It is widely maintained that a clinical trial is ethical only if some form of equipoise between the treatments being compared obtains. To be in equipoise between two treatments A and B is to be cognitively indifferent between the statement ‘A is strictly more effective than B’ and its negation. It is natural to claim that equipoise regarding A and B is necessary for randomised assignment to treatments A and B to be beneficent and non-maleficent and is sufficient for such an assignment to be fair. Cashing this out precisely is difficult, and various forms of equipoise have been discussed which consider whose equipoise is relevant to the decision. This is to make judgement of equipoise something to be managed socially, while its prima facie significance is supposedly cognitive. Recent reconstructions of equipoise-like concepts in epistemology give clues about how to understand equipoise cognitively. In this paper I examine some of this work and discuss how successful it has been. I suggest that while this work is promising, it still has far to go, and that while equipoise remains the best theory we have of the cognitive justification for clinical trials, it is nonetheless incoherent.