Previous studies of the magnetic field de- Solvent proton relaxation in protein solutions can be accounted for reasonably well within the framework of a simple model of water exchange between the bulk solvent, in which the protons are assumed not to interact with the solute, and regions on or near the protein where the rotational motion of the protein molecules is conveyed to the water molecules. If it is assumed that the water molecules in the latter region attach rigidly to the protein, then one can readily estimate (1) the number n of such attachment sites to be about 3 for lysozyme and to increase at least as fast as the surface area for larger proteins (K. Hallenga and S. H. Koenig, to be published). If the attachment is assumed not rigid, the derived value of n is greater by a small factor, but is still-and this is the major unsettling feature of this simple two-site model-only a few percent of the number of water molecules in the first hydration layer (1). The nature of such special sites is difficult to conceive.From the data for proton relaxation, one can only set fairly wide limits on the value (or distribution of values) of TM, the mean lifetime of a water molecule on the protein and a quantity which is a feature of the model itself. The application of the model gives TM > TR 10-7 sec, and TM < Tipr -10-4 sec, the relaxation time of the protons when on the protein. In an attempt to get a better measurement of TM, as well as to test other predictions of the models, we measured and report here comparisons of the magnetic field dependence of the relaxation rate of 170 in solutions of protein in 17O-enriched water, and deuterons in 2H20 solutions, with the proton data. Since deuterons in water relax about 10 times faster than protons, and 170 nuclei about 100 times faster than deuterons, the upper limit on the model value for TM is drastically reduced and the range for TM is drastically narrowed. We have also made measurements on an extremely large protein-hemocyanin associated to a M.W. of 9 X 106-in order to increase the lower limit of TrM. The experiments were quite successful in that -TM was squeezed into nonexistence: i.e., the data reported below cannot be explained by a simple two-region model with a defined life-