Jessica Pfeifer (UMBC)
It is commonly argued that probabilities in evolutionary theory result from abstraction (e.g., Sober 1984, Matthen and Ariew 2002, Matthen 2009). I delineate different modes of abstraction, and show these different modes affect how we think about fitness, selection, and drift. One might abstract from factors causally relevant to some effect for epistemic reasons, or one might abstract from causal factors because they obscure some feature of the world. Adequately representing these features requires that we abstract from the “gory details.” Gory details might obscure general patterns, as Kitcher (1984) and Sober (1984) have argued, or they might obscure underlying causal structure, an argument attributable to Mill (1872) and defended by Cartwright (1989). Following Mill’s and Cartwright’s line of reasoning, I argue that abstracting from factors that affect evolutionary outcomes allows us to represent two different types of causal processes: selection and drift. To make sense of this difference, I draw a distinction between selective environmental factors and non-selective environmental factors. Some environmental factors make a difference to evolutionary success partly in virtue of differences between the competing entities (the selective factors); some environmental factors make a difference simply because those factors are unequally distributed among the competing types in the population of interest (the non-selective factors). If we want fitness values to reflect this difference, as causalists ought to, then we ought to abstract from non-selective factors and make our fitness values relative to selective environmental factors.
This has important implications for understanding abstraction, for understanding of what it means for selection to act alone, and for understanding the relation between natural selection, drift, and population size. When we abstract from non-selective environmental factors, we are not considering what would happen were those factors absent. Instead, we are representing what would happen were the non-selective environmental factors equally distributed across the competing types of interest. What matters is whether the non-selective environmental factors make a difference to the relative success of the competing entities. The non-selective factors will not make a difference so long as they are equally distributed among the competitors. This marks a significant difference in the way that Mill thought about a causal factor acting alone. In the case of natural selection, selection can act alone even when non-selective causes are present, so long as the non-selective causal factors are equally distributed. Hence, when abstracting from non-selective causal factors, we are not thereby ignoring or subtracting the non-selective causal factors, but instead controlling for those factors. Drift, then, occurs whenever (though not only when) the non-selective causal factors are unequally distributed across the competing types of interest. Hence, whether drift occurs will not be purely a function of population size, but instead will depend partly on how the non-selective causal factors are distributed across the competing entities. Moreover, whether population size makes a difference to the likelihood of drift will be partly an empirical matter, not purely a mathematical truth.
