This study stems from our work training translators where we have noticed that trainee translators struggle with English hyphenated premodifiers. Such premodifiers come in a variety of different forms, e.g., N + ed-participle (pig-headed losers), adjective + ing-participle (a tight-fitting beret), NPs (the end-of-term reports) and verb phrases (a go-along-and-enjoy-yourselves gesture) (for an overview, see Biber et al. 1999: 534–5). As indicated in these examples, hyphens are used both with highly lexicalized premodifiers (Arnaud et al. 2008: 116) and ad hoc constructions. The aim is to investigate how professional translators translate these construction types into German and Swedish in the Oslo Multilingual Corpus, focusing on how the structural means of the two target languages affect the choices made. The results will also help to improve teaching materials for trainee translators, by providing an overview of the strategies used by professionals.

Previous findings suggest that premodification is more common in German than in Swedish source texts, which favour postmodification (cf. Fleischer & Barz 1995: 320–31; Teleman et al. 1999: III: 71–84). It can therefore be assumed that translations into these languages also have different preferences.

Our data show that different construction types are connected to different types of translation alternatives, and there are some indications of target-language-specific preferences. For example, ed-participles are generally rendered as similar adjectives in the translations (liver- coloured > leberfarbene/leverfärgad), and relative clauses are more common in Swedish translations (a market-analysis firm > ett företag som gjorde marknadsanalyser) which confirms the observation that Swedish is more prone to postmodification. In the German translations, on the other hand, complex premodifications are more often rendered as extended participial premodifiers (sea-washed stone > vom Meer glattgeschliffenen Stein). Other frequent translation strategies involve compounding and prepositional phrases. These premodifiers often lack lexicalized equivalents, and they are often omitted, restructured or rendered word-for-word.

Our results indicate that there is a relatively low degree of correspondence between the structures chosen in the target languages, at least regarding the more lexicalized instances. This suggests that the structural means of the individual languages affect the strategies used by either allowing or forcing translators into making different choices. The degree of lexicalization is a key factor when translating more freely or word-for-word.