This article aims at the characterization of specific features of translated texts. Taking a classroom experience as its starting point, the use of anglicisms in original and translated computing texts in Italian is examined. The corpus used for this purpose has three components: originals in Italian, comparable translations into Italian, and their English source texts. The frequency of three sets of English words – overt lexical borrowings, adapted borrowings and semantic loans, and morphosyntactic calques (plurals ending in –s) – is compared across the monolingual comparable subcorpus components. The parallel subcorpus is then checked to disprove the null hypothesis according to which observed differences are unrelated to the translation process. The results of the quantitative analysis, followed by careful qualitative observations, confirm that translators are more conservative in their choices and normalize more than writers, who seem to be more prone to interference from English as the lingua franca of the IT discourse community. Implications at the methodological, descriptive/theoretical and applied levels are discussed.