An experimental design was used to determine optimal conditions for refolding of a recombinant thermostable and alkaline active xylanase from Bacillus halodurans in PEG-phosphate two-phase system. The influence of different experimental variables on the enzyme recovery has been evaluated. To build the mathematical model and minimize the number of experiments for the design parameters, response surface methodology with a face-centered central composite design (CCF) was defined based on the conditions found by preliminary tests that resulted in the highest refolding yield. The adequacy of the calculated model for the response was confirmed by means of variance analysis and additional experiments. Analysis of contours of constant response as a function of pH, polyethylene glycol (PEG) molecular weight and concentration, and salt concentration for different enzyme loads revealed different effects of these five factors on the studied parameters. Recovery of more than 92% active xylanase was predicted for a system with 18.3% (w/w) PEG 1000, 14.4% (w/w) phosphate at pH 8.5, and enzyme load corresponding to a protein concentration of about 0.05 mg/g system. The yield of the refolded enzyme was found to be optimal at 22 degrees C. The validity of the response model was verified by a good agreement between predicted and experimental results.