With this article the author underlines the need for a corpus of translated texts, which reveals regularities of the text type of translation. These regularities arise from the unique context in which translation as a language activity is performed. Translated text is constrained by the source text, in another language; has to respond to the needs of the target audience, while taking into account the source readershi, and has to cope with its own changing social and textual status – i.e. translators tend to stay close to the typical patterns of the target language, as creativity on the part of the translator is easily confused with literalness. In the remainder of the article the author outlines the challenges for a translation corpus, and focuses on the issue of how to identify universal features of translation. The author lists the following hypothetical universal characteristics: explicitation, simplification, normalisation and levelling out. In the conclusion the need is stressed to define these features further.