Tuberculosis vaccine candidates are entering clinical studies in areas where BCG fails. This is a high-risk strategy. We suggest that geographical variation in the efficacy of BCG is related to the presence in developing countries of a cross-reactive background Th2-like response, probably attributable to exposure of mother and infant to helminths and environmental mycobacteria. Such Th2-like activity can stop Mycobacterium tuberculosis from being pushed into a latent state by the Th1 response, impair bactericidal functions and cause toxicity of TNF-alpha and pulmonary fibrosis. A successful vaccine, rather than driving a Th1 response, might need to suppress this pre-existing subversive Th2-like component.