A model of dynamic R&D behavior is presented in which participants in the race have imperfect information about the (true) "hazard rate" of the R&D process. In this model, a firm will be ambivalent about a rival firm's success at an intermediate stage. On the one hand, the probability of winning is reduced, since a rival firm is ahead and the technological gap is larger. This effect is always negative. On the other hand, the discovery could be a signal that the project is not as hard after all ("If you can do that, why not me?"), which could shorten the expected time needed for the discovery. This is a positive effect of a rival firm's success, one that is not present in existing models and hence has been ignored up to now. According to the relative magnitude of these two opposing effects, a much richer description of real-world R&D behavior is obtained. This article also provides a potential explanation of the strategic practice of innovation shelving.