Are women getting better looking?


Update: The author of the paper clears up confusions.

Update: Here's the paper. End Update

The British media is abuzz with another paper from Satoshi Kanazawa, the evolutionary psychologist who has great marketing savvy. I can't find the study online anyway, so here is the Times Online:

In a study released last week, Markus Jokela, a researcher at the University of Helsinki, found beautiful women had up to 16% more children than their plainer counterparts. He used data gathered in America, in which 1,244 women and 997 men were followed through four decades of life. Their attractiveness was assessed from photographs taken during the study, which also collected data on the number of children they had.


One finding was that women were generally regarded by both sexes as more aesthetically appealing than men. The other was that the most attractive parents were 26% less likely to have sons.

Kanazawa said: "Physical attractiveness is a highly heritable trait, which disproportionately increases the reproductive success of daughters much more than that of sons.

"If more attractive parents have more daughters and if physical attractiveness is heritable, it logically follows that women over many generations gradually become more physically attractive on average than men."

The Daily Mail has more numbers:

Their attractiveness was assessed from photographs taken during the study, which also recorded how many children they had. He found attractive women had 16 per cent more children, and very attractive 6 per cent more children than their less attractive counterparts.

Meanwhile the least attractive men had 13 per cent fewer children than other men. The findings build on a previous study which found attractive women are easier to find than handsome men because beautiful parents are more likely to have daughters than sons.
That study found that the beauty gap between men and women continues to grow with women becoming more attractive than men.

Researchers demonstrated that beautiful people are 36 per cent more likely to have a daughter than a son as their first born child.

The paper is built in previous research, which is a bad sign. Statistician Andrew Gelman pointed out several years ago major flaws in Kanazawa's methodology. Kanazawa didn't acknowledge the issues, in fact he kept going to the press publicizing his findings.

As for the idea that men and women are diverging in physical attractiveness so quickly, there is something strange about this. Remember: males and females inherit half their genes from an opposite sex parent. This has resulted in observations to the effect of very masculine men producing rather masculine females, so that there is a time-dependent trade off in fitness conditional on secondary sexual characteristics. This does not mean that biological differences between males and females can not emerge, obviously there is dimorphism in adult size, in addition to the sexual characteristics which are determined relatively early in life. But the genetic architecture of male-female differences takes a long time to emerge, trait value differences between the sexes take 10-100 times longer to emerge all things equal. See Genetics and Analysis of Quantitative Traits for the chapter on how the evolution of sexual dimorphism has to grapple with the problem of correlated response.

More straightforwardly, the rate of evolution due to selection is proportional to the reproductive variance, the heritability of a trait, and the correlation between reproductive variance and the trait value. In plain English, if tall people have many more children than short people, but whether someone is tall or short is mostly a function of environment, we are forced infer that there just won't be much directional genetic evolution on this trait since selection is not tracking much genetic variation. Similarly, if the reproductive variance is minimal the change in genetic frequencies from generation to generation will also be minimal. To have winners you need losers.

So how heritable is good-looks? I am sure that some readers know much more about the biology of appearance than I do, but I assume that there are two independent components of variation here: secondary sexual characteristics and symmetry. The former can often result in a physiological fitness drag in some species due to either the handicap principle or runaway sexual selection. In contrast symmetry presumably tracks overall genetic fitness and overlaps with some idealized "wild type." In The Mating Mind the evolutionary psychologist Geoffrey Miller posited that beauty which an outward signal of mutational load, and Armand Leroi relies on this model to explain why variance in beauty will always persist. Every human has a suite of deleterious alleles, some of them inherited from their parents, some of them new and unique. These alleles may not be fatal, but presumably they serve as a drag on development and physiological flourishing. If beauty is a signal of health, and health is an outcome of genetic normality, then the beautiful will likely be those with the fewest deleterious mutations. If the mutations of largest effect are relatively small in number, perhaps 100 or so, there will be enough variance among siblings for there to be a discernible range in outcomes (which we see). And if it is correct that every human has a few new mutations the well of potential ugliness is replenished every generation.

