Abstract: I introduce a framework for studying different interpretations of meritocracy. Each meritocracy has two components: a merit criterion, which determines when one individual is more meritorious than another, and a reward criterion, which determines when one outcome is more rewarding for a given individual. An allocation is meritocratic if more meritorious individuals are more highly rewarded. The framework distinguishes between meritocracy as an end, in which rewarding meritorious preferences is intrinsically valuable, and meritocracy as a means, in which rewarding meritorious actions is instrumentally valuable for achieving desirable outcomes. I show that these two conceptions are equivalent: any merit criterion over actions in a mechanism that yields meritorious outcomes must reduce to an underlying merit criterion over preferences. I apply the framework to two specific conceptions of meritocracy. Pareto meritocracy, according to which an action is more meritorious if it leads to a Pareto improvement, turns out to be vacuous, imposing virtually no constraints on allocations. Proportional meritocracy, according to which each individual’s consumption is proportional to labour input, is characterised by three conditions: monotonicity of the merit criterion in labour, scale invariance, and a welfarist reward criterion. This characterisation yields testable predictions for experiments where impartial spectators choose allocations.