Corpus-based approaches to critical discourse analysis usually move from establishing large-scale patterns and trends to examination of keywords and/or collocates to close concordance analysis. This presentation will demonstrate how ... 1.detailed concordance analysis can provide the motivation for large-scale analysis; 2.the examination of high-frequency content words (through multi-sorted concordances) can provide strong indications of the main topics in a specialised corpus of newspaper articles; 3.extending the analysis by examining mid-frequency words can provide a more comprehensive picture by establishing groups of words which a) indicate topics, b) specify contextual elements, and c) provide the co-text essential for the discussion of topics. The analysis forms part of the ESRC funded project, Representation of Islam and Muslims in the UK press, 1998-2009 (PI: Paul Baker; CI: Tony McEnery; RA: Costas Gabrielatos), using a specialised corpus of about 143 million words, containing 200,000 articles related to Islam, its devout, religious customs and practices from twelve national newspapers over twelve years.