The aim of this study was to test the hypothesis that the most appropriate model for studying the diffusional release of an active from a topical formulation is one in which the membrane offers minimal resistance to release and involves a receptor phase that presents the least possible interfacial discontinuity. Using ketoprofen as the active, a series of simple gels were prepared consisting of PEG400 thickened with Cabosil M5. Using Franz-type diffusion cells, three different types of membrane (two porous and one non-porous) were compared, as were receptor phases of PEG400 (component of formulation) and PBS. Of the membranes tested only 0.2 microm nylon provided consistent first order kinetics for a range of gel consistencies, indicating negligible influence of the membrane. The non-porous silicone membrane did not show first order kinetic profile confirming the diffusional nature of such a membrane. From the non-thickened formulations, diffusional release into a receptor phase of PEG400 was some 3x that into PBS, whereas from the formulation thickened with 5% Cabosil M5, diffusional release into a receptor phase of PEG400 was 6x lower than that into PBS. Diffusional release into PBS did not follow first order kinetics while diffusion into PEG400 did, suggesting that the existence of a discontinuity affected the release process. Although the importance of zero-resistance membranes is of perhaps obvious importance, it is often not stated in the literature. The existence of phase/hydrodynamic boundaries in release studies can be a source of significant inaccuracy.