A theory of the optimal number of banking relationships is developed and tested using matched bank-firm data. According to the theory, relationship banks may be unable to continue funding profitable projects owing to internal problems and a firm may thus have to refinance from nonrelationship banks. The latter, however, face an adverse selection problem, as they do not know the quality of the project, and may refuse to lend. In these circumstances, multiple banking can reduce the probability of an early liquidation of the project. The empirical evidence supports the predictions of the model.