Research & Data

This section holds only Nihongo to Japan's original research and publicly accessible aggregate data — things with a stated method, a sample size, a limitations section, and a citation format. Regular learning and travel articles are not published here.

⚠️ “Publicly accessible” is not the same as “openly licensed”. These datasets are not open data and carry no CC BY / CC0 licence. The figures may be cited with attribution and a link back; please contact us before reproducing our charts.

Datasets and reports

Japanese Learning Answer Dataset — August 2026

Published 2026-08-14 · version 1.1.0 | 56,613 anonymous answer events, 1,053 anonymous learner identifiers, 11,987 questions. Covers 2–13 August 2026 (12 days) — not a full-year study.

Raw accuracy broken down by skill axis, JLPT level and 39 knowledge points, with Wilson 95% confidence intervals, a minimum cell size rule, an analysis-set decomposition and a full limitations section. Downloadable as JSON and CSV, with APA, MLA and BibTeX citation formats. ⚠️ Not openly licensed: figures may be cited with attribution; please contact us to reproduce the charts.

Headline findings (quiz-context first attempts): grammar patterns 53.7% versus vocabulary 92.1%; the particle が 60.9% versus を 78.7%.

Read the full report → | JSON | CSV

How Taiwan Learns Japanese 2026

Published 2026-07-21 | Public data from the Japan Foundation, official JLPT statistics and JNTO, cross-referenced with this site's own search and learning-behaviour data.

What it answers: how Taiwanese learners actually study Japanese — the shape of lookup demand, why learners stall on nuance rather than vocabulary, and the shrinking classroom versus growing self-study picture.

Read the report →

How we handle data

An introduction for educators and librarians is here. For the site and its editorial standards see About and the editorial policy.