This research aims to find ways to optimize TTCT originality scoring system, using the generalizability
theory to identify origin of errors in various perspectives. For this, TTCT Figure A (only section 2) had
been carried out on 273 university students in metropolitan areas and scored separately with both
standard and simplified versions. The results are as follows. First, the results by two separate evaluation
guidelines have indicated that both guidelines had shown significantly high error variance between its
items and the subjects, whereas interrater reliability error turned out to be very low. Second, increasing
the number of test scorer was ineffective in seeking the optimal condition of originality grading on both
standard and simplified versions, but results of the study show that much more testing items are
necessary for the test to be more effective. The result suggests that using only 10 items of the TTCT
Section 2, for the sake of convenience, can result in distortion of test results both in standard and
simplified scoring versions. The present study also suggests the need for deep and multi-angular research
for the optimization of TTCT grading system using a more advanced statistical approach.