The Finnish plan for the final deposition of nuclear waste is deposition deep in the crystalline bedrock. In the safety case of the final deposition, matrix diffusion, along with sorption, is considered the most important retarding factor for radionuclide transport in the geosphere. We set out to measure matrix diffusion in the gas phase for two 80 cm long drill-core samples from Olkiluoto, Finland. One sample consisted of veined gneiss and the other of pegmatitic granite. The measurements were taken in the laboratory with an advection-diffusion measurement setup that uses nitrogen as the carrier gas and helium as the tracer. The measured breakthrough curve for the veined gneiss sample could be interpreted with a homogeneous mathematical model, and the result for the effective diffusion coefficient (), , agreed with previous results for veined gneiss from through-diffusion experiments in the gas phase. The breakthrough curve for the pegmatitic granite sample required a three-component model because of high variations in the local transport parameters (porosity, diffusion coefficient). The spatial distributions of porosities were characterized with the C-polymethylmethacrylate-autoradiography technique. The result, , for the most abundant component was also in accordance with the previous through-diffusion results. Comparison of the two measurements showed that diffusion phenomena can be very sensitive to the microstructure of rock even when macroscopic properties such as porosity remain the same. The retarding effect of matrix diffusion from flow was clearly observed for both samples, and knowledge of the sample structure, rock types, and mineralogy were concluded to be vital in analyzing and interpreting the results.