The kanji you studied this week might not be there next week.

You flip through flashcards and it's clicking. 飲 comes up, you think "drink" — right. 待 comes up, you think "wait" — right. You close the app feeling like you got something done. Then you open actual text, or someone asks you to write it, and it's gone.

Recognition and recall are two different cognitive processes. Most study methods train the first one. The second one is what you need to actually use kanji in context.

What recognition is

Recognition is what happens when a character appears and your brain matches it to something stored. The pattern looks familiar, you confirm it. When you study flashcards with the answer visible — or flip through cards reading both sides — that's mostly what's happening. Your brain gets the warm signal of "yes, that one" and it feels like learning.

Recall is the other direction. A blank, and you have to pull the answer out unprompted. These are not the same task, and practicing one does not build the other.

What the research says

Karpicke and Roediger tested this directly (Science, 2008). Students who practiced retrieval — actively trying to recall the answer before checking — retained significantly more after a week than students who re-studied the same material the same number of times. The re-study group felt more prepared going in. They remembered less.

The confidence is the trap. Going through cards where you can see the answer feels productive. But if the answer is visible while you process it, you're not retrieving anything. You're recognizing. You can spend an hour doing that and leave thinking you've learned twenty kanji, when what you've done is made them familiar.

Why kanji makes this worse

Characters like 特 and 持 look similar enough that passive review keeps both vaguely familiar without actually separating them. Your brain has seen each one enough times to confirm it when labeled, but not enough retrieval attempts to pull either one cleanly under pressure. So you study, feel like you're progressing, then blank when a character appears in a sentence with nothing nearby to confirm it.

I spent years on workbooks and textbook drills that were mostly reading and matching. Progress was slow. When I was working in Japan and had a teacher who had me write letters in Japanese, things changed fast. Not because writing is some special method — it's because I had to produce characters rather than recognize them. A lot of kanji I thought I knew turned out not to be there in that direction.

The fix

Force the recall. Card appears, you pause and attempt the answer before you can see it, then you flip. That pause is all retrieval practice is. It sounds like a minor adjustment. It changes what you're actually building.

SimplyKanji is built around this loop — see the character, try to recall, flip. N5 and N4 are free. But the principle applies to whatever you're using. The question to ask is whether the answer is visible while you're processing. If it is, you're doing recognition practice.

One setup thing worth fixing if you use bidirectional cards: studying "character to meaning" and "meaning to character" in the same session lets your brain link the pair during the session itself. You end up learning the association rather than training independent retrieval in each direction. Separating them into different sessions makes each direction stronger.

The uncomfortable version

Most people who feel like they study kanji regularly and see slow progress aren't failing because of effort. The effort is there. The task being practiced is recognition. What you need in actual use is recall. Most study setups don't make that distinction visible, so the gap never gets addressed.

What happens the next time you study if you actually pause to attempt the answer before flipping?