Ricochet is the best place on the internet to discuss the issues of the day, either through commenting on posts or writing your own for our active and dynamic community in a fully moderated environment. In addition, the Ricochet Audio Network offers over 50 original podcasts with new episodes released every day.
Homosexuality Facts #7: Small Genetic Correlation Established
I write regarding a major new paper released in Science on August 30, 2019, regarding the existence and extent of a hypothetical genetic contribution to the incidence of homosexuality. The paper is titled Large-scale GWAS reveals insights into the genetic architecture of same-sex sexual behavior, by Ganna et al. (paper here; supplement and figures here). As discussed in greater detail below, this paper establishes a small, but statistically significant, correlation between the genome and homosexuality.
“GWAS” means genome-wide association study, a type of analysis in which the entire genome, generally characterized by SNPs (single-nucleotide polymorphisms) is analyzed to identify correlations between the SNPs and the trait being studied. The data sources were extraordinarily large: a UK sample called the “UK Biobank” of about 500,000 individuals (about 400,000 of whom provided usable data regarding sexual orientation); an additional sample of about 75,000 individuals of European ancestry from 23andMe; and a few smaller samples.
The major conclusion was:
In aggregate, all tested genetic variants accounted for 8 to 25% of variation in male and female same-sex sexual behavior.
While results varied for different models and different data sets, they were all statistically significant at well above the 99% confidence level. This is a major contribution to our understanding.
Over about a week in March 2019, I published six Ricochet posts reviewing, summarizing, and critiquing the literature regarding two issues: (1) the genetic, biological, and environmental causes of homosexuality, and (2) the rate of reported change in homosexual orientation. My first such post (here, titled Homosexuality Facts #1: Baby You’re Not Born That Way) evaluated the evidence for a genetic cause for homosexuality, principally based on three major twin studies. My major conclusions were:
- The proper conclusion is that none of these [twin] studies demonstrates any genetic effect on homosexuality, while all of these studies demonstrate substantial non-shared environmental effect, generally in the 55-65% range.
- It is incorrect to draw the conclusion that there is no genetic component to homosexuality. This remains to be determined. But it is also incorrect to draw that conclusion that any such component has been established.
- A genetic basis for homosexuality has not been established, but there are indications of an effect in the 20-40% range, probably somewhat higher for males than females. For comparison purposes, the heritability of IQ appears to be in the range of 57-85%. The heritability of the Big Five personality traits appear to be in the 40-60% range. What we can say, with high confidence, is that homosexuality is not entirely an inborn trait. Baby, you’re not born that way. The environment, potentially including individual choice, is the principal cause.
This new paper by Ganna et al. provides evidence of a statistically significant, genome-wide contribution to the incidence of homosexuality (though perhaps indirectly, as discussed below). It finds no single “gay gene.” It confirms my tentative conclusion in March of a relatively small genetic contribution, indicating that it is even smaller than I expected — probably in the range of 15-20%, where I had estimated 20-40%. The effect seems about equal for males and females, though there is relatively low genetic correlation between the sexes (which means, if I understand the authors correctly, that different SNPs seem to affect males and females but that the overall estimated heritability is comparable).
There is a summary article about the Ganna et al. paper, at the Science website itself (here). It does report the main findings with reasonable accuracy, though with an absurdly misleading title and other misdirection. I will discuss the problems with reporting of these findings in Section III below.
I. The Major Results
From the Science press release:
- “When the researchers combined all the variants they measured across the entire genome, they estimate that genetics can explain between 8% and 25% of nonheterosexual behavior. The rest, they say, is explained by environmental influences, which could range from hormone exposure in the womb to social influences later in life.”
- “But the five DNA markers they found explained less than 1% of this behavior, as did another analysis that included more markers with smaller effects. As with other behavioral traits such as personality, there is no single ‘gay gene,’ says Broad team member Andrea Ganna. Instead, same-sex sexual behavior appears to be influenced by perhaps hundreds or thousands of genes, each with tiny effects.”
