How to Learn Vocabulary From Books (Without Ruining the Read)

Everyone is told to read more in their target language, and it is good advice. What nobody mentions is how few words actually stick from reading alone, and how small a change it takes to fix that without turning every novel into homework.

Reading in your target language is the best source of vocabulary you have. Words arrive in context, with the grammar around them and a situation attached, which is exactly the material a dictionary entry strips away. The problem is not the input. The problem is what happens after you close the book.

Read a novel over three weeks, meet four hundred words you did not know, and a fortnight later you will recognise a handful of them. That is not a failure of effort or memory. It is what happens by default, and once you know why, it takes about ninety seconds a chapter to change.

Why reading alone is a slow way to learn words

The technical term is incidental vocabulary learning: picking words up as a side effect of reading for meaning. It works, and it is how most of your first-language vocabulary arrived. It is also slow. Nagy, Herman and Anderson found in the mid-1980s that the chance of learning a given word from a single encounter in context is small, in the range of a few per cent. The reason it built such a large vocabulary in your native language is that you were reading for years, not that any one meeting with a word was efficient.

Rob Waring and Misako Takaki showed the same effect in a second language. Learners read a graded reader containing target words, and while a fair number were recognised right after reading, only a few survived three months later. Meeting a word once, understanding the sentence, and moving on leaves a trace that fades quickly.

So the honest picture is this: reading is unbeatable for meeting words, seeing how they behave and building the sense of a language that no deck gives you. It is a poor delivery mechanism for retention on its own. Treat it as the source, not the whole method.

Pick a book you can almost read

Everything downstream depends on this. Paul Nation's work on text coverage puts the threshold for comfortable unassisted reading of fiction at around 98 per cent of the running words known, which comes out at roughly two unknown words per hundred, or two or three per page. Below about 95 per cent, comprehension falls off badly and you are decoding rather than reading.

The practical test takes a minute in a bookshop or a sample chapter: read one full page and count what stops you. Two or three unknown words is the sweet spot. Ten and you have picked the wrong book, no matter how much you want to read that particular one. Put it back on the shelf for six months.

This is the same logic as comprehensible input and it is why graded readers exist. If unsimplified novels are still out of reach, a graded reader at your level is not a lesser choice; it is the choice that produces reading rather than translating. The extensive reading tradition is built entirely on this idea.

Do not take every word you do not know

The instinct on meeting an unknown word is to look it up and save it. Do that consistently and you will end up with two hundred cards from one novel, most of them words you will not meet again for years, and a review queue you abandon by February.

A word earns a place in your collection if it passes one of three tests. It is frequent, meaning it feels like something you have half-seen before and will meet again. It is blocking, meaning the sentence does not work without it. Or it is useful to you specifically, because it belongs to the subject you read, work or talk about. Everything else can be understood from context and left in the book.

As a rough target, five to ten words a chapter is plenty. Almost everyone who tries this takes far too many at the start and quietly gives up. Taking fewer is what makes the habit survive.

Capture without breaking the read

The reason most people stop is friction. Stopping mid-sentence, opening a dictionary app, reading three definitions, deciding on one, typing a card, then finding your way back into the paragraph costs a minute of attention and kills the flow the reading depended on.

The fix is to split capture from processing. While reading, do the cheapest possible thing: pencil dot in the margin, a strip of paper as a bookmark with words on it, or a photo of the page. Keep reading. Then at the end of the chapter, spend two minutes going back over what you marked, dropping the ones that failed the tests above and keeping the rest.

This also improves selection, because by the end of the chapter you know which words actually mattered and which you understood fine from context two lines later.

Keep the sentence, not just the word

When you do save a word, save the sentence you met it in. This is the single highest-return habit in the whole method and the one most people skip.

A word on its own gives you a translation to memorise. A word in its original sentence gives you the grammar it takes, the words it keeps company with, the register, and a small scene to hang the memory on. It is also the difference between recognising a word and being able to use it, which is the gap between passive and active vocabulary. Cards built from real sentences you read produce words you can actually deploy; bare translation pairs mostly produce recognition.

Trim the sentence if it is long, but keep enough that the word is doing real work in it. And test yourself on the sentence with the word blanked out, not on the isolated word, so retrieval happens in something like the conditions where you will need it.

Meet it again on purpose

Reading gives you the first encounter. Retention comes from the encounters after it, which is exactly what spaced repetition is for. Words need several spread-out meetings to move into long-term memory, and a novel will not reliably supply them: most of the words you meet once you will not meet again in that book at all.

So the loop is: read, mark, filter at the chapter end, save with the sentence, review on a schedule. The reading stays reading. The reviewing happens somewhere else, in a few minutes a day, on the twenty or thirty words a week that actually earned their place.

What to do with a word you meet

A quick decision rule for the moment you hit something you do not know.

Mistakes that make this fail

Looking up every word is the obvious one, and it converts reading into decoding. Saving every word you look up is the next, and it buries the words that mattered under the ones that did not. Both come from the same place: the feeling that skipping something means missing out. In a book with two unknown words per page, skipping is the method working, not a lapse.

The subtler mistake is reading something far too hard because the topic is interesting, and calling the resulting slog study. You will finish it, learn less than you would from an easier book read twice as fast, and be less likely to pick up the next one. Difficulty is not the ingredient. Volume of understood text is.

Making the capture step nearly free

The whole method stands or falls on how cheap the capture step is, which is why MindDory lets you point your phone at a page and pull words straight off it, sentence included. You mark as you read, scan at the end of the chapter, keep the five that matter, and they come back in your reviews with the sentence you actually met them in rather than a dictionary gloss. The broader technique is in the guide to memorising vocabulary fast, and the rules for writing the cards themselves are in how to use flashcards effectively.

None of this changes what reading is for. You read because the story is good and because a language only becomes real when you have spent hours inside it. The marking and the reviewing are a small tax on top, and they are the difference between finishing a novel with a vague sense of progress and finishing it with sixty words you can use.