Tracking the third code

Abstract Corpus-driven methods are a powerful heuristic for identifying features of translated language, particularly phraseological features, which have been largely neglected in crosslinguistic studies. This chapter presents the results of an n-gram based study of metadiscursive markers in original and translated English. Automatic comparison of three- to four-word recurrent sequences in the two corpora reveals a series of markers that are significantly overused or underused in translated language. One category of overused bundles, i.e. markers of contrast, is analysed in the light of translation universals and source language effects. In a wider perspective, the study highlights the benefits of a multi-corpus empirical basis for corpus-based crosslinguistic studies and draws the preliminary contours of a methodological framework, ‘Contrastive Translation Analysis’, which integrates comparable, translation and learner corpus data.