Japanese Counter Sound Changes: Why 一本 Is いっぽん, Not いちほん

Japanese counter sound changes are systematic shifts in pronunciation. They happen when a number joins a counter word, turning 一 (いち) + 本 (ほん) into いっぽん rather than the literal いちほん.¹² These shifts can look random to a beginner. But they follow a small set of rules that, once learned, help you predict the reading of a counter you have never seen.

Overview

A counter (助数詞, josūshi) is the short word Japanese attaches to a number to count a specific kind of thing: 本 for long objects, 匹 for small animals, or 杯 for cupfuls.² When certain numbers meet certain counters, the sound at the boundary changes to make the cluster easier to say.

That smoothing is called 音便 (onbin), euphonic change.¹² It is the reason one pencil is いっぽん, three cats is さんびき, and ten cups is じっぱい rather than the readings you would guess from the parts.

What "irregular" really means here

The shifts are not random exceptions to be memorized one by one. They are systematic euphony: regular sound changes at the number-plus-counter boundary that make the cluster easier to pronounce.¹²

Three processes are involved, and the same three recur across many counters: gemination (促音便, the small っ), half-voicing (半濁音, h becoming p), and voicing (濁音, h becoming b).¹² They come from phonology, but each counter fixes which process it uses. Once you know that process, the counter's whole reading pattern becomes predictable.

The deep phonetics of the geminate's "silent beat" and of sequential voicing belong in the dedicated articles on those topics. This article uses them descriptively and points there for the underlying mechanism, rather than re-deriving it.

Which numbers are the troublemakers

Only five numbers cause changes. The numbers 1 (いち), 6 (ろく), 8 (はち), and 10 (じゅう) trigger gemination, and 3 (さん) triggers voicing.²

The numbers 2 (に), 4 (よん), 5 (ご), 7 (なな), and 9 (きゅう) attach with no shift at all.²

The split has a phonetic cause

The numbers 1, 6, 8, and 10 end in a mora, a rhythm unit: ち, く, ち, and a う from the older じふ. That ending reduces to a small っ before a voiceless-obstruent counter. The number 3 ends in the moraic nasal ん, which causes voicing instead.¹² The set 1, 3, 6, 8, 10 is the predictive core of everything below.

The three sound-change mechanisms

Each of the three processes acts on a counter's initial consonant. Learn the trigger for each and you can read most counters on sight.

Gemination: the small tsu っ (促音便)

Before a voiceless-obstruent counter, meaning one in the k-, s-, t-, or h-/p-row, the numbers 1, 6, 8, and 10 drop their final mora and add a geminating っ.¹² The phonetic detail of that doubled, held consonant belongs in the sokuon treatment. Here it is enough to know that っ appears.

鉛筆えんぴつを一本いっぽんください。³
"One pencil, please."

Coverage differs across the four numbers. The numbers 8 and 10 geminate broadly across the k-, s-, t-, and h-/p-rows (八冊 はっさつ, 十冊 じっさつ, 八回 はっかい).³

The number 6 is choosier: it geminates before k- and h-/p-row counters (ろっかい, ろっぽん) but generally not before the s-row (六冊 ろくさつ, 六歳 ろくさい).³

Half-voicing: h becomes p (半濁音)

The h-row counters (は行: 本 hon, 匹 hiki, 杯 hai, 分 fun, 百 hyaku) harden their initial h to p when the preceding number geminates: after 1, 6, 8, and 10.²⁴ The two processes act together. The number adds っ, and the counter's h becomes p.²

So 本 ほん becomes いっぽん, ろっぽん, はっぽん, and じっぽん. 匹 ひき becomes いっぴき, ろっぴき, はっぴき; 杯 はい becomes いっぱい, ろっぱい, はっぱい; and 分 ふん becomes いっぷん, ろっぷん, はっぷん.²³⁴

コーヒーを一杯いっぱい飲のみました。³
"I drank one cup of coffee."

