Abstract  Van Emden executed numerous corrections in his edition of the Cambridge version of The Song of Roland (2005). However, these amendments are often insufficiently justified and cannot stand up to a rigorous and detailed examination from the standpoint of historical linguistics. In most cases, van Emden modified the text of the manuscript in order to « restore » metrically correct verses that, to his mind, should have been included according to the model followed by the Cambridge copyist. Van Emden accomplished his stated task by simultaneously rejecting and retaining modernized forms or words. Sometimes, he substituted an archaic form or word for a modernized term, even when the former had fallen into disuse long before the time the Cambridge manuscript was redacted. This paper does not merely list such contradictory or unsuccessful amendments. Instead, it shows that a common methodological problem is present in all the problematic corrections: van Emden was not entirely aware of the stage of textual evolution that he aimed to reinstate. By contrast, the author of the present paper began elaborating the methodology for his new edition of the Cambridge manuscript with precisely this question.