More on Buller and Evolutionary Psychology

This is an ongoing series of meditations on evolutionary psychology inspired by my recent reading of David Buller's piece in Scientific American.  I have been thinking quite a bit in the last year about the relationship between evolutionary psychology, human behavioral ecology, and evolutionary genetics, and maybe these ruminations will help me get my thoughts clear on these difficult topics.  Caveat utilitor: these are not fully formed ideas but the blog is a useful device for organizing my sketches.

I found an interesting  critique of Philosopher of Science and evolutionary psychology critic David Buller's book, Adapting Minds. Edouard Machery and H. Clark Barrett wrote an extended, critical review of Buller's 2005 book in the journal Philosophy of Science.

I must admit that I find myself torn on some of these debates. I am sympathetic to many of the criticisms voiced by Buller, but think that some of the rebuttals are quite compelling as well. For example, Buller is highly critical of work on child homicide by Martin Daly and Margo Wilson of McMaster University.  Daly and Wilson, in a series of famous studies, suggest that child homicide (a rare event) is much more likely to be perpetrated by step-parents (including boyfriends).  The explanation for why this might be relates to the existence of an anti-cuckoldry mechanism in men's brains. Given the enormous obligate investment -- generally on the part of two parents -- entailed in the successful recruitment of human offspring, cuckoldry represents a potentially enormous fitness cost for human men.

In one study of child homicides in Canada between 1974 and 1990, Daly and Wilson calculated a risk-ratio that child homicides are perpetrated by step-parents vs. (putative) biological parents of 123.7.  Buller suggests that such results might simply arise because of ascertainment bias in the reporting of child homicide.  Specifically, he suggests that the cause of death listed on a child's death certificate is far less likely to be homicide if the act was perpetrated by a biological parent. In support of this argument, he cites a paper by Crume et al. (2002) which compared cause-of-death as listed on the death certificate with the cause determined by a interagency multidisciplinary child fatality review team.  This team reviewed child deaths in the state of Colorado and found that a substantial number of likely homicides were not reported as such.  They were then able to investigate which attributes of (alleged) perpetrators made ascertainment more or less likely.  They found that homicides committed by non-relatives (including boyfriends) were 8.41 times more likely to be recorded as such than were those committed by parents. Of 152 death at the hands of parents only 65 were correctly ascertained while 87 were not.  For the 51 deaths attributable to non-relatives, 44 were correctly ascertained while seven were not.  This yields an odds ratio of (44*87)/(65*7)=8.41 that non-relatives will be correctly ascertained compared to parents (the OR changes to 8.71 following multivariate adjustment -- it is this number that is discussed in the various papers). This seems pretty damning (and suggests there are major problems detecting fatal violence against children).  However, one point from this paper that Buller does not note in his critique (at least his 2005 paper in Trends in Cognitive Sciences) is that the odds of ascertainment for non-parent relatives -- including step-parents -- is not significantly different from unity. That is, the group that includes step-parents is as likely to be ascertained as biological parents.  My understanding is that Daly and Wilson's analysis applies to step-parents as well as boyfriends.  The theory certainly predicts this.

My sense is that Buller is reaching a little too far in this critique. While I would hardly consider myself an expert on the topic, I have always thought quite highly of Daly and Wilson's demographic work on homicide.  One of my students is currently relying heavily on their Chicago mortality study published in BMJ.  That is something I do have some expertise in and I think it is excellent.  What I want to know is this: what is the counterfactual to the Daly & Wilson work?  How many child deaths would need to be re-classified in order to have ascertainment bias be sufficient to account for their observed differences?  Daly & Wilson (2007) do just this sort of counterfactual calculation.  They assume that step-fathers are always caught, whereas biological fathers are never caught.  According to their calculation, such a scenario would imply that there were 500 unaccounted-for paternal murders to yield the observed rates.  This is where the problem comes in.  There simply aren't 500 deaths each year to children under five in Canada in that period that aren't due to congenital defects or infectious disease.  Mortality among the young is rare in developed countries. Clearly, not all of the effect that Daly & Wilson report can be attributed to ascertainment bias.  There seems to be some there there.

I think that this over-reaching is a shame.  The critiques that Buller levels in his recent Scientific American piece are serious and deserve to be taken seriously. Here, I specifically mean the idea that an analysis of the Pleistocene will yield significant clues for understanding the design of the human mind and that evolutionary psychology will be much use in helping us understand unique and universal human traits.  The tone of this debate (on both sides) seems to preclude serious consideration of these important concerns.