- “As the researchers had reported last year, they also found people with these markers were more open to new experiences, more likely to use marijuana, and at higher risk for mental illnesses such as depression. LGBTQ people might be more susceptible to mental illness because of societal pressures, the researchers note.”
The main result appears at the bottom half of Table S11 in the supplement (here), which reports the results of 27 separate analysis — 9 different models and data sets, with each analysis performed on males separately, females separately, and the combined sample. The combined sample yielded lower estimates (ranging from 6.8% to 13.3%), with the authors explaining that “[e]stimates combining males and females are biased downwards because of the relatively low genetic correlation between the sexes.” This was consistent with my prior literature reviews, which indicated that there are significant differences between male and females with respect to both the causes of, and the stability of, sexual orientation.
My impression is that the most sophisticated of the 9 different models were the final two, which appear to be the most complex multivariate analyses applied to the largest samples. These are what I will call the “Multivariate MTAG” and “BOLT-LMM” models (don’t worry about the details; it’s just useful to have a label for them). These results were around the middle of the 8-25% range, as follows:
- Male: 15.36% – Confidence Interval 11.17-19.55% – p-value >99.9% (Multivariate MTAG)
- Male: 16.55% – Confidence Interval 12.01-21.10% – p-value >99.9% (BOLT-LMM)
- Male: 14.36% (average of all 9 models)
- Female: 13.25% – Confidence Interval 8.61-17.89% – p-value >99.9% (Multivariate MTAG)
- Female: 20.66% – Confidence Interval 15.42-25.91% – p-value >99.9% (BOLT-LMM)
- Female: 14.09 (average of all 9 models)
In my view, this provides extremely convincing evidence of a genetic contribution to homosexual behavior, though this contribution may be direct or indirect. It also provides very strong evidence that the genetic contribution is probably around 15% for both males and females.
The second major result was that, while the researchers identified 5 significant SNPs associated with homosexuality, but these were very minor factors, each explaining less than 1% of the incidence of homosexuality. This seems, to me, to demonstrate convincingly that there is no single “gay gene,” nor is there a small number of “gay genes,” which is the conclusion of the authors. Interestingly, of the 5 significant SNPs, 2 were significant in both males and females, 2 were significant in males but not females, and 1 was significant in females but not males.
The third major result is strong correlation of homosexuality with certain personality traits and other behaviors, including: (1) openness to experience (a Big Five personality trait); (2) trait neuroticism (a Big Five trait); (3) smoking; (4) cannabis use; and (5) “risk behavior” (a risk taking variable from the UK Biobank data set). There were also significant correlation between homosexuality and: (1) major depressive disorder, (2) ADHD, and (3) schizophrenia (this was smaller than the other two). These correlations were all statistically significant.
II. My Speculations Regarding Indirect Causation
The correlation between certain personality traits and other behaviors, on the one hand, and sexual orientation, on the other, suggest that the genome-wide association with homosexuality demonstrated in Ganna et al. may be indirect, at least in part. Understanding this would require further study.
As an example, consider openness to experience, one of the Big Five personality traits. The Big Five Wikipedia page describes openness as:
Openness is a general appreciation for art, emotion, adventure, unusual ideas, imagination, curiosity, and variety of experience. People who are open to experience are intellectually curious, open to emotion, sensitive to beauty and willing to try new things. They tend to be, when compared to closed people, more creative and more aware of their feelings. They are also more likely to hold unconventional beliefs. High openness can be perceived as unpredictability or lack of focus, and more likely to engage in risky behaviour or drug taking. Moreover, individuals with high openness are said to pursue self-actualization specifically by seeking out intense, euphoric experiences.
This 2018 paper by Bogaert et al. (abstract only – I can’t access the full paper) states that: “All nonheterosexual groups, particularly bisexual people, averaged higher in openness to experience than did heterosexuals.” This is consistent with the result of the Ganna et al. paper discussed above.
