Michael McCullough: "Two Cheers For Falsification"

Michael McCullough: "Two Cheers For Falsification"

HeadCon '14
Michael McCullough [11.18.14]

What I want to do today is raise one cheer for falsification, maybe two cheers for falsification. Maybe it’s not philosophical falsificationism I’m calling for, but maybe something more like methodological falsificationism. It has an important role to play in theory development that maybe we have turned our backs on in some areas of this racket we’re in, particularly the part of it that I do—Ev Psych—more than we should have.

[43:37 minutes]

MICHAEL MCCULLOUGH is Director, Evolution and Human Behavior Laboratory, Professor of Psychology, Cooper Fellow, University of Miami; Author, Beyond Revenge. Michael McCullough's Edge Bio page


I’m Mike McCullough. I’m a psychologist at the University of Miami. I want to talk a little bit about some thoughts I’ve been entertaining about falsification, and particularly its place in my bailiwick, which is Ev Psych.

Most of you, when you think about falsification, think about Karl Popper, who had this idea that is pretty compelling, which is that we can never have positive evidence for a hypothesis. Hypotheses give us predictions about how the world should be ordered. What we like to do is take the data from the world and then make inferences about whether the hypothesis is true. This is a problematic kind of reasoning. It’s not valid reasoning. As I'm sure many of you know, he suggested that there is a valid reasoning we can use for making inferences about the truth value of hypotheses from observations, but the way to do that is to look for predictions that are falsified in the world.


