When I read that Ruth Bader Ginsburg’s late husband was a wonderful man but less accomplished than his wife, I was reminded of “Ivy,” one of the most impressive students I ever had the privilege to teach. Ivy excelled in her coursework, won a prestigious scholarship for postgraduate study in England, went to a top-five law school, clerked for a Supreme Court Justice, and is now a law professor at a great university.
Like Ruth Ginsberg, Ivy married a man who is very nice but less intelligent than she. This is not an unusual situation…
Psychological theories abound but the true explanation is a statistical one: Regression to the mean. It also applies to many other choices in life …
When choosing mates, intelligence (no matter how it is measured) does matter because people generally enjoy the company of others of comparable intelligence. But intelligence isn’t the only thing that matters. Women may be attracted to mates who are caring, funny, athletic, sexy, share similar interests, whatever. The point is that a woman whose IQ is, say, 140 won’t restrict herself to soulmates with 140+ IQs and there are a lot more potential mates with IQs below 140 than above. So she will most likely choose someone with an IQ below 140. Similarly, a woman with an IQ of, say, 80 won’t restrict herself to mates with IQs of 80 and there are a lot more people with IQs above 80 than below.
To make the argument more pointed, consider this question: Why don’t women who have won a Nobel Prize marry only people who have also won Nobel Prizes? First, because they look for achievements other than Nobel Prizes in a future spouse’s resume. Second, because a Nobel Prize requirement would rule out virtually all potential mates. Gary Smith, “Why intelligent women marry less intelligent men” at Mind Matters News
That makes a lot of sense and Smith goes on to apply it to many other areas of life.
But here’s a question: If that’s a nearly universal tendency, how could evolution proceed via sexual selection? Outliers would tend to get reabsorbed far more often than not.
We asked our physics color commentator Rob Sheldon if he had heard of the problem and he replied,
Here we get to another of those controversies in statistics.
“Regression to the mean” is also equivalent to “increasing entropy”. That is, given the complications of life, finding a mate, the intentional randomizing of sexual reproduction, etc. is it more likely that the end result is random or pre-planned? Because if it is random, if it is a matter of chance, then the greater number of opportunities near the peak of the distribution will be more likely than the rare opportunities near the tails of the distribution.
This is the “frequentist” approach of standard statistics, always assume complete ignorance of the process, and pick the direction of maximum entropy. By contrast, the Bayesian will say, don’t assume random, instead ask whether past results have been random or not, and use your best guess about what sort of bias will happen in the future. Bayesians know that bias exists, and there is no guarantee that in a large ensemble the average individual biases will all cancel.
In the words of my daughter, who was married last month for the first time at age 32—“don’t settle.” Of course, the danger there is that she will never get married. Gary Smith didn’t put “high IQ women without husbands” in his statistics, because if he had, he wouldn’t have gotten regression to the mean.
What does this do for Darwinism?
Everyone acknowledges this is a problem. The statistics of evolution was the “Neo-Darwinian Synthesis” put together by mathematicians in the 1930’s to replace Darwin’s “gemmules”, and involves a lot of population genetics on Mendel’s invariant genes. Pop-gen is all about the statistics of “regression to the mean”, but the bias (toward some outlier spot) is attributed to (a) large variation in the statistics of small groups (founder effect, island isolation, hopeful monsters); (b) mating bias (sexual selection); (c) environmental pressures (asteroid impacts, climate change), (d) etc.
In my mind, (a) the “small group variation” arguments of islands being the centerpiece of evolution has played out long ago. Sexual selection (b) has had to carry the weight of Darwinist progress for decades, and despite it being an ad hoc way to get any result you want, it too has been crumbling. Probably (c) is doing the heavy lifting recently–especially as epigenetics shows how there can be a ratchet between information in the genome and information in the environment, leading to progressively greater information content. My guess is that this optimism will burn out in 10 years as people understand that epigenetics is not genetics, and that it is difficult for information to cross that barrier. But in any case, (c) might be already too distant from Darwinism to count as a solution, because Darwin saw evolution as the success of an individual, or at best, a species, whereas modern treatments see ecologies as a unified web of adapted subtypes progressing together the way a baby turns into an adult. Is it even proper to ask “How does an acorn evolve into a tree?” or “Did random processes convert the acorn into the oak?” But if you say “the future tree was the purpose of the acorn”, then why not, “The current Anthropocene was the purpose of those trilobites in the Carboniferous” ?
So bit by tiny bit, we are stepping away from Darwin and his 19th century reductionism, his frequentist obsession with randomness, and the consequent inevitable regression to the mean.