How NOT to Learn Chinese Characters

« previous post | next post »

There are many ways NOT to learn Chinese characters, but one that I just found out about today is probably the worst, even worse than T. K. Ann's Cracking the Chinese Puzzles.  It was written by Alison Matthews ("a statistician who has worked in the oil, aviation, tourism, medical and software industries") and Laurence Matthews (author of books that claim to help you find Chinese characters fast) and is called Learning Chinese Characters:  A revolutionary new way to learn and remember the 800 most basic Chinese characters.

You can find the Matthews' miraculous tome on Google books here.

If you start leafing through the book, as I did, you will find on any given page hilarious explanations such as the following:

"tree 木 + several 几 = machine 机":  It took several trees to provide enough wood to make the parts for the huge machine.           

This is accompanied by a picture of a wooden contraption behind which are four trunks of trees that have been felled and beyond that three trees that are still standing.  Whereas the Matthewses gushingly enjoin us to "see how the 'several trees' have indeed been felled to make the large 'machine' that is taking shape," this is actually a pictophonetic (or semantosyllabic) character in which MU4 is the semantic indicator and JI1 (not JI3 ["several") is the phonophore.

"wrap  勹 + a drop 丶 = ladle 勺":  When he had wrapped it up he put a drop of perfume on the package even though there was only a ladle inside.        

This tortuous explanation is accompanied by a picture of two hands in front of a belly; the right one is holding a perfume dropper out of which has come a drop of perfume that is wiggling in mid-air, while the left hand is holding the handle of a ladle that is wrapped in cloth or paper and tied with a string.  The ladle appears to be resting on a flat surface, or possibly partially submerged underwater.  In fact, this character goes back to the period of the oracle bones (earliest stage of the writing system, circa 1200 BC), at which time it depicted a ladle with a drop of liquid in it.  At the time of its creation, the character had absolutely nothing to do with "wrapping"; the idea of "wrapping" is an artifact of a later stage of evolution when the ladle was transformed into what became Kangxi radical 20 勹.

These are just the first two characters that I happened to turn to as I perused the book.  Looking further, I find that there are even more outlandish explanations for many other characters.  If one tried to use this method to learn 8 characters, it might work, but if one attempted to learn all of the 800 characters in the book this way, it would be a horribly frustrating experience.  If one did not go insane in the process, at the very least one would have lost hundreds, if not thousands, of hours in vain hopes of mastering the strokes, sounds, and meanings of so many characters in this absurd fashion.  And, if one should ever be so foolhardy as to try to employ the method of the Matthewses to learn 8,000 characters, one would certainly go stark, raving mad.

I pity anyone into whose hands this book falls and who actually tries to learn the characters by using it.  The only thing it will do for you is turn your brain into mush.

Hat tip to Jonathan Smith


  1. Andrew said,

    October 14, 2009 @ 2:49 pm

    Maybe I'm confused, but does the book claim that these constructions are genuine etymologies, or does it merely proffer them as mnemonics to help remember everyday uses? Because it sounds like they are doing the latter. If so, I think I might like a clearer explanation of the problem.

  2. Jonathan Badger said,

    October 14, 2009 @ 2:54 pm

    Exactly. I remember learning all sorts of silly mnemonics to remember the genders of German nouns, but it isn't the same thing as claiming that they were the reason for the grammatical gender. Not having tried to learn Chinese, I can't comment on whether these particular rules work to memorize characters, but I can attest that mnemonics do help in language learning in my experience. As long as nobody claims they are anything more than memorization tools, what's the problem?

  3. Jake said,

    October 14, 2009 @ 3:01 pm

    As a learning aid, this doesn't seem unreasonable to me. Many methods claiming to improve memories (especially learning lists, playing cards, etc) use similar sorts of techniques, and the outlandish scenarios serve to fix them in mind. I made good inroads into learning German vocab at school using a similar technique.

    Sure, it's not for everyone, but this criticism seems a bit harsh.

  4. Cale said,

    October 14, 2009 @ 3:04 pm

    Although I agree that learning 800 (or 8000) characters through tortuous mnemonic devices is probably a bad idea, I'm not sure it matters that they are not etymologically correct. As an example, I remembered that 'ostrov' meant 'island' in Russian because "Ostriches live on Islands." This is obviously false, has nothing to do with the etymology of the word, and really ostrich doesn't even sound much like ostrov, but I remember the word (now without having to think about the device).

