This was my second post on the work of the controversial sociologist Satoshi Kanazawa, author of noise-mining classics “Beautiful parents have more daughters,” “Engineers have more sons, nurses have more daughters,” “Violent men have more sons,” and so on.
My post had such a measured and polite tone:
Unfortunately, the new paper still has some problems. . . . Not to be picky on this, though, but . . .
I’m disappointed that Kanazawa couldn’t find a statistician in the Interdisciplinary Institute of Management where he works who could have checked his numbers . . . Just to be clear on this: we all make mistakes, I’m not trying to pick on Kanazawa. I think we can all do better by checking our results with others. . . .
Kanazawa is looking at some interesting things, and it’s certainly possible that the effects he’s finding are real (in the sense of generalizing to the larger population). But the results could also be reasonably explained by chance. I think a proper reporting of Kanazawa’s findings would be that they are interesting, and compatible with his biological theories, but not statistically confirmed.
My point in discussing this article is not to be a party pooper or to set myself up as some sort of statistical policeman or to discourage innovative work. Having had this example brought to my attention, I was curious enough to follow it up . . .
OK, maybe I should always be so polite. Whatever. Where I was wrong in the above post was to say that Kanazawa is looking at some interesting things. To put it another way, what I was missing was that he was studying tiny effects using very noisy measurements. It was an extreme example of the kangaroo problem. These were studies with N=3000 where you’d need something like N=1,000,000 to learn anything useful.
And I didn’t realize this! I just hadn’t thought of things that way. It took me years, first to think this problem through in the particular example of sex ratio research (see this paper with Weakliem) and then more generally (see this paper with Carlin). The concept of Type M errors had been around for awhile (this paper was published in 2000), but it took me, and the fields of applied statistics, many years to put it all together.
So back in 2006 I could write about published research that was nearly pure noise-mining and focus on incidental aspects such as multiple comparisons without realizing that this researcher was, as they say in poker, drawing dead.
I wrote that his findings were “interesting, and compatible with his biological theories, but not statistically confirmed.” It would’ve been more accurate to say “too noisy to provide any useful information; compatible with his biological theories in the same way that any set of random rolls of the dice would’ve been compatible with his biological theories.”
I was more polite back then because I was still under the impression that research such as in that beauty-and-sex-ratio paper could possibly advance science. The theoretical part could advance science, I guess, but the empirical part . . . you might as well just be flipping coins. But I didn’t realize it then. So I shouldn’t be surprised that some people still don’t realize it now. As the doll says, Math class is tough!
It sounds to me as if the recommendation is actually ‘don’t go it alone and try to reinvent everything’ because that’s what an autodictat with a textbook really ends up doing; but instead ‘enter a community of practice’ which works out ok unless you have a community of malpractice….
Benk:
Yes, a “community of malpractice” . . . well-put. In this case, it’s a community that’s supported internally via journals, promotion committees, etc., and externally via Freakonomics, Ted, NPR, etc. And the support goes both ways: the academic publications give the credibility which is enough to fool Freak, Ted, etc., and the public acclaim motivates the academics to continue to do and reward the bad work. And when these people are criticized, they will sometimes try to figure out what went wrong and do better, but more commonly they ignore or shrug off the criticisms (as in the case of the researcher mentioned above) or sometimes go on the offensive and attack their critics (for example, google this blog for “Stasi”).
Apparently, some people think these studies are interesting because of the evolutionary implications. Kanazawa wrote:
” …if beautiful parents have more daughters and physical attractiveness is heritable, then, over evolutionary history, women should gradually become more attractive than men. The analysis of the National Longitudinal Study of Adolescent Health (Add Health) confirm both of these hypotheses. Very attractive individuals are 26% less likely to have a son, and women are significantly more physically attractive than men in the representative American sample.”
Well that is interesting!
This quote manages to capture the zeitgeist of evo psych, and not much has changed. The basic method:
1. Find a human trait (anything works!).
2. Show variation in that trait across a population (no problem!).
3. Explain how that trait evolved based upon the axiom that only advantageous traits can predominate (go ahead and make up why it is advantageous, everyone else does).
The problem is that this approach is structurally identical to astrology, just replace “the axiom that only advantageous traits can predominate” with “how the stars were aligned when you were born.”
“Community of malpractice” is more right than wrong, but evo psych could not even exist without the fitness paradigm. If the most basic paradigm in your field is hokum, lots of malpractice will inevitable ensue.
So what’s the problem with the ‘fitness paradigm’? It kinda worked ok for Darwin et al. Or do you mean its bastardized version in the pseudo-science literature – the ‘just so’ stories?
“So what’s the problem with the ‘fitness paradigm’?”
Let’s start by unpacking the quote from Kanazawa.
“…if beautiful parents have more daughters and physical attractiveness is heritable, then, over evolutionary history, women should gradually become more attractive than men.”
