A Guide to Graded Reading

Why "level" works differently in Chinese

In a language written with an alphabet, the obstacle to reading is mostly vocabulary and syntax: once a reader can decode the script, the unknown words are the problem, and a graded reader is levelled by controlling how many of those appear. Chinese adds a second barrier in front of the first one. A reader can hold a large spoken vocabulary and still be unable to recognise the characters that represent it, which means a text can be completely understandable when heard and completely opaque when read. Level, in Chinese, is therefore at least two different things layered on top of each other: how well the reader recognises the characters, and how well the reader can handle the sentence structures once the characters are known.

Because those two are independent, a single number cannot describe a Chinese reader. Someone who grew up speaking Chinese at home may understand a complex argument when it is spoken but be blocked by ordinary characters on the page. An adult learner who has studied characters systematically may read a news paragraph with few unknowns and still miss what it means, because the syntax and the abstract vocabulary defeat them. Diagnosis has to separate the two, and the sections below do that deliberately. The first half of this page is about measuring; the second half is about choosing material and structuring the practice.

What actually makes a Chinese text hard

Four variables account for nearly all of the difficulty in ordinary reading material. They interact, but they can be judged separately, and knowing which one is causing the problem tells you what to change.

1. Known-character coverage

The single most important variable. Coverage is the proportion of characters on the page that the reader recognises without help. It is not the same as vocabulary, because many characters appear inside words whose meaning the reader could not derive, and it is not the same as comprehension, but it is the variable that determines whether reading is possible at all. Two thousand well-chosen characters cover the overwhelming majority of the characters in an ordinary text, which is why a reader at that level can work through a newspaper article slowly and a reader at eight hundred cannot begin.

2. Unknown-character density

The counting is right and the distribution is wrong — that is what density means, and it matters more than the total. Twenty unknown characters scattered evenly across a page of two thousand is manageable, because there is always context on both sides of each gap. Twenty unknown characters packed into three consecutive sentences stops the reading dead, because the reader loses the thread before the context can help. When you assess a passage, look for clusters, not just totals. A passage with a high overall figure but no clusters is easier than it looks; a passage with a low overall figure and one paragraph of unknowns is harder.

3. Sentence length and clause structure

This is the variable that experienced readers systematically underestimate, because it is invisible when your syntax is good. A forty-character sentence in Chinese can contain four nested clauses and dependent relationships marked by nothing more than word order. Reading it requires holding the whole structure in memory while the final particle resolves it. A beginner who knows every character in that sentence may still be unable to say what it meant. When a reader can read individual sentences aloud fluently but cannot summarise the paragraph, the problem is clause structure, not coverage, and the fix is shorter text rather than easier text.

4. Genre and abstraction

Narrative is the easiest genre for everyone, because events are ordered in time and the actors are concrete. Description is next. Exposition — explanation, instruction, argument — is harder than both, because it depends on connectives and abstract nouns that carry no image. News is a special case with its own difficulty: a single news paragraph may mix informal event reporting with formal written vocabulary, proper nouns, numbers, official titles and quoted speech, and the register shifts several times within the same sentence. For a reader who can handle a short story comfortably, a news item of the same length can be three times as hard.

The five-finger test, adapted for Chinese

The five-finger test in its original form works on the assumption that English text is measured in words. Chinese has no spaces, so the test has to be rebuilt around characters, and one further adaptation is needed for readers whose spoken Chinese is better than their reading.

The character version

1. Choose a passage of one hundred characters from the middle of the text, not the beginning — openings are usually easier than the body.
2. Read it aloud, or silently if you are assessing yourself, and mark every character you cannot read or whose reading you are unsure of.
3. Count the marks. Up to five unknown characters in a hundred is comfortable independent reading. Six to ten is a text you can read with some support. Eleven or more is decoding practice rather than reading, and you should either choose something easier or prepare the text with annotation first.

The threshold numbers are guideline rather than law, and the honest signal is how the reading feels: if you can finish the hundred characters and say what happened in them, the level is right regardless of the count.

The ear test, for readers with strong spoken Chinese

If the reader understands spoken Chinese well — the usual situation for a heritage-language learner — the character test alone will mislead you. Run a second test alongside it. Read the passage aloud, or play it, and ask the reader to say what it meant. If they understand the passage when it is spoken but cannot read it, the text is at the right level for comprehension and the gap is purely character recognition; that is a case for annotated text, described below, rather than for easier material. Conversely, if they cannot understand it even when it is spoken, the material is genuinely too hard and no amount of annotation will fix it. Running both tests takes two minutes and prevents the most common mistake in heritage-language reading practice, which is holding a reader back at picture-book level when their comprehension is years ahead of their character knowledge.

Annotated or unannotated text?

Annotation is not a level; it is a support that can be applied to any level. That distinction is worth stating clearly because it changes how you choose. Annotated text lets a reader work at a level above their unassisted level, because the pinyin resolves the character-recognition barrier while leaving the comprehension work intact. This makes it the right tool in two situations and the wrong tool in two others.

• Use annotation when the reader's comprehension is ahead of their character recognition. A heritage-language reader, or an adult learner who has studied a lot by ear, can read annotated material far above their unassisted level, and the reading is genuinely productive rather than a crutch.
• Use annotation for first encounters with new material of any kind. A first pass through an unfamiliar text is not the moment to be testing character knowledge. Annotated, it is a first pass; unannotated, it is an examination.
• Do not use annotation to make a genuinely difficult text easy. If the syntax defeats the reader, pinyin does not help; it only changes an incomprehensible text into a comprehensible-looking one, which is worse, because the reader cannot tell that they have not understood.
• Do not use annotation as the permanent form of the material. A text read only ever in annotated form becomes a text that can only be read in annotated form. The support has to be withdrawn, which is the subject of the next section.

