How to Learn Kanji: A Strategic Overview of Heisig, WaniKani, and Kanji-in-Context
Learning kanji is first a strategy question, not a technique question. Three macro-philosophies (kanji-meaning-first, structured radical-to-kanji-to-vocab spaced repetition, or SRS, and kanji-via-context) divide the 2,136-character jōyō (regular-use) load by order, unit of study, and time horizon.12 Picking among them up front matters more than picking a flashcard app. Every learner who has read three contradictory tutorials has already felt the cost of skipping that choice.
Overview
This article is a decision frame. It names the three macro-camps, explains how they differ on what gets memorized first and when readings enter, gives an explicit selection rubric, and shows how the canonical hybrids combine pieces of the three. Deep individual writeups of each method live in the linked sibling articles in this subdirectory.
The 2,136-character problem
The modern jōyō kanji list contains 2,136 characters. Cabinet Notification No. 2 of 30 November 2010 set that number by adding 196 characters and removing 5 (勺, 銑, 脹, 錘, 匁) from the previous list.12 The 文化庁 (Agency for Cultural Affairs) frames the list as 「現代の国語を書き表す場合の漢字使用の目安」, a guide for kanji use in writing modern Japanese, and as the permitted character set for official government documents.12
Reaching this baseline as a second-language learner is a multi-year undertaking by any of the macro-methods below. The U.S. Foreign Service Institute classifies Japanese as a Category V "super-hard" language. It requires roughly 88 weeks or 2,200 class hours of supervised study to reach ILR S-3/R-3 proficiency, with the kanji load named as a contributor to that classification.3
Scale, not per-character difficulty, is what makes the method choice strategic. Reading the same 2,136 characters is the deliverable of every camp described below; what differs is the order, the unit of study, and the time horizon.456
2,136 is the literacy baseline of compulsory Japanese education, not a guarantee that you can read a novel. Many published novels use kanji outside the list, and many high-frequency proper-name kanji are jinmeiyō rather than jōyō. The number anchors strategy; it does not bound the language.12
What this article covers (and does not)
Covered: the three macro-philosophies of second-language kanji study, the rubric for picking one, and the canonical hybrid patterns. Not covered: per-method deep dives, per-character etymology, and production-handwriting decisions, all of which live in dedicated articles.457
Claims here are anchored to primary sources for each camp's published self-description (Heisig's preface and publisher catalog page, WaniKani's landing page and knowledge base, Refold and AJATT-lineage published guides). Critiques are cited to the published critic, not to anonymous community opinion.485976
The three macro-philosophies
A taxonomy clears the field before any tool comparison. The three camps below are not three apps. They are three answers to the question "what is the smallest unit of kanji study, and when do readings attach to it?"
A. Kanji-meaning-first (the Heisig camp)
Remembering the Kanji (RTK) is a three-volume kanji-study series by James W. Heisig, published by University of Hawai'i Press. Volume I (writing plus a single English keyword) first appeared in 1977, and the 6th edition was issued in 2011 with updates for the 196 characters added by the 2010 jōyō revision.4810 Volume I teaches approximately 2,200 kanji using two design choices the publisher states explicitly: "the meaning and writing of the kanji but not their Japanese pronunciations," with readings deferred to Volume II.8410
The publisher's own description of the core principle is direct: the book begins "with writing the kanji because, contrary to first impressions, it is in fact simpler than learning how to pronounce them." Characters are organized by "component parts called 'primitive elements'" and recombined via "imaginative memory" stories.8 Volume II then teaches readings via phonetic-component groupings ("signal primitives"). Most online "I did RTK" reports refer to Volume I only.111012
The one-sentence thesis: assign one English keyword per kanji, learn 2,200 characters by meaning alone using primitives and stories, and handle readings as a separate later phase.48 For the deep walkthrough, see the dedicated Heisig method article in this subdirectory.
