Cellulose ethers have been increasingly used in the formulation of controlled release dosage forms; among them, compressed hydrophilic matrices for the oral route of administration are of special importance. Much interest has also been expressed in the study of the drug release mechanism from such swellable systems, in particular, in trying to explain deviations from Fickian diffusion. Thus, swelling-controlled transport is often invoked without any rationale. It is the purpose of the present work to provide independently determined diffusivity data for elucidating the drug release mechanism from the water-soluble cellulose derivatives. In the first part of this work, pulsed-field-gradient spin-echo nuclear magnetic resonance (PFG-SE NMR) was used to investigate the self-diffusion of the model solute sodium salicylate and, incidentally, that of water, in hydrogels made of hydroxypropyl methylcellulose (HPMC), hydroxyethylcellulose (HEC) and hydroxypropylcellulose (HPC) of varying polymer weight fraction and molecular weight in D2O. In parallel, the extent of bound water in the gels was determined using differential scanning calorimetry (DSC), and the presence of liquid crystals in the gels was examined by polarized light microscopy, as these are the structural factors capable of affecting drug diffusion. Solute diffusivity was not significantly affected by the substitution type of the cellulose ether, and an exponential polymer weight fraction dependence of the solute's self-diffusion coefficient was observed, ascertaining the validity of the free-volume theory, with extrapolated self-diffusion coefficient values similar to those in pure solvent. This also indicates that diffusion also takes place in the so-called bound water (which represents about 40% of the hydrogel weight). This questions the existence of thermodynamically different classes of water. Slightly reduced solute diffusion was measured in the HPC hydrogel of the highest polymer concentration (45 wt.%) where a liquid crystalline mesophase was observed. This structural factor could be of importance, especially in consideration of hydrogels of higher polymer fractions. The polymer molar mass (viscosity grade) of the cellulose ethers also did not affect solute self-diffusivity. The polymer matrix displayed the same retarding effect at equal weight fraction. This confirms that solute molecules can only diffuse in the void space occupied by the solvent. Solute self-diffusivity is dictated by the microviscosity of the system, i.e., the solvent viscosity, with the polymeric matrix increasing only the diffusion pathlength. Overall, because of the similar retarding effects of the hydrated cellulose ether network, at least at the polymer concentrations tested, it would be unwise to ascribe observed differences in drug release kinetics from this type of swellable system to differences in solute diffusivity in the hydrated gel layer (see Part II of the work [J. Control. Release, to be submitted]).