The partitioning of xylanase produced by Penicillium janthinellum in aqueous two-phase systems (ATPS) using poly(ethylene glycol) (PEG) and phosphate (K2HPO4/KH2PO4) was studied employing a statistical experimental design. The aim was to identify the key factors governing xylanase partitioning. The interactions of five factors (PEG concentration molecular weight, concentration of buffer K2HPO4/KH2PO4, pH, and NaCl concentration) and their main effects on the partition coefficient (K) were evaluated by means of a 2(5) full-factorial experimental design with four center points. The %PEG, %NaCl, and pH were the most important factors affecting the response variable (K). Response surface methodology (RSM) was adopted and an empirical second-order polynomial model was constructed on the basis of the results. The optimum partition conditions were pH 7.0, PEG = 8.83% and NaCl = 6.02%. Adequacy of the model for predicting optimum response value was tested under these conditions. The experimental xylanase partition coefficient (K) was 2.21, whereas its value predicted by the model was 2.33. These results indicate that the predicted model was adequate for the process. PEG molecular weight and phosphate concentration did not affect the xylanase partition coefficient.