You looked up a word, wrote it down, understood it perfectly, and could not produce it four days later. That is not a memory problem. It is a scheduling problem, and it has one of the better-replicated fixes in learning research.
The fix is old. Hermann Ebbinghaus ran the first version on himself in the 1880s, memorising nonsense syllables and measuring how much he lost over time. He found two things. Forgetting is steep at first and then flattens, and reviews spread out over days beat reviews packed into one sitting for the same total effort.
Nearly everything sold as a vocabulary method since then is a variation on getting those two facts right or wrong.
Why Rereading Feels Better and Works Worse
Open your notes, look at the word list, recognise every word, feel prepared. This is the trap, and it has a name in the literature: fluency is a bad cue for memory strength.
Roediger and Karpicke tested it directly in 2006. Students studied prose passages and were split into groups that either restudied the material or took recall tests on it. Five minutes later, the restudy group performed better and predicted they would do better. A week later the ranking had flipped hard: the tested group remembered far more, and the restudy group had confidently predicted the opposite.
The act of pulling a word out of memory changes the memory. The act of looking at it mostly changes your confidence.
For vocabulary this has a blunt practical meaning. A review where you see **ubiquitous** and its definition together is nearly worthless. A review where you see the prompt, feel the gap, struggle for three seconds and then produce it is the one that pays.
The Spacing Effect and the Size of the Gap
The second finding is that when you review matters as much as whether you review.
Cepeda and colleagues published a meta-analysis in 2006 covering 254 studies and more than 14,000 participants. Spaced practice beat massed practice consistently, and the advantage was not small. The follow-up question was the useful one: how big should the gap be?
Their 2008 experiment answered it by crossing study gaps against retention intervals. The result was that the best gap depends on how long you need the material to last, and it scales with that target. Short-term goals want short gaps. Long-term retention wants gaps measured in weeks and months.
The working approximation most people can act on:
| You need it in | Rough best gap between reviews |
|---|---|
| One week | About 1 day |
| One month | 3 to 5 days |
| Six months | 3 to 4 weeks |
| A year or more | 1 to 2 months |
Notice what this rules out. Reviewing a word every single day for a month is not thorough, it is wasteful. Each review lands before you had any chance to forget, so it costs time and adds almost nothing.
Desirable Difficulty
Robert Bjork's term for this is desirable difficulty: conditions that slow learning down in the moment and improve retention later. A review scheduled at the point where recall is effortful but still possible is doing more work than an easy one.
That is the whole logic of an expanding interval. Each successful recall buys you a longer gap, so the next review lands back near the edge of forgetting rather than well inside comfort. A typical ladder looks like this:
- New word: same day
- First recall: 1 day
- Second: 3 days
- Third: 1 week
- Fourth: 3 weeks
- Fifth: 2 months
- Then months apart, effectively maintenance
And when you fail a word, it drops back down the ladder. That drop is not a punishment, it is the schedule doing its job.
The Part Most Vocabulary Apps Get Wrong
Recognising a word is not knowing it. Researchers separate receptive knowledge, meaning you understand a word when you meet it, from productive knowledge, meaning you can retrieve it when you need it. The gap between the two is wide, and it is exactly the gap behind the complaint that you can understand English but cannot speak it.
Multiple choice reviews train recognition almost exclusively. Four options are on screen, three are obviously wrong, and elimination gets you the answer without ever retrieving anything. You will feel accurate and stay unable to produce the word in a sentence.
Three fixes, in order of how much they matter:
Review from meaning to word, not word to meaning. Seeing the definition and producing the word is the direction that actually gets used in speech.
Learn words inside a sentence. A single-word translation gives you a label. A sentence gives you the collocations and the register, which is what tells you that you make a decision but not that you do a decision.
Say it out loud. Producing the sound is a different retrieval route than reading it silently, and it is the route you need in conversation.
How Many Words, Realistically
The count that matters is not how many you add. It is how many reviews the additions generate.
Every new word adds a tail of reviews stretching months out. Add 30 a day for two weeks and the daily review queue quietly grows past an hour, which is the point where the habit dies and the whole deck goes stale.
Ten to twenty new words a day is the range that most people sustain. Over a year at fifteen a day, allowing for missed days and forgetting, that is a few thousand words. Against the roughly 3,000 word families that cover most everyday English, and the 8,000 to 9,000 that supports comfortable reading of unsimplified text, this is a serious number. We have covered the target counts in more detail in how many English words you need to be fluent.
Leeches: The Words That Refuse
Every deck develops a handful of words you have failed eleven times. These are called leeches, and they eat a disproportionate share of your review time.
Failing repeatedly means the card is wrong, not that you are. Usually one of three things:
- The word has no hook. Add a sentence you would actually say, or an image, or an etymology
- Two similar words are interfering. Study them together deliberately once, contrasted, rather than separately
- You do not need it. Suspend it. A rare word you will not meet for a year is not worth twelve reviews a month
What Does Not Help
Long single sessions. Two hours on Sunday loses to fifteen minutes a day, and the meta-analysis above is exactly why.
Highlighting and rereading notes. A 2013 review by Dunlosky and colleagues rated both as low utility techniques. Practice testing and distributed practice were the two rated high, which is the pairing spaced repetition happens to combine.
Enormous imported decks. A 5,000 word shared deck generates a review load you did not choose, full of words you will never use, and it collapses within a month.
Chasing the perfect algorithm. The difference between competing scheduling algorithms is small. The difference between reviewing on schedule and not reviewing is enormous.
Making the Schedule Survive Real Weeks
The mechanism is well established. The failure is always adherence: you skip four days, the queue is at 200, opening it feels awful, and the deck dies.
That is the narrow job Lingui: Learn English Words does. It handles the interval scheduling and daily review queue so the timing question is answered for you, and it teaches words in context rather than as bare translation pairs, which is the difference between recognising a word and being able to use it.
Be honest about the limits. No review app can give you the hundreds of hours of listening and speaking that turn a stocked vocabulary into fluent speech, and it cannot supply the situations where you have to produce a word under time pressure with someone waiting. What it does well is guarantee that the words you have already met come back at the right moment instead of quietly disappearing. Pair a fifteen minute review session with real reading and real conversation and you have both halves. Do only the reviews and you will end up with a large passive vocabulary and the same stuck feeling you started with.
The Short Version
- Testing yourself beats rereading, and the advantage grows the longer you wait to measure it
- Gaps should expand, and should scale with how long you need the word. Roughly 10 to 20 percent of the target interval
- A review that feels slightly hard is worth more than one that feels easy
- Review from meaning to word, in a sentence, out loud. Multiple choice trains recognition only
- Ten to twenty new words a day is sustainable. Thirty is a review queue you will abandon in a month
- Suspend or rebuild leeches instead of failing them forever
- The algorithm barely matters. Not skipping matters enormously