Analysing Texts through their Linguistic Properties: A Cross-disciplinary Study

The study reported here focuses on identifying linguistic complexity indices in discussion sections of research articles which can be used to discriminate among various disciplines. Two measures of lexical complexity including lexical diversity and lexical density and two measures of syntactic complexity including average sentence length and ratio of subordination were used as our text complexity model factors. Referring to three major disciplines (humanities, life sciences, and physics), we tried to examine this model. Our corpus consisted of 120 discussion sections of research articles (40 articles from each discipline). Wordsmith Tools software was used to calculate the four complexity measures of each piece of data. To examine the prediction power of our model, MANOVA and discriminant function analysis were run. The results demonstrated substantial predictability (68.3%) for the proposed model. Except for average sentence length, all of the complexity measures used in this study indicated significant contribution to discrimination among three disciplines. The study has implications for EAP and ESP teachers and practitioners.