Has Evolutionary Psychology Been Demolished? A Review of Buller, Chapter 4

Before I begin, I need to give a brief explanation of terminology (if I don't, I can guarantee you that Razib of Gene Expression will note that I should have). There are two senses of the label "evolutionary psychology," and since Buller is only addressing one, I will be as well. The first, broader sense, includes any psychology that takes evolution seriously, or includes evolutionary considerations in its explanations of psychological phenomena. The second, which is generally associated with Leda Cosmides, John Tooby, David Buss, and Steven Pinker, includes only those evolutionary psychologists who adhere to a certain set of tenets, including massive modularity, assumptions about the "environment of evolutionary adaptedness" (or the EEA), and a belief that the modern mind evolved in the EEA. To distinguish this second sense from the first, I will use Buller's convention or referring to it with capital letters (Evolutionary Psychology, Evolutionary Psychologists), and using only lowercase letters for the first sense.

If you've read through my archives (I can't imagine anyone has), or have been reading this blog for a while, you probably know that I am no fan of Evolutionary Psychology (EP). I'm not alone in feeling that way. Most cognitive scientists aren't fans. And we have good reasons not to be.

I'm also not a fan of attacking the scholarship of others without providing at least some semblance of a critique (in fact, I feel bad about doing just that in the previous post, even if everything I said is true, and I've offered more than a semblance of a critique of Lakoff in past posts). It's not always easy to avoid doing that where EP is concerned, as it stirs up a lot of emotions among cognitive scientists. To be honest, I've been guilty of some pot shots towards EP and its adherents in the past. It doesn't help that I harbor a deep personal animosity towards one of EP's most prominent figures (I won't mention any names, but you can ride him to work), an animosity built largely around my dislike for his research. However, even he is a very bright guy, and he works hard at what he does, as do most Evolutionary Psychologists. So, as scholars, I think we owe it to them, and any other scholars, to give reasons for any harsh words we may utter about their work. That's why I took Brian Leiter to task for his post entitled "Evolutionary Psychology Demolished." That post referenced a bad book review of a book that it's not clear Dr. Leiter has read, on a topic it's not clear Dr. Leiter is qualified to address (he may be, but he's given no indication that he is). But even my attack on Dr. Leiter's post made me feel a bit guilty, so I feel like I should say something a bit more substantial about the book reviewed in the article he linked.

It just so happens that I have read the book. My general policy is not to read books on EP, because in books, you can say just about anything you want, and Evolutionary Psychologists tend to do that more than most. Even carefully reviewed academic books are subjected to standards far below those of peer reviewed journals. I'm particularly reticent to read books on EP by philosophers (after reading the God-awful Furnishing the Mind, which I also got for free, I'm wary of any books on cognitive psychology by philosophers). But I got a copy of Buller's Adapting Minds for free, and I had heard about it through the grape vine, so I went against my general policy and read the book. It turns out that on the whole, it's a very good book. The first three chapters are excellent, and the chapters on Mating, Marriage, and Parenthood (chapters 5, 6, and 7) are very good, too.

But those areas of EP are its weakest. They are the areas with very little experimental work, and a heavy reliance on incredibly large and poorly controlled surveys. It's true that they are the most popular areas among the general public, but that's because they're the sexiest (sometimes literally). So, if you want to say that EP has been demolished, you have to do more than attack its weakest links. To do that, you have to critique its strongest points. And it goes without saying that from a methodological and theoretical standpoint, EP's strongest point is the work of Leda Cosmides, John Tooby, and their colleagues on social exchange. They use actual experiments, rather than surveys, and they've produced a great deal of challenging data. Buller addresses their work in Chapter 4 (which he co-wrote with Valerie Gray Hardcastle), and in my mind, that is the book's weakest chapter. If his weakest chapter is a critique of EP's best research, then it may be difficult to say that Buller has demolished EP. Furthermore, since Dr. Leiter wants "law-and-economics" folk to read this book to see how shoddy EP is as science, Chapter 4 is the most relevant to his purposes. It is the work on social exchange that has been seen as directly applicable to economics. Cosmides and Tooby have even written a paper on applications of EP to economics. So, if EP is to be demolished even for Dr. Leiter's purposes, Chapter 4 has to do the bulk of the work.

