The two pilot studies described in this article, both part of a larger on-going project investigating multilingualism in a translation context, deal with separation of languages in translation students. At the same time, they demonstrate how multilingualism research can be integrated into the translation classroom. Within the framework of Grosjean's model (1997, 1998, 2001), the first study tests the influence on translation performance of preferentially activating one of the languages for the translation version Spanish (L2) into German (L1). The second assesses the transfer of structures from the dominant language in translating from German (L1) into English (L2). Despite the relatively small database, various inferences can be made about multilingual language processing in trainee translators.