Investigating the L1-L2 relationship through grammatical complexity: A corpus-based study of student academic writing
(2026) Learner Corpus Research conference — Location: Prague (16.September.2026)
Files
No attached file found for this publication.
Details
- Authors
- Abstract
- Grammatical complexity, i.e. “the addition of optional structural elements to ‘simple’ phrases and clauses”, is a central indicator of academic writing performance (Larsson et al. 2023). It nonetheless constitutes a challenge for L2 writers, who struggle to meet the expected types of complexity across tasks and registers (ibid.). Various explanations have been proposed for this difficulty, including L1 influence. For example, previous work has found L1 background to be a strong predictor of L2 complexity, with learners from different L1 backgrounds shown to display different complexity profiles in their L2 writing (Wang et al. 2023, Ehret & Szmrecsanyi 2019, Staples & Reppen 2016). However, these studies infer L1 influence by comparing L2 production across learner groups with different L1 backgrounds rather than examining complexity in the learners’ L1 directly. In addition, very few studies have used paired within-writer data in two languages to understand L1 influence on complexity features (with some exceptions, such as Dirdal 2022). L1 influence claims are therefore largely based on indirect, group-level evidence, leaving the link between individual writers’ L1 and L2 complexity unestablished. Against this backdrop, this study examines the relationship between L1 and L2 grammatical complexity in student academic writing. The following research question is addressed: “To what extent does French-speaking students’ grammatical complexity in L2 English academic writing resemble their grammatical complexity in L1 French academic writing?”. The analysis draws on a corpus containing paired L1 French and L2 English research-based academic texts from 25 students in Literature and Linguistics (C1 English proficiency), carefully matched for discipline, text type (MICUSP classification; Römer & O’Donnell 2011), and text length (average: 3,365 words). This design allows for comparable cross-linguistic comparisons while maintaining ecological validity. Grammatical complexity is operationalized following Biber et al.’s (2011, 2025) framework, focusing specifically on features in Groupings 3A (finite dependent clauses functioning as clause-level constituents) and 1C (phrases functioning as noun-phrase modifiers) (see Biber et al. 2025: 150). These are selected because they exhibit the strongest co-occurrence patterns in general written discourse and represent stereotypical realizations of clausal and phrasal complexity, making them good indicators of register adaptation and academic writing performance. To ensure high precision and cross-linguistic comparability, features are extracted using a combination of automatic extraction and manual disambiguation. For English, the LexGrTagger (Kyle 2025), a Python package built on SpaCy, is used for automatic annotation, followed by manual verification. For French, the SpaCy model fr_dep_news_tfr is used for automatic extraction, with subsequent manual verification to ensure both precision and correspondence with the features captured in English. For both languages, recall rates for individual features (e.g. attributive adjectives) are calculated on a random sample of five texts, and extraction queries are refined whenever recall falls below 90%. Once all features extracted and disambiguated, frequency comparisons are made by means of the paired-samples t-test (with prior checks for normality). Results reveal a mixed picture. Some signs of L1 influence emerge in the clausal domain: verb + that and verb + wh complement clauses show no significant difference between L1 French and L2 English texts from the same individuals. This suggests that French academic writing’s preference for clausal subordination may carry over into the L2 writers’ English writing (De Clerq & Housen 2017, Hamilton & Carter-Thomas 2017). Some signs of adaptation to L2 norms are nonetheless apparent in the phrasal domain: English texts show significantly lower frequencies of of prepositional phrases, and higher frequencies of attributive adjectives, premodifying nouns, and other prepositional phrases. This pattern mirrors English academic writing’s preference for compact phrasal (pre)modification over French postnominal’s patterns (De Clerq & Housen 2017, Hamilton & Carter-Thomas 2017). Further studies should compare the learner data with reference corpora of French and English to better interpret these findings and clarify whether they reflect L1 influence or adaptation to L2 norms.
- Affiliations
Citations
Jadoulle, P. (2026, September). Investigating the L1-L2 relationship through grammatical complexity: A corpus-based study of student academic writing. Learner Corpus Research conference, Prague. https://hdl.handle.net/2078.5/281347