The pH and hydraulic retention time (HRT) of a chemostat reactor were varied according to a central composite design methodology with the aim of modeling and optimizing the conversion of starch into hydrogen by microorganisms in an anaerobic digested sludge. Experimental results from 23 runs indicate that a maximum hydrogen production rate of 1600 L/m(3)/d under the organic loading rate of 6 kg starch m(3)/d obtained at pH = 5.2 and HRT = 17 h. Throughout this study, the hydrogen percentage in the biogas was approximately 60% and no methanogenesis was observed. while the reactor was operated with HRT of 17 h, hydrogen was produced within a pH range between 4.7 and 5.7. Alcohol production rate was greater than hydrogen production rate if the pH was lower than 4.3 or higher than 6.1. Supplementary experiments confirm that the optimum conditions evaluated in this study were highly reliable; while a hydrogen production yield of 1.29 l H-2/g starch-GOD was obtained. An examination of the response surfaces, including hydrogen, volatile fatty acids (VFA) and alcohols production, led us to the belief that clostridium sp. predominated in the anaerobic hydrogen-producing microorganisms in this study. Experiment results obtained emphasize that the response of metabolites was a more useful indicator than hydrogenic activity for obtaining efficient hydrogen production. Furthermore, expressions of contour plots indicate that Response-Surface Methodology may provide easily interpretable advice on the operation of a hydrogen-producing bioprocess. (C) 2000 John Wiley & Sons, Inc.