Tuberculosis remains a global emergency because of our lack of understanding of the details of its pathogenesis. In the last 12 months there have been striking advances in the molecular genetics of the organism. Mutated strains of Mycobacterium tuberculosis have been used to study the genetic requirements for virulence and establishment of latency, and the biology of the interaction with host cells. Genes involved in lipid metabolism seem particularly important. The probable sites of latency within the host lungs have been identified by in situ polymerase chain reaction. The complex control by M. tuberculosis of apoptosis of T cells and macrophages has been somewhat clarified, and the data may suggest that M. tuberculosis causes death of a subset of T cells, while preserving some macrophages as hiding places with reduced microbicidal and antigen-presenting function. Similarly the demonstration of a very large relative increase in interleukin (IL)-4 and IL-13 expression, (together with IL-4delta2, the IL-4 splice variant), that correlates with lung damage, has been supported by data from flow cytometry and in situ hybridization, and indicates that a subversive T helper-2 (Th2) component in the response to M. tuberculosis may undermine the efficacy of immunity and contribute to immunopathology. Recently defined changes in metabolism of cortisol within the lesions may contribute to the development of the Th2 component. These findings underline the need to start testing vaccine candidates in models that mimic the situations in which bacille Calmette-Guerin fails, such as in the presence of latent infection, pre-existing Th2 responses to cross-reactive organisms, and stress.