Extended Katakana for Loanwords (ファ, ヴィ, ティ, トゥ, and the Full Small-Vowel System)
Extended katakana for loanwords refers to digraphs that pair a base katakana with a small vowel kana (ァ, ィ, ゥ, ェ, ォ) or a small ュ. They write sounds the 46-kana gojūon cannot reach.12 Without them, modern Japanese text with a foreign name, food, technology term, or brand would lose much of its phonetic detail.
Overview
What "extended katakana" means
An extended-katakana digraph is a base katakana followed by a small vowel kana. It is read as one mora, taking the base kana's onset consonant and the small kana's vowel.123 The technical name for these small kana is 捨て仮名 (sutegana). The same typographic family includes the small ャ, ュ, ョ that build yōon, but the function is different.2
The set is almost exclusive to katakana. Hiragana shares the small-vowel glyphs (ぁ, ぃ, ぅ, ぇ, ぉ), but it does not use them productively for loanword digraphs. The target sounds do not occur in the native vocabulary hiragana is designed to write.14
A small ャ/ュ/ョ on an i-row kana (キ + ャ → キャ) glides the consonant. A small ァ/ィ/ゥ/ェ/ォ on a base kana (フ + ァ → ファ) replaces the vowel. They belong to the same typographic family, but they work in two distinct ways.2
Why standard katakana cannot cover loanwords
The 46-kana gojūon, with its dakuon and yōon extensions, covers the native Japanese phonemic inventory. But it cannot represent /f/ outside フ, /v/, /ti/, /tu/, /di/, /du/, or the /wi/, /we/, /wo/ that hiragana lost in the 1946 spelling reform.143 Vance (2008) lists /ɸ/ and /ts/ as phonemes that occur "only marginally, in loanwords". Extended-katakana digraphs are the spelling counterpart to that phonemic expansion.5
The 1991 Cabinet Notification on the writing of loanwords formalises the workaround. A base kana plus a small vowel reads as one mora whose vowel is the small kana's, so フ + ォ → フォ is read /fo/.673
Register and audience
The 1991 Cabinet Notification 外来語の表記 (Naikaku-kokuji No. 2, June 28 1991) is the prescriptive standard for writing loanwords in modern Japanese. It is also the source most learner materials cite, directly or indirectly, when they show the extended chart.687 It frames itself as a よりどころ (reference) for 「一般の社会生活において現代の国語を書き表すための外来語の表記」, the writing of loanwords in general social use.87
The Notification is widely characterised as a 目安 (guideline) rather than a 強制 (binding rule). Specialist fields and long-established 慣用 (conventional) spellings are tolerated.7 It superseded the 1954 国語審議会 試案, which had served as the de facto reference until the 1986–1991 deliberations produced the current text.7
The small-vowel mechanic
The five small vowels and one small ュ
The productive small-kana inventory for extended-katakana digraphs is ァ, ィ, ゥ, ェ, ォ, plus the small ュ used in テュ, デュ, フュ, ヴュ.672 These are the same glyphs used in the yōon set and the historical small ヮ, visually reduced to about half height. They belong to the larger 捨て仮名 (sutegana, small kana) family.2
How a base kana plus a small vowel forms one mora
A digraph counts as one mora, exactly like a yōon combination. フォ in フォト ("photo") is two morae (フォ + ト), not three. ティ in ティー ("tea") is two morae (ティ + ー).23
The pronunciation rule is: read the base kana's onset consonant, then the small vowel's vowel. フ (/ɸu/) + ァ → ファ (/ɸa/); テ (/te/) + ィ → ティ (/ti/); ト (/to/) + ゥ → トゥ (/tu/).93
Kubozono (2015) describes the same mechanism in prescriptive terms. He notes that the post-1991 guidelines recommend innovations in the katakana syllabary that "assist speakers in approximating phonemes not occurring in the Japanese language, /f/ and /v/ for example."3
写真のフォトをアップロードします。9
"I'll upload the photos."
パーティーは明日です。9
"The party is tomorrow."