As Andrew points out below, beauty and fecundity – the latter being important to evolution – are not even related. An attractive woman can have a hundred sex partners and no children. A homely woman can have one homely sex partner and fifteen children. The bigger point is that evolution is so messy that the idea that beauty will increase in the described manner is one of those “just so” stories.
There is also a “piranha problem.” There are lots of human traits, evolution cannot be driven by all of them at once. The biggest problem with the fitness paradigm is that there is no evidence that it works the way Darwin suggested. Goldschmidt claimed that variation within an interbreeding population is essentially meaningless from an evolutionary perspective, and does not form the raw material from which speciation ensues. Eighty years later, we are still looking for evidence that new species arise by the gradual accumulation of advantageous traits. In fact, there is now good evidence that speciation does not happen that way, and if it doesn’t, Goldschmidt was right. And if Goldschmidt was right, then the entire field of evo psych, which is based upon the foundational premise that within-breeding-group variation drives future evolution, is an exercise in navel-gazing.
I cannot give a full answer to the question “what is wrong with the fitness paradigm?” in a blog comment, but I highly recommend Oliver Sacks “The Island of the Colorblind.” The understanding gained from this book is that the actual outcomes of evolutionary processes are messy beyond our wildest imagination.
One last point goes back to something I have mentioned on this blog before. The effect sizes that Kanazawa attests are meaningless in his formulation. The concept of “will win out” in an evolutionary sense applies to even the tiniest advantage in fitness. So if beautiful parents are only 1% more likely to have girls rather than 26%, it doesn’t even matter, the pretty girls still “win out.” It’s nonsense.
Like you I don’t have time to write something meaningful and brief. So, like yours, this will have to do….
Much of evo-psych, like evo-med, belongs in the fiction section of the library, period.
The quote that Andrew shared about beauty and fecundity is missing a boat-load of essential evidence and argument that would be needed to make it a serious scientific (or intellectual) point.
Your n-of-2 example of the relative fecundity illustrates biological variation but does disprove Darwin’s theory. One would have to know the average difference in fecundity, and provide evidence that beauty (whatever that is) has a genetic basis. Most theory, models and understanding of evolution are based on average effects sometimes with a good dose of stochasticity. (For better or worse.) Examples of exceptions don’t really probe the rule because even an incremental difference can be meaningful despite with striking exceptions. The ability of subordinate males to breed does not disprove the general advantage of dominance. The paradigm obviously allows for exceptions simply because of the messy nature of biology. As Sewell Wright showed even a heavy doses of chance (drift) complicates but does not compromise the paradigm.
Goldschmidt was more right than wrong, but so was Darwin.
‘All the human traits at once’ – as Lewontin and others showed (I am setting R Dawkins for now), gene interactions (epistasis connects many traits (pleiotropy) so that selection is based on individuals, which represent the cumulative action of all traits, and not on individual traits (or genes). Genes and traits do not act in isolation but instead emerge from the entire genetic background and from all environmental experiences. (Of course the various genes and environments are not equally important, they just can’t be dismissed as irrelevant.)
I’m just not convinced that the Darwin’s evolutionary paradigm belongs in the dust bin. But that’s a discussion for another day…..
But for now I agree with you and Andrew that work like Kanazawa’s is not worth wasting ATP or risking brain damage. And lots of evo-psyc and evo-med is just silly. Poor science, poor thinking don’t deserve serious consideration, unless lessons can be learned, which is pretty rare. But that doesn’t mean the evolutionary paradigm is wrong.
I thought people accepted punctuated equilibrium by now. Ie, the biggest drivers of evolution are local or global catastrophes that wipe out a large fraction of the species and/or create barriers to interbreeding between different communities.
ouf! Matt S, sorry. My first sentence is pretty snarky. But it was aimed at me rather than you, though it doesn’t read that way. You said your earlier reply would be too long for this blog. I simply meant that I didn’t have time or space to write anything very meaningful, or brief (short is harder to write than long). So I apologize that what I said came across badly, it did. Your reply to me was thoughtful, and I hope the rest of my reply to you was reasonably thoughtful. The issues Andrew and you raised are interesting, and like many topics here deserve more discussion than a blog affords.
We are off topic so this will be my last post here, but I wanted to clarify a couple of things.
When we consider human evolution, what we have are species, from Australopithecus to various Homo species. What we lack is any evidence of gradual change causing one species to become another. For a long time the best explanation was that the fossil record was lacking, and there are still very few human fossils. But meanwhile, the evidence has been growing that those intermediary genotypes may not have existed at all. Instead, human species and most (all?) other species may have evolved saltationally. To get back to the topic at hand, what is the evolutionary significance of beauty and sex ratios “winning out” if speciation is mostly or entirely saltational? This is precisely the point that Goldschmidt made in his book “The Material Basis of Evolution.”
I certainly don’t mean to imply that selection does not occur or that fitness is not important. Just that you cannot tease out how evolution occurred in the past based upon the distribution of extant traits and some squishy relationship between competition and fitness.
Joe wrote:
“I’m just not convinced that the Darwin’s evolutionary paradigm belongs in the dust bin.”
