About Rationally Speaking

Rationally Speaking is a blog maintained by Prof. Massimo Pigliucci, a philosopher at the City University of New York. The blog reflects the Enlightenment figure Marquis de Condorcet's idea of what a public intellectual (yes, we know, that's such a bad word) ought to be: someone who devotes himself to "the tracking down of prejudices in the hiding places where priests, the schools, the government, and all long-established institutions had gathered and protected them." You're welcome. Please notice that the contents of this blog can be reprinted under the standard Creative Commons license.

Friday, October 31, 2008

So much for absence of natural selection in modern human populations

I have recently criticized British geneticist Steve Jones for his claim that modern human societies are no longer subject to natural selection. My arguments were based on basic principles of evolution and population genetics. Now a new study shows that Jones is wrong on the basis of the available empirical evidence and, ironically, that evidence comes from research on the British population!

The paper in question was published by Daniel Nettle and Thomas Pollet, of Newcastle University, in the prestigious journal American Naturalist (November 2008). Nettle and Pollet took advantage of a large database called the National Child Development Study, an ongoing longitudinal survey of all people born in the UK in a particular week of March 1958. Note that this is not a sample, with all the statistical uncertainties that follow, but the entire population of the nation for a given slice of time.

The authors set out to disentangle the effects of education and wealth on number of progeny for both men and women, because most previous studies -- which typically found a negative relationship between education and offspring number -- are biased by the inability to separate these two factors. The results are simply stunning. There is a strong selection coefficient relating men’s wealth and the offspring they produce, meaning that the wealthier men do in fact have more children. This is despite a negative effect on (and therefore selection against) education, again in men. In other words, natural selection in contemporary British society is favoring wealthy but under-educated men (though the negative effect of education disappears at very high levels of wealth).

The data are equally clear for women, but the pattern is completely different. Selection is again strong, but it favors low income, with education having a negative effect when income is low and a positive effect when it is high. That is to say, natural selection is favoring women who both forgo education and do not accumulate wealth -- although if you really want to be educated as a woman, you better be rich for your education to have a small but positive effect on the number of progeny you have.

There is bit of empirical consolation for Jones, however. Nettle and Pollet compared their data to estimates of selection coefficients to a variety of other samples, both historical and contemporary. They found that the strongest coefficient of selection are detected in highly polygynous populations (i.e., not in Western-style industrial societies). Nettle and Pollet suggest that this is because polygynous groups have a higher variance in the number of offspring, and it is a well known principle of evolutionary biology that increased phenotypic variance makes selection more effective. (Jones’ argument, by the way, was different, and had to do with the changing age of reproduction in Western society, not with polygyny or lack thereof.)

Two caveats, of course, need to be kept in mind. First, this is by all means not a suggestion that women should aim for low paying jobs and drop from school to hunt for rich husbands. To go from a factual statement about what is happening to a value judgment about what ought to happen would be to commit the naturalistic fallacy that David Hume has warned us against, and which crops up regularly on this blog. Needless to say (or is it?), the British government should not look at these results and embark on a program to keep women from achieving equal pay on the job, or to discourage girls from entering higher education, just so that natural selection can do its job.

The second caveat is more subtle and interesting. A classic evolutionary biologist would point out that there is a difference between selection and evolution: the latter happens only if the traits under selection (in this case education and income) are heritable from one generation to the next. We do not know the extent to which male and female traits affecting education and wealth are genetically heritable (and I’m not too fond of so-called twin studies for a variety of reasons). But we do know that they are culturally heritable. Cultural inheritance does affect evolution, and in fact does so at a much higher rate than genetic inheritance, because cultural changes are much more rapid than genetic ones. There is nothing in Darwin’s theory that specifies what kind of inheritance is necessary for evolution: any mechanism that reliably passes traits to one’s offspring is good enough. Moreover, cultural inheritance can have a hitchhiking effect on the genetic makeup of the human population: even if the entire response to selection on wealth and education is due to culturally inherited factors, the next generation will still carry on a likely non-random subset of genetic markers of the British population, which means that biological evolution in the stricter sense of changes in gene frequencies will still be happening. Again, pace Jones.


