In many domains risky health behavior is still only poorly understood. Analysis is often plagued by incomplete data and a general lack of information. In this study; we try to understand the determinants of helmet use among motorcyclists in Delhi; a context in which road safety is very low. We use a very detailed data set collected especially for the purpose of the study. To guide our empirical analysis; we rely on a simple model in which drivers decide on their speed and helmet use. The empirical findings suggest that risk averse individuals are more likely to wear a helmet. We do not find any systematic effect of risk aversion on speed. Both findings are coherent with our theoretical model. Helmet use also increases with education. Drivers who show a higher awareness of road risks; because for instance; they are better informed about Delhi's actual road traffc accident fatality and injury rates; are both more likely to wear a helmet and to speed less. In turn; those drivers who show a high level of unawareness take the highest risks. Controlling for risk awareness; we observe that drivers tend to compensate between speed and helmet use. The most obvious solution to India's road safety problem and the related high social costs that result from it is to enforce the helmet law and speed limits. An alternative strategy; and probably more feasible in the current context; is to design interventions which raise awareness of road risks. Improvements to the road infrastructure are also a possible solution but these measures bear the risk that drivers will react to the improved road safety by either increasing speed or lowering helmet use.