It is widely held that individuals who are unable to provide informed consent should be enrolled in clinical research only when the risks are low, or the research offers them the prospect of direct benefit. There is now a rich literature on when the risks of clinical research are low enough to enroll individuals who cannot consent. Much less attention has focused on which benefits of research participation count as ‘direct’, and the few existing accounts disagree over how this crucial (...) concept should be defined. This disagreement raises concern over whether those who cannot consent, including children and adults with severe dementia, are being adequately protected. The present paper attempts to address this concern by considering first what additional protections are needed for these vulnerable individuals. This analysis suggests that the extant definitions of direct benefits either provide insufficient protection for research subjects or pose excessive obstacles to appropriate research. This analysis also points to a modified definition of direct benefits with the potential to avoid these two extremes, protecting individuals who cannot consent without blocking appropriate research. (shrink)