On the physical interpretation of fractional diffusion