Mandarin -n vs -ng: The Nasal Finals Even Natives Merge

A colleague introduces herself as 陈 Chén. You repeat it back. She says it again, slightly slower. You repeat it again, identically, and she smiles and lets it go.

The surname you might have said instead is 程 Chéng, which was 44th on the mainland frequency list in 2019, while 陈 is in the top five. Same initial. Same second tone. The only thing separating them is which of the two Mandarin nasal finals ends the syllable, -n or -ng, and that ending is the part your ear is least equipped to catch.

Every pronunciation guide hands you the same rule. Tongue tip on the ridge behind your teeth for -n, back of the tongue against the soft palate for -ng. The rule is accurate.

It also quietly implies that the sixteen nasal finals are one problem of equal size, and they are not. Two pairs carry almost the entire difficulty. Four are decisions you never have to make.

Your ear is aimed at the ending because English aimed it there

Say "sin," then "sing." In both words your soft palate drops early and the vowel picks up nasality from the consonant coming after it. English shifts the vowel before -ng too. In English that shift is fully predictable from the consonant, because nasality never separates one English word from another. On the standard description of English vowel nasalization it is a byproduct rather than a signal, and listeners learn to weight the consonant and treat the vowel as noise. In Mandarin that same information is the signal. The habit is what you brought with you.

Mandarin's two endings sit at wide ends of the mouth. Wai-Sum Lee and Eric Zee's IPA Handbook description of Beijing Mandarin calls syllable-final [n] apico-laminal denti-alveolar, tip and blade pressed forward against the teeth, further front than the alveolar English /n/ you are substituting. At the other end, Ohio State's Mandarin IPA chart notes the velar nasal is sometimes pronounced farther back, as a uvular nasal. Nobody has measured the two languages against each other directly, but the descriptions do not overlap.

Two endings described in very different places in the mouth, and they still reach your ear sounding close together. The answer is not in the consonant. It is in the vowel in front of it.

山 and 商 are not the hard kind

Twenty-five Shanghai-raised speakers read a wordlist into a microphone. When the target syllable ended in -ang, the intended -ng came out as -n 0.67% of the time. In relaxed interview speech it climbed to 2.1%. For practical purposes, this pair does not merge.

The reason is that -an and -ang are not carrying the contrast on the ending alone. The vowel shifts too. 山 shān (mountain) has a fronter [a]; 商 shāng (commerce) has a backer [ɑ]. Yihan Guan, who ran that study, puts it plainly: speakers "do not confuse two nasals in the production of the pair [an]-[ɑŋ] because of the existence of the stronger cue [a]-[ɑ] contrast."

The same vowel shift should protect 连 lián against 良 liáng, and 关 guān against 光 guāng, though Guan's wordlist only tested -an against -ang directly. It is probably why 旁观 pángguān (to look on) and 膀胱 pángguāng (urinary bladder) are funny on paper and rarely a problem out loud.

One honest limit on the phonetics. "Fronter" and "backer" are deliberately vague here, because the exact vowel symbols are not settled. Lee and Zee do not transcribe a difference between the two at all, and describe the pre-nasal [a] as retracted regardless of what follows it. The cue is audible; its precise value is still argued about.

Chen vs Cheng: one cue, and it is the weak one

Run those same twenty-five speakers on a vowel other than /a/ and the numbers stop being reassuring. When the syllable ran on /ə/, the intended -ng came out as -n 48% of the time in careful wordlist speech, and 78% in casual conversation. On /i/ it was 58% careful, 79% casual.

Nothing changed except which vowel sat in front of the ending. Guan's explanation is the whole post in one sentence: in [ən]-[əŋ] and [in]-[iŋ], "the only obvious cue is the nasal difference," and the nasal is "a relatively weak cue compared to the vowel difference."

So 陈 chén and 程 chéng have no vowel working for them. Neither do 分 fēn and 风 fēng, or 全民 quánmín (all the people) and 全名 quánmíng (full name). These are the pairs where the ending is the entire contrast, and the ending is quiet.

The collapse also runs mostly one direction, -ng toward -n, so the word that goes missing is usually the -ng one. And casual speech is far worse than careful speech, at least among the Shanghai speakers Guan recorded, which means the person who pronounces 程 crisply for you in a classroom may be a different speaker at dinner, at speed, over a bad phone line. Nothing about your ear changed between those two rooms. The signal did.

You can hit this one on a menu. 清真 qīngzhēn and 清蒸 qīngzhēng are one ending apart, and both get printed on real signage:

In vs ing: even northern listeners failed this test

Fourteen Shanghai Mandarin listeners and fourteen northern Mandarin controls sat through an AXB discrimination task on [in] versus [iŋ]. The Shanghai group performed near chance, which the researchers expected. The northern controls did too. Their write-up says it flatly: "The coda nasals are apparently indistinguishable after /i/ for control and Shanghai Mandarin speakers." The northern controls even showed an unexpected bias toward hearing /ŋ/.

