Abstract This study uses a novel approach to describe modern English academic texts in terms of the amount and distribution of syntactic subordination, coordination and phrasal structures. Inconsistent results have been reported in previous scholarship regarding the trajectory of syntactic complexification based on English language backgrounds and linguistic proficiency of writers. For the first time, a combination of two predictive statistical modelling methods of mixed-effects modelling and supervised machine learning modelling with random forest is used to examine the extent to which the type, amount, and distribution of these syntactic structures can be attributed to a text-intrinsic feature (sub-genres or rhetorical sections of academic register) and a text-extrinsic factor (English language background based on academic context) in a corpus of master’s dissertations. To revisit the theories of English L1 vs. L2 texts, strong predictors of syntactic complexity across academic writings of EFL, ESL, and English L1 postgraduate students were identified. The findings show that EFL texts are dominantly subordinate in nature, English L1’s dominantly phrasal, and the ESL dissertations exhibited similar amounts of subordination and phrasal structures. Distinct syntactic structures also characterise rhetorical sections: abstracts are predominantly phrasal, literature reviews distinctly subordinate, and conclusion sections have noticeable amounts of verb phrases.

中文翻译：

---

研究生英语学术写作更多的是小句还是短语？流派、熟练程度和统计建模交叉路口的句法复杂化

摘要 本研究采用一种新颖的方法来描述现代英语学术文本的句法从属、协调和短语结构的数量和分布。关于基于英语语言背景和作家的语言能力的句法复杂化轨迹的先前学术研究报告了不一致的结果。首次将混合效应建模和监督机器学习建模与随机森林两种预测统计建模方法相结合，用于检查类型、数量、这些句法结构的分布和分布可以归因于硕士论文语料库中的文本内在特征（学术注册的子流派或修辞部分）和文本外在因素（基于学术背景的英语语言背景）。为了重新审视英语 L1 与 L2 文本的理论，确定了 EFL、ESL 和英语 L1 研究生学术著作中句法复杂性的强预测因素。研究结果表明，EFL 文本本质上是从属的，英语 L1 的主要是短语，而 ESL 论文表现出相似数量的从属和短语结构。不同的句法结构也是修辞部分的特征：摘要主要是短语，文献评论明显从属，