Describing and modelling a machined surface require the selection of relevant roughness parameters. However, this selection is difficult since a machined surface morphology can be described by a large number of roughness parameters.This investigation focuses on the roughness of metallic surfaces taking for two applications: a) the description of machined surface morphologies produced by grinding b) the relationships between machined surface morphologies (grinding or cold-rolling) and their brightness level when irradiated by the white light beam of an optical glossmeter used for industrial surface quality control. For each application, the aim is to determine, from an objective quantitative point of view, the relevance of one hundred or so surface roughness parameters. To reach this objective, a specific software program has been developed to determine a ranking of relevance thanks to the calculation of a computed statistical index of performance.The statistical results of this study show that the fractal dimension estimated by an original method is the most relevant roughness parameter to describe the surface morphology after grinding or rolling. Because of this relevance, this roughness parameter has also to be taken into consideration in models showing the interactions between machined surfaces and an optical wave. The methodology presented in this study can be a useful tool in the quality control phase to keep under control the fabrication process parameters of manufactured objects in industrial environments. (C) 2003 Kluwer Academic Publishers.