This chapter focuses on describing, illustrating and critiquing the keywords technique, which is used to automatically identify lexical salience when comparing multiple corpora. Arguably, keywords present researchers with words that they may not have chosen to analyse in advance, thus helping to reduce researcher subjectivity. 1 I illustrate how the identification of keywords enables researchers to embark on interesting research journeys, through examples taken from an analysis of the representation of Islam and Muslims in a corpus of British newspaper articles. However, when using large corpora, even with high cut-off points for statistical salience, hundreds of keywords may be produced, meaning that researchers need to make decisions regarding which words are worthy of detailed focus. The chapter ends with an illustrative analysis where I revisit six of my own keyword studies, arguing that researchers should consider the benefits of giving a more reflexive account of their own decision making procedures around keywords. © 2018 John Benjamins Publishing Company