This 2017 paper by Lo et al. indicates that there is a genome-wide correlation with openness, with a heritability estimate of around 10-12% (to see these figures, you need to download the supplemental table at the bottom of the paper).
It stands to reason that people who are more adventurous and curious, more willing to try new things, and more interested in a variety of experience, would be more likely to experiment with homosexuality. Some might find that they like it (or even become hooked). If there is a genome-wide contribution to openness, and if openness makes one more likely to be homosexual — both of which appear to be true — then part of the genome-wide correlation with homosexuality could be indirect.
There may be other avenues of indirect causation, and it is possible that genetic factors could be triggered by environmental factors to result in a same-sex orientation. It’s complicated, to say the least.
III. Concern About Media Reporting
Reading the Ganna et al. article makes it clear, frankly, that the authors were terrified of running afoul of the SJW mobs. The paper actually concludes with this explanation:
Science Communication Strategy
To communicate the results of the study to the broader audience, we engaged with different LGBTQIA+ (lesbian, gay, bisexual, transgender, queer, intersex, asexual, and other+) and science communication organizations and created multimedia materials for a lay audience.
As noted above, the major result of this study was an estimate that the genome-wide genetic correlation with homosexuality is in the 8-25% range, probably close to 15%, for both males and females. This was acknowledged in the Science press release (here, again) — but guess what the headline reads?
Genetics may explain up to 25% of same-sex behavior, giant analysis reveals.
Yeah, and it may explain as little as 8%. And even that may be indirect.
There were also two special, gray-highlighted boxes in the paper (on pages 2 and 6), which included the following text:
We acknowledge that the grouping phrase “nonheterosexuals” has the potential to present messages of othering (that is, undesirable marginalization of another person or group on the basis of their sexual expression) — by defining an “outgroup” in reference to an “ingroup” and implying that “nonheterosexual behavior” may have a negative connotation, whereas “heterosexual behavior” may have a positive one. We wish to make clear that our choice of language is not meant to forward messages of othering on the basis of sexual behavior.. . .As is common in genetic analyses, we dropped individuals from our study whose biological sex and self-identified sex/gender did not match. This is an important limitation of our analyses because the analyses do not include transgender persons, intersex persons, and other important persons and groups within the queer community.We hope that this limitation will be addressed in future work.. . .The topic explored in this study is complex and intersects with sexuality, identity, and attraction and potentially has civil and political implications for sexual minority groups. Therefore, we haveEngaged with science communication teams,
Engaged with LGBTQIA+ advocacy groups nationally and within our local institutions, and
Tried to make clear the many limitations and nuances of our study and our phenotypes.We wish to make it clear that our results overwhelmingly point toward the richness and diversity of human sexuality. Our results do not point toward a role for discrimination on the basis of sexual identity or attraction, nor do our results make any conclusive statements about the degree to which “nature” and “nurture” influence sexual preference.
These disclaimers raise concern about either the ideological predisposition of the authors (which I do not know), or their fear of reprisal, or both. However, they are to be commended for going forward with this research and publication.
Published in General
Jerry,
Very interesting. Thank you.
H.
Most quotable line!
But more seriously – very interesting write up.
Agreed. A reasonable, fair and fairly intricate explanation of the article. I’m not sure why either side of this issue would push the idea of one simple solution, a “gay gene”. It never took anything as decisive as that.
Jerry’s posts, though they support ideas unwelcome in many parts of the media, never fell for that fallacy, and neither did the writer who is his counterpart, the opposition in this discussion, gay writer Andrew Sullivan, whose position is surprisingly similar to his: there’s a genetic component but there’s more than that. For that matter, on the other hand Jerry said, I think correctly, that it’s pretty clear that many or most homosexuals (I’d say nearly all of them, but why quibble about numbers?) had no choice and are no more conscious of a moment when they “decided to be gay” than you or I decided to be straight.
Another point of agreement (I’m tempted to write ‘broad’ agreement): women are different. They seem more flexible and changeable, more prone to experimentation, though only half as many of them (compared to men) remain gay.
