For English speakers, I know of three heavyweights in the ``Kanji
- ``Nelson'' -
The Modern Reader's Japanese-English
by Andrew Nelson, Tuttle, 1962.
The old edition
is still available, but there's also a new edition
out, but I have not seen it.
- ``S&H'' -
The Kanji Dictionary
by Mark Spahn and Wolfgang Hadamitzky.
First published in 1989 by Nichigai Associates under the name
``Japanese Character Dictionary -
With Compound Lookup via Any Kanji,''
revised and republished in 1996 by Tuttle.
- ``Halpern'' -
by Jack Halpern, published in 1990 by Kenkyusha and in 1993 by NTC.
There is also a learner's version (online here).
(See it at Amazon.com)
When considering these and others -- which to buy, which are good, which to
keep nearby on the shelf, etc. -- there are two primary concerns to keep in
I'll address the second point first.
- What indexing method or methods are used to find the entry for a
- What information is available once you get there.
Information a ``Kanji Dictionary'' Provides
It is important to keep in mind that there are several completely
different types of reference works called ``Kanji Dictionaries.'' In
Japanese, this is usually written 字典 (jiten)
which happens to be pronounced the same as 辞典,
which is a more traditional ``dictionary.''
Works might have their PRIMARY GOAL to
I've never seen one work that encompasses everything.
provide meanings of words -
Provides translations of words from Japanese to English, but has them
ordered and grouped via kanji rather than pronunciation. Examples are
Nelson, S&H, edict, and the Canon Word Tank (which is also a normal
dictionary as well),
provide an understanding of how a kanji is used -
Provide information helpful for the student of Japanese kanji, for study.
This might well include words (as is the primary goal of the type above)
and data (primary goal of the type below), but this information is only
provided as support toward the overal goal of showing how the character is
used in Japanese. Important also are discussions of the ``feel'' a
character might give a word, examples of use with relevent grammatical
comments, notes on common mistakes and areas of confusion, and perhaps
stroke-order diagrams. An example is Halpern.
provide data about kanji -
Some might describes in detail information related to a kanji. Its
readings, perhaps its encodings via various human and computer encoding
standards. Perhaps variant forms, old forms, and forms from other
languages. Perhaps its etymology and stroke ordering as well. An example is
kanjidic and the JIS X 0208 standard.
If you merely want to look up words
For looking up words, S&H or Nelson are probably what you want. I have
never used Nelson, but it has been ``the standard'' since long before I was
For the most part, these kinds of ``Kanji Dictionaries'' are really ``Word
Dictionaries'' with the entries arranged via kanji. In fact, the Japanese
title of S&H translates to ``Kanji-English Word Dictionary'', which is
really more appropriate than ``The Kanji Dictionary''.
Often, these works limit their attention to jukugo, words
composed of combinations of kanji, for the most part giving light or no
treatment to the kun'yomi meanings of a character. For example, the
character 掛 is the ka in `kakeru',
`kake', `kakaru', and `kakari'. Yet the many meanings of these
stand-alone words are often glossed over or ignored. Contrast this with
For the most part, English translation is simply a word or two -- this is
usually enough to give a good indication, but once the reading is known, a
``real'' dictionary should be checked if there are still questions. This
makes these works ``glossaries'', not ``dictionaries'', but few make the
distinction (heck, I call my server a ``dictionary server'' but it uses
edict, which is certainly a glossary).
If you want to study kanji
If you, as an English speaker, want to study kanji -- get a feel for their
practical use across the span of the language, then you want Halpern. The
purpose of Halpern is not to provide an exhaustive list of words, but
to provide a representative ``feel'' for how a character is used in
Halpern is not designed for bulk translation work; it is designed for
study. He presents detailed ``core meanings'' of each character, and
partitions a character's actual use in the language into senses that helps
the student absorb usage patterns. He gives equal attention to all types of
uses, including kun'yomi, on'yomi, independent, and
special readings. Examples are a mixture of not only isolated words, but
phrases and sentences with appropriate grammatical points of interest.
(See the dictionary's home page for a lot more information on
what it provides.)
To extend on the kakeru example from above,
consider what S&H says about this oft-used word:
hang (tr.); put on top of; turn on, start;
spend; multiply; (as suffix) begin to, start ...ing
Contrast this with Halpern's page-and-a-half on kakeru,
giving 42 examples on this word alone (the entire entry for
掛 uses 132 examples). Here are a just few of the examples:
None of these are covered in S&H.
