Focusing on non-Fickian diffusion in biosourced materials, we have recently developed a new formulationthat can model this behavior using memory functions. To identify those memory functions, an experimentaldevice based on a magnetic suspension balance is used. Very thin samples of low-densityfiberboard, medium-density fiberboard, and Spruce in the longitudinal direction are subjected to variationsin ambient relative humidity (RH) while their mass is recorded. The imposed RH and temperatureare used as boundary conditions in a comprehensive model to solve the heat and mass transfer equationsin porous media. A computational code solving the new formulation was used to determine the memoryfunctions by inverse analysis. The use of very thin samples enables macroscopic parameters to be consideredas negligible compared to microscopic effects.