accuracy, based purely on photos of their face. For male photos, the hit rate was 0.57; for women it was better with an accuracy of 0.65.
However, that's on a scale where you get 0.50 by flipping a coin. So saying that gaydar is '65% accurate', as almost everyone has, is misleading. Still, the numbers seem solid. The sample sizes were large and the effect was replicated very convincingly in two experiments.
However... this tells us very little about real world "gaydar", and it wasn't intended to. There are reasons to think it could underestimate the accuracy:
Most importantly - people only saw the pictures for 50 milliseconds each. 1/20th of a second. Followed by a backward mask. That's right on the threshold of conscious perception, almost 'subliminal' but not quite. With longer viewing times, they might have done better.
All the faces were black and white photos with the hair and ears cropped out (see above - and I think those two photos from the paper are the authors, although I may be wrong!). Anyone with facial hair, glasses, or any other 'accessories' wasn't used. In the real world, we have that extra information.
In real life, we get clues from facial expressions, body language, voice, clothes. You could argue that these are being used (consciously or not) specifically as signals of sexuality, so they don't count as 'gaydar' - but more on that later.
But it could also overestimate gaydar's powers:
These were photos that people chose for their Facebook profiles. We all know how much effort some people put into that choice. We also know that different photos of the same person can often seem like two different people. Your Facebook pic is probably the most "selected" photo of you in existence. It would be better - but also much harder - to use passport photos.
All of the gays in the study were out of the closet: they broadcast their sexuality on Facebook. But lots of gay people don't do that. Now those cases are probably where 'gaydar' is most likely to be of interest to most people, I think; those people might be harder to spot.
As far as I can tell, this study wasn't intended to "prove that gaydar works". It was meant to examine how it works, by seeing whether it works very quickly (yes - in 50 ms in some cases). The authors also tested how accuracy was changed by flipping the photos upside down; this reduced accuracy but it was still well above chance.
Ultimately, we need to ask what "gaydar" means and why we find it so interesting.
On a superficial level, it just means being able to sense, from someone's appearance, if they're gay. That certainly does 'work' - if you see a guy coming out of a gay club in a tight pink Boy George t-shirt then yeah, he's probably gay. But he's (effectively) told you so, by being in that club and wearing those clothes, so that's not very interesting. That's an extreme case, but clearly people advertise their sexuality (and much else of course) all the time. Gaydar, in a weak sense, is just perception.
I think what makes "gaydar" intriguing is the stronger idea that it can go beneath such adverts. That we can see who's really gay, whether or not they admit it, even to themselves. If that were possible, then it would seem to mean that homosexuality is part of the essence of some people - in other words, that it's a biological trait.
So gaydar in a strong sense is risque. It calls to mind un-PC ideas such as physiognomy and would seem to validate various stereotypes which are the stuff of dirty jokes more than polite discussion.
Does gaydar in this strong, exciting sense exist? That's another question. This study doesn't tell us.
Tabak, J., and Zayas, V. (2012). The Roles of Featural and Configural Face Processing in Snap Judgments of Sexual Orientation PLoS ONE, 7 (5) DOI: 10.1371/journal.pone.0036671