The Heisig Method: How Remembering the Kanji Teaches Meaning Before Reading
The Heisig method for kanji is the study procedure laid out in James W. Heisig's Remembering the Kanji (RTK). RTK is a three-volume series whose first volume teaches the writing and an English keyword for roughly 2,200 characters, while deliberately postponing readings to a later volume.12 For a learner comparing methods, the decision hinges on accepting that trade: spend three to six months building writing recall from meaning before learning to read a single kanji aloud.
Overview
What "Remembering the Kanji" actually is
Remembering the Kanji is a three-volume series by James W. Heisig, a permanent research fellow at the Nanzan Institute for Religion and Culture in Nagoya.34 Volume I was first published in 1977. The 6th edition appeared in March 2011 from University of Hawai'i Press and was updated to include the 196 characters added by the 2010 jōyō revision.132
The three volumes split the work into distinct passes rather than teaching kanji in one go.
| Volume | What it teaches | Approximate coverage |
|---|---|---|
| I | Writing and one English keyword per character; no readings | ~2,200 kanji12 |
| II | Official on'yomi and kun'yomi, grouped by signal primitives where possible | The Volume I set52 |
| III (with Tanya Sienko) | A further set of writing/keywords plus their readings | ~800 additional kanji, for ~3,000 across the series62 |
When learners say "the Heisig method," they almost always mean the Volume I procedure specifically.78
The four assumptions the method is built on
The method rests on four design choices. These distinguish it from frequency lists, school-grade lists, and vocabulary-anchored systems.
First, meaning and pronunciation are treated as two separable obstacles. Heisig's rationale is that asking the brain to bind shape, meaning, and sound at the same time is double work. Isolating writing and meaning into one pass lets each piece settle before the next.79
Second, every kanji is treated as something that can be broken into "primitive elements." These include traditional Kangxi radicals, full kanji used as parts of other kanji, and stroke clusters Heisig names himself when no traditional name exists.102
Third, a vivid invented story links the primitives' keywords to the target kanji's keyword. The story is the connective tissue. The act of building it is what stabilises the memory.1210
Fourth, one English keyword per kanji is enough as an anchor. Heisig deliberately picks unusual or narrow English words so that the 2,200-kanji keyword set stays collision-free.28
How the Heisig method works
Primitives: the building blocks
A primitive element, in Heisig's terms, is any recurring visual unit used to build RTK kanji. The Wikipedia summary captures the relationship: kanji "are analyzed by components, Heisig terms these 'primitives,' which may be traditional radicals, other kanji themselves, or a collection of strokes."2
This is not the same set as the 214 Kangxi radicals (部首, bushu). Some primitives are traditional radicals. Some are full kanji repurposed as components. Others are stroke clusters Heisig named himself, with his "fish-guts" primitive as the most cited example.102 The radical-versus-component distinction is the cleanest way to see where Heisig sits in the lexicographic landscape.
A learner who memorises the Heisig primitive name for a stroke cluster and then tries to look the kanji up in a 部首 (bushu, radical) dictionary will often fail to find it. The conventional dictionary indexes by Kangxi radical, not by Heisig's coined primitives.10
The order of presentation is built around primitive accumulation. Simpler characters and their primitives are introduced first, so later, denser characters can be built from pieces the learner already knows.102
Keywords: one English word per kanji
Each of the ~2,200 kanji in Volume I is paired with exactly one English keyword. Each non-kanji primitive also receives its own keyword.12 The keywords are intentionally distinctive, sometimes archaic or narrow, so the whole keyword set stays collision-free across the volume.28
A keyword is a memory hook, not a translation. The keyword "man" for 男 is a single anchor. The kanji itself appears in Japanese compounds covering "male," "son," and various kinship readings, none of which the single English word captures.27
The case learners cite most often is 生.
生、生、生きる、生まれる、生える、生、生7
"sei, shō, ikiru, umareru, haeru, ki, nama: a sample of 生's attested readings."
The kanji's senses span "life," "birth," "raw," "growth," "student," and "fresh," among others.78 One English keyword cannot carry that semantic load and is not intended to.
