Ateji (当て字): Kanji Chosen for Sound, Not Meaning
Ateji (当て字) are kanji assigned to a word for their sound rather than their meaning. This spelling choice explains why 寿司 (sushi) and 珈琲 (coffee) use the kanji they do.1 Learners usually meet the term after on'yomi and kun'yomi are stable and irregular loanword spellings start to demand a name.
Overview
What ateji means
The term 当て字 reads literally as "applied" or "assigned characters," from the verb 当てる ("to apply, assign"); ateji is the noun for "characters one applies" to a word.2
English Wikipedia defines ateji as kanji "used to phonetically represent native or borrowed words with less regard to the underlying meaning of the characters."1 Japanese Wikipedia frames the same operation as 「字の本来の用法を無視して、当座の用のために異なる語の表記に転用した漢字などの文字」, characters repurposed ad hoc to write a different word while ignoring the character's original function.3
Inside that single term, Japanese Wikipedia identifies two senses. 借音 (shakuon, "sound-borrowing") ignores the character's meaning and assigns kanji on sound alone. 借訓 (shakkun, "meaning-borrowing") ignores the character's reading and assigns kanji on meaning alone.3 The sound-borrowing sense is the primary one and the focus of this article.
Modern dictionaries list the term as 当て字. The forms 宛字 and 充て字 are minority variants of the same word, not different concepts.32
Where ateji sits among kanji-reading phenomena
Ateji operates on orthography: which kanji you write. Jukujikun operates on reading: what sound the compound takes. The two phenomena can co-occur in a single compound. The JMdict data model tags them in separate fields: ke_inf on the kanji element for ateji, re_inf on the reading element for gikun.45
English Wikipedia uses a broad-vs-narrow framing. In that view, ateji is the umbrella for irregular kanji-to-reading assignments, and jukujikun is the meaning-driven sub-case; the sound-driven case is called "phonetic ateji."1 Japanese Wikipedia takes the adjacent-categories view, listing 熟字訓 in 当て字's 関連項目 (related items) section.3 The two taxonomies disagree on hierarchy, not on facts. A learner can pick either framing and arrive at the same per-word classification.
The 文化庁 (Agency for Cultural Affairs) 答申 (official report) for the 常用漢字表 付表 introduces both phenomena together as 「いわゆる当て字や熟字訓など、一字一字の音訓として挙げにくい語句」, "so-called ateji and jukujikun, and similar word groups difficult to list as the on or kun reading of a single character."6
How sound-driven ateji works
寿司 (sushi): pure sound, auspicious gloss
The word sushi derives from 酸し (sushi, "sour"), the archaic shi-terminal form of modern 酸い (sui, "sour"). Its literal sense is "vinegared rice."78 The word existed first, in kana, with no fixed kanji.
The 寿司 spelling is a sound assignment added later. 寿 supplies su (its base meaning is "one's natural life span; longevity"). 司 supplies shi (its base meaning is "to administer, preside over").17 Neither character's meaning relates to the food.
Wiktionary classifies the spelling explicitly: "The kanji are an example of ateji (当て字)."7 sljfaq.org concurs, noting that 寿司, 鮨, and 鮓 are all "characters given to a word after it was formed."8 Jisho.org surfaces the JMdict flag on the form 寿司 itself as "Ateji (phonetic) reading"; the alternative kanji 鮨 and 鮓 are unflagged.9
The auspicious "preside over longevity" gloss is sometimes presented as the reason those kanji were chosen. It is a post-hoc rationalization of an already-existing pairing. The etymology runs sound to kanji, not meaning to word.78
今夜、寿司を食べに行きませんか。10
"Do you want to go for sushi tonight?"
あなたは寿司と天ぷら、どっちが好き。10
"Which do you like better, sushi or tempura?"
