Intuition versus Corpus Data: Assessing the Effectiveness of Lexical Judgments in Game-Based Materials Selection

Abstract While game-based learning offers potential benefits to the language learning process (i.e., novelty, randomization, constrained choice), massed-practice conditions inherent in gameplay necessitate careful material selection. Data-driven approaches, readability research, and vocabulary learning principles highlight the importance of matching text to support learning development and mitigate frustration, disinterest, and disengagement. In game-based learning, however, mismatched text poses an additional risk: erroneous form/meaning mappings through fossilized errors. While teacher mediation through pre-teaching and debriefing can help address these issues, not all game-based learning contexts have direct teacher support (e.g., self-access centers or out-of-school learning). Thus, it is essential to understand how well learners are able to assess such learning materials. Using data from a card-based language game, this study developed a lemmatized high-frequency word list to establish at- and above-level thresholds and examined differences, reliability, and classification performance in learners’ at- and above-level lexical judgment scores. The relationship between judgments and word frequency was further analyzed beyond these corpus-based thresholds. Findings suggest that learners can contribute to material selection, particularly with at-level items. However, their poor classification performance of above-level items underscores the continued importance of teacher mediation in game-based learning.