["Bioethics, EarlyView. ", "\nABSTRACT\nEthical human subjects research must have an acceptable risk‐benefit ratio, which requires that the net risks participants face be proportionate to the research's social value. Yet existing scholarship does not explain what makes risks proportionate to social value. Here, I distinguish internal and external senses of proportionality and argue that it is internal proportionality that is the proper object of IRB/REC review. I propose understanding internal proportionality in terms of a literal ratio of expected health effects; I also argue against the alternative of attempting to assess proportionality through intuitive comparisons. I illustrate the potential implications of this account through a discussion of the proposal to engage in controlled human infection studies with the novel coronavirus. This case highlights that trials with enormous numbers of potential beneficiaries can be internally proportional even when they are very unlikely to succeed.\n"]