As I mentioned in my previous post, I find the latter problem particularly troubling because it suggests that there are some things we can never know about human evolution in a scientific way.  Depending on the question, one possible solution to this problem is something Marc Hauser used to talk about in Science B-29 at Harvard.  The problem was how to use evolutionary tools to explain the unique phenomenon of human language.  While human language is clearly a unique, derived trait -- and therefore in a difficult position with respect to scientific explanation -- there are features of human language (e.g., those described by Hockett in his design features of human language) that are shared across multiple species, making them amenable to the comparative method.  If we limit ourselves to specific autapomorphies -- as Buller apparently wants us to when it comes to Human Nature -- then we are sunk.  If we can find features of our cognition that are shared across species and look, as Darwin first suggested, at convergent solutions to similar problems across species, then we may have some hope of understanding the unique whole of human cognition. Of course, we can't do this for cognitive features that have arisen since the Pleistocene because we only have one remnant of the hominin clade left (us).

Regarding our ability to understand the design of the human brain based on our knowledge of the environment of Pleistocene hunter-gatherers, Machery and Barrett (2006: 236) write that Pleistocene hominins experienced a "reduction in sexual dimorphism in body size due to increased pair bonding and male investment in offspring and corresponding reduction in male-male competition." While I happen to agree with this point (and have two new papers either submitted or in prep elaborating my take on this particular phenomenon), it is, in fact, conjectural.  There is nothing to stop us from forming hypotheses about the mechanisms or functional consequences of human behavior that result from this conjecture, and there might be substantial value in doing so.  Nonetheless, I think it's important to note that it is hardly certain that the cause of the reduction in sexual dimorphism among Pleistocene hominins (something we are pretty sure of) was pair bonding.  I'm afraid to say that I am not the least bit confident that we will ever know this for certain.

Why do we think that Pleistocene hominins were "pair bonded"?  We know that sexual size dimorphism is a correlate of mating system.  Polygynous mammals tend to be sexually dimorphic.  The more polygynous, the more dimorphic.  Presumably, this arises through intra-male mating competition, where size matters for the outcome of agonistic encounters.  As detailed in our 1999 paper, the best paleontological evidence we have suggests that there was a substantial reduction in both sexual size dimorphism and dimorphism in canine teeth (another strong correlate of polygyny among Primates) with the emergence of the genus Homo.  This reduction in sexual dimorphism is attributed by many authors, ourselves included, as a signal of a change in mating system toward increased monogamy.  Does monogamy necessarily mean pair-bonding?  Not necessarily. (again, I do think it's true in this case and hopefully, I will finish the paper in which I discuss the details of this argument soon)  There is also the issue that humans are not much different in terms of sexual size dimorphism from chimpanzees, whose mating system is completely promiscuous.  Our teeth may rescue us here.  Chimpanzees are quite sexually dimorphic in their canine teeth.  But how do you weigh the importance of canines as a weapon in a species that makes tools, including weapons that allow it to kill from a distance?

My point here is that there is a good deal of uncertainty about basic aspects of Pleistocene hominin behavior.  This uncertainty is unlikely to ever be completely resolved.  As a result, I'm not convinced that looking for clues about human behavior and the design of the human brain in the behavior of Pleistocene hominins is necessarily the most efficient of productive avenue for understanding our psychology. I don't take the absolutist position that Buller seems to take that there is nothing to be learned about the present by studying the deep past (i.e., it is more than "pure guesswork").  I like the iterative approach of working between hypothesis generation and empirical test that Machery and Barrett describe and think that it sounds an awful lot like the process that most scientists employ in their work and it sounds like the way individuals adapt to dynamic environments.

I'll end this ramble with a question: Do you have to be an evolutionary psychologist to believe in Human Nature?  Buller seems to think so and to think that it's a bad idea.  I don't think of myself as an evolutionary psychologist, but I do think there is such a thing as Human Nature.  I am struck by the fact that despite the dizzying array of cultural diversity that is manifested by our species, a smile is a smile, embarrassment is embarrassment, and a look of consternation is a look of consternation.  We might find different things amusing, mortifying, or distressing but pretty much people everywhere experience these emotions and, because of our theory of mind, recognize them in others.  The work of Eckman, Eibl-Eibesfeldt, and Fernald, to name a few, is pretty compelling in this regard.  Do we have a cheater-detection module that was engineered in the Pleistocene?  Maybe.  Honestly, I don't care that much, but I do think that denying the existence of Human Nature is done at our collective peril.