  1. Selection against education? Are you saying that we're heading towards a time where Luke Wilson is the smartest man alive?

  2. Funny you mention that. I saw that movie because it was suggested by one of my graduate students, and didn't like it. Perhaps it was too close to the truth? :)

  3. I like the idea of cultural selection... memes having a phenotypic effect. The one thing we never know is: which characteristics will be advantageous in the future. I don't envision overall changes of the extent portrayed in Idiocracy even in 500 years. Unless. Unless there is a major human population catastrophe. And I have to say, I'm not discounting that at all. Or, an argueably more likely situation; we will use genetic engineering to "improve" ourselves to the point that we will appear to have evolved. Too bad I (we) won't be around to find out.
    Eat well, stay fit, Die Anyway.

  4. I'm curious to hear more about your thoughts on twin studies. I presume part of the problem is the variability of heritability by environment and gene x environment interactions.

    Have you written about this previously?


  5. If you think you're just "not a numbers person", you might be correct. It seems that some people are born with a naturally better sense of numbers than others - although that doesn't mean education can't improve your mathematical abilities. Being good at maths is thought to depend on two factors: the inherent sense of numbers that children, and some animals, possess from a very young age, and the formal education they receive at school.


  6. dtae,

    yes I have written on nature-nurture issues before, a whole book in fact (the "Phenotypic Plasticity" one advertised on the right side of this blog). The last chapter of that book gets into details about the difficulty of studying genotype by environment interactions in human begins.

    The chief problem with twin studies is that they tend to confound genetic and environmental effects. Even when one compares twins reared apart, the problem is that adopting agencies tend to give twins to families of comparable cultural and socio-economic background.

    Moreover, there are serious problems with the concept of heritability in quantitative genetics, particularly when applied to humans. An unsurpassed classic on this topic is: Lewontin, R. C. (1974). The analysis of variance and the analysis of causes. American Journal of Human Genetics 26: 400-411. That's the paper that inspired me to dedicate my professional career to gxe studies!

  7. Thanks Massimo. Both your book and Lewontin's article are on my to-read list to find out more.

    I suppose your main critiques don't apply as much to MZ/DZ twin studies without adoption?


  8. dtae,

    MZ/DZ studies suffer from other problems, chiefly the ones affecting heritability (which refers to genetic/environmental *variation* not to be confused, as it often is, with degree of causal determination).

    What I think is fair to say is that twin studies do suggest strong genetic influences on human behavioral traits. But we also have strong suggestions of environmental effects from research in social science. The key would be to find out about the interaction component, the elusive GxE. Unfortunately, to do that would require controlled crosses, clearly undoable in the case of humans, both for logistical and ethical reasons.

  9. The second caveat is more subtle and interesting. A classic evolutionary biologist would point out that there is a difference between selection and evolution

    Phew, I thought you wouldn't get to that part, but you didn't disappoint as usual! :-)

    Anyway, I think I still don't buy the "hitchhike by cultural evolution". As I was reading your post and the sirens were playing in my mind (about the non-inheritability you mention), I for a second considered the hitchhike scenario. OK, maybe there are no "genes for income/education", but everyone else rides along the body anyway.

    But on a second thought, I discarded that, and please let me know if anyone here thinks this is not the case, and why.

    My idea was the following: if wealth/education are not genetically linked, they should be distributed normally among the genotypes present in the population, specially a big population like United Kingdom's (so no bottleneck effect here). Now, if wealthy/educated men (or poor/uneducated women) are having more kids, I'd suppose they are ALL doing so, so the relative frequencies of the different genotypes would not change. Ergo, no evolution. If the population is/was small enough, we could of course have evolution by genetic drift, but our whole beef here is with evolution by selection to begin with, so there we go...

    What, if anything, am I missing?

  10. The importance of commas, illustrated by my own ineptitude in post reviewing, hahaha. :-D

    I meant "but you didn't disappoint, as usual", of course.

