The third tone is rarely pronounced the way textbooks describe it. In isolation, sure, mǎ dips down and rises back up, the full 2-1-4 contour. But in actual speech, the full dipping third tone almost never happens. What comes out of people's mouths is either a second tone or something chopped in half. This is the single biggest reason learners sound unnatural even when their individual tones are correct.
The core rule is simple. Two third tones in a row, the first becomes a second tone. 你好 is written nǐ hǎo but spoken as ní hǎo. Same for 很好 (hén hǎo), 可以 (ké yǐ), 所以 (suó yǐ), 水果 (shuí guǒ). Every two-syllable third-tone pair follows this rule, no exceptions. The pinyin never changes to reflect it, textbooks, dictionaries, HSK exams all show the original third-tone marks. The spelling is nǐ hǎo forever. The sound is ní hǎo forever. This gap between what's written and what's spoken, it's really just colloquial speech doing its thing. English has the same kind of thing, no?
When three third tones line up, word boundaries decide which ones change. 我很好 (wǒ hěn hǎo) splits into 我 and 很好, so only 很 changes — wǒ hén hǎo. 展览馆 (zhǎnlǎnguǎn, exhibition hall) splits as 展览 and 馆, so the first two become second tone, zhán lán guǎn. Memorizing groupings isn't necessary. The ear picks up the pattern over time, the same way English speakers learn "want to" becomes "wanna" without ever being taught a rule for it.
The third rule covers the most common case. When a third tone sits before a first, second, or fourth tone, which happens constantly, it becomes a half third. The rise gets dropped entirely. The voice goes low and stays there. 很多 (hěn duō) sounds like hěn stays in the basement while duō goes high and flat. 美国 (měi guó) has měi low and uncut, guó rising. 好看 (hǎo kàn) starts low then drops further. This is the rule used most often in everyday speech, because third tones are far more likely to appear before other tones than next to another third tone.
The practical summary is two lines. 3rd + 3rd → 2nd + 3rd, the first goes up. 3rd + anything else → half 3rd, cut the tail, stay low. That's it. Everything else is practice. Reading pinyin aloud without applying these rules produces the robot voice that native speakers politely pretend not to notice. Recording oneself and playing it back next to native audio catches it instantly.
I remember sitting in a coffee shop in Beijing when a classmate said wǒ hěn hǎo, full third tones, dips and all, and the barista paused a full second before replying. She understood. But she had to decode it. Tone sandhi isn't decoration. It's the difference between being understood immediately and being understood after a pause.
Train your ear →Practice distinguishing tones with our free Tone Quiz — hear a syllable, pick which tone it is, and build the reflex that makes tone sandhi feel natural. No signup.
FAQ
Does this rule apply to all third-tone words?
Every two-syllable combination where both characters are third tone follows this rule, no exceptions. Common ones: 可以 (ké yǐ), 所以 (suó yǐ), 理解 (lí jiě, to understand), 影响 (yíng xiǎng, influence).
Do native speakers know they're doing this?
Most don't consciously think about it — just like English speakers don't think about how "want to" becomes "wanna" in fast speech. It's automatic. But when pointed out, they'll immediately recognize it's true.
Does tone sandhi affect HSK listening tests?
Yes. HSK listening sections use natural speech, which means speakers apply tone sandhi. If only the written pinyin has been practiced, 可以 won't be recognized when heard as ké yǐ in the recording. Listening to real speech matters — the Tone Quiz is a good place to start.