Japanese Kanji characters have various degrees of consistency of character–sound correspondences in multicharacter words. A word was classified as consistent, inconsistent typical, or inconsistent atypical with reference to the most typical pronunciations for constituent characters among words sharing the same characters. A nonword was classified as consistent, inconsistent biased, or inconsistent ambiguous according to the degree of pronunciation typicality of its constituent characters in real words. A word-naming experiment yielded a significant Frequency × Consistency interaction, and a nonword-naming experiment yielded significant consistency effects. In addition, both word frequency and lexicality exerted strong effects on efficiency of naming Kanji character strings. These results demonstrate the influence of Kanji print–sound correspondences both at subword and whole-word levels.