How to Choose Reading Material at Your Level
In short
- Count the unknown words on one page. Two or fewer per hundred words means the text is right for fluency practice; five or more means it is study material, not reading material.
- The 98% figure comes from Hu and Nation (2000), who manipulated known-word coverage at 80, 90, 95 and 100% and found most learners reached adequate unassisted comprehension only at around 98%.
- Background knowledge compensates for linguistic difficulty. A text about something you already know is effectively easier than its readability score suggests.
- Graded readers are the most reliable source of level-appropriate text, and the most neglected by adult learners because they feel like a step backwards.
- Re-reading a text you have already understood is one of the cheapest ways to build reading speed, because the second pass is spent recognising words rather than working them out.
Most abandoned reading plans were not a discipline failure. They were a text-selection failure. The learner picked something one or two levels too hard, ground through eight pages, felt stupid, and stopped. The method was fine. The book was wrong.
Choosing correctly takes about thirty seconds and one rule.
The thirty-second test
Open the text at a random page in the middle — not the first page, which is usually written to be inviting. Read a hundred words. Count the words you cannot understand at all, including from context.
| Unknown words per 100 | What it is | What to do with it |
|---|---|---|
| 0-2 | Comfortable reading material | Read it fast, no dictionary. This is where fluency is built. |
| 3-4 | Slightly stretched | Fine, but expect it to be slow. Good for a weekly stretch text. |
| 5-9 | Study material | One page at a time, with a dictionary. Not a book you will finish. |
| 10+ | Too hard | Put it down. You will learn less per hour here than from an easier text. |
The threshold in the top row is not arbitrary. It comes from research on lexical coverage — the proportion of running words in a text that a reader already knows. Marcella Hu and Paul Nation's 2000 study gave learners the same fiction text with known-word coverage set at 80%, 90%, 95% or 100% and measured comprehension. Most learners reached adequate unassisted comprehension at around 98%; at 95% only a minority did; at 80% almost nobody did. Ninety-eight per cent means two unknown words in a hundred: roughly one every two to three lines.
That number is repeated everywhere, so it is worth knowing where it stops. The original regression ran on 66 university students in New Zealand, and a replication in Language Learning has since revisited exactly that limitation. It is the best threshold available and a thin base to rest a whole method on. In practice this does not change what you should do — count the unknown words, aim low — but it should stop you treating 97% as a failure.
That is stricter than almost anyone expects. It is also why "just read a novel, you will pick it up" is bad advice below B2. At five unknown words per hundred, you are not picking anything up; you are decoding, and decoding does not produce reading fluency.
If you want a second opinion on a specific text, paste a page into the readability checker — it returns Flesch-Kincaid, Gunning Fog and an estimated CEFR band, which is a useful cross-check against your own count.
Familiar topics are effectively easier
There is a second variable the word-count test does not capture, and it is large: how much you already know about the subject.
Background knowledge does a lot of the work that vocabulary would otherwise have to do. If you are a nurse reading an article about a clinical trial, you can survive a much higher unknown-word density than the 98% rule suggests, because you can predict what the sentence is going to say. If you are that same nurse reading about cricket, you cannot.
The practical consequence is that "level" is not a property of a text on its own. It is a property of a text and a reader together. Which means:
- Start with what you already know. Your own field, your own hobbies, a book you have read in your first language, a film you have already seen. The comprehension scaffolding is free.
- Do not mistake topic difficulty for language difficulty. If you find an article hard, check whether you would find it hard in your first language too. Plenty of learners conclude their English is worse than it is after failing to understand an article about macroeconomics.
- A translated book you already love is a very strong option. You know the plot, so unknown words resolve themselves from a context you already hold.