B. Structured radical to kanji to vocab SRS (the WaniKani camp)
WaniKani is a paid web SRS that sequences study in a fixed three-stage gated order: radicals first, then kanji, then vocabulary. Across 60 levels, it teaches roughly 2,000 kanji and 6,000 vocabulary words by completion.59 The cognitive-load rationale is stated explicitly on the platform's own forum: "the kanji are building blocks to the vocabulary," and one taught reading per kanji at the lesson level is the deliberate scoping choice "so as not to overwhelm you".13
The gating is mechanical. Radicals at a given level must reach the Guru SRS stage before that level's kanji unlock. Kanji must reach Guru before that level's vocabulary unlocks. This structure distinguishes WaniKani from a plain flashcard deck.913 The official knowledge base states that "it's possible to get through all 60 levels in just over a year," but calls that "breakneck speed." It names "a year and a half or two years" as the "much more reasonable speed."14
Pricing as listed on the WaniKani landing page is free for the first three levels, then $9 per month, $89 per year, or $299 one-time lifetime.95 The one-sentence thesis: WaniKani is a curated, gamified pipeline that teaches radicals, then kanji, then vocab inside one SRS, on a fixed 60-level path.59 The dedicated WaniKani sibling article in this subdirectory unpacks the SRS mechanics.
C. Kanji-acquired-via-context (the AJATT / Refold / mining camp)
All Japanese All The Time (AJATT) was founded by Khatzumoto in 2006. It advocates immersion-based language acquisition with "the focus on learning sentences instead of isolated vocabulary and grammar." Khatzumoto retired from AJATT in 2023, and Tatsumoto Ren now governs the AJATT method as the named successor.15 Refold is the modern restatement of the same immersion-first, vocab-in-context position. Its published kanji guidance is to "start by recognizing the most common 500 characters" and then to stop studying kanji as a separate object: "after this, you don't need to study Kanji separately."7
The operational principle is shared across the lineage. Refold: "by treating each word as a separate unit, you will improve your listening skills and reading ability in Japanese."7 Tatsumoto: "kanji do not exist in a vacuum." Recognition fluency is reached by meeting characters inside the words that contain them, not by drilling them in isolation, with the goal of pushing the learner "to the point where recognizing a kanji feels similar to recognizing a human face."6 Tofugu's taxonomy names this camp "Vocabulary & Experience" and lists it as one of seven recognized kanji-learning approaches.16
The one-sentence thesis: do not study kanji as isolated objects. Absorb them through high-volume vocabulary and reading, via mined sentence cards or pre-built decks.76 To operationalize this approach, J-Compass recommends Amenokori. It surfaces each kanji's reading and meaning inside the vocab entry it appears in. It also keeps everything on a single FSRS schedule across the jōyō range and the JLPT N5 to N1 vocabulary, so characters are learned in context exactly as the camp prescribes.1718 The dedicated kanji-via-vocab sibling article covers it in depth, and the choosing-your-first-resources article compares it against other SRS tools. The deep treatment lives there, not here.
How the three camps actually differ
The taxonomy above tells you the camps exist. The five axes below turn it into a side-by-side comparison you can act on.
| Axis | Heisig | WaniKani | Kanji-via-context |
|---|---|---|---|
| Day-one unit | one kanji + one English keyword | one radical name, later a kanji + one reading | whole word with kanji form, reading, meaning |
| Readings | deferred to Volume II | one per kanji from first kanji card; others arrive via vocab gating | bundled with the word from day one |
| Context | not supplied; assumes a parallel track | engineered vocab inside the SRS | the entire method |
| Typical pace | ~3 months fast-pass (Vol. I); +6–12 months for Vol. II | ~1 year fastest, 1.5–2 years reasonable | scales with reading volume; no fixed end |
| Cost floor | ~$30 book + free SRS | free to level 3, then $9/mo, $89/yr, or $299 lifetime | free Anki + Yomitan + community decks |
What you memorize on day one
Heisig: one English keyword per kanji, plus a story that assembles the keyword from primitive-element keywords. No Japanese reading is attached. The deliverable of one day's study is the ability to write the kanji on paper from the English keyword.48
WaniKani: the first lesson gives a radical name in English. Once enough radicals reach Guru, kanji lessons unlock, and the learner adds one English meaning plus one taught reading (typically on'yomi) per kanji. The deliverable is recognition of the kanji on the SRS card with that taught reading attached.5913
Kanji-via-context: a whole word, with its kanji form, its reading, and its meaning bundled into a single SRS card. The kanji is acquired as part of the word, not as an isolated object. The deliverable is recognition of the word in context.7617
When readings enter the picture
Heisig: not in Volume I. Volume I leaves the learner unable to read a single Japanese kanji aloud, by explicit design; readings are the scope of Volume II.4111012
WaniKani: from the first kanji card, but locked to a single reading at a time. The other readings appear later via vocabulary that uses them, which the gating mechanism enforces.513
Kanji-via-context: always, because the word is the unit. There is no separate "reading phase." The reading is part of every card from day one.76 The two-reading system that all three camps eventually have to internalize is unpacked in the on'yomi-versus-kun'yomi article.
