How to actually learn English vocabulary by reading
Words are picked up from reading, but only under conditions most people never meet: a text you mostly understand, and enough repeat encounters for a word to survive. Here is how to arrange both.
"Just read more" is the most common piece of vocabulary advice in English, and it is true in the same way that "just eat less" is true: correct as a direction, useless as an instruction. Plenty of people read English every day for years and their vocabulary barely moves. The reason is that incidental vocabulary acquisition — picking up words as a by-product of reading for meaning — only works under conditions that ordinary reading rarely supplies.
How a word is actually picked up
Learning a word from reading is not one event. The first time you meet an unknown word in context you usually extract something vague and partly wrong: a rough sense of whether it is positive or negative, whether it is a thing or an action, roughly what area of life it belongs to. That fragment is fragile and, on its own, will be gone in a day.
The second encounter, in a different sentence, does something the first could not: it contradicts or narrows your guess. By the fourth or fifth time in genuinely different contexts you have collocations, register, grammar, the shape of the word in use. Nobody can tell you the exact number of encounters required, and the honest research position is that it varies enormously by word, by learner and by how much the context gave away — but everyone agrees that one encounter is not enough and that the encounters must be spread out.
This is the fact that quietly disqualifies most reading. In a random article, a given unfamiliar word appears once. You look it up, you feel like you have learned something, and by the time it turns up again a month later you have no memory of the first meeting. The problem is not your memory. It is the interval.
The 95–98% rule of thumb
There is a figure that circulates in reading research: to read a text comfortably and pick up words from it, you need to already know something like 95 to 98 percent of the running words on the page. Below that, comprehension drops away and you spend your attention decoding instead of reading.
Treat it as a rule of thumb rather than a constant. The exact figure depends on the study, the text, the reader’s background knowledge and, crucially, how comprehension was measured — different research groups have landed on somewhat different thresholds, and the number quoted at you is usually a rounded consensus rather than a single result. What is robust is the shape of the finding: the coverage you need is very high, much higher than most learners assume, and the difference between 90 and 98 percent is the difference between work and reading.
Put it in concrete terms. At 95 percent coverage there is roughly one unknown word in every twenty — about one per line of a paperback. That is already demanding. At 90 percent it is one in ten, and comprehension collapses: you cannot use context to infer a word when the context itself is full of holes. That is why the "read a novel with a dictionary" plan defeats so many people. It is not lack of discipline. The method has a coverage requirement, and the novel does not meet it.
What to do when the text is too hard
You have four honest options, and only one of them is bad.
- Read something easier. Graded readers, children’s and young-adult fiction, simplified news, and writing on subjects you already know well all raise your effective coverage. This is the option people resist because it feels like a demotion, and it is the one that works.
- Read the same text twice. The second pass through a difficult text has much higher coverage than the first, since the first pass paid for the vocabulary.
- Pre-learn the load-bearing words. If a chapter turns on fifteen unknown words, deliberately learning those fifteen first turns an impossible text into a possible one. This is where list work earns its keep.
- Push through anyway, looking up everything. This is the bad option. It is slow, it does not produce durable vocabulary, and it reliably ends with the book abandoned around page forty.
A note on that resistance: people avoid easier material because it feels like it is not teaching them anything. It is teaching them the thing hardest to acquire — fluency of recognition, the automatic reading of words you already half-know. That automaticity is what frees up the attention you then spend on the genuinely new word.
How to make encounters repeat
Since the mechanism needs multiple spaced meetings and ordinary reading does not supply them, the practical question is how to manufacture them. A few things work.
- Read a lot within one subject or one author. Vocabulary clusters by topic, so five articles about climate policy recycle the same words in a way five random articles never will.
- Keep a short list of the words you actually met, and re-meet them deliberately rather than waiting for chance.
- Say the word aloud when you look it up. It costs a second and adds a second encoding route.
- Write one sentence with it, the same day. Producing a word is a much stronger encounter than recognising it.
- Re-read. Repeat reading is underrated precisely because it feels like it is not progress.
This repeat-encounter problem is what Lexi is built around. Every sentence it gives you contains exactly one word you have not met, which pins your coverage where the reading research says it needs to be instead of leaving it to whatever you happened to open. The sentence library ships inside the app, so this works with no signal at all, and the word lists — Everyday, Phrasal Verbs & Set Expressions, Gaokao, CET-4, CET-6, Postgraduate, IELTS, TOEFL, GRE and Business — let you keep the material inside the domain you actually need. At the end of a session it also writes a short English piece built around the words you met that day, which is the cheapest second encounter there is: the same words again, in sentences you have not seen, on the same day.
None of this replaces reading real English. The point of controlled sentences is to get your coverage high enough that real English stops being a translation exercise, and then to go and read some.
Frequently asked questions
How many times do I need to see a word before I remember it?
More than once, spread out, and nobody can give you a reliable single number. Estimates in the research vary widely because it depends on the word, how much the context gave away, how similar it is to something you already know and how far apart the encounters were. The useful takeaway is structural: one look-up is almost never enough, and encounters spread over days beat the same number of encounters in one sitting.
What percentage of words do I need to know to read a book in English?
The rule of thumb from reading research is 95 to 98 percent of the running words for comfortable reading and for picking up new vocabulary along the way. The exact number moves with the study and with how comprehension was measured, so treat it as a range rather than a threshold. Below about 90 percent, context stops being reliable enough to infer meaning from and reading turns into decoding.
Is it better to read or to use flashcards?
Both, for different jobs. Reading is where depth comes from — collocation, register, the grammar a word takes — and it is the only realistic route to a large vocabulary. Deliberate practice is far faster for getting a first, rough meaning attached to a specific set of words, which is exactly what you need before an exam or before a difficult text. Using one and refusing the other is the common mistake.
Should I look up every unknown word while reading?
No. Looking up everything destroys the reading flow that makes the words stick in the first place, and it is the single most common reason people abandon an English book. Read the sentence, guess, and look up only the words that block comprehension or that keep coming back. If more than a couple of words per paragraph are blocking you, the honest fix is an easier text.
Does reading subtitles or social media count?
It counts, but it is thin fuel. Both are heavily repetitive within a narrow band of everyday vocabulary, so they build fluency in words you already have and introduce very few new ones. They are not harmful — any comprehensible English is better than none — but if your vocabulary has plateaued despite hours of English every day, this is usually the reason.
Try the method for ten minutes
Lexi hands you one short English sentence at a time in which exactly one word is new to you: tap it for IPA and meaning, tap again for the translation, and hear either read aloud. The library is on your phone, so it works with no signal, and there is no account to create. Thirty sentences a day, every word list and every dictionary language are free.