Immobilized protein solute, approximately 20 wt %, alters the longitudinal and transverse nuclear magnetic relaxation rates 1/T1 and 1/T2 of solvent water protons in a manner that makes their values indistinguishable from those of a typical human tissue. There is now a quantitative theory at the molecular level (S.H. Koenig and R. D. Brown III (1993) Magn. Reson. Med. 30:685-695) that accounts for this, as a function of magnetic field strength, in terms of several distinguishable classes of water-binding sites at the protein-water interface at which significant relaxation and solute-solvent transfer of proton Zeeman energy occur. We review the arguments that these several classes of sites, characterized by widely disparate values of the resident lifetimes tau M of the bound waters, are associated with different numbers of hydrogen bonds that stabilize the particular protein-water complex. The sites that dominate relaxation-and produce contrast in magnetic resonance imaging (MRI), which derives from 1/T1 and 1/T2 of tissue water protons-have tau M approximately 10(-6)s. These, which involve four hydrogen bonds, occupy < or = 1% of the protein-water interface. Sites that involve three bonds, although more numerous, have < or = 20% smaller intrinsic effect on relaxation. The greater part of the "traditional" hydration monolayer, with even shorter-lived hydrogen-bonded waters, has little influence on solvent relaxation and is relatively unimportant in MRI. Finally, we argue, from the data, that most of the protein of tissue (a typical tissue is mostly protein) must be rotationally immobile (with Brownian rotational relaxation times slower than that of a 5 x 10(7) Da (very heavy) globular protein). We propose a functional basis for this immobilization ("cytoplasmic order"), and then indicate a way in which this order can break down ("cytoplasmic chaos") as a result of neoplastic transformation (cancer) and alter water-proton rates of pathological tissue and, hence, image contrast in MRI.