A corpus-based study of the mediation effect in translated and edited language

This paper presents the results of a study investigating the hypothesis that the recurrent features, or universals, of translated language are primarily the result of a mediation process that is shared among different kinds of mediated language, rather than the particularities of bilingual language processing. The investigation made use of a comparable corpus consisting of a subcorpus of English texts translated from Afrikaans, a subcorpus of comparable edited English texts, and a subcorpus of comparable unedited (and also untranslated) English texts. The frequency and distribution of linguistic features associated with three of the universals of translated language (explicitation, normalisation/conservatism, and simplification) across the three subcorpora were analysed. The study was guided by the hypothesis that the frequency and distribution of linguistic features associated with the universals of translated language would demonstrate similarities in the two subcorpora of mediated text (i.e., the translated and edited subcorpus), as compared to the subcorpus of unmediated text (i.e., the unedited subcorpus). However, the study yields almost no evidence for a mediation effect that is shared by translated and edited language, at least not along the linguistic features investigated. There is, however, evidence for what appears to be a separate translation-specific effect, which seems likely to be more unconscious, more proceduralised and more related to the linguistic level alone. This offers some support for the hypothesis of universals of translated language that are unique to this kind of text mediation specifically. Furthermore, the findings of the study suggest that editing may involve a different kind of mediation effect altogether, which frequently remains invisible in conventional corpus-based studies comparing translated and non-translated language, and which requires further investigation.