Your hypothesis says the world should be structured this way, and when you find evidence that it’s not structured that way you’re then in a position to make a valid conclusion about how the world is structured. More specifically, you’re in a position to say this is a valid conclusion about how the world is not structured. That's valid reasoning. Modus Tollens works.
The thing is, science has done pretty good with basic induction. Most scientists feel this way anyway. Taking observations about the world that are true and then making inferences about hypotheses has been a pretty decent way to do science. For most practicing scientists, affirming the consequent looks like a reasonable way to approach our jobs.
What I want to do today is raise one cheer for falsification, maybe two cheers for falsification. Maybe it’s not philosophical falsificationism I’m calling for, but maybe something more like methodological falsificationism. It has an important role to play in theory development that maybe we have turned our backs on in some areas of this racket we’re in, particularly the part of it that I do—Ev Psych—more than we should have.
In general, we don’t do much falsifying. As an empirical matter this is old news. Everyone knows that most of what gets published in journals is supportive of hyptheses: "The prediction supported the hypotheses." There are lots of methodological, sociological, and cultural reasons for that. I’m not that interested in that other than to say we are comfortable with doing confirmatory research, and there may be some non-methodological reasons why we like that. One is that falsification is just hard. We don’t do it very well. Think about something like the basic Wason selection task. We know from that work by Wason that when people have cards with even or odd numbers on them and the other sides of the cards have either the color red or the color brown, and they’re asked to determine whether a rule is being followed—if the number is even, it must have a red color on the other side—people don’t choose well. They don’t choose optimally. They’re good at saying, "We should turn over the even cards and see if there’s red to see if that condition is being fulfilled," but then we want to turn over red cards, and see if there are even numbers. We don’t do Modus Tollens very well. It just doesn’t come naturally in these basic problems that are more like scientific problems.
There are two other reasons why we are not as good as we should be at falsification. One is that we tend to have weak methodological beliefs. This has been beaten to death over the past three years. We tend to imagine lots of ways in which the methods might not have been good enough to provide a decent falsification of the hypothesis. I’m more interested in the obverse side of that, particularly as it concerns EP, which is we tend to have very strong theoretical beliefs. This hits home in Ev Psych, where I live, because evolutionary psychology is working on the theory of natural selection, which is one of the strongest theories in all of science because it has just a couple of basic ideas that you get very powerful axioms from.
If you assume you’ve got these replicating units in the world—let’s just create a planet and put these things on it that have found a way to take material from around them and hack off copies of themselves, that’s all you need to get evolution by natural selection. If you want to slip in there that the replication has to be imperfect, you can put that in as a second condition, but that’s life. Everything is imperfect.
Once you’ve got this world with replicators in it, we know exactly what’s going to happen. They’re going to go through time cloaking themselves in these nifty design features that increase their reproductive rates. It’s axiomatic; we don’t need to prove this in any sense. It’s an algorithmic process that’s going to happen.
As a function of this building of design features, critters are going to come to look like they were built to optimize something. That’s one thing we can take as canon. We don’t know quite what they’re optimizing until we get to Bill Hamilton, and then finally we see what it is these design features are, in a sense, for—to maximize inclusive fitness.
This is a strong theory. It allows you to predict that there are features in the world that have purpose to them. The purpose is increasing inclusive fitness. Here's the problem: Once you get to that point there are no more axioms to be had. You don’t know what design features to expect natural selection to provide you with. Thomas Nagel had us wondering why don’t pigs have wings. It seems like from a certain way you could think about what’s good for pigs, "Gosh—they should have had wings by now! Wouldn’t wings be great? They could fly, they could skip over empty troughs, and they could go find the muddiest mud holes. It would be fantastic."
Why don’t they have them? The reason they don’t have them is many-fold. There were powerful constraints against it; every gene that would have helped to build wingedness was downward from an optimal peak. All of those reasons they don’t have wings are highly contingent on natural history, highly contingent on phylogenetic constraints. As soon as we start trying to predict the design features that are going to be out there, we’re stuck. Natural selection only takes us so far as a theory. It cannot tell you, for any organism, what you're going to find. You know what the things you’re going to find are for with reasonable certainty, but you can't postdictively predict them.