Voicing: h becomes b after 三 (濁音)

After 三 (さん), an h-row counter's initial h voices to b, because the moraic nasal ん conditions voicing rather than gemination.¹² So 三本 is さんぼん, 三杯 is さんばい, 三百 is さんびゃく.²⁴

猫ねこが三匹さんびきいます。³
"There are three cats."

This is the same family of process as rendaku, the sequential voicing seen in compound words. The deep mechanism is treated there. Two important non-uniformities sit inside the "三 voices to b" rule and must be stated, not generalized.

三分 is さんぷん, not さんぼん

The minute counter 分 takes a hardened p after 三, giving 三分 さんぷん, not the b-form the general rule would predict.⁵ The reading さんぶん is a different word, "three parts." Likewise 三階 (floors) is conventionally さんがい, where the k of 階 voices to g after 三, though さんかい is also accepted.⁶

These cases show that 三 generally causes voicing of the following obstruent (h to b, k to g). But the exact result is fixed for each counter, and 分 resists it in favor of the p-form.⁵⁶

The rules by consonant row

The cleanest way to predict a reading is to sort counters by their initial consonant. Each row reacts to the trigger numbers in its own consistent way.

h-row counters (本, 匹, 杯, 分)

This is the messiest row, because both gemination plus hardening (after 1, 6, 8, 10) and voicing (after 3) apply.²⁴ It is also where いっぽん, さんぼん, and ろっぽん appear. The reading column carries the kana, so the kanji headers stay bare.

#	本 (hon)	匹 (hiki)	杯 (hai)	分 (fun, minutes)	Process
1 いち	いっぽん	いっぴき	いっぱい	いっぷん	gemination + h→p
2 に	にほん	にひき	にはい	にふん	none
3 さん	さんぼん	さんびき	さんばい	さんぷん	voicing h→b (分: h→p, see note)
4 よん	よんほん	よんひき	よんはい	よんぷん	none (分: よんぷん)
5 ご	ごほん	ごひき	ごはい	ごふん	none
6 ろく	ろっぽん	ろっぴき	ろっぱい	ろっぷん	gemination + h→p
7 なな	ななほん	ななひき	ななはい	ななふん	none
8 はち	はっぽん	はっぴき	はっぱい	はっぷん	gemination + h→p
9 きゅう	きゅうほん	きゅうひき	きゅうはい	きゅうふん	none
10 じゅう	じっぽん／じゅっぽん	じっぴき／じゅっぴき	じっぱい／じゅっぱい	じっぷん／じゅっぷん	gemination + h→p

The minute counter 分 has one quirk among the regular numbers: the 4-row is よんぷん, with a hardened p, not よんふん.³⁵ The counter 百 follows the same pattern in its irregular cells (三百 さんびゃく, 六百 ろっぴゃく, 八百 はっぴゃく), while other hundreds keep ひゃく. Note that 100 alone is simply ひゃく, never 一百.⁴

k-row counters (個, 回, 階)

The numbers 1, 6, 8, and 10 geminate to っ + k. The number 3 does not voice 個 or 回 (さんこ, さんかい), but 階 takes voicing to g (さんがい).²³⁶

#	個 (ko)	回 (kai)	階 (kai, floors)
1	いっこ	いっかい	いっかい
2	にこ	にかい	にかい
3	さんこ	さんかい	さんがい／さんかい
4	よんこ	よんかい	よんかい
5	ごこ	ごかい	ごかい
6	ろっこ	ろっかい	ろっかい
7	ななこ	ななかい	ななかい
8	はっこ（はちこ）	はっかい	はっかい
9	きゅうこ	きゅうかい	きゅうかい
10	じっこ／じゅっこ	じっかい／じゅっかい	じっかい／じゅっかい

The counter 個 allows both はっこ and はちこ in the 8-row.³ Compared with the h-row, the k-row is simpler: gemination doubles the k but adds no hardening, and only 階 shows any voicing after 三.