    Not knowing anything about Chinese, I suppose it is possible that learning symbols this way would seriously impair your ability to learn other symbols. I have no idea.

  5. David said,

    October 14, 2009 @ 3:12 pm

    This looks similar to Hesig's method in "Remembering the Kana" and "Remembering the Kanji" for Japanese (You can see "Remembering the Kanji" complete with it's introduction and some sample chapters on Google Books). The basic idea is to create short stories that act as mnemonics for remembering the characters (not even all of the meanings or sounds, just the basics so that they are easier to remember). These stories do not necessarily have any relevance to what the radicals involved necessarily actually mean.

    For example, for "hi" (日, word given is "day") we have the story "This kanji is intended to be a pictograph of the sun. Recalling what we said in the previous frame about round forms [no circular shapes, so we represent them with boxes], it is easy to detect the circle and the big smile that characterize our simplest drawings of the sun–like those yellow badges with the words, 'Have a nice *day*!'"

    Not meant to say that the kanji actually represents the sun smiling down, but it is meant to help the student recognize and remember the kanji based on that story.

  6. Leonardo Boiko said,

    October 14, 2009 @ 3:12 pm

    Is this a Chinese version of Heisig’s “Remembering the Kanji”, or is it the other way around? Heisig employs a similar method; he assigns a single mnemonic keyword to each character component, and then weave convoluted stories to tie together those mnemonics to an English meaning-keyword, one per character. The book is quite popular, and many Japanese learners vouch for it (probably, I think, because it’s more fun than the traditional method of writing the characters repeatedly while drilling half a dozen on– and kun-yomi each).

    I hate these mnemonic methods with a passion, especially because it’s all a big lie. It’s teaching you a fake system with fake explanations for fake components grouped in fake meanings. It has nothing to do with the real history of characters nor with the actual motivations for their makeup, it simplifies their semantics to the point of falsehood, and it subscribes heavily to the ideographic myth, ignoring the phonetic components altogether (and 85% of the characters have a phonetic component…)

    Myself, I’ve reached a kind of middle-way to study: first, I drill whole-words from real, non-textbook Japanese, often whole sentences from a story so that the context is clear; second, I supplement the drills with careful study of Henshall’s dictionary, which actually tells the real story and etymology of characters. It’s often complex and not easy to commit to memory, but in practice there’s no reason to; I just want to get to know the characters better. At least I’m burning neurons on the actual system behind Chinese characters, not some made-up synchronic pop-tale.

    I like to think this way of study will help me later when I’m finally ready to learn grass script or Old Japanese or kanbun, but I admit that’s a post-hoc justification —the real reason for the extra work is, I like the truth better than lies.

  7. David said,

    October 14, 2009 @ 3:16 pm

    This is the BEST way to learn Chinese characters. I learned 2000 Kanji during a few months using exactly those kinds of silly mnemonics, provided by the excellent book "Remembering the Kanji".

    Of course these days I can hardly remember what the mnemonics were, but without them I would have despaired and given up long before reaching reading proficieny.

  8. Sili said,

    October 14, 2009 @ 3:43 pm

    Like the "pro nun", "Auntie, see! Dent!" described by Liberman a while back, these convoluted mnemonics have always struck me as being more work than just rote learning.

    But obviously that's not the case for everyone, and as attested some people swear by them.

    Of course, I'm 1) horribly unimaginative, 2) laze as all fsck and 3) rubbish at languages.

  9. Jake K said,

    October 14, 2009 @ 3:44 pm

    To call a system like this a "lie" is a bit of a stretch. I've studied Japanese for about 2 years and lived in Japan for 5 months, and I've read Remembering the Kanji. I never studied that book in an in depth way, but I do know some people who have used it as their primary kanji book and are WAY more literate in Japanese than I am. Judging by the assessments of my friends who have used Remembering the Kanji and are able to read Japanese very well, much better than people who have used other methods, methods like these can be an effective way to learn Japanese characters. That doesn't mean they work for everyone, and it also doesn't mean that the system is supposed to provide education in the roots of Japanese words. It doesn't claim to do that in any way. For people who want to actually use kanji in their everyday lives, methods like these can be great. People who want to study the history and semantics of the Japanese language should use another book and not criticize people who learn languages in other ways.

