Non-Fickian or anomalous diffusion is frequently observed when the absorption of moisture by a polymer is being studied. Different models have been presented in the literature that can accurately predict the trends of the weight-gain curves. However, it is not always clear which of these models yield good predictions of moisture distribution. This article presents a time-resolved moisture distribution study of an epoxy sample immersed in deuterated water (D2O) at 70 degrees C over a period of 2.5 months. The moisture distribution was measured during that period with a novel high-resolution magnetic resonance imaging technique that is well adapted to the imaging of thin plates. The experimental results showed that the concentration of D2O at the surface of the sample increased with time, even after 2.5 months. These results were used to evaluate the performance of several standard diffusion models. Although this study is phenomenological, it appears that a model featuring time-varying boundary conditions yields the best representation of moisture absorption for these samples. (C) 2008 Wiley Periodicals, Inc.