Distorted perceptions of competence and incompetence are more than regression effects

Students inaccurately assess their own skills, especially high- or low-performers on exams. This study assessed whether regression effects account for this observation. After completing the Infection and Immunity course final exam (IIF), second year medical students (N=143) estimated their performance on the IIF in terms of percent correct and percentile rank. Second year grade point averages (M2GPAs) were combined with the IIF results to form five subgroups: 1=true-low (lowest third on both IIF and M2GPA, 2=false-low (lowest third on IIF only), 3=middle (neither lowest nor highest third on IIF), 4=false-high (highest third on IIF only), 5=true-high (highest third on IIF and M2GPA). The false-low and false-high groups were considered more susceptible to regression effects due to likely group misclassification. Differences between self-assessment and actual performance within each group and between the five groups were used to estimate what portion of observed differences is due to general tendencies versus regression effects. Results found that students accurately assessed their percent correct, but inaccurately assessed their percentile rank. No statistically significant differences existed between the true and false-low subgroups nor the true- and false-high subgroups. Percentages of mean differences suggest that while regression effects resulted in 50-75% over/under-estimates of scores by students who were misclassified, when they were merged with the true-low/high groups, they do not account for more than 14% of low performer over-estimates of their performance and high performer under-estimates of their performance. Accurate percent correct assessments and distorted percentile rank assessments are challenges in using instructional methods dependent on student self-assessments of their learning needs. Identifying and helping students with distorted perceptions of their test performances may be a key issue in such instructional approaches.