This study investigates how syntactic features of learners' interlanguage develop nonlinearly. Longitudinal data (T = 7) from Saudi learners of English (N < 90) analyzed using GAMMs and Bayesian mixed effects models showed nonlinearity and systematic variability in developmental pathways, and demonstrated the effect of task iteration in learner development.
Learners' interlanguage development is nonlinear and characterized by phases of stability alternating with high degrees of variability that accompany rapid development (e.g., Lowie & Verspoor, 2019). Variability is thought to be a necessary feature of L2 development (Verspoor & de Bot, 2021), and is a source of meaningful information both quantitatively, because such data show fluctuating levels of the variables of interest, and qualitatively, in the sense that multidimensional combinations or juxtapositions can also be illustrated for each measurement occasion (Larsen-Freeman, 2006). Within an overall trajectory of development, variability can be investigated in the context of task performance--particularly task iteration. Task iteration is repeated engagement with a task at a given interval that generates variation. Task iteration creates variation because learners orient to it differently at each iteration. This creates options in learners' language resources and gives them choices for making meaning. In this paper we report a longitudinal study of how the syntactic features of learners' interlanguage develop over time. Eighty nine Saudi learners of English in their foundation year of college were asked to compose a descriptive text in English on a topic of personal choice. Participants did this every two weeks over the course of a semester (T = 7) as part of their 3 hours per week of language instruction. Text type was controlled for by providing a writing prompt where the field of writing was varied but the tenor and mode of writing were fixed across all waves. Syntactic complexity was examined through length of production units (MCL = mean clause length). Using Generalized Additive Mixed Models (GAMMs; Murakami, 2016; Wieling, 2018) and Bayesian mixed-effects negative binomial location-scale models, we examined whether learners' development show nonlinearity, stability, and variability; what this stability and variability in the data illustrate about learners' development; and, whether iteration of the same task procedure produces different effects over time.