Anti-reductionist philosophers have often argued that mental and linguistic phenomena contain an intrinsically normative element that cannot be captured by the natural sciences which focus on causal rather than rational relations. This line of reasoning raises the questions of how reasons could evolve in a world of causes and how children can be acculturated to participate in rule-governed social practices. In this paper I will sketch a Wittgensteinian answer to these questions. I will first point out that throughout his later (...) philosophy Wittgenstein draws a sharp distinction between "teaching" and "training": newly-born children are trained (conditioned) to react to specific stimuli in specific ways, which then allows them to acquire concepts and follow rules. I will then show that this picture presupposes a strong analogy between concepts and capacities, which is also present in Wittgenstein's later philosophy. In the last section I will point out that Wittgenstein only discusses the ontogenetic question of how individual children can acquire speech, but not the phylogenetic question of how rule-governed behavior could evolve in the first place. I will argue that this strategy should not be seen as a shortcoming, but rather as an expression of Wittgenstein's approach that can be characterized as naturalistic in a wide sense. (shrink)