Dr. WONG Man Ho, IvyIvyDr. WONG Man Ho2026-07-022026-07-022025Wong, M. H. (25-28 Jun 2025). Beyond p-values: A bayesian perspective on cognitive linguistics-inspired instruction for English prepositions. The 34th Conference of the European Second Language Association (EuroSLA 34), Tromsø, Norway.http://hdl.handle.net/20.500.11861/27899This study explores the intersection of cognitive linguistics-inspired instruction and Bayesian statistical modelling in second language (L2) acquisition research. A close replication of Wong, Zhao, and MacWhinney (2018), it investigates how cognitive linguistics-informed feedback strategies impact the learnability of English prepositions among 81 Chinese-speaking learners of English. Uniquely, this study not only collected new data but also re-analysed data from the original study, using it to extract prior information for the Bayesian models and to enable direct comparisons between the original and replication studies. Additionally, unlike the original study, which only included pre- and posttests, this replication incorporated a delayed posttest to assess the long-term retention of learning. By replacing traditional null hypothesis testing with Bayesian mixed-effects logistic models, this research provides a nuanced understanding of both immediate learning outcomes and sustained effects.<br> Cognitive linguistics (CL) emphasizes the conceptual links between spatial and abstract meanings of prepositions, effectively presenting these ubiquitous meanings in a systematic fashion for L2 instruction. Through schematic diagrams, learners can visualize the underlying conceptual relationships between spatial and non-spatial prepositional polysemes, such as spatial containment (e.g., in the box) and emotional states (e.g., in love). The study compared three feedback conditions—schematic diagram feedback, rule-and-example feedback, and corrective feedback—with a focus on how each method promoted receptive and productive knowledge differently. The CALL environment provided a consistent platform for delivering feedback and instructional materials, ensuring full transparency and relatively balanced pedagogical richness across the experimental groups. This replication revealed that cognitive linguisticsinformed instruction fosters both immediate and sustained gains in L2 learners’ productive and receptive knowledge of prepositions. Specifically, animated schematic diagram feedback was most effective in facilitating the learning of spatial and nonspatial preposition polysemes, particularly among learners from lower-performing schools.<br> From a methodological perspective, the adoption of Bayesian mixed-effects models represents a significant advancement in SLA research, particularly in replication studies. Unlike frequentist methods, the Bayesian approach uniquely incorporates prior knowledge from the original study, enabling the integration of valuable insights into the replication process. This allows researchers to directly calculate the probability of a parameter or condition being better than another, offering actionable and intuitive information for real-life decision-making in instructional design. By moving beyond binary significance testing, Bayesian modelling provides richer, probabilistic insights that are both transparent and interpretable.<br> While mixed-effects modelling itself allows for the disentangling of fixed and random effects and the accommodation of individual variability, the Bayesian framework extends these capabilities further. For instance, the inclusion of prior distributions enhanced the precision of parameter estimates, and posterior predictive checks confirmed the robustness and validity of the models. This replication study demonstrates how Bayesian inference not only strengthens the credibility of findings but also improves the ability to generalize results across different contexts. Moreover, Bayesian analysis thrives in scenarios with small sample sizes, a common limitation in SLA studies, by leveraging prior information to maximize statistical power and reliability.enConference Paper