The Heart Has Its Reasons That Reason Knows Not Of

I.

Psychoanalysts argue that sons are attracted to women who look like their mothers, because they imprint on their mothers and use them as a schema for their ideal woman.

(and probably something similar for daughters and their fathers, though the psychoanalysts don’t usually get around to talking about that)

I’ve counterargued that sons get half their genes from their fathers, who apparently were attracted to women who looked like the sons’ mothers. Since study after study shows genes having unexpectedly big effects on behavior, and early childhood experiences with parents having unexpectedly small effects, maybe this would be a better explanation for the effect (if it even exists).

But the research doesn’t really bear me out.

Goose-ologist Konrad Lorenz raised goslings from birth. When the goslings grew up, he found they tried to mate with humans, especially “Caucasian men with white beards”. He concluded that they imprinted on their adoptive parent (him) and learned to prefer mates who looked like that parent.

More formally, Bischof et al got some male zebra finches and arrange to have them raised by a closely-related species, Bengalese finches. Then they put them in cages with both female zebra finches and female Bengalese finches and observed which females the birds tried to court. The results were pretty striking; they overwhelmingly went for the Bengalese finches who looked like their mothers, not the zebra finches who were genetically more suitable. Spence & Smith replicated this finding with zebra fish raised by differently-colored zebra fish.

So the research shows conclusively that sexual selection is based on learned imprinting, at least in animals whose names start with the string “zebra fi”. What about humans?

I can’t find the study itself, but multiple reviews cite Jedlicka 1984, who looked at children of mixed-race couples (white and native Hawaiian). They found that both men and women were more likely to marry someone of the race of their opposite-sex parent than of their same-sex parent (eg if you’re a woman with a Hawaiian mother and white father, you’re more likely to marry a white person). This is consistent with some kind of social imprinting where your opposite-sex parent serves as a template for future romantic interest. It’s not consistent with a simple genetic theory where you just get both parents’ genes. It might be consistent with a more complicated genetic theory where mate preferences are on a sex-appropriate chromosome or get chromosomally imprinted such that you only care about your father’s preferences for women and your mother’s preferences for men, but this is hard and I haven’t seen any analysis of whether it’s evolutionarily worth it.

Enquist, Aronsson, Ghirlanda, Jansson, and Jannini (2010) starts its Methods section with “We obtained data through newsgroups alt.sex.fetish and alt.sex.fetish.breastmilk”, so you know it’s gonna be interesting. They test another feature of men sexually imprinting on their mothers: suppose you’re a man with a sibling a few years younger than yourself. That means your mother was pregnant or lactating during the supposed critical sexual imprinting window. So if men with younger siblings are more likely to have pregnancy and lactation fetishes, that suggests that sexual imprinting on mothers is really a thing. This is indeed what they found: when a person with a pregnancy or lactation fetish only had one sibling, there was a 66% chance (compared to expected 50%) that their sibling would be younger than they, p < 0.0001 in their sample of 560 such people. This was true if and only if they were between age 1.5 and 5 during their sibling’s birth, hinting at the span of the imprinting window. Of course, this is still a really poor predictor: 33% of such people got the fetish without any younger siblings, and of course most people with younger siblings don’t end up with the fetish at all. But it does look like something is going on.

(I wonder what’s up with adult baby fetishes; if you could find a similar pattern among them that would suggest they’re imprinting on the baby sibling instead of the mother, which would be fascinating. Maybe I should try to survey the appropriate subreddit.)

Hefferman and Fraley did a very similar study. They find that people born to older parents, when compared against people born to younger parents, find older faces more attractive. This was true even after controlling for the age of the participants themselves. The effect size was small but pretty consistent across different groups and measurements. I am not quite as happy about the quality this study as in the ones above, but nothing raises huge red flags.

So there’s a lot of circumstantial evidence suggesting some kind of imprinting process is going on. But what about the original question – do people choose mates who look like their opposite-sex parent? And how genetic versus environmental is it?