- 薦を掛ける [komo o kakeru] spread a mat
- 計りに掛ける [hakari ni kakeru] weigh on a scale
- 眼鏡を掛ける [megane o kakeru] wear glasses
- 火を掛ける [hi o kakeru] set fire
- 電話を掛ける [denwa o kakeru] make a phone call, telephone
- 謎を掛ける [nazo o kakeru] pose a riddle
- 保険を掛ける [hoken o kakeru] insure (something)
- 迷惑を掛ける [meiwaku o kakeru] impose trouble
- 心配を掛ける [shinpai o kakeru] cause someone to worry
- 気に掛ける [ki ni kakeru] have one's mind upon, take to heart
- 錠を掛ける [jyou o kakeru] fasten a lock
- 医者に掛ける [isha ni kakeru] entrust to a doctor (for treatment)
- 時間を掛ける [jikan o kakeru] spend time (on doing something)
- 八掛ける二 [hachi kakeru ni] 2 times 8
He also provides synonyms and homophones, and occasionally includes
interesting side notes on how interesting compounds have been formed. For
example, the component characters of 傾国
(keikoku) tend to mean ``ruin'' and ``country'', yet the word means
``beautiful woman'' and/or ``harlot''. Halpern notes:
woman; courtesan' derives from an ancient Chinese legend of a beautiful
siren that brought on the downfall 傾 of a nation
(国) with her voluptuous charms.
Halpern also provides generous usage notes. For example the four words
変える、代える、換える、替える are all pronounced
kaeru and all generally mean ``to exchange''. Yet their meanings
have some not-so-subtle differences. He gives examples using each
word. Their English are:
(see page 930 of Halpern for the full treatment).
I find these extremely useful.
- Change a horse into a cow (by magic)
- Use a horse in place of a cow
- Exchange a horse for a cow
- Replace an old horse with a new one.
Halpern also provides a stroke-order diagram, variant
and old forms, various calligraphic (handwritten) styles (square,
semi-cursive, cursive), common handwritten abbreviations, both Ming and
Gothic typeface glyphes, as well as Chinese forms and readings of the
character. Japanese readings are indicated as approved (by the
Ministry of Education), unapproved but common enough to include,
name-only, or special.
Other per-character information includes the traditional
radical (with special notations for ``lost radicals''),
the frequency-of-use in modern printed Japanese, and The
Ministry-of-Education grade designation.
As I said, Halpern is for studying kanji. It is perhaps the one for which
the name ``Kanji Dictionary'' is truly applicable. I often use Halpern
merely to look up words because I like the indexing method (see below)
better than S&H, but I realize that it is hit'n'miss for this. Halpern uses
30,841 distinct words as examples when describing the senses that a
character can impart to a word, but the coverage is spotty. For example,
字典 is listed under
典 but not 字.
He apparently found that other words better represented the meanings of
How to Find What You're Looking For
A reference work that provides all the information you want is useless
unless you can actually find it. To find the entry for a particular
character, any one work generally has several indexing methods, with one
main method that sets the ordering of all the characters.
All works should provide an index via reading. Thus, if you know a reading
of the character in question, you can check the index at the back which
will then point you to the proper entry.
The real question is, what if you don't know the reading but can only see
the character? First of all, most methods require that you know how to
count strokes when you see a character properly written. If you've never
looked at this stuff before, you might be surprised to find that
口 has 3 strokes, while 市 has five (but
not the five you might think at first glimpse -- the vertical line is
actually two strokes!) Smaller works will often have a stroke-count index,
but this is simply too inefficient for larger works. For example, Halpern
has 266 characters with eight strokes, Nelson has 381, and S&H has 426.
Traditionally, kanji have been indexed via their radical and stroke
count. Halpern's primary index is his SKIP method, which I
find extremely useful (see further comments on the radical
Looking up words
S&H lists each word in the entry for each component character. For example,
if faced with 世論 (seron or yoron),
even if you had trouble finding 世 (S&H lists it in the five-stroke
``no-radical'' group), you could still find the word by looking up
論, which should be quite easy to find because the radical is obviously
With Nelson (which I don't use, so maybe I'm wrong), I believe words are
only listed via their first component character (as is common with
Japanese-Kanji dictionaries). The traditional radical for 世 is
一, but Nelson lists it under ｜,
neither of which I find apparent.
