Sound Changes in Numbers (三百, 六百, 八百)

By now you've been warned repeatedly that 三百 isn't san-hyaku but さんびゃく sanbyaku, and 六百 isn't roku-hyaku but ろっぴゃく roppyaku. It's time to stop memorizing these as isolated exceptions and see them for what they are: the output of two regular phonological forces that operate at the joint between a number and whatever follows it. The huge payoff is that these are the same forces that govern the whole counter system — learn them here on 百 and 千 and you've simultaneously learned why 一本 is ippon, 三本 is sanbon, and 十分 is juppun. This is one rule set doing an enormous amount of work.

The two forces

Every irregular number-reading in Japanese comes from one of two euphonic changes at the boundary:

1. Gemination (促音化): After 一, 六, 八, 十 — numbers that historically ended in a stop — the boundary tightens into a small っ and the following consonant doubles (and if it's h, it hardens to p). This is why 六 + 百 → ろっぴゃく roppyaku.

2. Voicing (連濁 / nasal assimilation): After the moraic ん — which lives at the end of 三 (san) and 何 (nan) — a following h softens to b (or, for a few counters, p), and s softens to z. This is why 三 + 百 → さんびゃく sanbyaku and 三 + 千 → さんぜん sanzen.

Everything below is these two forces applied to specific place-words and counters. The numbers 2, 4, 5, 7, 9 never trigger either change — they're phonologically "quiet," so 二百 ni-hyaku, 五百 go-hyaku, 九百 kyū-hyaku are all perfectly regular.

The hundreds (百)

百 hyaku begins with h, so it's a prime target for both forces. Watch the three that change:

Notice the tidy split: 三 and 何 voice the h to a soft b (because they end in ん); 六 and 八 geminate and harden it to a doubled p. Two forces, four irregular hundreds.

The thousands (千)

千 sen begins with s, so voicing turns it to z and gemination doubles it. But because s isn't h, the pattern is slightly different from 百 — and the difference is exactly where learners over-correct.

Here is the trap. With 百, both 六 and 八 geminate (roppyaku, happyaku). With 千, only 八 geminates (hassen) — 六千 stays regular roku-sen. Why the asymmetry? Gemination before h is very strong (so 六百 must double), but before s it's weaker and 六 simply doesn't trigger it. There's no way to reason this out from meaning; you note that 六 doubles before 百 but not before 千, and move on.

That last one packs the contrast into a single number: 六万 is a clean roku-man and 六千 is a clean roku-sen — 六 changes nothing here, which is exactly what makes 六百 roppyaku feel like the odd one out.

一 before counters: pure gemination

The number 一 ichi is the great geminator. Before almost any counter beginning with k, s, t, or h/p, it drops its -chi and fuses into a small っ that doubles the counter's first consonant:

Honest complication: h → b or h → p after 三?

Here is a genuine irregularity, not a clean rule — so learn it as a fact. After ん (三, 何), most h-initial counters voice to b: 三本 sanbon, 三匹 sanbiki, 三杯 sanbai, 三百 sanbyaku. But the minutes counter 分 breaks ranks and hardens to p: 三分 is さんぷん sanpun, not sanbun. Same environment, different outcome — and no meaning-based reason for it.

The reason to teach the sound changes on 百/千 is that this whole grid falls out of the same two forces — so once 三百/六百/八百 are automatic, 三本/六本, 三分/六分, 三匹/六匹 need almost no extra effort. The full counter grid lives on Counter Sound Changes: The Master Pattern, and the h-counter you'll use most is on 〜本 (Long Cylindrical Things).

That 十分 juppun (ten minutes) is 十 geminating before 分 — the very same force behind 六百 roppyaku. (Watch out: 十分 read as jūbun means "enough / sufficient" — a different word that happens to share the kanji.)

What doesn't change: 万, 億, and vowel-initial 億

Two of the big place-words are immune. 万 man begins with m (a nasal), which resists both forces, so every multiple is clean: 三万 san-man, 六万 roku-man, 八万 hachi-man — no sanban, no roppan. And 億 oku begins with a vowel, so there's nothing for gemination to double: 一億 is a plain ichi-oku, not ikkoku. But 兆 chō begins with a consonant, so it does geminate: 一兆 → いっちょう itchō, 八兆 → hatchō. Same logic, opposite result, decided entirely by the first sound of the place-word.

A fossil worth knowing: 八百屋

One charming exception you'll actually meet: 八百屋 "greengrocer" is read やおや yaoya, not happyaku-ya. Here 八百 is an ancient idiom for "countless / a great many" (the shop with countless kinds of vegetables), and it keeps a frozen native reading やお rather than the Sino happyaku. It's a fossil, not a counterexample to the rules — but it's the kind of thing that trips up a learner who's just drilled 八百 = happyaku.

Common mistakes

Reading 六百 as regular roku-hyaku is the single most common number-pronunciation error. It's roppyaku, with a held っ and a p.

The mirror error: once you know roppyaku, you assume rossen. But 六 geminates before 百, not before 千. Six-thousand is the plain roku-sen.

Most h-counters voice to b after 三 (三本 sanbon), so learners produce sanbun. But 分 is the maverick that takes p: 三分 is sanpun.

Gemination needs a consonant to double. 億 begins with a vowel, so 一億 stays a clean ichi-oku — unlike 一兆 itchō, where the ch does double.

Key takeaways

• Two forces cause every irregular number reading: gemination after 一/六/八/十 and voicing after 三/何/ん. The numbers 2, 4, 5, 7, 9 never change.
• Hundreds: 三百 sanbyaku, 六百 roppyaku, 八百 happyaku (the rest regular).
• Thousands: 三千 sanzen, 八千 hassen — but 六千 is regular roku-sen.
• The same forces power counters: 一本 ippon, 十分 juppun, 三分 sanpun — learn them once here, use them everywhere.
• Behavior depends on the first sound of the place-word/counter: h-initial changes most; m-initial (万) and vowel-initial (億, 円) never change.