It depends upon how you define many, most, or nearly all, especially if you consider bi-sexual behavior. I learned this weekend that a cousin who married and fathered a daughter is now exclusively gay. Another male cousin never married, came out in the 1960’s with a boyfriend, which was rare at that time. So my experience tends to back Andrew Sullivan and Jerry.
Thanks for reporting on this. I’ve been swamped lately and missed any related headlines.
I’ve always thought proponents of a “gay gene” were setting themselves up. If a genetic predisposition to homosexuality is found, gayness could be labeled a genetic abnormality. Furthermore, genetic testing in utero would lead to abortion based on projected sexual preference.
Which reminds me:
Q: How do you tell the sex of a chromosome?
A: Take down its genes.
It is my highly non-professional opinion, based on decades of entirely amateurish research and armchair speculation, that male and female homosexuality and bisexuality are vastly different things. There are strictly homosexual men and strictly homosexual women but, when it comes to bisexuality, men approach it from the homosexual side, whereas women approach it from the heterosexual side. That is, most women — I like to say 85%, because it annoys my female friends — are to some degree bisexual. Whereas, among men, only gays are bisexual.
Women are great.
[ You know you haven’t contributed anything worthwhile to this thoughtful and substantive thread, right? Yes, I know. ]
Henry, the data and studies surveyed in my 6 prior posts (in March) did indicate that there were significant differences between males and females on issues relating to sexual orientation.
Two of the papers that I surveyed included empirical evidence on the incidence of different sexual orientations. They were Mock 2011 (here) and Savin-Williams 2012 (here). Both were based on large longitudinal studies.
Mock 2011 involved over 2,500 American adults, with initial ages between 24 and 74, with an average age of 47 at the first report (called a “wave”) and 56 at the second wave. The incidence of nonheterosexuality was quite similar between the sexes (this is from Table 1, in the first wave):
The paper didn’t provide a summary at the second wave, but it can be calculated from Table 2, and the results are similar (my calculations):
These are very small differences between the sexes, inconsistent with your anecdotal suggestion.
Savin-Williams 2012 involved over 12,000 young adults (previously sampled as teens), reporting changes between wave 3 (average age 22) and wave 4 (average age 28). It added categories for “mostly heterosexual” and “mostly homosexual,” in addition to the standard three categories used in Mock 2011. The results (Table 1) are as follows (with the addition my combination of the “100%” and mostly” categories into additional “combined heterosexual” and “combined homosexual” categories:
This provides support for your impression that women are less likely to be completely heterosexual, though the overall rates of combined heterosexuality remain in the 96-97% range. Females are 3-4 times more likely to be “mostly” heterosexual.
Wave 3 was conducted in 2001-02 and Wave 4 was conducted in 2007-08. Note the quite large shift away from the 100% heterosexual category, concentrated among females — the decline was from 85.6% in Wave 3 to 80.2% in Wave 4.
Recent data (from 2017) suggests that this trend has accelerated among the young, though the data that I’ve seen on this issue did not differentiate between males and females.
Jerry, thanks again for serious, fact-filled comments on an interesting topic.
As regards my “anecdotal suggestion,” as you put it, two comments in its defense: (1) I’ll rely heavily on the “to some degree” in my observation that most women are “to some degree bisexual,” and (2) most of them don’t even know it, since girls are so completely unlike guys in how they interact with the same sex.
So, among my female acquaintances, 85% bisexual seems a lot closer to the mark than the roughly 18%* in your figures.
Of course, I could just be lucky that way.
* (I’m lumping mostly-this and mostly-that in with bisexual because, come on, what else could it mean?)
Henry — I meant no criticism. I happened to have empirical evidence on the question. Your general impression was supported by the data, though at a lower rate than you suspected, and though I would not have phrased it as you did.
I would say that while males and females in our country, historically, have been overwhelmingly heterosexual (about 96-97% or more), there is more of a tendency toward same-sex attraction among women than among men. Most women having some degree of same-sex attraction nevertheless self-identify as “mostly heterosexual.”
