Optimization of traditional processing of soybeans using response surface methodology (RSM) to achieve a minimum level of antinutritional factors (ANFs) in kinema. Central composite rotatable designs were used to optimize the processing stages of kinema preparation. In each stage, the linear or quadratic effects of independent variables were significant in minimizing ANF levels. The predicted optimum condition for soaking was when the raw beans-water ratio was 1 : 10, and the soaking temperature, time and pH were 10°C, 20 h and 8·0 respectively. Here, tannins content (TC), phytic acid content (PAC) and trypsin inhibitor activity (TIA) decreased (P < 0·05). While haemagglutinating activity (HA) level remained unchanged (P < 0·05), total biogenic amines content (TBAC) increased. The optimum condition for cooking was optimally soaked beans-water ratio of 1 : 5, and cooking pressure and time were 1·10 kg cm(-2) and 20 min respectively. Here, TC, PAC, TIA and HA decreased (P < 0·05), but TBAC remained unchanged compared to optimally soaked beans. TC and HA went below the level of detection. The optimum condition for fermentation was obtained when inoculum load was 10(3) total cells g(-1) grits, and fermentation temperature and time were 37°C and 48 h respectively. Fermentation of optimally cooked beans caused a reduction (P < 0·05) of PAC. While TIA remained unchanged (P < 0·05), TBAC increased. In kinema, TC, PAC, TIA and HA decreased (P < 0·05) over raw beans by 100, 61, 71 and 100% respectively. Good agreement was observed between predicted values and experimental values. The processing treatments significantly minimized the level of ANFs in soybeans. RSM was successfully deployed to obtain the optimum condition for kinema-making with a minimum level of ANFs without impairing sensory attributes of the product. The results are useful for commercial production of kinema.