Stories: imaginative memory in practice
Heisig calls the central technique "imaginative memory": the faculty of building mental images vivid enough to anchor a keyword to its primitives.12 The learner is expected to invent their own story for each kanji rather than reuse a canned one. The act of invention is what stabilises the link.102
The most-repeated worked example in RTK exposition is 男.
The kanji 男 decomposes into 田 ("rice field") above 力 ("power"), with the kanji's own keyword being "man."2 A representative RTK-style story is: a MAN is the one who uses his POWER in the RICE FIELD.2
The decomposition aligns with the etymological account from oracle-bone analysis. In that account, 男 depicts a strong arm or plough (力) at work in the irrigated paddy (田), denoting the person who does the field labor.11 The mnemonic and the historical etymology happen to converge here, which is part of why 男 is the canonical worked example.112
A second worked example shows the same mechanism with a primitive Heisig coined himself. The kanji 肌 ("texture") decomposes into the body-flesh primitive ⺼ ("part of the body") plus 几 ("wind"). A typical RTK story is ever notice how the TEXTURE of your face and hands is affected by the WIND?7
Mature Heisig learners report that the elaborate story scaffolding falls away after weeks of consistent review. What remains is direct keyword-to-kanji recall; the story has done its job and dissolved. Heisig himself frames stories as "temporary scaffoldings."9
The order of learning
Volume I orders the ~2,200 kanji by primitive build-up: a kanji is introduced only after all of its primitives have been.210 This order is explicitly not based on frequency, school grade, or JLPT level.27
The practical result cuts both ways. Visually complex kanji whose primitives appear early, such as 鬱 ("gloom"), can land relatively early in the sequence once their components are in place. Common kanji whose primitives appear later, such as 校 ("school"), land much later than they would in a school-grade-ordered curriculum.210
The typical RTK timeline
The pace math
Heisig's own preface says he personally learned the writing of "some 1,900 characters" in roughly a month of intensive study after arriving in Japan.1 This is the source of the popular "three months to 2,000" claim that circulates in the learner community.12
Plain arithmetic against the 2,200-kanji Volume I target gives the two reference paces.
| Target window | New kanji per day | Profile |
|---|---|---|
| 90 days | ~24 to 25 | The 3-month sprint pace12 |
| 180 days | ~12 to 13 | The 6-month sustainable pace12 |
Community-reported starting points sit around 20 new kanji per day for learners with "a couple of hours" of daily study. The same reports explicitly advise dropping to 15 per day rather than burning out at 25 or 30.12 Treat these numbers as community-reported norms, not as research findings. For the underlying load math and a generic three-block schedule, see the daily kanji study routine.
Review load is the real cost
The pace math counts only new cards. Under any spaced-repetition schedule (Anki, kanji.koohii's built-in SRS, or similar), each new card generates a review tail. That tail peaks several weeks after introduction, when early kanji come due at the same time as fresh ones.13
An SRS layer is assumed, not optional, in nearly every documented RTK workflow. SRS means spaced repetition software, such as Anki or kanji.koohii.137
The community-canonical failure mode is not "I cannot remember the kanji" but "I fell behind on reviews in week 5 or 6." That is an SRS-queue management problem driven by the review tail, not a memory problem. Reducing the new-cards-per-day rate before week 5 prevents it. Trying to grind through the backlog after it appears rarely works.12
What "finished" actually means
Finishing Volume I means: given a printed English keyword, the learner can write the corresponding kanji on paper from a kana-free prompt, for approximately 2,200 kanji.110
It does not mean the learner can read Japanese. After Volume I, the learner still does not know any on'yomi or kun'yomi readings of the kanji they have just learned to write.2107 This is the explicit design of Volume I, not an oversight. Readings are the scope of Volume II.52
What RTK does well
Writing recall from a meaning prompt
The deliverable unique to RTK is production from meaning: given an English keyword, the learner can write the kanji on paper.27 Most competing kanji curricula (WaniKani, JPDB, and textbook-anchored systems) train recognition with vocabulary context. They do not produce this output reliably.78
A shared component vocabulary
