This paper raises some concerns relating to universality in translation and to the methodology adopted in the search for translation “universals”. The term itself may be misleading if applied to corpus-based research, where the emphasis should be first on the relations between translated texts and the socio-cultural constraints under which they where produced and then on the cognitive processes underlying translation activities. Taking examples from the CEXI corpus, the paper illustrates the working of some such constraints on corpus design. As to non-fiction texts, it shows how two different cultures reciprocally select for translation texts belonging to different textual typologies, resulting in the possibility of skewed distributions within comparable corpora. As to fiction texts, is shows how Italian texts translated into English tend to be canonical high-brow ones, contrary to English texts translated into Italian. The authors propose not to neglect the effect of such contextual variables in translation research, and to set up corpus resources so as to allow multiple comparisons across subcorpora.