In a field trial in Gadchiroli, India, we trained 30 paramedical workers (PMWs), 25 village health workers (VHWs) and 86 traditional birth attendants (TBAs) from 58 villages to diagnose childhood pneumonia and treat it with sulfamethoxazole+trimethoprim. Continued training, the development of a breath counter, and educative supervision progressively reduced errors in case management made by the TBAs. Over the 3.5-year period 1988-91, 2568 attacks of childhood pneumonia were managed and the case fatality rate was 0.9%, compared with a rate of 13.5% in the control area. The case fatality rates for the three types of worker were similar. The TBAs were superior to the other workers in terms of their availability, outreach, access to neonates, and cost. Satisfaction with the VHWs, and PMWs was expressed by 85%, 69% and 18% of users, respectively. In the intervention area the mortality rate attributable to pneumonia among neonates declined by 44% (P < 0.01) while the total neonatal mortality fell by 20%, presumably because of the involvement of TBAs in the control of acute respiratory infections (ARI). If adequately supported by the health system, TBAs can successfully manage childhood pneumonia in villages at the lowest possible cost and with a high degree of community acceptance. TBAs and VHWs are the most suitable community-based health workers for ARI control programmes in developing countries.