To see why it's important to critique the work of Cosmides and Tooby in order to show how misguided EP really is, it's important to understand the goals of EP, and how Cosmides and Tooby's work is integral to the accomplishment of those goals. Dan Sperber and Vittorio Girotto put it this way¹:

Evolutionary psychology--in its ambitious version well formulated by Cosmides and Tooby (e.g., Cosmides & Tooby 1987, Tooby & Cosmides 1992)--will succeed to the extent that it causes cognitive psychologists to rethink central aspects of human cognition in an evolutionary perspective, to the extent, that is, that psychology in general becomes evolutionary. The human species is exceptional by its massive investment in cognition, and in forms of cognitive activity--language, metarepresentation, abstract thinking--that are as unique to humans as echolocation is unique to bats. The promise of evolutionary psychology is thus to help explain not just traits of human psychology that are homologous to those of many other species, but also traits of human psychology that are genuinely exceptional and that in turn help explain the exceptional character of human culture and ecology. (p. 197)

However, most of the work done in evolutionary psychology so far is on aspects of human psychology that are not specifically human except in their details. Showing, for instance, how human preferences in mate choice are fine-tuned in the way the theory of evolution would predict is of great interest (see e.g., Buss 1994) but it can be done on the basis of a relatively shallow psychology.

Thus, Sperber and Girotto argue, if EP is to accomplish its goals, work like that of Cosmides and Tooby (CT from now on) on specifically cognitive, and specifically human characteristics must be at the forefront. I would further argue that the primacy of work like that of CT is paramount if EP is to accomplish the goal of making psychology evolutionary. If psychology were to become evolutionary in the manner of work by people like Buss, it would truly be a regression, rather than a progression, in psychological science. Even the area most adjacent to the EP research on mating and parenting, social psychology, has been using experimental methods for nearly a century. Thus, whatever theoretical gains the evolutionization (yes, I made that word up) of psychology might bring us, it would cancel them out with losses in methodological rigor. I've already seen this happening within the realm of EP itself. I've attended many job talks by graduate students and post docs. Those that presented experimental research were so undertrained in experimental methods that, when they presented their work, everyone in the audience who was not an Evolutionary Psychologist (and perhaps some who were) could do little more than roll their eyes. And these were the best applicants to the best EP program in the country, and thus were the best that EP's next generation had to offer. Thus, from both a theoretical and methodological standpoint, the work of CT, and their intellectual progeny, is integral to the aims of EP itself.

OK, enough of all that preliminary stuff. On to Chapter 4 itself. The chapter is not just about the work of Cosmides and Tooby, but about modularity in general (it also discusses theory of mind, which I'll get to at the end of this post.. It begins quite well, arguing that massive innate modularity is unlikely, if not unlikely, from the perspective of developmental neurobiology. He gives a detailed description of the development of the brain in the human fetus and soon after birth, and argues that the facts of this development argue strongly against the existence of a large number of innate cognitive modules. He also addresses some of the arguments for massive modularity, such as the one which Donald Symons once characterized by saying,

It is no more probably that some sort of general-purpose brain/mind mechanism could solve all of the behavioral problems an organism faces (find food, choose mate, select a habitat, etc.) than it is that some sort of general-purpose organ could perform all the physiological functions (pump blood, digest food, nourish an embryo, etc.).²

Buller points out that neural plasticity, and the way that it allows brain circuits to be shaped by experience, allows the brain to evolve as a general-purpose organ. In fact, given the dynamic nature of the environment of our evolutionary ancestors, neural plasticity is likely a more adaptive solution than massive modularity, which restricts the brain to a specific set of tasks. It's also a more likely solution, since modules would quickly become obsolete as the environment changed. He also addresses Pinker's argument for modularity, which compares all of cognition to language in the first section, noting that language is likely unusually complex, cognitively, and that we do many incredibly complex things that we obviously do not have modules for (like playing chess). Language is likely a bad analogy, then.

