This paper utilises a unique data set to study the scale and density properties of rail infrastructure maintenance, distinguishing between different technologies existing within national rail systems; namely, urban, high-speed, and regional lines. We find differences in the position and shapes of the alter native technologies, with returns to scale and density properties varying by network type. Such technology-disaggregated cost structure characteristics have not previously been derived in the literature, and have implications for track access charges and rail way maintenance organisation. Additionally, the widely used translog functional form is shown to be insufficiently flexible to deal with the different technologies under consideration.