Knowing a Chinese word and being able to read it are two separate achievements, and neither one delivers the other. The writing does not record the sound, so the route from a spoken word to its character and the route from a character to its sound are different pieces of memory, built by different practice, and each has to be maintained on its own.
That is the real shape of the difficulty, and it is a shape rather than a height. Nothing here is intellectually hard. There is simply a second memory system running alongside the first, and a learner who does not know that is building one of the two and assuming the other will follow.
Three facts that decide how Chinese is learned
- there is no conjugation at all. No tenses, no plurals, no agreement, no cases. A verb has exactly one form for ever, and time is expressed by separate words. The grammar is genuinely easier than any European language on this list.
- four tones and a neutral one. mā is mother, má is hemp, mǎ is horse, mà is to scold. The tone belongs to the word, and this is where the difficulty was moved to.
- characters are not an alphabet. Each one is a syllable with a meaning, and reading requires knowing them individually. This is the other place the difficulty went, and it is why speaking runs far ahead of reading here.
A character is a small number of reused parts
Characters look like an unbounded set and are not. Most of them are built from a meaning component and a sound component, and both components are themselves characters you already have. The character for mother is the woman component beside the character for horse, which contributes nothing to the meaning and almost all of the pronunciation: horse is ma, and so is mother.
That same horse component sits inside the question particle and inside the verb to scold, and all three are said ma. Which is the point: the sound component is unreliable and it is not useless. It narrows several thousand possibilities to a handful, which is the difference between recall and a lottery. What follows is a practical rule that saves months: learn components rather than strokes, and learn them before the characters that contain them, so that every new character is two known things in an arrangement rather than fourteen lines in an order. Radicals, components, and how to remember a few thousand is that work, and it is the largest single block of time in this subject.
The same logic runs one level up. Most modern words are two characters, and the pair is usually transparent once both halves are known, so two character words, and guessing the meaning of a new one becomes a decoding skill rather than a memory load, exactly as the components did below it.
The tones survive the flashcard and die in the sentence
This is the stall, and it arrives late enough to be demoralising. A learner drills the four tones on isolated syllables, scores well, and is then not understood in a shop. The drill was not wrong. It was measuring a different thing.
The four tones on that one syllable are the standard demonstration, and they are four real words: mother, hemp, horse and to scold. Said slowly in a quiet room, all four are easy. Three mechanisms then take them apart. A tone is a shape in your own pitch range rather than an absolute note, so it moves with the sentence and with how loudly you are talking. Tones change each other: two third tones in a row turn the first into a second tone, and the common words bu and yi shift depending on what follows them. And connected speech compresses the contours, with the second syllable of many words losing its tone entirely.
So tone is a property of the sentence and not of the syllable, which means it has to be practised in sentences from the first week. Four of them, and what happens when you get one wrong covers the failures, and hearing tone at conversational speed is the receiving end of the same problem, which usually lags production by a year.
Pinyin deserves a warning of its own here. It is a precise system and it was not designed for English speakers, so several letters do not say what an English reader expects: q, x, c and zh are the usual casualties, and a learner who reads pinyin as though it were English installs four wrong sounds in the first week. The Roman spelling, and where it misleads English readers is a half hour that prevents a year of correction.
One syllable
Four words, and the only difference is the shape of the pitch
A tone is not decoration applied to a syllable. It is part of which word you just said, and each of these is defined against the other three.
- Levelma held high and flat, and it is mother
- Risingma rising the way an English question rises, and it is hemp
- Dippingma falling and then coming back up, and it is horse
- Fallingma dropped sharply from high to low, and it is to scold
All four are easy said slowly in a quiet room. Not one of these shapes reaches a sentence unchanged, which is why the drill and the shop disagree.
No conjugation is not no grammar
Nothing changes shape. Verbs do not conjugate, nouns do not take plurals, adjectives do not agree, and none of that means the information disappeared. It moved into word order and into a small set of particles, and those particles are where the actual difficulty is. The particle le alone carries completion and change of state and sits in two different positions with two different meanings, and it is the single most argued about item in the language.
Measure words are the other surprise: you cannot count a noun without the right classifier between the number and the noun, and the choice is partly logical and partly historical. Word order carries everything, because nothing else changes is where those get their proper treatment, and the honest summary is that Chinese moved the difficulty rather than removing it.
One dictionary that everything else is built on
Pleco is the dictionary, and in 2026 it still has no real competitor. Look up by pinyin, by handwriting, by camera, or by drawing a component you half remember, and it will give you the character, the stroke order, audio and the words that contain it. It replaced the paper dictionary and its radical index, which required knowing the radical and counting the remaining strokes before you could begin looking.
Underneath Pleco and most of its competitors sits CC-CEDICT, a free and openly licensed Chinese to English dictionary that volunteers have maintained since the 1990s. It is worth knowing that it is there, through a front end such as MDBG, because it means the core data is not owned by any app you might have to leave.
Anki matters here for a Chinese specific reason. Because the two memories are separate, a card that shows a character and accepts a meaning tests only one of them, so build the reverse direction as its own card and expect it to be harder for a long time. For the certificate, the official scale, and the vocabulary each band requires is the thing employers and universities read; the standard was revised in 2021, so check which version your test is being set against before buying a word list.
Whether the characters or the tones brought you here
The question at the end of the first year is which of the two systems you actually want, because they lead to different places. Three writing systems, one grammar, and a route through both reuses these characters and gives them several readings each, so the character work transfers and the sound work does not. An alphabet designed to be learnt in a morning drops the characters entirely and keeps a layer of vocabulary borrowed from Chinese, which is a discount you can only spend after you have paid here. Both live in every tongue this atlas speaks, and the method, which is the same whichever language you picked has the scheduling argument that a few thousand characters makes unavoidable.