This is not surprising, frankly, as they are so much better looking than we are.
The above is based on historical data. The rates of LGBT identification are increasing substantially among the young. I will address this in a separate post (in progress).
Actually, I appreciate your willingness to put up with my tongue-in-cheek observations. ;)
Because women in our culture are much more free to express same-sex physical closeness than are men, the social and identity cost of being to some degree bisexual is, I think, much lower for them. Also, as you point out, they’re just intrinsically more attractive, so they have a greater incentive. Given all that, it would be weird if women weren’t more inclined toward same-sex experimentation.
It is surely akin to other neurological variations? For instance, bipolar disorder is considered highly heritable, and yet the disorder can be very mild (in which case, the “sufferer” is moody and generally quite talented) or completely debilitating.
I had a wonderful conversation once with Kay Jamieson, author of An Unquiet Mind. She is a psychiatrist, researcher and psychotherapist who, herself, is afflicted with bipolar disorder and came close to dying of it. She said there is a distinct link between creativity and bipolar disorder (the usual examples being Vincent Van Gogh and Lord Byron) but the “winners” in the equation are the siblings of the person with full-blown bipolar. It would seem that they have enough of the nature+nurture to be artists, but not so much that it is actually disabling.
This brings up an interesting question with regard to the Darwinian “advantage” of a genetic predisposition for homosexuality. If, say, openmindedness is adaptive and associated, perhaps intimately, with a stronger tendency toward homosexuality, mightn’t siblings of the homosexual person get the advantage without the homosexuality? Given that homosexuality isn’t actually disabling—not only can homosexuals be happy and even live (if not entirely comfortably) as heterosexual men, with wives and offspring—the adaptive elements could easily be more than worth the evolutionary cost of a somewhat reduced potential for reproductive success.
Somewhat, I say, because I know a number of gay men who fathered two, three, four and even five kids before realizing that heterosexuality just wasn’t doing it for them anymore.
By the way—men are wonderful, and much, much better looking than women.
About the last comment, I’ll only point out that you haven’t met me yet. :)
Analytically, the discussion of Darwinian advantage or disadvantage shouldn’t be overstated. The genetic contribution to nonheterosexuality is very small. Environmental influences are much more significant. Personally, I believe that individual choice is involved.
I don’t mean that people consciously choose to be nonheterosexual. In fact, I was tickled that Zafar (in cmt #2 above) seemed to like my line: “Some might find that they like it (or even become hooked).” I suspect that our friend Zafar’s idea was like the old Chinese restaurant commercial in Tucson — “I know if you try it, you will like it!” My idea was more like — don’t try crack or meth, not even once, you might get hooked.
There are some darker possibilities, and I’m working on a follow-up post on the issue. There are indications, for example, that homosexuality is associated with childhood sexual abuse specifically, and with a risky childhood home environment more generally. I don’t expect these to provide a complete explanation, but my initial impression is that the contribution of these factors may be approximately equal to the genetic contribution (and again, the genetic contribution may be indirect). I haven’t done the full calculations. More to come.
A variation of a 1972 commercial:
This is also true of bipolar disorder, however. Certainly the severity of the illness can have a lot to do with the environment. (Sexual abuse, for example, doesn’t help).
I’d be interested in knowing how many other variants in human life have a similar ratio of nature to nurture?
I loved that gay sex is like crack.
My opinion: not to heterosexuals.
Situational homosexual behavior (prisons, the army) suggests that it might be the other white meat on occasion for some.
This could be an indicator. ??
The early political gambit “born that way,” was a counter to the then politically potent “Moral Majority” or “religious right.” Within the grossly ill-taught (lousy/lazy shepherds (preacher/pastor/priest)) American “religious right,” “God made me this way” might reasonably be expected to have persuasive power. To reject this claim would be for the “believer” to condemn his/her self in their own biblical failings.
