Most modern Chinese words are two characters, and the pair is usually readable once both halves are known. Electricity plus speech is a telephone. Electricity plus vision is a television. Fire plus vehicle is a train. That is not a mnemonic, it is how the vocabulary was built, and it means new words are decoded far more often than memorised.
The decoding works on the page and it works much less well in the ear, because the same syllables belong to many different words, and that asymmetry is the real shape of vocabulary in this language.
Six examples of why the vocabulary is cheaper than it looks
- two characters make most words in modern Chinese, so a word is usually a small compound rather than an arbitrary string.
- 电脑 is an electric brain, a computer. Once both characters are known, the word costs nothing to learn and nothing to remember.
- 火车 is a fire vehicle, a train, named for the steam engine and never updated.
- 手机 is a hand machine, a mobile phone, and the same second character appears in a long list of devices.
- 学 gives 学生, 学校 and 学习, student, school and to study. One character learned pays for a cluster.
- compounds are transparent far more often than in English, which is the compensation for the cost of the characters themselves and is why vocabulary growth accelerates sharply after the first year.
Four ways two characters combine
The pairs are not arbitrary and there are a small number of relationships between the halves. Two near synonyms can sit together and mean roughly what each does alone. One character can modify the other, which is the commonest pattern and the most transparent: the second half says what kind of thing it is and the first narrows it. A verb and its object can be welded into one word. And a verb and a result can combine, so the second character says what happened to the action.
Knowing which pattern you are looking at is what makes a guess reliable. Modifier and head is the safest, so a two character word whose second half is a familiar noun is usually exactly what it looks like. Verb and result is the one that surprises people, because the compound can mean something the halves do not obviously add up to, and it is a large class.
The residue is real and it is worth stating. Some compounds are opaque, some are borrowings written for their sound rather than their meaning, and a number of very common words are single characters whose sense has drifted a long way. Those opaque compounds are the items that genuinely need a card, and they are a minority of the vocabulary.
You can read a word and not hear it
Here is the stall and it is caused by the writing system doing its job. Mandarin has a small inventory of possible syllables, so an enormous number of words sound identical or nearly identical, and the characters keep them apart on the page without any work at all. In speech nothing keeps them apart except the context and the other syllable in the pair.
So a learner who has built vocabulary by reading has built exactly the half the writing supports. They meet a word in conversation, fail to recognise it, and discover afterwards that it is a word they have known for months. More reading does not fix this, because reading is what produced the imbalance, and the repair is to meet the same words in audio deliberately: a card with sound on the front rather than a character, or a graded audio text where you already know the vocabulary. The general form of that argument, that two directions of the same word are two separate acquisitions, is in why you forget, and the schedule that beats forgetting.
From a character to a word you can use
Where the vocabulary is lost between the page and the ear
Each band removes a share of what the one above it contained, and the losses are not where a reader expects them.
- Characters you recogniseThe count that grows fastest and is easiest to measure
- Compounds you can decodeMost of them, if both halves are known
- Compounds you have actually metWhich pair Chinese uses is a fact about the vocabulary
- Words you recognise in speechWhere the syllable inventory is small and the homophones arrive
- Words you produce with the right tonesThe narrowest band, and the one nobody measures
There are no numbers on the bands because nobody has counted your vocabulary. The claim is the order of the losses: a learner reporting a vocabulary size is describing the second band and using it in the fourth.
Two things that decide what to learn and in what order
The published HSK word lists are free and they are the closest thing to a frequency ordering that anybody agrees on, which matters because employers and universities read the same scale. The standard was revised in 2021, so check which version a list is built against before working through it. It replaced the thematic vocabulary list, which is ordered by topic and has no relationship to what turns up in a sentence.
Mandarin Companion is the other half: graded readers written inside a stated character count, so a whole book can be read at a density where almost nothing needs looking up. It replaced trying to read native material too early, which in Chinese fails harder than in other languages, since an unknown character cannot be sounded out and asked about the way an unknown word can in an alphabetic language. Paid, and cheap relative to what it does, which is to make reading a way of making vocabulary automatic rather than a way of acquiring it.
The mistake: learning characters and expecting words
A learner who has worked hard on characters can know both halves of a compound and not know the compound, and conclude that their character study has failed. It has not. Knowing the parts makes the word cheap to learn and does not make it already learned, because which pairs exist is a fact about the vocabulary rather than about the characters.
The useful version of the rule is that characters make words guessable and words still have to be met. A compound guessed correctly from its halves is not yet a word you can produce, since you do not know that this particular pair is the one Chinese uses for that idea; there are usually several possible pairings and only one of them is current. So the character work buys speed of acquisition rather than acquisition, which is a large return and a different one from the one people expect.
What the words still need
Every one of these words carries a tone on each syllable, and nothing about the characters or the compounding will remind you of it, which is four of them, and what happens when you get one wrong. Countable nouns also carry a classifier that has to be stored beside them, in word order carries everything, because nothing else changes. The halves being combined are themselves built from components, and learning those first is what makes any of this cheap, in radicals, components, and how to remember a few thousand. And the asymmetry between reading a word and catching it is the whole subject of hearing tone at conversational speed.