Attractiveness: logarithmically perceived, normally distributed, sought for genetic benefits
Our intuition of space and time is to perceive them logarithmically: we place a bunch of tick-marks near “here” and “now,” and only measure orders of magnitude as we move outward. The linear scale used by scientists places a tick-mark at evenly spaced intervals. For example, between “here” and 100 miles away, humans may have a bunch of words for nearby distances — “right here right here,” “right here,” “heeeeere,” “here,” and “there” — while we would represent the majority of the distance that is not near with a few words, such as “theeeeere,” “over there,” “way over there,” “way way over there,” etc. Ditto for time. John Hawks reviewed a recent study, which itself contains many references, er, right here right here.
I think we perceive physical attractiveness in the same way. (Although quick Googling didn’t turn up anything, I’d be surprised if this idea were original — at least, it’s not a common idea.) The reason is the same as the above: when we use something like the popular “1 to 10″ ratings, we seem to finely slice up the attractiveness space near the “good-looking” end and place tick-marks increasingly farther apart as we move farther away, like this:
Just think about it — have you ever split hairs over how incredibly ugly a person was, like 1 vs. 2? Probably you have never done so, but I’ll bet you and your friends get into regular arguments about whether Jessica Alba, or anyone else like her, is an 8.5, 9, or 9.5. You probably save the “1″ from the “1 to 10″ scale only for the most distant monsters, humanoids so freakish you could not expect to reach them in a dozen lifetimes. And anyone in a large vicinity of that spot would be compressed into the “1″ category.
Why do we make these compressions — why not keep the fine structure of the space, like rate people from 1 to 1 million? Because we have limited vocabularies and cognitive resources, and because not all regions of the space are as attention-worthy as others for surviving and reproducing. We care a lot about what’s going on near us because the goings-on of the other side of the world, until very recently, had no bearing on our survival and reproduction. The same is true for time: until very recently, the very long run did not matter at all, so why bother measuring the next millenium in yearly intervals? Only the somewhat near future has mattered.
Continuing the analogy, then, it must be that it has been the good-looking rather than the ugly people who have been most worthy of our attention during our evolution, since that’s where the density of tick-marks is greatest. That is not a tautology. Indeed, the great evolutionary biologist William Hamilton had an idea that probably is too crazy to be true — that animals pay attention to attractiveness in order to avoid getting infected by parasites while mating, attractiveness signaling the mate’s parasite-free status and ugliness signaling their being bug-ridden.
But if that were true for humans (and I’m only talking about us now), then we would finely slice up the attractiveness space near the ugly end, reflecting our worry of getting infected: “definitely swimming in bugs,” “pretty buggy,” “buggy enough that I’ll be scarred for life,” “buggy but I won’t be too compromised by sleeping with them,” “low bugginess,” etc., and compress the vast expanse of attractive people into a few categories like “probably not buggy” and “definitely safe.”
So, it doesn’t look like we avoid the ugly but that we pursue the attractive, and that jibes better with the alternatives to Hamilton’s “parasite avoidance” hypothesis, namely the genetic hypotheses. First is Alfred Russell Wallace’s “good genes” hypothesis: we pursue good-looking people because their good looks signal having genes that have protected them against the ravages of pathogens, or whatever else may damage their health. And second is R.A. Fisher’s “sexy sons” hypothesis: we pursue good-looking people because, whether we find them good-looking or not, the potential mates of our offspring will, so we’d improve their reproductive success by giving them genes for attractiveness. The idea that attractiveness is logarithmically perceived doesn’t decide between these two genetic theories, but I think it does go against the “parasite avoidance” hypothesis.
Last, I used to think that attractiveness was lognormally distributed — that, due to the synergistic effects of different body parts, most people are ugly, and only a handful are good-looking. (And anyone who says otherwise is being polite, fooling themselves, or not looking at a broad spectrum of human beings.) However, that’s just based on my perception — perhaps attractiveness really is normally distributed. If our mind re-scales attractiveness using logarithms, then it will transform a normal into a lognormal distribution too.
Normally distributed traits suggest an additive genetic basis, whereby small effects across a large number of loci are added together to determine the phenotype.* A lognormally distributed trait, such as “genius” in the sense of “eminence in the arts and sciences,” has low heritability. The reason isn’t relevant here; what’s important is that we wouldn’t expect a lognormally distributed trait to have a mostly additive genetic basis.
But the one study that estimated the heritability of attractiveness, McGovern et al. (1996), found that the monozygous concordance rate was virtually twice the dizygous rate (0.65 vs. 0.33, respectively), and that means that the genetic variance in the trait is almost entirely additive. So the quantitative genetic evidence fits into the bigger picture of a normally distributed trait in reality, but which the human mind transforms logarithmically.
