This study is based on an analysis of 1354 translated metaphor examples drawn from a corpus consisting of the official published translations into French, Italian, German, Russian and Polish of 62 Scientific American articles that appeared between January 2003 and July 2004. It investigates what happens to metaphor in scientific discourse when translated into another language, on both micro- and macro-levels. Since one of the main advantages of a data-rich multilingual study of this kind is that it can potentially produce results that allow us to draw conclusions about this aspect of scientific translation at a high level of generalization, particular attention is paid to tendencies that appear to be common to translators regardless of the target language. The study distinguishes between macro-level mappings and micro-level metaphorical expressions and examines individual mappings and clusters of mappings in the English source text and their renderings into all five languages. It adopts a bottom-up approach, in that all mappings and other high-level structures are posited on the basis of the metaphorical expressions identified rather than trying to fit the metaphorical expressions into a pre-determined framework of categories.