The consistency of students' self-assessments in short-essay subject matter examinations

This longitudinal study compares the accuracy of self-assessments of 22 students across four examinations during their first 2 years of medical school. The four examinations used a similar short-essay format and covered many of the same basic science disciplines at similar levels of difficulty. Immediately after answering an average of 20 questions on each examination, students predicted their performance on those questions. After assigned subject matter experts had scored the questions, the differences between students' predictions and the experts' scores were calculated for each question. The degree to which students had over- and underestimated their performance across all questions was determined by separately averaging all positive and negative differences between students' and experts' assessments on each examination. The results of the study indicated that accuracy in self-assessment improved from examination 1 to examination 3 (with less overestimation) and dropped on examination 4 (with more underestimation). The results revealed no relationship between self-assessment estimations and actual scores received. Furthermore, the self-assessment estimations tended to be statistically correlated between contiguous examinations (i.e., examinations 1 and 2, 2 and 3, etc.) but not between non-contiguous ones (i.e., examinations 1 and 3, etc.). The results of the study are interpreted to suggest that the students in the study have a self-assessment tendency towards over- or underestimation that is somewhat stable but that gradually evolves over time with experience, maturity and self-assessment practice. The most frequent direction of change is towards decreased overestimation and increased underestimation. These results are consistent with the findings of other recent longitudinal self-assessment studies.