When drawing up a contract, it is often impracticable for the parties to specify all the relevant contingencies. In particular, they may be unable to describe the states of the world in enough detail that an outsider (the courts) could later verify which state had occurred, and so the contract will be incomplete. The parties can make up for this incompleteness to some extent by building into the contract a mechanism for revising the terms of trade as they each receive information about benefits and costs. One striking conclusion of our analysis is that because the parties can rescind the original contract and write a new one, severe limitations are placed on the form the revisions can take. Moreover, these limitations depend crucially on what means of communication the parties have at their disposal during the revision process. We characterize an optimal contract in two cases. First, when a contract is being used to facilitate trade between two agents who must undertake relationship-specific investments, it is generally not possible to implement the first-best. For a particular example, we are able to confirm the idea that the second-best outcome will involve under- investment. Second, when a contract is being used to share risk, and there are no specific investments, we find that it is possible to implement the first-best provided messages sent between the agents can be publicly verified.