Compound Nouns

Japanese builds new nouns by gluing nouns together, and it does this constantly and freely. 電話（でんわ）"telephone," 手紙（てがみ）"letter," 花火（はなび）"fireworks" — all are two smaller words fused into one. Because compounding is so productive, most of the "big words" you will ever read are transparent once you can see the seam and know two rules: how the pieces are read (which shifts the sound, sometimes voicing the second half) and how they are structured (the last piece is always the boss). Master those, and a fearsome-looking word like 東京大学入学試験 becomes a short, decodable phrase.

Two main types, sorted by reading

A compound's flavor depends on whether its parts use their Chinese-derived readings (on'yomi) or their native Japanese readings (kun'yomi). See On'yomi and Kun'yomi Readings for the underlying split.

Sino-Japanese compounds (jukugo) — on'yomi + on'yomi

These are two-kanji (or more) words read entirely with borrowed Chinese readings. They form the bulk of the formal, technical, and written vocabulary.

This type is such a large topic — including how the readings sometimes change at the boundary (学校 gaku＋kō → gak*kō*) — that it has its own page: Compound Words (Jukugo).

Native compounds — kun'yomi + kun'yomi, often with rendaku

When two native Japanese nouns combine, the second element's initial consonant frequently voices — this is rendaku. It is the single most important sound rule for reading native compounds correctly.

手紙 is 手 te ("hand") + 紙 kami ("paper"), but the kami voices to gami: tegami "letter." The same thing happens in 花火 (花 hana "flower" + 火 hi "fire" → hanabi "fireworks") and 青空 (青 ao "blue" + 空 sora "sky" → aozora "blue sky").

Notice 花火大会 is itself a compound of a compound: 花火 ("fireworks") + 大会 ("big gathering, festival"). Rendaku is conditioned — it doesn't fire every time — and the details (Lyman's Law, etc.) live on Rendaku: Sequential Voicing. For reading purposes, just expect the second half to voice and check when it doesn't.

Mixed compounds — the reading types blend

Some compounds mix an on'yomi with a kun'yomi. These "hybrid" readings even have names: 重箱読み（じゅうばこよみ）for on＋kun and 湯桶読み（ゆとうよみ）for kun＋on. They can still rendaku.

台所 is 台 dai (on'yomi) + 所 tokoro (kun'yomi), with tokoro voicing to dokoro: daidokoro. Modern life also produces native-plus-loanword hybrids:

消しゴム = 消し keshi (native, "erasing") + ゴム gomu (a loanword from Dutch, "rubber").

The head-final rule — the last noun is the boss

Every Japanese compound is head-final: the rightmost element is the head (the thing the whole compound fundamentally is), and everything before it modifies that head. This is the structural key.

日本語学校 = 日本語 ("the Japanese language") + 学校 ("school"). The head is 学校, so the whole thing is a kind of school — specifically, a school for Japanese. It is emphatically not a kind of Japanese. Good news for English speakers: English noun-compounds are head-final too ("a Japanese-language school" is also a school), so your instinct here transfers directly. The trap is only when learners over-think and reverse it.

Decoding long compounds: chunk right-to-left

Because compounding is productive and head-final, you can chain pieces indefinitely — and you can read any such chain by working right-to-left, peeling off the head, then the next modifier, and so on. Take a genuinely intimidating example:

Parse 東京大学入学試験 from the right:

1. 試験 (shiken) = exam ← the head; whatever else is here, it's an exam.
2. 入学試験 (nyūgaku shiken) = 入学 ("entering school") + exam → an entrance exam.
3. 東京大学 (Tōkyō daigaku) = 東京 ("Tokyo") + 大学 ("university") → the University of Tokyo.
4. Whole thing → "the University of Tokyo entrance exam."

Each chunk is itself a small head-final compound (大学 is a university, 試験 is an exam), so the strategy is fractal: find the rightmost head, then treat the block before it as a modifier, recursing leftward. This turns wall-of-kanji vocabulary into something you can dismantle on sight — a decoding skill worth practicing deliberately.

Pronunciation: one word, one accent

A compound is a single phonological word, not two words spoken in a row. It takes one unified pitch-accent pattern, and the individual parts lose their original accents. English speakers tend to stress each element separately (like "AIR-PORT"), but 図書館 is one smooth toshokan, not "TO-SHO-KAN" with three beats of stress. Treating the compound as one word — not chopping it into pieces — is half of sounding natural.

Compounds that become verbs

Many Sino-Japanese compound nouns double as verbs by adding する ("to do"): 勉強（べんきょう）→ 勉強する "to study," 電話 → 電話する "to phone," 旅行（りょこう）→ 旅行する "to travel." These are so systematic they get their own treatment on する Nouns.

Common mistakes

Missing rendaku is the number-one compound error for learners. Expect the second half of a native compound to voice: tegami, hanabi, aozora — not tekami, hanahi, aosora.

Respect the head-final rule: the rightmost element is what the compound is. Reversing it inverts the meaning.

A fixed compound is one lexical unit; don't glue its parts back together with の. (の makes a possessive phrase, which is a different construction — see The Possessive Particle の.)

A compound is one phonological word with one accent contour. Over-segmenting it into equally stressed syllables is a giveaway English-transfer error.

Key takeaways

• Japanese freely stacks noun on noun: jukugo (on'yomi, 電話, 会社), native compounds (kun'yomi, often rendaku: 手紙, 花火, 青空), and mixed forms (台所, 消しゴム).
• Compounds are head-final — the last element is what the word is; earlier elements modify it. English shares this, so the instinct transfers.
• Long compounds decode right-to-left: peel off the head, then read each earlier block as a modifier (東京大学入学試験 → the Tokyo-University entrance exam).
• Expect rendaku (voicing of the second half) in native compounds, and pronounce the whole compound as one word with one accent.