Abstract

This paper defends global realism about the units of selection, the view that there is always (or nearly always) an objective fact of the matter concerning the level at which natural selection acts. The argument proceeds in two stages. First, it is argued that global conventionalist-pluralism is false. This is established by identifying plausible sufficient conditions for irreducible selection at a particular level, and showing that these conditions are sometimes satisfied in nature. Second, it is argued that local pluralism – the view that while realism is true of some selection regimes, pluralist conventionalism holds for others – should also be rejected. I show that the main arguments for local pluralism are consistent with global realism. I also suggest that local pluralism offers an unacceptably disunified view of the metaphysics of selection. It follows that we should accept global realism. But this leaves open the question of how to classify so called ‘multi-level selection type 1’ (MLS1) processes, such as Wilson’s classic trait-group model for the evolution of altruism: should they be interpreted as particle selection or collective selection? On the assumption of global realism, at most one of these is correct. I argue, against global realists such as Sober, that MLS1 processes should be understood as particle, not collective, selection, due to three features of MLS1: the reducibility of collective fitness, the absence of collective reproduction, and the dispensable role of collectives.