A firm that wants innovation must attract workers whose suitability for exploration it cannot observe. Any contract that attracts explorers is also open to exploiters. In a two-period model with hidden actions, explorers and exploiters generate different performance histories. The firm therefore screens by placing pay on histories relatively more likely under the intended activity than under mimicry, so the timing and path dependence of compensation become instruments of selection. Exploiters are paid for each success; explorers mainly for long-run success and, only when early failure is sufficiently diagnostic, after early failure. Commitment, termination, and feedback screen by the same logic. A common contract can help overcome a lack of commitment; separate contracts earn their keep when termination or feedback must differ. Two predictions follow. Exploration-oriented incentives change who enters, remains in, and chooses exploratory work, not merely innovative output. Competition for talent shifts where pay sits across histories, not merely its level.