How to Learn Vocabulary from Reading (Without Making 4,000 Flashcards)
In short
- A word is rarely learned from one encounter. Webb gave learners unknown words 1, 3, 7 or 10 times in context: every increase helped, ten encounters produced sizeable gains, and full knowledge of a word needed more than ten.
- Spaced review is one of the most robustly supported findings in learning research. Cepeda and colleagues synthesised 839 assessments across 317 experiments and found distributed practice reliably beats massed practice at the same total study time.
- Reading supplies the encounters and the context; spaced review supplies the schedule. Neither one alone is efficient.
- Most words you meet do not deserve a flashcard. The ones that do are high and mid-frequency words you have already met more than once.
- A sustainable rate for an adult learner is roughly 15 to 30 new cards a week, not the 100 people attempt in week one and abandon in week three.
There are two standard approaches to vocabulary and both of them are half right.
The first says: just read a lot, the words will come. This is true and it is slow. The second says: drill flashcards, that is how you build a vocabulary fast. This is also true and it produces a peculiar kind of knowledge where you can translate a word but cannot use it, and cannot recognise it when it appears with a different preposition.
The efficient version is the combination, with a filter in between. Here is how the pieces actually fit.
Why reading alone is slow
Incidental vocabulary acquisition — picking up words from context without trying to — is real and well documented. It is also expensive per word.
Stuart Webb tested the repetition question directly with 121 Japanese learners, substituting nonsense words into authentic text so that each target appeared 1, 3, 7 or 10 times, and measuring ten separate aspects of word knowledge — orthography, association, grammatical functions, syntax, and meaning and form. Every increase in repetition improved at least one aspect. Ten encounters produced sizeable learning gains; developing full knowledge of a word, he concluded, takes more than ten.
Rob Waring and Misako Takaki's study shows the decay side of the same problem. Fifteen Japanese learners read one 400-headword graded reader with 25 invented words substituted in at frequencies from 1 to 18 occurrences, and were tested immediately, after a week and after three months. About 40% of the words were remembered on the immediate multiple-choice recognition test. Three months later, even the words that had appeared more than eighteen times had only a 10-15% chance of the meaning being remembered. More frequent words were both more likely to be learned and more resistant to decay.
One book does not teach vocabulary. That is not an argument against reading — it is an argument about volume. A hundred books does teach vocabulary, and teaches it far better than any deck, because each encounter arrives with collocation, grammar and register attached. But if you want a specific word available in six weeks rather than in two years, reading alone will not get you there.
Why flashcards alone are shallow
Paul Nation's framework for what it means to know a word, set out in Learning Vocabulary in Another Language, splits it three ways — its form (spelling, sound, word parts), its meaning (form and meaning, concept and referents, associations), and its use (grammatical functions, collocations, constraints of register and frequency) — each of those in a receptive and a productive version. That is nine cells, and a translation pair fills one of them.
A bare two-sided flashcard teaches receptive form-to-meaning recognition and nothing else. That is genuinely useful; it is also one cell out of nine. It is why learners who have drilled 5,000 cards still write sentences that are individually correct and collectively wrong: they know the meanings and not the company the words keep.
The fix is not to abandon cards. It is to put context on them.
The spacing effect, briefly
The reason to use a scheduled review system rather than re-reading a word list is one of the older and more solid findings in psychology. Hermann Ebbinghaus documented the forgetting curve in 1885; Cepeda, Pashler, Vul, Wixted and Rohrer synthesised 839 assessments of distributed practice across 317 experiments in 184 articles and confirmed that spacing reliably beats massing at the same total study time.
Their more interesting finding, and the one most spaced-repetition advice skips: the gap between study sessions that produces maximum retention grows with the retention interval. If you need to remember something next week, review it in a few days. If you need it in a year, the optimal gaps are much longer than intuition suggests.
The practical version: reviewing a word four times across three weeks beats reviewing it four times in one evening, and it is not close. Spaced repetition software automates the scheduling — the Leitner box system does it with physical cards and five boxes, and modern algorithms do the same thing with computed intervals — but the scheduling is the mechanism, not the software.
The important limitation: a review system only rehearses what you put in it. It has no way of strengthening the four hundred words you half-know but never carded. That is what reading is for.
