In the modern Japanese logographic writing system, both Kanji and Kana are used. Since they have quite different characteristics, quite different speech processes in the two hemispheres of the brain are involved. Some previous studies show that there are different neural mechanisms linked to the memorization of Kanji and Kana when Japanese native speakers are concerned. In order to illustrate the neural mechanisms related to the process in which the Chinese Japanese learners memorize Kanji and Kana, this paper investigates the differences between encoding and retrieval of Kanji and Kana during the memory process, by comparing the Event-Related Potentials (ERPs) components which are elicited by Kanji and Kana during the process of memorizing and recollecting them by Chinese Japanese learners. In the study phase, significant differences are found in considering the amplitude and the latency of the ERPs components elicited by Kanji and Kana condition, and the scalp distributions are different, too, which indicates the different aspects of the memory storage after Kanji and Kana are encoded. The encoding of Kanji is faster and deeper than that of Kana. In the test phase, the old items of Kanji elicit lower amplitude and shorter duration of FN400 and higher amplitude and shorter duration of the Late Positive Component (LPC) than the old items of Kana. The old items of Kanji are easier and quicker to be recalled than the old items of Kana. In addition to this, there are obvious hemisphere advantages for encoding and retrieval of old items of Kanji, which demonstrates that the neural mechanism related to Kanji is different from that related to Kana. It is believed that because the Kana's memory processing is more complicated than Kanji, which increases the memory load, leading to inferior effects than Kanji. In this way, the illustration of the memory feature of Kanji and Kana is instructive to Japanese learning.