Eleven packets carry two consonants underneath. You say one of them — until a vowel shows up, and then you say both.
Korean allows two consonants in the bottom slot. Your mouth cannot make two unreleased seals at once, so in isolation only one of the two comes out. Which one is fixed per cluster, and there is a rough pattern: the one that survives is usually the one you can hold.
| Cluster | Sound that comes out | Example |
|---|---|---|
| ㄳ | [k̚] | 몫 — a share |
| ㄵ | [n] | 앉다 — to sit |
| ㄶ | [n] | 많다 — to be many |
| ㄺ | [k̚] | 닭 — chicken |
| ㄻ | [m] | 삶 — life |
| ㄼ | [l] | 넓다 — to be wide |
| ㄽ | [l] | 외곬 — a single path |
| ㄾ | [l] | 훑다 — to strip, to scan |
| ㄿ | [p̚] | 읊다 — to recite |
| ㅀ | [l] | 싫다 — to dislike |
| ㅄ | [p̚] | 값 — price |
The reason a cluster loses a letter is purely mechanical: there is nowhere to put the second consonant. Give it somewhere to go — a following packet whose first slot is the silent circle — and it goes there, and you say both.
On its own, only the first survives. Add a vowel and the second one moves into the next packet.
Same mechanism: the ㄱ has a vowel to attach to now.
And the same again.
Compare with 읽고, where no vowel follows.
HanDa corrects every sentence you write or say — in real conversation, not flashcards.