Estimates of the incidence of pulmonary tuberculosis in developing countries are based on case reporting from local health laboratories or the annual risk of tuberculin skin test conversion. Because these methods are problematic, the authors used a multiple case ascertainment method to estimate the incidence of pulmonary tuberculosis from 1989 to 1993 in a Peruvian shantytown of 34,000 inhabitants. Two methods, face-to-face interview of all local inhabitants and examination of local laboratory smear records, were used for case gathering. The number of missed cases was estimated by capture-recapture analysis. Survey cases with positive smears were matched to age- and sex-matched controls and interviewed about socioeconomic conditions. The average annual incidence per 100,000 population was 364 (95% confidence interval 293-528) by capture-recapture methods. For the city encompassing the shantytown, the Peruvian Ministry of Heath reported an average annual incidence of 134 cases per 100,000 population. The authors conclude that, in Peru, alarming clusters of pulmonary tuberculosis are masked by government reports that pool zones of disparate incidence. Existing estimators of pulmonary tuberculosis incidence based on tuberculin conversion rates may be invalid in such areas. Within these hyperendemic areas, persons suitable for intensive prophylaxis efforts cannot be reliably identified by housing and socioeconomic risk factors.