Count one mora per digraph, plus one for each chōonpu ー. パーティー is four morae: パ + ー + ティ + ー. This matters for pitch accent, song lyrics, and 5-7-5 metric forms.23
Why this trick works in katakana but not in hiragana
Katakana is used for foreign-word transcription, onomatopoeia, and emphasis. Extended digraphs target phonemes that arise almost exclusively in those domains.14 Hiragana writes native vocabulary and grammar, neither of which contains /f-/ outside /ɸu/, /v/, /ti/, /tu/, /di/, or /du/.14
There is no productive use case in native vocabulary, so the small-vowel mechanic was never extended to hiragana, even though the small kana glyphs exist there. The single exception is the rarely used hiragana ゔ (vu). It is encoded in JIS X 0213 so that IMEs can render a hiragana counterpart to ヴ, but it is essentially unused in print.10
The full extended-katakana chart
The 1991 Cabinet Notification organises the extended kana into two tables.687 第1表 (Table 1) lists the 五十音 plus 濁音 plus conventional extensions already in wide use. These are frequent enough that the Notification treats them as the default for general writing. 第2表 (Table 2) lists optional extensions used 「原音や原つづりになるべく近く書き表そうとする場合」, when one wishes to write closer to the source-language sound or spelling.7
Table 1 conventional extensions: シェ, ジェ, チェ, ツァ, ツェ, ツォ, ティ, ディ, ファ, フィ, フェ, フォ, デュ.7
Table 2 optional extensions: イェ, ウィ, ウェ, ウォ, クァ, クィ, クェ, クォ, ツィ, トゥ, グァ, ドゥ, ヴ, ヴァ, ヴィ, ヴェ, ヴォ, テュ, フュ, ヴュ.7
The Notification also gives a substitution rule: Table 2 kana may be rewritten with Table 1 kana. トゥ may be written as ツ or ト, ヴァ may be written as バ, and so on.67
The f-row: ファ, フィ, フェ, フォ
The f-row is formed from フ plus a small ァ, ィ, ェ, or ォ, and sits in Table 1 as conventional.7 /f/ targets are common in English-origin loanwords: ファミリー (family), フィルム (film), フェリー (ferry), フォト (photo).9 Phonetically, Japanese フ is [ɸu], a voiceless bilabial fricative before /u/. The f-row digraphs spread that consonant onto other vowels, giving [ɸa], [ɸi], [ɸe], [ɸo].5
今日はファミリーレストランで夕食を食べます。9
"Today we'll have dinner at a family restaurant."
このフィルムは古いです。9
"This film is old."
The t-row gaps: ティ, トゥ
ティ (テ + ィ) and トゥ (ト + ゥ) fill the slots the native t-row leaves vacant. テ /te/ alternates with /tʃi/ (チ) for /ti/, and ト /to/ alternates with /tsu/ (ツ) for /tu/.793 ティ is in Table 1 as a high-frequency conventional extension. トゥ is in Table 2, where the older convention ツー or トウー coexists.67
パーティーで友達と話しました。9
"I talked with my friend at the party."
ティラミスはイタリア語で「私を元気づけて」という意味です。11
"Tiramisù means 'cheer me up' in Italian."
The d-row gaps: ディ, ドゥ
ディ (デ + ィ) sits in Table 1; ドゥ (ド + ゥ) is Table 2.7 These mirror the t-row pattern for voiced /d/. Examples include ディズニー (Disney), ディナー (dinner), the older アイドル (idol, with traditional ド), and ヒンドゥー (Hindu, with Table 2 ドゥ).9 Under the Notification, ドゥ may be rewritten as ド where 慣用 prefers it.7
The w-row revivals: ウィ, ウェ, ウォ
ウィ (ウ + ィ), ウェ (ウ + ェ), and ウォ (ウ + ォ) are Table 2 entries. They take over the function the obsolete ヰ, ヱ, and ヲ used to serve for foreign /wi/, /we/, and /wo/.712 The historical ヰ and ヱ were removed from general use in the 1946 spelling reform. ヲ survives as a particle marker but is not productive for foreign sounds.12
Common examples are ウィキペディア (Wikipedia), ウェブ (web), and ウォーター (water). Stylised retention of ヰ in proper nouns survives in trademarks (the older ウヰスキー for "whisky") but is otherwise rare.12
