Direct microscopy is the only available method for diagnosis of tuberculosis in most centres in developing countries. Methods to improve the sensitivity of direct smear are an urgent requirement. Sputum specimens artificially seeded with known concentrations of Mycobacterium tuberculosis were liquefied and decontaminated with sodium hydroxide-sodium citrate-N-actyl-L-cysteine solutions. They were subjected to different centrifugation forces and centrifugation times after which the centrifuged deposits were examined by smear and culture. Statistical analysis of results was carried out using EpiInfo version 6.0. The optimum relative centrifugal force (RCF) and centrifugation time combination was 4000 g for 15 min. The sensitivity of detection at an RCF of 4000 g for 15 min was 5000 organisms/mL and 500 organisms/mL for smear and culture, respectively. When results of 163 clinical samples were analyzed after centrifugation at 4000 g for 15 min sensitivity of the direct smear improved from 63% to 92% (P < 0.05) and negative predictive value from 30.5% to 45% (P < 0.05) when culture was considered the 'gold standard'. With the concentrated smear there was a reduction in specificity from 82% to 60% (P > 0.05). As most laboratories are equipped with a simple centrifuge, smear sensitivity can be improved with this simple modification. The other advantage is that the same centrifuged deposit can be cultured, in contrast to when sodium hypochlorite is used for liquefaction.