Using corpus linguistic techniques in (critical) discourse studies reduces but does not remove bias: Evidence from an Arabic corpus about refugees

Abstract

Abstract Critical Discourse Analysis (CDA) investigates the relationship between language, power, and society. Corpus linguistics (CL) is the study of language based on examples of real life language use. Over the last two decades, various scholars have combined some approaches and notions of CDA with the analytical framework of CL to examine the representation of several phenomena in relatively large texts. This study follows a corpus-assisted (critical) discourse analysis approach to investigate a 2.5 million word corpus of Arabic news articles by Jordan’s News Agency (PETRA). It demonstrates how some researchers following this approach may make some decisions, at some stages of their analysis, which are likely to affect their findings. These potential decisions may include selecting what statistical measures to use, what threshold to consider, what terms from the frequency, cluster, and collocation results to further investigate, which concordance lines to include in their study, and some others. In this study, I argue that some of these decisions can be made to suit the researchers’ preconceived assumptions and pre-existing hypotheses. The study concludes that using corpus linguistic techniques to discursively analyze large data reduces but not completely removes researchers’ bias.