Dakuten and Handakuten: The Voicing Marks on Hiragana
Dakuten and handakuten are the two small marks at the upper-right corner of a hiragana character. They change the consonant it spells: the two-dot dakuten (゛) voices a consonant (k → g, s → z, t → d, h → b), while the single-circle handakuten (゜) turns the h-row into the p-row.1 Adding these marks to the 46 base hiragana gives 25 additional kana. Together, they round out the standard chart you will see in every textbook.21
Overview
Dakuten and handakuten are the only diacritics in everyday hiragana. They are also the smallest change the script makes: the base character stays the same, and one small mark in a fixed position rewrites the consonant.2 Both marks work the same way in katakana, so the rule you learn here transfers directly to the second syllabary.1
What the marks are called
The two-dot mark is formally 濁点 (dakuten), literally "muddying mark," from 濁 ("muddy / cloudy") and 点 ("mark"); colloquially it is called 点々 (tenten, "dots").13 The small-circle mark is formally 半濁点 (handakuten), "half-muddying mark," and colloquially 丸 (maru, "circle").13
In Unicode, the character-encoding standard, the combining forms are U+3099 (◌゙ combining voiced sound mark) and U+309A (◌゚ combining semi-voiced sound mark). The spacing standalone shapes used in text are U+309B (゛) and U+309C (゜), with fullwidth katakana variants at U+FF9E and U+FF9F.1
Where they sit on the kana
Both marks are written at the upper-right of the base character: two short diagonal strokes for the dakuten and a small circle for the handakuten.41 The placement is the same in hiragana and katakana, and the same base character takes the same mark in both syllabaries (か → が in hiragana, カ → ガ in katakana).1
Adding a dakuten or handakuten does not change the base character's stroke count or stroke order; it is a separate mark added after the base is finished.2 The companion article Hiragana Stroke Order: Why It Matters Even If You Type covers the base stroke rules; this article only adds the diacritic.
Which rows they affect
The dakuten applies to the k, s, t, and h rows of the gojūon (the traditional 5×10 hiragana grid), plus a few marginal cases.1 The handakuten applies only to the h row, producing the p-row. It is the only diacritic in the script with a one-row scope.1
The 46-character base inventory (fixed by the 1900 kana reform), 20 dakuten kana (5 each from the k, s, t, and h rows), and 5 handakuten kana from the h row together give the 25 additional kana that complete the standard hiragana table alongside the base set.21 The full layout, including yōon (small-ゃ/ゅ/ょ) contractions, is collected in the companion article The Complete Hiragana Chart (Gojūon): How to Read All 46 Base Kana, Dakuten, and Yōon.
The voicing transformations
The five sections below walk through the chart row by row. Each one gives a small reference table you can return to. The worked examples after each table show the kana doing real work in everyday vocabulary.
k-row to g-row (か→が, き→ぎ, く→ぐ, け→げ, こ→ご)
The dakuten on the k-row voices the voiceless velar stop /k/ to the voiced velar stop /ɡ/.5
| Base | Voiced | Romaji |
|---|---|---|
| か | が | ka → ga |
| き | ぎ | ki → gi |
| く | ぐ | ku → gu |
| け | げ | ke → ge |
| こ | ご | ko → go |
In some varieties of Japanese, notably older Tokyo speech and NHK announcer training, word-medial が is realised as the nasal allophone [ŋ], called が行鼻濁音 (ga-gyō bidakuon). Pronunciations with [ŋ] are less frequent among younger speakers, so this is increasingly a marked, broadcast-register feature rather than a default modern realisation.56
学校。6
"School."
元気です。6
"Well; in good spirits."
午後。6
"Afternoon; p.m."
s-row to z-row (さ→ざ, し→じ, す→ず, せ→ぜ, そ→ぞ)
The dakuten on the s-row voices /s/ to /z/, with one apparent irregularity that comes from the base row itself.5
| Base | Voiced | Romaji |
|---|---|---|
| さ | ざ | sa → za |
| し | じ | shi → ji |
| す | ず | su → zu |
| せ | ぜ | se → ze |
| そ | ぞ | so → zo |
し is phonetically [ɕi] ("shi"), so its voiced counterpart じ is [(d)ʑi] ("ji"), not "zi." The voicing follows the actual base consonant, not the romanization column. し is consistently transcribed shi and じ as ji in Modified Hepburn.572
雑誌。6
"Magazine."
