Imagine a patient, Bill, who has lost decision-making capacity. An important treatment decision must be made, but Bill cannot make it. Assume there is no advance directive or anything equivalent. There is no way to take a statement made by Bill, when competent, and apply it to the current setting. A standard move in cases like this is to have the patient’s surrogate decision-maker—Bill’s life-partner, for example—enter the picture and answer the question, “In this situation, what would Bill choose?” Bill’s partner believes she knows: given the life he has lived, he would not want to live with severely limited mental capacities, and he would decline a procedure likely to leave him in this condition. The medical team is ready to accept her decision. Something seems very right about taking this route, but what makes this the right thing to do? Bioethics orthodoxy sees the patient’s autonomy as at stake here, but in fact this is not so, at least not as “autonomy” is usually understood in clinical contexts. Instead, what underpins the moral force of the “What would the patient choose?” question is a value that is related to—but not the same as—either autonomy or that other common bioethical consideration, the patient’s best interests.