In a pseudo-longitudinal study that aims to understand the development of complexity in L2 writing, we demonstrate the utility of an approach to studying complexity using information-theoretic measures based on Kolmogorov complexity and complexity contours that track the progression of complexity within a text.
In this paper, we address two major challenges in L2 complexity research: The first challenge concerns the fact that widely used complexity metrics suffer from what Ortega (2012:128) refers to as ‘concept reductionism’ and that they may yield inconclusive and/or misleading results (Ehret & Szmrecsanyi, 2019, for a discussion). The second challenge relates to the fact that complexity research has typically relied on a ‘global’ assessment of text complexity, obscuring the considerable degree of variation of complexity within a text. Building upon recent work (Ehret & Szmrecsanyi, 2016, 2019), in addition to traditionally used complexity measures (CMs), we employ information-theoretic CMs based on Kolmogorov complexity that utilize compression algorithms to provide holistic measures of structural surface redundancy that have the potential to avoid the limitations of traditional metrics mentioned above. Moreover, rather than applying complexity metrics to the entire text - through a single aggregate score, we track the progression of complexity within a text using a sliding window approach. We demonstrate the utility of this approach in a pseudo-longitudinal study that aims to understand the development of complexity in L2 writing. The data for this study come from a relatively large corpus of authentic writing samples produced by 690 secondary school children across grades 7-12. These data also include information from the LEAP-Q (Marian et al., 2007), a questionnaire used to obtain information on self-rated proficiency for three language areas (reading, understanding and speaking) and self-rated current knowledge of L2 English and exposure to the L2, as well as indicators of receptive vocabulary (Vocabulary Levels Test; Schmitt et al., 2001, LexTALE; Lemhöfer & Broersma, 2012). Two research questions are addressed: (1) what is the relationship between traditional and information-theoretic trajectories of complexity development, and (2) how are these trajectories affected by individual differences in the language background variables.
