Trait-centered vs. fitness-centered definitions of natural selection

During the past few decades, fitness-centered and trait-centered definitions of natural selection have coexisted in the philosophical literature. The former render natural selection definitionally dependent on the presence of fitness differences, where “fitness” is understood as a distinct property from actual reproductive success. On the other hand, trait-centered definitions see selection as definitionally dependent on the presence of a causal relation between a trait (not necessarily fitness) and reproductive success. Interestingly, endorsers of these definitions have rarely–and usually only cursorily–critically engaged the views of the other camp. Therefore, a critical comparison of the two kinds of definitions is lacking in the literature. This paper starts filling this void by opening a discussion about which of the two kinds of definition is more appropriate. I first argue that fitness-centered definitions have difficulties in accommodating cases of opposing selection on correlated traits, whereas trait-centered views have no such problems. To do so, I revisit an old argument put forth by Elliott Sober and I show that recent attempts from the fitness-centered camp to reply to Sober’s charge are unsuccessful. I then show that fitness-centered views also have problems with a different type of case, namely opposing selection on a single trait; trait-centered views, on the other hand, may accommodate such cases if, as I propose here, we specify that the causal relation that figures prominently in them is understood as a relation of contributing causation. These arguments suggest that trait-centered definitions of selection are preferable to fitness-centered ones.