Reliable evaluation and grading of students' clinical and laboratory performance continue to be a problem for dental education.
In this article, 31 samples of tooth carving product selected from 138 whole product at random were evaluated under 5-grading-system and reliability of the rating grades, especially interrater variance, was studied. Furthermore, in order to enhance the reliability, two modified rating methods were presented. The results were as follows:
1. When rating was left to raters' criteria, large interrater variance came out in assigning carving product to each grade.
2. Correlation coefficient between 10 raters' grades for the same product varied from 0.20 to 0.88 (mean 0.60).
3. Carving product evaluated with large interrater variance had always an inharmonious part which deranged the whole integrity and was generally above average-ranked one. On the contrary, carving product evaluated with consistency was very poor one.
4. When rating was repeated within the same raters, the rating grades were comparatively reliable; 0.64-0.94 (mean 0.79) of correletion coefficient.
5. In the viewpoint of reliability of rating grades, it was recommendable in the rating process to classify the carving product into two or three groups.
6. When approximate numbers of carving product at each grade were provided on the assumption that carving skill of the dental students indicates normal distribution, the reliability of rating grades was improved; 0.43-0.85 (mean 0.71) of correlation coefficient.