珈琲 (kōhī, coffee): borrowing the Western word into kanji
Wiktionary classifies the spelling as ateji and credits the coinage to 宇田川榕菴 (Udagawa Yōan), an Edo-period Rangaku (Dutch-studies) scholar working in the first half of the 1800s.1112 The earliest known appearance is in his manuscript 蘭和語彙集 (Ranwa Goishū), but its precise date is unrecorded.12
The two kanji's base meanings have no relationship to coffee. 珈 names a kind of ornamental hairpin worn by women; 琲 names a string of pearls.12 Udagawa chose them for sound (kō, hī).
A popular gloss describes 珈 as a coffee branch bearing red berries (the 琲) and presents this as Udagawa's reason. Japanese Wikipedia attributes that reading not to Udagawa, but to 奥山儀八郎's 1940 publication. It is a 20th-century reinterpretation laid over a 19th-century sound-based assignment.12 The branch-and-berry story is folk etymology, not authorial intent.
The 19th-century spelling was unsettled before 珈琲 stabilized; documented variants include 可否, 架非, 加非, and 咖啡.12 The modern dictionary form is the katakana コーヒー. 珈琲 survives as a graphical alternative used on cafe signage and packaging for atmosphere.11
私が朝起きて最初にすることは、コーヒーを淹れることだ。13
"The first thing I do after waking up in the morning is brew coffee."
Udagawa Yōan's reason for choosing 珈 and 琲 was the sound, not a botanical picture. The branch-and-berry story is traceable to 奥山儀八郎's 1940 publication, almost a century after the spelling was coined; popular sources sometimes present it as the original motive.12
倶楽部 (kurabu, club): phono-semantic matching
Wiktionary classifies the spelling as "ateji (当て字), apparently from the early Meiji era."14 Each kanji supplies sound and meaning at once: 倶 (ku, kun "together"), 楽 (ra, kun "to enjoy"), 部 (bu, "part / division," carrying "a sense of 'smaller group of people'").14
This is the canonical phono-semantic match, where sound and meaning both fit. The sound ku-ra-bu reproduces English club while the assembled meanings ("together-enjoy-place") describe what a club is. Most surviving ateji do not achieve this dual fit.114
The modern dictionary form is the katakana クラブ. The kanji form is the historical / formal variant, and Jisho.org tags it as "Ateji (phonetic) reading," JLPT N1.1415
私は音楽クラブに入っています。10
"I am in the music club."
Country names: 亜米利加, 仏蘭西, 独逸
Wiktionary's 亜米利加 entry tags all four kanji as ateji: 亜 (a), 米 (me), 利 (ri), 加 (ka), assembling Amerika across four characters.16 English Wikipedia lists 亜米利加 among its ateji examples. The parallel 仏蘭西 (France) appears in the Man'yōgana article's catalog of ateji-style sound-borrowed renderings.117 The other pre-katakana national-name spellings (独逸 for Germany, 英吉利 for England, 露西亜 for Russia) follow the same construction.
These long forms are obsolete in modern Japanese. Wiktionary labels 亜米利加 itself as obsolete, and the live forms are the abbreviation 米国 plus the katakana アメリカ.1618
The abbreviations survive. Single-kanji forms recur in newspaper headlines and compounds: 米 for America, alongside 仏 (France), 独 (Germany), 英 (England), 露 (Russia), and 伊 (Italy), following the same pattern. Their kanji meanings ("rice," "Buddha," "alone," "outstanding") carry no semantic content here. The character was lifted from the ateji string for its position, not its sense.
The 米 abbreviation is the worked case in the sources. Wiktionary's 米国 entry describes the kanji as a shortening of "the phonetic kanji spelling 亜米利加 (Amerika, 'America')." It also notes an alternative source candidate, 米利堅 (Meriken), used from the 1854 treaty era through the early 1900s. In either derivation, the 米 character is doing phonetic, not semantic, work.18 The other single-kanji abbreviations follow the same pattern from their respective ateji strings, though primary sources enumerating each derivation are thinner.
米国連邦準備銀行は金利を引き下げました。13
"The Federal Reserve lowered interest rates."
日米関係について話し合う。10
"(They) discuss Japan-US relations."
The Meiji translation wave
Before katakana monopolized loanwords
Pre-Meiji and Meiji-era writers had no convention forcing foreign loanwords into katakana. A natural move was to assign kanji to the foreign sound and sometimes layer meaning on top.