Affricate and palatal gaps: ジェ, シェ, チェ
ジェ (ジ + ェ), シェ (シ + ェ), and チェ (チ + ェ) are all Table 1 conventional extensions.7 They cover /dʒe/, /ʃe/, and /tʃe/, which the native j-, sh-, and ch- rows lack. Japanese natively has only ジャ/ジュ/ジョ, シャ/シュ/ショ, and チャ/チュ/チョ for those consonants.9
Frequent examples include ジェット (jet), シェフ (chef), and チェック (check).9
The ヴ-row: ヴ, ヴァ, ヴィ, ヴェ, ヴォ
ヴ is the base ウ with dakuten, used to target /v/. The character is attributed to 福澤諭吉, who proposed ウに濁点 in 『増訂華英通語』 (1860) as 「試にウワの假名に濁點を附けてヴヷと記したるは當時思付の新案」.10 The five-cell row (ヴ, ヴァ, ヴィ, ヴェ, ヴォ) sits in Table 2 of the 1991 Notification. It appears under the heading 「原音や原つづりになるべく近く書き表そうとする場合に用いる仮名」, kana used when writing as close as possible to the original sound or spelling.7
The phoneme /v/ is not part of native Japanese phonology. Even where ヴ-row spellings appear in writing, speakers typically realise the consonant as [b]. Some use the bilabial approximant [β], but not the labiodental [v].1045 The hiragana counterpart ゔ exists in JIS X 0213 but has essentially no productive use in print.10
バイオリンを習っています。104
"I'm learning the violin." (same meaning, ハ-row spelling)
Lower-frequency combinations across both tables: イェ, クァ, グァ, ツァ, ツィ, ツェ, ツォ, テュ, デュ, フュ, ヴュ
These digraphs sit across both Tables 1 and 2 of the 1991 Notification. They are generally lower-frequency than the f-row, t-row, d-row, w-row, and ヴ-row combinations covered above. ツァ, ツェ, ツォ, and デュ are Table 1 entries, frequent enough to count as conventional. ツィ, テュ, フュ, ヴュ, クァ, クィ, クェ, クォ, グァ, and イェ are Table 2 entries.7 All of them cluster in specialist transliteration: place names, personal names, classical-music vocabulary, and culinary terms.49
Examples include ヴェネツィア (Venice, written with ツィ; also appears as ベネチア or ヴェネチア), ツェッペリン (Zeppelin), デュッセルドルフ (Düsseldorf), クァルテット (quartet, alongside カルテット), and フュージョン (fusion).91314 The Notification provides Table 1 fallbacks: クァ → クア, グァ → グア, ヴ → ブ or ハ-row, ツィ → チ, テュ → チュ, フュ → ヒュ, and so on.7
Reading the chart at a glance
The combinatorial rule is the same across every row. A base kana column plus a small-vowel row produces a digraph read as the base kana's onset consonant plus the small vowel's vowel. It counts as one mora.6723 If you see an unfamiliar digraph, such as クォ, you can decompose it (ク = /ku/, ォ = /o/) and pronounce it /kwo/ or, where 慣用 prefers, /ko/.9
When each extension is used
Anglicised vocabulary (most common path)
English-origin loanwords drive most of the f-row, the t-row gaps, the d-row gaps, and the w-row revivals. /f/ targets pick up the ファ-row in ファイル (file), ファン (fan), フィット (fit), フェア (fair), and フォーマット (format).49 /ti/ and /tu/ take ティ and トゥ in ティーチャー (teacher), ティーン (teen), and トゥナイト (tonight). Older convention used チ for /ti/, surviving in チーム (team), and ツ for /tu/, surviving in ツアー (tour).79
/di/ and /du/ take ディ and ドゥ in ディスク (disc) and ディフェンス (defense). ヒンドゥー coexists with the older ヒンズー.79 /wi/, /we/, and /wo/ take ウィ, ウェ, and ウォ in ウィキペディア, ウェブサイト, and ウォーキング, with Table 1 ウオーキング as the older alternate.712
新しいファイルをディスクに保存します。9
"I'll save the new file to the disc."
Place names and personal names
Table 2 extensions cluster in proper nouns because faithful transcription of foreign names matters more than ease of Japanese pronunciation.74 ヴェネツィア (Italian Venezia), デュッセルドルフ (Düsseldorf), ツェッペリン (Zeppelin), and モーツァルト (Mozart) are typical examples.