Where to actually find text at your level
Ranked by how reliably each source gives you something at a known difficulty, with the honest downside attached.
| Source | Typical level | Strength | Real downside |
|---|---|---|---|
| Graded readers | A1-C1, published per level | Vocabulary controlled against a headword list. You know what you are getting, and Victoria University of Wellington publishes a set free at 4,000, 6,000 and 8,000 words. | Simplified stories. Adults often find them thin. |
| Graded news sites | A2-B2 | Current events rewritten at controlled levels, updated daily. | Short. Not enough volume on their own. |
| Children's and middle-grade fiction | A2-B1 for adult learners | Long, cheap, plentiful, genuinely enjoyable. | Idiom and invented words can spike unexpectedly. |
| Young adult fiction | B1-B2 | Bridges to adult fiction well. Strong narrative pull. | Slang dates fast and is rarely in dictionaries. |
| Simple-language encyclopedias | A2-B1 | Free, factual, huge topic range. | Reference prose. No narrative momentum. |
| Non-fiction in your own field | B1+ if the field is yours | Background knowledge does half the work. | Technical vocabulary is unavoidable. |
| General news and magazines | B2+ | Authentic, current, endlessly available. | Much denser than learners expect. Idiom-heavy. |
| Literary fiction | C1+ | The eventual goal for many people. | Deliberate difficulty. Not a training text. |
| Generated passages at a chosen level | Any | Exact level, exact topic, unlimited supply. | Only as good as the generator. Not a substitute for real books. |
Graded readers deserve a defence, because adult learners skip them almost universally. The objection is that they feel like a demotion. But the alternative to a graded reader at your level is usually a real novel above your level that you will not finish, and forty pages of a hard book teaches you less than four easy books. Day and Bamford put it first among their ten principles: reading material should be easy. Extensive reading programmes commonly target about one graded reader a week; at that rate you will pass through the whole graded range in a year and arrive at unsimplified fiction able to actually read it.
If cost is the obstacle, the graded readers written by Gerry Meister and Paul Nation are out of copyright and free to download at 4,000, 6,000 and 8,000 word levels.
Build a ladder, not a jump
Rather than picking one book, pick a sequence. A ladder that works for most people sitting around B1:
- Two or three graded readers at your current level. Confirm the level is right using the thirty-second test, not the label on the cover.
- The same number one level up. Expect this to feel slower for the first book and normal by the third.
- A middle-grade novel in a series. Series are ideal: the vocabulary, characters and world repeat, so book three is markedly easier than book one even though the level is identical.
- A non-fiction book in your own field. Background knowledge carries you through the vocabulary spike.
- Unsimplified adult fiction, chosen for plot rather than prestige.
At each rung, run the thirty-second test rather than trusting your sense of how it is going. Learners systematically over-estimate what they can read comfortably.
Two techniques that are almost free
Re-read. Once you have understood a text, reading it again costs very little effort and does something the first pass cannot: it lets you spend the whole time recognising words rather than working them out. That is exactly the mechanism behind reading speed, and it is the same principle as a controlled-vocabulary speed reading course — Sonia Millett publishes free ones where every text is 500 standard words so the times are directly comparable. It requires no new material. There is more on the speed side in reading speed in a second language.
Read the same book twice at different distances. First pass extensive: fast, no dictionary, plot only. Second pass intensive on one chapter: look everything up, notice the structures. You get both methods out of one text. The split between them is covered in extensive vs intensive reading.
If you do not know your level yet
You cannot pick text at your level without a level. A free English level test based on reading comprehension gives you a CEFR band in a few minutes, and a reading-based estimate is the relevant one here — a general placement test averages in your speaking, which is usually behind.
Once you have a band, the thirty-second test does the rest. And where finding a genuine supply of level-appropriate text is the bottleneck rather than knowing your level, LingoRead generates passages at a CEFR band you choose, on topics you pick, with comprehension questions and a vocabulary list drawn from the passage — which removes the selection problem entirely for daily practice, while you work through real books alongside it.
Sources
- Hu, M. and Nation, P. (2000). Unknown vocabulary density and reading comprehension. Reading in a Foreign Language, 13(1), 403-430. Source for the 80/90/95/100% coverage conditions and the 98% threshold.
- Kremmel, B. et al. (2023). Unknown vocabulary density and reading comprehension: replicating Hu and Nation (2000). Language Learning. The replication and the source for the caveat about the original sample.
- Day, R. and Bamford, J. (2002). Top ten principles for teaching extensive reading. Reading in a Foreign Language, 14(2). Source for "reading material should be easy" and for learner choice.
- Meister, G. and Nation, I. S. P. Graded readers, Te Herenga Waka - Victoria University of Wellington. Free out-of-copyright readers at 4,000, 6,000 and 8,000 word levels.
- Millett, S. Speed reading and listening fluency, Te Herenga Waka - Victoria University of Wellington. Free 500-word timed reading courses.