Overall, the theoretical and developmental discussion of massive modularity is very good. However, after that, he turns to the evidence for modularity, which consists almost entirely of CT's work on social exchange. Note, first, that while it may be that the brain is not massively modular (it isn't!), it is still possible that there are a few individual modules for specific cognitive tasks that did not change over long periods of our evolutionary history. It could be that a cheater detection module did evolve, even if it's not accompanied by a lot of other domain-specific modules. Thus, Buller's arguments against the massive modularity thesis of EP can't be used as a critique of social exchange theory³. Also, as Buller himself notes, theoretical arguments are great, but they only go as far as the empirical data, and as any good scientist will tell you, no matter how foolproof the logic of a theory sounds, if in the face of that theory the data remains recalcitrant, we have to let it go. If the data is most consistent with the existence of a cheater detection module like the one that social exchange theory posits, even though theoretical arguments against the existence of cognitive modules make a lot of sense, we, as scientists, are going to have to go with social exchange theory.

Buller devotes the bulk of chapter 4 to examining the evidence for a cheater detection module. Before I get to his critique of that evidence, though, I should probably summarize social exchange theory and the evidence for a cheater detection module. I've done it before, but since I'm too lazy to go back and find the posts, I can't expect anyone to actually read them.

In his classic 1971 paper titled "The evolution of reciprocal altruism,"⁴ Robert Trivers argued that in order for reciprocal altruism to evolve, two other capabilities must have been present: the ability to detect cheaters, and the tendency to reward reciprocators. This insight has been confirmed by several game-theoretic simulations⁵. Drawing on these arguments, Cosmides developed social exchange theory, which is designed to explain the evolution of reciprocal altruism in humans. As part of this theory, she hypothesized that humans have evolved a cheater-detection module that allows them to detect those who do take benefits without reciprocating. This cheater-detection mechanism would allow humans to satisfy Trivers' two requirements for the evolution of reciprocal altruism.

There is evidence of a cheater detection algorithm in several nonhuman species. The most frequently cited example (perhaps because it's just plain cool) comes from vampire bats. Apparently vampire bats will regurgitate excess blood so that other members of their group can eat it. Furthermore, they can tell when another bat is not regurgitating excess food, by looking at its fat belly. A fat belly and no regurgitation equals cheating, and bats will not regurgitate their food for cheaters⁶. In humans, however, it's a bit more difficult to detect cheating (a fact which is unfortunate for researchers studying cheater detection). According to CT, the complexity of detecting cheaters means that we must have evolved a mechanism specifically designed to do so. This mechanism would recognize social exchange situations, and activate the appropriate reasoning principles in order to detect any cheaters.

The evidence for the existence of such a module comes almost exclusively from experiments using multiple variants of a single task, the Wason selection task (see this post for my short description of this task). There are two relevant versions of this task, the descriptive and the deontic. The descriptive task uses an indicative conditional of the form "If p, then q," and the deontic form uses a deontic conditional of the form "If you do p, then you must also do q," or some variant thereof. One of the classic findings from Wason task experiments is the "content effect," in which people perform quite poorly on the descriptive version, and quite well on the deontic version. Cosmides argues that the increased performance in the deontic version is due to the fact that it activates the cheater-detection module, and thus allows people to use domain-specific reasoning mechanisms to solve the task, whereas the descriptive version requires the use of domain-general reasoning mechanisms.

It is to this argument which Buller replies in Chapter 4. He offers two criticisms of this research, one of which is a "big picture" criticism, and the other of which is a "devil in the details" criticism. Here is his description of the first, from pp. 171-172:

Cosmides derives the hypothesis of a cheater-detection module from Trivers's argument that reciprocal altruism is evolutionarily unstable unless parties to reciprocal exchanges of benefits have the ability to detect when someone is taking a benefit from them without providing them with a benefit in return. Reciprocal altruism concerns what Cosmides calls social exchanges, which occur when two individuals perform acts that benefit one another at a cost to each. Social exchanges are thus relations between two individuals, and cheating in social exchange involves benefiting from an act performed by another individual without performing an act that benefits the individual. Virtually all of the experimental results that purportedly provide evidence of a cheater-detection module, however, derive from selection tasks involving what Cosmides and Tooby call social contracts, which is a much broader class of phenomena than the class of social exchanges. [Emphasis in original.]

