Towards a Validated Discourse Analysis of Scientific Text

Abstract

This study reviewed analyses of the discourse structure of scientific experimental-article introductions, and subjected a well-known one (Swales 1981) to empirical test. A corpus of 96 articles was selected by stratified random sampling from 12 journals identified according to post-1980 citation frequency. Raters coded one quarter of the corpus according to the original Swales system, and found the system inadequate. Following revision of the system, 75% of a second quarter of the corpus was able to be rated at a satisfactorily high level of interrater agreement. Results were consistent with the existence of four basic units of discourse. Some subject-specific variation in structure was found. Social science article introductions were more complex than others. Implications for EST materials writing are drawn. The importance of empirical validation of discourse analyses is stressed. This study reviewed analyses of the discourse structure of scientific experimental-article introductions, and subjected a well-known one (Swales 1981) to empirical test. A corpus of 96 articles was selected by stratified random sampling from 12 journals identified according to post-1980 citation frequency. Raters coded one quarter of the corpus according to the original Swales system, and found the system inadequate. Following revision of the system, 75% of a second quarter of the corpus was able to be rated at a satisfactorily high level of interrater agreement. Results were consistent with the existence of four basic units of discourse. Some subject-specific variation in structure was found. Social science article introductions were more complex than others. Implications for EST materials writing are drawn. The importance of empirical validation of discourse analyses is stressed.