This symposium takes place at the intersection of an acute success (the extraordinary recent developments in genetical research relevant to mycobacteria) and a chronic failure (the embarrassing failure of Bacillus Calmette-Guérin [BCG], the world's most widely used vaccine, to control tuberculosis). I will summarize the arguments that the variable efficacy of BCG is attributable, at least in part, to genetic differences in either the vaccine, the host populations or Mycobacterium tuberculosis. I will then address the problem of building upon our experience with BCG in order to develop and evaluate an improved vaccine against tuberculosis. The fact that the great burden of tuberculosis is associated with pulmonary disease in adults, much of which is associated either with reactivation of old infection or else with relatively recent reinfection, poses an immense problem for vaccine strategy and evaluation. One approach may be to vaccinate early in life, prior to initial infection with the tubercle bacillus, and then to follow up individuals for many decades. The alternative is to develop vaccines to boost appropriate protective immune mechanisms in individuals who have already been infected, and/or who have already been vaccinated (e.g. with BCG), and/or who may also have been sensitized to various environmental mycobacteria that influence the immune response to tubercle bacilli. The latter approach is novel, and may ultimately prove impossible, but its potential logistic and public health advantages are so great that it warrants serious attention by the research community before being abandoned.