Abstract This study investigates the syntactic complexity, as measured by both large-grained and fine-grained measures, of 410 narrative writings across four writing proficiency levels written by beginner and intermediate L2 English learners. By exploring the differences in syntactic complexity between writings at different proficiency levels, the study is purposed to find out the measures that can best discriminate and predict writing proficiency. The L2 Syntactic Complexity Analyzer and the dependency syntactically-annotated corpus are used respectively to collect the data for large-grained and fine-grained measures. With regard to large-grained measures, it is found that students with higher writing proficiency tend to produce longer language units, more subordinate clauses, more coordinate clauses, and more noun phrases in their writings; mean length of T-unit, mean length of sentence, and dependent clauses per clause can better predict writing proficiency than other traditional large-grained measures. As for fine-grained measures, it is found that three types of subordinate clauses, that is, adverbial clauses, complement clauses and relative clauses, and two types of noun modifiers, that is, prepositional phrases and adjectival relative clauses, occur more frequently in the writings of more proficient learners; the frequency of compound nouns correlates negatively with writing proficiency.

中文翻译：

---

EFL 学习者写作中的句法复杂性发展：来自依赖句法注释语料库的见解

摘要 本研究调查了由初级和中级 L2 英语学习者撰写的跨越四个写作水平的 410 篇叙事写作的句法复杂性，通过大粒度和细粒度度量来衡量。本研究旨在通过探索不同水平的写作在句法复杂度上的差异，找出最能区分和预测写作水平的指标。L2 Syntactic Complexity Analyzer 和依赖句法注释语料库分别用于收集大粒度和细粒度度量的数据。从大粒度的测量来看，写作能力较高的学生往往在写作中产生更长的语言单元、更多的从句、更多的并列从句、更多的名词短语；T-unit 的平均长度、句子的平均长度和每个从句的从句可以比其他传统的大粒度度量更好地预测写作水平。至于细粒度的措施，发现三类从句，即状语从句、补语从句和关系从句，以及两种名词修饰语，即介词短语和形容词关系从句，出现频率更高。更熟练的学习者的著作；复合名词的频率与写作能力呈负相关。补语从句和关系从句，以及两种类型的名词修饰语，即介词短语和形容词关系从句，在更熟练的学习者的作品中出现的频率更高；复合名词的频率与写作能力呈负相关。补语从句和关系从句，以及两种类型的名词修饰语，即介词短语和形容词关系从句，在更熟练的学习者的作品中出现的频率更高；复合名词的频率与写作能力呈负相关。