Tone Numbers or Tone Marks: ma3 and mǎ
Use numbers for typing and marks for reading. ma3 survives keyboards, filenames, URLs, database columns and forum posts; mǎ sits on the vowel and is read at a glance, which is why every textbook uses it. Pleco is the everyday proof - it takes numbers in the search box and gives marks back in the results. There is also a third convention, and it is the one that catches people out.
The same syllable, written two ways
马 is mǎ in a textbook and ma3 in a forum post, and the two mean exactly the same thing: this syllable carries the third tone. Nobody ever explains why both exist, so learners tend to pick one at random, feel vaguely unserious about it, and stick with it everywhere. The honest answer is that neither is better - they solve different problems, and the reason to know both is that each fails badly in the other one's territory.
Numbers are a transport format. Marks are a reading format. Almost everything about the choice follows from that one distinction, and once you have it the rest is a matter of knowing which situation you are in.
Numbers are for anywhere a diacritic breaks
A tone mark is a non-ASCII character sitting on top of a letter, and there is a long list of places where that is a liability rather than a feature:
- Typing. On a plain keyboard, ma3 takes three keystrokes and mǎ takes a detour through a character picker. Every Chinese input method accepts the numbered form, and most accept it silently as an ordering hint.
- Filenames and URLs. A filename containing ǎ can be stored as two different byte sequences depending on which operating system wrote it - the letter and the mark combined into one character, or held as two - so the same name compares unequal across machines. Numbered pinyin is plain ASCII and has none of this.
- Databases and search. Sorting, indexing and matching all get simpler when the tone is a digit. It is also what your users will actually type: very few people put marks into a search box.
- Forums, chat and code. Anywhere the person on the other end may not have an input method configured, numbers are the format that arrives intact.
This is not a fringe convention. CC-CEDICT, the dictionary file underneath a great many Chinese tools, stores its pinyin with numbers. Pleco takes ni3hao3 in the search box quite happily. If you have ever wondered why the notation you type is not the notation you are shown, this is why: the machine wants the digit and the reader wants the mark.
Marks are for reading, and the reason is where they sit
Marks win on the page for two reasons that are easy to demonstrate on yourself. The first is position: the mark is on the vowel, inside the syllable, so the eye takes the syllable and its tone in one movement. A number sits after the syllable and is read separately - hao, then 3, then a small act of recall to turn 3 into something. Across a page of text that is thousands of extra glances.
The second is that the marks are iconic and the numbers are arbitrary. The macron is flat and the tone is flat. The acute rises and the tone rises. The caron dips and comes back up; the grave falls. Nobody has to memorise that mapping - it is visible. The digit 3 tells you nothing at all until you have separately learned that 3 means low and dipping, which is one more link in a chain you are trying to make automatic.
This is why every textbook, every dictionary entry meant for human eyes, and the official pinyin standard itself use marks. If something is going to be read rather than parsed, it gets marks. Where the marks go on the vowel is its own small rulebook, covered in the post on tone marks and placement.
The numbers do not agree with each other
Marks have one convention. Numbers have several, and they collide quietly.
The neutral tone is the common one. Some sources write it 5, some write it 0, and some leave the digit off entirely, so 妈妈 is ma1ma5, ma1ma0 or ma1ma depending on where you found it. All three are standard somewhere. If you are building a deck or a dataset, pick one and be consistent, because a mixture is unsearchable.
The sharper collision is with linguistics. Phoneticians also number tones, but their numbers describe the pitch contour on a five-point scale rather than naming a category - 1 is the bottom of the speaker's range and 5 the top. In that system Mandarin's first tone is 55, the second is 35, the third is 214 and the fourth is 51.
| Tone | Mark | Learner number | Contour number |
|---|---|---|---|
| First | mā | ma1 | ma55 |
| Second | má | ma2 | ma35 |
| Third | mǎ | ma3 | ma214 |
| Fourth | mà | ma4 | ma51 |
So a single digit names a category and a string of digits draws a shape, and you will meet both under the same word, "tone number". Notice also which column carries more information: only the contour numbers say anything about the sound. ma3 and mǎ are two spellings of the same label. ma214 is a small claim about what the pitch actually does, and it is the only one of the three that could be wrong.
One last quirk in the same family: since ü is awkward to type, numbered pinyin conventionally uses v for it, so 女 is nv3. v is unused elsewhere in pinyin, so the letter was free - the same reasoning that produced several of the consonant assignments in why pinyin is not pronunciation.
The third convention: as written, or as spoken
Here is the one that almost nobody names, and it matters more than the choice between a digit and a diacritic. Before you can write a tone down at all, you have to decide which question you are answering: what tone does this syllable have, or what tone did the speaker just produce?