But wait, that's not how CT define social exchange. Here is how they define it⁷:

[A] situation in which, in order to be entitled to receive a benefit from another individual or group, an individual is obligated to satisfy a requirement of some kind (often, but not necessarily, at some cost to him- or herself). Those who are rationing access to the benefit impose the requirement because its satisfaction creates a situation that benefits them.

This definition is completely consistent with Trivers' analysis of reciprocal altruism, and completely consistent with the problems used to detect the cheater-mechanism in CT's experiments. If that doesn't seem obvious from the "beer" example in the post I linked above, here are some more to help make it clear:

• If you go canoeing on the lake, then you have to have a clean bunk house.
• If a man eats cassava root, then he must have a tattoo on his face. (Accompanied by a story in which cassava root is an aphrodisiac, and tattoos are a sign that a man is married.)
• If I give you some potatoes, then you must give me some corn.

In fact, the last of these, which was used in one of the papers that Buller cites, is perfectly consistent with Buller's definition as well. If CT are actually testing the definition of social exchange that they offer, and derived from Trivers, and even use problems that are consistent with Buller's definition, then why does Buller offer this groundless criticism? I have no idea, but even more baffling than that is his second criticism.

Buller's second criticism, which he spends 16 pages describing in reference to several different experimental results, is simple. He points out, correctly, that the logic of an indicative conditional is different from that of a deontic conditional. For an indicative conditional, ~q is the negation of the consequent. However, for deontic conditionals, the equivalent of the "~q" card in the Wason task doesn't actually negate the consequent (such a negation would be something like, "doesn't have to have a tattoo on his face," for the cassava root example above). Instead, it simply says that condition of the rule has not been obtained (e.g., "no tattoo"). So, solving the deontic version of the Wason task involves using a different set of logical principles. Buller argues that this explains the difference in performance in the descriptive and deontic versions (the "content effect,") and CT and their colleagues have never realized this.

Upon concluding these 16 pages, all anyone who is familiar with the literature on social exchange theory can do is say, "Huh?" Cosmides herself noted this in the 1989 paper in which she first described social exchange theory⁸. In fact, a debate has been raging for the last 16 years about whether the content effect can be explained by reference to a mechanism that uses the logical principles of deontic conditionals (Buller's explanation), or whether a cheater-detection mechanism is needed⁹. Sixteen years of debate, and Buller treats the issue like he is the first to notice the problem. In fact, CT believe that they have addressed this issue, by showing that social exchange deontic conditionals produce better performance on the Wason task than non-social exchange deontic conditionals (e.g., in Fiddick, et al. 2000, reference in footnote 7). Since Buller cites this paper, without acknowledging that they address his criticism, I can't believe that he's simply unaware of their answers to it. He doesn't cite, and therefore may not be aware of the neurological evidence that Cosmides and her colleagues have offered to demonstrate that social exchange and non-social exchange deontic conditionals are understood using different brain regions¹⁰. They showed that a patient with brain damage to an area of the brain associated with social reasoning was unable to perform social exchange versions of the Wason task, but performed quite well in non-social exchange deontic versions.

None of this is to say that the Wason tasks actually do provide evidence for the existence of a cheater-detection module. In fact, reasoning researchers and other cognitive scientists have been poking holes in this research for the better part of two decades. Most notably, Sperber and Girroto (e.g., their 2003 paper, cited in footnote 1) note that relevance theory, a theory that states that people use pragmatic cues to search for information that is relevant to their perspective, actually predicts CT's findings, along with another set of Wason task findings that social exchange theory cannot predict. They show that by manipulating perspective and the availability of relevance cues, you can make people perform well or poorly in deontic and descriptive versions of the Wason task. The domain of the problem is irrelevant. Thus, they argue convincingly that the Wason selection task cannot provide evidence for the existence of a cheater-detection module. Sperber and Girroto therefore arrive at the same conclusion that Buller does, but unlike Buller, they do so by actually addressing the claims and research of Cosmides and Tooby.

The neurological data is problematic as well, but Buller doesn't even address it, so readers of his book would be unaware of the problems. They only had an n of 1, which means they were unable to show a double dissociation, and his brain damage was to the medial orbitofrontal cortex and anterior temporal cortex (both bilateral), two regions which are associated with a wide range of social reasoning abilities. Their Wason task is unable to dissociate cheater-detection from general social reasoning, and is therefore not evidence for their theory.

