There is a longstanding debate over the implications of natural and vaccine-induced delayed type hypertensivity for protective immunity to mycobacterial infections. The identification of correlates of vaccine-induced protective immunity should help explain the inconsistent behaviour of BCG vaccines in different populations and assist in efforts to devise improved vaccines. More than 70,000 subjects in Karonga District, northern Malawi were skin tested with soluble antigens of the tubercle and leprosy bacilli, and then followed up for five years for tuberculosis and leprosy incidence. Incidence rate ratios were calculated to compare subjects with different levels of prior skin test sensitivity, after controlling for the effects of age, sex and previous BCG vaccination. BCG vaccination protected against leprosy without persistent delayed-type hypersensitivity to tuberculin or to soluble antigens of the leprosy bacillus. In subjects who had not received BCG, hypersensitivity to tuberculin or to antigens of the leprosy bacillus was associated with strong protection against leprosy. In BCG-vaccinated and unvaccinated subjects, there was a J-shaped relation between hypersensitivity to tuberculin and subsequent rates of tuberculosis, with lowest rates associated with low grade sensitivity (induration 1-10 mm). This study shows that delayed-type hypersensitivity to mycobacterial antigens has different implications for tuberculosis and leprosy: low-level hypersensitivity (probably attributable to environmental mycobacteria) is associated with protection, but persistent vaccine-associated hypersensitivity to mycobacterial antigens is not a correlate of vaccine-derived protection against mycobacterial diseases.