Using the comparison task of two numerals presented in Arabic, kanji, hiragana and katakana, the study investigated the mechanism of kana and kanji processing. The 108 student sat Grade 4 to 6 were chosen for the experiment on the basis of an equal proportion of gender, grade and reading level. Numerals in the script mismatched conditions (i.e., kanji vs. Arabic, hiragana vs. Arabic and katakana vs. Arabic) were processed less accurately in comparison to the script matched condition of Arabic vs. Arabic. Numerals in Arabic and kanji were processed faster than ones in hiragana and katakana. The numeral length of kana symbols did not show any effect in numeral processing. Therefore, the numerals presented in both hiragana and katakana would be processed in the higher degree of indirect (phonological) processing for numeric judgments with no effects of numeral kana-symbol length, while the same numerals presented in Arabic and kanji would be processed in the higher degree of direct (orthographic) processing. Furthermore, the study clearly illustrated strong influences from the subject's grade and reading level in the efficiency of numeral processing among the four different scripts.