The epidemiology of tuberculosis remains poorly understood. We investigated the relative importance of within-household and community transmission of infection among children aged 6 months to 14 years living in a Peruvian shanty-town. The prevalence of Mycobacterium tuberculosis exposure among 175 contact children (sharing a household with a person who had confirmed pulmonary tuberculosis) and 382 control children (living in nearby households free of active tuberculosis) was defined as the proportion of children with a positive purified protein derivative (PPD) skin-test. 97 (55%) contact children and 129 (34%) controls were PPD positive. Living in a contact household (odds ratio 1·74, 95% Cl 1·11-2·73) and age (1·11, 1·06-1·18) were significant risk factors for PPD positivity. We calculated the community infection ratio (CIR) as the odds ratio of PPD-positive controls to PPD-positive contacts: CIR=Prevalence in controls/(1-prevalence in controls) Prevalence in contacts/(1-prevalence in contacts) A low CIR therefore suggests mainly household spread of infection, whereas a high value suggests frequent transmission outside the household. The adjusted odds ratio (for age, sex, within-household correlation, and household size) was 0·40 (95% Cl 0·26-0·64), compared with values of 0·18-0·37 in studies elsewhere. Currently recommended tuberculosis control strategies are suitable for areas with low CIRs. Different strategies may be needed for areas, such as that we studied, with high values. © 1995.