Why one kanji has several readings

Two layers of language

Japan had a spoken language before it borrowed a writing system from China. When kanji arrived, each character brought a Chinese reading with it and was also attached to the Japanese word that already meant the same thing.

So 山 is san from the borrowed reading and yama from the native word. Both are correct, and which one a word takes is a property of that word.

The borrowed reading is the on'yomi. The native one is the kun'yomi.

The pattern that mostly holds

• A kanji standing alone usually takes the kun reading. 山 alone is yama. 水 alone is mizu.
• Two kanji together usually take on readings. 火山 is kazan, not hiyama. 水道 is suidou.
• A kanji with a hiragana tail usually takes kun. 食べる is taberu; the 食 in 食事 is shoku.

Where it breaks

Some words mix both — 場所 is basho, an on reading and a kun one. Some names take readings you cannot derive from the characters. And some words are simply exceptions: 今日 is kyou and does not decompose into anything.

This sounds worse than it is. The exceptions cluster in very common words, and very common words are learned by repetition rather than by rule.

What to do about it

Learn readings as words, not as characters. Knowing 火山 is kazan is immediately usable; knowing that 火 has the readings ka, hi and bi is not yet usable for anything.

This is why typing whole sentences beats typing isolated words. Readings come with context, and context is what tells you which reading is the right one.