    Uff, good thing I caught that one in time.

  11. J,

    you are not missing anything, though that *all* rich men reproduce at the same rate seems a big assumption to me. But I do consider cultural inheritance a legitimate form of inheritance (see Jablonka, E. and M. J. Lamb (2005). Evolution in Four Dimensions: Genetic, Epigenetic, Behavioral, and Symbolic Variation in the History of Life. Cambridge, MA, MIT Press). If so, then we are still talking about evolution.

  12. Note that this is not a sample, with all the statistical uncertainties that follow, but the entire population of the nation for a given slice of time.

    I'm not certain I agree with this. Would immigration and emmigration dilute this to a sample of a population, rather than the population per se? Additionally, you seem to be assuming that no errors have been introduced to this dataset in the last 50 years. British birth records are presumably pretty good, but are they perfect?

    Having said that, would re-defining this dataset as a sample rather than a population have much of an impact on the results of the analyses? I doubt it.

  13. The authors themselves point out that this is not a sample. This doesn't mean that there is no measurement error, but only that there is no, well, sampling error.

    However, you are right, it wouldn't make much difference to the results if we were talking about a sample instead of an entire population.

  14. The following statement is key:

    "A classic evolutionary biologist would point out that there is a difference between selection and evolution: the latter happens only if the traits under selection (in this case education and income) are heritable from one generation to the next."

    While wealth is clearly *potentially* heritable, that has not been shown here. It has not been shown that wealthy men have more kids that are wealthy and they have more kids, etc. Hence it has *not* been demonstrated that selection (in the evolutionary sense) on wealth is occurring. In fact, we find little evidence of that in men in traditional communities (including those practicing polygamy) in Indonesia.


    To reiterate - it is premature to claim that wealth/status/etc is a positive selective force in humans, traditional or modern. Heritability is KEY!


  15. bhall,

    I disagree on two counts: first, this study does demonstrate selection, but it does not demonstrate evolution because it does not directly address the issue you raise, of the inheritance of wealth. That was the sense of my quote.

    However, do you really believe that, on average, the kids of wealthy people do not in fact inherit wealth??

  16. #1. Semantics

    #2. Sometimes, but evolution *has* as to be demonstrated at the population level and has not been. That is the achilles heal of many papers "showing selection". Who cares if the selection doesn't persist and the trait is not (very) heritable. I've been looking at traits that are partially heritable for a while now - the surprising result is how quickly evolution at the population level is canceled out by stochastic effects when the heritability goes down. I suspect that is true for most cultural traits.

    My main point is simply that we can correlate traits and fertility until the cows come home but that doesn't really show how selection is working in human populations. Show me a SINGLE study that looks at this over many generations and finds evidence for actual evolution due to such selection. I gave you one, we looked, and well, we didn't see it.


  17. I should add that the issue with heritability is so important because the group with the advantage is the minority.

    Here, in the case of wealth, not only are the wealthy a small minority but wealth is not perfectly heritable. The majority of offspring are born to the poor. Upwards social mobility and imperfect heritability combine to functionally erase the selective advantage == no evolution. I'd be happy to accept it if it is shown, but the burden of proof here is on those claiming selection/evolution. Neutrality has been the null model since Kimura.

  18. bhall,

    the distinction between selection and evolution is most certainly *not* a matter of semantic, as Fisher pointed out almost a century ago.

    Second, the fact that the majority of offspring is born to the poor doesn't affect the argument that there is (cultural) heritability of wealth. Again, just look at any society and tell me that wealth is randomly redistributed after every generation.

    So, the observation of heritability coupled with the measurement of selection *do* make a very strong argument for evolution taking place.

    Finally, Kimura's theory applies at the molecular level (and even there, not exactly, since the talk is of "quasi-neutralism"). We're considering high-end phenotypes here, a whole different ballpark.

  19. Let me clarify- I agree the semantics point is important, I meant that there is some variation in the usage of the term 'selection' as applied to human studies because of the way people look for it. So rather than argue about it, I will strictly adopt your usage and say more in a moment.

