The introduction of the sulphonamides in the 1930s was seen as an immensely important advance in the treatment of a wide range of bacterial diseases. But there had been so many false alarms and disappointments with other treatments in the past that it was clear that formal clinical trials were needed to assess these new drugs. At the time, most clinical trials relied on comparing morbidity or mortality in current patients treated with a new treatment with that in apparently similar, past patients – ‘historical controls’ – who had either received another treatment, or no treatment. Comparisons of current patients given new treatments with ‘historical controls’ can yield trustworthy results when the differential effects of the two treatments are dramatic.1 But it began to be recognized during the 1930s that such comparisons were not a reliable indicator of moderate treatment effects and differences. In these circumstances, deliberate steps were needed to ensure that the patients in treatment comparison groups were similar, so that ‘like would be compared with like’. Reflecting a growing understanding of this principle, a growing number of clinical trials reported during the 1930s and 1940s used alternation or random allocation of patients to treatment comparison groups, with the objective of ensuring that ‘like would be compared with like’. The history of research on the sulphonamides during the 1930s and 1940s neatly illustrates the need for these different kinds of clinical trials. Studies using historical controls were sufficient to provide convincing evidence that sulphonamides had a dramatic effect on mortality from puerperal fever2–4 and meningococcal meningitis.5 More carefully controlled trials, using alternation of patients to treatment and control groups, showed that these new drugs also had worthwhile effects on erysipelas,6,7 pneumonia8–10 and plague.11 However, studies using alternation also suggested that sulphonamides were unlikely to be useful for scarlet fever12,13 and some gastrointestinal disorders.14,15 In brief, the design of fair tests of the effects of sulphonamides depended on how large an effect, if any, these drugs had on each of the diseases for which they were considered. When the role of bacteria in infective diseases had become firmly established in the late nineteenth century there were, broadly speaking, two approaches to treatment. The first, and by far the largest, was a series of attempts to attack bacteria by techniques of passive or active immunization, and many of the earliest trials using alternation to generate comparison groups were done to assess the effects of such immunization.16–21 Passive immunization consisted of injecting a patient suffering from a bacterial disease with serum from an animal (most often a horse) which had been given a series of injections of the bacterium in question to provoke the production of antibodies. The horse was, in effect, used as a factory for making antibodies which were given to the patient as injections of horse serum. The major snag of such passive immunization was that horse serum was apt to produce dangerous anaphylactic reactions, known at the time as ‘serum sickness’. Active immunization consisted of injecting the patient with bacteria which had been attenuated in such a way that they retained their ability to provoke antibody formation, but were too weak to provoke illness. Here, there was no danger of serum sickness, but it takes a considerable time for the body to produce antibodies. Active immunization of healthy people is therefore an effective way of preventing a disease, but an ineffective form of treatment since the patient may die before active immunization has had time to produce immunity. Immunotherapy dominated the treatment of infective disease during the first thirty years of the twentieth century, but another and totally different approach was being developed simultaneously. It was an approach which became known as chemotherapy – a term which was first used by one of its greatest proponents, Paul Ehrlich (1854–1915). Although Ehrlich had begun his research with antitoxins, he soon developed a line of reasoning which went something like this: it was known that dyes could be used to stain bacteria on a microscope slide; if dyes could stain bacteria in the laboratory, there might be dyes which could be administered as drugs and stick to bacteria in the human body, killing the bacteria in the process but doing no harm to the patient. Such a drug would, Ehrlich suggested, resemble a ‘magic bullet’. Ehrlich had limited success in his search for a magic bullet, but his reasoning explains what at the time seemed very odd indeed – that there might be a link between bacteriology and dyes which were manufactured for the purpose of colouring fabrics and clothing.