I can’t do any empirical tests like those in the study that John Hawks reviewed because we haven’t yet found an objective way to measure attractiveness. I don’t just mean that we can’t trust what people say — even if you measured a person’s attractiveness by taking the average level of physiological arousal from subjects who viewed the person’s picture, that still is a reflection of the subjects’ perception. However, I’m more of a theorist than an experimentalist, so maybe a clever doohicky-rigger out there can think of something better. If they do, the prediction is that true attractiveness can be measured on a linear scale just like time and space, and that on this scale, humans would place their “1 to 10″ tick-marks in a logarithmic spacing, as in the earlier picture, the same way that innumerate tribes do with their number words.
* This is unlike, for example, eye color, where only a few genes make most of the difference, and where eye color is mostly a recessive trait. That is, most people have dark eyes and only a few have light eyes — whereas a normal distribution of eye color would show most having green and equal, smaller numbers having blue or brown.
Labels: babes and hunks, Genetics
You’ve misconstrued Hamilton here. The thesis you attribute to Wallace was the one he favored. I’ll be re-reading “Mate Choice Near or Far” later on tonight to make sure.
From memory though, the idea came out of his work on Red Queen dynamics — if the purpose of sex is to rapidly churn out new recombinants in order to keep a step ahead of pathogens, then sexual selection should favor hard-to-fake signs of health as signifiers of genes that are currently “hot” for resistance to the commonest bugs. He did a study with Zuk that backed this idea up by correlating degree of “showiness” in male birds of various species with prevalence of blood parasites.
If you’re looking at human behavior, you have to account for the fact that both men and women have an incentive to misrepresent how choosy they really are. Making a great show of finding fault with relatively-attractive-but-not-perfect members of the opposite sex is a standard way of trying to boost your own perceived choosiness and thus desirability. This distorts the picture.
What is “true attractiveness” supposed to be (as opposed to the supposedly untrue 1 to 10 ratings people give)? How would one measure such a thing?
it doesn’t seem like a particularly interesting point to me. basically what you’re saying is that people are more concerned with small differences in people they find attractive relative to people they find unattractive, dressed in mumbo jumbo that makes it sound more precise or scientific
You’ve misconstrued Hamilton here.
Google “parasite avoidance,” and the last entry on the first page (from the book *The Bowerbirds*) says what I said. I’m not an expert on Hamilton — maybe he didn’t believe the theory deep down but just proposed it for investigation, or maybe only intended it for birds (as in the Hamilton & Zuk 1982 article).
What is “true attractiveness” supposed to be
Some objective measure of good looks. It sounds funny to say that there is attractiveness before perception, but maybe it’s just a composite of health and symmetry. Something like that.
basically what you’re saying is that people are more concerned with small differences in people they find attractive relative to people they find unattractive, dressed in mumbo jumbo that makes it sound more precise or scientific
The difference between a linear and log scale is not mumbo jumbo. It has real consequences too: you can’t do ordinary arithmetic when the variable is measured on a log scale. Simple case — taking the average. Any study that asks subjects to rate a person’s attractiveness, and then takes the arithmetic mean of these ratings, is probably doing something wrong.
You’d need to transform these ratings so that they were on a linear scale first (ideally by finding out what base the log scale is on, and then exponentiating the ratings using that base).
In the future, if you don’t know what you’re talking about, don’t comment.
Just based on personal observation, this makes sense to me (I apologize for stream of conciousness writing below).
This goes for many other things to: for example a “strong” radio station may be well over a million times stronger than a “weak” one, where as on a 1-10 scale “strong” may be rated as 8 or 9, and “weak” 2 or 3. A station rated as having strength of “5″ may be 1,000 times stronger than the weak, but only 1/1000 as strong as the “strong” station. Of course for heavily geek populated groups like radio dx’ers, a logarithmic scale is already established…for example “strong” might mean ~90 dBu (w/”very strong” upwards of 110 dBu), medium ~60 dBu, weak ~30dBu.
The biggest factor I can think of here is that most people find it a lot easier to think in terms of linear, rather than logarithmic scales, even though i think most things in nature are best seen in terms of logarithmic scales. I guess that’s why they call a*10^b notation “scientific” notation.
IQ is probably another thing best seen in logarithmic terms. By the way I believe the Gaussian distribution is designed for a logarithmized scale; this would make sense as something like height can’t have a negative value, but log (height/avgheight) can, and indeed does about 50 percent of the time.
I think attractiveness is on a pretty smooth continuum. I also think the upperbound is pretty low. The most attractive kids at highschool weren’t too different from the most popular faces you see in movies.
[Give me the best guys and girls from your highschool, Johnny Depp and Angelina Jolie, and some fresh judges, and then tell me some major path dependence isn’t at play here.]
The reason I think you see more hair-splitting about 9s and 10s, etc, is because attractiveness isn’t too objective after a certain point above the average. No men agree about the “hottest” celebrity because 10 is a myth. Endless subjective preferences make the 10.
[The most popular pornstars often look like transsexuals to me. The most popular fashion models look like adolescent males to Erik Holland. And Holland’s “scientifically” attractive glamor models often look like inbred trailer skanks to me.]
I don’t think any of those cross-cultural experiments or baby experiments, etc, would find any kind of uniformity in preference if we were just comparing pictures of, say, 9s and 10s. Perhaps even 7s, 8s, and above.