When to take the pinyin off the page

The decision rule is not about time but about attention: remove the support when the reader is no longer looking at it. Three passes over the same text, spaced a few days apart, do this reliably.

1. Pass one, annotated at full size. Read the text with the pinyin at roughly half the character height. Use the pinyin whenever a character fails. Note how many characters you needed it for — that number is the baseline and the most useful measurement you will take.
2. Pass two, annotated but faded. Reprint the same text with the pinyin shrunk and its colour dimmed. The readings are still there, but they cannot be read without effort, so the eye tries the character first. If you find yourself reading the grey pinyin instead of the characters, that is the signal to fade it further, not to go back to full size.
3. Pass three, plain text. A week or so later, read the source text with no annotation at all. Whatever you cannot read on this pass is the genuine residue, and the number should be much lower than the baseline.

Three signals tell you it is time to move to plain text even without waiting for a third pass. If the number of characters you needed pinyin for on pass one was under five per hundred, the text was already within reach. If a reader stops noticing the pinyin at all — reading a whole page and only realising afterwards that they never looked up — the support has become decoration. And if the reader is a child who has re-read the same book several times and is reciting the story from memory, the characters are being carried by the memory rather than by the page, which is the moment to reintroduce the same story in a harder form: same content, less support, larger vocabulary.

Choosing material at each stage

Rather than asking what level a particular series claims to be, ask what each kind of material demands of the reader and how much support it can carry.

• Picture books. Very few characters per page, concrete nouns and verbs, heavy illustration that supplies the context for any unknown characters. The natural first material, and the easiest to annotate because the text is short. Their limitation is the inverse of their strength: the vocabulary is narrow, so a reader can stay at picture-book level indefinitely without widening it.
• Graded readers and simplified retellings. Text controlled for coverage, usually around a few hundred to a thousand distinct characters, with short sentences and a limited grammar. The best material for building genuine reading volume at a known level, and the most predictable to plot progress against. Two cautions: the simplification is sometimes achieved by removing the connectives that make sentences flow, so the text can be choppy; and a retelling of a story the reader already knows in another language reduces the comprehension work to zero, which means it exercises character recognition only.
• Short stories and essays aimed at general readers. The step after graded readers, and the largest jump in the list. Sentences lengthen, register opens up, and vocabulary stops being predictable. The right approach is to read a short story annotated for the first pass and unannotated for the second, rather than reading new short stories unannotated and abandoning most of them.
• News and current-affairs writing. The hardest common material, as described above: mixed register, proper nouns, numbers, quoted speech. Worth reaching for once a short story is comfortable, because it is the material that appears in daily life. Choose a single short item rather than a page, annotate it, and read it twice.
• Social and functional text. Messages, listings, product descriptions, recipes, menus. Frequently overlooked in reading programmes and genuinely useful, because the register is closer to speech, the vocabulary is high-frequency, and there is an obvious purpose that carries the reader past the hard parts.

A sample weekly reading plan

The shape below assumes roughly fifteen to twenty minutes a day and works for an adult learner or for a parent running a session with a child. The principle behind it is that re-reading is where consolidation happens, so every text is met three times before it is retired; four texts a week is the ceiling, and two is enough to make progress.

1. Monday — new text, annotated, read aloud. Choose something at ninety-five per cent coverage by the character test. Read it once at full pinyin size and mark the characters you needed help with.
2. Tuesday — questions and vocabulary. Answer two questions about the text without looking at it first: one literal, one inferential. Write five items of vocabulary as they appeared in the text, each with its pinyin and the sentence it came from.
3. Wednesday — second pass, faded pinyin. Same text, pinyin shrunk and greyed. Read aloud again and count how many characters you now read without looking up.
4. Thursday — new short text, annotated. Keep it short — a message, a paragraph, a recipe. Annotate, read once, stop.
5. Friday — third pass on Monday's text, plain. Read the unannotated source. Anything unreadable on this pass goes into a running list, and that list, not a textbook, is what you review from.
6. Weekend — free reading. Something chosen purely for interest, annotated or not depending on how it feels. No questions, no vocabulary list, no counting. The point of this slot is to keep the association between Chinese and voluntary reading.

Keep a one-line log per text: the date, the count of characters needing pinyin on the first pass, and the count on the last. The trend across a month is a far better measure of progress than any level label, and when the third-pass count stays near zero for several texts in a row, raise the difficulty rather than the volume.

Making your own levelled material with the tool

The practical difficulty with levelled reading in Chinese is that published material is scarce and rarely matched to one particular reader. The solution is to level text yourself, and the online pinyin annotation tool is what makes that feasible. You supply any text — a news item, a page of a story, a message from a relative, something you wrote — and the tool returns the same text with pinyin above every character, at a pinyin size and colour you control, ready to print or export as a PDF or Word file. That gives you the first and second passes of the three-pass cycle described above from one source text: annotate at full size for pass one, annotate again with the pinyin shrunk and dimmed for pass two, and print the untouched source for pass three. Characters with more than one reading are flagged rather than silently resolved, so the small number of characters that would otherwise teach a wrong reading can be corrected in a click before you print — worth doing every time, since a levelled sheet reused across a term teaches whatever is on it, right or wrong.

For the routine that these sheets fit into, see reading Chinese with kids, which covers the same cycle at each age and stage, and making your first bilingual book, which applies the technique to a complete book. For a step-by-step walkthrough of building a reading session around one passage, continue to graded Chinese reading practice; for the characters worth the deliberate effort at each level, see 500 essential Chinese characters. The rest of the series is indexed in the Tutorials.