時間。6
"Time; an hour."
全部。6
"All; the whole lot."
t-row to d-row (た→だ, ち→ぢ, つ→づ, て→で, と→ど)
The dakuten on the t-row voices /t/ to /d/, with the same surface-form caveat as the s-row. ち is [tɕi] and つ is [tsɯ], so their dakuten forms ぢ and づ are phonetically identical to じ and ず respectively in modern Standard Japanese.5
| Base | Voiced | Romaji |
|---|---|---|
| た | だ | ta → da |
| ち | ぢ | chi → ji (merged with じ) |
| つ | づ | tsu → zu (merged with ず) |
| て | で | te → de |
| と | ど | to → do |
Because of that merger, modern kana usage (現代仮名遣い, Cabinet Notification No. 1, 1986) writes じ and ず by default for the "ji" and "zu" sounds. It keeps ぢ and づ only in two narrow contexts: same-morpheme repetition (同音の連呼), as in ちぢむ "to shrink" and つづく "to continue"; and transparent compounds whose second element undergoes rendaku from ち or つ (二語の連合), as in 鼻血 hanaji "nosebleed", 底力 sokojikara "reserve strength", and 三日月 mikazuki "crescent moon".8
The 1986 notification keeps ぢ and づ alive only where the writer or reader can still see the original ち or つ inside the word: in repetitions of the same morpheme (ちぢむ, つづく) and in compounds where the second element clearly came from a ち- or つ-word and got voiced. Everywhere else, write じ and ず.8
大学。6
"University."
電車。6
"Train."
鼻血。8
"Nosebleed."
h-row to b-row (は→ば, ひ→び, ふ→ぶ, へ→べ, ほ→ぼ)
The dakuten on the h-row yields the voiced bilabial stop /b/, articulated as a fully voiced [b] in modern Standard Japanese, not a /v/ or any fricative.59
| Base | Voiced | Romaji |
|---|---|---|
| は | ば | ha → ba |
| ひ | び | hi → bi |
| ふ | ぶ | fu → bu |
| へ | べ | he → be |
| ほ | ぼ | ho → bo |
This pairing looks arbitrary if you read the chart synchronically, as a snapshot of modern Japanese: why would "h" voice to "b"? Historically, though, it is regular. The modern h-row descends from a row that was once pronounced with a bilabial stop, and /b/ is the voiced counterpart of that older /p/. The next section explains the shift in detail.910
病院。6
"Hospital."
文房具。6
"Stationery."
勉強します。6
"I will study."
h-row to p-row with handakuten (は→ぱ, ひ→ぴ, ふ→ぷ, へ→ぺ, ほ→ぽ)
The handakuten is the only diacritic in the script that targets a single row. It produces /p/, the voiceless bilabial plosive [p], made by closing and releasing both lips without vocal-fold vibration.51
| Base | + handakuten | Romaji |
|---|---|---|
| は | ぱ | ha → pa |
| ひ | ぴ | hi → pi |
| ふ | ぷ | fu → pu |
| へ | ぺ | he → pe |
| ほ | ぽ | ho → po |
In native vocabulary, the bare p-row appears mostly in mimetic / sound-symbolic words and a handful of Sino-Japanese loans. In everyday Japanese, it is very common in foreign loanwords (ピアノ, ペン, パン).2
鉛筆。6
"Pencil."
散歩。6
"A walk; a stroll."
切符。6
"Ticket."
Edge cases (ゔ for v, ゐ゙ and ゑ゙)
ゔ is う + dakuten, romanised "vu." It is a modern addition used to represent /v/ in foreign words and is far more common in its katakana form ヴ. Even when it is written, Japanese speakers typically realise /v/ as [b], so Venus is "typically transliterated as ビーナス (bīnasu) instead of ヴィーナス".1
The historical kana ゐ (wi) and ゑ (we) were removed from standard use by the post-war kana reforms. In principle, they can take a dakuten to give ゐ゙ and ゑ゙, but these forms are confined to historical or stylised text and have no role in modern reading or writing.11 Recognise that ゔ exists, mainly so you can read loanword signage. You can ignore the obsolete dakuten forms.2
Why the h-row gets two marks
Every other affected row gets exactly one diacritic: k pairs with g, s with z, and t with d. The h-row is the exception. It produces both b (with dakuten) and p (with handakuten). The reason is historical, and once you see it, the apparent irregularity becomes the most regular pairing on the chart.