Government style diverged from this convention in 2019. The 在外公館名称位置給与法 was amended that March to remove ヴ from Japan's official country names, taking effect April 1 2019. カーボヴェルデ became カーボベルデ, and セントクリストファー・ネーヴィス became セントクリストファー・ネービス. These were the last two country names in Japanese statutory text that had used ヴ.15161718
Technical, culinary, and musical vocabulary
Specialist domains preserve original-language phonology more aggressively, drawing on Table 2 extensions.49 Culinary writing gives シェフ (chef, Table 1 シェ)9 and ティラミス (tiramisù, Table 1 ティ).11 Musical writing gives ヴィオラ (viola, alongside ビオラ),10 クァルテット (quartet, alongside カルテット),1314 and フュージョン (fusion).9
When the older non-extended form persists
The 1991 Notification explicitly preserves 慣用 spellings even where Table 2 would give a closer transcription.7 バイオリン coexists with ヴァイオリン. ベートーベン coexists with ベートーヴェン. ハロウィン and ハロウィーン alternate without a clear winner.710
The same word may circulate in two spellings indefinitely. The Japanese Wikipedia entry on ヴ notes that バイオリン and ヴァイオリン continue to be used in parallel.10 The Notification cites ウオーキング / ウォーキング, バイオリン / ヴァイオリン, and ヒュージョン / フュージョン as canonical Table 1 vs. Table 2 alternate pairs.7
The ヴァイオリン vs. バイオリン question
Why both spellings circulate
The ヴ-row is prescriptively permitted by Table 2 of the 1991 Notification for closer-to-source transcription of /v/. But the underlying phoneme /v/ does not exist in spoken Japanese. Speakers typically realise both spellings with a [b]-initial.1045 The spelling distinction therefore has no consistent phonetic correlate in everyday speech. It is a visual and stylistic choice, not a pronunciation one.105
Both spellings appear in modern published text, and dictionary entries treat them as variants of the same word.1014
What official sources do
文化庁 (Bunkachō) has not replaced or rescinded the 1991 Notification; the ヴ-row remains a permitted optional transcription under Table 2.687
外務省 (Ministry of Foreign Affairs), by contrast, removed ヴ from country names in 2019. The 在外公館名称位置給与法 amendment passed the 衆議院 on March 19 2019 and the 参議院 on March 29 2019, taking effect April 1 2019.151617 After the change, カーボヴェルデ became カーボベルデ, and セントクリストファー・ネーヴィス became セントクリストファー・ネービス. These were the last two country names in Japanese statutory text using ヴ.15161718
Then-Foreign Minister 河野太郎 characterised the change as making country names 「なるべくわかりやすい」, as easy to understand as possible.16 The two policies are not in conflict. MOFA scoped its change to official country names in statutory text, not to general writing or the Notification's chart.15
Practical rule of thumb
For reading, accept either spelling as the same word. Pronunciation is [b]-initial regardless.105 For writing general loanwords, the ハ-row (バ, ビ, ブ, ベ, ボ) is the modern default. The ヴ-row is reserved for stylised contexts and proper-noun precision where the original /v/ matters.71015 For official text, government and broadcast style prefer the ハ-row for country names after 2019.1516
The 1991 Cabinet notation in one page
What the notation is
The Notification is titled 外来語の表記 (Notation of Loanwords). It was issued by the Cabinet of Japan as 内閣告示第二号 (Cabinet Notification No. 2) on June 28 1991, based on the 国語審議会 answer of February 7 1991.687 Its stated scope is 「一般の社会生活において現代の国語を書き表すための外来語の表記」. Its stated status is a よりどころ (reference, guideline), widely characterised as a 目安 rather than a 強制.87
Table 1 vs. Table 2
Table 1 (第1表) contains the 五十音 plus 濁音 plus the conventional extensions シェ, ジェ, チェ, ツァ, ツェ, ツォ, ティ, ディ, デュ, ファ, フィ, フェ, and フォ. It also includes 拗音, 撥音 ン, 促音 ッ, and 長音符 ー.7 Table 2 (第2表) lists the optional extensions used for closer-to-source writing: イェ, ウィ, ウェ, ウォ, クァ, クィ, クェ, クォ, ツィ, トゥ, グァ, ドゥ, ヴ, ヴァ, ヴィ, ヴェ, ヴォ, テュ, フュ, and ヴュ.7
Together with the gojūon, dakuon, and yōon base, Tables 1 and 2 produce the corpus of approximately 113 katakana digraphs that Kubozono (2015) cites as the post-1991 recommended inventory.3
| Table | Status | Examples |
|---|---|---|
| 第1表 | Conventional, default for general writing | ファ, フィ, フェ, フォ, ティ, ディ, ツァ, ツェ, ツォ, デュ, シェ, ジェ, チェ |
| 第2表 | Optional, closer-to-source transcription | ヴ-row, ウィ, ウェ, ウォ, クァ-row, グァ, トゥ, ドゥ, ツィ, テュ, フュ, ヴュ, イェ |
What it does and does not legislate
The Notification sets a 仮名 inventory and spelling guideline for writing loanwords, not a pronunciation standard.87 It explicitly preserves long-established 慣用 spellings such as バイオリン and ベートーベン where they have crystallised.7
It permits substitution. Table 2 kana may be rewritten with Table 1 equivalents (トゥ as ツ or ト, ヴァ as バ, and so on), so writers may downgrade to Table 1 for simpler audiences.67 It does not legislate proper-noun transliteration rules in detail. Those live in separate style guides, including NHK, MOFA, and individual publishers.715
Good to know
Small vowels are not yōon
Yōon uses small ャ, ュ, ョ on an i-row kana to palatalise the consonant, as in キャ /kya/.2 Extended katakana uses small ァ, ィ, ゥ, ェ, ォ on various base kana to retarget the vowel, as in フィ /fi/. The two mechanics share the same typographic family of small kana, but they produce different phonological results. A learner reading フィ as a palatalised /fyi/ on analogy with キャ is misreading the block.2
フィルムを買いました。9
"I bought a film." (フィ is /fi/, one mora, not a palatalised /fyi/.)