Once the learner has internalised the roughly 300 primitives that recur across Volume I, each new kanji becomes a small puzzle whose pieces are already known. The cognitive-load drop reported in learner accounts is consistent with the building-block ordering of the volume.27
Decoupling memory work from grammar work
Because Volume I uses no Japanese reading, the meaning-and-writing pass can run in parallel with, or even ahead of, grammar and vocabulary study. This makes the method manageable for learners with prior CJK exposure or a heavy upfront time budget.714
The criticisms
No readings until Volume II
The strongest factual criticism is structural. Finishing Volume I leaves the learner unable to read a single Japanese kanji aloud.278 Volume II teaches readings via phonetic-component groupings (signal primitives). However, learner accounts widely report that many people who finish Volume I never start Volume II, leaving them with writing recall but no reading ability.72
Keywords are not Japanese meanings
A keyword is a memory hook, not a translation. Heisig's keyword "man" for 男 does not capture 男's full Japanese semantic range. The keyword "life" for 生 captures only one of that kanji's many senses.278
A learner who has internalised "生 = life" and then encounters 生ビール (nama bīru, "draft beer," literally "raw beer") will misread the sentence. The keyword is a memory anchor for retrieval of the character's shape, not a Japanese meaning the learner can substitute into native text. The fix is to treat the keyword as a label only and learn Japanese senses through vocabulary study.7
No vocabulary, no grammar, no context
Volume I gives the learner isolated characters with English keywords. There are no example sentences, no compound words, and no grammar.127 The method explicitly assumes a parallel grammar-and-vocabulary track that the book itself does not provide.714
Opportunity cost
The strongest opinion-side critique, articulated most clearly by the AJATT and immersion-first camps, is opportunity cost. The three to six months a learner spends on a meaning-only pass is time they are not spending on listening, speaking, or reading Japanese.147 Khatzumoto, the originator of the AJATT method, recommends RTK but explicitly advises learners to cut it off early if it stops paying for itself.14
Academic critique: meaning without sound
J. Marshall Unger, Emeritus Professor of Japanese at Ohio State University, critiques the broader frame behind the method: treating characters as visual symbols of meaning detached from spoken Japanese.159 His summary, paraphrased on Language Log, is that Heisig's method is "a thorough-going technique for memorizing the equivalent of a dictionary," not a route to learning Japanese as a language.9
Victor Mair, Professor of Chinese Language and Literature at the University of Pennsylvania, makes the same structural point on Language Log: the Heisig method "placed characters on a pedestal of visuality / iconicity without integrating them with spoken language."9 Unger also acknowledges Heisig's own self-description of the stories as "temporary scaffoldings, which fall away as the student learns more and more Japanese."9
Who the method fits, and how to run it
Good fits
Learners with a heavy upfront time budget (sabbatical, semester abroad, gap year, unemployment window) are good fits if they can absorb the two-to-four-hour daily review tail.127 Learners who have hit the kanji wall in conventional study also tend to do well, especially if they want a structured climb with a tangible deliverable: handwritten production from a keyword prompt.78
Learners with prior CJK exposure (speakers of Mandarin or other Chinese-writing-system languages) start the method with most primitives already familiar as shapes, which shortens the timeline.27 Learners who plan to read native Japanese intensively and want a writing-recall foundation as a stepping stone are the fourth common profile.78
Poor fits
Learners on a 30-minute-a-day budget will not finish the volume in any realistic timeframe, especially if they need every minute to build sentences and listening hours.714
Learners who only need recognition for digital reading (browser overlays, popup dictionaries) and never plan to handwrite get less return on the production training.7 The JLPT itself does not test handwriting at any level, which removes one common justification for the production work.16
Learners chasing a near-term JLPT deadline are also a poor fit. RTK's order does not align with JLPT levels, so a learner cannot pre-load the kanji their next test will ask for.7
The standard companion stack
Three components recur across nearly every documented RTK workflow.
