This study examines how vocabulary assessment outcomes are jointly shaped by task format, learner characteristics, and lexical properties. Vocabulary measures often assume that different task formats capture equivalent knowledge; we challenge this assumption by comparing synonym and definition identification tasks across learners with varying language proficiency. Over 6,000 linguistically diverse middle school students from an urban California district completed a 50-item synonym task and a randomly assigned 12-item definition form, with overlapping target words across formats. Explanatory item response theory models incorporated person-level predictors (language proficiency, grade, gifted status, socioeconomic status, reading comprehension) and item-level lexical features (frequency, complexity, proximity, polysemy, diversity), estimating both main and interaction effects. Students were generally more accurate on synonym items, but this advantage was absent for Limited English Proficient learners. Words with multiple meanings and broad contextual use were easier, particularly on synonym tasks. Results demonstrate systematic interactions between task, learner, and word, underscoring the need for vocabulary assessments that account for task effects and learner diversity.