How does this align with the thesis posited by the authors above? I don't know since I don't have the paper in front of me. But I'm rather skeptical that the correlations between beauty and mutational load, and the variance in reproductive fitness correlated with beauty, are high enough to have the effects that are posited above. Additionally, selective transmission of mutational load (or lack of transmission) seems really bizarre (though I'm sure you can posit some sort of distorter at work in the Mendelian process). If mutational load is dropping, then sons should be getting better looking concomittantly. Thinking hard about the proximate mechanisms of biological inheritance as well as evolutionary genetic dynamics might give fruit to some interesting ideas, but relying just on the "logic" of evolution is too thin a methodology me thinks.

Note: It's an interesting point that the thesis promoted by this paper goes against the dominant strain of concern among evolutionary biologists who have mooted the issue of evolution, fitness and selection over the past few centuries, that is, one of degredation or increased genetic load. W. D. Hamilton was the most famous worry-wart on this issue.

More like this

I haven't read the paper, so maybe this is off-base, but it seems to me that the scientific value of the observations the perception of "beauty" is often dependent on non-genetic factors like the use of cosmetics, hair styling, and body modification such as breast implants or liposuction.

If Kanazawa's observations are indeed statistically significant, it seems like he could actually be measuring the effect of skill in presenting oneself within the currently fashionable beauty norms rather than "beauty" itself. I think that's an important distinction, because adherence to social feminine beauty norms can vary by class and education level, which would throw some unaccounted-for variables into Kanazawa's analysis.

Kanazawa is incredibly sloppy as a scientist. He's a bad stereotype of an evolutionary psychologist.

peggy, i'm giving the benefit the doubt, obviously. but yeah, the stuff i'm talking about is probably downstream of what you'd want to start scrutinizing.

Kanazawa is incredibly sloppy as a scientist. He's a bad stereotype of an evolutionary psychologist.

i think he might very well be doing more harm to evo pscy than not. yes, he popularizes some provocative ideas, but they're so flimsy upon first inspection that i think it'll discourage people from digging deeper into the area of research.

> If mutational load is dropping, then sons should be getting better looking concomittantly.

Don't you think sibs are somewhat correlated for beauty across gender? In other words, don't you think the sons are getting better looking concomitantly?

There are reasons I can imagine why the effect might be a little stronger on women.

> I'm rather skeptical that the correlations between beauty and mutational load, and the variance in reproductive fitness correlated with beauty, are high enough to have the effects that are posited above.

Your skepticism seems conditioned on r(FITNESS, BEAUTY) being the same today as it was before the demographic transition, and indeed before agriculture. It seems to me that this r could be very different today.

I'm guessing these data were not corrected for infertility - I see no reason to do so. "About 10 percent of women [...] in the United States ages 15-44 have difficulty getting pregnant or staying pregnant, according to the Centers for Disease Control and Prevention." Disease/syndrome prevalence stats often seem quite high, and I think they have a tendency to use rather broad definitions. Nevertheless, infertility, if related to beauty, could explain a part of the modern r(FITNESS, BEAUTY). Probably a small part.

But what causes the rest of it? Well, a few percent are debilitated by disease and some of those will choose not to reproduce. This could explain another small part if the beautiful are less diseased.

And beautiful women might have sex more, leading to more accidental pregnancies.

But, intuitively, the biggest determinant of fitness today, by far, seems to be the sheer desire to have more kids. One would think that was a very small factor indeed before 1800, and certainly before agriculture. That's why it's easy for me to imagine r(FITNESS, BEAUTY) being much higher in the past.

We find a womanâs lifetime parity dominates her decision to have another child, while stepfamily status does not alter the fertility schedule, holding constant a womanâs lifetime parity. [I did not read this whole paper or even the whole abstract]

Lifetime parity could have varied with beauty, and of course, lifetime parity is not fitness itself. The vigor of the offspring also matters - and mattered manyfold more pre-1800.

So, why is there still (maybe) a beauty-parity correlation today? Hmm. Estrogen or something? I got nothing with light googling.

I'll fess up my "prior convictions." It's hard to see how the large differences in female attractiveness can exist, without taking a beauty to wife having a huge and very real payoff. I don't see why the ratio of the fitness values of two women (to me as wives) should be much different from the ratio of their attractiveness. If I may be perfectly callous, I'd gladly trade both Jessica Alba and Megan Fox, provided I haven't gotten too attached to them yet, for Gillian Anderson. I'm assuming they are all 22 years old, and equal in wealth, talent, IQ, and apparent health.

