Suppose that you're interested in the evolution of human behavior, and specifically in the strategies which shape our choice of long-term mates. Suppose, more specifically, that you're a pair of behavioral scientists at Cornell named Peter Buston and Stephen Emlen. You would then then take 978 heterosexual undergraduates and ask them how important they thought nine different traits were in a potential long-term mate ("financial resources", "physical attractiveness", "faithfulness", "parenting qualities", "social status", "health", "desire for children", "devotion", "ambition", "strength of family bonds"); you would also ask them to rate themselves on those qualities. You might find that students who say they are, e.g., very attractive claim to demand very attractive mates, and vice versa. This would be an interesting finding, and it would certainly be reasonable for it to be published someplace respectable, like the Proceedings of the National Academy of Sciences.
It would not be reasonable, however, to say that you are actually studying "the cognitive processes underlying human mate choice", for several reasons.
The first should be obvious but apparently isn't; even if some of the students in the sample have found long-term mates, no data was collected on this. So we don't know that people who think they are attractive end up with more attractive mates, or even that they try to get more attractive mates, just that they want them.
But the second point is I think the real killer for this study, or anything like it. There is no data on revealed or behavioral preferences, just on stated preferences among the available categories. Notice that many things which are socially acceptable to admit influence such choices, like age or religion or intellectuality, just aren't on the list. Notice also that nothing impolite is on the list, however important or even determiniative it may be in real life, e.g., "race", "quality of sex", "frequency of sex" or "feelings about leather boots".
But let's ignore that issue for a moment and think just about what we're actually learning from the answers to the questions about the nine categories. We are not seeing what people want in those lines, but what they are willing to tell the questioner they want. Who were the questioners? "Each of the 122 students enrolled in extra-credit discussion sections of the Cornell University Introduction to Animal Behavior course were asked to administer questionnaires to 10 other people." In other words, undergrads were gettting asked by their friends about their mate preferences. I don't think it's very daring to conjecture that there is a certain pressure, in such a situation, to say the right thing; which is not necessarily the truthful thing. We are not getting at the secrets of the heart here, just more presentation of self in everyday life. Similarly for point (3) above: what's treated as self-evaluation is actually what the subjects presented to someone else as their self-evaluations.
Grant that these sophomores sincerely told their friends what they think they want in a mate. They may be wrong. What they think is important to them may not in fact be important to their behavior, and it's behavior which matters evolutionarily. I used to have a friend who said, and I think meant it, that what he wanted was intellectual companionship from a fellow Catholic; his actual behavior could be explained entirely as a function of breast size. Most of us probably know similar cases. That there is often a gap between what people say they want, even what they think they want, and what their acts reveal they want, should not have to be explained to anyone with their eyes open, certainly not to behavioral scientists. But the conclusions of the paper --- that people seek mates who are similar to how they see themselves, because this leads to stable partnerships, which are more reproductively successful --- only makes sense if this gap is closed.
Perhaps scientists, ivory-tower dreamer types that we are, might miss this; but surely hard-eyed journalists will see the gap? Evidently not; I offer in evidence Nathalie Angier's write-up on the paper in the New York Times. As usual with Angier, it is amusing, cleverly written, and shallow; she even manages to make it sound like the Cornell authors are reporting a correlation between how people are and what they want, rather than between how they see themselves and what they want. She also manages to mis-represent the position of most evolutionary psychologists, but this is so routine with her as to hardly be worth noting.
A better study would look at revealed preferences, in the form of what people pursue in their mates. It would be hard to design, and harder still to design in a form which would get past a human subjects committee. Certainly I don't have any bright ideas this afternoon. But we either need such studies, or some way of showing that preferences stated in this way reliably indicate actual behavior, or we need to leave the field to novelists writing comedies of manners.
Update, 5 August 2003: A number of people (some of them not on the editorial board of Social Text) have asked me how an article with such a glaring logical flaw could make it through the scientific community's vaunted process of peer review. The answer is, it didn't. More exactly, articles in PNAS either go through the normal process of peer review, or they are "communicated" by members of the Academy, and don't; this paper was communicated. (See here for more details.) Many people, understandably, are not happy about this system.
Posted by crshalizi at July 12, 2003 12:54 | permanent link