Résumé : This paper analyses several issues that arise when measuring technological specialisation with patent data. Three starting choices are required regarding the data source, the statistical measure and the sectoral aggregation level. We show that the measure is highly sensitive to the data source and to the level of sectoral aggregation. The statistical analysis further suggests that the most stable and reliable measures of technological specialization are obtained with patents applied at the EPO, with Gini or C20 as statistical measure and the 4-digits aggregation level of the IPC classification system.