s-row and t-row counters (歳, 冊, 通)

On the s-row (歳 sai, 冊 satsu), the numbers 1, 8, and 10 geminate, but 6 generally does not (六歳 ろくさい, 六冊 ろくさつ), and 3 does not voice (さんさい, さんさつ).²³

#	歳 (sai, age)	冊 (satsu, books)
1	いっさい	いっさつ
2	にさい	にさつ
3	さんさい	さんさつ
4	よんさい	よんさつ
5	ごさい	ごさつ
6	ろくさい	ろくさつ
7	ななさい	ななさつ
8	はっさい	はっさつ
9	きゅうさい	きゅうさつ
10	じっさい／じゅっさい	じっさつ／じゅっさつ

The t-row counter 通 (tsū, for letters) follows the s-row pattern for gemination: 一通 いっつう, 八通 はっつう, 十通 じっつう／じゅっつう, with 三 left unvoiced (さんつう).³ One s-row form is a pure memorize-it exception: 二十歳 is read はたち. This is a special lexical reading outside the regular number-plus-歳 series, not a euphonic rule.³

Counters that barely change (枚, 人, 台)

Some counters attach with no gemination, voicing, or hardening at all. They are the regular comparison case: the number keeps its plain form, and the counter is unchanged.²³

#	枚 (mai)	台 (dai)	人 (nin)
1	いちまい	いちだい	ひとり (special)
2	にまい	にだい	ふたり (special)
3	さんまい	さんだい	さんにん
4	よんまい	よんだい	よにん (special)
5	ごまい	ごだい	ごにん
6	ろくまい	ろくだい	ろくにん
7	ななまい	ななだい	しちにん／ななにん
8	はちまい	はちだい	はちにん
9	きゅうまい	きゅうだい	きゅうにん／くにん
10	じゅうまい	じゅうだい	じゅうにん

The counters 枚 and 台 are fully regular.²³ The counter 人 is regular for the consonant of にん, but it has its own number-side irregularities (ひとり, ふたり, よにん). Those are number-reading exceptions rather than counter euphony, and they belong with the treatment of numbers and the 人 counter. They are noted here only for contrast.

Reading a counter you have never seen

The rules above combine into a short procedure you can run in your head. It predicts the default reading, but a handful of lexical exceptions still need a dictionary check.

A quick decision procedure

The diagram below captures the procedure as a single decision path. The reading depends on two sequential yes/no checks: which row, then which number. A flowchart makes that branching clearer than prose.

Step 1: identify the counter's initial consonant row: k-, s-, t-, h-/p-, or a safe row such as m-, d-, or n-.² Step 2: check whether the number is in the gemination set {1, 6, 8, 10} or is the voicing trigger 3.²

Step 3: apply the matching shift. For 1, 6, 8, 10 before a k-, s-, t-, or h-row counter, geminate with っ. For an h-row counter, also harden h to p. For 3 before an h-row counter, voice h to b. For 2, 4, 5, 7, 9, attach plainly.¹²

The procedure predicts the default, not every cell

The number 6 often skips gemination on the s-row (六冊 ろくさつ), and individual counters fix their own readings (三分 さんぷん, 三階 さんがい). Confirm a new counter against a dictionary; the steps give you the expected reading, not a guarantee for every lexical exception.³⁵⁶

The master irregularity table

This consolidated grid is the main takeaway. Rows are the numbers 1 to 10; columns are one representative counter per consonant-row class. The irregular readings cluster in the 1, 3, 6, 8, and 10 rows.²³⁴

