Bryan Caplan  

Nurture and Orientation Reconsidered

Health Care Summit "Mid-Mortem... Grading the Health Care Summit...
A couple days ago, I mentioned that gays' adopted siblings are gay at six times the normal rate - and called this finding a "smoking gun" proving that family environment has a modest effect on sexual orientation.  This finding has bugged me for a while, so I finally turned to Michael Bailey, one of the original researchers.  Is it possible, I asked, that this "effect" merely reflects response bias?  Maybe gays were more likely to respond if they had gay siblings - biological or adopted.

Bailey's answer: Not only is response bias possible; it's his "best guess" for the finding.  He then pointed out something that I hadn't previously noticed (or perhaps had forgotten): The original 1991 study explicitly noted the possibility of "ascertainment bias":
There is an indirect indicator of proband cooperation: whether the proband consented to have his relative contacted... [P]robands with heterosexual adoptive brothers were significantly less likely to consent than probands with heterosexual twins; cooperation did not differ notably if relatives were homosexual.  If similar factors affect probands' decisions regarding (1) allowing their relatives to be contacted and (2) their initial participation in the study, then our results would suggest that the proportion of heterosexual relatives was underestimated in the adoptive brothers, compared with twin subsamples.
Bottom line: I'm rewriting the section on family effects on sexual orientation.  The adoption evidence raises the possibility of a family effect, but contrary to what I originally wrote, there's no "smoking gun."

Comments and Sharing

COMMENTS (5 to date)
dWj writes:

I wonder whether this would be less of a problem if the study were replicated today, especially if it focussed on younger people, who are more likely to be in social circles where homosexuality's stigma is much smaller than would typically be the case 20 years ago or even somewhat today among older people.

Daniel Kuehn writes:

That's very interesting.

I think there also might be a difference between the labels we use to describe reality and reality itself. My understanding is sexuality exists along a spectrum, right? Not only is there the bisexual issue, but there are degrees of bisexual preferences that range from being much closer to heterosexual or much closer to homosexual.

Take two adopted siblings of an identical type of bisexuality. One that lives with a gay sibling may be much more willing to label their orientation "gay", whereas the one without a gay sibling may label their orientation as "straight, with some confusing urges every once in awhile". So it's not even a response bias per se. Both are perfectly honest with themselves - it's not misreporting. It's just a different understanding and labeling of the same phenomenon. The very term "homosexual" or "heterosexual" is actually only a rough proxy for what we're really trying to get at, which is sexual orientation.

Loof writes:

Still see a smoking gun dialectically inherent, Bryan. Even with response bias, see negative nurturing playing a role, possibly a starring role in sexual orientation: not so much parental influence; primarily sibling influence.

Thesis: “Parents have a small effect on sexual orientation.” Passé conclusion: “caused by overprotective mothers and distant fathers.” i.e. negative nurturing. Prevailing conclusion: “sexual orientation is a preference inherent in our genes.”

Anti-thesis: “Yet genes are far from the whole story - if you're gay, your identical twin is usually still straight. Furthermore, family environment clearly affects sexual orientation. The smoking gun: Adoptive brothers of gay men and adoptive sisters of gay women are about six times more likely to be gay that you would expect from chance”

Synthesis: Parents have a small effect on sexual orientation. Sexual orientation is a preference inherent in our genes. Since same gene and adoptive siblings are usually straight they could be the main influence in gayness. Further, enhanced causation could be overprotective siblings with distant parents.

Heredity may still be a star of the story of sexual orientation; negative nurturing appears as a co-star, if not the star. Family environment may still have a leading role: directly with overprotective siblings and, indirectly, with distant parents.

Joey Donuts writes:

I see a pretty big swing in conclusions, "Smoking gun" to "possibility." The comments by the authors of ascertainment bias is, on close examination, a pretty weak statement.

First I don't know what "were significantly less likely to consent" means. Does it mean that the likelihood of consent was .2 for the adoptive siblings group and .8 for the twins? Or does it mean likelihood of consent of 0.75 for the sibling consenters is statiscally different than 0.9 for the twins. Statistical significance does not necessarily imply large bias only that there is a strong probability of bias.

Furthermore, to state that "IF (emphasis mine) similar factors affect probands' decisions regarding (1) allowing their relatives to be contacted and (2) their initial participation in the study." Then there is a possible undercount of heterosexual siblings. That's a pretty big if and some tests of the proposition can be made perhaps.

What percentage of Homosexuals asked to participate had hetero sexual adopted siblings? Perhaps that question was never asked before asking the person to participate. But, it could be asked now. Once we have an estimate of the percentage of homosexulas with adopted siblings. We can compare it to the percentage of homosexuals with adopted siblings in the study. Maybe there is another estimate of the percentage that exists in another study.

In any even participation rates would have to be 1 sixth of the possible adoptive sibling homosexual group in order for the undercount to off set all of the difference. If the participation rate was only half, the study would still have shown 3 times as many homosexual adopted siblings of homosexuals in the study. Would this be a smoking gun? Where on the continuum between smoking gun and possibility does the evidence fall?

By claiming the study shows a "possibility" of family influence, implies, at least to me, that the study only raises a question. A little further digging may show that the study provides much stronger evidence of a family influence.

Floccina writes:

How about we start using congenital rather than genetic to discus this. It seems a better fit for this subject because there is a theory that it happens in the womb. (which by the way could imply a drug could help to avoid it).

Pronunciation: \kən-ˈje-nə-təl, kän-\
Function: adjective
Etymology: Latin congenitus, from com- + genitus, past participle of gignere to bring forth — more at kin
Date: 1796

1 a : existing at or dating from birth b : constituting an essential characteristic : inherent c : acquired during development in the uterus and not through heredity
2 : being such by nature

Comments for this entry have been closed
Return to top