Bereczkei, Gyuris, and Weisfeld addresses this question directly. They get a sample of grown-up adopted daughters and their adoptive fathers. They show subjects pictures of the adoptive fathers taken when their daughters were 2-8 years old, and then photos of four similar-aged men, one of whom is the daughter’s current husband. The subjects’ task is to guess which of the photos is the husband based on which of them looks most like the adoptive father. And subjects are in fact able to do this better than chance; they pick the correct husband 37% of the time compared to 21% of the time for (each of the three) incorrect husbands, which in their sample of 242 subjects is a p-less-than 0.001 result.

Somebody else check if I’m wrong here, but I’m a little concerned about the way they calculated p-values in this study. Suppose that their 242 subjects are always accurate in identifying the husband who looks most like the father; no randomness or noise here. But now the chance for randomness and noise comes in how many of the women might have picked husbands who look like their fathers by pure coincidence. It looks like there were only 26 women used in the study, so there’s a lot more opportunity for those twenty-six people to coincidentally choose husbands who happened to look like Dad than for the 242 subjects to coincidentally choose the correct husband even if they all looked alike. Yet as far as I can tell, the p-value calculation was only done on the latter possibility.

(an analogy: suppose that I want to prove I have ESP. I predict that a coin will land heads, then flip the coin. It does indeed land heads. Then I survey one thousand observers to ask them to judge whether the coin landed in accordance with my psychic prediction. All 1,000 of them say yes, a p = 0.000001 result that could not possibly be by coincidence. This does not prove at the p = 0.000001 level that I have ESP!)

Rantala and Marcinkowska do a really excellent review of this field that found a lot of the same papers I did. They acknowledge the Bereczkei/Gyuris/Weisfeld study as very interesting, but note that the first author “replicated” these results in a 2009 study using more scientific facial measurements. That study was found to be extremely flawed (found by Rantala himself, in fact) and was retracted under a cloud of accusations of misconduct and/or inexcusable error. It’s still not clear exactly what happened there, and nobody has formally accused the original father/adopted-daughter study of anything, but it’s kind of awkward to hang your theory on a study by somebody accused of scientific misconduct for faking a very similar study that supported the same hypothesis.

(side note: I wish that people would explain retractions better. Right now I don’t feel like I have a good understanding of exactly how the 2009 study got retracted, so I haven’t learned anything that will help me spot similar problems in the future)

So in summary: there’s a lot of good evidence that animals learn mate preferences by imprinting on their parents. There’s some circumstantial evidence that humans might do this too, across traits as diverse as race, age, and lactation status. There is, however, as yet no smoking gun.

II.

Wait, what about twin studies? Those are usually pretty great! Can we just do a twin study and sort this out once and for all?

Zietsch et al look at “a large community-based sample of twins and their partners and parents (N > 20,000 individuals) to test for genetic and family environmental influences on mate choice”. They find:

…near-zero genetic influences on male and female mate choice over all traits and no significant genetic influences on mate choice for any specific trait (!!!!!). A significant family environmental influence was found for the age and income of females’ mate choices, possibly reflecting parental influence over mating decisions.

This might be the first twin study I’ve ever seen which unambiguously breaks Turkheimer’s First Law Of Behavioral Genetics (every trait is somewhat heritable). Such a striking finding should increase our confidence in all of the above experiments a lot. So okay, I guess this issue is solved, it’s definitely just sexual imprinting on the opposite-sex parent, thank goodness, for once we have a perfectly clear noncontradictory result and we can all just go home and –

We also tested for evidence of sexual imprinting, where individuals acquire mate-choice criteria during development by using their opposite-sex parent as the template of a desirable mate; there was no such effect for any trait.

Sigh.

Okay, fine, let’s look at this a little more closely. They were analyzing a bunch of data from a big Australian survey of twins. This included the twins, the twins’ family, and the twins’ mates “height, BMI, age, education, income, personality, social attitudes, and religiosity”, which were the dimensions along which they tried to predict mate choice. They figured out whether imprinting was involved by checking whether twins’ mates were more like their opposite-sex parent than like other members of their family (their same-sex parent, their other twin). There was no sign of this being true, not even a nonsignificant trend.

