Arsenic contamination of drinking water is a serious problem in many parts of the world. The precipitation/coprecipitation method was used for arsenic removal from drinking water by ferric chloride, ferric sulfate and ferrous sulfate as coagulant. A Box-Behnken statistical experiment design method was used to investigate the effects of major operating variables such as initial arsenate concentration (10-1000 microg L(-1)), coagulant dose (0.5-60 mg L(-1)) and pH (4-9) were investigated. Experimental data were used for determination of the response functions coefficients. Predicted values of arsenate removal obtained using the response functions were in good agreement with the experimental data. Fe(III) ions were more effective and economic than Fe(II) ion due to required lower coagulant dose and pH. In the low initial arsenate concentrations, the highest arsenate removal efficiency was required high ferric chloride and ferric sulfate dose of 50 and 40 mg L(-1), while in the high initial arsenate concentrations, the highest arsenate removal efficiency was provided at low ferric chloride and ferric sulfate dose of 37 and 32 mg L(-1), respectively. This study showed that Box-Behnken design and response surface methodology was reliable and effective in determining the optimum conditions for arsenic removal by coagulation and flocculation.