This article presents a study of polyethylene glycol 1000 (PEG1000)/potassium phosphate aqueous twophase systems (ATPSs) for Bacillus subtilis NS99 alkaline protease extraction. The objectives were to evaluate effects of system pH (7.5, 8.5, 9.5, and 10.5), and NaCl concentration (0, 4, 7, and 10% (w/w)) on ATPS binodal curves, effects of system pH, NaCl concentration, and tie-line length (TLL) on alkaline protease partition coefficient (K) and yield (Y%) at room temperature (30±2 oC). Casein hydrolysis was used for determination of alkaline protease activity. It was revealed that system pH had the slightest effect on locations of binodal curves (except at pH 10.5). In contrast, addition of NaCl appeared to have a significant effect on phase characteristics since binodal curves of systems with NaCl (4-10% (w/w)) shifted significantly towards the origin in comparison to the ones without NaCl. Increased NaCl concentration from 4 to 10% (w/w), however, showed trivial influence on locations of the binodal curves. Changes of system compositions due to variation in system pH, TLL, and NaCl concentrations obviously resulted in varied obtainable K and Y% of alkaline proteases. Longer TLL and higher pH generally resulted in higher K. In contrast, the lower NaCl concentration, the higher K. Since the same phase volume ration (1 : 1) was used throughout the experiments, Y% depended solely on K. The most suitable PEG1000/potassium phosphate ATPS was determined at pH 9.5, and comprised PEG1000, potassium phosphate, and NaCl 18.0, 13.0, and 0% (w/w), respectively. This system resulted in considerably high K, and Y% of 20.0, and 95.1%, respectively. Information from this study will be important for further development of an ATPS extraction unit for alkaline protease recovery.