The paper deals with computing key words by comparing and processing data retrieved from two corpora: reference corpus and a specifically designed corpus or a text. The main aim of the research revolves around the idea of equipping linguists, teachers and students with the practical skills to analyze the language they teach statistically. Thus, the research is based upon statistical procedures provided by standard corpus software. We set out to find out whether the key words computed automatically, with the help of software WordSmith Tools 6.0, could provide a reliable basis for selecting the most important lexical units for learners of English. The learners in question were undergraduate students of Area Studies (the UK). In our experiment we compared three lists of words: chosen by two ESP teachers and the one computed automatically. The article elaborates on the procedure and the relevance of the three word lists within second language acquisition context. In conclusion, it is stated that corpus tools could be of significant help for ESP professionals.