Building a word list that survives
The schedule decides when you see a card. What is written on it decides whether seeing it was worth anything.
Not every word is worth the same effort
English vocabulary is distributed brutally unevenly. The most frequent couple of thousand word families account for something in the region of four fifths of the running words in ordinary written English; everything after that buys you progressively less coverage for the same effort. This is the single most useful fact about vocabulary learning, and it cuts both ways.
Early on it is an argument for frequency lists: if you do not yet have the common two or three thousand, nothing else you learn will pay off as quickly. Later it is an argument against them. Once the frequent core is in place, the words that remain are rare almost by definition, and which rare words matter depends entirely on what you read and who you talk to. A generic list cannot know that. This is the point at which harvesting from your own reading starts to beat any list somebody else compiled.
The figures usually quoted for comfortable unassisted reading — somewhere around 95 to 98 per cent of the running words in a text known — are worth knowing mostly for scale. They explain why a text that feels almost readable can still be exhausting: at 90 per cent you are stopping roughly every tenth word, which is often enough to destroy the thread.
What "knowing a word" actually involves
Treating vocabulary as a set of word–gloss pairs is what makes so much of it evaporate. A more honest breakdown, following the standard framing, is that knowing a word means knowing its form, its meaning, and its use — and that these are learned separately and at different speeds.
- Form. How it sounds, how it is spelled, what it is built from. A word whose pronunciation you have never heard is a word you will not catch in speech, however well you know its definition.
- Meaning. Not one gloss but a range, plus the concepts it does not cover. Most of the errors that survive into advanced use come from a range learned too narrowly.
- Use. What it goes with, what register it belongs to, what grammar it demands. This is the part that a translation pair cannot store at all.
The practical consequence is that the unit worth recording is rarely the bare word. It is the word inside a short piece of language that carries its use along with it.
The shape of a card that works
Four rules, arrived at by getting each of them wrong first.
One card, one thing
A word with three distinct senses is three cards, or one card and two abandoned senses. Cards that ask for several things at once fail as a unit, and you can never tell which part failed.
Prompt with a sentence, not a word
The front should be a sentence you actually met, with the target removed. Context does most of the retrieval work, it stores the collocation for free, and it makes the card unambiguous — you are asking for one specific word, not for whichever synonym comes to mind first.
Record the collocation explicitly
Not tentative — предварительный but a tentative agreement / tentative conclusion. Words that are individually known and jointly unusable are the most common failure of advanced vocabulary, and collocation is where the problem lives.
Keep the source
A line saying where the sentence came from costs nothing and rescues the card later, when the meaning has gone fuzzy and you need to reconstruct why you thought it was worth keeping.
| Word | Collocation kept | Note to self |
|---|---|---|
| brittle | a brittle smile; brittle bones | Not just physical — the figurative use (of a mood, a peace, a laugh) is the common one in prose. |
| hedge | hedge a claim; hedge against risk | Verb sense only. The garden noun was already known and does not belong on the same card. |
| outstrip | demand outstripped supply | Almost always about quantities growing past one another. Rarely about people racing. |
| tentative | a tentative agreement; tentative steps | Two ranges — provisional, and hesitant. Split into two cards after failing it four times. |
| byword | a byword for inefficiency | Needs for. Without the preposition the word is unusable, which is why the bare gloss never stuck. |
Do not learn similar words together
This is the counterintuitive one, and the mistake nearly every self-made list contains. Sets of semantically related words — synonyms, opposites, or thematic groups like the clothing words, the kitchen words — are noticeably harder to learn together than the same words learned apart. The items interfere: each one becomes a competing candidate at the moment you try to retrieve another, and the whole set stays mutually confusable long after it should have settled.
It feels efficient to group them, which is why textbooks do it and why every beginner's list is organised by topic. If you have a persistent pair you keep swapping, this is very often the cause. Separate them by a few weeks and one of them usually resolves on its own.
Pruning
A list is not an archive. Cards earn their place:
- If a word has failed repeatedly and rewriting it has not helped, delete it. You will meet it again if it matters, and next time it will arrive with better context.
- If you cannot remember why you added it, delete it.
- If it is so rare you will not meet it again this year, it belongs in a note, not a review queue.
The deck exists to be finished each day. Anything that makes that less likely is costing more than it returns.