For the last thirty years, string theory has played a highly influential role in fundamental physics without having found empirical confirmation. The presentation will analyse reasons for the high degree of trust many physicists have developed in a theory that, according to classical standards of theory assessment, would have to be called an unconfirmed speculation. It will be argued that the case of string theory suggests an extension of the concept of theory confirmation that allows for confirmation by observations that are not predicted by the theory in question (to be called “non-empirical confirmation”). The relevance of non-empirical confirmation reaches far beyond the context of fundamental physics, however, and may throw new light on a number of core debates in the philosophy of science.