The diffusion coefficient of tritium in its liquid form through cementitious materials is an indicator of durability in the domain of nuclear waste disposal. Because the cement-based materials designed for the waste storage have very high confinement properties, diffusion tests tend to last several months if not years in some cases. Therefore, basing the determination of the tritium diffusion coefficient on the experimental data measured during non steady state is tempting. The aim of this work is to demonstrate that the determination during the transient regime of the effective diffusion coefficient and the material porosity leads to non-negligible underestimation of these quantities. Moreover, it appears that, when using a numerical model for the determination, their validity can only be judged on the basis of more than one criterion, whether those criteria are obtained from the experimental data or from a numerical model. (C) 2015 Elsevier Ltd. All rights reserved.