PZ Myers Clarifies Criteria For Distinguishing Genuine Hypotheses from “Just So Stories”Published 28 December, 2011
PZ Myers recently wrote a blog entry about the answer to the question, “Why Do Women Menstruate?” In the piece, he went through a number of candidate answers and summarized a recent paper that addresses this question. My interest after having read the post was how to reconcile Myers’ discussion of the possible evolved function of menstruation with his dim view of evolutionary psychology, since it seemed to me that the structure of the arguments he entertained was the same in the menstruation case as in evolutionary psychology.
A commenter (bromion, #23) on Myers’ site had the same question, writing:
I’m wondering why you support this kind of evolutionary explanation of a phenomenon that is difficult to prove directly (we can’t observe an evolutionary fossil record here), but dismiss the entire field of evolutionary psychology as “just so” stories.
Myers (#32) answered as follows:
This hypothesis has the advantage of being based on actual comparative data and the physiology of decidualization. It has an explanation based on the cell and molecular mechanism, not just derived from a phenotype. And it makes a specific, testable prediction about how the shift could have been made.
It isn’t about looking up the fossil record (fossils suck for evaluating most evolutionary hypotheses), but about using the molecular evidence to evaluate the answer. We have molecular evidence. We can also get more much more readily than we can dig up a fossil uterus.
This is really useful because Myers has made explicit his criteria for distinguishing a legitimate scientific hypothesis from a “Just So Story.” (The “Just So Story” label, for the uninitiated, is meant to imply that the hypothesis in question cannot be shown to be false, and therefore is not a legitimate scientific hypothesis.) Myers seems to have committed to what appear to be three reasons having to do with 1) comparative data (which I think is the same as “actual comparative data”), 2) cell & molecular mechanism, 3) predictions about “the shift,” which in context I believe refers to the evolutionary pathway from the ancestral condition to the present condition.
I thought I would apply these three criteria to some hypotheses to see how they fare on the Myers Story Scale (MSS). Let’s take a hypothesis Jerry Coyne proposed in a blog entry some time ago, about why sloths come down to the ground to defecate instead of doing so in the tree. His favored hypothesis seems to be that this is to attract mates; the dung pile is a signal to potential mates that a sloth is present in the dung-adjacent tree. Does it satisfy the three criteria Myers points to?
- Comparative data. Nope. The post was just about sloths.
- Based on the cell and molecular mechanism. Fails. Just poop.
- Makes a prediction about evolutionary course. Fails. No discussion of this.
This hypothesis, then, is 0 for 3 on the MSS, and so is an untestable Just So Story, according to the Myers criteria. Coyne, however, thinks this is a perfectly fine hypothesis, writing that “in principle these theories are testable. We could see, for example, how sloths manage to find each other at mating time. We could also do mock-defecation studies from branches, using model sloths, to see if the noise attracts predators.” From this conflict, we can see either that the criteria are wrong, or Coyne is wrong. I don’t see another option.
To take a second example, in Myers’ post, he entertains and rejects one explanation for female menstruation, writing that one possibility is that “humans have rather aggressive embryos that implant deeply and intimately with the mother’s tissues, and menstruation “preconditions” the uterine lining to cope with the stress. There is, unfortunately, no evidence that menstruation provides any boost to the ‘toughness’ of the uterus at all.”
He rejects this hypothesis, clearly, but he doesn’t reject it because it’s an untestable Just So Story. He accepts it as a legitimate hypothesis because he (implicitly) reasons as follows. If the function of the trait is coping with future embryos, then the trait should have properties (toughness) that contribute to this function. From the observations of toughness (however that is measured), he rejects the hypothesis. The key idea to note is that the claim about the function is, in itself, enough to render the hypothesis testable. The hypothesis did not seem to need comparative data, etc.
Finally, at the end of the post, in a postscript, PZ Myers tried to become an evolutionary psychologist, trying his hand at understanding one aspect of human social behavior, why some people are pro-choice and others pro-life. Here is his explanation:
The maternal-fetal conflict is also a conflict between males and females: it is in the man’s reproductive interests to have his genes propagated in any one pregnancy, while it is in the woman’s reproductive interests to bail out and try again if conditions aren’t optimal for any one pregnancy. This conflict is also played out in culture, as well as genetics — pro-choice is a pro-woman strategy, anti-abortion is a pro-man position. Sometimes, politics is a reflection of an evolutionary struggle, too.
First of all, it’s obviously very interesting that Myers seems to be trying to explain a psychological phenomenon – here, moral variation – by thinking about the relevant fitness interests. It’s a bit surprising that the people commenting on his blog, many of whom roundly condemn evolutionary psychology, seem not to have had a concern about his proposal here.
It’s not precisely clear, but this remark seems to me to be proposing that the reason for variation in views on abortion has to do with reproductive interests and, in turn, the key variable that underlies these differences is one’s sex, with women gaining an advantage by adopting the pro-choice position. Now, the proposal does poorly according to the MSS — no comparative data, no cellular mechanisms, and no word about the evolutionary trajectory of the trait – but the proposal does seem to make a prediction, that a person’s sex will be the key predictor variable for views on abortion. As it happens, there are a large number of datasets that Myers could have consulted in making this proposal, nearly all of which would have shown that there is no sex difference, and, when there is one, it accounts for a tiny amount of the variation in abortion views.
Now, the part that Myers might be right about is the idea people’s views on sexual issues can be systematically predicted by considering their strategic reproductive interests. But the crucial variable isn’t sex, it’s reproductive strategy. In my opinion, the best work on this is by Jason Weeden, who has gathered and analyzed a tremendous amount of data to interrogate his explanation for variation on abortion views. (Weeden’s dissertation is, with respect, one of the only dissertations I’ve even come across that’s actually worth reading. Note that in the spirit of disclosure, I should mention that I’ve collaborated with Weeden and we have made a similar proposal regarding views on recreational drugs.)
In terms of the broader theme here, my guess is that Myers doesn’t really believe that those three criteria distinguish a testable hypothesis from a Just So Story. As I’ve shown here, he himself produces hypotheses which fare poorly according to his own criteria. It is easy to find other hypotheses from the animal and human literature, measure them against the MSS, and show that hypotheses that fail on this measure are treated as perfectly legitimate hypotheses. The reason is that functional claims entail predictions about the structure of traits, whether physiological or behavioral, as illustrated by the three cases above. The criteria Myers enumerates are not required for making a hypothesis falsifiable, and from his own writing, it seems that he implicitly believes this.
In short, Myers’ proposal shows that he thinks that evolutionary reasoning can form the basis for hypotheses about the human mind and human social behavior. Yes, I wish that he would police himself a bit more, and dip his toe in the pools of data available to evaluate the claims he makes before writing about them, but I’m very happy to look past that because I think it’s encouraging that he’s starting to write about how evolutionary ideas can inform hypotheses about human behavior. He didn’t do all that well this time, but I’m confident that, with helpful policing from others, he can do better.
Weeden, J. (2003). Genetic interests, life histories, and attitudes towards abortion. Dissertations available from ProQuest. Paper AAI3087480.