The generation effect and vocabulary: produce it to keep it
Oct 7, 2026 · 12 min read
Read a word off a list and it can be gone by evening. Fill in the missing letters yourself, or guess a word from its sentence and then check it, and something different happens: the word becomes yours in a way a copied definition never quite is.
The generation effect is the finding that information you produce yourself is remembered better than information you only read or hear. In vocabulary work it points to one plain change of order: guess first, check second.
What the generation effect is
The classic demonstration comes from Slamecka and Graf (1978). One group of participants read complete pairs of words; another saw the same pairs with part of the second word missing and had to supply it. Asked later to recall the pairs, the group that produced the missing words remembered more of them. The researchers found that producing an item, rather than simply meeting it, made the difference.
The usual explanation is that producing a word forces fuller processing. You notice its shape, its sound and its meaning, and how it sits in the sentence, and you build your own route back to it. Reading it leaves only recognition, without that work.
Four ideas that sit close by
These four terms are often used as one. They are not.
| Idea | In one line | What it actually covers |
|---|---|---|
| Generation effect | You keep what you make | The first encounter: producing a word beats reading it |
| Testing effect | Retrieval strengthens | Later recall of something already met, as in Roediger and Karpicke (2006) |
| Desirable difficulties | Harder can be better | A general principle from Bjork (1994), covering generation and spacing |
| Depth of processing | Meaning beats surface | The older account from Craik and Lockhart (1972) of why generation pays |
Hard practice helps only when it is the right kind of hard. A word so far out of reach that you produce nothing is a blank wall, not a useful difficulty.
Why it suits vocabulary in particular
A word is at least four things: how it is spelt, how it sounds, what it means, and which words it keeps company with. Producing a word touches all four, because you must build the form and place it in a context. Copying a definition touches the spelling and sometimes the meaning, and leaves the sound and the company of the word alone.
That is why a produced answer makes a better flashcard than a copied one, and why collocations belong in the practice itself. A gap such as "reach a ______" makes you search for a word that fits the meaning and the grammar at once.
Four exercises you can start today
- Guess from the sentence. Read a sentence with one unknown word, decide what it must mean, and write the guess down before you check. The writing makes the correction land. This is using context clues with the checking step made deliberate.
- First-letter retrieval. Write d_c_s_on and produce decision. Two or three missing letters are enough to start the search, and the search is the practice.
- Answer the definition. Read "make up your mind" and produce decide or decision before you look anything up. A definition is a cue, not an answer.
- Fill a collocation gap. Take "make a ______" and produce decision, or "keep a ______" and produce record. Done in pairs, it also shows which word matches the register.
When not to use it
Generation works best when the answer is within reach. If a word has never appeared in anything you have read, a guess is a lottery, and a wrong form you never correct is worse than no attempt: you rehearse the mistake. The rule is not "always guess" but "guess and then check".
Two cases call for reading first. New learners should meet plenty of words in sentences before producing them, because the input supplies the models. Long, low-frequency words that appear once in a text rarely repay a guess. Mixing the material, as in interleaving practice, keeps the guessing honest.
Where it meets spaced repetition
Answers you generate are better review material than definitions you copy, because they carry the trace of your own search. That matters when the word returns in spaced repetition: a card that asks you to produce the word is a retrieval, while a card that shows word and meaning together is a second reading.
So build the card from your own guess and correction. Keep the prompt tight enough to have one clear answer, and far enough from options that you must produce the word. The practical details are in using flashcards effectively, and the reasoning behind the retrieval step is in active recall for vocabulary.
A twenty-five-minute week
Six minutes a day, four days running, is enough to test the habit. Use one set of twenty words for the whole week.
| Day | Six minutes |
|---|---|
| 1 | Read the twenty words in sentences and guess each meaning before you check |
| 2 | Write twenty cards from your own guesses and corrections, not from the dictionary |
| 3 | Cover the answers and produce each word from its definition or its gap |
| 4 | Test yourself cold: write as many of the twenty as you can recall, then check |
Day four shows whether any of it worked. Keep the sheet: the words you missed start next week.
How to tell whether it is working
Take twenty words and split them into two sets of ten. Meet one set by guessing and checking, and copy the other straight from the dictionary. A week later, write down as many as you can from each set without looking, and compare the two scores. Ten words settle nothing, but it puts your own two ways of working on equal terms.
Look beyond the total. Note which words you could produce inside a sentence and which you could only recognise, and whether the produced set held up better with no review. Keeping the record beside a long-term review routine makes the comparison meaningful.
Misunderstandings that get in the way
- "Guessing wrong is harmful, so I should not guess." A wrong guess that is corrected is useful, because it gives you a reason to notice the right answer.
- "Generation replaces practice volume." It changes the order of the work, not the amount. Twenty words still need the same number of meetings.
- "It only works for abstract ideas." Concrete words respond just as well. What matters is the quality of the cue, not how easy the word is to picture.
- "Produce it once and it is fixed." One successful production is a good start. The gains come from producing the word again after a gap.
Frequently asked questions
What is the generation effect?
It is the finding that a word you produce yourself, by filling in letters or guessing before checking, is remembered better than one you only read or hear. Slamecka and Graf set it out in 1978.
Is the generation effect the same as active recall?
No. Generation concerns how you meet a word the first time; active recall concerns retrieving it later. Practice that uses both is stronger than either one alone.
Does guessing a word wrongly make things worse?
Only if you never check. A corrected wrong guess gives you a reason to notice the right answer, so the check matters more than the guess.
Can I use this with flashcards?
Yes. Write the prompt so that you must supply the whole word rather than choose from options, and keep the answer on the back for checking.
How long before I notice a difference?
Judge it over a week or two of steady practice, not after one session. Compare a set you generated with a set you copied, then test both.
This page sets out one way of ordering vocabulary practice. How much of a word stays with you depends on how often you produce it and how the reviews are spaced.