Buller, D. J. (2005). Evolutionary psychology: the emperor's new paradigm. Trends in Cognitive Sciences, 9(6), 277-283.

Crume, T. L., DiGuiseppi, C., Byers, T., Sirotnak, A. P., & Garrett, C. J. (2002). Underascertainment of Child Maltreatment Fatalities by Death Certificates, 1990–1998. Pediatrics, 110(2), 1-6.

Daly M, Wilson M (2007) Is the "Cinderella effect" controversial? A case study of evolution-minded research and critiques thereof. In C Crawford & D Krebs, eds., Foundations of evolutionary psychology. Mahwah NJ: Erlbaum.

Machery, E., & Barrett, H. C. (2006). Essay Review: Debunking Adapting Minds. Philosophy of Science, 73, 232-246.

Wilson, M., & Daly, M. (1997). Life expectancy, economic inequality, homicide, and reproductive timing in Chicago neighbourhoods. British Medical Journal, 314(7089), 1271-1274.

12 thoughts on “More on Buller and Evolutionary Psychology”

  1. Great series of posts. I'd like to make a few small comments.

    1 as far as I remember the tables in Buller's chapter on Daly & Wilson give inconsistent figures.
    2 When one takes thief plausible figures in this chapter, one finds a small but real Cindirella effect.
    3. Buller mischaracterizes Buss's hypothesis about jealousy.

    These 3 examples reveal what is theosf frustrating aspect of the book: a lack of impartiality, a desire to falsify EP instead of reaching a balanced evaluation of the strengths and weaknesses of the field. In my mind, the ltter would have been a more useful contribution. But of course you would not attract as much attention.

    4. I take your point about the evolution of dimorphism. But still the essential point is that between the things we know for sure and things we don't know at all, there is a range of more or less plausible hypotheses about the evolution of human behavior. And there is no reason not to use these hypotheses heuristically to develop hypotheses about the components of human psychology.
    5 I agree with your point about human nature. I have a small paper on my website on this topic.
    6 I also agree with the point made in the other post that EPists have underestimated how difficult it is to provide good evidence that something is an adaptation.
    6 finally, I feel that the controversies between different traditions within the evolutionary aproach to human behavior is really a remnant from the 1990s and that it should be in part analyzed from a sociological point of view as some kind of turf war between different group of scientists competing for jobs, grants etc.

  2. One of the major problems I have with EP is that it is too simplistic. Behaviors are much too complicated to be programmed at the level of a single gene, and evolution has to be able to exert selection influence at the single gene level.

    In the case of infanticide, there are (perhaps contrived) circumstances where the most "evolutionarily advantageous" action might be to kill your child. There may be situations such as Sophie's choice, where a parent has to choose which child lives and which child does not. An offspring must be supplied with sufficient nutrients from conception to when the offspring becomes self-sustaining. If there are insufficient food supplies to sustain two children, but sufficient food to sustain one, the "evolutionarily correct" response is to kill one child and raise the other to maturity rather than let both children die.

    Over evolutionary time, the average number of children each parent had survive and reproduce was 2. If the number was less than 2, humans would have gone extinct, if the number was more than 2 the population would have reached levels that we know did not happen. Over evolutionary time a capacity for a behavior doesn't need to increase the average number of reproducing children very much to have a large impact on the population after thousands of generations. If a trait increases the survival by 5%, from 2 to 2.1, or one extra child every 5 generations, then over 1000 generation the trait becomes universal in the population. (1.05^1000 = 10^21) What this arithmetic also tells us is that there is no trait that has increased the average number of reproducing children to 2.1 over 1000 generations because the human population never reached 10^21.

    Evolution will tend to increase a trait until the marginal increase in reproductive success matches the marginal decrease in reproductive success. In other words, evolution will increase a trait until the sum of non-reproduction from not enough of the trait and non-reproduction from too much of the trait is minimized. An example I like to use is anaphylaxis. Is anaphylaxis a genetic disorder? Are deaths due to anaphylaxis the result of a genetic defect? No, evolution has configured the immune system to minimize the sum of deaths from infection due to an insufficient immune response and deaths from anaphylaxis or septic shock from too much of an immune response. The capacity to mount such a strong immune response that it might kill you may save you from an infection that would kill individuals with lesser immune systems.

