This paper studies the role of goal bracketing to attenuate time inconsistency. When setting non-binding goals in multi-stage project, an agent must also decide how and when to evaluate himself against such goals. In particular, he can bracket broadly by setting an aggregate goal for the entire project, or narrowly by setting incremental goals for individual stages. In the presence of loss aversion and uncertainty over outcomes, this decision involves a trade-off between motivation and comparative disutility due to ex-ante uncertainty. Narrow goal bracketing can be used as an instrument to counteract the self-control problem, while broad goal bracketing can itself generate apparently erroneous behavior such as the sunk cost fallacy. The sequential nature of decision-making introduces a differential reaction to outcome uncertainty based on its timing, which determines the optimal bracketing choice. * I thank my adviser, Roland Bénabou, for guidance and encouragement. I also thank