How to explain spaced repetition to a friend in 2 minutes

A friend watches you tap through cards on your phone at the table and asks what you are doing. You say "spaced repetition." They nod politely. The conversation moves on, and nothing you know has moved across.

Here is a script that fits in about two minutes of talking, the five questions people ask back, and one thing you can set up with them the same day.

Start with the forgetting, not the algorithm

Open with the problem the method solves, because your friend already has that problem and does not know it has a fix.

Say this, which runs about 20 seconds out loud:

"You know how you look up a word, use it once, and it's gone by Thursday? Every brain does that with anything it meets once. Yours is working correctly. And the forgetting happens on a schedule, which means you can get in front of it."

This is the part that earns you the next ninety seconds. Most people carry a private theory that they personally have a bad memory for languages. Telling them the loss is normal and timed makes them curious instead of defensive.

Then give them the mechanism, another 20 seconds:

"So instead of reviewing a word every day forever, you review it right before you'd forget it. Get it right, and the next gap gets longer: two days, then five, then two weeks, then two months. Get it wrong, and the gap shrinks back down."

Expanding gaps, driven by whether you got it right. Now ask them to say it back to you in their own words. If they can, they have it. If they stall, the sentence they stall on is the one to say again.

Add the second half, which most explanations skip

Almost everyone explains the spacing and stops. The missing half is the half that does the work.

Say this, about 22 seconds:

"The other half is that you have to actually pull the word out of your head. Say it out loud, from memory, before you see the answer. Recognizing a word on the page feels exactly like knowing it, and the two come apart the moment you have to speak."

If they look sceptical, name the study. This takes eight seconds:

"Researchers ran this on real school classes for a whole term. The facts students got quizzed on came back better months later than the facts they only sat through in class."

Background, if they want more than the eight-second version. Roediger, Agarwal, McDaniel and McDermott published three experiments in the Journal of Experimental Psychology: Applied in 2011. The setting was a sixth-grade social studies course in a suburban middle school. The material was the real course material. The measures were the real chapter exams and semester exams the students were graded on. Some facts got short low-stakes quizzes during the term. Other facts got the same class coverage with no quiz. Quizzed facts scored higher on the chapter exams and again on the semester exams at the end of the term. One experiment went further and compared quizzed facts against facts the students were simply shown a second time. Being quizzed still won. The second look at the page did not do what the act of answering did.

Those three quoted lines are the whole two minutes: 139 words, roughly a minute of talking, and the rest is the pauses where your friend reacts. The five answers below are not part of the two minutes. Say them only if asked, one at a time.

Have the five answers ready

People push back in the same five ways. That count is my own observation from having this conversation, not a survey finding.

Each objection below gets three parts, and they are for three different moments. The first is the line you say out loud. The second is one sentence naming where it comes from, for when they ask how you know. The third is a longer version to paste into a message if they ask you to send it to them in writing. Keep the third one under fifty words, because nobody reads a paragraph on a phone.

"Isn't that just flashcards?"

Say: "Same cards, different timing. Paper flashcards make you review the whole stack at one rate, so you spend most of your time on the ones you already know. This sorts the stack by what is about to fade."

If they want the evidence, say: "This one is arithmetic about how the two stacks get ordered, so there is no study to point at."

In writing: with 200 paper cards you turn all 200 or none. With the same 200 scheduled, the ones you know sit months out, so maybe 15 come due today. Same words, five minutes instead of an hour, and the five minutes land on the shaky ones.

"How long does it take, and how much of my day?"

Say, about 20 seconds: "The US government's own language school takes about seven months of full-time classes to get a diplomat to working Spanish, with a teacher, every day. So months, not weeks. The daily part is small though, and it depends entirely on how many words you add. Anki's manual says twenty new cards a day settles at about two hundred reviews a day. Add five a week and you are nowhere near that."

If they want the evidence, say: "The seven months is the Foreign Service Institute's own published figure, and the card ratio is written in the Anki manual."

In writing: the Foreign Service Institute puts Spanish and French at 30 weeks, 552 to 690 class hours, on 23 hours of class and 17 of homework a week. Anki's manual says 20 new cards a day settles at 200 reviews a day. The project is long, the day is short.

"Does it actually work, or is it just a thing people say?"