The combinations are not in the gojūon chart
Learners often hunt for ファ on the 46-kana table and conclude they have missed a row. The gojūon table is finished at 46; ファ is a digraph built on top of the table, not a base kana within it.71
Input-method tips
Most IMEs (Microsoft IME, Google Japanese Input, macOS Japanese input) accept direct romaji for the conventional extensions: fa → ファ, fi → フィ, fe → フェ, fo → フォ, ti → ティ, tu → トゥ, di → ディ, du → ドゥ, vu → ヴ, va → ヴァ, vi → ヴィ, ve → ヴェ, vo → ヴォ. If an IME does not auto-resolve a particular digraph, a leading x or l types the small kana directly (xa → ァ, xi → ィ, and so on). You can then assemble any digraph by typing the base kana and then the small vowel.
ヴ-row in plain prose for non-musical, non-foreign-name vocabulary
Outside proper nouns, classical-music vocabulary, and stylised writing, the ヴ-row reads as overly precise or affected. The ハ-row is the unmarked modern default. Government style explicitly prefers the ハ-row for country names after the April 2019 in-force date of the amended 在外公館名称位置給与法.1516
The ヴ character is a 19th-century invention
福澤諭吉 proposed ウに濁点 in 『増訂華英通語』 (1860) specifically to transcribe the foreign /v/ sound. The move is documented as 「試にウワの假名に濁點を附けてヴヷと記したるは當時思付の新案」.10 The character is therefore younger than the rest of the katakana inventory by roughly a millennium, and it was prescriptively endorsed only by the 1991 Cabinet Notification.10
Treating an extended digraph as two morae
A frequent counting mistake is reading フォト as three morae (フ + ォ + ト). The correct count is two (フォ + ト). A base kana plus a small vowel occupies one mora, exactly like a yōon combination.23 This matters for pitch-accent placement, song lyrics, and 5-7-5 metric forms.
写真を撮るのが趣味です。フォトが好きです。9
"Taking pictures is my hobby. I like photos." (フォト is two morae, not three.)
A historical accident, not a script-design choice
The asymmetry between hiragana and katakana extension capacity is not a deliberate design decision. It is a downstream consequence of how the two scripts came to specialise. Katakana absorbed the loanword job, so the small-vowel mechanic accreted there over centuries of ad hoc transcription before the 1991 Notification ratified it.71 Hiragana never developed an equivalent inventory of extensions because nothing in its domain pulled for one. ゔ (hiragana vu) sits in JIS X 0213 mainly to round out the encoding rather than to serve real text.102 The earlier section on why this trick works in katakana but not in hiragana covers the mechanics in detail.
See also
- What Is Gairaigo? A Guide to Loanwords in Japanese
- The Japanese Consonant Inventory: Phonemes, Allophones, and the Kana Chart
- The Japanese Vowel Inventory: Five Vowels, Done Right
- How to Pronounce つ (tsu) in Japanese: The Voiceless Alveolar Affricate English Lacks
- Hiragana vs. Katakana: How to Tell Them Apart and Use Both
- The History of Katakana: From Heian Monks' Shorthand to the Modern 46-Character Set