As I've noted, Halpern is not designed to be a dictionary of words,
but as it turns out, you can find 世論 listed under
世, 論, and
操, this last one because of the entry
世論を操る (seron wo ayaturu)
``manipulate public opinion.''
Other Lookup Methods and Features
S&H has only its primary radical indexing, and a reading index in the back.
It has a fair amount of internal cross references to ease common lookup
errors. For example, 辛, whose radical is
立, has a cross-reference under the
亠 radical since it would be a reasonable mistake
Other lookup features are
- The outside margins of the page show the radicals with a mark
indicating where you are at the moment. This is very useful for
quickly finding a radical section -- you just rifle through the
pages until the mark comes to the radical you want.
At the start of a new radical, a table lists all the characters
in the group (including cross references). Even though it adds an
extra step to go to the head of a radical's group to scan the
table for the character, I find it speeds things up for me.
One I locate the character in the table, I can then turn directly
to its entry.
As I've noted, I don't use Nelson nor have one handy for reference.
However, it appears the new Nelson has an
index which will let you find a character via any component radical, even
if it isn't the radical. This could be very, very cool.
Halpern's main indexing is his SKIP system, which is augmented
by an extensive set of cross references for common lookup mistakes and
Of course, there is the standard reading index, and in addition, Halpern
also provides a number of interesting index methods:
The new S&H has some of the above, and others, but inexplicably does not give
the characters' index into the main dictionary. So although you can see the
character in the list or grouping, you're no closer to actually finding the
character's entry in the main dictionary. Too bad.
- Via Radical --
For the traditionalist, there is a radical index giving the 214
traditional radicals and the characters that are
indexed by them.
- Radical --
For actually finding entries for the radicals themselves.
Halpern gives radicals a main entry, just as he does characters.
This affords him the ability to list detailed information about the
radical, such as stroke order, a prose description, etymology, etc.
The Radical Chart lists the 214 traditional radicals, with
variants, by stroke count. The radical name (with an English name as
well) are provided, as well as the radical's main character entry in
- Kanji Synonym Groups --
A powerful feature, sort of like a kanji thesaurus. Characters
(and their indices in the main dictionary) are presented in
groups of related meanings. For example, the grouping
``blue and purple colors'' shows
with core meanings of blue, deep blue, lapis lazuli
(bright blue), dark blue, indigo, and purple.
[Note: the English tags that this server's kanji lookup presents are from
kanjidic, not Halpern, so will not necessarily match exactly]
Another example is ``long time periods'', which lists also lists
six characters, with meanings such as dynastic period and era.
All in all there are hundreds of these thesaurus-like entries. I
don't know of any other work that even comes close to paralleling
- Frequency of Use --
The 2135 most-frequently used characters are listed in order, with
English core meanings and main dictionary index number.
- Jouyou Kanji --
The (as of March 15, 1989) 1945 characters designated for general use
by the Japanese Ministry of Education, arranged by the grade level that
the character is taught in Japanese public schools.
- Historical Tables, etc. --
Perhaps of marginal use, he lists information about certain groups of
characters, all with their index into the main dictionary.
- The Signs of the Oriental Zodiac
- The Five Elements
- The Ten Calendar Signs
- The Pentatonic Scale
- Abbreviation of Place Names
I bought S&H when it first came out (1989), and have used it a lot since. I
also bought Halpern when it first came out (1990). I liked it so
much, I spent three months of my life entering a lot of the information
into kanjidic. When copyright worries came up, I tracked down
Jack Halpern, to ask about it. He told me I was an idiot for
spending so much time -- he would have given me a disk (which he did).
Since then I have become friends with Jack, and collaborate with him on
some of his projects. Because our meeting was a result of my enthusiastic
recommendation of his book, I don't feel there's any ``conflict of
interest'' if I continue to enthusiastically recommend his book. Still, I
mention our relation here because it seems the thing to do.
Comments appreciated [Return to Description of Edict]
[Return to Main Page]
[Jump to Index]
(this page's master source last modified 9 years, 10 months ago)
This reply to request 120,665,938 made just for you Fri July 28th 2017 7:49pm JST [load currently averaging 38400 requests/day over a 117-second sample]