Comparison of L1 and L2 writing fluency on the basis of a corpus of screencast videos and keystroke log files

Gilquin, Gaëtanelle [UCL] d’Herbais de Thun, Valérie [UCL]

Writing research has moved away from product-oriented approaches to also consider writing as a process, involving, among others, pauses and revisions (Chenoweth & Hayes, 2001). This complex process is what we intend to examine, with a focus on writing fluency in L1 and L2. The study uses the PROcess Corpus of English in EDucation (PROCEED) (Gilquin, 2021), a new kind of corpus which includes three types of data complementing each other: (i) final written products, namely argumentative texts from university students with at least two texts per participant (one in their L1, i.e. French, and one in their L2, i.e. English), as well as (ii) screencast videos and (iii) keystroke log files, which make the writing process visible and which have rarely been combined with each other in writing research. PROCEED also comes with rich metadata, for example on students’ typing skills or linguistic profiles. The analysis of writing fluency in PROCEED will be twofold. First, it will involve a quantitative approach combining different fluency measures (e.g. inter-key intervals or pause locations). Second, in a more qualitative perspective, it will rely on the thorough annotation of the data (cf. Laporte & Gilquin, 2018) and on the exploitation of the metadata to identify different fluency profiles (based on revisions, pauses, etc.) and to see how they correlate with learners’ proficiency and their overall text quality, but also how they correlate between learners’ L1 and L2. With these different metrics and qualitative data, we aim to uncover how L2 writing fluency compares with that of L1 and what potential differences can tell us about the cognitive dimensions of both L1 and L2 writing processes. By adopting a corpus-based method, our study will add to the existing cognitive literature on writing processes (e.g. Kellogg, 2001) with a high degree of generalizability and systematicity, as well as sophisticated techniques of data extraction and analysis.