The role of printed and self-rated usage frequencies of Chinese characters was investigated by comparing the effects of printed and self-rated frequencies on the reaction time for a cognitive task. The self-rated frequency used in this study was a subjective estimate of the usage frequency of a Chinese character in daily life. Participants rated the Chinese characters once or three times. Then, they judged whether a presented character was a Chinese character or not. The reaction time for making the judgment was recorded. When rating the character only once, low-frequency characters were judged more slowly than high-frequency characters, irrespective of whether the printed or self-rated frequency was used. When rating the character three times, the difference in reaction times between high and low frequency characters disappeared when reaction times were arranged according to the rating value of the first occasion. In contrast, the relationship between reaction time and frequency did not change when the printed frequency was used. Results indicate that frequency estimate was changed by repetition of rating frequency. The theoretical implications of these findings in terms of the judgment of frequency are discussed.