Previous studies of the magnetic field dependence of the magnetic relaxation rate of solvent protons in protein solutions have indicated that this dependence (called relaxation dispersion) is related to the rotational Brownian motion of the solute proteins. In particular, the dispersion of the longitudinal (spin-lattice) relaxation rate 1/T1 shows a monotonic decrease with increasing field, with an inflection point corresponding to a proton Larmor frequency which is inversely proportional to the orientational relaxation time of the protein. We have now compared the relaxation dispersion of solvent 1H, 2H, and 17O In aqueous solutions of lysozyme (molecular weight 14,700) and 1H and 2H in solutions of hemocyanin (molecular weight 14,7 00) and 1H and 2H in solutions of hemocyanin (molecular weight 9 x 10(6)). The main experimental observation is that the dispersion of the relaxation rates of the three solvent nuclei in lysozyme solutions, normalized to their respective rates in pure water, is essentially the same. This is also true for 1H and 2H relaxation in hemocyanin solutions. These results confirm that entire solvent water molecules, rather than exchanging protons, are involved in the interaction. We have been unable to deduce the correct mechanism to explain the data, but we can eliminate several interaction mechanisms from consideration. For example, all observations combined cannot be explained by a simple two-site model of exchange, in which water molecules are either in sites on the protein with a relaxation rate characteristic of these sites, or else in the bulk solvent (the observed relaxation rate being the weighted average of the two). Also eliminated is the class of models in which the protein molecules induce a preferential partial alignment of neighboring solvent molecules, for example by electrostatic interaction of the electric dipole moments of the water with the electric fields produced by surface charges of the protein molecules. In addition, the idea that relaxation of solvent nuclei is due, in the main, to interactions with protein protons is precluded. Rather, it appears that the protein molecules influence the dynamics of the motion of solvent water molecules in their neighborhood in a manner that imposes on all the solvent molecules a correlation time for their orientational relaxation which equals that of the solute proteins.