A statistical thermodynamic model for the prediction of precipitation curves of globular proteins using nonionic polymers has been proposed. The model accounts for protein-polymer, polymer-solvent, electrostatic, and hydrophobic interactions as well as the entropy of mixing and employs simplifying assumptions such as spherical globular protein molecule with uniform surface properties and linear, homogeneous polymer uniform with respect to molecular weight. The proposed model can only be employed to predict precipitation curves of charged proteins at sufficiently high ionic strengths since it does not account for electrostatic protein-protein interactions due to overlap of electrical double layers. The model predictions of precipitation curves of human serum albumin (HSA) at the isoelectric point using polyethylene glycol (PEG) for different initial protein concentrations and molecular weights of PEG agreed well with the experimental data. Higher polymer concentrations were found to be required to precipitate proteins for lower molecular weight polymers, lower initial protein concentrations, and more favorable protein-polymer interactions. The HSA-PEG interaction parameter, obtained by fitting the model to experimental data for one molecular weight PEG, was found to be 0.122. Solubility of HSA in PEG solution was found to decrease with increasing salt concentrations, this effect being more pronounced at lower PEG concentrations. The net charge on HSA was found to result in a maximum in its solubility at intermediate salt concentrations as a result of competing salting-in and salting-out effects.