Routine results of direct examination of sputum smears for acid-fast bacilli from 34 laboratories in Tanzania were analysed. These represented 8 regions providing 94 laboratory-years of work; 61,580 tuberculosis suspects were evaluated with the aid of 141,371 smears. The average proportion of cases found among suspects was 18.9% (range 14.3-23.8% in the 8 regions). The number of cases missed among suspects with incomplete examinations was calculated based on the number observed among suspects with a complete set of 3 smears examined, and an incremental yield of 83.4% with the first, 12.2% with the second, and 4.4% with the third smear was estimated for the total number of expected cases. These data suggest that (i) the method frequently employed for calculating requirements for laboratory supplies in low income countries, based on the estimate that 10 suspects need examination to identify one case of sputum smear-positive tuberculosis, is generous in the context of Tanzania and (ii) under routine conditions the incremental yield from a third smear examination after 2 negative examinations is relatively small.