Ev Psych is a big idea, and big ideas get criticized for good reason—they’re big ideas. But the criticism of some of it has crystallized around a few substantive areas of research. It's been interesting to think about what’s going on in these areas that make them such lightning rods for criticism. Some of it has to do with falsification or maybe even a lack of appetite for falsification.
We’re going to have to discover adaptations ultimately, not deduce them from first principles. One of the areas where this has become a PR problem for Ev Psych is in an area of research on how women’s behavior varies as a function of their ovulatory state. There’s a lot of work coming out these days based on the idea that when women are not fertile they have a set of preferences, a set of emotions, a set of motivations, and a set of behaviors that are quite different from when they’re ovulating. They prefer to wear red when they’re ovulating. They find certain features in men more sexy, attractive, and worthy of pursuit when they’re ovulating than when they’re not ovulating.
One of these hypotheses behind this idea—and it’s in the news as we speak, the Internet is burning up with commentary on this—is a hypothesis called the ovulatory shift hypothesis. The hypothesis is that women who are not fertile have a different reproductive agenda than women who are currently in the fertile part of their ovulatory cycle. The idea goes something like this: When you’re not fertile, what interests you in men—among other things, but particularly these things—are traits that might make them valuable, long-term mates. When you’re ovulating, what you shift your focus toward are indicators that that man has got good genes that are free of a heady mutation load, and that the man is in good condition.
The idea is that you want to get those good genes, so you’re looking for features in men that, ancestrally, would have been correlated with the possession of a low mutation load, got enough to eat while developing, were free of a lot of blunt force trauma as an adolescent, any features that would indicate that this is somebody who’s carrying around a decent set of modules for me to merge with my modules.
This could be right, this hypothesis. There is something I like about it. I find it a provocative idea. What I realized about a week ago is it’s not inevitably right. It’s not axiomatically right. There’s lots of ways we can think about natural selection having worked on women’s psychology that would have been inclusive fitness maximizing, but that would not have involved an ovulatory shift in preferences, behaviors, emotions, and motivations.
One thing you can imagine is that the same genes that make you not so attractive as a man—when there are mutations that come out in reducing your body symmetry, your attractiveness, or masculinity; What if it’s the case that those same defects, those same genetic problems also make you a worse dad? Why aren’t the dad modules equally run down by these genetic defects that are piling up in a visible way in reducing your condition, or your masculinity, or your symmetry? That's one possibility that would say maybe there is no ovulatory shift. Maybe women want and find attractive in men the same things throughout their entire ovulatory cycle.
I don’t think these are outlandish possibilities. They are hypotheses that you should put on the table. Another one is maybe men were sufficient, but not necessary co-provisioners for children, ancestrally. There’s some decent evidence that if you’ve got another co-provisioner around, men might be facultative rather than obligatory co-parents. If that’s the case maybe what women want all the time is just the good genes. That’s what they’re interested in 24/7 because male parenting effort is not something that’s making a marginal, unique contribution to women’s inclusive fitness.
All to say, there are data to suggest that those hypotheses are not laughably, absurdly wrong. They're reasonable alternatives. Here's what is happening in some of these areas that have become so fractious, like this research on ovulatory cycle shifts. For those of you who don’t know what’s going on, there are a set of reviews and meta-analyses by some researchers that have looked at all of the evidence for the ovulatory shift hypothesis. They’ve meta-analyzed it very carefully, and on some measures of this shift in women’s preferences for certain traits in men in certain perceptual contexts, the data looked pretty supportive. They're not overwhelmingly supportive, but they’re supportive enough that you’ve got to give the hypothesis some serious thought.
There is another group of researchers who think that the literature is so beset by methodological problems that we ought to ignore it and start over with other studies. Then there’s a sub camp of that camp who says the hypothesis itself is fundamentally flawed. What has happened, in part, to make this so inflammatory is that some of the EP researchers who are working on this hypothesis—in a lot of different areas, not just in mate choice, but also in advertising one’s own qualities to prospective mates—maybe have come to feel that the hypothesis they’re supporting is the Darwin hypothesis. It’s the hypothesis that’s available that has to be defended if Darwin is right. If there’s not something profoundly wrong with the theory of natural selection, this inevitably has to be right. It has to be right in some way.

