Examining the Impact of Linguistic Features of Item Options on Response Patterns and Item Functioning

초록

This study investigated how the linguistic features of multiple-choice options influence examinees’ response patterns and psychometric functioning in English reading comprehension assessments. Using 502 items and 2,008 options from the Cambridge Multiple-Choice Questions Reading Dataset, we extracted computational linguistic indicators, including semantic similarity, lexical overlap, syntactic complexity, and readability. Option-level regression and item-level multilevel analyses revealed that distractors were more likely to be selected when they exhibited greater lexical overlap with passages, whereas correct options were more likely to be selected when they showed stronger semantic similarity. Readability complexity exerted differential effects across option types, such that moderately complex distractors improved discrimination, whereas excessive complexity in correct options weakened it. Overall, these findings provide empirical implications for evidence-based item development, distractor analysis, and automated item generation.

키워드