This paper is about the implications of a cross-linguistic generalization on theories of genericity. The generalization is that if kind-denoting and generic-characterizing plurals are non-uniform in definiteness, then definiteness is better with the former compared to the latter. This poses a challenge to the prediction of uniform definiteness between the two sorts of plurals by the neo-Carlsonian and optimality-theoretic approaches to genericity. We meet the challenge by integrating the observation that kind-reference involves a narrower notion of maximality than generic-characterization.