We are well past that. Yet, a demonstration that minority sexual identities are NOT closely tied to a discreet, targetable gene/DNA strand is a defense against parents, patients, or the state targeting the genetic code for reprogramming. See Downs Syndrome for a horrific counter example.
Well, sexual gratification is always popular with people, the question is how to get it in ways that don’t lead to negative social consequences.
Of course the issues surrounding homosexuality in modern society aren’t really about sex or how one achieves sexual gratification but about the the legitimacy of homosexual romance. After all one doesn’t need “marriage” to have sex (straight or gay), but if one wants their romantic passions socially validated then officially sanctioned marriage is the ultimate expression and endorsement our society can provide in that respect.
Ultimately though none of this research really proves that people can choose their sexual preferences, as environment is no more under our control than our genetics (for the most part). To me it seemed obvious that what ever the biological mechanisms underlying human sexual variations are they would end up being developmental (as sexual preference arises late in human development with puberty). Also as human sexuality is a very complex trait the possibilities that its various expressions could be tracked down to one or even a few genes seemed so astronomically small as to be practically zero.
As a good rule of thumb one should expect the general trait (desire of sex in this case) to have a stronger genetic component than any specific variation (desire for sex with people of the same sex).
I admit I haven’t read through all the great information you linked, but I’m not sure I understand where some of your conclusions are coming from.
For example, the Swedish study you mentioned did not find “no statistically-significant genetic component” for male homosexuality. It found a heritable (I assume this can be read as genetic) component of between .34 and .39 — a confidence interval well above the threshold for statistical significance.
Also, I believe the GWAS you discuss here finds 8-25% correlation from among the loci investigated. In other words (using midpoints) if the true genetic component is 0.36 and the loci investigated explain 0.16, then other genes presumably explain the remaining 0.20. It sounds like your comments are saying something different.
Again, I apologize if I’ve missed “a few, uh, provisos, ah, a couple of quid pro quos” or a caveat. (I can only ignore the day job for so long!)
Given the pretty consistent findings of heritability of behavior and personality, shouldn’t our prior be 50%? At this point, I’d be surprised if genes explain something much different from that — assuming we all agree on definitions and whatnot of course.
Thank you for a very thorough post.
Where did this quote (my bold) come from? It is nowhere in this post.
Where does your 0.36 come from? The 8-25% is the range for the complete genetic component.
You seem to have injected information not found in the post. Please clarify.
Hey, Phil.
I was attempting to refer to your first bullet above, which actually says, “The proper conclusion is that none of these [twin] studies demonstrates any genetic effect on homosexuality, while all of these studies demonstrate substantial non-shared environmental effect, generally in the 55-65% range.” I suspect the quote I mangled (or something like it) comes from one of your other six posts. I do apologize for being sloppy, there. My bad.
As I stated, the 0.36 is roughly the midpoint of the range for heritability in the Swedish study you linked in post #1. The abstract says, “…in men, genetic effects explained .34 to .39 of the variance….” I was making the point that twin studies capture total heritability (presumably genetic), whereas GWAS capture the effect of specific genes. In other words, the 8-25% range from the GWAS is only that part of the genetic component that can be trades to the loci they considered.
I’m sure my data are out of date, but there is a similar thing going on with IQ. Twin studies suggest high heritability, as you noted. GWAS have identified genes responsible for some small fraction of it — something like 7%, as I recall.
Darin, good questions in #19 and #21.
(1) About the Swedish twin study
You are incorrect about the Swedish twin study (here). You quote from the abstract, but you assume that the statement in the abstract — “in men, genetic effects explained .34 to .39 of the variance” — is reporting a confidence interval. It is not.
Table 1 reports the analytical results for two different dependent variables: “any lifetime same-sex partner” and “total number of same-sex partners.” The point-estimates were 0.39 and 0.34 respectively. Neither was statistically significant.
This is not the only place, in the technical literature on nonheterosexuality, in which I have observed an abstract reporting a result that is not statistically significant, or subsequent papers relying on results that were not statistically significant. I actually find this to be academic malpractice, at least arguably.