If we want to play the dangerous game of trying to explain differences between contradictory studies instead of just dismissing everything as noise, I might argue that this looked at some pretty different variables compared to the last set. Instead of looking at facial similarities, it’s looking at things like social attitudes and religiosity; young children trying to imprint on their mother’s image can maybe be forgiven for not knowing her opinion about Asian immigrants (one of the “social attitudes” questions they asked).

The authors took a different tactic and pointed out that most of us don’t marry the first person we have a crush on, or even the second or third. Sometimes we marry the tenth person we really like, sometimes we settle for people we only like a little, and sometimes we get drunk, have sex in a cheap motel, and have the person’s parent threaten us with a shotgun unless we go to the chapel right now. So maybe the person we end up marrying isn’t a good proxy for our mate preferences per se.

(this is starting to get kind of depressing)

They investigate this hypothesis in a followup study where they directly ask twins about their preferences for an ideal mate. Here they’re able to get some more immediately visual data – preferences like tall/short, long-hair/short-hair, beards/clean-shaven, even big-breasts/small-breasts. They find…well, some things come out heritable, but the confidence intervals are really wide. There are a lot of suspicious things like hair-length preferences being impressively and significantly heritable for women but not heritable at all for men, and on one hand the two findings’ 95% confidence intervals do overlap, but on the other hand that’s because all of the confidence intervals are super-wide anyway. Overall I guess it’s nice that this study doesn’t blatantly break Turkheimer’s First Law, but I’m not going to draw any sweeping conclusions off of it.

I have one more twin study here, Lykken and Tellegen (which coincidentally opens with the same Pascal quote as this post). This experiment has an interesting design – they ask dizygotic and monozygotic twins to rate how attracted they are to their co-twin’s spouse! (I hope these researchers went above and beyond in keeping all their data confidential!) Despite those human interest stories where two identical twins separated at birth rediscover each other and find that they have both married blonde Corgi-owning optometrists named Theo, in this study there was very little resemblance between the spouses of either identical or fraternal twins. On average identical twins correlated with each other at r = 0.57 on some long list of variables, but their spouses correlated with each other only at r = 0.14 (identical) and r = 0.11 (fraternal), and most of this was just similar religious and educational backgrounds that don’t surprise us (we already know people tend to marry others of the same religion and social class, and twins are no exception).

This is interesting because it suggests a minimal role of either genetics or shared environment in mate choice, even in the most extreme circumstance (identical twins raised together). And in fact, the authors remind us that people who have had multiple long-term relationships often choose people who don’t resemble each other in more than the most superficial ways (if the authors had exposure to polyamory, they might note that most people’s simultaneous partners aren’t very similar either).

Maybe this just supports Zietsch et al’s hypothesis that mate preferences, whether imprinted or genetic, don’t really matter because we don’t have a whole shopping aisle of mates lined up to choose from and we’ve got to take whatever we can get. Yet even so there’s room for surprise. Even if only 5% of the opposite sex is interested in you, that still leaves the average person hundreds of different choices over their lifetime. Surely there should be some degrees of freedom for people to pick who they end up with. So what’s going on? If not by genes or shared environment, how do we choose our significant others?

The authors’ guess: true love, a mysterious and magical thing totally beyond scientific understanding.

If we provisionally accept our interpretation of these data, we are left with a curious and disquieting conclusion: Although most human choice behavior lawfully reflects the characteristics of the chooser and of the choice, the most important choice of all, that of a mate, seems to be an exception…we outline a theory that is compatible with these interpretations, namely that human pair bonding is relatively adventitious, based on romantic infatuation which, as Stendhal observed, “is like a fever that comes and goes quite independently of the will.”

Sure. Sounds plausible. Let’s just consider this whole matter closed.