Meanings are established in individual languages by contrasts of similar items in semantic fields and a consequence, semantic structures do not match cross-linguistically (Gorlach 2003: 93). This broad statement about the formation of meaning can be adopted as a theoretical principle underlying the study of lexical borrowing, a phenomenon whereby new vocabulary acquired through language contact tends to present in the receptor language a lexico-grammatical profile that differs to varying degrees from the corresponding patterning in the donor language, this being the case particularly with polysemic words. Moreover, given that in translation meaning relations between expressions are evaluated as a normal kind of linguistic activity rather than as part of some metalinguistic, philosophical or theoretical reflection (Dyvik 1998: 51), translated texts can serve as semantic mirrors reflecting meaning across languages (Johansson 2003: 136) and become an invaluable resource for investigating loan words across donor and receptor languages. The aim of this paper is twofold. Firstly it intends to put forward a corpus-based methodology for the study of anglicisms in business discourse, which is devised on the basis of the principles expressed by Gorlach's, Dyvik's and Johansson's statements. Secondly it reports on findings that focus on the lemma business and have been obtained from an initial application of the proposed methodology. After reviewing recent Italian studies of anglicisms (Pulcini 2002a, 2002b), I will define the current object of study, deal with corpus design and compilation, outline the corpus processing tools and procedural steps and finally illustrate two sets of findings obtained from the analysis of a bilingual comparable corpus, ComIC&ComEC (Commercial Italian Corpus & Commercial English Corpus) and a bilingual parallel corpus, BusiPC (Business Parallel Corpus).