How can it be rational to work on a new theory that does not yet meet the standards for good or acceptable theories? If diversity of approaches is a condition for scientific progress, how can a scientific community achieve such progress when each member does what it is rational to do, namely work on the best theory? These two methodological problems, the problem of pursuit and the problem of diversity, can be solved by taking into account the cognitive risk that is involved in theory choice. I compare this solution to other proposals, in particular T. S. Kuhn's and P. Kitcher's view that the two problems demonstrate the epistemic significance of the scientific community.