A crude analysis of the observed thickness dependence of the conductivity in fine-grained metals shows a very weak connection with the resistivity caused by grain-boundary scattering. An ``effective intrinsic mean free path'' for the whole polycrystal, as introduced in the well-known theory of Mayadas and Shatzkes, cannot explain the observed size effect. It is impossible to define a mean free path for the whole polycrystal, especially if the resistivity is governed by grain-boundary scattering.