#	本 (h→p)	杯 (h→p)	個 (k)	回 (k)	歳 (s)	冊 (s)
1 いち	いっぽん	いっぱい	いっこ	いっかい	いっさい	いっさつ
2 に	にほん	にはい	にこ	にかい	にさい	にさつ
3 さん	さんぼん	さんばい	さんこ	さんかい	さんさい	さんさつ
4 よん	よんほん	よんはい	よんこ	よんかい	よんさい	よんさつ
5 ご	ごほん	ごはい	ごこ	ごかい	ごさい	ごさつ
6 ろく	ろっぽん	ろっぱい	ろっこ	ろっかい	ろくさい	ろくさつ
7 なな	ななほん	ななはい	ななこ	ななかい	ななさい	ななさつ
8 はち	はっぽん	はっぱい	はっこ	はっかい	はっさい	はっさつ
9 きゅう	きゅうほん	きゅうはい	きゅうこ	きゅうかい	きゅうさい	きゅうさつ
10 じゅう	じっぽん／じゅっぽん	じっぱい／じゅっぱい	じっこ／じゅっこ	じっかい／じゅっかい	じっさい／じゅっさい	じっさつ／じゅっさつ

Good to know

"1, 6, 8, 10 hit hard; 3 softens"

For h-row counters, the numbers 1, 6, 8, and 10 produce a hard っ + p (いっぽん, ろっぽん, はっぽん, じゅっぽん), while 3 softens to a voiced b (さんぼん).²⁴

The hook works because it groups the four gemination triggers against the single voicing trigger. That is exactly the split the phonology makes: mora-final reduction for 1/6/8/10 versus moraic-nasal voicing for 3.¹

十 was once じふ, which is why じっ is the standard form

The number 十 historically read じふ. Before か-, た-, and は-row sounds, the フ reduced to っ, giving じっ (じっぽん, じっかい).⁷

Most speakers now read 十 in isolation as じゅう and naturally produce じゅっ-. That is why both readings circulate, though じっ remains the historically regular form.⁷

じっ and じゅっ are both accepted, with a datable shift

Most dictionaries treat じっ (じっかい, じっぽん) as the standard reading and じゅっ as a variant.⁷ For learners, the practical advice is to produce either and recognize both: tests and dictionaries lean じっ, everyday speech leans じゅっ.⁷⁸

The November 2010 revision of the 常用漢字表 added the remark 「『ジュッ』とも」 to the 十 entry, formally admitting じゅっ alongside ジッ because the reading had become too widespread to ignore.⁹⁸

The 四 / 七 / 九 escape hatch

A common beginner error is to treat 4, 7, and 9 like the trigger numbers and harden or geminate the counter, inventing a form such as よんぼん. None of these numbers is in the trigger set, so their counters keep the plain consonant.

The correct readings attach plainly:

鉛筆えんぴつが四本よんほんあります。³
"There are four pencils."

The readings よん, なな, and きゅう are themselves preferred partly because they avoid the awkward shifts that し, しち, and く would invite. The number-internal choice of よん over し belongs with the numbers material. Here it matters only as the contrast that keeps these counters regular.²³

三分 is さんぷん, not さんぶん

Applying the generic "3 voices h to b" rule to the minute counter produces さんぶん, but that spelling is a different word, "three parts."⁵ The minute counter 分 is the one common h-row counter whose 三 cell takes the hardened p:

三分さんぷん待まってください。⁵
"Please wait three minutes."

This is a lexically fixed exception to the otherwise reliable "三 to b" pattern, so it is worth fixing in memory as its own item.⁵

See also

• Geminate Consonants (Sokuon っ): The Silent Pause
• Rendaku: When K Becomes G in Compound Words
• Dakuten and Handakuten: The Voicing Marks on Hiragana
• The Mora-N (ん) and Its Four Allophones
• Japanese Numbers: How to Count from 1 to 100,000,000 (and Beyond)
• Counters by Category: A Reference Index
• 枚 (Mai) Counter: Flat, Thin Objects in Japanese