  10. Leonardo Boiko said,

    October 14, 2009 @ 3:47 pm

    @Andrew and others: to see what is the problem with mnemonics, imagine you’re required to learn ALL French or German words through a mnemonic. Imagine further that the mnemonics completely ignore not only the real etymology, but also the living grammar and relationships between words.

    This is basically what the mnemonic books do for hànzì. Writing systems don’t exist in a vacuum and, contrary to popular opinion, Chinese characters don’t stand up for a simple isolated «meaning» that you can symbolize with a single English word. After all the time you spend to learn 800 or 2000 mnemonics, all you’re left with is an arbitrary, simplistic mapping of a «meaning» for each character. But there’s a real system, a «grammar» behind these characters, driven in no small part by Chinese phonetics, and the more you ignore it, the more trouble it will give you later. You’re much better off spending your time with real Chinese or Japanese IMO. (related discussion).

  11. Nathalie said,

    October 14, 2009 @ 3:51 pm

    My Chinese teacher once told us that the character for electricity was a picture of a kite in the rain (Benjamin Franklin holding a kite with a key on it during a storm). I told all my friends about it, until one native Chinese speaker informed me that that was only a mnemonic, not the actual history of the character.

  12. John Lawler said,

    October 14, 2009 @ 4:00 pm

    There really isn't any best way. People are incredibly variable in their learning strategies. This may be just the ticket for some folks, and a horrible waste of time and effort for others. This is especially true for adults, and most especially true for illiterate adults (in Chinese/Japanese, that is) who are trying to learn to read logographs in a language that's not their native one.

    Executive summary: Whatever works for you, works. Try'em all and see.

  13. Ray Girvan said,

    October 14, 2009 @ 4:15 pm

    Reminds me of Pelmanism.

  14. Gregory Karber said,

    October 14, 2009 @ 4:34 pm

    I memorized the US state capitals like this, and never once labored under the assumption that the reason Bismark was the capital of North Dakota was because it was named after the bees marking the Norse duck quota.

    Mnemonics certainly don't allow someone to have a firm grasp of the grammar of a language, no doubt, but if it helps someone form a foundation to base later understanding upon, then I say more power to them.

    Maybe I'm confused about what you're upset about.

  15. Chris said,

    October 14, 2009 @ 4:45 pm

    Heisig's book is quite explicit about the mnemonics being fake etymologies, and the English keywords you associate with them being very very approximate. It's just a crutch to cram a 2000-character dictionary into your head, to eliminate that hurdle during the later, normal learning of Japanese vocabulary. You're not meant to forever associate "several trees" and "machine" or whatever — that's just meant as scaffolding that falls away once you learn it the normal way. How well it works, I don't know, because I never got to that point. But the just-so arbitrariness of the mnemonics is a feature, not a bug: they are designed to be more regular, vivid, and distinguishing of subtle differences than the real etymologies would be.

  16. Carl said,

    October 14, 2009 @ 5:42 pm

    So what is the best resource for learning Chinese characters? (Besides taking a university course.)

  17. Jake said,

    October 14, 2009 @ 6:06 pm

    @Leonardo Boiko:

    "to see what is the problem with mnemonics, imagine you’re required to learn ALL French or German words through a mnemonic."

    That's just being silly. No teacher would advocate a single learning method to be applied to everything – it's inappropriate and ineffective. But to infer therefore that mnenonics are therefore useless is plain barmy. The overwhelming opinions and experiences in these comments are testament to that.

  18. Franz Bebop said,

    October 14, 2009 @ 6:18 pm

    I agree with Gregory and Chris and the others, above.

    Mnemonics are not supposed to make logical sense. In fact, sometimes it's better if they make no sense at all, that way, they are easier to remember.

    When I was a kid leaning musical notation, they told us to remember the notes of the five lines on a staff (E-G-B-D-F) using the mnemonic "every good boy does fine." What logical sense does that make? None. But it's easy to remember, which is why it was useful for a young beginner.

    If the image of the wooden contraption doesn't help you, then you can find some other mnemonic which does help you. Mnemonics are highly individual.

    @Leonardo: It's not an either-or thing. Use of mnemonics does not preclude learning characters (or words) using etymology or grammar.

  19. TB said,

    October 14, 2009 @ 6:43 pm

    I quite agree with John Lawler. There is no best way, and there is no worst way, there are just ways that work for you or don't. It is much easier for me to remember a picture than a sound, so for me the kanji itself is a mnemonic for the word. Kid's books written only in hiragana are really hard for me! (I wish French and Spanish were written in Chinese characters; remembering their vocabulary was always impossible for me in high school.)

