A CORPUS-BASED ANALYSIS OF DISTRIBUTIONAL AND FUNCTIONAL DIFFERENCES IN DISCOURSE MARKERS ACROSS SPOKEN AND ACADEMIC ENGLISH

Özet

This corpus-based study examines the extent to which selected discourse markers (well, you know, actually, I mean, anyway, in fact, indeed) differ in distribution and functional realization across spoken and academic written English in the Corpus of Contemporary American English. Using the spoken (126,135,576 words) and academic (119,790,456 words) sections in the corpus, the analysis extracted raw token counts and normalized frequencies per million words, and computed spoken-to-academic ratios as a register-sensitivity index. To address polyfunctionality, a targeted concordance validation was conducted for well and actually, sampling 10 lines per register (20 per marker) and coding each instance as discourse-marker/discourse-marker-like versus non-discourse marker, with an indicative adjusted discourse marker rate derived from observed proportions. Results show strong register differentiation: you know, and I mean display the largest spoken skew, anyway, is also spoken-preferential, while well and actually show moderate spoken skews in raw counts. In fact exhibits only a modest spoken preference, whereas indeed is the sole marker favoring academic prose. Functional validation indicates that raw frequencies substantially overestimate DM use for polyfunctional items, especially actually, whose discourse-pragmatic uses are minority instances in both registers. Pedagogically, the findings support corpus-informed instruction that treats discourse markers as register-bound resources and incorporates concordance-based activities to build learners’ pragmatic and register competence.