To be of use in the control of tuberculosis, any new drug must be capable of shortening the duration of treatment by accelerating sterilizing activity, that is the rate at which Mycobacterium tuberculosis is killed in the lesions. The most difficult to kill are the extra-cellular bacilli in cavities. Persistence during therapy arises because there is a proportion of slowly metabolising bacilli (persisters) in the cavitary bacterial population at the start of treatment. Bacterial growth is slowed by low oxygen tension, quorum sensing and old age, but probably not by cellular immunity, since there are few professional phagocytic cells in cavities. The degree of phenotypic resistance to the bactericidal action of drugs can go through several stages: (i) the non-replicating stages 1 and 2 of micro-aerophilic adaptation, described by Wayne; (ii) a "tolerant" population that survives exposure to high rifampicin concentrations and is capable of growth in liquid medium but not on solid medium; and (iii) a population found in the sterile state of Cornell model mice which cannot grow initially in either liquid or solid medium but will eventually cause re-activation of tuberculosis. In all of these stages the bacilli are phenotypically resistant; there is no selection for genomic drug resistance. Rifampicin and pyrazinamide are the two drugs largely responsible for sterilizing activity during current treatment. Pyrazinamide is unique amongst anti-tuberculosis drugs in having no genomic site of action and having greater bactericidal activity as bacillary metabolism slows down; it is remarkably effective in human disease. The development of a new drug with a similar mode of activity might be very fruitful, especially if there were no need for an acid environment. Current methods advocated for drug development pass through a number of complex stages: choice of a genomic target, development of an in vitro assay, high throughput screening and identification of lead compounds, often with scaling up of synthesis of the molecule and preliminary studies of toxicity and animal pharmacology before tests are done for sterilizing activity. If the drug is not good at sterilizing, all of this initial work will be largely wasted as it would only have a very limited role in the treatment of MDR disease. One of the most important steps necessary is the development of rapid and simple tests to screen for sterilizing activity. Of tests currently available, none of those employing mice seem adequate, though a screen using a streptomycin dependent Mycobacterium tuberculosis seems the most hopeful. A set of in vitro tests is described. There is an urgent need to develop these tests further since the factors slowing growth are closer to those in tuberculous cavities than in mouse models. They have the advantages of simplicity and require only small amounts of a new molecule.