A corpus linguistic approach to L2 writing processes: Opportunities and challenges

Corpus linguistics, thanks to its focus on authentic language and its use of techniques that uncover language patterns, has provided important insights into writing, including L2 writing (e.g., Römer et al. 2020). So far, however, it has almost exclusively investigated the written product, neglecting the processes through which texts are written. While these processes have been the subject of empirical studies (e.g. Roca De Larios et al. 2002), few of them have been (explicitly) anchored in the framework of corpus linguistics. This presentation will discuss the challenges and opportunities of adopting a corpus linguistic approach to L2 writing processes. It will introduce the Process Corpus of English in Education (PROCEED; Gilquin 2022), a learner corpus of academic writing in which each text (corresponding to the final product) is accompanied by a screencast video and a keylog file (representing the writing process), as well as information about the learner’s sociolinguistic and cognitive profile. It will be explained how this corpus compares with more traditional learner corpora and makes it possible to go further in the study of learner language acquisition by giving access to writing process data. Two cases studies based on PROCEED will be presented, one on the use of online writing tools (which exploits screencast videos; see Gilquin see Gilquin forthcoming). Both case studies rely on tools and techniques taken from corpus linguistics, such as the ELAN software (Brugman & Russel 2004) for corpus annotation and the VariAnt (Anthony 2017) and AntConc (Anthony 2022) programs for data extraction. It will be shown that these tools and techniques facilitate the analysis of the data, through the (partial) automation of certain steps in the analysis and the identification of recurrent patterns, but that they still make the language production sufficiently visible to allow for detailed qualitative studies. Some of the difficulties linked to the exploitation of these data will also be described, such as the alignment of the different types of data and the interpretation of some observed phenomena.