How context is handled
Heisig defers context. Volume I supplies isolated characters with English keywords. It gives no example sentences, no compound words, and no grammar. The method explicitly assumes a parallel grammar-and-vocabulary track that the book itself does not provide.41012
WaniKani engineers context. The platform supplies vocabulary cards inside the SRS so that each kanji is reinforced by words that use it. The community phrasing is that "the point of the vocab in WK is reinforcement of the kanji pronunciation and meanings; it's up to you to go elsewhere to find it in context."1913
Kanji-via-context is built on context. The method is built around real-text input, mined from native material via Yomitan, jpdb, or curated decks like Kaishi 1.5k.76202122
Timeline and pace
Heisig fast-pass: 2,200 kanji over 90 days is roughly 24 to 25 new kanji per day. Over 180 days, it is 12 to 13 per day. The per-day arithmetic is derived from the Volume I frame count.4 Community-reported starting points for serious learners sit in the 20 to 25 new kanji per day range, with the caveat that this figure is forum-aggregated rather than supplied by a controlled study or by Heisig himself. The daily-kanji-routine article covers pacing math in depth.
WaniKani full path: the platform's own knowledge base names "just over a year" as the breakneck-pace floor and "a year and a half or two years" as the reasonable target.14
Kanji-via-context: timeline scales with reading volume. Refold's stated structure is an initial recognition pass on the most common 500 characters, after which kanji are absorbed as the learner reads.7 Without a fixed curriculum, the timeline has no upper bound but also no lower bound. The discipline cost, not the calendar, is the load-bearing constraint.6
The FSI's 2,200 class-hour estimate to reach ILR S-3/R-3 in Japanese is the supervised-study equivalent of roughly 88 weeks at full-time pace, with kanji study one component inside that envelope rather than the whole envelope.3
Cost
Heisig: one roughly $30 paperback (RTK Volume I, University of Hawai'i Press) plus a free SRS layer, such as Anki with a community RTK deck or the SRS built into kanji.koohii.4823
WaniKani: $9 per month, $89 per year, or $299 one-time lifetime, with the first three levels free.59
Kanji-via-context: free for Anki plus free decks. Kaishi 1.5k is publicly available under a community license, and Yomitan is GPL-3.0 open source. Paid options such as Amenokori Premium or jpdb add a modest cost.20212218
Inside the kanji-via-context camp, the paid option is a vocab-context SRS such as Amenokori ($150 lifetime) or jpdb, both of which sit below WaniKani's $299 lifetime price.18920 Currency and promotional pricing shift the comparison, but the structural point holds: a paid path exists inside every camp.
Selection criteria: which camp fits which learner
Five questions move the choice from philosophy to action. None of them is "which method works best." All three work to completion. The question is which method matches the constraints the learner actually has.
Your timeline
Six months or less to recognition: Heisig fast-pass is the only camp that hits a sub-six-month window for roughly 2,000-character meaning recognition. The explicit cost is zero reading ability at the end of that window.41012
Twelve to eighteen months: any of the three camps works inside this window. WaniKani's knowledge-base-named "reasonable" completion lands here.14 An RTK Volume I plus Volume II readings pass lands in this range. A disciplined vocab-context workflow accumulates substantial kanji recognition here as well.7
Two or more years with reading as the goal: kanji-via-context is the long-game default. The method scales with reading volume. It is the only one of the three that compounds outside the SRS.76
Your budget
A free-only constraint rules out WaniKani after level 3.59 Heisig (one book plus free Anki) and kanji-via-context (Anki, Yomitan, and community decks) both work for under $35 lifetime.42122
The paid kanji-via-context path sits below WaniKani's lifetime price: Amenokori's lifetime is $15018 versus WaniKani's $2999. The comparison shifts with currency and promotions. The structural point is that "paid" does not equal "WaniKani" in this space.518
Your learner personality
The "I need a track laid for me" learner (minimum decision overhead, daily lesson queue, social proof from level-up celebration culture) maps onto WaniKani.51319
The "give me the system, I will run it" learner (high discipline, comfortable inventing mnemonics, willing to absorb the review-peak load around weeks 4 to 6) maps onto Heisig.424
The "I learn by consuming the language" learner (high tolerance for early-stage slow progress, immersion material already lined up, willing to mine) maps onto kanji-via-context.76
Personality match has to survive past month three. Tatsumoto names retention drop-off as the structural risk for any isolated-kanji pass: retention "falls to 45 to 50 percent several months later" after completing an isolated deck, if reading does not reinforce it.6 The personality question and the related "what order do I study in" question are paired naturally with the school-order-versus-frequency-order debate, which is covered in the dedicated learning-order article.
