The aim of this paper is to explore some of the resources of written language (as opposed to spoken language) in English-Korean translation with a view to assisting trainee translators in delivering ‘translationese’-free natural translation in different types of texts. Trainee translators in their transition from the ‘general language (GL)’ translation phase to the ‘special language (SL)’ one (terms used following Snell-Hornby, 1990: 33) tend to confuse translation strategies for the two different phases. Consequently, they may inadvertently apply the translation skills they have developed in the former to the latter, often producing ‘awkward’ translation. In English-Korean translation, this is particularly true in terms of how much written-ness their translated text should involve in comparison with translations rendered by professional translators. In this paper, written-ness is measured using the concept of ‘grammatical metaphor’ whose central contributor is nominalisation (Halliday, 2004), along with two other criteria: number of clauses per sentence and lexical density. The main data of this pilot study includes two GL texts and two SL texts and their translations by trainees (TT1) and professional translators (TT2). The results of the analysis indicate that TT1 of GL texts should be more spoken than as is while that of SL texts needs to be considerably more written.