Learning vocabulary that stays

Vocabulary is the largest single component of language learning by volume, and most people approach it with methods that guarantee forgetting.

Why words are forgotten

The mechanism.

Memory research consistently finds that retention decays rapidly after learning unless the material is retrieved.

Recognition — seeing a word and remembering its meaning — decays more slowly than recall, which is why people can read far more than they can say.

And words met once are almost never retained; the estimates for how many encounters are needed for a word to stick from reading alone generally run into double figures.

Which means the problem is not effort at the moment of learning but what happens over the following weeks.

Retrieval rather than review

The most important distinction.

Re-reading a word list feels productive and produces very little.

Attempting to recall the word before seeing the answer produces substantially better retention, which is one of the most robust findings in learning research.

Which means every review should be a test rather than a reading.

And the effort of retrieval is the mechanism, so easy reviews are less useful than difficult ones.

Spacing

The second principle.

Reviewing at expanding intervals produces better long-term retention than the same total time spent in one session.

Which means a word reviewed after a day, three days, a week, a fortnight, a month is retained where the same number of reviews in one afternoon is not.

Spaced repetition software automates the scheduling and is the practical implementation.

Ten to twenty minutes daily is enough for most learners, and consistency matters more than session length.

What to put on a card

Where most systems fail.

A word in isolation with a single translation is the weakest form.

Better: the word in a sentence, with the target word blanked, so the context supplies the meaning.

Include an image where the word is concrete.

Include audio, since a word learned only in writing is not recognisable in speech.

One card per meaning rather than a list of meanings.

And make cards from material you have actually encountered, which means they arrive with context already attached.

Which words

The prioritisation.

Frequency lists are the efficient starting point: in most languages a few thousand words cover the large majority of everyday text.

The first thousand are worth deliberate study.

Beyond that, words from your own reading and listening, since they reflect what you actually consume.

Plus the vocabulary of whatever you will talk about, which is personal and is what makes conversation possible.

And not obscure words, however satisfying, which will not recur and will not stick.

Collocations and chunks

Which are more useful than single words.

Words occur in predictable partnerships, and learning them together produces natural-sounding output.

Learning "make a decision" rather than "decision" avoids the characteristic errors of learners who assemble sentences word by word.

Which means noting phrases rather than words when reading.

And it is why input matters: collocations are absorbed rather than studied.

Depth of processing

What makes a word stick faster.

Words connected to something — an image, a personal association, an emotion, a story — are retained better than words memorised abstractly.

Which means using a new word in a sentence about your own life, or noticing where it comes from, or connecting it to a similar word you know.

The keyword method — linking a foreign word to a similar-sounding word in your own language via an image — has good experimental support for initial learning, though the link fades and the word must still be consolidated by use.

Production

The step that is skipped.

A word you can recognise is not a word you can use.

Which means deliberately using new words: in writing, in conversation, in the diary you keep in the language.

Aim to use a new word within a few days of learning it.

And accept the errors this produces, since a word used wrongly and corrected is learned better than a word never risked.

Maintenance

The long game.

Vocabulary decays without use, including in your own languages.

Which means continued input rather than continued study: reading and listening maintain vocabulary far more efficiently than review does.

Spaced repetition is best used for acquisition and for words that do not recur naturally.

And the ideal end state is that most maintenance happens through using the language for something you would do anyway.