In the funding of R&D, push mechanisms, such as research grants, subsidize research input, while pull mechanisms, such as innovation prizes, reward research output. By rewarding research output, pull mechanisms create strong incentives for researchers to devote non observable inputs to R&D. Push mechanisms, in contrast, may reward a researcher independently of her output. In the presence of moral hazard, it might seem that push mechanisms generate weak incentives for non observable inputs from the researcher and, absent risk-sharing considerations, would be inferior to pull mechanisms; it is the aim of this paper to critically assess this hypothesis. I analyze a principal-agent model in which a funder encourages R&D activity through a push incentive (a grant) and/or a pull incentive (a prize); R&D input consists of both an observable and non observable component. In contrast to the stated hypothesis, it is shown that a grant may emerge as an optimal means of funding as a result of the interaction between adverse selection and moral hazard. The model also helps to explain the use of matching grants, it is shown that such grants serve as an effective sorting device in the presence of adverse selection.