Since their appearance in the translation field, computer-assisted translation (CAT) tools and, notably, translation memories (TMs) have drawn the attention of academia. Research evidence has, for instance, pointed towards an increase in translators’ productivity when using TMs, and some scholars have warned about possible implications of their use. The TRACE project, carried out by the Tradumàtica research group at the Universitat Autònoma de Barcelona, set out to explore the possible effects of these tools on the translation product. This article focuses specifically on linguistic interference, a phenomenon that, it has been suggested, might be a translation universal. Through experimental research, using a multi-methodological approach and a combination of different data-gathering resources, translations were conducted, with and without TMs, by 90 subjects. The experiments provide interesting data on the distribution of interference according to the environment in which they are carried out, as well as on the differences between different translator profiles.