The rules for how to determine the radical of a kanji, together with the radical table, constitute the radical system of a character dictionary. While the radical table lists the characters and character-components that could be a radical (“radical candidates”), what determines where in the dictionary any given kanji is to be found are the rules that determine which of the several radical candidates in the kanji actually is its radical. In those cases in which the kanji is itself a radical or contains only one component that is listed in the radical table, these rules are not needed. But since about 90% of all kanji contain between two and seven components that are listed in the radical table, it is indispensable to have rules for deciding which of them is the kanji’s radical. Theoretically, a kanji could be listed under every one of the radical candidates it contains, thus obviating the need for choosing among them. But in practice, all character dictionaries list each kanji under only one of its components, lest the dictionary become too bulky.

This limitation to just one radical per character does not apply to electronic character dictionaries, which, along with their systems for looking up a kanji, are discussed in separate articles.

Most monolingual Japanese character dictionaries are arranged according to the traditional radical system. Their use is made considerably more difficult by their lack of rules for determining the radical of a kanji. Instead, they all follow the character dictionary 康熙字典 Kangxi zidian (in Japanese, Kōki jiten), which was published in China in 1716, and which does not explain by what standards the sequence of listed kanji is determined. In other sources than the dictionaries themselves it is stated that the meaning-bearing part of a kanji is its radical. Apart from the fact that one character will have multiple meaning-bearing components, such a rule assumes that anyone trying to look up a kanji in a character dictionary already knows what the kanji means, which is just the information that the user is consulting the dictionary for. Thus the user is left in a paradoxical lurch, because a reference to the Kangxi zidian, which is known to most Japanese only by name and which moreover was compiled in Chinese, is of little practical use, especially considering that its rules for determining the radical of a character are never clearly stated. If despite these drawbacks a user can find the kanji he is looking for, it is only because the most important meaning-bearing component of a kanji is usually found in a definite position within the kanji, namely on its left or top, or as an enclosure around the rest of the kanji. If a kanji has two possibilities for which part might be its radical, for example the left-hand part of the kanji and its right-hand part, or a top part and a bottom part, then the left takes precedence over the right, and the top over the bottom. Every user of a traditional dictionary soon notices this and abandons the search for the meaning-bearing part of the kanji, determining the radical instead by its position within the character.

The fact that the radical correlates well with its position within the kanji, rather than according to its relevance to the meaning of the kanji, has been exploited by Western lexicographers and been made into a general rule. In 1924 Arthur Rose-Innes, and later in 1962 Andrew N. Nelson, developed explicit rules for determining the radical of a kanji in their character dictionaries. These rules in turn were a model for the 79-radical system of the character dictionaries that have appeared since 1989, which make it even simpler and quicker to look up a given kanji.

The following page presents an annotated overview of the rules for determining the radical of a kanji in the 79-radical system, along with hints (one can hardly call them “rules”) for determining the kanji of a radical in dictionaries that are based on the Kangxi zidian. Relevant Japanese-English works are discussed first, followed by the two monolingual character dictionaries of the most important kanji lexicographer of the 20th century and compiler of the most comprehensive character dictionary of all time (about 49,000 kanji, in 13 volumes), the Japanese Sinologist Morohashi Tetsuji (1883–1982).


M. Spahn, W. Hadamitzky: The Kanji Dictionary. Tuttle 1996

The above table is a summary of the actual rules in which the individual steps for determining the radical of a kanji are explained in detail.

These rules apply not just to the large dictionary but also to:
M. Spahn, W. Hadamitzky: The Learner’s Kanji Dictionary. Tuttle 1998
W. Hadamitzky: Japanese, Chinese, and Korean Surnames and How to Read Them. Vol. 1, 2. Saur 1998
W. Hadamitzky, M. Spahn: Kanji & Kana. Revised edition. Tuttle 1997 (the radical index is arranged according to these rules)


A. N. Nelson: Japanese-English character dictionary. Completely rev. by J. H. Haig.
Tuttle 1997

Quotation from the Foreword, page vii:
“... the reversion to the traditional radical system, more than compensate for the loss of the Radical Priority System, ...”