Your prior CJK exposure
Speakers of Mandarin or other Chinese-writing-system languages already recognize most jōyō shapes. The primitive-installation step that Heisig formalizes, and that vocab-context implicitly requires, is largely done before they start. Kanji-via-context usually wins because isolated-keyword study is wasteful when the learner already has a character inventory.106
Speakers of Korean with hanja literacy bring partial character-shape exposure but no Japanese readings. A partial WaniKani pass (for systematic on'yomi installation) or an RTK Volume II pass (for signal-primitive phonetic groupings) maps onto this profile. This is a reasoned recommendation from the structural mapping, not a cited research finding.1156
Zero CJK exposure (the modal English-speaking learner) is the case where explicit primitive scaffolding pays for itself. Refold's "first 500 characters" floor exists for the same reason: below that floor, the component-pattern intuition that lets vocab encounters compound is too thin.7
Your end goal
Reading novels and news: kanji-via-context or WaniKani both deliver. WaniKani's 6,000-vocab payload during the kanji pipeline jump-starts the word-level recognition that newspaper and novel reading actually require.59 Kanji-via-context delivers the same outcome more slowly but more transferably, because the SRS cards are the words the learner is reading.76
Writing kanji by hand and passing Kanken: this goal calls for structured study with an explicit production component. WaniKani trains recognition, not production. The kanji-via-context camp does not train production at all. Heisig trains production from a meaning prompt and is the closest match, though Kanken's reading sections still require additional reading-phase study.46 The dedicated handwriting-debate article frames the production decision in full.
JLPT N3 or above by a fixed date: the JLPT has published no official kanji list at any level since the 2010 redesign,2526 but community per-level estimates put N3 at roughly 600 to 700 kanji and N1 at roughly 2,000. A hybrid that front-loads recognition (RTK fast-pass or the first 25 WaniKani levels) and then pivots to JLPT-targeted mining is the most defensible fixed-deadline route. The JLPT has no handwriting section, so the production-versus-recognition question is moot for this goal.26
Designing a hybrid strategy
Most serious learners end up running a hybrid. The three patterns below are the canonical shapes documented in the published method literature. The anti-patterns are the documented failure modes.
Pattern 1: RTK fast-pass, then mining
This is the AJATT-lineage hybrid canonized in the named-method literature. Tatsumoto names "isolated study (KanjiDamage or RTK) used as a preparatory bridge" as one of three legitimate paths to kanji recognition fluency. The other two are vocab-only from day one and the JP1K-style hybrid with hidden readings as training wheels.6
Khatzumoto's own AJATT guidance explicitly tells learners to "cut off" RTK early if it stops paying for itself rather than insisting on the full 2,200-frame pass, in the post "Early Cut-Off Kanji." The instruction is not "do not do RTK." It is "do not be a hostage to finishing RTK."24
Operational shape: two to three months of Heisig Volume I (writing plus keyword for roughly 1,000 to 2,200 kanji, depending on cut-off). Then make a permanent switch to sentence-mining with a vocab-first SRS such as jpdb, Anki plus Yomitan, or Amenokori, for readings, vocabulary, and real usage.62420 Volume II of RTK is widely treated as redundant in this hybrid because "kanji readings should be learned in context," and Pattern 1 learners typically skip it.6
Pattern 2: WaniKani plus parallel reading
This is the documented WaniKani-community hybrid. The learner runs WaniKani for the structured radical-to-kanji-to-vocab spine, but begins immersion (graded readers, manga, visual novels) early rather than waiting for completion. The forum-stated consensus is to "begin immersion immediately, but scale difficulty as you progress rather than waiting for a specific WaniKani level."19
The WaniKani-community framing of vocab's role inside this hybrid is that the in-platform vocabulary is "reinforcement of the kanji pronunciation and meanings; it's up to you to go elsewhere to find it in context."19 Pattern 2 takes that framing literally by supplying the external context in parallel.