Before I wrap up this review of chapter 4, I should say something about his short discussion of the research on theory of mind. I don't really want to, because I'm genuinely embarrassed for Buller. If he intentionally or unintentionally ignored large portions of the literature on social exchange, he at least made an attempt to look like he was familiar with that literature. In his discussion of theory of mind, however, he makes no such effort. In fact, while there is a huge body of literature, written almost exclusively by non-Evolutionary Psychologists, which demonstrates fairly conclusively that humans have at least some innate theory of mind abilities (i.e., abilities to reason about the thoughts and desires of others), Buller only addresses one finding. He doesn't address the infant gaze-following literature, the language-learning literature, or many of the developmental reasoning tasks showing that from early in infancy, children reason about animate objects differently than they reason about inanimate objects. The finding that he does address, that autistic children tend to perform worse on false belief tasks than non-autistic children, has been criticized extensively already, because the false belief task may not detect theory of mind at all. Buller repeats this, and then argues baselessly that this indicates that there are no innate theory-of-mind mechanisms.

So, if you've read this far, you've probably realized by now that Buller doesn't demolish Evolutionary Psychology. In fact, while he does a very good job of showing just how shoddy EP's shoddiest research, that on mating and parenting, really is, he doesn't even touch EP's strongest research. Of course, even this strongest research has been thoroughly dealt with in the literature. For most cognitive scientists, social exchange theory is either falsified or evidenceless. Even those most willing to accept social exchange theory must admit that the reliance on a single experimental paradigm which has been shown, over the last 30 years, to cause several difficulties in interpretation, is problematic. But one would never know this from Buller's book. All we get, in chapter 4, is baseless criticisms. It might be fair to say that Buller failed to demolish a theory that was already demolished to begin with.

So, while I would definitely recommend this book to the general public, so that they might read the excellent chapters 1-3, and the good chapters 5-7, I wouldn't recommend it to anyone who's interested in the scholarly debate of these issues. I also wouldn't recommend it to economists and legal scholars. While the general public is most interested in what EP has to say about sex, they tend to be interested in what it has to say about social exchange, and since Buller's book is so far off on the social exchange research, it would be worthless to them.

¹Sperber, D. & Girotto, V. (2003). Does the selection task detect cheater-detection? In Fitness, J. & Sterelny, K. (Eds.). From Mating to Mentality. Macquarie Monographs in Cognitive Science, Psychology Press.
²Symons, D. (1992). On the use and misuse of Darwinism in the study of human behavior. In Barkow, J.H., Cosmides, L., and Tooby, J. (eds.), The Adapted Mind: Evolutionary Psychology and the Generation of Culture (pp. 137-159). New York: Oxford University Press.
³Of course, Cosmides and Tooby also tend to treat evidence of a cheater detection module as evidence of massive modularity. However, evidence for one module is not evidence of massive modularity. Buller successfully argues that point when he counters Pinker's analogy to the language module.
⁴ Trivers, R.L. (1971). The evolution of reciprocal altruism. Quarterly Review of Biology, 46, 29-47.
⁵Axelrod, R. & Hamilton, W.D. (1981). The evolution of cooperation. Science, 211, 1390-1396.
⁶Wilkinson, G.S. (1984). Reciprocal food sharing in the vampire bat. Nature, 308, 181-184.
⁷Fiddick, L., Cosmides, L., & Tooby, J. (2000). No interpretation without representation: The role of domain-specific representations and inferences in the Wason selection task. Cognition, 77, 1-79.
⁸Cosmides, L. (1989). The logic of social exchange: Has natural selection shaped how humans reason? Studies with the Wason selection task. Cognition, 31, 187-276.
⁹See e.g., Cheng, P., & Holyoak, K. (1989). On the natural selection of reasoning theories. Cognition, 33, 285-313.
¹⁰Stone, V.E., Cosmides, L., Tooby, J., Kroll, N., & Knight, R.T. (2002). Selective impairment of reasoning about social exchange in a patient with bilateral limbic system damage. Proceedings of the National Academy of Sciences, 99(17), 11531-11536.