Self-control problems constitute a potential explanation for the underinvestment in preventive health in low-income countries. Behavioral economics offers a tool to solve such problems: commitment devices. We conduct a field experiment to evaluate the effectiveness of different types of theoretically motivated commitment contracts in increasing preventive doctor visits by hypertensive patients in rural India. Despite achieving high take-up of such contracts in some treatment arms, we find no effects on actual doctor visits or individual health outcomes. A substantial number of individuals pay for commitment but fail to follow through on the doctor visit, losing money without experiencing health benefits. We develop and structurally estimate a prespecified model of consumer behavior under present bias with varying levels of naiveté. The results are consistent with a large share of individuals being partially naive about their own self-control problems: sophisticated enough to demand some commitment but overly optimistic about whether a given level of commitment is sufficiently strong to be effective. The results suggest that commitment devices may in practice be welfare diminishing, at least in some contexts, and serve as a cautionary tale about their role in health care.