Operational shape: WaniKani as the spine (fixed lesson queue, fixed gating), plus a separate input loop (Yomitan-supported reading and native media) running alongside from early levels. Grammar proficiency is the documented gating variable for the external loop, not WaniKani level.19
Pattern 3: Kanji-via-vocab with a radical-literacy primer
Skip the dedicated-kanji-deck step entirely. First spend roughly two weeks installing the high-frequency radicals and components so that mined characters do not feel like noise. Refold's published "first 500 characters" floor is the operational version of this primer; below that floor, vocab-first acquisition has too little component-pattern surface to compound on.7
The primer is content-light by design. It is not RTK. It is the radical and component vocabulary, not the full character pipeline. The compounding effect (semantic and phonetic component intuition) is what the primer unlocks, not specific characters.76 For the radical-frequency list itself, see the dedicated top-50-radicals article.
Operational shape: two weeks of radical study (any high-frequency-radical resource), then immediate vocab-first SRS plus mining, with kanji surfaced as a byproduct of the words in the SRS.7617
Hybrid anti-patterns
Running two heavy SRS decks in parallel (for example, RTK on Anki plus full WaniKani, both at peak new-card pace) is the first anti-pattern. The review tails stack. The documented failure mode is review-load collapse, not memory failure. Tatsumoto names the same shape from the opposite direction: 18 months on WaniKani is the published time-cost objection. Stacking RTK on top of it doubles the bill.27
Switching camps mid-stride at the 4-to-6-week peak-review window is the second. New-card grading at week 5 of any SRS-anchored pass coincides with the first major review pileup. Abandoning the deck in that window is the modal drop-out point reported across all three camps.6
Abandoning a meaning-only pass three weeks in because "no readings" feels wrong is the third. The Heisig design explicitly defers readings; the discomfort is the design, not a defect.41012 The decision to skip Heisig has to be made on day zero, not day twenty-one. Mid-pass abandonment loses the upfront investment without buying the reading-first speed.24
The supporting techniques every camp needs
Some layers of kanji study are method-agnostic. If you pick a camp, you still have to invest in mnemonics, a sustainable daily routine, and a lookup workflow for the day you start reading real text. Picking the camp does not pick these for you.
Mnemonics as a discipline, not a method
Mnemonics are a retention technique used inside all three macro-camps. They are not themselves a method. Heisig formalizes the imaginative-memory story per kanji (the publisher's own phrasing: "the powers of 'imaginative memory'" applied to "the setting for a particular kanji's 'story'").8 WaniKani pre-writes a canned mnemonic for every radical, kanji, and vocab item.59 Kanji-via-context uses mnemonics opportunistically, mostly for visually-confusable pairs.6 The dedicated kanji-mnemonics article covers when to invent and when to stop.
The canonical community story bank for Heisig learners is kanji.koohii. It hosts crowd-sourced mnemonic stories indexed against Heisig's keywords, with permission to use the RTK keyword index, alongside its own SRS.23 Pattern 1 hybrids commonly use kanji.koohii alongside RTK during the fast-pass, and the dedicated kanji-koohii article covers the workflow.
A daily routine you can sustain
The load math (new items per day, the review peak around weeks 4 to 6, and the burnout failure mode) is method-agnostic. Whichever camp you choose, the same pacing constraints apply. A new card today produces a review tail across multiple SRS intervals, and the peak load lands several weeks after the new-card start.623 The dedicated daily-kanji-routine article runs the math for each camp.
FSRS (Free Spaced Repetition Scheduler) is the open-source successor algorithm to SM-2. It models three memory variables (difficulty, stability, retrievability), is the default scheduler in Amenokori, and is an officially supported scheduler inside Anki from version 23.10 onward.281718 The scheduling algorithm is independent of the camp.
A lookup workflow for real text
Once kanji study leaves the SRS and meets a novel or news article, the bottleneck shifts to fast lookup. Yomitan is the in-browser pop-up dictionary that handles the lookup step. It is the named successor to Yomichan, open-source under GPL-3.0, and available on Chrome, Firefox, and Edge.21 jpdb is the pre-mined alternative. It extracts vocabulary from text the user pastes in, hosts prebuilt decks for thousands of anime, light novels, and visual novels, and "automatically synchronizes progress" across decks that share words.20
Both lookup tools work for any of the three camps. They become load-bearing once daily kanji study turns into daily reading.2021 The dedicated "looking up an unknown kanji" article covers hover, handwriting, OCR, and radical lookup in full.