The SRS layer: Anki with a community RTK deck (several decks track the 6th edition order specifically), or the SRS built into kanji.koohii.137
The shared-story bank: kanji.koohii's community-contributed stories, indexed against Heisig's keywords. The site has operated since 2006 from Belgium and runs with explicit permission from Heisig to use the RTK keyword index.1317
A tracking spreadsheet or app, for daily new-card count and review backlog.12
Common hybrid strategies
Three hybrid patterns dominate the reports.
The "fast-pass then mining" pattern: complete Volume I in three months, drop Volume II, and pick up readings via vocabulary cards mined from native input.1412
The partial-Heisig pattern: do the first 500 to 1,000 RTK frames to break the primitive barrier, then switch to a vocab-context method.147
The full-three-volume pattern: rarer, mainly pursued by learners who want exhaustive coverage and accept the extra year. Most online "I did RTK" reports refer to Volume I only.7
Good to know
"Primitive" is Heisig's word, not a linguistics term
Many Heisig primitives are not 部首 (bushu, Kangxi radicals). Some are stroke clusters Heisig coined himself, with his "fish-guts" primitive as the most cited example.102 Confusing Heisig primitives with the conventional radical system causes lookup trouble later, when the learner tries to find a kanji in a 部首-indexed dictionary. The mental model to keep: Heisig primitives are a teaching inventory, not a dictionary one.
Treating an RTK keyword as a translation
A learner who assumes the keyword is the Japanese meaning will misread native text. The keyword "life" anchors 生 for retrieval. The kanji's Japanese senses span at least the following.
生きる、生まれる、生、生7
"ikiru (to live), umareru (to be born), nama (raw), sei (student / a person, in compounds)."
Heisig explicitly assigns one keyword as a uniqueness anchor. The keyword is not a Japanese translation.27
Equating "I finished RTK Volume I" with "I know kanji"
Finishing Volume I means the learner can write ~2,200 kanji from an English keyword prompt. It does not mean the learner can read Japanese. Readings, vocabulary, and grammar are out of scope for Volume I by design.127 The honest summary is: "I finished RTK Volume I, so I can write the characters; I still need to learn the readings and vocabulary."
"I forgot the stories" is not the failure mode it looks like
Mature RTK learners commonly report that the elaborate story scaffolding falls away within weeks of consistent review, leaving direct keyword-to-kanji recall as the residue.9 Heisig's own framing treats stories as temporary scaffolding. A learner who panics because they "cannot remember the story" for a kanji they are nevertheless writing correctly is seeing the design work, not a memory leak.
Volume II is a separate decision
Most online "I did RTK" reports mean Volume I only. Volume II teaches readings through phonetic-component grouping and feels structurally like a different book. It does not use "imaginative memory" the way Volume I does.5210 A learner planning to "do RTK" should decide before starting whether they mean Volume I, Volumes I and II, or all three.
The 6th edition and the 2010 jōyō revision
Volume I's 6th edition (March 2011) was updated to include the 196 characters added to the jōyō list by the Cabinet 告示 (kokuji, official notice) of 30 November 2010. That revision set the jōyō list at 2,136 characters after also removing 勺, 銑, 脹, 錘, and 匁.1181920 Older RTK editions list approximately 2,042 kanji and predate jōyō entries like 鬱. When two learner reports disagree on what "finishing RTK" covers, the edition is part of the explanation.
Production versus recognition
RTK trains handwriting production from a meaning prompt; digital Japanese reading (websites, apps, e-books) mostly demands recognition with dictionary support.7 The JLPT itself does not test handwriting at any level.16 Whether the production training pays for itself depends on the learner's lifestyle. A learner who plans to take handwritten notes, fill out paper forms, or write Japanese by hand benefits. A learner who only reads on a screen with a popup dictionary captures less of the value.
See also
- The Six Categories of Kanji (六書): Pictographs, Ideographs, and Phono-Semantic Compounds
- Semantic Components in Kanji (意符): What the Water, Person, and Tree Radicals Tell You About Meaning
- How Many Kanji Do You Need? A Realistic Count
- Beyond Anki: SRS Tools and Approaches Compared
- Top 50 Kanji Radicals by Frequency: The 70% Coverage List for Jōyō Kanji
- How to Look Up a Kanji You Don't Know: Hover, Handwriting, OCR, and Radical Lookup