References

Footnotes

1. Vance, Timothy J. The Sounds of Japanese. Cambridge University Press, 2008. (Reference for the phonological terms 促音便 sokuonbin / gemination, 濁音 voicing, and 半濁音 handakuon, and for the historical じふ-class reduction behind the geminating numbers.) ↩ ↩² ↩³ ↩⁴ ↩⁵ ↩⁶ ↩⁷ ↩⁸ ↩⁹

2. "Japanese counter word." Wikipedia. https://en.wikipedia.org/wiki/Japanese_counter_word (Section on euphonic changes: gemination of 1/6/8/10 before voiceless-obstruent counters, the h→p shift, and the h→b voicing after 三; statement that ji- is the older reading of 十 replaced by ju- in recent generations.) ↩ ↩² ↩³ ↩⁴ ↩⁵ ↩⁶ ↩⁷ ↩⁸ ↩⁹ ↩¹⁰ ↩¹¹ ↩¹² ↩¹³ ↩¹⁴ ↩¹⁵ ↩¹⁶ ↩¹⁷ ↩¹⁸ ↩¹⁹ ↩²⁰ ↩²¹ ↩²² ↩²³ ↩²⁴ ↩²⁵

3. "本 (Wiktionary)" and counter paradigm cross-checks: Wiktionary entries for 本, 匹, 杯, 分, 百, 個, 回, 歳, 冊, 階. https://en.wiktionary.org/wiki/本 (Per-number kana readings of the listed counters.) ↩ ↩² ↩³ ↩⁴ ↩⁵ ↩⁶ ↩⁷ ↩⁸ ↩⁹ ↩¹⁰ ↩¹¹ ↩¹² ↩¹³ ↩¹⁴ ↩¹⁵ ↩¹⁶ ↩¹⁷ ↩¹⁸

4. "百 (Wiktionary)." https://en.wiktionary.org/wiki/百 (Readings 三百 さんびゃく, 六百 ろっぴゃく, 八百 はっぴゃく; other hundreds keep ひゃく.) ↩ ↩² ↩³ ↩⁴ ↩⁵ ↩⁶ ↩⁷

5. 三省堂『大辞林』, entries 三分（さんぷん）and 三分（さんぶん）, as reflected in dictionary aggregations (Nihongo Master / JapanDict). The minute counter 分 takes 三 + ぷん (さんぷん), distinct from the homographic noun 三分 さんぶん "three parts". https://www.nihongomaster.com/japanese/dictionary/word/28738/ ↩ ↩² ↩³ ↩⁴ ↩⁵ ↩⁶ ↩⁷

6. 三省堂『大辞林』, entry 三階, recording both さんがい and さんかい, with さんがい the conventional reading. https://www.nihongomaster.com/japanese/dictionary/word/28582/ ↩ ↩² ↩³ ↩⁴

7. 漢字文化資料館 (大修館書店). 「子どもが『十回』は正しくはジッカイと読むのだと習ってきたのですが…」 漢字Q&A Q0133. https://kanjibunka.com/kanji-faq/old-faq/q0133/ (Etymology: 十 was historically ジフ; before か/た/は-row sounds the フ reduced to っ, giving ジッ. じっかい is the form most dictionaries treat as standard; じゅっかい is a widespread modern variant.) ↩ ↩² ↩³ ↩⁴

8. 光村図書出版.「『二十回』は『ニジュッカイ』と読んでいいの？」教科書の言葉Q&A 第20回. https://www.mitsumura-tosho.co.jp/webmaga/kotoba-to-manabi/kotoba/detail20 (Confirms that じゅっ was admitted as an accepted reading alongside じっ when the 常用漢字表 was revised in November 2010, reflecting that the reading had become too widespread to ignore.) ↩ ↩²

9. 文化庁. 『常用漢字表』(平成22年内閣告示第2号), 2010. The entry for 十 lists the on-readings ジュウ・ジッ with the remark 「『ジュッ』とも」 added in the November 2010 revision. https://www.bunka.go.jp/kokugo_nihongo/sisaku/joho/joho/kakuki/14/tosin02/ ↩