This paper examines the role of spillover learning in shaping the value of exploratory versus incremental R&D.  Using data from pharmaceuticals, we document a new stylized fact: relative to those that are incremental, novel drug candidates generate more dynamic spillovers.  Despite being more likely to fail in the development process, novel drugs are more likely to inspire the development of subsequent successful drugs.  Motivated by this, we develop a model in which firms have less information about the viability of novel projects, but undertaking such projects generates information even if the project fails.  Our model then provides an empirical diagnostic for assessing the value of increased certainty today versus increased learning tomorrow: if firms value learning, then they should be willing to accept lower direct revenues when investing in novel drugs.  Empirically, we show that this is not the case.  Novel drugs are substantially less likely to enter development and, if approved, tend to generate higher direct revenues.  We provide evidence that firms are concerned about appropriability: in more competitive markets, firms place even less value on spillover learning.