In this article, through the use of several corpora containing original and translated texts, we investigate the use of non-canonical word order in EN-FR texts translated by students in order to determine whether target language norms are respected. We first provide information on cross-linguistic differences between original English and original French, and then compare these results with the characteristics of students’ translations. Our goal is clearly pedagogical: we aim to use data from electronic corpora to sensitize students to differences between their use of the French language in their translations and original French. This is particularly important because of the latest developments of machine translation, which make it crucial for translators to determine their added value over machine-translated texts; we therefore also analyze a sample of machine-translated texts. We conclude on the importance of respecting target language norms in relation to word order to increase naturalness and translator invisibility.