Japanese Wikipedia describes the surge: 「明治から始まる言文一致と文明開化による西洋文化の流入によって、欧文音写したカタカナにその意味から漢字を当て字する例も増えた」, "with the genbun itchi movement (unification of written and spoken style) and the influx of Western culture during the bunmei kaika (civilization and enlightenment) of Meiji, examples of applying ateji to katakana transcriptions of European words on the basis of meaning also increased."3
Most surviving ateji entered the language during this window. 珈琲 dates to Udagawa Yōan in the first half of the 1800s.1112 倶楽部 dates to the early Meiji era.14
浪漫 is popularly attributed to Natsume Sōseki, though Wiktionary's own entry flags the attribution with a "can this etymology be sourced?" caveat; "popularly attributed" is the honest framing.19
The 麦酒 spelling pre-dates the Meiji wave (documented from 1724), but gained breadth as Western beer arrived in volume in the late 19th century. The live form today is the katakana ビール, borrowed from Dutch bier.20
Surviving Meiji ateji in everyday Japanese
A handful of Meiji-era assignments entered the standard vocabulary. They are read on sight today, with their etymology no longer surfacing for most readers.
浪漫 (rōman, "romance") is marked ateji on both kanji in Wiktionary, with a "phono-semantic matching of French roman" note. The per-character meanings ("wave; unrestrained" and "free; inundate") gloss the romantic-novel sense.19
麦酒 (bīru, "beer") sits on the meaning-driven branch: 麦 ("barley") + 酒 ("alcohol") assembles to "barley liquor," and the pronunciation bīru comes from Dutch bier.20 煙草 (tabako, "tobacco") works the same way: 煙 ("smoke") + 草 ("grass / plant") assembles to "smoking plant," with tabako from Portuguese tabaco.21 Both compounds are sometimes classified as jukujikun rather than ateji, depending on the taxonomy used. English Wikipedia files them under ateji "with meaning only," and the 常用漢字表 付表 lists 煙草 / タバコ among the compounds difficult to assign per-character readings.1212223
The pattern is consistent. Pure sound-ateji clusters around proper nouns (country names) and food / beverage names where no Sino-Japanese equivalent existed. Meaning-driven ateji clusters around concepts that already had a native or Chinese-style word the translators could lean on.13
The 20th-century handoff to katakana
As the katakana-for-gairaigo convention solidified across the 20th century, new loanwords stopped getting kanji assignments. The ateji inventory froze at its existing entries while katakana became the productive loanword channel.24
The surviving status of 寿司, 珈琲, 倶楽部, and 麦酒 is therefore not productivity but legacy. The words entered the standard vocabulary before the handoff. New gairaigo (loanwords) like パソコン (pasokon), スマホ (sumaho), and アプリ (apuri) take katakana by default.124
Modern ateji: where you still meet it
Restaurant and shop signage
English Wikipedia notes that ateji "appear on traditional store signs and menus."1 寿司 on shop curtains and 珈琲 on cafe awnings carry a retro register that the katakana spelling would not. The kanji form is doing atmospheric work.1119
The original character meanings, even when historically post-hoc, contribute to the atmosphere. The hairpin-and-pearls reading of 珈琲 may not be Udagawa's reason, but it can still suggest an atmosphere when a cafe puts those kanji on its sign.712
Manga, lyrics, and creative orthography
English Wikipedia identifies the productive corner of modern ateji: a practice "frequently employed in manga and song lyrics by pairing kanji with furigana for creative effect and to add layers of meaning."1
An author writes a kanji compound that carries the meaning they want fixed on the page, then prints furigana giving a different reading, often a katakana loanword. The reader sees both layers at once: the orthography fixes the meaning, while the furigana fixes the sound.1 In dictionary terms, this overlaps the JMdict gikun tag rather than the ateji tag, because the reading element carries the irregularity rather than the kanji element.4
Names and brand orthography
Personal-name kanji choice uses the same mechanism: a sound (the name) and a kanji string assigned to it, typically with an auspicious-meaning gloss layered on the chosen characters. Japanese Wikipedia treats 人名用漢字 (name-use kanji) selection as an adjacent application of the same orthographic principle, though the technical literature distinguishes nanori-readings as their own category.3
This branch is the live, productive use of ateji in 21st-century Japanese. The loanword-naming branch is largely historical.13
The boundary with jukujikun
Which dimension is irregular?
