We assessed the utility of bronchoalveolar lavage (BAL) in the diagnosis of pulmonary tuberculosis (PTB) in 50 consecutive HIV-negative patients with clinical and radiographic findings suggestive of PTB, but with negative microscopy for acid-fast bacilli (AFB) on sputum smear. Patients were grouped, using a scoring system, into relative likelihoods of having PTB (I-IV, in descending probability). Patients were started on anti-tuberculosis treatment according to the BAL results. Bacteriological diagnosis of PTB was confirmed in 22/50 BAL; 11 (91.6%), seven (37%) and four (40%) of groups I-III, respectively. In 13 cases, an early diagnosis of PTB was made by positive microscopy for AFB on BAL; an alternative diagnosis was made in six cases (bacterial pneumonia 4, carcinoma 2). A decision analysis model was created to assess the overall utility of BAL. This suggested that in a region of high PTB prevalence, and when the clinical diagnosis of PTB is likely, empirical treatment is the best course of action, with BAL being reserved for further investigation of non-responders. Early BAL should be considered when the diagnosis of PTB is uncertain.