Abstract
The bioethics literature on "tainted data" (medical knowledge produced through atrocity) has largely settled one question: such data may be used, if at all, only with explicit condemnation of the methods that produced it. It has left a second question unanswered: what operational form should that condemnation take? This article addresses both questions through paired histories. Fritz Haber's nitrogen fixation, which feeds much of the world, and the frostbite and pathogen experiments of the Japanese Imperial Army's Unit 731, purchased by the United States with grants of immunity, mark the two poles of the problem. I first propose methodological dependence as the criterion of severability: science is separable from the wrongdoing of its makers only when the wrongdoing does not constitute the finding itself. Haber's chemistry passes the test; Unit 731's calibrations cannot. I then argue that the required condemnation should be operationalized grammatically: citations of methodologically dependent data must refuse the passive voice, naming perpetrator, act, and victim. Against the objections that such a norm would chill citation or calcify into ritual, I argue that it prices citation honestly while keeping perpetration permanently legible in the scientific record.