The early bactericidal activity (EBA) of an antituberculosis agent is arbitrarily defined as the fall in log(10) colony forming units (cfu) of Mycobacterium tuberculosis per ml sputum per day during the first 2 days of treatment. Determining the EBA is an important preliminary step in the clinical evaluation of an antituberculosis agent. We review the results of eight published studies of the EBA of different antituberculosis agents, the impact of these results on our understanding of the actions of the respective agents, the clinical characteristics and sputum findings of patients included in these studies, and explore sources of variation in the EBA results. Patients in these studies had a mean age of 31-36 years, a mean weight of 50-57 kg, 67% were male and 56% had lung involvement covering an area of more than one lung, and 90% had multicavitary disease. None of these findings were related to EBA in any study. The mean log(10) cfu per ml sputum in the first specimen was 6.474. This was related to radiological extent of disease and cavity size in one study (p < 0.001) and, in the case of isoniazid to EBA with a rise in EBA of 0.094 (95% CL 0.029-0.158) for each tenfold rise in cfu counts/ml sputum. The overall variation in EBA in these studies was 0.0303, that due to laboratory processing of specimens was 0.0011, and due to patient characteristics and sputum sampling 0.0212. The EBA is a reproducible investigation that has contributed significantly to our knowledge of the actions and characteristics of both established and new antituberculosis agents. The greatest source of variation in EBA results appears to be that due to interpatient variation in disease characteristics and sputum sampling.