The h-row used to be a p-row
Comparative reconstruction from Ryukyuan languages and earlier Japanese sources gives proto-Japanese *p as the ancestor of the modern h-row. Old Japanese p reflected an earlier voiceless bilabial stop *p.10 The lenition path, a weakening of a consonant over time, is the standard textbook chain p → ɸ → h. Word-initial p had become a voiceless bilabial fricative [ɸ] by Early Modern Japanese. Modern Standard Japanese has /h/ word-initially, with [ɸ] surviving as the allophone before /u/ (which is why ふね fune is pronounced [ɸɯne]).105
Word-medially, the same consonant followed a different path. It became [w] in Early Middle Japanese and has since disappeared except before a, which is why は as the topic particle is read wa today.10
Voiced /b/ is the voiced counterpart of /p/. Both are bilabial plosives, differing only in voicing. When the h-row inherited /b/ as its dakuten partner, it inherited the voiced counterpart of the older /p/, not of the modern /h/. The h → b mapping is historically regular, even though it looks irregular on the modern chart.910
The handakuten was invented to recover p
After the lenition shift away from /p/, Japanese no longer had a clean way to write a "pa, pi, pu, pe, po" syllable using bare kana. The は-row characters had drifted to ɸ / h, and the dakuten on those characters produced /b/, not /p/.9
Sixteenth-century Jesuit missionaries needed to transcribe Portuguese and Latin words containing /p/ in their Japanese-language teaching materials. The Rakuyōshū (1598), printed by the Jesuit Mission Press, was the first movable-type dictionary to systematically use a small circle at the upper right of an h-row kana to mark the /p/ sound. The convention has survived to this day.4
The dakuten generalises a voicing alternation that has been productive in Japanese for centuries (k/g, s/z, t/d, h/b). The handakuten was added in the early modern period for one specific job: writing a /p/ sound that the h-row had lost through historical change. The marks share a chart cell, but their stories do not overlap.94
Why a circle and not a third dot
As a functional contrast, the small circle is visually distinct from the two-dot dakuten. It signals to the reader that a different kind of change is in play. It is a separate mark added after the base, not a variant of the dakuten.1 Modern Japanese speakers perceive both marks as part of one unified diacritic system on the kana chart. Still, their historical stories are different, as the previous section described, and the visual contrast preserves that difference at a glance.94
A brief history of when these marks became standard
Pre-modern usage was optional and contextual
Dakuten-like marks are attested from the late Heian period. They were adapted from the 声点 (shōten, "voice points") used to annotate pitch and other phonological features in Buddhist Sanskrit and Chinese-text recitation. The system itself drew on Chinese tone-mark conventions and the Siddhaṃ anusvāra.1
Through the classical and Edo periods, dakuten usage on kana was sporadic and reader-dependent. Voicing was often left unmarked and recovered from context, which is why classical poetry manuscripts and Edo-period woodblock prints look "naked" by today's standards.111 In pre-war kana usage, even on signage, the dakuten was frequently omitted, and readers were expected to supply the voicing themselves.11
Standardization came in the 20th century
Consistent marking of voiced sounds in official documents became normative only with the post-war kana reforms. The Cabinet promulgated 現代かなづかい (Modern Kana Usage) in 1946, replacing the historical 歴史的仮名遣い. This was superseded by the current 現代仮名遣い, promulgated as Cabinet Notification No. 1 in 1986, which is the standard in force today.118
A frequently cited example of the older convention is the Imperial Surrender Rescript of 1945, whose text was written without dakuten in the contemporary official style. The 1946 reform marks the boundary after which marked spellings became the default in government and education.11
The same reform tightened ぢ/づ vs. じ/ず spelling. ぢ and づ are retained only in same-morpheme repetition (ちぢむ, つづく) and in transparent rendaku compounds (鼻血, 三日月). Otherwise, the merged-pronunciation forms じ and ず are written.8
A first look at rendaku (sequential voicing)
What rendaku is in one sentence
Rendaku (連濁, "sequential voicing") is a pronunciation change seen in the middle of some Japanese compound words. It replaces a voiceless consonant such as /k s t h/ with its voiced counterpart /ɡ z d b/, using the same dakuten you just learned.12
The clearest single example is 人 (hito, "person") + 人 (hito, "person") → 人々 (hitobito, "people"). The second hito takes a dakuten on its first consonant and becomes bito.1213
Two more everyday examples
折り紙。12
"Origami; paper folding."
