Conditional Inference Tree Modelling of Competing Motivators of the Positioning of Concessive Clauses: The Case of a Non-native Corpus

• Published in: Journal of quantitative linguistics Vol. 24; no. 2-3; pp. 89 - 106
• Main Authors: Rezaee, Abbas Ali, Golparvar, Seyyed Ehsan
• Format: Journal Article
• Language: English
• Published: Lisse Routledge 03.07.2017; Taylor & Francis Ltd
• Subjects: Adverbials; Clauses; Concessives; Inference; Pragmatics; Second language writing
• ISSN: 0929-6174; 1744-5035
• DOI: 10.1080/09296174.2016.1265799

Concessive adverbial clauses may precede or follow their associated matrix clauses. Previous research has revealed that multiple discourse-pragmatics and processing-related factors affect the ordering of these clauses. The present study purports to investigate the relative importance of competing motivators of the positioning of concessive clauses in a non-native discourse community. To this end, a corpus of 100 research articles written by Iranian non-native writers was used and all the finite concessive clauses accompanied by their main clauses were gathered. Random forest of conditional inference trees is the main statistical modelling in this study. The results demonstrated that Iranian writers use pre-posed concessive clauses more frequently than post-posed clauses. In addition, subordinator choice and bridging, which are discourse-pragmatics constrain on clause positioning, are far better predictors of the ordering of concessive clauses in this corpus. Finally, random forest modelling turned out to be slightly more efficient than logistic regression for the task of prediction.