Introduction

There has recently been a surge of interest in phraseology by linguists and language pedagogues alike. Fuelling this has been the realization that ritualized phrases are a common feature of everyday conversation (Wilkins 1976) and that a high percentage of spoken and written text is made up of recurrent patterns which constitute single choices for the language user (Gross 1988; Sinclair 1991; Feilke 1996). Computer-based analyses of large amounts of data have cemented this newly emergent picture of language, especially with regard to English (e.g. Sinclair 1991; Howarth 1996; Moon 1998; Biber et al. 1999; Hunston and Francis 2000) and French (Siepmann 1999; Gréciano 2000). Part I of the present research has gone some way towards further filling in the picture; there it was seen that both academics and journalists cherish a specific type of conventionalized multi-word unit which I have called the ‘second-level discourse marker’, or ‘SLDM’ for short.