A cured epoxy originating from a stoichiometric mixture of a diglycidyl ether of Bisphenol A and 1,3-phenylenediamine has been studied by proton NMR including multiple-pulse techniques. In order to address the question of the uniformity of water distribution based on possible variations in crosslink density, spin diffusion experiments on a dry epoxy were performed over a wide temperature range in order to probe the distance scale of cross-link heterogeneity. The actual experiments measured the minimum distance scale (4 nm) within which sample-wide variations in the multiple-pulse relaxation time, T-1xz, are included. Experiments were performed to indicate that these variations in T-1xz were not primarily associated with aliphatic versus aromatic protons. Whether the variations in T-1xz indicate variations in cross-link density versus fluctuations in local packing in the glass remains an open question. Water chemical shift versus water content was also measured. It was inferred from these data that the majority of the water was molecularly dispersed as opposed to being aggregated into voids. Given that the existence of voids is often invoked in the epoxy literature, the NMR data would limit the volume of any voids to a volume where only one or two water molecules would fit in each void. No evidence was seen in the chemical shift data which would support the idea that there were two different sites with substantially different affinities for water. Also, water line widths were measured at 75 degrees C for samples of different average water concentrations and relatively uniform concentration profiles. It was anticipated that this measurement would provide a calibration curve whereby the existence of water concentration gradients could. be identified for samples of known average water content. This anticipated result was not demonstrated conclusively. Finally, the line-width results are considered as they apply to the behavior of water in the later stages of water uptake.