← Blog

Why last Monday's words are already gone

The research says the problem is not how much you studied. It is what happened after you got them right.

You had the lesson. You wrote the words in your notebook, characters on the left, pinyin and meaning on the right. Your tutor asked you to say them back and you said them back, correctly, all of them.

Then Monday became Thursday. And now it is next Monday, you are looking at 小笼包 on a menu, and you have the specific feeling of having met this word before and having nothing else to show for it.

The usual explanation is that you did not study enough.

That is almost never what happened.

Everyone in the experiment learned the words

In 2008, Jeffrey Karpicke and Henry Roediger ran a study that is uncomfortable to read if you own a notebook full of vocabulary.

Students learned 40 Swahili-English word pairs. Everyone studied and everyone was tested, in alternating rounds, until they had the whole list. The four groups differed in one thing only: what happened to a word after the student got it right for the first time.

By the end of the session, all four groups had learned the list. The paper is blunt about it: performance was "virtually perfect by the end of learning," and there were "no differences in the learning curves of the four conditions." Same effort, same result, same confidence walking out the door.

A week later they came back for a final test.

Read those numbers again with your notebook in mind. The group that kept studying the words they had already got right, over and over, ended up at 36%. Restudying a word you have already recalled correctly bought them almost nothing. What separated 80% from 36% was whether they had to pull the word out of their own head again.

Karpicke and Roediger had found the same thing the year before with English word lists, in a paper whose title leaves no room for interpretation: "Repeated retrieval during learning is the key to long-term retention." In that experiment, when learners kept being tested on words they had already recalled, one-week memory was more than twice what it was when those words were dropped from further testing. 44% against 21%. Meanwhile, continuing to study the already-recalled words instead of dropping them made no measurable difference at all.

Almost everything you do after a lesson is restudying

Here is the part that stings.

Reading back through your notes is studying. Highlighting the vocabulary list is studying. Copying the characters out a second time, neatly, is studying. Scrolling the photo you took of the menu is studying. Watching the word go past in a video with the meaning under it is studying.

All of it feels like effort. All of it feels like the responsible thing to do. And on a one-week horizon, the evidence says it buys you close to nothing you did not already have.

Retrieval is different, and it is uncomfortable in a way that restudying never is. It means the word is not in front of you and you have to produce it anyway. Cover the right-hand column and say what 外滩 means before you look. Walk past a shop and try to remember the character you learned for it, then check. Try to say the sentence your tutor gave you without opening the notebook.

The word you just got right is the one you are about to lose

This is the trap, and it is built into how we all study.

The moment a word feels solid is the moment you take it off the list. It goes in the "done" pile. You stop quizzing it and you move to the words that still feel shaky, because that feels like the efficient use of your evening.

But "I got it right once" was the exact condition in the experiment that fell to 36%. Those students were not lazy and they were not undertrained. They had recalled the word correctly. Then the word stopped coming back to them, and a week later a third of it was left.

So the practical change is small and slightly counterintuitive: the words you got right on Monday need to come back on Wednesday, and again a few days after that. Not to be read. To be answered, from memory, before you check.

You do not need an elaborate system for this. You need the words you met this week to come round again while you still half-remember them, and you need to try to produce each one before you see the answer. That is the whole mechanism. Ten minutes of being asked beats an hour of being shown.

Kezi does the first half of this. You paste the text you actually met today, it fills in the pinyin and the meaning, and you are practising in about ten seconds instead of spending twenty minutes making cards.

It does the second half too, quietly. Open a lesson you have already finished and run it again, mastered words included. Swipe left on anything that has gone soft and it drops straight back into your practice pile.

What Kezi will not do yet is tap you on the shoulder and tell you which words are due. We are building that.

Until then you can do it yourself, and it takes about two minutes. Open a lesson you have already finished and practise it again. Everything comes back, the mastered words included. Swipe left on anything you cannot produce, and it drops into your practice pile, where you can guess, reveal, and guess again until it feels solid.

That guessing is not the part to get through. That is the part that works.

Because Monday-you could say it. That was never the problem.