By the by, one study did find something like a nonadditive attractiveness (not in reported ratings, but in revealed ratings), but for males only:
“Overall, the relationship between outcomes and looks is similar for men and women. However, there is a surprising ?superstar effect? for men. Men in the top five percent of ratings receive almost twice as many first contacts as the next five percent; for women, on the other hand, the analogous difference in outcomes is much smaller.”
Fisher’s theory sounds circular to me.
But if that were true for humans (and I’m only talking about us now), then we would finely slice up the attractiveness space near the ugly end, reflecting our worry of getting infected…
Above study aside, I think there is perhaps some evidence that this is how attractiveness is perceived.
Dean Hammermesh has three studies looking at wages, performance evaluation, and attractiveness, and in all three there were greater penalties for looking bad, than rewards for looking good.
So if we define attractiveness by revealed treatment instead of survey opinion, then the differences at the ugly end are more finely divided than the differences at the attractive end.
Fisher’s theory sounds circular to me.
Google “sexy sons fisher” and you can get more detail. Basically, it’s a peer pressure argument. In 1994, you may not have found baggy jeans attractive, but if you wanted to fit in, you had to wear them. In Fisher’s view, if the potential mates of your offspring find X attractive, you’d do well to mate with someone who could give X to your offspring.
and in all three there were greater penalties for looking bad, than rewards for looking good.
That just follows from the perceived attractiveness scale being logarithmic: on the objective, linear scale, there are larger differences between perceived 1 and 2 than between perceived 9 and 10. So there should be greater penalties as you get uglier since the objective distance separating the tick-marks gets bigger and bigger.
So if we define attractiveness by revealed treatment instead of survey opinion, then the differences at the ugly end are more finely divided than the differences at the attractive end.
Again, they were measured on a subjective scale, so harsher treatment of uglies may simply mean that people respond to objective qualities (health, symmetry, etc.), in the same way that our visual system guides our behavior by responding to objective cues of distance (objects being fuzzier, closer together, etc. when farther away).
But perception tells us what was worth paying attention to: the near-term, the nearby, and the good-looking.
I don’t need to Google, I have his collected papers right here and have actually read them — just re-read and yep, I’m not crazy. When you find a secondary source attributing a dumb theory to a genius, it’s a good idea to go check and see if that’s what he actually said.
ok now, let’s not degrade our within-blog asabiya! ;-)
For the log nature of 1 to 10 scales, check out the IMDB.com movie rankings by readers: a 6 is an OK run of the mill movie, 7 is pretty good, 8 is very good, a 9 only comes out once every few years.
The ratings are reasonable to within, say, 0.5: there’s a pretty good chance you’d like a 6.7 more more than you’d like a 6.2.
Lots of data there to work with.
Another one of those academic urban legends, I guess.
For the log nature of 1 to 10 scales, check out the IMDB.com movie rankings by readers
That’s another good example. Statisticians warn psychologists all the time about these subjective ratings (especially the ubiquitous Likert scales), since you can’t take averages etc. etc. — but they still do it anyway.
academic urban legends
To clarify, who originated the idea. The idea is out there, people discuss and test it, but they wrongly attribute it. My interest is always in the ideas themselves. Just don’t make me anyone’s biographer.
It sounds to me like this would be an easy thing to test. Any grad students out there looking for a research paper topic? Why not simply take a group of random people and have them rank pictures of others according to their perceived attractiveness? Then it’d be a simple matter to see which type of curve best fits the data. I’m sure you could also glean more hypothesizes from evaluating how the distribution changes from examining subsets of random people (i.e. all normal weight 19-22 year olds, etc), or how rankings of the same picture by the same person change based on pictures they’ve seen prior.
Fisher’s theory sounds circular to me.
This is a good point; it is indeed a recursive process. But there are ways to “initialize” the self-reenforcement by arriving, by other means, at a weak sexiness for a given trait. The runaway sexual selection (AKA sexy son) mechanism might then amplify drastically both the trait and the tendency to find it sexually attractive.
How to initialize? One can use one of the other proposed mechanisms of sexiness and sexual selection – such as pre-existing sensory bias, Zahavi handicap, etc.
“academic urban legends”
Yeah, it’s a misattribution of an idea of William Freeland’s that Hamilton specifically argues against in the introduction to “Mate Choice Near or Far” in _Narrow Roads_. It’s unlikely that the person who wrote the bowerbird book actually read the H&Z paper, since the blood parasites they used as an indicator of pathogen load for each species are only transmissible by mosquitoes.
This is interesting, I’ve actually thought about this problem before. I tend to also use a logarithmic sort of scale for ranking girls, and that seems to be the general method for most guys, but then I’ve also known guys who gave out 10s to almost any attractive girl – maybe they were reserving the fine detail for the lower end of the spectrum (perhaps that was the best they could get)?
On a side note, recently an acquaintance of mine was confused by an American TV show she saw (I live in China) where an ugly man was called a 3 by women and an attractive man an 8, and she had no idea what the numbers meant – evidently the 1 to 10 scale for attractiveness does not exist here.