    I think my point has been lost, however. My point is this:

    I ~agree~ that they found evidence for differential reproductive success in a modern human population for a single generation and that it is correlated with wealth.

    I ~agree~ that wealth is heritable, though imperfectly.

    What I disagree with is the following:

    "So, the observation of heritability coupled with the measurement of selection *do* make a very strong argument for evolution taking place."

    Here's why.

    Selection is a population level phenomenon. There are two approaches to showing it - this type of study, where one correlates reproductive success with pheno/genotype over a generation or two and population level selection tests that look for evidence of deviation from neutrality in some test statistic or distribution (Ewens exact test, HKA, MK, Tajima's D, coalescent methods, etc...). One might call the second type of tests 'evolution tests' because the necessarily look over many generations, however nobody does that. Hence the variability of the usage of 'selection' as it applies to humans. In other words, in usage, selection is used both for 'selection' and 'evolution', at least in my field of human population genetics.

    When heritability is absolute, as in the case of animal studies, equating the two methods (and words) is justified because the first leads to the second - differential reproductive success + heritability leads to a non-neutral distribution over time at the population level or 'evolution'.

    This case is different because that is not strictly true. Thus equating the two methods, while intuitively appealing, is not intellectually justified. Ironically, I am the one advocating the skeptical position here. That is, I am saying, "now hold on a minute, just because over one generation wealthy men had more kids on average does NOT equal evolution based on differential fitness due to wealth". You appear to be arguing the opposite. On my side, I submit our recent PNAS paper on this very topic. The point of the paper is that which I just made - you can not equate selection and evolution shown using the first method in the case of imperfect heritability of a trait. It is a somewhat counterintuitive result - but that does not mean it is false.

    As for the neutral theory, I am not sure why you would suggest it applies only at the molecular level. That misses the essential intellectual contribution! I agree that its introduction was made in that context, but the ramifications are much broader. Functionally, it established neutrality as a statistical null model against which selection is looked for. In ALL contexts. It is the basis for ALL population level selection tests. Pick a selection test - the derivation comes from looking at the distribution of the test statistic or distribution under the standard neutral model and then looking at your value of the statistics and then trying to pry apart the effects of demography from selection in the case of a deviation. That way of thinking - showing selection as a deviation from neutrality - is key, and exactly what we aimed to do in our recent work that goes from this level, that of differential reproductive success in a single generation to the effects in the population over time. I stand by my claim - it is harder than you'd initially guess to get from what this paper showed to 'evolution' because if the heritability criterion. Take a look!


  20. bhall,

    wow, I disagree on *so* many levels...

    "I will strictly adopt your usage and say more in a moment."

    It isn't my usage, it is the standard in evolutionary theory.

    "Selection is a population level phenomenon."

    No, it isn't. Evolution is a population-level phenomenon (see the importance of the semantics?). Selection happens at the individual level (save for group selection, which is not what we are talking about), but is often *measured* as a statistical effect. Jonathan Kaplan and I dedicate two chapters of Making Sense of Evolution (see side bar on the blog) to this very problem.

    "When heritability is absolute..."

    I don't know what it means for heritability to be absolute, it is not standard quantitative genetic language.

    "As for the neutral theory, I am not sure why you would suggest it applies only at the molecular level. That misses the essential intellectual contribution!"

    Not at all, the theory, to my knowledge, has never been moved up to the phenotypic level where, frankly, it would make very little sense considering that it is easy to show that a lot of phenotypic characteristics are adaptive.

    "it established neutrality as a statistical null model"

    Which Kaplan and I think is a really bad idea. Null models are a big disservice to scientific inquiry, because they shift the burden of proof a priori. More and more people have realized this over the last few years, which has caused a move toward model comparisons-type approaches.

  21. Hmmm... kinda ironic for me to be arguing with a 'skeptic' for overstating the implications of data...

    You are arguing word choice not the claim. We do not disagree on the definitions. I understand evolutionary theory, though I admit I discuss it somewhat loosely, but not opaquely. So I'm not sure why you find it so difficult to understand me? Perhaps you don't have an actual counter claim?

