A thermodynamic relationship is obtained which links the effect of a cosolvent on the denaturation equilibrium of a protein to the effect of the cosolvent on the change in partial molar volume (pmv) of a protein on denaturation. The relationship uses the concept of preferential interactions and is exact for an infinitely dilute protein. Analysis of the literature data on protein volume changes suggests that many of the observed volume changes are thermodynamically inconsistent with the corresponding free energy changes, especially at low cosolvent concentrations. It is argued that the most reasonable explanation for this involves cosolvent induced changes in the degree of protein-protein association.