How do we know what we know? Sexual selection, in humans and in lizards.

Over the last few months, there’s been a slow-boiling battle underway between Holly Dunsworth and Jerry Coyne about the evolution of sexual dimorphism in humans, surrounding the question of why male and female humans, on average, differ in size. The battlefield ranged from blogposts to twitter to magazine articles. In a nutshell, Coyne argued that “sexual dimorphism for body size (difference between men and women) in humans is most likely explained by sexual selection” because “males compete for females, and greater size and strength give males an advantage.” His whole argument was motivated by this notion that certain Leftists ignore facts about the biology of sex differences because of their ideological fears, and are therefore being unscientific.

Dunsworth’s response to Coyne’s position was that “it’s not that Jerry Coyne’s facts aren’t necessarily facts, or whatever. It’s that this point of view is too simple and is obviously biased toward some stories, ignoring others. And this particular one he shares…has been the same old story for a long long time.” Dunsworth went on to propose, seemingly off the cuff, alternative hypotheses for sexual dimorphism in body size in humans that were focussed not on men but on women, as examples of the kind of hypothesis that is relatively rarely considered or tested in this field.

Though on the surface this battle may seem to be about specific biological facts (Coyne certainly tries to win by treating it that way), in reality this disagreement is, as Dunsworth argues, about the process by which hypotheses are tested and about how knowledge comes into existence. About which hypotheses are considered for testing in the first place. As a result, the two ended up arguing past each other quite a bit.

As I followed this whole exchange, I shook my head at the timing–I had a paper in preparation that was SO RELEVANT to the centre of this debate! That paper is now available as a preprint, so I can try to outline why I think that Dunsworth is right, and Coyne is being short-sighted. My argument has *nothing* to do with humans, however–I don’t know the human sexual selection literature well enough to weigh in on that. Instead, my argument is by analogy with our knowledge of mating systems in Anolis lizards.

Until relatively recently, it’s been widely accepted, on the basis of behavioral studies, that anoles are territorial and polygynous. This description is so prevalent, in scientific as well as popular accounts of these lizards’ biology, that anyone who knows anything about anoles wouldn’t stop to think twice about it.

Following a somewhat circuitous path (which you can hear about at or after my thesis defence on March 28th!), the main project of my Ph.D. ended up being an investigation of exactly this “fact”–are anoles really territorial? Genetic evidence, which has shown evidence for females mating with multiple males as well as complex spatial relationships between mating pairs, led me to wonder if territoriality was a useful description of these lizards’ mating system. And while an important part of my work is empirical, my best sampling was restricted to a single population of a single species. So a complementary and equally important endeavour has been reviewing all of the evidence we have in support of the conclusion that anoles are territorial. And an outcome of this endeavour has been realizing that even things that we *think* are well-supported scientific facts–like territoriality in Anolis–may in reality be based on very little evidence.

I’m not going to rehash the whole argument of the review paper (co-written with my advisor Jonathan Losos) over here. The gist is that the earliest studies concluded that anoles are territorial based on strange and limited data, but the idea caught on. Most subsequent studies, therefore, ended up assuming territoriality implicitly or explicitly. This assumption affected choices made in sampling design, analysis, and interpretation, such that it became unlikely that studies would consider important, or even be able to detect, behaviours that were not quite territorial but were still potentially important for reproduction.

To illustrate what we meant, I am going to excerpt a couple of paragraphs below (edited for out-of-context clarity and to remove examples). In this section,we argued that if studies are designed with the assumption, implicit or explicit, that individuals remain in relatively small, exclusive areas (i.e. they are territorial), then these studies end up being designed such that they will not detect, or won’t consider important, evidence suggesting otherwise:

Because by the 1970’s the consensus seemed to be that anoles are territorial, research at this time was not often designed to explicitly test if these lizards behave territorially, i.e. to show that they stay in the same place and maintain exclusive areas. Specifically, territoriality was an almost foregone conclusion in studies with a limited spatial and temporal extent of sampling.

If the sampling period of a study of social behavior is not long enough, then relatively infrequent but reproductively consequential departures from territorial behavior are unlikely to be detected often enough that they are considered signal and not noise. This includes not only occasional forays away from and returns to a fixed territory, but also shifts in territory location that may take place only a few times per breeding season—neither would be detected by studies with short durations.

Moreover, if a study of social behavior does not sample over a large enough area and a sampled individual disappears from the study site, researchers cannot know if the individual has died or simply moved. Thus, studies with limited sampling areas will be most likely to sample only those individuals who stay in the same place, that is, animals whose behavior appears territorial.

You can read the paper for details and examples of limited sampling, as well as other ways in which research choices were shaped by assumptions of territoriality. But in sum, because of this dependence on territoriality throughout research on anole social behavior, the facts we have come to hold about these lizards’ biology look very different than they may have if biologists had started out with different assumptions, or had clarified what their assumptions were. The upshot is that, at this point, we simply don’t know if anoles are territorial or not–they may well be, but we don’t yet have good evidence for it.

If you read the paper, you may glean that I now think that instead of arguing about whether or not anoles are territorial, it’s more fruitful to ask if territoriality is a useful way to describe behavior. In the case of Anolis, I don’t think it is. Others of course may disagree, and our disagreement could be resolved by testing predictions emerging from territoriality against predictions that do not depend on territoriality. But this would be very different from considering the predictions made by two hypotheses that both reside within a territorial framework. And this, I think, is Dunsworth’s point–which hypotheses we consider and how we decide to test them not only shapes what facts we have the capacity to discover but also depends an awful lot on what we think we already know. What we think we know in turn depends an awful lot on the particular trajectory that a body of research has followed. Consequently, the hypotheses that get tested do not emerge from a vacuum. All hypotheses emerge from assumptions, whether we recognize them or not.

While our paper, by design, deals primarily with assumptions about Anolis territoriality originating within science, it hasn’t escaped my attention that the existing description of these lizards’ social behavior are positively Victorian. This struck me most clearly when I explained my empirical research to non-biologist friends and family. “They sound so old-fashioned!”, I was told, which made me realise that this might well be because the science on which the descriptions are based originated in the social milieu where these now-old-fashioned ideas were a given *. It also hasn’t escaped my attention that most of the research that suggested departures from territorial behavior in Anolis remained unpublished in scientific journals, and that three of the four genetic studies showing female multiple mating were conducted by women scientists. These observations wouldn’t sway anyone who believes that science is 100% objective, and it’s certainly possible that animal behavior could conform precisely to Victorian ideals, but I think the coincidences are at least worth pondering.

In this world where the very concepts of knowledge, facts, and scientific expertise are under dispute, I fully recognize the danger of writing about how science can do an incomplete or even incorrect job of discovering truths about the natural world. But it isn’t doing science any good for us to ignore how our processes of discovery can be blinkered by unwarranted assumptions, assumptions that can originate either inside of science or outside it, from myriad sources of bias that afflict every single one of us. The future of science cannot depend on us pretending that we scientists are infallible.

*Of course, none of this is new to sociologists or historians of science–check out, as just one of many examples, Erika Milam’s Looking for a Few Good Males ^.