One of the great merits of On Complicity and Compromise is that it wades into specific swamps where ordinary theorists fear to slog. It is persuasive that in general it can be right sometimes to be complicit in wrongdoing by others through causally contributing to the wrongdoing, but not sharing its purpose, if by being involved one can reasonably expect to lessen the extent of the wrong that would otherwise be suffered by the victims. I focus on whether the book's general thesis is applicable to torture, which depends on what torture and the torture situation are in fact like. I focus on the case to which the chapter several times refers: the innovative CIA paradigm of torture. First, to the extent that the paradigm, which is predominantly mental, or psychological, torture succeeds in its goal of producing regression to a compliant state, the physician would be unable to rely on the torture victim's expressions of preferences or interests as authentically his own. Second, since disorientation plays such a large role in the CIA's style of torture (adopted at Guantanamo by the military), the authorities would refuse to allow a stable relationship to be built up with any one doctor by any victim, making comprehension of the victim's preferences difficult. Third, even if the doctor could somehow judge what the victim's genuine interests were, the control of the situation is much too totalistic to allow the physician any action independent of what the torture regime requires.