We examine incentives in research teams where the market, such as the scientific community, attributes credit for success based on its inference of individual efforts. A social planner who could commit to credit ex ante would induce more effort from higher ability agents in exchange for less credit per unit effort. Lacking such commitment, the Bayesian market assigns credit proportional to perceived effort. This inability to distort credit per unit effort leads to an incentive reversal across projects. For "easy" projects with a concave marginal cost of effort, in the unique interior equilibrium, higher ability agents work less and receive lower credit/utility, while the opposite holds for "difficult" projects with a sufficiently convex marginal cost of effort. Moreover, equilibrium may involve over-investment by some team members who expect to receive most of the credit. The incentives to team up and the implications of effort observability on credit attribution are also investigated.