Say: "Yes, and it is one of the better-tested findings in memory research. Hundreds of experiments, going back decades, and they point the same way: the same study time spread out beats the same study time piled up."

If they want the evidence, say: "There's a 2006 review of over 300 experiments. The spacing effect is not controversial."

In writing: Cepeda and colleagues reviewed 839 assessments across 317 experiments in 2006. Their 2008 paper taught over 1,350 people and found the best gap shrinks with the horizon: 20 to 40% of a one-week delay, 5 to 10% at a year.

"I already tried Anki and quit."

Say, about 18 seconds: "You almost certainly added too many cards in the first week. Twenty new cards a day compounds to about two hundred reviews a day, roughly ten reviews for every card you add. Start again with five cards a week and cap new cards at two a day, and let the backlog you already have run itself out."

If they want the evidence, say: "The Anki manual warns about exactly this. It says more than one user has excitedly studied hundreds of new cards over their first few days and then become overwhelmed by the reviews."

In writing: the fix is one setting, not a new deck. Lower new cards per day and leave the old cards alone; the review count falls on its own as they get scheduled further out. Two a day settles near 20 reviews, about four minutes.

"What if I keep getting them wrong?"

Say: "Missing a card is the machine working. Guessing wrong and then seeing the answer beats being handed the answer."

If they want the evidence, say: "Kornell, Hays and Bjork showed that across six experiments in 2009."

In writing: that 2009 paper found failed retrieval attempts still improved later learning. It used trivia questions and word pairs, not foreign vocabulary, so do not stretch it. It is enough to say a missed card is not a sign you are doing it wrong.

Do not oversell it, and say who should skip it

Spaced repetition builds recall of specific items. It does not build the ability to hold a conversation. A person with 800 well-scheduled words can read a menu, follow a slow sentence, and name what they want. Put them in a real exchange and they stall, because they have practised producing one word on cue and not stringing words together while somebody waits. Tell your friend the cards sit underneath conversation practice and feed it. They still have to go and talk to somebody.

Then tell them who should ignore this entirely. Somebody going to Lisbon in nine days does not need a review schedule. They need forty phrases and a phrasebook. The method pays off over months, and the payoff is that words stop leaking away between lessons. If your friend has no months to spend, hand them a phrasebook and let it go.

Give them something to do today

Close with a specific action, and pick the one that fits them.

If they want to try the method with no software and no money, give them these six steps:

1. Take five words from whatever they are learning.
2. Write each on an index card: the word on one side, a full sentence using it on the other.
3. Test the stack tomorrow, saying each word out loud before turning the card.
4. Test it again at three days, then at a week.
5. Move any missed card to the front and test it again the next day. Any card they get right three times running goes to a second pile they test once a month.
6. Stop hand-running the stack at four weeks, because past that the gaps get too long to track on paper. That is the moment to move to software.

If they already use software and want the scheduling done for them, point them at Anki and give them these four steps:

1. Install it. Anki is free on desktop, Android and the web, and US$24.99 one-time on iOS.
2. Make one deck. One is enough; more decks now just gives them something to tidy instead of study.
3. Add five cards through the Add window using the Basic note type, which has two fields, Front and Back.
4. Open that deck's options and turn on FSRS, in a section near the bottom of the page. Anki ships SM-2 as its default scheduler and the newer FSRS one is opt-in, so this step will not happen by itself.

Warn them about the content before they start: Anki ships no language material of its own, so every card is one they build or download.

Learn alongside them instead of recommending at them

Agree on this with them: each of you picks five words a week from something real you met, a lesson, a show, a conversation. Each Sunday you send each other your five, in a sentence, not as a bare list. Then during the week, when either of you uses one of the other person's words, you say so.

Slamecka and Graf's 1978 work on the generation effect found that material a person generates is remembered better than the same material read. Writing your own sentence for a friend is generation. So is having to produce their word later in the week. The social part earns its place mechanically: it adds two more retrieval events per word.

Writing five sentences takes about six minutes, and reading theirs takes two. Eight minutes a week is the whole cost of the pact. Set the day now rather than agreeing to do it soon, because a weekly exchange with no named day is a weekly exchange that starts never.

So send them your first five tonight, each one inside a sentence, and ask for theirs by Sunday.