I wouldn’t put it that way. “On the Origin of Species” is a landmark in human thought, and that won’t change. Darwin can’t be faulted for not figuring out speciation, it is a topic that remains contentious to this day. And he pretty much knew he hadn’t figured it out.
Matt –
Nice comment (even if I did have to look up saltational).
It’s remarkable how confidently some people (e.g., Jordan Peterson, Brett Weinstein, Heather Heying to just name some of the currently most popular) reverse engineer from their favored ideology to figure out what was favored by evolution.
On the other hand, given how much money can be made by doing so, maybe it shows that inventing Just-So stories about survival of the fittest is a kind of proof of concept for survival of the fittest?
Saltation is a horribly vague way of attempting to explain gaps in the fossil record. It begs the question of the genetic events involved (that produced Goldschmidt’s “hopeful monsters”). In very deep evolutionary past in archaic organisms there is circumstantial (formation of mitochondria by ingestion of a bacterium in a proto-eukaryote) and rather solid (yeast whole genome duplication) evidence for events that would be considered saltational (whole genome duplication also happened in some early fish). But gaps in the human fossil record are much better understood in terms of punctuated equilibrium (which is an often misrepresented idea partly due to the flowery and obscure prose of Gould whose writings I quite like!).
So you generally don’t see that many intermediate forms since (a) speciation was mostly happening “elsewhere” in reproductively isolated populations that (b) accumulated significant genetic changes as a result of adaptation to a particular environment (which might produce some more rapid genotypic and phenotypic changes) and genetic drift (which could involve mutations in genes involved in developmental control of a sort compatible with Goldschmidt’s ideas). If at some point a descendant population (which might be a new species) arising from of the isolated population migrated back to the location of the original population/species then the fossil record at that location would show an apparent jump in form without evidence of intermediate forms.
You can see some of this gradual drift towards speciation by acquisition of small genetic diversity in so-called “ring species” where a circular migration (over very long periods) around a massive natural obstacle like a vast mountain range produces forms where spatially adjacent populations can interbreed but as you go further around the barrier to complete the circle the populations arriving back at the original point has become reproductively isolated from the original population. ID of ring species isn’t straightforward but genome sequencing of populations within the ring has provided support for this mechanism. That seems a more plausible means of speciation IMO and one that is supported by real world evidence.
The question about evolutionary consequences of very marginal fitness advantages especially within changing environments/contexts has been pointed out to have some dependence on whether evolution has a gradualist or saltational underpinning. IMO there’s massive evidence for gradualism and very little for saltation – whatever that actually means at the mechanistic level – but would be happy to be convinced otherwise. I’d certainly grant Goldschmidt some kudos for being perhaps unwittingly ahead of the game when it comes to modern Evo-Devo ideas and how mutations in toolkit genes can have major developmental consequences. Anyway was interested enought to look at a couple of his papers.
One thing that stands out from Goldschmidt’s writing is his style of argumentation. He assembles a series of “facts” and proceeds to argue for what he considers is a watertight deduction (1950’s style perhaps!). Mostly he takes his “facts” and argues to a completely wrong conclusion. His writing is full of assertions like:
So in “Repeats” and the Modern Theory of the Gene” PNAS 1950 he says:
In fact we know that Bridges was correct and gene/genome duplication is one of the most important drivers of eukaryotic complexity and origin of new genes.
And in “Position Effect and the Theory of the Corpuscular Gene” Experientia 1946 he says (referring to enzymes – i.e. the “specific catalyzer” of a step in a metabolic pathway):
This is followed by a list of “facts” from which he deduces an entirely incorrect conclusion (genes as commonly envisaged don’t exist):
And from the same paper:
Whereas we know that point mutations are largely the result of “a chemical change in a side-chain”.
I like the fact that there are curmudgeons throughout the development of scientific fields – and perhaps Goldschmidt spurred his colleagues to redouble their efforts to establish evidence for their (mostly correct) interpretations.
Matt, Joe:
The theory is potentially interesting. The trouble is that it falls apart for quantitative reasons (as discussed in my paper with Weakliem). Also it’s an example of the fallacy of the one-way bet, in that the theory doesn’t consider the possibility that the effect, such as it is, could be positive in some settings and negative in others, indeed it could change over time in feedbacky ways that could make the predicted outcomes not happen at all.
Past tense. Barbie doesn’t say that anymore.
Mark:
Believe it or not, I thought about that when writing that sentence!
In 2006 the culture was a different one. We psychologists could still get by with “sexy findings” based on cherry-picked studies with ridculously small Ns (relatively speaking, with regard to noise), and I guess even statisticians like you who knew better were nevertheless in the thrall of that culture. Things changed by 2010 to 2015, the revolution years. The culture has changed substantially, and it’s ok now to name things clearly now. We’re probably still years away from a state where you could say: “Yes, now the majority of reserach in the behavioral sciences is doing a decent enough job, statistics-wise.” But my hope is that we’ll be getting there eventually. (But what will you blog about then?)