The Letter ㅇ: Silent Onset vs. Final [ŋ]

The letter ㅇ leads a double life, and understanding it is what makes the whole liaison system click. In one position it is completely silent; in the other it is a full-bodied consonant. That split is not a quirk to memorize — it is the exact reason some syllables relink and others cannot. Once you see that "liaison" really means "fill the empty ㅇ onset slot," it becomes obvious which final consonant is the one that can never move.

ㅇ as an onset: silent, a placeholder for the vowel

At the start of a syllable, ㅇ has no sound at all. It is a graphic placeholder — Hangul requires every syllable block to have an initial position filled, so when a syllable begins with a vowel, ㅇ sits in that slot and stays mute. 아 is just "a," 이 is just "i," 우 is just "u." Nothing is pronounced for the ㅇ; it is a silent chair holding the vowel's place.

Every ㅇ in 아이, 우유 is silent — these are pure vowels, [a-i], [u-yu]. That silent onset is the whole point: because the slot is empty, it is exactly the slot a preceding batchim can slide into. When you say 밥을 as [바블], the ㅂ moves into the empty onset of 을. Liaison is simply a batchim filling that vacancy.

ㅇ as a batchim: the real sound [ŋ] — and it can't move

At the end of a syllable, ㅇ is not silent at all. It is [ŋ], the "ng" in English "sing" or "song." And here is the decisive fact: Korean has no [ŋ] at the start of a syllable. No Korean word begins with an "ng-" sound. So when a syllable ends in ㅇ and a vowel follows, the [ŋ] has nowhere to go — a batchim can only relink into an empty onset slot, and [ŋ] is not allowed to sit in an onset. It therefore stays put as the coda, and the following vowel keeps its own silent ㅇ.

종이 is [종이], jong-i — the [ŋ] holds its place at the end of 종, and 이 keeps its own silent onset. It is not re-syllabified to a "jo-ngi." Compare this with what a single ㄱ or ㅅ would do in the same spot: those would glide forward. The [ŋ] refuses, because the onset door is closed to it.

Watch the contrast inside one sentence. 강은 stays [강은], gang-eun — the [ŋ] does not link. But 물이 relinks to [무리] (the ㄹ slides over), and 맑아요 relinks to [말가요] (the cluster splits, more below). The [ŋ] of 강 is the one consonant sitting still while everything around it glides.

방이 → [방이], bang-i, the [ŋ] frozen in place, while 넓어요 → [널버요] shows a cluster relinking freely. The pattern never fails: a final ㅇ is a real sound that stays; every other single batchim is a consonant looking for an empty onset to fill.

The cleanest demonstration puts both behaviors inside a single word. In 공원 ("park"), the ㅇ of 공 and the ㄴ of 원 sit side by side — and only one of them can ever move:

공원에서 is [공워네서]: the [ŋ] of 공 stays welded to its own syllable, but the ㄴ of 원 links straight onto 에서. Same word, two batchim, one rule — the ㄴ has an empty onset to fill, and the [ŋ] does not.

Double-batchim: the cluster splits, the second member links

A syllable can end in a two-consonant cluster — a 겹받침 such as ㄺ, ㄼ, ㄵ, ㅄ, ㄻ. In isolation you hear only one of the two (see double-batchim for which member survives before a pause). But when a vowel follows, the cluster does something elegant: it divides. The first consonant stays behind as the coda of its own syllable, and the second consonant slides forward into the empty onset next door.

읽어요 has the cluster ㄺ (ㄹ + ㄱ). It splits: the ㄹ stays as the coda of 일, and the ㄱ links onto 어 — [일거요]. You hear both consonants: neither is lost. (책을 also relinks normally: [채글].)

앉아서 has the cluster ㄵ (ㄴ + ㅈ). The ㄴ stays as the coda of 안, the ㅈ moves to onset the 아 — [안자서]. Again both survive.

없어요 has the cluster ㅄ (ㅂ + ㅅ). The ㅂ stays as the coda of 업, the ㅅ links forward — but because it now follows the obstruent [p], it tenses to [ㅆ], giving [업써요]. (That tensing is 경음화, which the Revised Romanization convention does not spell out — the romanization stays "eopseoyo," while the bracket shows the true [업써요].) The key point stands: the cluster split, and both consonants are heard.

The commonest error with these clusters is to collapse them to a single sound — to read 읽어 as if it were 이거 [이거], dropping the ㄹ entirely, or 앉아 as 아자, dropping the ㄴ. Korean keeps both. That is precisely what the spelling is telling you by writing two batchim letters.

Putting the two ㅇ facts together

The reason the double-batchim rule and the ㅇ rule belong on the same page is that they are the same principle seen from two sides. Liaison fills empty onset slots. A single non-ㅇ batchim fills the slot and moves. A cluster's second member fills the slot and moves, while the first stays. And the batchim ㅇ — [ŋ] — never moves, because it is not looking for a slot; it is already a fully realized coda consonant, and its target onset position is forbidden to it in Korean. Silent-onset ㅇ opens the door; coda ㅇ is the one guest who cannot walk through it.

Common Mistakes

1. Relinking a final ㅇ. The single most common ㅇ error: treating 강이 like 감이 and saying "ga-ngi." The [ŋ] stays as a coda; 이 keeps its own silent onset.

• ✗ 강이 read as "ga-ngi."
• ✓ 강이 [강이], gang-i — the [ŋ] does not move.

2. Pronouncing an onset ㅇ. An onset ㅇ is silent. 아이 is [a-i], never "nga-i" or "ha-i" — there is no consonant there at all.

• ✗ 아이 given a hard consonantal onset.
• ✓ 아이 [a-i] — pure vowels, silent ㅇ.

3. Collapsing a double batchim to one sound. Reading 읽어 as [이거] drops the ㄹ; reading 앉아 as [아자] drops the ㄴ. Both cluster members must be heard.

• ✗ 읽어요 read as "i-geo-yo."
• ✓ 읽어요 [일거요] — ㄹ coda + ㄱ onset, both present.

4. Splitting the cluster in the wrong order. The first letter stays as coda and the second moves forward — not the reverse. 넓어 is [널버] (ㄹ stays, ㅂ moves), not ×[넙어].

• ✗ 넓어요 read as "neob-eo-yo," moving the wrong member.
• ✓ 넓어요 [널버요] — ㄹ stays, ㅂ links.

5. Forgetting the linked ㅅ tenses. In 없어요, 값이, 몫이, the ㅅ that links forward comes out tense: [업써요], [갑씨], [목씨]. Don't soften it to a plain [s].

Key Takeaways

• Onset ㅇ is silent — a placeholder for the vowel and the empty slot a batchim relinks into. This is what makes liaison possible.
• Batchim ㅇ is [ŋ], a real consonant. Because Korean has no syllable-initial [ŋ], a final ㅇ can never relink: 종이 [종이], 강이 [강이], 방이 [방이].
• Reframe liaison as "fill the empty ㅇ onset slot," and the exception explains itself.
• A double batchim splits before a vowel: the first consonant stays as coda, the second links forward — 읽어 [일거], 앉아 [안자], 넓어 [널버], 없어 [업써] (the linked ㅅ tenses to [ㅆ]).
• The cardinal cluster error is collapsing it to one sound — Korean keeps both consonants.