Two Zombie Papers in Human Behavioral Ecology

This is a box from a book chapter that I wrote with Jeremy Koster (volume). In it we comment on unscientific and counter-productive norms in our field. We also suggest constructive reforms. Since the chapter will be buried in a book, I’m reprinting the best bit here on my blog. You can download a PDF of the whole chapter.

The purpose of this chapter is to examine the general conditions and development of the field of human behavioral ecology. We do not focus on particular papers. However, there are two papers which have been published often enough now that they deserve special consideration. These papers contribute little to cumulative science but are difficult to eradicate.

The first paper is entitled something like “Social Networks, Reproductive Success, and the Evolution of Human Sociality.” After an introduction that promotes the banal claim that social relationships and fertility may influence one another, the number of self-reported friends is shown to be associated with past fertility. The analysis adjusts for a basket of perfunctorily measured demographic variables, and every coefficient is interpreted as a total causal effect. The paper claims that this finding demonstrates the value of an evolutionary approach to human behavior, mainly because economists have ignored it. But since no evolutionary model predicted the result, the size and direction of the association have no impact on any evolutionary model. Any evolutionarily relevant variable like relatedness or age at first birth can be deployed in this paper to imitate, but not actually to perform, behavioral ecology. The paper is highly cited.

The second appears under the banner “The Evolution of Human Cooperation: A Bayesian Phylogenetic Analysis of Kinship and Pudding Recipes.” A cross-cultural sample of stereotyped features of human societies is fed into software designed for the analysis of biological species. The procedure identifies an association between uxorilocal residence and the prominence of dessert. No model predicts this association nor its size. But rejecting a null model of no association somehow illustrates the value of an evolutionary approach. A few fatal confounds and methodological flaws flirt with the reader in the twilight paragraphs. The result has implications for the origins of human society and the possibility of world peace.

We have likewise written these papers, but we are trying to stop. If we could all avoid producing, reviewing, and reading these papers ever again, we would divert a substantial amount of talent toward productive research.

The authors of these papers could make a contribution by instead advancing an optimization model that resembles the models that are described in other chapters. Then the model’s predictions could be made algorithmically and contrasted with similarly precise alternatives. Data collection and analytical procedures could be designed that have some hope of causal identification. But at least the assumptions that justify a causal interpretation would be clear. Making zombies does not train us to perform any of these tasks. Our students and their students deserve better.