本棚。12
"Bookshelf."
In each case, the second element's initial consonant is voiced by adding a dakuten: kami → gami, tana → dana. The mark is the same one you just learned on the chart. Only the trigger, a compound seam, is new.
Why this article stops here
Rendaku does not apply to every compound. It is constrained by Lyman's Law, which blocks rendaku when the second element already contains one of the consonants /ɡ z d b/ (so yama + kado does not become *yamagado, because kado already contains /d/). It also interacts with the native (和語) vs. Sino-Japanese (漢語) vs. loanword (外来語) origin of the second element. And it has well-documented lexical exceptions where rendaku may fail even with no definite blocking factor present.1214
A full treatment of those rules belongs with pronunciation and with kanji-compound reading. This section is here for one purpose: when you meet the dakuten in 人々, 折り紙, or 本棚, you can recognise it as the same mark you just learned, doing real work in real vocabulary.
Good to know
"Muddy" and "half-muddy": what the names actually say
The kanji 濁 means "muddy" or "cloudy," and 半 means "half." So 濁点 reads literally as "muddying mark" and 半濁点 as "half-muddying mark." Japanese metalanguage, or language about language, conceptualises voicing as making a sound "muddy" rather than "loud" or "heavy." This is a useful frame when you hear teachers describe voiced kana as 濁った音 ("muddy sounds").3
Writing はなじ instead of はなぢ for 鼻血
In modern Japanese the "ji" sound is written じ by default, so a learner who has only met じ may reach for はなじ when spelling 鼻血. The correct spelling is:
鼻血。8
"Nosebleed."
The 1986 現代仮名遣い rule retains ぢ and づ only in two contexts: same-morpheme repetition (ちぢむ, つづく) and rendaku in a transparent two-element compound (鼻血, 底力, 三日月). 鼻血 is transparently 鼻 + 血, and the ち voices to ぢ, so the historical spelling is preserved.8
Hearing ば versus ぱ in fast speech
Beginners often expect the ば / ぱ contrast to feel like the English "b" vs. "p" they already know. But the two Japanese sounds are made at exactly the same place: both are bilabial plosives, produced with the lips. They differ only in whether the vocal folds are voicing (/b/ = [b]) or not (/p/ = [p]). To hear the pair at speed, listen for voice-onset timing at the same place of articulation, rather than for a "harder" or "softer" quality.57
A mnemonic that pairs the mark with the articulation
Tenten "turns the voice on": two dots add voicing, so k → g, s → z, t → d, h → b. Maru adds "a puff of air" on the lips: the small circle on the h-row gives p. Pairing the visual mark with the articulatory feature it represents is easier to remember than memorising five conversion tables in isolation.2
The が行鼻濁音 [ŋ] is an older / broadcast feature
Word-medial が was traditionally realised as the nasal [ŋ] in Tokyo speech (が行鼻濁音, ga-gyō bidakuon) and is still drilled in NHK announcer training. But the nasal realisation is less frequent among younger speakers and is now best treated as a marked, broadcast-register feature.56 Recognise it when you hear it on older broadcasts. You do not need to reproduce it yourself.
Typing the marks on a Japanese IME
On a software keyboard or PC IME (input method editor), you almost never insert ゛ or ゜ as separate Unicode characters. Type the romanised voiced form directly (ga for が, pa for ぱ), or type the base kana and then a diacritic key on a kana keyboard. Either path produces the precomposed kana your reader expects to see.1
The dakuten is a separate mark, not a stroke
Adding a dakuten or handakuten does not change the base character's stroke count or stroke order. The mark is written after the base is finished. Practice the base hiragana for stroke order, then add the diacritic as a single, separate gesture on top.2
See also
- The Complete Katakana Chart (Gojūon): How to Read All 46 Base Kana, Dakuten, and Yōon
- Long Vowels in Hiragana: How to Read and Write ああ, いい, うう, ええ/えい, and おう/おお
- The Small つ (Sokuon): How to Read and Pronounce the Geminate Consonant
- The Mora-N (ん) and Its Four Allophones
- Japanese Vowel Devoicing: Why です Sounds Like "Des"