In most developing countries, bacterial meningitis (BM) is associated with a high case-fatality rate. The search for a simple, convenient, and inexpensive antibiotic treatment remains a priority. In this study, a non-blinded, multicentre, randomised clinical trial of 528 cases of BM was done in two hospitals in Mali and Niger, between March, 1989, and May, 1990, to see whether a double injection of long-acting chloramphenicol (on admission to hospital and 48 h later) is as effective as a course of intravenous ampicillin (8 days, 4 times a day). The cumulative case-fatality rate on day 4 (principal end-point) among the chloramphenicol (254 patients) and ampicillin (274) groups were, respectively, 28% and 24.5% (relative risk 1.14, 95% confidence interval 0.86-1.52). No outbreak occurred during the study period. The hospital case-fatality rate was 33.1%. Main risk factors for death were associated with clinical condition on admission--ie, altered consciousness, convulsions, or dehydration. The case-fatality rates were 13% (21/161) for Neisseria meningitidis, 36.1% (48/133) for Haemophilus influenzae, and 67% (77/115) for Streptococcus pneumoniae. In a multiple logistic regression model, controlling for the differential distribution of potential risk factors (including bacterial species), there was no difference between treatment groups. Our findings suggest that long-acting chloramphenicol is a useful first-line presumptive treatment for BM in high-incidence countries.