A Bayesian approach to implicit L2 learning: Web-based replication of Williams (2005) and Kim et al. (2023) with underrepresented learners

Abstract

Expanding participation beyond university samples is increasingly seen as key to more inclusive and generalizable SLA research. This study examined the effectiveness of web-based experimentation for L2 grammar learning among adults without post-secondary education—an underrepresented population in the field. Building on Williams (2005) and replicating Kim et al. (2023), we implemented a fully remote, researcher-supervised design to enhance accessibility and participant engagement. Forty-nine participants without college degrees completed a semi-artificial language learning task. Bayesian analyses indicated comparable overall learning outcomes, with greater variability in training accuracy and item reliability in the web-based condition. Crucially, no evidence of implicit learning was found among unaware learners—replicating Kim et al. (2023) and highlighting limits to generalizing such effects to non-traditional populations. These findings underscore the importance of inclusive research designs that expand research access while safeguarding data quality.