Multiple-choice exams : student explanations for answer choices

This study examined relationships between three variable domains: (a) students' explanations for their answers to difficult multiple-choice items, (b) generic critical thinking, and ( c) performance on different types of exam items. Students took a critical thinking test at the beginning of a large human development course and 5 multiple-choice exams during the course. Prior to turning in each exam, students prepared written explanations for their answers to the 4 items judged most difficult by the instructors. Written explanations were reviewed for accuracy of responses, number of options considered, and the presence of faulty thinking strategies. Specific questions addressed by the study included the following: (a) What is the correlation between explanatory dimensions (i.e., accuracy, options, error) and different types of exam items (i.e., total, recall, comprehension, mixed, and target)? (b) What is the correlation between base-level critical thinking and the explanatory dimensions? ( c) To what degree do the explanatory dimensions predict exam performance and critical thinking scores? ( d) How well does critical thinking as a singular predictor account for scores on item types? and ( e) What differences in the explanatory dimension, critical thinking, and item-type scores exist between high-, average-, and low-performing students on the exams? Accuracy and options significantly and moderately correlated with all item types. The error dimension significantly, but weakly, correlated with all item types except the recall items. Critical thinking significantly and weakly correlated with accuracy and options, but did not correlate significantly with the error dimension. All explanatory dimensions significantly predicted exam performance measures, but the number of 111 options explained proved to be the strongest predictor of exam measures. Options also was the only explanatory dimension that significantly predicted critical thinking. Critical thinking as a singular predictor accounted for a small amount of the variance in all item types, but accounted for the highest percentage of variance in the comprehension and total items. Significant differences existed among all groups with respect to all dependent measures except for the error dimension between the low- and average-performing groups.