Despite over 30 years of intermittent investigations, there is no definitive evidence that macrophages kill Mycobacterium tuberculosis. This is largely because all attempts to demonstrate such killing by macrophages in vitro have produced only rather unconvincing results. How then can I discuss how macrophages kill tubercle bacilli when it is not proven that they do? The answer is that just as there is a large and growing body of indirect evidence that makes it highly likely that macrophages can kill tubercle bacilli, there is also a body of evidence that makes it likely that they do so by producing hydrogen peroxide.I will spend most of this lecture summarising the more recent body of evidence that macrophages kill tubercle bacilli by producing hydrogen peroxide but I will also refer briefly to the question whether macrophage lysosomes have any direct activity against tubercle bacilli, for it is unlikely that macrophages are dependent solely on one antimycobacterial product. Different macrophage products may be of crucial importance under different circumstances.The peroxide story, in relation to host defence against M . tuberculosis, begins around 1953. After the chemotherapeutic drug isoniazid had been in use for a few years, reports began to appear of the isolation of isoniazid-resistant mutants of M . tuberculosis that had low virulence in the guinea-pig, were catalase-negative and were susceptible to killing by hydrogen peroxide (Barnett, Bushby and Mitchison, 1953;Cohn et al., 1954;Mitchison, 1954;Morse et al., 1954;Peizer and Widelock, 1955). At that time hydrogen peroxide production by mammalian tissues had not even been demonstrated but the implication that peroxide might reach tuberculocidal concentrations inside leucocytes was not missed by Coleman and Middlebrook (1956). This idea was strengthened considerably in the early 1960s by reports that a substantial proportion of isolates of M . tuberculosis from patients in the Indian subcontinent were of low virulence in the guinea-pig and peroxide-susceptible even though they had a normal catalase content and isoniazid susceptibility (Subbaiah, Mitchison and Selkon,