What I wish I knew about Anki before I started

Point #1: The Generation Effect. It’s a robust finding in an extremely wide range of contexts across basically all of memory research. Generating anything from your own mind just does so much more to build recognition of that thing than practicing recognizing it does. Output shouldn’t be something you postpone until you think you’re ‘ready’ to become more fluent or whatever, because it is the thing that actually makes you get there. And the hurdle of initial extra time and effort you’re worried about will be more than compensated for by how much less you will be forgetting things and drowning in reviews months down the line—because the scientific evidence universally shows that it is dramatically more effective.

I felt like I was sludging through WaniKani until I decided to pick up Ringotan and test myself on handwriting kanji from memory. I experimented with handwriting the upcoming WK level’s kanji a day or two in advance, and though it seemed like that was going to be more work… that little bit of extra effort just makes each recall so much more effective that it actually feels like the lazy approach in the end.

After some levels of doing this, I quit WK because it felt like it was throwing me underwater with leg shackles on—I was like, how many times do I need to repeat this stuff to get to something new? I know these already! And when that feeling just got worse for weeks in a row, I realized it was time to move on.

At some point I spent a couple weeks going through the next 20+ levels of WK in Ringotan and “finishing it.” (I now have 2700 kanji maintained in Ringotan with 2350 “mastered”. 2700 is the point at which I felt learning new kanji instead of learning new words finally took a very noticeably steep drop in value—but I was still coming across things outside of Jōyo or WK like 朦朧 or 閏年 or 一瞥 or 蹂躙 or 躊躇 or 滲む quite frequently until then. Suffice to say that at the rate I was going doing recognition practice only, I thought I would never get here. I’m honestly still shocked at how different things were after I changed focus to output…. and every bit of evidence I can find suggests it works that way for everyone, because that’s just how memory works.)

Point #2: You can easily make your own WK-esque cards—better ones, in fact, because they can do everything that WaniKani does right and then some—with only basic Anki knowledge.

Here’s what WaniKani does right that your average Anki deck doesn’t: it gives you atomic questions where you either get this singular piece of knowledge right or you don’t, and it splits all the information it’s testing you on up separately and repeats each piece according to how well you pass it individually. This doesn’t allow you to deceive yourself, and it doesn’t leave you in analysis paralysis guessing “hmm, did I remember enough about that to pass it?”

Here’s all you need to know to incorporate that into your own card making. As in, the whole thing is in the next paragraph. Everything after the next paragraph is just elaborating. So:

When you click "Add Card," what you are actually creating is a "Note." Usually this also means one card, because most notes are configured to produce one card. But notes are just an abstract storage of information—cards are built out of this information, and cards are what you actually review.

So if you fill in a note with {English definition}, {Kanji}, {Reading} fields, here are some examples of what you can do.

You can have one card generate with {English definition} on the front and a line saying “handwrite it!” with {type:Kanji} on the back.

But I might remember how to write 狩人 (hunter) and forget that it has that weird reading かりゅうど.

Voila—we click Cards… while we’re in Add Note with the note type we want to modify on, we put in a few simple instructions and now another card generates instantly off the same note where the front gives your native language cue and you’re asked to write かりゅうど, or type it or speak it out loud.

If you want, another one has {Kanji} on the front and {type:Reading} on the back—although in my experience outputting both the kanji and the word separately is so much more effective it makes this type of reading practice largely redundant (and you get it by going out and reading something real, anyway).

If you’re early in your study and you’re struggling to recall the kanji from memory and need some help at first, or maybe just help distinguishing synonyms whose nuances you don’t fully grasp yet? You can make a card type with {English definition} and {Reading} on the front and {type:Kanji} on the back. And then later on, you can just remove {Reading} from the front of the card and it will take the extra clue off every single cars you’ve ever made just like that. (Or you can put it back on all of them later just as easily.) Starting to see how creative you can get with even the most simple note, yet?

You can even add another field that asks whether you want the reading to be on the front of your kanji card or not.

Then you can have one card type generate if you put an x there, or the other type generate if you put nothing. You just wrap the front of the card in {{#Field}}{{/Field}} to make it generate only if Field has something in it, or wrap it in {{^{Field}}{{/Field}}} to make it generate only if Field is empty.

One way I’ve incorporated this is by adding kanji1 through kanji6 fields into my notes. So let’s say I want to remember 死亡推測時間—しぼうすいそくじかん, estimated time of death. I have cards that only generate if I fill these extra kanji1 kanji2 fields in on my note, and if I do that, I get six new cards.

One shows me ?亡推測時間 and asks me to write 死. The next shows me 死?推測時間 and asks me to write 亡. Or I throw 死亡 together into the first kanji1 slot and 時間 together into the last and only practice 推 and 測 separately.

Now what I can do is, if I feel like I don’t need these cards… the next time I see one, whether I’m on mobile or desktop, I can edit the note right there on the spot and just erase all the info from the kanji1+ fields, fill in the top Kanji field, and get back the one card asking me to handwrite 死亡推測時間 from memory all at once (and at this point it’s easy).

Or if I thought a kanji compound was easy when I made my card but I realize I keep fumbling only on a specific part in my reviews, I can quickly punch that part into the kanji1+ field and get a card to autogenerate testing me only on that bit.

Regardless: any way you think of implementing this, once you have it set up, it’s doing its thing for you automatically, from now on, forever, with every single new card you make.

Not only are you testing yourself on the information more atomically and devoting exactly as much time as you need to each specific aspect, you’re generating cards more efficiently because you might be typing in a few fields on a note and getting half a dozen atomic cards.

And now because the system knows the cards are testing the same information—they’re formed out of the same note so they’re what Anki calls “siblings”—two simple clicks will have it disperse them so it will never test you on different aspects of the same word on the same day. If I have this on all my decks and I make a new card for hunter and I practice writing 狩人 today, it will ask me to say かりゅうど tomorrow. More importantly, it will disperse the reviews, and you can disperse the reviews but not the learning cards if you want.

You actually can go even further from here with some add-ons that let you make deeper conditionals for some cards unlocking others—so you could do what I did with fill-in-the-individual-kanji cards that are programmed to automatically delete themselves and generate the “handwrite 死亡推測時間” card without a great deal of work. But everything above is so fast and easy that you’re way past the most bang for your buck at that point.

By the way: it’s extremely easy to make premade decks more useful with only basic knowledge about this too. For a simple example, if you download a deck that isn’t asking you to input anything, you can just edit the front of the card to have {type:Field} and it will check your input against Field. Want to go through the 2K deck but also handwrite the specific kanji you find in each new word?

Just open the note, click cards, and add a card type to the note with audio or whatever you want on the front and add {type:Whatever the Kanji field is called}. Boom. It’s done. All 2000 cards are going to generate for you, just like that.