Corpus studies of translated language usually analyse corpora of published texts. While such corpora have the strength of drawing on actual data, they hide the influence that agents other than the translator exert on the text in the process from manuscript to publication. The corpus study presented here investigates the linguistic action that editors take on translated articles. A parallel corpus of English business articles and their published German translations is complemented by a corpus of the corresponding manuscript translations, which allows an analysis of the translated text before and after the editing process. The focus of this talk will be passive constructions and passive alternatives. Passive constructions are a hallmark of scientific language cross-linguistically, but at the same time style guides discourage writers from using them. This conflict leads to a tension in the translation of business articles, which often report empirical results but nevertheless aim to be comprehensible. The passive voice may increase textual metaphoricity, but also allow serve as a tool to improve the thematic discourse structure of the text. Editors’ work plays a major role in navigating this conflict, but generally remains understudied in linguistics. I will discuss English-German functional contrasts that may affect language users’ likelihood to use the passive, depending on whether the language provides other means of achieving the desired discourse organisation. Preliminary findings suggest that editors regularly turn passive constructions in the translated manuscripts into active constructions. This seems to hold true for grammaticalised German passive forms using auxiliaries such as "werden", modal infinitives and passive alternatives such as the modal passive or the impersonal pronoun "man". Consequently, I will argue that the common practice of making claims about translated language based on published texts, or attributing phenomena observed in the final products solely to translators, may be problematic.