I mean, the other two are very pretty, but the "smoldering" sexy look is not for me. I very much go in for the "concerned, tender" look (which I suspect may be a morph that apes the transient concerned facial expression all people have).

More generally, I think most men would "trade" two 30th percentile wives for one 60th percentile one, in the artificial thought experiment above (which I put forth just to have a way to quantify physical attractiveness in a non-relative way). Or, if not 60th vs 30th - surely 70th to 20th.

To absolutize it in another way. If I were a marriageable bachelor of 20 years, I would do dangerous things to get Ms Anderson instead of a median wife or 75th percentile one. Really dangerous, in fact - because, remember, in this scenario I am a pre-modern male. I'm inured to being around great danger, around violence, war, vicious animals, hunger, etc - quite regularly.

Why would instinct steer us wrong on this all-important matter? What kind of weird hitch in our evolution could possibly make us go wrong?

Well, weird disconnects between instinct and one's best fitness interest CAN happen. But they can't last. So if this were some kind of weird hitch it would be 10-100 times more likely to evolve sexually antagonistic selection than it would be not to - because sexually antagonistic selection is 10-100 times slower to finish its work.

By Eric Johnson (not verified) on 27 Jul 2009 #permalink

I wrote (roughly): "Weird disconnects between an instinct and the fitness value for the corresponding behavior can occur, of course. What they cannot do is LAST."

That's assuming, of course, that the fitness value of the behavior isn't constantly moving/oscillating in a significant degree.

By Eric Johnson (not verified) on 27 Jul 2009 #permalink

This story has been all over the interwebs, and it's infuriating. Not only is the study probably seriously flawed ( I, also, am having a hard time finding it), but it's reminiscent of social darwinism. I thought we'd hashed this whole thing out decades ago.

I actually did look up the h^2 of attractiveness -- it's in a post or comment section at GNXP. The original source I can't recall, but I read it in the refs to one of Kanazawa's articles on Trivers-Willard. Try searching for relevant terms, and it should turn up. It was about 0.36 (narrow-sense), iirc.

Yeah, the slow dimorphism is a real killer here. Women evolving to look better than men over hundreds of thousands of years -- sure. But so much so quickly? Smells fishy.

What's your exact thought, Carman? That it can't be true? Isn't true? Shouldn't be true? No one should study it?

I can respect some of those opinions, to a greater or lesser degree.

Look, I'm not exactly that hot of a commodity myself. But one already knows that. It already is what it is. How does it make things worse to understand why?

And what if one day, every face could be fair to see? Why should there be diseases? This isn't our world, but we can make it ours. No - I don't mean by rounding up and machine gunning all the ugly people. That's just so passe. I mean by zero-coercion liberal eugenics, where parents freely choose what they consider to be the best embryo out of 20 or 30, before implantation.

I'm sure you want to help suffering and weak people. You think their problems can't possibly have much to do with their genotypes and the right program and culture will balm all wounds. Or that liberal eugenics can't possibly be carried through without destroying society. But have you really thought about it? As the "hip" young folks in SoHo and Greenwich Village like to put it these days, "I beseech you Sir, in the bowels of Christ, think it possible you may be mistaken." And maybe sample a good transhumanist writer like Dave Pearce of

By Eric Johnson (not verified) on 27 Jul 2009 #permalink

Were zero-coercion liberal eugenics even possible, we don't know enough yet to make it "safe" were too many to use it. I don't know how many too many is.

Most parents have no idea how to choose and that level of education isn't likely to hit the general public until years (if ever) after the scientists get to it... and I'm sure they are not there yet.

Please don't interpret that as being anti-science or anti-research. It's just a strong desire for lots of evidence before something is implemented. The law of unintended consequences hasn't yet been repealed, has it?

The zero-coercion part is not possible because we're social animals with a tendency to conform and to apply pressure to achieve some level of it.

Donna, you're probably right about zero coercion.

If one were to just use embryo selection, it would be the same as in vitro fertilization, which has been done for a long time. With the addition of taking tissue from the embryo for genotyping (unless tissue is naturally shed) - I'm not certain just how the safety of that is studied.