    This kind of mnemonic would never work for me, but unlike the authors of many language textbooks (including Heisig), I know that everyone learns differently, and would never claim my way is the "best way" or the "one true way" or whatever.

  20. Jim said,

    October 14, 2009 @ 7:02 pm

    "I've studied Japanese for about 2 years and lived in Japan for 5 months, "

    This isn't about kanji. Any nonsensical system will work for thiose as weel as any other nonsesniical system, because using them for Japanese is nonsensical.

    This is a stupid way to learn hanzi. The best way is to memorize enough simple ones with phonetic values that you can start to learn more complex ones with similar phonetic values and then build form there. So you start with kou3 – mouth and go on to gu3 – ancient and then to qi2- strange and then to qi3 – straddle; ride a horse and so on.
    Obviously this only works in Chinese because the rebuses that form the system are Chinese.

    A side benefit to this method is that when you run across a "wrong character" you are prepared to puzzle it out, because native-speakers will erroneously substitute homophonous characters. As a side side benefit, that may even clue you into what their home dialect is by guessing from the miss-substitutions.

  21. TB said,

    October 14, 2009 @ 7:17 pm

    See, David and Jim both claim to know the best way, and this being Language Log, a place for scientists who love Science, I wonder if they can show me what studies or data or evidence they have to back up their claims. Because I have just realized that I have nothing to back up my claim that there *is* no best way. I should say, rather, that "as far as I can tell, there is no best way."

  22. adjusting said,

    October 14, 2009 @ 7:38 pm

    An interesting post that addresses many of the criticisms against using Heisig's method for learning chinese characters:

  23. James said,

    October 14, 2009 @ 9:16 pm

    The Introduction to this book (viewable on Google books) is very explicit that these are stories with accompanying images, and NOT etymologies. The stories and images are simple memory techniques, not "truth" about the characters themselves. The authors mention Heisig's work directly: "The systematic use of imagery for learning the meanings of characters was pioneered by James Heisig (for Japanese characters)."

    This learning method is most definitely not for me, however.

    I managed to learn all 1945 常用漢字 in Japanese using the old black Hadamitzky & Spahn, along with newspapers written for elementary school kids, mystery novels, magazines, and newspapers written for adults. Actual context speeds up the learning process considerably, though I also spent hours making my own kanji flash cards on colored construction paper sitting on my tatami in the evenings.

    Using an electronice dictionary actually SLOWED my progress, because it became a crutch. The act of leafing through a book, seeing plenty of other characters surrounding the one I was looking up, added to the learning process. I started to remember entire pages of the dictionary, and of course writing the characters out provided a tactile experience which drove the learning home.

    In other words, a variety of methods focusing largely on language in actual use worked great. Go figure!

  24. Brendan said,

    October 14, 2009 @ 9:44 pm

    I'm a little more sympathetic to the idea of mnemonics for absolute beginners, but this clearly gets into diminishing returns before too long. (醫: "The 医 doctor started 殳 hitting the 酉 sauce pretty hard, so he was promoted to an administrative position.")

    That said, the best way to learn characters, in my experience, is just to write the suckers again and again and again until they get into muscle memory.

  25. GAC said,

    October 15, 2009 @ 12:56 am

    I've made my own mneumonics for particularly difficult characters. I still think it's better to learn them in the context of words.

    As far as I know Japanese do fairly well with kanji. While it's not ideal for that language and it might be better for Japanese to fall back on it's native syllabic systems (and may very well as less common kanji are forgotten by more people).

    However, they are far from ideal for Chinese, as most of the rebuses are not based on any modern Chinese language. Your example betrays this — do you really expect most people to link 奇 with 口 (I can't even see a 古 in there). I don't, but I know it's part of 奇怪 and I remember it associated with that word most of the time.

    Chinese characters, in whatever language they are used, are a "good enough" system, as are most writing systems.

  26. Lorelei said,

    October 15, 2009 @ 1:09 am

    For what it's worth, I think Jim was using "口" and "古" as one set, and "奇" with "骑" as another set.
    At least I hope that's what he was doing…

  27. Richard Wein said,

    October 15, 2009 @ 4:04 am

    I can't comment on Chinese, which I don't know at all. But I found this sort of mnemonic system useful when I started learning German. Of course, I didn't use it for words that had clear relationships with words I already knew (in English, German or some other language). In those cases it was easier to make use of that relationship. But for words that were completely unfamiliar, mnemonics provided a way of memorising the word more quickly and securely than rote repetition. After a time I found myself using the mnemonic system less and less, but I would still use it occasionally with words that I was having difficulty remembering.

