Are homophobes secretly homosexual?

It's an idea that's been around since Freud - but is there any decent evidence for it?

Does anyone remember Ted Haggard? He was the Evangelical Christian megachurch pastor from Colorado who famously did a testy interview with Richard Dawkins in 2006, at the beginning of the “New Atheism” craze. As well as preaching creationism—the main subject of his clash with Dawkins—Haggard was big on “family values” - a marriage should be between a man and a woman, gay sex is immoral, that kind of thing.

And then, of course, it turned out Haggard had been having extra-marital affairs with not one but two male sex workers - causing him to lose his hugely prominent church position (after this was all discovered, he changed his view to support gay marriage).

The fact that I said “of course” in the previous paragraph—and the fact you knew what I was about to say—speaks to how well-known this idea is: homophobes are often themselves secretly homosexual. This idea goes all the way back to Freud, whose idea of “reaction formation”, where we deal with anxiety or threat by acting in a completely opposite way from the thing that’s worrying us, along with the idea of “latent homosexuality”, would seem to explain Haggard’s anti-gay sermons.

It’s not just Haggard, either: there are enough “Homophobic Leaders Who Turned Out to be Gay or Bi” to make entire listicles, and there’s a website called gayhomophobe.com that provides all the links (the one most familiar to me as a British person is Cardinal Keith O’Brien, the head of the Catholic Church in Scotland, who turned out to have sexually harassed and assaulted multiple male church members who were junior to him over a period of decades).

Christopher Hitchens phrased it most memorably—though perhaps not in the most politically-correct manner—in his memoir:

Whenever I hear some bigmouth in Washington or the Christian heartland banging on about the evils of sodomy or whatever, I mentally enter his name in my notebook and contentedly set my watch. Sooner rather than later, he will be discovered down on his weary and well-worn old knees in some dreary motel or latrine, with an expired Visa card, having tried to pay well over the odds to be peed upon by some Apache transvestite.

There’s no shortage of anecdotes. But is the idea true in general? Is it really the case that homophobia is related to latent, repressed, or closeted homosexuality? Every so often you see one of two or three studies being passed around on social media that claims that this is true - but are these studies any good? I’ve always been interested in taking a look - and since it’s Pride Month, it seems like a particularly good time for it.

The really famous study

The most-cited and most-shared study on this topic is a pretty eye-catching one. That’s because it used penile plethysmography. A plethysmograph is a kind of strain gauge that lets you measure differences in the circumference of some body part - for example, in people with vascular disease they’re sometimes used to measure blood circulation to the limbs. But in this case, they were used to measure the “tumescence” of the participants’ penises.

This was a 1996 study, from the University of Georgia. The setup and the findings are irresistible: they took 35 homophobic straight men and 29 non-homophobic straight men. They showed them straight, gay, and lesbian pornography. And, as the study’s Abstract has it:

Only the homophobic men showed an increase in penile erection to male homosexual stimuli.

So straightforward. So telling! But let’s not get too excited [note to self: think of less unfortunate phrasing here]. Let’s have a look at the actual numbers they report.

How the measurements were made in the 1996 study (not really! But not actually that far off)

They measured the two groups (homophobes/non-homophobes) with the three different types of pornography (straight/gay/lesbian) over six different “time blocks” (each lasting 40 seconds). So the statistical result of interest is a group×type×time interaction: the penile circumference rises faster over time in the homophobic group, to the gay porn type only. And that’s what they found… except it wasn’t super-impressive: it’s just reported as “p < .05”, but I make it (you can use one of these little calculators online) to be p = .02.

Now, that’s technically a “significant” result - but like the dog-racism results we saw last time, it’s the kind of border-zone significant result you wouldn’t want to bet any amount of money on until you saw a replication. That’s especially true since this isn’t exactly a massive sample, and the complex three-way interaction effect could be driven by just a few stray datapoints. It could, in other words, have been pure luck.

There are also some weird decisions in this study, like the way they categorise the men into only two groups. People aren’t either homophobic or non-homophobic: they run the whole range from extremely accepting, to indifferent, to totally bigoted. The reason they did this split, they say, is that they couldn’t find “an adequate number” of men (they don’t say how many) who had very low levels of homophobia according to their questionnaire, so they lumped everyone into two categories. That’s depressing in a social sense, but it’s still a weakness in the study - stronger evidence would have come from a dose-response relation where higher and higher homophobia levels were related to more and more arousal.

