OddThinking

Gaydar Alert!

During the play’s interval, Sarah confided, “I think the husband is gay.”

“Who?” I asked, “The character or the actor playing him?”

“The actor!”

To me that seemed like a bold claim – to be able to reasonably determine the sexuality of a person who you have only seen adopting a very masculine character role in a play. I was a little skeptical.

I’ve seen “gaydars” go wrong before. Several times, women have informed me that male friends of mine are indubitably gay. Several times, I’ve had to reply: “Really? Perhaps you should warn his girlfriend then.” So, I am dubious about claims of gaydar accuracy.

I explained that it would be a good idea to have an experiment – put 50 straight men and 50 gay men in a room, and let 20 women mingle with them and chat, and come back and give a rating. We could test the accuracy of gaydar; it is probably better than random, but how much?

Today, it occurred to me that this scientific experiment has probably already been done. Here’s what I found.

The Studies

Visual Cues

In 1999, Ambady, Hallahan and Connor tested the abilities of straight and gay men and women to pick the sexual orientation of straight and gay men and women based on photographs, 1-second silent film clips and 10-second silent film clips. Their results were “more accurate than chance” but I haven’t forked over the money (nor got up and visited my nearest university library) to read their journal article to find the true accuracy.

William Lee Adams was an undergraduate who replicated their results in 2005. I haven’t found his results published – Hey, he’s was an undergraduate; it probably didn’t appear in a peer-reviewed journal – except a mention in Psychology Today and a rather informal and chatty Weekend America NPR radio show. The media shared plenty of anecdotes, but not many numbers that explained the strength of people’s gaydars. Adams came across with some extraordinary claims about his own gaydar, supported only by anecdotes. The closest claim I found to gaydar strength was a statement that “some people had remarkable accuracy of 90%.” He also claimed that the gaydar acted fast (less than 2 seconds) and subconsciously, and that gay people were better judges than straight people.

Audio Cues

Rogers and Smyth found that subjects listening to straight and gay men could get it right 62% of the time.

Odour Cues

Martins, Preti, Crabtree, Runyan, Vainius and Wysocki found that the pleasantness of other people’s odour depends on both your gender and sexual orientation and that of the other person. An interesting result, but unlikely to be relevant for most “gaydar” scenarios. I would think that asking someone about their preferences is slightly less socially taboo than sniffing their armpits and evaluating the result.

Putting This Together

So, if I hear a woman declare that some man is gay, based on a brief meeting, how much stock should I put in that? Looks like it is time to whip the back of the envelope for some calculations.

First, let’s be very generous to the claimant by assuming that she is gay (because they are better judges, according to Adams) and that she is one of these top performers based on visual cues (i.e. 90% accurate).

Now, let’s be extremely generous, and assume that the extra audio information can work successfully (at the same 62% rate) on the 10% of errors of the visual cues, giving an overall accuracy of 96.2%). In practice, this is too high – I would only expect a 2-3% advantage by adding the audio cues.

I don’t have any data available differentiating the specificity or sensitivity of the gaydar, so I assume that they are both the same (i.e. 96.2%)

Using what I learnt from Gigerenzer, let’s imagine this scenario played out one million times.

1.1% of the million men (or 11,000 men) are gay. Of those, 96.2% or 10582, would be successfully detected by the gaydar.

The rest of the men (989,000 men) are straight, of which 3.8% (37,582 men) would be falsely detected as gay.

So in every million men, 48,164 men are declared gay, and only 10,582 of those men actually are gay – about 22%.

To conclude: if a woman tells me a man is gay, there is a 78% chance she is wrong, even if I make absurdly generous assumptions about her gaydar ability.

Caveats

• This article is unreservedly gender-biased, purely because I do not ever recall hearing any of my male friends seriously declare one of other friends to be gay based on a short meeting, but I have heard it many times from my female friends. If the situation did reverse, I would be equally skeptical of the claim.
• The figures would look slightly better for gaydar if the 1.1% figure is too low, especially in the group you are sampling (e.g. in this example, actors). At the 5% rate, the claimant is only wrong 43% of the time; don’t forget to substitute lesbian rates if you are reversing the scenario.
• This discussion takes an overly-simplistic black and white view of sexual orientation – you are either heterosexual or you are homosexual. I let myself fall for the fallacy of the false dilemma because (a) the concept of “gaydar” makes the same assumption and (b) the scientific data I had did too. I consider it preferable to exclude the middle ground than to permit claimants proven wrong to get away with “Oh, well he’s still gay but he doesn’t admit/know it.” Statements like that makes the whole claim unfalsifiable, and therefore not worth discussing scientifically.

Coda

After the second act, I was able to point out to Sarah that the actor was having a drink with friends in the bar. He was wearing a wedding ring; not proof that he wasn’t gay, but a far better predictor than “gaydar”.

[Update: Added the third caveat which I accidentally omitted.]