Correction: The new edition is not a “reversion” to the traditional system, but a conversion to it. A “reversion” implies that the work once arranged its entries in a traditional order, which it never did.

The Radical Priority System that Andrew N. Nelson (1893–1975) introduced in the 1962 first edition for determining the radical of a kanji has been done away with in the 1997 revised edition (see above), without being replaced with new rules. The characters are now arranged according to the traditional radical system, but without revealing what the traditional rules for determining the radical of a kanji are. The beginner or occasional user can only guess which of the up to seven components of a kanji is the radical it is listed under.

The publisher of the new edition has obviously recognized the problem. To be sure, it is explained that the tried-and-true Radical Priority System has been done away with and that the characters are arranged in a traditional way. But in the chapter “How to determine the radical of a character” (pages 1234–37), the “12 Steps” for determining the radical of a kanji (except for Step 12) that constitute the Radical Priority System are again introduced, and indeed this is the only set of rules that appears in the new edition. In contrast to the old edition, the term “Radical Priority System” is no longer used, and the rules themselves are accompanied by the warning, “The rules are not absolute, but the guidelines will generally lead the user to the correct location.”

Here is a summary of the “12 Steps” from the 1997 new edition, without the explanation and examples given in the book:
“1. All?
2. Lone?
3. Enclosure?
4. Left?
5. Right?
6. Top?
7. Bottom?
8. NW? [northwest]
9. NE?
10. SE?
11. SW?
12. Inside?

Bottom line: The rules for determining the radical of a kanji that were worked out by Nelson and have been tested out over decades of use are practically the same in both the old and the new Nelson. The difference is that in the earlier editions the ordering of the characters followed these rules 100%, while in the new edition these rules lead to finding a kanji on the first try less than 90% of the time. It can be assumed that this is hardly what the author, who died in 1975, intended.


大漢和辭典　諸橋轍次他著　全13巻 大修館書店 1984–1986.
Dai Kan-Wa jiten. Morohashi Tetsuji et al. Revised edition. Volumes 1–13. Publisher Taishūkan 1984–1986 (1st edition 1955–1960)

Quotation from chapter heading (from volume 1, user’s guide, page 4; English translation): 
“Arrangement of characters and words
1. The characters listed in this work are arranged mainly as in the Kangxi zidian. That is, all the characters are first classified in the order of the 214 radicals, and within each radical the characters are a listed in sequence according to their stroke count. ”

As is usual in traditional character dictionaries, the directions for use refer to the Kangxi zidian, without explaining its rules for determining the radical of a kanji. These rules are assumed to be known.


廣漢和辭典　諸橋轍次他著　全４巻　大修館書店 1981.
Kō Kan-Wa jiten. Morohashi Tetsuji et al. Volumes 1–4. Publisher Taishūkan 1981

Quotation from Volume 1, page 4, Directions for use (English translation; annotation in brackets)
 “10. The head characters are listed by radical, and within the same radical by stroke count.
11. The selection and sequence of the radicals are based on the Kangxi zidian, with the following modifications for easier lookup:
a) [the three strokes at the top of 単] has been introduced as a new radical. ...
Characters that have lost their original radical are listed under another radical of the traditional set. ...
b) The following radicals are listed separately:
[This is followed by a list of radicals that formerly were taken to be variants of each other but now are listed as separate radicals.]
c) The following radicals have been combined:
[Pairs of radicals (like 日 and 曰) that could be confused with each other because of their similarity are combined under the more frequently occurring radical; that is, some of the relatively rare radicals in the Kangxi zidian are taken to be variants of a more common similar radical.]
d) Some characters, whose radical classification in the Kangxi zidian is illogical or would make the character difficult to find, have been reclassified under a different radical. ...
[Also, some radicals now are listed under a different number of strokes.]”

Although the directions for using this dictionary give a relatively detailed explanation of the deviations from the traditional radical system of the Kangxi zidian, that system itself is not explained but is assumed to be known. This is especially true of the criteria for determining the radical of a kanji, which apart from the radical table is the most important part of a radical system.


October 2005, W.H. with Mark Spahn