Instead, we have this group difference, just from splitting the measure in half: people with a score of 0-50 were non-homophobic and 51-100 were homophobic. Is someone with a score of 51 really that different in their level of homophobia from someone with a score of 50? Well, this study would have counted these people as entirely, categorically different. Is it really that meaningful if you find a difference between groups created in this way, even if it is statistically significant? It all depends on the actual distribution of the data - and of course, this being an “old” paper (from 1996, remember), there’s no way of getting access to the dataset.

And while we’re talking about interpreting the results, here’s another problem: what’s an increase in penile tumescence actually measuring? It could be sexual arousal, sure - but the (small amount of) evidence suggests that at least in some cases, “arousal” as measured by a plethysmograph might increase when someone is feeling anxious. So if the more homophobic men felt more anxious or uncomfortable when seeing the gay pornography, this could’ve been reflected in their physiology in a way that didn’t actually reflect a secret attraction.

So, the famous plethysmography study, beloved of people making a point online, provides only very weak evidence for the “homophobes are homosexual” theory. And here’s another thing: nobody has ever published a replication of it. Sure, it’s a bit of an awkward study to run, but many sex-research labs and clinics do have the necessary equipment, and you’d think someone over the past 26 years would’ve given it a go. One possibility is that someone has tried to replicate it, and didn’t find anything, so never bothered to send it for publication.

But even if there aren’t any other studies using the penis-measuring technique, there are a bunch of studies that attempt to look at this same question in other ways. Let’s take a look.

All the other studies

Here are all the other studies I could find that attempted to answer the question of whether more homophobic people are actually more likely to have same-sex attraction:

• A 2006 study of 44 undergrad students. They were asked to rate pictures, some of which were of gay couples. Secretly the amount of time they looked at each picture was recorded. The more homophobic students—but only if they had high levels of self-deception—looked at the gay-coded pictures for a significantly shorter time.
  
  The authors’ interpretation is that “defensive” homophobics do actually have something like a “phobia”, where they find it very unpleasant to look at gay images for long. It certainly doesn’t look like a “secret attraction”. Nevertheless, this is a tiny study that has results that are based on an interaction, so again would require replication. It doesn’t seem super-reliable to me.
  

• A 2011 study of 132 male psychology students that attached electrodes to the muscles under their eyes to check their startle response when loud noises were paired with erotic gay or straight images. Less of a startle indicated a more positive response to the picture; the idea was that if people high in homophobia had an initially positive response to photos of gay couples, and then hid it with a later negative response, it might indicate that they had an unconscious homosexual desire that was being repressed. They found that more homophobia related to a more negative startle response, and it didn’t change over time, indicating that the more homophobic people might’ve just been more genuinely offended by the pictures.

• A 2012 study using subliminal methods. This one got a lot of coverage in the media at the time. Across 4 studies (n’s = 89, 181, 189, and 184, respectively), they derived an “implicit sexual orientation” - which often didn’t relate very much at all to a person’s stated, explicit orientation. They did this by seeing how quickly the person reacted to words like “gay” or “straight” after being subliminally primed with the word “me” or the word “others” (being quick to react to “gay” after being primed with “me” means you’re more likely to be secretly gay, according to this paper).
  
  They repeatedly found that people who stated they were gay had no, or a negative, relation between their implicit homosexuality and their level of homophobia. But for those who stated they were straight, the more implicitly, secretly gay they were, the more homophobic beliefs they held.
  
  The p-values for these results (all statistical interactions, again) are .04, .01, .03, and .03. On the one hand, they do all replicate that same result. On the other, none of them are very convincing: if the result was real (and not due to noise or p-hacking), it would be unlikely that every p-value would be so close to the p < .05 threshold. In any case, even if we had a solid result, it has similar interpretation problems to the plethysmograph study: what if homophobic people react differently to the IAT not because they’re secretly gay, but because the idea of homosexuality makes them upset or angry or nervous - because they really genuinely hate it? A Slate article by Daniel Engber from the time discusses this and some other important criticisms.