    One of the major deficiencies of pop EP is to consider a very limited range of behaviors under a very limited range of circumstances; usually modest and relatively insignificant perturbations of modern life which is completely different than the conditions that occurred over evolutionary time (which are mostly unknown and to many unimaginable). The circumstances that are most important are life-threatening circumstances. Those circumstances cannot be tested on humans in the laboratory. People have difficulty imagining them because to imagine them invokes the mental state where such behaviors are doable, and many of those behaviors are extremely maladaptive except under near death circumstances (i.e. infanticide under lactation induced metabolic stress).

    What we should be considering are not behaviors, but rather capacities for behaviors. If a behavior is only adaptive under rare circumstances, but is highly adaptive under those rare circumstances (such as the ability to make a Sophie's choice under conditions of severe famine every 200 years), pop EP can't consider it no matter how important it might be in the "wild".

    Essentially all behaviors have a dynamic range. The genotype needs to generate a phenotype through the process of development that can support the necessary dynamic range of that behavior and choose the correct behavior at the correct time depending on circumstances. Systems comprising a few coupled non-linear interactions are chaotic and are completely unpredictable. That is much more complicated than EP has the capacity to consider.

    I think the same problem holds for physiology. Researchers are considering too narrow a physiological range to be "normal", and everything outside that range to be "pathological". Is anaphylaxis pathological? If you have bacteria floating around in your blood stream, and don't have access to effective antibiotics, you want an extremely strong immune system response because that is the only thing that might save your life. An X% chance of death from anaphylaxis is acceptable if it gives you a 100-X% chance of living because without it your chance of death is 100%.

    (In case it wasn't obvious, my comment on the evolution of manualist behavior was meant to be humor following the heuristic: ask a silly question, get a silly answer.)

  3. Honestly, my reading of your comment is one of fairly orthodox evolutionary psychology. Specifically, the statement "What we should be considering are not behaviors, but rather capacities for behaviors" could have easily come out of Tooby & Cosmides (1989).

    Your point about extrema of environmental conditions is well taken. This is why I think it is especially important to understand the general calculus of risk employed by people in fitness-critical domains (e.g., foraging, reproductive decision-making, self-preservation). Pen-and-paper tests of such proclivities are, indeed, useless and I have concerns about even the interesting experimental approaches made popular by the current wave of MIT experimental macroeconomists and used by evolutionary anthropologists such as Richard McElreath, Lawrence Kuznar, and Joe Henrich (though these are much, much better in general).

    I am all about demographic considerations. This is one of my big beefs with EP. According to some of the foundational documents of the field (the exchange that occurred in Ethology and Sociobiology in the late 80s/early 90s starting with Symons's "Critique of Darwinian Anthropology"), demographic considerations (i.e., measures fitness) are irrelevant to understanding adaptation -- and may even be misleading.

    I do think that your demographic arguments are a little simplistic however. In a variable environment, the rate of increase of the average life history can be positive but the realized rate of increase zero or even negative depending upon the covariance structure of (1) the life history itself and (2) the sequence of environments.

    Glad to learn the last comment was a joke...

  4. I feel that the third commentnin histhread is characteristic of what is wrong with much of the criticisms of EP: it is simply misinformed and fights strawmen.

    1. I doubt many EPists believe that "behavior are programmer at the level of a single gene."

    2. EP insists that cognitive mechanisms and not behavior are the proper level of analysis.

    3. Daly and Wilson do not assert that a disposition to kill stepchidren was selected for. Rather it is meant to be a by-product of our love for the children we believe to be our own.


  5. I'm not sure what the third comment is.

    The gene-behavior mapping is reminiscent of Marshall Sahlins's lame (1976) critique of sociobiology, "The Use and Abuse of Biology," which was really a horrid mis-characterization of what the early sociobiologists really thought.

    Surely, no one really believes that any behavior is encoded by a single gene! How the different sub-disciplines of human evolutionary biology deal with genetics -- or really fail to deal -- is very interesting to me. I think that the difficulty of reconciling genic-level explanations (e.g., Hamiltonian kin selection) with complex behavior may have contributed to the schisms between EP, HBE, and (for lack of a better term), dual inheritance theory.