(2) About GWAS
I think that you are misinterpreting the result of this GWAS study. This is not a matter of an analysis using “the loci investigated.” GWAS stands for “genome-wide association study, which involves all loci (i.e. all SNPs).
Here is a JAMA article titled “How to Interpret a Genome-wide Association Study,” which explains:
(3) About a priori assumptions
You write: “Given the pretty consistent findings of heritability of behavior and personality, shouldn’t our prior be 50%?” I say no, and I think that this represents an unscientific viewpoint.
Colloquially, I would agree that, before investigation, we should presume that any behavior or characteristic is partly nature and partly nurture, though even this traditional dichotomy is deeply flawed (as it is based on a purely materialist presupposition and implicitly denies both the existence of individual free will and the possibility of a spiritual or metaphysical explanation). Even given these implicit flaws in the model, we should not then start with a 50-50 assumption, as this is purely an empirical question.
If one is proposing a nonzero role for nurture (i.e. environmental causes), then scientific and statistical methodology requires that we assume no relationship as the null hypothesis, compare this to the observations, and evaluate whether the observations demonstrate a statistically significant relationship. If not, then we conclude that the relationship has not been established.
The same holds if one is proposing a nonzero role for nature (i.e. a genetic cause). Thus, we should assume that there is no genetic component to nonheterosexuality, until evidence demonstrates a contrary result. The Ganna paper demonstrates such a statistically significant result, though it is small and may be indirect, as explained at length in the OP.
Thanks for the feedback.
(1) Yes, I see what you mean, and I agree. (In my defense, the abstract is amazingly poorly written!) I understand your “not proven” comment.
Table 2 refers to a bunch of similar studies, all of which show non-statistically-significant by positive genetic correlation. I wonder what a meta-analysis would show. I don’t know of one.
(2) GWAS. Yes, but there’s always a sample-size problem. What we are finding in the case of other traits is that there are many genes each with a small contribution. You wouldn’t expect a study of a few thousand people to uncover more than a tiny fraction of the responsible genes. That’s what I was getting at when I mentioned the example of IQ.
It’s possible there’s also a definitional issue compounding the problem. A history of same-sex partners may be a blunt measure if sexuality exists on some kind of continuum.
Thank you for linking the JAMA paper. I will definitely read it.
(3) I wasn’t suggesting that we should start with a 50% prior because there are two possible causes, but because, very roughly, 50% genes and 50% non-shared environment seems to be what we find for all sorts of traits of personality and behavior (here). I agree that what our prior should be is an empirical question, but empirical based on what? Should the fact that many other traits have turned out to be roughly 50/50 influence us? I say yes.
Regardless, I would not agree that “not proven” by statistical convention does not imply we have no insight into the genetic component of homosexuality. If the Swedish study were all we had to go on, we’d presumably have a posterior estimate of 0.34 with wide uncertainty. But we’d believe that zero is less likely than 0.34. Assuming there’s no genetic element, as you suggest, simply because we haven’t yet seen the magic 0.05 p-value means ignoring information that we do in fact have.
If there are spiritual or metaphysical causes, or free will, they are presumably lumped into the non-shared environment. Maybe the terminology could be better, but I don’t see a technical problem. Do you?
Thanks for a thoughtful discussion.
I think that we need to be a bit more careful about terms like “exclusively gay”. The Kinsey Scale (https://en.wikipedia.org/wiki/Kinsey_scale) runs from 0 to 6, and outside of the two ends of the continuum, behavior/conduct is essentially a decision of (depending on what is the perceived as the desired behavior) varying degrees of difficulty.
Darin, thanks for the excellent feedback. I have comments about a few of your points.
I don’t know of one, either. There is a problem with meta-analyses, which I have not previously considered, actually raised by one of the principal authors of the Ganna paper (Brendan Zeitsch, who is listed last, but is the corresponding author and is identified elsewhere as one of the lead researchers).