For people in that other camp who may have motivations to find problems with this literature, maybe they feel that the hypothesis is so offensive to what we understand about cultural variability and the problems that women have faced traditionally that we need to undo, that it’s become for them a cause that is a righteous one because they view it as the propagation of the mistreatment of women using the tools of science.
What I want to suggest is this thing can be unwound. I’m confident that the empirical matters of fact are going to get settled one day. What I want to suggest is that methodological falsificationism could have a huge role to play here for people who care about the theory of natural selection and applying it to understand human behavior. I’m going to go one step forward axiomatically and say there are some forced moves. Dennett talks about forced moves—things that are inevitable consequences of natural selection. You’ve got to have a skin. There’s got to be a place where you stop and the other organisms start. That’s a forced move. It’s going to happen to all replicating things. You have to know where your interests end and the next guy's begins. That’s a forced move.

Another forced move is you’ve got to get something to eat. Every organism has got to reverse entropy locally. For sexually reproducing species I think it is the case, there is a forced move, which is to get the best modules you can from your mate. We should expect the female ovulatory cycle to be well tuned to operating the way it should have ancestrally to maximize women’s inclusive fitness.
I’m going to go a little bit axiomatic on you in that respect. That’s the way we should expect that system to be designed, but I don’t know that we know enough about the initial conditions. I’m not so worried about available genetic variance, but I do worry about phylogenetic constraints, social ecology. There are a lot of possibilities on the table—I just made up those fun alternative hypotheses—that could be, if stated more seriously, reasonable alternatives.
What I would like to see is the next wave of research on that involve some evolutionarily interested, engaged researchers who take other hypotheses like those seriously as alternatives, so we cannot evaluate the fit of a certain model in absolute terms, but we can say which of the viable models fit these data better and which ox gets gored by the data. We should be in a position at the end of the day to take some hypotheses off the table through a greater psychological comfort with disconfirmation, with seeking out falsifying evidence. To do that comfortably for somebody who cares about using the tools of natural selection to understand human behavior, you’ve got to have multiple hypotheses on the table.
There are a couple of areas where this is happening in a healthy way. I’m not a mating guy. I mean I’m a mating guy, but I don’t study mating 24/7. Probably, I do, but not for pay. That's an area of what the putative functions of life after menopause is about. There’s not just a single hypothesis that’s being taken seriously by evolutionary demographers, psychologists, and modelers about why it is women live past their reproductive years. We’ve got at least two plausible hypotheses. One, which is the very famous one, is the grandmother effect. Women improve their inclusive fitness by withdrawing their reproductive effort from their own direct fitness maximization and then toward providing care to their daughters’ offspring. It’s a good hypothesis. It’s a winsome hypothesis. It’s not the only hypothesis.
There’s also a hypothesis that menopause came along to minimize parent-offspring conflict, which is a distinction with a difference. They lead in some ways to some interesting differences and predictions. If you ask me about them after the camera’s off I’ll tell you I’m not able to describe them to you in a whole lot of detail. That’s one area.
Whatever hypothesis gets gored at the end of the day, there’s going to be an evolutionary hypothesis still standing, and it might be a functional one. Gould told us to be pluralistic about all of the possible mechanisms that could generate design features—phenotypes. What I want to say is we need to be adaptive pluralists as well. We’ll do a lot better as a science with adaptive pluralism in the same way that we are multi-evolutionary force pluralists.
The other area where there’s some nice work going on that’s falsificationist is in the area of cooperation. Fifteen years ago there was an idea that Ernst Fehr championed and Herb Gintis and a few others, which suggested that humans are as nice as they are, as cooperative, as disposed to punish bad guys as they are because, through some group selection—either cultural group selection, or gene-culture evolution, or genetic group selection—we developed this propensity they called strong reciprocity. Because it relied on certain flavors of group selection that some evolutionary psychologists found objectionable, there was a lot of grumbling in the field for a lot of years. It took a while, but 7 to 10 years later people figured out how to put together some empirical horse races for some of the claims that would leave one ox gored. Somebody was going to get bloodied, some hypothesis was going to be damaged, and another one hopefully would be less damaged.
We are far from solving that problem, far from resolving those debates because it’s still true that theories die funeral by funeral in science. We at least know how to frame debates as empirical matters that ultimately could enable us to adjudicate hypotheses that give proper props to Darwin, give proper props to the only force we know in the universe that can build complex functional design, and also help us build a better science.


SARAH-JAYNE BLAKEMORE: Do these kinds of hypotheses—Ev Psych hypotheses and particularly the ovulation shift hypothesis—only apply to humans? Is there research on other species that form pair bonds, like prairie voles? Do the women prairie voles prefer ripped men when they’re ovulating?

MCCULLOUGH:  I don’t know. There is a lot of sexual choice.

BLAKEMORE: Isn’t that a key question? Whether it applies across species?

MCCULLOUGH:  It’s an interesting question, but it’s not decisive. It wouldn’t be decisive because they’re not humans. They don’t have our same history.

BLAKEMORE: In a way Ev Psych does just apply to humans. Well, humans are what you're interested in—human behavior and human psychology.

MCCULLOUGH:  Yes. It is an attempt to unite evolutionary biology and behavioral ecology more closely with psychology.

LAURIE SANTOS:  The problem that you described, about sticky hypotheses, is that folks don’t want to propose other ones (like this ovulatory shift hypothesis) in the face of a lot of folks who said that the data are not there. It stuck around, and folks haven’t proposed alternatives. But I don’t think the same has been true in cases of behavioral ecology, where you'd imagine the same constraints apply. Behavioral ecologists believe in inclusive fitness theory, they believe in sexual selection, yet when some life history variable does work for one individual species, ecologists propose a ton of other alternative hypotheses. So why in EP did we get stuck? That's an interesting sociological question because you don’t see it with folks who have exactly the same commitments about the meta-theories.

MCCULLOUGH:  You took the words right out of my mouth. One of the people who works with me as a primatologist, and when we were talking about this issue at a lab meeting she said, "This is just astonishing to me." She’s working with humans in the lab for the first time. She said exactly what you said. We see something unusual, we come up with three, four, or five hypotheses, potentially, and you see which of those hypotheses gets a better fit from the data, and you throw the worst ones away. It is interesting to me that these are sticky. I would love to know why that is. I wonder if there’s an interesting set of attractors that have to do with how we intuit about humans.