Good to know
Match the failure mode, not the success story
Every camp's success story looks the same on a celebration thread. Every camp's failure mode is different. WaniKani's failure mode is subscription fatigue and a stall around the mid-levels.1319 Heisig's failure mode is the "I know 2,000 keywords but cannot read a menu" gap at the end of Volume I.624 Kanji-via-context's failure mode is slow early progress that feels like nothing is happening for three to six months.6
The selection question is not "which method works best," because all three work to completion. The question is which failure mode you can survive. Pick the camp whose failure mode you can sit with for six weeks without abandoning the deck.131962724
Treating "I finished RTK" as "I can read"
The "I finished RTK so I can read Japanese now" claim conflates Volume I with the full RTK arc. RTK Volume I delivers the ability to write roughly 2,200 kanji from an English keyword prompt. The learner still needs the readings (Volume II or a separate reading-phase track) and the vocabulary before they can read. Volume I deliberately omits readings. This is design, not defect, and most online "I did RTK" reports refer to Volume I only.41012
Treating an RTK keyword as a Japanese translation
Treating Heisig's English keyword as a translation is the other common Volume I misreading. "Life" is the memory anchor RTK assigns to 生, but the kanji's Japanese senses include 生きる (ikiru, "to live"), 生まれる (umareru, "to be born"), 生 (nama, "raw"), and 生 (sei, "student" or a suffix for a person), among others. Heisig assigns one keyword per kanji as a uniqueness anchor. The keyword is not a Japanese translation, and using it as one breeds errors in compounds.1012
Stacking two full-bore SRS decks in parallel
The full WaniKani lesson queue plus a full RTK Anki deck, both at peak new-card pace, is the canonical stacked-SRS anti-pattern. Review tails compound, not add. The modal failure mode of the stacked hybrid is review-load collapse in week five or six. That is a load-management failure, not a memory failure. The better shape is to pick one method as the spine and use the other only as a finite preparatory bridge (Pattern 1 or Pattern 3) before the spine starts.627
"Vocab-first" without immersion is not vocab-first
The kanji-via-context method assumes input that the learner can read or watch. Without immersion material, mining produces an empty loop. The learner ends up running a pre-built vocab deck with no contextual reinforcement, a kanji-via-vocab name without the kanji-via-context mechanism. AJATT's foundational claim is "doing as much Japanese as you can every day." The learner who runs vocab-first without input is not running the method.156
Why "the 2,136 number" keeps appearing
The 2,136-character target dates to Cabinet Notification No. 2 of 30 November 2010. That notice added 196 characters and removed 5 (勺, 銑, 脹, 錘, 匁) from the prior list.12 Older sources cite 1,945 (the previous list, 1981) or 1,850 (the post-war tōyō list, 1946). When pacing claims, deck sizes, or "complete coverage" promises disagree across articles or vendors, the year of the source's jōyō baseline often explains part of the mismatch.12
Production handwriting is a separate decision
Every camp above is a recognition method, with the partial exception that Heisig Volume I trains production from a meaning prompt. Production handwriting is a layer the learner adds or skips, not a method choice. The JLPT has no handwriting section, digital reading does not require handwriting, and the production-versus-recognition trade-off operates separately from the philosophy choice.426
"Kanji are impossible" is mostly a sequencing problem
The "kanji are impossible" framing is mostly a side effect of bad sequencing, not the characters themselves. The opposite claim, "I will just learn kanji in context naturally," is true at scale but unreliable at the six-month horizon. Refold itself names a "first 500 characters" recognition floor before pure context-acquisition starts compounding, which is why the method literature includes a primer phase rather than a pure cold-start.76 The realistic statement is simple: all three camps work, none of them feels good in week five, and the difference shows up at month six.
See also
- How to Learn Japanese: The Complete Roadmap from Zero to Fluency
- The History of Kanji: From Oracle Bones to the Jōyō List
- The Six Categories of Kanji (六書): Pictographs, Ideographs, and Phono-Semantic Compounds
- Kanji Radicals (部首): The 214 Kangxi Indexing System
- Phonetic Components in Kanji (音符): The Hidden Reading Hint in 75% of Kanji
- How to Predict the Reading of an Unknown Kanji Compound: The On+On Default, Jūbako, Yutō, and the Look-It-Up Bucket