I suppose it's possible side effects might arise in time, due to progressively concentrating the alleles that code the desired traits. I think this would take generations to become an issue. If you selected only for IQ, it would only go up ~12 points per generation. If you select also for health and beauty at the same time, the slopes would be gentler still. But perhaps it could be an issue, in time. Larger head size might be one problem.

I admit, it's hard to totally rule out "medical" (broad sense of the word) side effects with empirical studies. Common side effects can be detected. Obvious ones too. It's rare, subtle side effects that are quite difficult to pick up - depending on just how rare and just how subtle you mean.

Unintended consequences - definitely a concern.

To me it's at least very, very hard to think we shouldn't even study the issue at all (I know you aren't stating favor for that position). Above all because what Hamilton was worried about, as mentioned by Razib, might be true. The accumulation of deleterious alleles because purifying selection is now relaxed. If it's true, there would be no choice but to do something about it sooner or later. And it would be, in practice, by far the most important fact ever discovered - no matter whether the correct timescale for concern (which is hard to nail down) is generations or millennia.

My opinion is that it is quite likely to be true, and to be concerning in magnitude. Fortunately, it can always be fixed later on, even if the average genome becomes worse for a while. Not even one of the better alleles will ever be lost permanently.

By Eric Johnson (not verified) on 27 Jul 2009 #permalink

Every face can't be perfectly attractive, because every child isn't an exact copy of its parents. I predict that humans are already as attractive to each other as they can possibly be, and all that the disproportionate reproductive success of pretty people versus ugly people does is "tread water" against the tendency for some children to be a bit uglier than their optimally-pretty parents.

I read that something like 90% of each generation's sage grouse are the offspring of like 10% of the male sage grouse of the previous generation. The remaining 90% of male sage grouse are effectively locked out of posterity because they're not pretty enough for the ladies. That should make each generation of sage grouse more awesomely macho than the last, but that doesn't happen.

I have found that the women who rejected me (or would have rejected me if I had had the courage to ask them out) were far more good looking than the women that would go out with me. Now that I'm a member of AARP and rather overweight there are extremely few nubile women that are interested in going out with me, so naturally I find that virtually all women are far more good looking. So, over the course of a mere 50 years the proportion of very good looking women has gone up dramatically.

I asked Markus Jokela for his paper and found that the article does NOT claim that women or men are becoming more attractive (it suggests this possibility only in passing in the discussion at the end and indicates that the positive gradient found is less than half that observed in studies of non-human animals). The article only claims to have found that women who were rated as attractive, were likely to have more children overall than women who were not rated as attractive. The difference was small but statistically significant according to the author. Kanazawa's findings about attractive women having more daughters are discounted in the article (on the strength of Gelman's evidence). So, I do not see how the so-called science editor in the Sunday Times could warrant dragging Kanazawa's findings into the footlight again based on Jokela's article.
Jokela incidentally also touches on the weaknesses of his study in that it is based on a sample from mainly white Amercicans from the generation born around 1940. Nowadays, attractiveness is rated differently and can be articificially enhanced as can fertility, while marital duration is lower on average.

By P.S.Paaskynen (not verified) on 28 Jul 2009 #permalink

Cool. If you have time, would you mind giving the the citations to papers on the same phenomenon in animals?

I wonder how they decided if animals were attractive or not.

Lead author plus paper title is sufficient for googling something, so there would be no need to type out the whole citation.

By Eric Johnson (not verified) on 29 Jul 2009 #permalink

Statistician Andrew Gelman pointed out several years ago major flaws in Kanazawa's methodology.

Gelman's paper was an enjoyable read and reminds me of my maxim (things are always reminding me of my maxims): "Statistical tests are not a substitute for actual thought."

Of course, actual thought is a tall order for some people so perhaps we should have some sympathy for Kanazawa, who is possibly doing his best with the tools at hand.

Oh wait, this is the guy who said that we should have responded to 9/11 by killing over 300 million people? Nevermind, cancel the sympathy.

I just wanted to mention to add on to PSP's point. The paper also claimed to find unattractive men have less reproductive success, and discussed directional gradients for both sexes.

By DarkLayers (not verified) on 01 Aug 2009 #permalink