SANTOS: Another question I always have for EP is basically Sarah-Jayne's question. There are other species that have parts of our life history, which, if we feel strongly that some set of hypotheses should be true of humans, we should expect to those traits in them. But it depends on the specific hypothesis about some aspects of a species’ life history. If we have a hypothesis about mating and pair bonding species, then we should expect to find it in other pair bonding primates. But if our hypothesis is about being a species where men have access to resources and females don’t, we should look to a species like pair-bonding primates, and expect to see that trait there.

As an animal person it’s always felt like there’s a reluctance in EP to even speculate about what’s going on in animals for the same answer you just gave, which is: "We’re just interested in humans," but if you’re interested in the mechanisms that shape humans, then those same life history variables should be of interest broadly. It feels like they’re not.

MCCULLOUGH:  One example that is of interest to me in this particular debate is some researchers who have taken up the hypothesis that women advertise their ovulation through dress, which I find interesting. I don’t see a lot of evidence that they have evolved other, more direct ways to advertise their ovulation. In fact, I don’t know if they’ve evolved to conceal it, but they certainly don’t seem to have evolved much more straightforward ways to advertise it. Relative to things that they could be doing, if you take into account all of the interesting things primates have managed to do to advertise their ovulation.

I don’t mean to say don’t tell me about those other animals. Personally, I love that work.

SANTOS: It seems like that could be a useful way to do falsification in EP. In part because sometimes there are experiments that are hard to do with humans, but you can do in other animals. You can do stuff with the pair bonded prairie voles that IRB-wise is a little bit sketchy to get approved with undergrads.

DAVID PIZARRO:  Facebook would do it.

SANTOS:  Not Facebook, but OkCupid. It seems that other animals provide a useful possibility for falsification. One of the things I teach in my EP course is this idea that children should look more like their fathers at first. There is another evolutionary psych hypothesis that’s out claiming that males won’t murder children who look like them because they think they are the fathers of the kids. This claim makes a strong prediction in primates that when the females are mating with a bunch of dudes and it doesn’t matter who your kid looks like, you should see that those primate kids look less like their dads than human kids look like their dads. We should be able to do a comparative study. It seems like sometimes other species could form this sort of beneficial outgroup, and I’ve always been curious why EP folks seem frustrated to use that as a potential falsification.

MCCULLOUGH:  Maybe it’s just a matter of getting the groups together who have the resources to make that work happen. That’s a reasonable first approximation. Psychologists go to school and learn to do studies in the lab with Homo sapiens, and to the extent that they know researchers who have access to non-human animals it might be fantastic, but perhaps the network isn’t there.

PIZARRO:  Just listening to who’s been doing this work, it strikes me that as you set out the task of evolutionary psychology coming up with maybe some empirical horse races to pit hypotheses against each other, there are two ways in which this seems to be done. One is by people who are committed to the view that, say, in this case sexual selection would have placed specific pressures on female mate choice or ovulation. How did that manifest itself? As you were laying out two potential hypotheses, but both of those share the commitment that the way the natural selection worked was by changing women’s judgment about sexual selection during ovulation.

MCCULLOUGH:  No. Not necessarily during ovulation.

PIZARRO:  Yeah. But as a change in whether or not they are fertile.

MCCULLOUGH:  Well, no. It may be that they just like a certain set of traits all the time, 24/7. My alternatives were to explicitly not shift hypotheses.

