{{first_name | みんな}}さん、こんにちは ☕

You open your SRS app. The same 20 words are back. 止とめる or 止とまる? (We covered why those pairs trip everyone up — learning them with their particle is the fix.) Was it くう or そら for 空そら? You hit "Again" for the eighth time on the same card and start wondering if your brain is just... broken.

It's not.

Every leech card — a card you fail so often it starts consuming a disproportionate share of your review time; most SRS apps (like Anki) tag and suspend these automatically after a threshold number of failures — is trying to tell you something. There are exactly three things it might be saying. Today we crack all three.

In our SRS Triage edition, we gave you the system-level triage: when to add cards, when to stop, when to change decks. Today we go one level deeper — card-level triage. The stubborn ones that survived your first purge? Those are the ones we're after.

Here's the thing most learners get wrong about leeches: they assume a leech means the word is hard.

Sometimes a word genuinely IS harder — abstract concepts, sound-alike pairs, kanji with overlapping readings. That's real. But the card failing eight times in a row? That's almost always a card problem, not a brain problem. The word 届とどける is not inherently harder than 食たべる. Your card for 届ける just failed to create the conditions for encoding.

And the default SRS experience gives you zero help. Anki's leech system? It tags the card, suspends it, and then leaves you staring at it with no guidance. WaniKani detects leeches but offers no framework for fixing the underlying cause. The tools identify the symptom without treating the disease.

That's where most learners get stuck. They either brute-force through (doesn't work — shallow processing creates unstable memory traces that crumble under pressure), or they mass-delete everything (loses words they actually need).

Neither approach asks the only question that matters: why did this card become a leech?

Every leech you've ever had falls into one of three categories. Diagnose the type, apply the matching fix. That's it.

Type 1: The Ghost Card (encoding failure)

The word never formed a strong memory trace. You added it from a word list or pre-made deck without context. You "learned" it by seeing the answer, not by understanding it.

The signal: You see the card and feel nothing. No partial recognition, no tip-of-the-tongue. Just blank.

Example: You added 届ける from a frequency list. You see "to deliver" and think... nothing. No image, no sentence, no situation where you'd ever say this.

The fix: Context injection. Add a sentence you care about: 荷物にもつを届とどけてくれますか？ "Can you deliver the package for me?" Add an image or personal connection — we covered how the right image transforms a card in a previous edition. If you've never encountered this word outside your SRS — that's the problem. Wozniak's first rule of knowledge formulation: "Do not learn if you do not understand."

One more trick that works surprisingly well: learn the kanji that compose the word. If you understand the components — what each kanji means, how it contributes to the compound — you build structural memory that supports word-level recall. 届ける becomes less of a stranger when you recognize 届 as "reach/deliver" from its components.

Type 2: The Evil Twin (interference failure)

You know this word exists. You can almost remember it. But your brain keeps swapping it with a similar word — same reading, similar kanji, overlapping meaning. This is proactive interference — old memories blocking new ones, or similar items competing for the same mental slot.

The signal: You see the card, you think you know it, and you're confidently wrong. You give the answer for the other word.

Example: 聞きく vs 聴きく — both きく, but one is passive hearing and the other is active listening. Or the classic 思おもう vs 考かんがえる — both "to think," but your card asks for one and you always produce the other.

The fix: Distinction card. Do NOT study the two words separately and hope your brain sorts it out. Create a new card that puts them side by side and asks: "What is the difference between X and Y?" This is discrimination training — and once the distinction is explicit, both original cards often become easy.

Here's a mindset shift worth making: not all cards in your deck need to be vocabulary, kanji, or grammar cards. You can create "question yourself" cards — cards that test your understanding of a distinction, not just recall of a single item. Your deck is a thinking tool, not just a memorization tool.

And just like with Ghost Cards, studying the kanji that compose each word helps here too. When you understand why 聞きく uses 耳みみ "ear" + 門もん "gate" while 聴く uses 耳 + 十 + 四 + 心こころ "heart," the distinction stops being arbitrary and starts being structural.

Type 3: The Wrong Shape (format failure)

The card is testing the wrong thing, or testing too many things at once. You know the word. You use it in conversation. But something about the card trips you up every time.

The signal: You know you know this. The card just doesn't let you prove it.

Example: A card asks "What does 早はやい mean?" and the answer is "early/fast." You write "fast" — wrong, it wanted "early." But the real issue: the card is testing two things at once. If the problem is actually a near-twin (早い vs 速はやい), treat it as an Evil Twin instead.

The fix: Redesign the card. Split cards that test multiple things. Change direction (recognition instead of production, or vice versa). Add acceptable answer variants. If the card tests something you don't actually need (like handwriting kanji when you only type), delete it. The Minimum Information Principle: each card should test exactly one thing.

This is the part we wish someone had told us two years ago. Not a quick hack — a real system for cards that won't behave.

After diagnosing the type, one more question: "Do I actually need this word in the next 3 months?" If no — suspend it. Not delete. Suspend. You might need it later.

If yes — here's our five-step protocol.

The Leech Triage: Diagnose Your Top 5

Open your SRS app (Anki, WaniKani, Bunpro, Kitsun — whatever you use). Find your leech cards. In Anki: search tag:leech in the browser. In WaniKani: sort by most failed or use the Leech Training userscript. In other apps: find cards you've failed 5+ times.

Pick your top 5 worst offenders. For each one, ask: why did my brain reject this?

(Answers aren't pass/fail here — but we'll share example diagnoses next week so you can check your thinking!)

→ Starter: Pick 3 cards. Just label each one: Ghost, Evil Twin, or Wrong Shape. No fixes yet — diagnosis only.

→ Standard: Pick 5 cards. Diagnose each one, then apply the matching fix from the triage. Time yourself — most card edits take under 3 minutes.

→ Challenge: After diagnosing and fixing 5 cards, create a dedicated leech deck. Move your diagnosed cards into it and set tighter intervals (1-day graduating, 3-day max). Review daily for one week, then report back.

You'll know you did it when: You can point at each of your worst leeches and say exactly which of the three types it is — and you've done something different to each one.

Last week's answers (Kanji Radicals):

Last week we asked you to X-Ray 5 kanji — find the parts, spot the meaning hint and the sound hint. There's no single right answer, but here's what to look for: the semantic component (meaning hint) is often the 部首ぶしゅ, and the phonetic component (sound hint) is the other piece. In 語ご, the 言 tells you it's about language, and 吾 predicts the ゴ reading. In 清せい, 氵says "water/purity" and 青 predicts セイ. The pattern holds more often than you'd expect — and now you can't unsee it.

You don't need to fix every leech today. You don't even need to fix five.

If all you did was look at one stubborn card and think "oh — that's a Ghost" instead of "I'm bad at this" — that's the shift. You just went from blaming your memory to diagnosing your cards. Everything changes from here. 💡

Further reading (optional):

This week, triage one leech. Just one. Diagnose it, fix it, and reply to tell us: Ghost, Evil Twin, or Wrong Shape? We read every reply. (And yes, we get unreasonably invested in other people's leech stories.)