We move from something that does work and has much to tell us about molecular mechanism, to something that does neither, but is (astonishingly) widely used in the NHS [4]. In the present issue Paris et al. describe a well constructed add-on randomised controlled trial with 3 arms: a double blind homeopathic remedy versus placebo, plus an open label non-interventional control group [5]. Treatment was administered the evening before knee ligament surgery and continued for three days. The primary endpoint was based on the amount of morphine delivered by patient controlled analgesia during the first twenty-four hours. Outcomes in treated and placebo groups were similar both for the primary endpoint and for secondary endpoints (subsequent morphine intake, visual analogue pain score and quality of life). These parameters were also similar in patients enrolled in the open label non-interventional arm. Proponents of homeopathy would probably object to the conclusion that homeopathy is not better than placebo in reducing morphine consumption after surgery, acceding only that a certain homeopathic remedy fails to be effective for a certain type of surgical pain. In an accompanying commentary, Edzard Ernst argues that if one were to accept such pleading, one would pointlessly divert considerable effort and resource. Instead, common sense and existing knowledge tell us that homeopathy is biologically implausible, that its predictions are incorrect and that the clinical evidence is largely negative. Observational data attesting to the apparent effectiveness of homeopathy are explained by the variable nature of the natural history of disease and the potential of homeopathy to act as a placebo. He concludes his commentary by referring to a strange historical episode relating to a research programme conducted during the Nazi era. The report survived the war but disappeared subsequently. Reference to this has perturbed one reader and we publish his letter together with a response from Ernst [6,7,8]. The question what to do when sound data are obtained in an unethical manner is an important and complex issue, and we may revisit it. What is, however, straightforward is that pharmacology is based in physics, chemistry, physiology, biochemistry and cell biology. Any future submissions to BJCP on this subject will require a scientifically plausible hypothesis as well as valid methodology if they are to be taken seriously: we do not anticipate a spate of publications.