This paper analyzes the emergence of norms from an evolutionary perspective. It is hypothesized that norms diffuse not necessarily because they are functional for the communities in which they are embedded but because in certain selection environments they are evolutionarily advantaged: norm-like strategies can be evolutionarily stable while all strategies that are not norm-like are unstable. Further, we show that, contrary to classical functionalist theories, there are norms that are both evolutionarily stable and “dysfunctional” (Pareto-deficient). However, functionalism is partially redeemed by our quantitative finding that in a wide class of situations (games) functional-i.e., Pareto-optimal-norms are more stable than dysfunctional ones. We then apply and extend these results in order to analyze the relative evolutionary stability of ascripitive norms, which take into account people’s ascriptive identities (e.g., caste norms), versus the stability of universalistic norms, which ignore identity and focus only on people’s actions. Although ascriptive norms can be evolutionarily stable, in a wide class of situations universalistic norms turn out to be more stable in several quantitative respects.