The first multicomponent diffusion data ever determined in protein-polymer systems are presented for the system lysozyme(l)-PEG 400(2)-water. Although there are no specific interactions between protein and polymer, the cross-term diffusion coefficient D-21, that links the PEG flow to the protein concentration gradient, is up to 35 times the main-term diffusion coefficient of the protein. This observation can only be due to a "crowding effect" and not to specific interactions, such as electrostatic ones. The exclusion effect is also qualitatively confirmed by the measured counter-flow associated with the protein motion. On the base of a hard core potential, our recent predictive equations are used to predict diffusion coefficients in this ternary system, and a good agreement with the experimental D-21 is obtained. The PEG concentration dependence of the main-term diffusion coefficient of the protein cannot be interpreted exclusively by the excluded volume effect. Some dielectric effect or aggregation phenomena must be invoked to completely describe diffusive behavior in protein-PEG systems. A strong dielectric constant decrease and an anomalous pH dependence on PEG concentration in this system have been observed. We have extended to this nonelectrolyte system a recent procedure for extracting thermodynamic data from ternary diffusion coefficients that uses the Onsager reciprocal relations and the coupling of D-ij and second virial coefficient data. Thus we obtained the change of the lysozyme chemical potential with increasing PEG concentration. We emphasize that it is incorrect to neglect the nonideality of PEG-water systems, as was done in some previous preferential solvation analyses.