It has been suggested in comparative translation studies that translated texts are more like each other than comparable non-translated texts, i.e. the range of variation between them is smaller. For simultaneous interpreting, a translation universal describing such a change has been postulated: an equalizing universal, or the flattening of the orality-literacy continuum. This means that inherently oral genres, e.g. interviews, will show more literate features when interpreted, while inherently literate genres, e.g. an opening speech delivered by the President of the General Assembly at a United Nations General Debate, will receive more oral features than the originals.

In this project, I investigate collocativity in the English component of a 230,000 word corpus of bidirectional Russian-English simultaneous interpreting (SIREN), using ukWaC as a reference corpus. Collocativity is the degree to which a text relies on pre-patterned chunks characteristic of the language in question. To measure collocativity in SIREN, I apply the methodology developed by Bernardini (2015) that uses larger corpora to overcome the data bottleneck caused by small interpreting corpora. The results show statistically significant differences in the frequency and range of three POS chains: Adjective-Conjunction/Preposition-Adjective, Verb-Adverb, and Adverb-Adjective. When these POS chains are investigated in two other corpora of written (BAWE) and spoken (OANC) English, they also exhibit statistically significant differences with the direction of shift expected on the basis of the equalizing universal. In this manner, the study provides further evidence for the validity of the equalizing universal in simultaneous interpreting and suggests a possible explanation to sometimes contradictory results of earlier studies of linguistic variation in interpreted speech. ?