The filter: which words deserve a card
This is where most vocabulary systems fail. Learners card everything they look up, the deck grows past what they can review, reviews back up into the hundreds, and the whole thing gets abandoned.
Card a word only when all three of these are true:
- You have met it at least twice. First encounters are noise. If a word matters, it will come back.
- It is in the high or mid-frequency range — roughly the first 9,000 word families of English. Below that, a word is rare enough that context and guessing strategies are a better investment than memorisation.
- Not knowing it actually blocked you. If you understood the sentence anyway, the word can wait.
| Word type | Card it? | Why |
|---|---|---|
| High-frequency word you keep missing | Yes, immediately | The first 2,000 words cover about 80% of a general written corpus. Gaps here are expensive. |
| Mid-frequency word met twice | Yes | This band is where most adult learners are actually stuck. |
| Academic word in your field | Yes | High return if you study or work in English. |
| Technical term outside your field | No | You will not meet it again. |
| Proper noun | No | Not vocabulary. |
| Word you guessed correctly from context | No | You are already most of the way there. |
| Rare literary word | Usually no | Enjoy it, do not schedule it. |
In practice this filter removes about four out of five words you look up. That is the point.
A weekly shape that survives contact with real life
| When | What | Time |
|---|---|---|
| Daily | Review whatever the system schedules | 5-10 minutes |
| 3-4 times a week | Extensive reading at your level, no dictionary, flag words with a pencil | 20-30 minutes |
| Once a week | Go through the flagged words, apply the filter, add 15-30 cards with their sentences | 15 minutes |
Note what is deliberately absent: no dictionary during the reading session. Stopping to look up and card a word mid-page destroys the reading, and the reading is doing half the work. Mark it, keep going, deal with it at the weekend. The split between fast reading and close study is covered in more depth in extensive vs intensive reading.
What to actually put on the card
Front: the word, and the sentence you met it in with the target word blanked out.
Back: the meaning in whichever language you think in, the original sentence complete, and — if the word has one — the collocation it usually appears in.
That last part matters more than it looks. Knowing that "research" pairs with "conduct" and not with "make" is a large fraction of sounding fluent, and no two-sided translation card will ever teach it.
Measuring whether any of this is working
Two measurements are worth taking, roughly every three months.
Vocabulary size. A vocabulary size test samples across frequency bands and gives you an estimate. It is a blunt instrument — it measures recognition, not use — but tracked over time the direction is informative, and it tells you which frequency band you are currently stuck in. What the resulting number does and does not mean is covered in how many English words you actually need.
Reading comfort. Take a text at the level above your current one and count unknown words per hundred. If that count is falling, the vocabulary is doing its job. If your deck is growing but the count is not moving, you are probably carding words that are too rare to matter.
For the daily reading half of this, LingoRead generates passages at a CEFR level you choose and extracts the vocabulary from the passage itself, so the words you review are the words you actually just met in context — which removes the collection step that is usually the first thing to fall apart.
Sources
- Webb, S. (2007). The effects of repetition on vocabulary knowledge. Applied Linguistics, 28(1), 46-65. Source for the 1/3/7/10 encounter design, the ten measured aspects of word knowledge, and the conclusion about ten encounters.
- Waring, R. and Takaki, M. (2003). At what rate do learners learn and retain new vocabulary from reading a graded reader? Reading in a Foreign Language, 15(2), 130-163. Source for the 40% immediate recognition figure and the 10-15% three-month retention for words met more than eighteen times.
- Cepeda, N. J., Pashler, H., Vul, E., Wixted, J. T. and Rohrer, D. (2006). Distributed practice in verbal recall tasks: a review and quantitative synthesis. Psychological Bulletin, 132(3), 354-380. Source for the 839 assessments and for the interaction between study gap and retention interval.
- Uchihara, T., Webb, S. and Yanagisawa, A. (2019). The effects of repetition on incidental vocabulary learning: a meta-analysis of correlational studies. Language Learning, 69(3). A wider synthesis of the repetition question.
- Nation, I. S. P. Learning Vocabulary in Another Language. Cambridge University Press. Source for the form/meaning/use framework; a free summary of much of the underlying work is on his resources page.
- Nation, I. S. P. and Waring, R. (1997). Vocabulary size, text coverage and word lists. Source for the 2,000-word coverage figure.