Chomsky and Lasnik (1993) argue convincingly that PRO has null Case, checked by nonfinite T, and suggest that this may explain PRO's narrow distribution. However, their analysis falls short of reaching this goal. Here, I refine the theory of null Case so as to fully account for the distribution of empty and lexical subjects of nonfinite clauses, concluding that this minimalist analysis is more explanatory than earlier ones based on the theories of binding and government. In particular, I argue that whether or not nonfinite T can check null Case depends crucially on its temporal properties and present a number of empirical arguments supporting this conclusion.