Ateji and jukujikun describe irregularities on different axes. Ateji says the kanji choice is irregular: the kanji were picked on the basis of sound rather than meaning. But the reading attached to the compound is still composable from each character's on or kun reading (寿 su + 司 shi = sushi).17
Jukujikun says the reading is irregular: a native Japanese word has been assigned to a multi-character compound as a whole, with no per-character sound contribution (大 + 人 → otona, not derivable from any reading of either character).1 Both can occur in the same compound. JMdict's two-flag system was built to handle that case.45
How JMdict separates them
The JMdict DTD (Document Type Definition) defines two separate entities: <!ENTITY ateji "ateji (phonetic) reading"> and <!ENTITY gikun "gikun (meaning as reading) or jukujikun (special kanji reading)">.4 The two flags attach to two different XML elements, which is what makes them independently addressable.
The ateji tag lives in <ke_inf>, the kanji-element-information field. The DTD describes it as carrying "a coded information field related specifically to the orthography of the keb."4 The gikun tag lives in <re_inf>, the reading-element-information field, "general coded information pertaining to the specific reading."4
The EDRDG (Electronic Dictionary Research and Development Group) editorial policy makes the contrast explicit. The ateji tag "is used with kanji forms where one or more of the kanji are used to phonetically represent native or borrowed words with less regard to the underlying meaning of the characters." The gikun tag "is used with readings that are based on the meaning of the term and not the readings of the kanji form."5
Roughly 550 JMdict entries carry an ateji ke_inf flag on at least one kanji form. The canonical 出鱈目 (detarame, "nonsense") is an example, where the kanji are read de + tara + me.5 Typical gikun re_inf entries include 今日 / きょう, 田舎 / いなか, 明日 / あした, and 煙草 / タバコ.5
A single entry can carry both tags. When both dimensions diverge, the same compound is ateji on its kanji element and gikun on its reading element.45
| Phenomenon | What is irregular | JMdict field | JMdict tag |
|---|---|---|---|
| Ateji | The kanji choice (sound-driven) | <ke_inf> (kanji element) | ateji |
| Jukujikun / gikun | The reading (assigned to the whole compound) | <re_inf> (reading element) | gikun |
| Both | Kanji choice and reading both diverge | both fields | both tags |
The two competing taxonomies, reconciled
English Wikipedia treats jukujikun as a meaning-driven sub-case under the broader ateji label: "phonetic ateji" (寿司) versus "semantic ateji" (大人, more commonly called jukujikun).1 The 文化庁 答申 likewise groups them under one heading in the 付表 preamble.6 The 116-entry appendix mixes both phenomena rather than separating them into columns.2223
JMdict resolves the ambiguity tag by tag in separate fields. For a learner trying to read what a dictionary entry is telling them, the JMdict framing is the durable one because the tags survive in downstream tools. Jisho, Yomichan, and JapanDict all carry them through.45915
Japanese Wikipedia takes the adjacent-categories framing instead: 熟字訓 sits in 当て字's 関連項目 section rather than under it.3 The taxonomies disagree on hierarchy, not on facts. The per-word classification is the same on either framing.134
A two-question decision flow
Two questions resolve any candidate compound.
Q1 asks: "Are the kanji chosen for their sound rather than their meaning?" If yes, the compound carries an ateji ke_inf flag; the orthography is sound-driven.45
Q2 asks: "Is the reading assigned to the whole compound rather than derivable from each kanji's on or kun?" If yes, the compound carries a gikun re_inf flag; the reading is meaning-driven on the cluster.45
Both yes is possible: ateji on the kanji side and gikun on the reading side. Both no means the compound is a standard on'yomi or kun'yomi compound, so this diagnostic does not apply.