Suyuan Liu and Matthew Faytak raise the possibility of a near merger for the northern controls: people who have no difficulty producing an /in/ versus /iŋ/ contrast but cannot perceive it. Their ultrasound data confirms the northern controls really do keep the two apart in the mouth. They just cannot hear it done.

A Shanghai speaker may stop producing the contrast. A Beijing speaker keeps producing it and still cannot hear it, so you cannot fix your ear by finding a better speaker.

Victor Mair has been reporting the same thing from a teaching chair for years. Writing on Language Log, he notes that some of his students from Taiwan and the mainland, "including those who have graduated from the best schools, routinely mix up -n and -ng," and that on the rare occasions he points it out, "they can't tell the difference." In the comments, the linguist John Rohsenow describes a Taiwanese student who made the distinction only when reading the two Zhuyin symbols aloud, "and nowhere else."

That is anecdote rather than data, one teacher's decades of noticing. But it lines up with the lab work, and the practical consequence is concrete: a native recording of 陈 and 程 is not automatically ground truth.

The front and back nasal finals nobody needs to police

In the full finals table, -ong and -iong only ever take -ng. The contrast does not exist, so you cannot get it wrong. 中 zhōng, 用 yòng, 穷 qióng: no decision to make.

Here is the honest inventory. Eight finals end in -n, eight end in -ng, and five pairs genuinely contrast:

Four finals have no partner at all. Alongside -ong and -iong, neither -üan nor -ün has an -ng counterpart, so those four are decisions you never have to make.

A sixth pair technically exists, 温 wēn and 翁 wēng. It runs on the same weak /ə/ cue as 陈 and 程, so it is not easy, it is only rare: -ueng never combines with an initial, which makes 翁 wēng the single syllable on that side of the pair.

The first three pairs in the table have the vowel doing the work. The last two do not. That is your priority list, and it is shorter than the sixteen finals you were bracing for.

盆友: the merger already got into the writing

盆友 pényǒu is a real word on the Chinese internet, a pun on 朋友 péngyou that spells the merger out, with 小盆友 for 小朋友 close behind. A pronunciation that ignores the ending produces a pinyin that matches a different real character, and the joke wrote itself.

Chinese-language teaching materials treat the same slip without the joke. A guide for the 普通话水平测试, China's national Mandarin proficiency exam, lists 明星 míngxīng read as 明新 míngxīn among its standard error types, and says plainly that the dialects of many places conflate the two endings, usually reading the back-nasal finals as front-nasal ones. That is the -ng toward -n direction the Shanghai recordings showed. The exam's syllabus lists 前后鼻尾音 among its 难点音, its named categories of difficult sounds, and every mispronounced syllable costs points.

None of this is your ear failing. It is a difficulty Mandarin's own examiners have a name for.

The drill that tests the right thing

Find a native recording of 山 shān and 商 shāng and trim each clip before the nasal, so only the vowel is left. If you can still tell them apart, the vowel is doing its job and that pair is safe for you. Run this on somebody else's mouth rather than your own, since your own two vowels will differ because you made them differ.

For production, say a syllable ending in -n and freeze on the -n. Your tongue tip should be pressed forward against your teeth, and it should stay there. If it is floating, you produced a nasalized vowel and no -n at all. That is the failure mode English coarticulation makes easy, and the one Mandarin punishes.

Then put 金 jīn and 京 jīng into YouGlish's Chinese search, which plays clips from real video and filters by region. Run each one on Mainland, then run it again on Taiwan. You are listening for a spread of pronunciations, not for one correct sound.

One limit worth stating before you spend a month on this. On -en/-eng and -in/-ing, reliable production plus context is a reachable goal and perfect perception is not, because the target is not stable across the speakers you will meet. On -in/-ing, northern listeners in a controlled lab could not do it either. Spend the effort where it converts.

What to do the next time someone tells you their surname

Back to 陈 Chén. You are not going to recover that ending from a casual introduction. The information often is not in the signal, and if it is, the speaker may not hear it themselves.

So stop replaying the syllable and ask: 哪个 chén？ Nǎge chén? Which chén? Chinese speakers answer that constantly by naming the parts of the character: 耳东陈 for 陈, 禾呈程 for 程. Nobody finds it rude, and it settles in three seconds what your ear was never going to settle.

Then say the name back with the vowel and the tone right, and let context carry the ending, exactly the way the person you are talking to already does.