The effectiveness of semantic clustering on vocabulary learning: A multilevel meta-analysis

How to organize and introduce novel vocabulary to second language learners effectively has sparked intense discussion. The semantic clustering method, which involves grouping words based on their semantic relatedness, is a prevalent practice of presenting words in both L2 classrooms and textbooks. Despite extensive research investigating its effects in enhancing vocabulary learning, the findings remain inconclusive. The absence of research synthesis motivated the current study to investigate the overall effectiveness of semantic clustering, and potential moderators to account for the inconsistency in previous findings. A three-level meta-analytic model was fitted to the dataset obtained from 27 eligible primary studies. Contrasting the summary results of trials-to-criterion studies with those of posttest-design studies reveals a large negative effect of semantic clustering during the initial learning stage, as evidenced by trials-to-criterion metrics showing that significantly more trials were required to reach successful recall ( g = 1.02, p = .0002). However, this detrimental effect does not carry over to the later word retention stage, as indicated by the non-significant results synthesized from immediate posttests ( g = 0.01, p = .96). The subsequent moderator analysis examined 12 potential moderators and identified two significant moderators (i.e., training time per session, and provision of context). Specifically, studies with longer training time per session and studies presenting target words in meaningful contexts tend to find semantic clustering beneficial for word learning. This study highlights the need for a more nuanced understanding of the effectiveness of semantic clustering in specific contexts.