Contrasting Fixed‐ and Mixed‐Effects Modeling in Vocabulary Research: Reanalyzing Laufer (2024) and McLean et al. (2020)

Abstract Analyses in vocabulary research should avoid the language‐as‐a‐fixed‐effect fallacy, whereby no statistical evidence is provided to support claimed generalizations beyond the words tested in the sample. Although mixed‐effects models are widely adopted in social sciences to avoid this fallacy, second language vocabulary researchers primarily conduct potentially problematic fixed‐effects analyses. In the present study, two published vocabulary studies relying on fixed‐effects modeling were re‐analyzed with generalized linear mixed‐effects models (GLMMs). Consistent with prior research comparing these approaches, effect sizes in the GLMMs were reduced by 36% to nearly 80%. Crucially, one study's claims were not fully substantiated with GLMM re‐analysis. The findings suggest that second language vocabulary researchers should strongly consider mixed‐effect models to avoid the language‐as‐a‐fixed‐effect fallacy. Furthermore, replications of earlier studies that employed fixed‐effects only analyses should be conducted to verify that their effect sizes were not overstated.