A Study of Enhancing Reliability of Classroom Self-Assessment in University

Abstract

The usefulness of self-assessment for evaluation purposes has been controversial among language testers and teachers, pointing to the long-held psychometric concerns of its low-leveled reliability for assessment. This study aimed at exploring the possibility of enhancing the reliability of self-assessment used in a college English classroom (n=32) by diagnosing a cause of deteriorating the reliability quality of self-assessment and seeking measures to offset the negative factor for future use. In order to achieve the goals of the study, internal consistency reliabilities were obtained before and after an improvement was made in the self-assessment. Students’ feedback on the self-assessment was sought through a focus-group interview. Next, a multiple regression analysis was conducted to see if self-assessment can be a significant predictor in students’ final grades in the course. The independent variables used for this study were final exam, midterm exam, and five self-assessments. The results of the analyses are as follows: 1) from the focus-group interview on the 6th week of semester, a subjectivity problem of self-assessments was found and a new set of items in a quiz format were added for its modified version, 2) Cronbach’s α’s of self-assessments have drastically gone up from its original ones (.514-.546) to modified ones (.614-.850), and 3) four variables were included in the final model predicting students’ final grades: final exam, midterm exam, third and second self-assessments. Based on the findings, a new perspective on reliability issues in self-assessment was proposed.