For most words the two answers agree. For a large class of words they do not. 你好 is two third-tone syllables in the dictionary and is never said that way - a third tone before another third tone comes out as a second, so the word is spoken ní hǎo. Written as numbers, the same word is ni3 hao3 under one convention and ni2 hao3 under the other. Neither is a mistake. They are answers to different questions.
| Word | As written | As spoken | Meaning |
|---|---|---|---|
| 你好 | ni3 hao3 | ni2 hao3 | hello |
| 水果 | shui3 guo3 | shui2 guo3 | fruit |
| 一半 | yi1 ban4 | yi2 ban4 | half |
| 不错 | bu4 cuo4 | bu2 cuo4 | not bad |
Dictionaries answer the first question, because a dictionary is telling you what a word is. Tingyin answers the second, because it is telling you what a clip contains. Its corpus is 637 clips and every one is labelled with the tone as spoken, which produces a fact that reads as a bug until you see why: across all 637 clips there is not one tagged third-then-third. Not because such words are missing - they are in there - but because none of them is ever said that way. Sandhi is applied in 55 clips and 21 contain 一, which shifts its tone according to what follows it.
For a listening trainer this is the only defensible choice. If a clip of 你好 were labelled 3+3, a learner who correctly heard a rising first syllable would be marked wrong for hearing what is there. The full set of rules behind this is in the tone sandhi post, which is worth reading before you conclude that a dictionary and your ears are in disagreement.
A rule of thumb you can apply without thinking
| Where | Use | Why |
|---|---|---|
| Reading, notes, textbooks | Marks | Read in one glance, and the shape is the meaning |
| Typing into an input method | Numbers | Plain keys, no character picker |
| Filenames, URLs, spreadsheets | Numbers | ASCII survives every system unchanged |
| Forums, chat, code comments | Numbers | Arrives intact whatever the reader has installed |
| Anki and flashcard decks | Both | Numbers in the sortable field, marks on the card face |
And whichever you use, be explicit about the third convention when it could matter. A deck of two-syllable words labelled as written will quietly train you to expect a sound that never occurs; a deck labelled as spoken will disagree with your dictionary in a way that looks like an error. Both are fine as long as you know which one you are holding.
None of this notation, in any of its forms, tells you what a tone sounds like - it is bookkeeping for a category, and the category has to be filled by ear. That is the whole argument of why tones are hard to hear rather than hard to say, and it is what Tingyin exists to drill: a clip plays, you pick the tone, it tells you at once. Level 1 is 80 single-syllable minimal pairs and is free with no account at all; level 2 adds 200 more and is free once you sign in. Levels 3 and 4 are paid and move to real words - 237 two-syllable, 107 three-syllable and 13 four-syllable clips - with progress sync and offline downloads, at $2.49 weekly, $5.99 monthly or $15.99 once for lifetime access on the web. Every clip is a human recording, and each one is labelled by what is actually spoken rather than by what the written form would predict - which for a notation argument is the distinction that matters. You can start the drill before deciding whether any of that is worth paying for.
Frequently asked questions
What do the numbers in Mandarin tone numbers mean?
A single digit after a syllable names the tone: 1 high level, 2 rising, 3 low dipping, 4 falling, and either 5, 0 or nothing at all for the neutral tone, depending on whose convention you are reading. So ma3 is the same syllable as mǎ. Watch for the other system: in linguistics a string like 214 describes the pitch contour on a five-point scale rather than naming a category.
Should I learn pinyin with tone marks or tone numbers?
Learn to read marks, because everything written for human eyes uses them and the shapes carry the meaning. Learn to type numbers, because keyboards, filenames and search boxes all prefer plain ASCII. It takes an afternoon to become fluent in both and you will use each of them daily. Pleco demonstrates the split neatly: you type ni3hao3 into it and it shows you nǐ hǎo.
Why does the same word have different tone numbers in different places?
Usually because one source is writing the tone as written and another as spoken. 你好 is ni3 hao3 in a dictionary and ni2 hao3 in a transcription of somebody saying it, since a third tone before another third tone is pronounced as a second. Tingyin labels all 637 of its clips as spoken, which is why none of them is tagged third-then-third. Neither convention is wrong; they answer different questions.
How do I type ü and tone marks on a normal keyboard?
For ü, type v - it is unused elsewhere in pinyin, so nv3 gives you nǚ in most input methods and dictionary search boxes. For the marks themselves, most people do not type them directly at all: they type numbered pinyin and let the software render the marks, which is exactly the division of labour the two notations exist for.
Does knowing the tone number help me hear the tone?
Not on its own. A number is a label on a category, and the category is filled by listening to a few hundred examples of it, not by learning the label. Tingyin drills exactly that identification step and nothing else - it teaches no characters, no grammar and no vocabulary, and it cannot hear you speak. For feedback on your own production you need a person, which is what italki and a tutor are for.