PIZARRO:  It’s still framed as accepting that there would be direct natural selection pressure on female mate choice in a way that ought to be obvious with the right test, so you can have three hypotheses. It strikes me that the people who disagree with this, and the majority of people who find this controversial or object to it, it’s not because they think you have the wrong view of how sexual selection drove female mate choice and it’s also not that they don’t believe that evolution shaped the human mind, it’s simply that they disagree about whether or not natural selection or sexual selection, in this case, happened to shape something like sexual selection in the way that evolutionary psychologists say.
You have people, for instance, in cognition in general, who believe that the way natural selection works was by giving us a lot of general-purpose learning mechanisms. They’re not disagreeing with the view that evolution shapes the mind, but they are disagreeing with a very specific view about how evolution shapes the mind—very specific, functional, modular—and it’s rare to see somebody in this camp who has the commitment to saying that natural selection shaped the mind so specifically across human cognition, or who accept as evidence the studies that the people in the other camp do. It's almost what you take to be axiomatic is different than what they take to be axiomatic about natural selection, therefore, it generates very different predictions. Then there’s this tension because they think you guys aren’t paying any attention to specific hypotheses about how culture works, or how learning might cause diversity, so they accuse evolutionary psychologists of that. Then evolutionary psychologists accuse them of just being misguided about how sexual selection had to have shaped the mind with the specific modular decision-making mechanisms.

How do you get those two to agree on the terms by which you can say it turns out sexual selection has nothing to do with the way that women choose their mate or choose their clothing? (Maybe not just with their mate, but a large variety of other decisions that women might make.)

MCCULLOUGH:  The claim that I want to make is that mate choice is pretty close to the engine of natural selection. Organisms are going to be under selection pressure to identify reliable cues to mate quality. That’s about as far as I want to go without some more information, please. I just don’t see how natural selection can build general-purpose things. It can build a lot of special-purpose things that combine together to make you look awfully general purpose. But that, to me, is something that I have found useful for helping me.

PIZARRO:  Right. And that’s the heart of the controversy.

MCCULLOUGH:  It is absolutely the heart.

DAVID RAND:  It seems to me that part of the point that I’m taking from what David was saying is that if there’s going to be a concept of horseracing, all the different sides have to agree on who the horses are.
MCCULLOUGH:  Yeah. Absolutely. That’s right.

SANTOS:  And if it's domain specificity versus domain generality that’s going to look like a different horserace than: When in the cycle do they move?

JOSHUA KNOBE:  This point people are making with the horses seems helpful, in terms of thinking generally about what falsification is. There’s this picture that you started out with, from Popper, where you take one individual hypothesis, and there’s just some fact about whether they did an experiment that falsifies it. Even if you had no other hypothesis, you just have to abandon it.

Then there’s this other metaphor that looks much less like a trial and much more like a race. The idea is that whether one horse wins depends on its status relative to the other horses; it’s not just some individual fact about that one horse. In the latter way of thinking about falsification, the only way something can be falsified is dependent on how it did relative to these other hypotheses. It’s not that you can know from a given experiment whether it falsified this theory about the ovulatory cycle. It depends on whether you’re comparing it to other hypotheses that also involve adaptation, or to other hypotheses about the general mechanisms.

MCCULLOUGH: There’s another problem here, which is methodological, which is what if this stuff is so deeply buried by culture, by cultural variability, by time lags, modern environment being so different, that it’s so deeply buried that we can’t find it with the current methods, in any case. I share what tends to be a healthy skepticism about many layers of the feasibility of this enterprise. It's all quite valid.

RAND:  Another issue with the horse racing is that not only does everyone have to agree on which horses are in the race, but they have to agree on the details of the horses. In a lot of situations, people make evolutionary hypotheses using words that mean one thing to the person that said the hypothesis, but different things to other people. Then there’s a lot of disagreement. That’s where formal models, like from evolutionary game theory, can be useful. You say, "Look, this is exactly what I mean, and we can say exactly from this set of assumptions that this is what the result should be."

What that reveals is, from however many decades of evolutionary game theory models, one of the basic lessons is that what happens depends on the details of the assumptions that you make. You can falsify one specific model, but then people can think that framework just wasn’t the right group selection model, or it just wasn’t the right individual selection model. You can’t be like, "Look, this is an experiment that distinguishes between these two broad classes of theories." I guess you could if people that wanted one side to be right could try every single model they could possibly think of and never come up with a model that worked. Based on my understanding of models, that’s probably not going to happen. For any given observation, you’re always going to be able to come up with some individual selection and some group selection model that can make that prediction. So, it's hard. 

