Researching L2 writers’ use of metadiscourse markers at intermediate and advanced levels

Abstract

Metadiscourse markers refer to aspects of text organisation or indicate a writer's stance towards the text's content or towards the reader (Hyland, 2004:109). The CEFR (Council of Europe, 2001) indicates that one of the key areas of development anticipated between levels B2 and C1 is an increasing variety of discourse markers and growing acknowledgement of the intended audience by learners. This study represents the first large-scale project of the metadiscourse of general second language learner writing, through the analysis of 281 metadiscourse markers in 13 categories, from 900 exam scripts at CEFR B2-C2 levels. The study employed the online text analysis tool Text Inspector (Bax, 2012), in conjunction with human analysts. The findings revealed that higher level writers used fewer metadiscourse markers than lower level writers, but used a significantly wider range of 8 of the 13 classes of markers. The study also demonstrated the crucial importance of analysing not only the behaviour of whole classes of metadiscourse items but also the individual items themselves. The findings are of potential interest to those involved in the development of assessment scales at different levels of the CEFR, or to teachers interested in aiding the development of learners. Highlights • Advanced L2 writers use fewer metadiscourse markers than intermediate writers • Advanced L2 writers can express text organisation without explicit markers • Higher-level writers use a greater range of metadiscourse markers • Higher-level writers abandon simple markers in favour of other makers

Interactional Meta-Discourse Resources in Oliver Twist

“…Less frequent use of IMRs in the simplified stories can be attributed to the essence of simplification since the narrators are aware of the codes of simplification and they try to adopt an impersonal style. Bax, Nakatsuhara, and Waller (2019) claim that in academic writings, there is a threshold level for using meta-discourse markers beyond which fewer meta-discourse resources will be used. That is, at the beginning levels more of these resources will be used while advanced writers use fewer, this decline might be due to using more sophisticated words to write a text instead of relying on explicit markers.…”

Relationship between persuasive metadiscoursal devices in research article abstracts and their attention on social media

“…Due to the multiple linguistic markers of the interactive and interactional types, we employed Text Inspector, an award winning text analysis algorithm developed by Prof. Stephen Bax [81], which has been used by major publishers including Cambridge University Press and students and staff at 145 universities and colleges in over 100 countries [82], for the automated analysis of the large data set collected. Text Inspector adapted Hyland's taxonomy on metadiscourse analysis and is often used in research related to metadiscourse and EVP analysis [81,83,84]. However, as noted in a previous study, Text Inspector's outputs are not "perfectly accurate" [81].…”

“…Text Inspector adapted Hyland's taxonomy on metadiscourse analysis and is often used in research related to metadiscourse and EVP analysis [81,83,84]. However, as noted in a previous study, Text Inspector's outputs are not "perfectly accurate" [81]. To ensure an accurate counting of metadiscoursal devices in the 241 research article abstracts collected, we conducted a four-stage text analysis in which both automated analysis [81] and manual checking [85] were employed to enhance the accuracy of marker identification and avoid the risk of overlooking other markers.…”

“…In order to examine the association of metadiscourse devices and AAS (RQ1 and 2) which had a very different corpus base, we standardized the raw count on the metadiscourse markers using z score for investigating the correlation between specific metadiscourse devices, AAS [86] and mentions from various sources including news, blogs, Twitter, and Facebook. Since the research article abstracts collected were of varied lengths, ranging from 25 to 1097 words, the comparison between markers count in the raw dataset (RQ3) could confound the results [81]. We therefore standardized the data using the "frequency per 100 words" [81]for ANOVA analysis.…”

“…Since the research article abstracts collected were of varied lengths, ranging from 25 to 1097 words, the comparison between markers count in the raw dataset (RQ3) could confound the results [81]. We therefore standardized the data using the "frequency per 100 words" [81]for ANOVA analysis. Subsequently, all statistical analyses were conducted on the standardized dataset.…”