In this article, I provide a new account of the basis of medical researchers' ancillary care obligations. Ancillary care in medical research, or medical care that research participants need but that is not required for the validity or safety of a study or to redress research injuries, is a topic that has drawn increasing attention in research ethics over the last ten years. My view, the relationship-based approach, improves on the main existing theory, Richardson and Belsky's ‘partial-entrustment model’, by avoiding its problematic restriction on the scope of health needs for which researchers could be obligated to provide ancillary care. Instead, it grounds ancillary care obligations in a wide range of morally relevant features of the researcher-participant relationship, including the level of engagement between researchers and participants, and weighs these factors against each other. I argue that the level of engagement, that is, the duration and intensity of interactions, between researchers and participants matters for ancillary care because of its connection to the meaningfulness of a relationship, and I suggest that other morally relevant features can be grounded in researchers' role obligations.