This study applied a computer program to analyze the descriptors of Japanese characters, including 56 Hiragana, 56 Katakana, and 98 Kanji characters. An experiment was designed to test the legibility of these characters by 40 Japanese students studying in university. Subjects were asked to identify characters displayed in the screen center, shown initially at an illegible size and gradually enlarged until the subject recognized them. The results illustrated that the Hiragana characters were more legible than the Katakana, and the Katakana characters were more legible than the Kanji. Although the legibility thresholds were increased with the number of strokes in these three kinds of Japanese characters, the effects of the number of strokes on legibility among them were different. Increasing one stroke in a Kanji character would necessitate a 0.4‐arcminute enlargement to maintain legibility, whereas no linear stroke effect was found in Hiragana and Katakana characters. This may because of the features of stroke of Hiragana and Katakana are quite different from that of Kanji. Furthermore, for Hiragana and Katakana, adding a semivoiced consonant sign ‘o’ to a common character would necessitate a 7‐arcminute enlargement to maintain legibility and adding a voiced consonant sign ‘ ” ’ to a common character would necessitate a 10.3‐arcminute enlargement to maintain legibility. The results can be used as reference for information and interface design.