Statistical optimization of culture conditions for production of a widely suited detergent protease from Bacillus sp. RGR-14 was carried out using a two-step approach. A quick identification of the important factors with simple screening experiment was followed by application of complex response surface design for further optimization. The production of extracellular alkaline protease by Bacillus sp. was favored in the presence of complex carbon and nitrogen sources, viz. starch, casamino acid and soybean meal. A reduced quadratic model was found to fit the alkaline protease production. Response surface analysis revealed the significant role of phosphate ions in determining alkaline protease production. A steep, stretched out response surface showed direct relation between the level of protease production and casamino acid and starch concentration in the medium. A 12.85 fold increase in protease production could be obtained within the design space. Protease production was found to be repressed in the presence of high concentrations of casamino acid. The model could be validated in up to 21 shake flasks (3914 U ml(-1)). The same statistical design could explain economic protease production in cost-effective medium as well. (C) 2003 Elsevier Ltd. All rights reserved.