As I approached WK level 40 or so, I had a problem that a lot of folks here are pretty familiar with: how do I actually go from “remembering kanji and words and grammar” to “consuming actual native material comfortably enough to reinforce it”? How do I find a way to meaningfully motivate myself to stare at sentences until the meaning starts to osmose into my brain? How do I pass through the final bootstrapping stage? How do I get my reading fast enough that I can actually parse whole sentences at speed instead of wracking my brain on each character? How do I do all this with absolutely zero classroom structure (I have never taken a Japanese class). And to note, I’m not someone who is normally very good at the whole “healthy motivation” thing, so I’m starting, relatively, on hard mode here.
So, since I found something that worked for me, I figure I’ll share it for anyone else interested.
The first step was to find something I wanted to read, preferably something that “contained its own motivation” and could be consumed in small bites. In this case, I picked a mobile gacha game for a franchise I enjoy, since those tend to be pretty good at manipulating human compulsions to make you want to keep playing. Some critical aspects: the material isn’t available in English, and the material isn’t available in a form where I can copy and paste it.
Next, I transcribed everything I wanted to read from the game (mostly story dialogue, etc)-- by keyboard – into DeepL, word by word. If I didn’t know a kanji, I looked it up, if I didn’t know a reading, I looked it up, etc. Then I typed it by hand once I had looked it up (no copy/paste!!!). Important thing here: transcribing it all manually forced me to rigorously read every single mora of what I was reading. No skimming. I had to read every last bit! And since copy/paste was unavailable, there was no temptation of cheating.
You can imagine how much this reinforces the readings of kanji!!
The next part of this: DeepL is a pretty good translation engine… for a computer. It’s still not very good though. It will give terrible translations a lot of the time. Most of the time, it’s pretty obvious when it does. But it will often hint heavily at the overall meaning and structure of a sentence, even if it gets things wrong. So, in this way it provides “help”, while strongly incentivizing me to not rely completely on it because of how much it garbles things at times.
So, after many hours of doing this (and enjoying playing through a game I wanted to play), I noticed I was parsing out the meaning from sentences before I even looked at the output from DeepL. My brain knew DeepL was often going to give me garbage a lot of the time, so it was starting to figure it out as I transcribed it, just on its own. Of course, I still only understood some percentage of the content, but that’s fine; perfect is the enemy of good.
When I started, reading Japanese was a painful “one character or maybe word at a time” affair that could take a minute for a sentence. I think my first 1000 characters of reading/transcription took at least 2-3 hours.
After about the first ~10,000 characters of reading, I think my reading speed had at least doubled.
After ~40,000, it had increased by another factor of 2 or so.
After ~100,000, I could read fast enough that, so long as I didn’t encounter an unfamiliar kanji, I could often read a sentence fast enough to say it at-speed out loud, with minimal pausing. This felt absurd compared to where I started! I was also reading through multiple sentences at a time without looking at the “translation” window from DeepL – not because I chose to, but subconsciously, because my brain just didn’t “need” to look at the translation window.
I also started adding new words I learned from new content into an SRS, as well as periodically looking up new grammar and adding it to BunPro, and so on.
Some other things I picked up during this time using the same method:
- Animal crossing!
- Various manga and doujinshi I had been sitting on
- Final Fantasy XIV (yes, you can set it to Japanese! it’s quite intimidating at first, but now I can read through a lot of quest dialogue unassisted).
I’ve now reached the point where for a lot of new content, I just drop the DeepL step entirely, though I go back to it whenever I find my brain being tempted to skim just a little bit too much and I want to force it to actually read every character.
I think I’ve spent about ~100-150 hours on native material at this point (in the past ~1.5 months). I’ve actually slowed down on WaniKani quite a bit because at this point I’m running into far more new kanji through native material than WK; I’m at level ~50 now, and by volume, most of the new kanji I’m running into are actually not in WK at all, so I gotta give myself time for all those.
The most fascinating experience in all this has been when I realized I was starting to understand subconsciously – that is, reading my way through a sentence without any “conscious thought” about the sentence, and yet at the end having a good idea of what it said despite not making any conscious “attempt” to read it. It just… happened! Just like English! But this still only happens sometimes, and it took 100+ hours of reading native material to really start to get to this point, at least for me.
Anyways, this is what worked for me. It successfully tricks my brain into actually reading the entire sentence mora-for-mora, provides tons of reinforcement on learning readings, and gives me some “training wheels” to help Actually Get There. Maybe most importantly, it lets me read things I actually want to read, so I don’t have to slog through material I’m not actually motivated to look at.
A critical requirement for this process, I think, is knowing enough kanji that you can comfortably work your way through most things without having to constantly be stopped in your tracks. In this way, it works really well for people who take a more “kanji-first” approach, and less well if you find yourself with 5 unfamiliar kanji every sentence.
