Economic theory suggests that, when designing aid programs, ordeal mechanisms that impose differential costs for rich and poor can induce self-selection and hence improve targeting ("self-targeting"). We first re-examine this theory and show that ordeal mechanisms may actually have theoretically ambiguous effects on targeting: for example, time spent applying imposes a higher monetary cost on the rich, but may impose a higher utility cost on the poor. We then examine these issues empirically by conducting a 400-village field experiment within Indonesia's Conditional Cash Transfer program. Targeting in the program is usually conducted by automatically enrolling candidates who pass an asset test. We compare whether instituting an ordeal mechanism, where villagers come to a central application site to apply and take the asset test, improves targeting over the existing automatic enrollment system. Within self-targeting villages, we find that the poor are more likely to apply, even conditional on whether they would pass the asset test. On net, self-targeting villages have a much poorer group of beneficiaries than status quo villages. However, marginally increasing the ordeal does not necessarily improve targeting: while experimentally increasing the distance to the application site reduces the number of applicants, it screens out both rich and poor in roughly equal proportions. Estimating the model structurally, we show that only one would need to increase the ordeal dramatically (e.g. tripling wait times to 9 hours or more) to induce detectable additional selection. In short, ordeal mechanisms can induce self-selection, but marginally increasing the ordeal can impose additional costs on applicants without necessarily improving targeting.