Contrasting languages and varieties with translational corpora

Translations are invaluable sources for contrastive research. Yet they possess specific linguistic features, which arise from the nature of translation as a form of language contact. This paper explores some aspects of the specificity of translated texts: social and cognitive determinants of translations are briefly discussed, and the linguistic consequences are evaluated in view of recent findings of corpus research based on translational corpora, parallel as well as monolingual comparable corpora. It is argued that we cannot hope to ‘purify’ translational data so as to have them reflect two fully independent languages, unless we sacrifice the use of naturally-occurring data. Translational corpora are biased for genre, because translation situations are asymmetrical, and they are biased for equivalence, because the situation necessitates a search for equivalence more than might be natural in spontaneous production. However, translation and other language contact are ubiquitous in our globalized world; thus entirely independent, ‘pure’ languages are more illusory than real, thus hardly worth hankering after.