In this paper we will focus on the scribal variation in manually copied medieval texts. Using a lazy machine learning technique, we will argue that it is possible to discriminate between scribes, implying that they did adapt texts when copying them. Consequently, we will assess to what extent scribal interventions compromise our ability to detect the original authorship of medieval texts. It will be shown that, if the right features and weighting methods are used, the automated discrimination of both copyists and authors is possible for medieval texts. The case studies presented suggest that scribes only corrupted the original texts in a shallow and superficial way, leaving authorial features generally intact on deeper levels. This result will be of interest for research into e.g. contemporary newspaper articles when trying to detect editorial interventions.