FIERY CUSHMAN:  For at least the first specific case that you brought up—the ovulatory cycle effects—we’re being our own conceptual sophistication in thinking about Popper and about a model comparison approach. It seems like what the meta-analyses disagree on is the question: If you assess women’s mate preferences over the course of the ovulatory cycle, do they change? The meta-analyses are not even addressing the issues of whether given the existence of such a change natural selection provides the right answer, or which form of natural selection, whether it’s to remain general or remain specific, that’s not what’s at issue. What’s at issue is: Is this a whole pile of false positives or not? It seems like we don’t have to engage in thinking about Popper or model comparison. What we need to do are some replications.

Everyone else has framed it as the question where there’s much bigger issues at stake. Am I missing something?

JOSHUA KNOBE:  It’s just because people are thinking about examples other than ovulatory cycles.RAND:  Say you take something where there is evidence, then the question is what can you conclude from that evidence.

CUSHMAN:  Is that a problem that you think characterizes evolutionary psychology? That is, there’re lots of instances where the evidence is clear, there’s no dispute over the effect, unlike the ovulatory cycle issues. The fundamental problem that we face is an adductive one—what is the best explanation for that evidence?

MCCULLOUGH:  The latter. I don’t wish to characterize all of evolutionary psychology as having this particular issue. Some areas of Ev Psych do, some areas of lots of psych do. In fact, it’s possible that most areas of psych do suffer from this problem. There’s a nice opportunity here for Ev Psych to take a cue from behavioral ecology in that way and remember that it’s good to train students to pick up another hypothesis and play with it, and hold them all equally dear and equally hostile. Then you can do a horse race that may end up with a more interesting finish.

RAND: One thing that occurs to me is that something that happens a lot in Ev Psych, and also is criticized about economics a lot, is that in both cases there is an overarching theory, a very clear overarching theory, and then there’s an attempt to see how much of the data fits with the theory, or in what ways can you contort the theory to try and make it fit with the data. My sense is that in a lot of the rest of psychology, there just is no overarching theory. It’s like we did some experiments, we got an explanation for these experiments. These other guys did an experiment, they have some explanation for their experiments. An overarching theory causes all kinds of issues, and maybe people bend it a little extremely sometimes, but that is maybe one of the things that comes along with trying to have an overarching theory of things.

MCCULLOUGH:  That’s why I made the point I made about the very strong theoretical beliefs that evolutionary psychologists have. There’re two or three beliefs that are very "Father, Son, Holy Spirit," they’re very strong.

HUGO MERCIER:  To some extent when we are talking about having experiments that can distinguish between the very broad classes of models, such as group selection versus individual selection, or even when we’re referring to Popper's falsificationism, which isn’t meant for this kind of thing, we have to give up this physics envy that we have. We have to put much more weight on the preponderance of the evidence. You have this whole framework, and you can look at things, not only experiments, but real life data. This is completely undervalued in psychology. You have to have that one experiment that shows what you want to demonstrate, and everything else just gets thrown out of the door. This is done partly, for good reasons, because we want to be more scientific, we put a lot of weight on these things, but we might be slightly off sometimes.

MCCULLOUGH:  The people who work in this area do take what they can get from what we know about the life history and demography of ancestral humans and they try to interact with it. To the extent that I portray them as being insensitive to other kinds of data in their theory development, I probably did them wrong. They do take that other work seriously and try to interact with it, but there are still holes in the record. We don’t have a complete, perfect evolutionary reconstruction of the human ancestral social ecology any more than we do for any other critter. There are going to be multiple ways of getting from there to here.

MERCIER:  Then it is not true that science progresses funeral by funeral. In sciences that work, you don’t need people to die. Revolutions happen all the time. People change their minds all the time to adopt the best theories. We have less good evidence, so it makes sense that we change our minds more slowly. In physics, clearly people change their mind all the time very quickly, in biology as well, so it's our problem rather than the problem of science.