Noun Phrase Complexity in EFL Academic Writing: A Corpus-based Study of Postgraduate Academic Writing

Abstract

Noun phrase (NP) centered structures are distinctive syntactic devices in academic discourse. The commonly employed subordination-based complexity measures cannot adequately capture the development of syntactic complexity of noun phrases expected of advanced student academic writing (Biber, Gray, & Poonpon, 2011). Following the call for more research in this area (e.g. Lu, 2011, p. 57), the current study compared noun phrase complexity in two corpora: one is a corpus of MA dissertations written by Chinese EFL students and the other comprises published research articles in applied linguistics journals. The study examined overall noun phrase complexity using an automatic syntactic complexity analyzer and specifically identified features of one aspect of NP complexity: NP postmodification. The quantitative results were further contextualized in a textual analysis of excerpts from the two corpora for demonstrating the significance of NP complexity to the establishment of discourse coherence. Results of the analyses showed significant underdevelopment of NP postmodification complexity in student writing relative to published texts, meanwhile explicating the circumstances under which the difference is meaningful. Implications of the findings for the teaching of EFL academic writing were also discussed.

The Effect of L2 Proficiency on Post-editing Machine Translated Texts

“…While we know that advanced learners are more capable of detecting and correcting a higher number of errors (Kol et al, 2018), we do not yet know what kind of translation errors are detected and corrected by learners of varying proficiency levels. Previous research makes it clear that when writing in the L2, advanced learners are more likely to produce longer texts (Crossley & McNamara, 2012;Jung et al, 2019), display greater lexical variety (Grant & Ginther, 2000;Jarvis et al, 2003), and generate syntactically complex and accurate texts with a higher number of nominalization/subordination (Grant & Ginther, 2000;Liu & Li, 2016) and error-free T-units (Hwang, 2012). Whether such linguistic features can also be observed for MT post-editing remains to be seen.…”

Genre-based Analysis of Syntactic Complexity in L2 College Students’ Writing: Pedagogic Scope and Directions

“…In the case of argumentative essays, the results of all 14 syntactic complexity measures were not statistically significant across the proficiency levels. According to previous related research by Liu and Li (2016), Lu (2011), Park (2012), and Kim (2014), there were statistically significant differences between the levels either negatively or not at all. Even though there was not a statistically significant difference of the syntactic complexity in the argumentative essay across proficiency levels in the present study, it was possible to find syntactic differences between level 1 and level 3 as well as between level 2 and level 3 respectively.…”

Syntactic Complexity Measures and Academic Writing Proficiency : A Corpus-based Study of Professional and Students’ Prose

“…Lu's classification of the measures was "drawn from WolfeQuintero et al (1998) and Ortega (2003) inter alia" (Kyle, 2016, p. 52). As Lu (2011) put it, they are comprehensive complexity indices commonly used in writing development (e.g., Bae & Min, 2020;Bailey & Judd, 2018;Kyle & Crossley, 2018;Lee, 2021;Liu & Li, 2016). The measures included in this study are mean length of T-unit (MLT), mean length of clause (MLC), clauses per sentences (C/S), dependent clauses per T-unit (DC/T), and complex nominals per T-unit (CN/T).…”

“…We choose SCA because of its free availability, inclusion of large number of measures (5 types of measures, each of which is assumed to capture one dimension of syntactic complexity), batch processing, and high reliability indices which are reported to range from 0.83 to 1.00 for structural unit identification, and 0.83 to 1.00 for correlation with human annotators (Lu & Ai, 2015). Initially, SCA uses Stanford parser to segment the text into individual sentences, which are then tokenized, and part-of-speech (POS) tagged (Liu & Li, 2016). Then using Tregex, the nod matches within a Tree are identified, and finally 14 measures of syntactic complexity are calculated.…”