I study the behavior of individuals with present biased preferences who are involved in costly, long-run projects. By using generic cost and reward functions, I characterize the behaviors of the sophisticated, partial naive and naive types. It is shown that there may arise cases where naives needlessly put effort on projects they never complete. Moreover, in endogenous total cost projects, the naive types always end up completing projects of lesser quality than originally intended. By introducing a bonus motive, it is shown that agents with higher self-control problems should be given a higher bonus to prevent inefficient procrastination. I, then, characterize the behavior of partially naives who potentially learn self-preferences. It is found that without learning self-preferences, partial naives behave either like sophisticates or naives depending on the level of naivete; with learning, if the learning pace is fast enough, procrastination until the deadline does not occur. (C) 2011 Elsevier B.V. All rights reserved.