Worked cases:
- 寿司 → Q1 yes (寿 and 司 supply sounds), Q2 no (sushi is composable from those readings). Pure ateji.79
- 大人 → Q1 no (大 and 人 carry their normal meanings), Q2 yes (otona is not derivable from either character). Pure jukujikun.15
- 倶楽部 → Q1 yes (sound-driven), Q2 no (kurabu is composable from 倶 / 楽 / 部 readings). Ateji.1415
Good to know
The 当て字 / 宛字 / 充て字 orthographic triplet
All three are written forms of the same word. 当て字 is dominant in modern Japanese dictionaries and reference works. 宛字 and 充て字 survive as variants.32 Wiktionary additionally records 當て字 as the kyūjitai (pre-reform) form and 当字 as a simplified variant.2 The variation is orthographic, not semantic.
Do not reverse-engineer the meaning of an ateji compound
A common trap is to assume 寿司 must somehow mean "preside over longevity" because the kanji say so. The word means "vinegared rice / sushi." It existed first (from 酸し sushi, "sour"), and the kanji were assigned to it on the basis of sound.78 The character meanings on the page are post-hoc.
The right gloss for the existing word is the natural Japanese one:
今夜、寿司を食べに行きませんか。10
"Do you want to go for sushi tonight?"
Ateji etymology runs sound to kanji, not kanji to meaning.17
The 珈琲 hairpin-and-pearls story is not Udagawa's reason
A popular gloss says Udagawa Yōan chose 珈 (hairpin) and 琲 (string of pearls) to depict coffee branches bearing red berries. Japanese Wikipedia attributes that gloss to 奥山儀八郎's 1940 publication, almost a century after Udagawa's coinage.12
The original assignment was based on sound. The branch-and-berry story is a 20th-century reconstruction: useful as decoration, but not as authorial intent.
Phono-semantic matching is the elegant case, not the norm
倶楽部 is the textbook example because ku-ra-bu reproduces club and 倶 ("together") + 楽 ("enjoy") + 部 ("place / group") describes what a club is.14 Most surviving ateji are pure sound: 寿司, 珈琲, and 亜米利加 work as orthography only at the sound level, with any meaning gloss layered on afterwards.1716 For learning, expect sound-only fit by default. Phono-semantic matching is the trophy case, not the everyday inventory.
Man'yōgana is the historical ancestor
In the Man'yōshū era (8th century), all kana-equivalent writing was sound-borrowed kanji. Hiragana and katakana later developed as organic simplifications of man'yōgana. Ateji is the surviving fragment of the older convention that did not collapse into a kana script.17 English Wikipedia describes the link directly: ateji is "A phenomenon similar to man'yōgana... where words (including loanwords) are spelled out using kanji for their phonetic value," with 倶楽部 and 仏蘭西 cited as cases.17 A learner who sees ateji as weird isolated cases misses the historical depth. The cases are not weird; they are the conserved tail of a once-general convention.
Why dictionaries flag ateji
A learner using a Japanese-Japanese (J-J) or Japanese-English (J-E) dictionary sees small parenthetical "(ateji)" tags next to entries like 寿司 and 倶楽部.915 That is the JMdict ke_inf flag surfacing in the user interface.45
The flag signals "do not try to read the kanji individually." Reading 寿司 as kotobuki-tsukasa or 倶楽部 as gu-raku-bu by composing standard readings over-applies regular on'yomi composition. The dictionary's flag is telling the reader to look the cluster up as a unit.
See also
- What Is Gairaigo? A Guide to Loanwords in Japanese
- Pre-English Loanwords in Japanese: Portuguese, Dutch, German, and More
- Jinmeiyō Kanji (人名用漢字): The 863 Name-Use Characters Beyond Jōyō
- The Six Categories of Kanji (六書): Pictographs, Ideographs, and Phono-Semantic Compounds
- Jukugo (熟語): How Kanji Combine to Form Japanese Words
- Mixed Script: How Japanese Combines Kanji, Hiragana, and Katakana