    Claim: what has been shown is that men with more money have a few more kids on average. Period.

    More generally, showing selection over a generation or two in a population for a trait that is not always inherited does not translate into evolution as easily as people assume.

    Show otherwise.

  22. bhall,

    I'm not sure what the problem is either. I've made plenty of arguments in this exchange, as clearly as I can make them. More than this, we might just have to agree to disagree.

    "showing selection over a generation or two in a population for a trait that is not always inherited does not translate into evolution as easily as people assume."

    If there is selection (which the paper we are discussing demonstrated) and there is heritability (which is undeniable, again I don't know why imperfect heritability is a problem here,or even what you mean by imperfect) then it *must* follow that there is/will be evolution.

  23. Sigh, perhaps language truly is a barrier here...

    Imperfect heritability = not transmitted 100% of the time

    Absolute heritability = transmitted 100% of the time

    Here, wealth is NOT transmitted 100% of the time to offspring. In an upcoming paper we show, under a variety of models, how when there is imperfect heritability, the
    "must follow that there is evolution" part breaks down quickly. We show this with real data in PNAS using traditional communities where there definitely is selection on wealth, i.e. wealthy males demonstrably have more children. We do not see evidence of evolution in these populations. The heritability part is where it breaks down. This part is somewhat non-intuitive and needs to be demonstrated mathematically/via simulation, which I can not do here. But the significance is big, in that papers such as Nettle's lead to a big 'so what' if there is no evolution. The heritability assumption has not been considered so much because in most biological systems the traits are definitely heritable. It's only when we start to look at things in a cultural context where this assumption needs reconsidering.


  24. Here, for clarity I'll add some other aspects of our findings:

    1. Being wealthy is a minority designation. More wealth = fewer people in that class. So if fertility scales with wealth then smaller and smaller groups are producing more offspring on average for their size but little in terms of the whole population.

    2. Wealth inheritance cannot be perfect - the more children you have the less each one can get. So there is a diffusion of wealth that diminishes the returns even if all the wealth is passed on. Societies often structure this in different ways - first son gets all, more equal divisions etc. but the chosen system has clear implications for evolution.

    3. The wealthy class is not closed, people can enter it who were not wealthy before and people can leave that class by losing their money. That is fundamentally different than say selection on a gene - you can't change (yet ;-) ) what gene you have midway through your life.

    4. The overall economy is a major force in wealth allocation and huge changes can occur rapidly that quickly erase evolutionary gains. While this is true for all environments, the timescale is relatively short in modern economies.

    In the more typical circumstance of 100% inheritance (e.g. actual genes), those men who inherit the trait will have more kids, and those kids will have more kids leading to an increase in that trait over time. But wealth can't act like that for the above reasons. Thus we can't have the normal progression of selection -> evolution in the exact same way.

    What we think happens is something more like this - certain patrilines definitely have more offspring over certain periods (a few generations), but overall there is enough turnover in who is wealthy so that the 'signal of evolution' at the population level is snuffed out and the populations are indistinguishable from neutral ones. That does not mean that wealth isn't structuring things, culturally, demographically or genetically. Nor does that does negate the original finding, that wealthy men have more kids. It just means that the evolutionary consequences are subtler and more interesting. In general, at this point, we think that there is enough stochasticity in the system overall to make it very difficult for a cultural trait like wealth to really structure things evolutionarily in the way natural selection does in other contexts.

  25. bhall,

    now I finally understand what you mean by 100% heritability. You are actually referring to *inheritance*, not heritability (again, the darn semantic!).

    Heritability is a statistical measure of genetic to phenotypic covariances, and it makes no sense to talk about perfection (or lack thereof) in that sense.

    As for inheritance, if your argument were correct, probably life could have not gotten started, since surely early replication systems were imperfect. I need to look at your upcoming paper to see the technical arguments, but at the moment I'm highly skeptical of your conclusions.


Note: Only a member of this blog may post a comment.