• A 2013 study of 237 male and female undergrads, using the Implicit Association Test (another subliminal-ish measure which is supposed to reveal any automatic associations you’ve made, say between words describing your outgroup and negative, derogatory terms). Using this, they also derived a measure of “implicit sexual attraction” - reacting more quickly to pairs of words that mentioned the opposite sex and words to do with sexual attraction indicated you were implicitly heterosexual, and the other way around for implicitly homosexual. They failed to replicate the effect in the 2012 study: there was no evidence that explicit heterosexuals with higher implicit homosexuality were more homophobic (in fact, if anything there was evidence for the more straightforward view that people who have more implicit homosexuality are less homophobic).

• A 2013 PhD thesis (I don’t think published in a journal anywhere) that examined the question using a kind of electrical brain response. There are particular “appetitive” brain responses that we know about—responses that come up when people get things they want or like. It would be interesting if more homophobic people had these kinds of responses when they saw homosexual erotic pictures. But in this n = 48 study, that wasn’t the case.

• A 2016 study where 38 students did a looking-time task, like the one in the 2006 study above, and had their eyes tracked so the researchers knew they were looking at the relevant parts of the images. They also did a “manikin task”, where you drag a little on-screen figure towards or away from photos of gay or straight erotica. They found that “homophobic men may indeed have a higher sexual interest toward homosexual rather than heterosexual stimuli… but only if they have strong impulses to approach homosexual related stimuli”. This sounds a bit strange and circular. In any case, a later study by the same group (also small-sample; n = 36) concluded that homophobia might relate to “concerns about sexuality in general and not homosexuality in particular”.
  
  (One weird thing about this study is that you can see on the university’s website that it was a Master’s thesis before it was published in a journal. That’s fine, but the name of the Master’s student who did the research is nowhere in the final published version - not even in an Acknowledgements section. So the student’s supervisors and other scientists have written up and published the student’s work without any reference to the student herself. I wonder what the story there is.)

Overall, it’s a mess: only the implicit-association studies have any replications (the plethysmograph, eye-tracking, and startle-response studies are all one-offs), and even then, the evidence is far from knockdown. Unless I’m missing something, the amount of evidence is way out of whack with the amount of attention and publicity that this idea regularly gets.

Dare not speak its name

The “homophobes are actually homosexual” idea is so compelling because it’s counterintuitive: it’s an initially-unexpected, clever-sounding point that surprises those who haven’t heard of it, and allows you to get one over on the homophobes. Turns out they’re actually total hypocrites like Ted Haggard! They’re repressing their own desires while lashing out at those who are open and honest and free about their sexuality!

But think about it. If it’s true, it implies that homophobia—or at least, a lot of homophobia—ultimately stems from gay people themselves. When you put it like that, it seems a lot less appealing.

But that’s a moral argument - how appealing an hypothesis sounds should have nothing to do with whether we accept it or not. We’re warranted to believe in an hypothesis if there’s good-quality evidence behind it. A look at the salad of studies above should tell you that these studies really don’t meet this criterion.

It shouldn’t be like this! I shouldn’t read through so many studies and find that they’re so low-quality that they can’t sway our views much in either direction; I shouldn’t find that over 600 studies have cited a paper (like the 1996 one) but not a single one of them has tried to replicate it. Of course, this is hardly a problem that’s unique to this question: it’s a basic failure of the scientific literature and it underlies the replication crisis. And don’t get me wrong: it’s fine for studies to be inconclusive because they genuinely find mixed results - it’s just less fine for them to be inconclusive because they were never designed well enough, or never had enough participants, to give them a hope of answering the question in the first place.

But until some better studies come along, all we can do is conclude that there’s not much behind the “homophobes are gay” idea. In my view, the evidence looks much more consistent with the idea that those who say they dislike or hate gay people actually do hate them, and it’s not some kind of repressed reaction to their own psychology. That’s not to say that some homophobes are mainly driven by this kind of self-hatred - but my bet is that this is such a small proportion of people that you’d rarely find them in a study (and especially not a tiny one with 40 people in it).

Sorry, Dr. Freud - sometimes a cigar is just a cigar, and sometimes a homophobe is just a homophobe.

Image credits: Getty