    Daly and Wilson are probably seen as easy targets for would-be critics because, to someone who hasn't read their work or seen them speak, their research topics probably sound horrific. Of course, as you note, what they are really saying is actually quite different from what an uninformed first impression would be.

  6. I don’t consider myself very well informed on the literature of EP and don’t really think much in terms of psychology, rather I tend to think in terms of physiology and how physiology epigenetically programs organisms to have a certain phenotype.

    I see the problems of EP more as mentioned in the post, imagining what conditions were like in the Pleistocene, then trying to figure out what that means in terms of human psychology. I see that as bassackwards. We don’t know what the conditions were like in the Pleistocene or earlier. We know what human psychology is like now (far better than we will ever know what the Pleistocene or human ancestors in the Pleistocene were like).

    I think it is legitimate to take present human physiology and try and understand under what circumstances that physiology would be sufficiently adaptive for it to have been selected for. I think this approach is under utilized because there is a default to pathologize what ever has adverse effects not to try and understand the circumstances where those adverse effects might be advantageous.

    For example I see what is called the cycle of violence as an adaptive feature that (probably) all humans are capable of exhibiting. I see it as simply changing the threshold at which an individual initiates violence depending on how much violence that individual has experienced. I don’t think that an EP argument is needed because the “cycle of violence” is pretty well established as something that affects the life course of humans. If we can observe something like the “cycle of violence”, then there must be physiology that transduces the signal of “exposure to violence” into “more violent phenotype”.

    We don’t know the physiology that mediates the cycle of violence. There has been some work on monoamine oxidase A, but I think that is too simplistic. That was what I was thinking about in terms of one gene affecting behavior. There is significant thought that MAOA status does affect behavior. I think that MAOA is likely only working on concert with other pathways and that it is a threshold effect. That is that the threshold for invoking complex behaviors (such as initiating violence) is changed by a change in MAOA levels.

    Essentially all organs are known to be epigenetically programmed in utero, including the heart, liver, vasculature, endocrine system, etc. It would be beyond preposterous to suppose that the most important organ, the brain was not epigenetically programmed in utero even if we don’t understand the mechanisms.

    What I am working on is basal nitric oxide, which I feel is one of the signals that does transduce the state of stress of an organism into short term and long term adaptation, including epigenetic programming in utero and across the human life span.

  7. I couldn't agree more with your comment on the problematic pathologizing of physiological diversity. This has been a rallying cry for Peter Ellison and his crew at Harvard who have been working on the natural variation in human ovarian function for years now.

    There has been a great deal of work recently on so-called fetal programming and its possible consequences in human evolution. My friend Chris Kuzawa at Northwestern has done some nice work on the implications of fetal programming for growth and later metabolic disease.

    MAOA is an interesting case as well. Epigenetics is one of the frontiers of evolutionary biology and it greatly complicates our understanding of inheritance, ontogeny, plasticity, and adaptation. There is an interesting paper by Massimo Pigliucci in which he expounds his ideas for the possibility of a new "modern synthesis" in evolutionary biology.


    Ellison, P. T. 1994. Advances in Human Reproductive Ecology. Annual Review of Anthropology. 23:255-275.

    Kuzawa, CW, 2007. Developmental origins of life history strategy: growth, productivity and reproduction. American Journal of Human Biology. 19(5):654-61.

    Pigliucci, M. 2007. Do We Need an Extended Evolutionary Synthesis? Evolution. 61 (12):2743–2749.

  8. I'm enjoying your series...

    Something that critics sometimes overlook: Tooby and Cosmides (and others) have argued repeatedly that the reconstruction of the EEA is meant to be heuristic. That is, it's useful as a starting point in generating hypotheses (what Reichenbach would have called "the context of discovery") which are then tested with psychological experiments. EP, however, is not evolutionary biology: it does not attempt to reconstruct the exact sequence of selective pressures, random events and much else that caused our current suite of psychological adaptations. The aim is to understand current psychology, not reconstruct evolutionary history. The features Pleistocene is therefore irrelevant to the "context of justification".

    I do agree, however, that many evolutionary psychological experiments are badly done. (If you want to have a hernia, read "Music and Dance as a Signalling System"). But that's a problem we can do something about...

  9. Pingback: More on Diamond

Leave a Reply

Your email address will not be published. Required fields are marked *