Identifying and explaining L2 growth spurts: A tutorial on generalized additive models for time-intensive longitudinal data in applied linguistics research

Abstract

Recent years have seen a marked increase in the use of time-intensive longitudinal designs in applied linguistics, particularly in second language acquisition (SLA) research, where individual developmental trajectories have become inferential targets in their own right. Generalized additive mixed-effects models (GAMMs) have emerged as a powerful tool for modeling between- and within-person variation, for disentangling linear from nonlinear relationships, and for assessing the predictive power of (relatively) static (e.g., gender, educational attainment) alongside time-varying (e.g., socioaffect, cognition) predictors on developmental pathways. Setting our tutorial apart from other GAMM resources in computational linguistics, phonetics, and sociolinguistics, we present the first in-depth application of GAMMs to identify and explain periods of significant change (growth or decline) in longitudinal datasets with repeated measurements. We draw on a novel micro-development study comprising 43 older adult L2 learners who completed a battery of L2, socioaffective, and cognitive tasks in 30 consecutive waves across a period of two years (i.e., 30 measures per participant per task). Our contribution guides readers through the computational steps to identify periods of statistically significant change. We then illustrate how findings can be interpreted and supplemented with both qualitative introspective data and quantitative measures of individual learner differences. While our tutorial focuses on SLA research, the methods are applicable to any number of disciplines in the social and natural sciences where developmental patterns are of direct inferential interest.