Sources and method 典拠
Every source named, and the editorial decisions behind them - including what we deliberately left out, and why.
A learning app is a set of claims about a language. It is fair to ask where those claims came from, and most cannot answer - the material arrived from somewhere, was reworded, and the trail stops. This page is the trail.
Character histories
The etymologies follow classical 六書 the six categories of character formation analysis of the oracle-bone and seal forms - the traditional scholarly framework for asking what a character was originally a picture of, and how the modern meaning arrived from there.
This is the deliberate opposite of an invented mnemonic. There is a genre of kanji teaching built on made-up stories - a character is “a man with a hat next to a box” because that is memorable, regardless of whether the hat was ever a hat. Those stories work, briefly, and then they are a second thing to remember that happens to be false. Every one of the 2,136 Jōyō kanji here carries its actual history instead.
Where scholars disagree, the honest answer is that they disagree, and a real history has more of those than a manufactured one does.
Example sentences
Hand-written. They were originally researched with reference to Jisho.org, Tagaini.net, ejje.weblio.jp, iKnow.jp, kanjialive.com and Okjiten.jp - and we do not distribute their material. The sentences you see are ours.
Vocabulary
The word list draws its lexical facts from the Core 10k frequency deck: the word, its kana, its meaning, its part of speech, its JLPT level and its frequency rank. Facts about a language are not anybody's property.
What we deliberately did not take is that deck's creative content - its example sentences, its audio and its images. Those are somebody's work and not ours to redistribute, so they were skipped at import rather than quietly reused.
Two editorial rules fall out of that import. A hand-written word always wins over an imported duplicate. And a word containing a kanji outside the Jōyō set is skipped entirely - because a word can only open once you have mastered the characters inside it, and a word containing a character we do not teach could never honestly open at all.
Stroke order
Stroke order and the component groupings come from KanjiVG by Ulrich Apel, used under CC BY-SA 3.0. It is a genuinely excellent dataset and it deserves the credit by name rather than a footnote.
Pronunciation
Audio is synthesised with VOICEVOX：四国めたん. Synthesised rather than recorded, and said plainly here so that nobody mistakes it for a native speaker in a studio. It is consistent, it covers the whole corpus, and it costs nothing per play - which is why it can be given away.
Scheduling
Reviews are scheduled by FSRS, run on the server rather than in your browser, with randomised interval fuzz switched off so that the same history always produces the same schedule. The detail is on the scheduling page.
Why this page exists
Because a good deal of language-learning material on the internet now has no traceable origin at all, and the difference between a fact somebody checked and a plausible sentence is invisible from the outside. Naming sources is the only way to make it visible.
And because a claim about a language is only as good as the person making it: EdoQuest is built by a team led by Stijn, who has spent most of his life in Japan and reads the language fluently - which is why the etymology decision above was available to us at all. A named author is also somebody who can be corrected, which is the point of the paragraph below. Who we are 由緒.
If something here is wrong, we would like to know - corrections go to the box 目安箱, and a good letter still changes this app.