This research compares pupils' self-assessments made from watching their own videos with the self-assessments they made from watching their own videos and a model video simultaneously. The research also examines whether the pupils showed differences in recognizing improvements in their moves and techniques, and it studies the characteristics pupils displayed while recognizing these improvements in the two kinds of self-assessment. We compared the two different self-assessments made by fourth-year elementary school pupils in a mat exercise class. On the basis of the comparison, we found that the pupils extensively reported the improvements in their moves and techniques in the self-assessment of their own videos. This showed that they were liable to report what they wanted to modify consciously and what they noticed during the trial. From the self-assessments made using the dual-video comparison, we inferred that they were liable to raise the evaluation criteria of the exercise taught in the class, that they easily identified improvements they had not recognized from their own videos, and that they were likely to disregard any recognition they made by themselves.