Dr. Zeitsch points out that: “Given widespread bias toward publishing positive results, including only published findings raises the possibility of a meta-analytic result biased in the direction of the hypothesized effect.”
The Ganna paper uses a very large database — close to 500,000 people.
I agree that individual choice (free will) would be lumped into the non-shared environment. Metaphysical causes could effect either the non-shared environment estimate, or the genetic estimate, or perhaps even the shared environment.
Huh. Not my quote at all. How did you do that?
Ok, but not my post either.
Jerry answered these…
My technical commenting skills have been exposed today. I have mis-quoted and conflated conversations with at least two people. I have no defense, only an apology.
Hopefully the questions I’ve tried to raise can be taken in good faith despite my own shortcomings.
Thanks, Jerry. A few thoughts in response.
Meta-analyses, limitations and biases of
Absolutely agree. Although I’m not sure what “positive result” means in this case. You seem to suggest that positive would mean some genetic component other than zero, but I’m not so sure. What is the hypothesized effect, in this case?
Ganna study, sample size
Half a millions sounds like a lot until you consider how many possible loci there are. And how few homosexuals there are. And the general result that there’s not “a gene” for most anything, but many genes with small effects. Again, I’m thinking of the IQ example, where heritability (presumably genetic) is over 50% but loci accounting for only a small fraction have been found, and with far larger samples. My assumption is that we’ll continue to identify genes until the GWAS roughly match the twin studies for whatever trait. Do you disagree?
Metaphysical causes
You say, “Metaphysical causes could effect either the non-shared environment estimate, or the genetic estimate, or perhaps even the shared environment.” Hm. I’d be interested to hear more about this. Maybe you could suggest a metaphysical cause and how it could fall into one of those buckets. This bumps up against my mental limitations, I admit.
Darin, don’t be discouraged at all. These are very, very complicated issues. I’m doing my best to interpret highly technical information, and I am not a technical expert in these areas.
Ask away, and I’ll do my best to respond.
(I’m a bit of an “expert,” depending on what you mean by “expert.” I do have an educational background in probability and statistics, quite a long time ago, and I have professional expertise as a lawyer in marshaling, understanding, and criticizing technical experts and their reports. I think that this makes me an “expert” as generally defined by the law, which is someone having greater knowledge than the average juror. Obviously, I know far less than the authors of the various studies that I reference — though, if it came up, I do think that I could quite effectively cross-examine them, given a bit of preparation time.)
In general, a “positive result” would mean finding a correlation. In the present case, for example, it would mean finding a correlation between nonheterosexuality and the “genetic” component in a twin study.
I don’t think that it’s a matter of identifying genes. I think that all of the genes are included in the GWAS analysis.
Which IQ studies are you referencing? I found this one, published in 2017, using the same “UK Biobank” data of about 500,000 individuals that was used in the Ganna paper. It addressed many characteristics, including IQ. Table 1 shows a heritability estimate of 0.233 for fluid intelligence, with a standard error of 0.011, with N=34,491. This implies a 95% confidence interval of about 0.211-0.255. Do you have a larger one in mind?
Returning to the main point, I think that GWAS gives us the actual genetic correlation, while other methods of assessing “heritability” are indirect, and are essentially the result of correlation between the IQ scores of relatives (particularly parent/child). Prior to GWAS, the heritability estimates that I’ve seen for IQ ranged from 0.4 to 0.8.
The disparity between GWAS results and other heritability results may suggest a problem in the “ACE” model generally used in twin studies. The ACE model assumes an additive effect, so the different results obtained by GWAS studies may indicate a problem with the additive model. One possibility, I think, would be non-additive relationships between the genetic, shared environmental, and nonshared environmental factors (which are represented by the ACE letters).
Regarding your statement regarding sample size and the relatively low incidence of nonheterosexuality: I think that this is addressed by the confidence interval calculations. You cannot evaluate, in a vacuum, whether a sample is of sufficient size. But a large sample will give relatively tight confidence intervals, and this is what is shown in the Ganna study.