    I don't think anyone would suggest that such a system should be used for learning every word (or even most). But it can be very useful for getting started.

  28. Troy S. said,

    October 15, 2009 @ 4:45 am

    This reminds me of stories of Euposean missionaries to China, who would make rebuses out of the character system, noting, for example, that the character for "righteous" looked like the character for "I" under the character "lamb" and drawing from this a mnemonic "I am righteous under the Lamb." Obviously this is not true or serious etymology, but the missionaries found it nevertheless a useful means of relating the ideas of Gospels, a sort of cultural foot in the door. Of course, the stories themselves may be aprocryphal, have you heard of such a thing?

  29. Carl said,

    October 15, 2009 @ 5:37 am

    Troy, the stories aren't apocryphal. The missionaries aren't doing it as mnemonic though. They're crazy enough to believe that they have uncovered the hidden proof that the Chinese were secretly refugees from Noah's flood or whatever. See their idiotic website:

  30. Carl said,

    October 15, 2009 @ 5:44 am

    Here's a little more insanity:

    And a rebuttal by sane people:

  31. Fergal Daly said,

    October 15, 2009 @ 6:26 am

    I've used Heisig to learn kanji. I spent years of futzing around and trying and then forgetting and basically coming to the conclusion that I will never read and write Japanese without devoting several years of my life to it. In the last 9 months I've learned to write from memory almost 1900 kanji. I have a full-time job (and then some), as does my wife and we have 2 kids under 3, so my time is extremely limited. I managed to learn 600 kanji in 4 weeks while my kids were out of town but often I'm lucky to do 30 in one week. There is absolutely no way I would have been able to learn this via drilling in the time available. With drilling I would have to write each character 100s of times. The average number of times I have written each of my 1900 kanji is 10!

    Now, learning a new Japanese word is almost as simple as learning a new French or German word. It has a pronunciation and a "spelling". The link between the two is weaker in Japanese than in French however in Japanese the "spelling" has a strong tie to the overall meaning which is often missing in French or German (knowing Latin helps in some cases but not all).

    As for this system missing the phonetic information that is present in Chinese (and muddled in Japanese), I'm not so sure. The phonetic information comes from the phonetic "half" of the character and (due to the system's ordering) all components are learned before the characters that contain them so you should already know the component pronunciation and spot the similarity when you get to characters that use it.

    The fact is that Heisig's system works incredibly well for some people and in many of the negative reviews I've seen of it, the reviewer has had a fundamental misunderstanding of it, either of the method (e.g. complaining that the stories are not historically accurate) or in their expectations ("what? this doesn't teach we any grammar!").

    Does this Hanzi book give stories for all 800 characters? If so I'd say that's a mistake. One of the under-emphasised values of Heisig is that it trains you to make your own mnemonics. This is even more important for Chinese where you have many thousands more characters to learn when you get to the end of the book.

    Basically if you want to assert that the method is terrible then you need to explain why it has worked very well for so many people (or show that they are in fact deluded and produce a class of students who have done much better with some other method).

  32. Aaron Davies said,

    October 15, 2009 @ 6:31 am

    Random semi-related question: do people using computers to read Chinese generally use the same fonts and font sizes as people reading alphabetic languages? In my browser's default setup (14-point Times New Roman), a character like"醫" is an almost impenetrable smudge.

  33. GAC said,

    October 15, 2009 @ 7:26 am

    I looked it up. 義 (simplified variant 义) does in fact mean "righteous" (according to nciku and it does look a bit like 羊 (sheep) on top of 我 (first-person pronoun), though I'm not clear enough on different forms of radicals to know whether that is really a 羊 or another similar character.

    Anyway, yeah, I don't see anything in the Answers in Genesis crap. I don't know enough to fully refute it, but it's pretty obvious they don't have a clue as to how Chinese characters work (they never once mentioned a phonetic element in the parts I skimmed through).

  34. Randy Alexander said,

    October 15, 2009 @ 8:17 am

    I can't really knock the method per se, but I have to agree with the above commenters that mnemonic success is an extremely personal thing. A book of mnemonic images or stories is likely most useful only to the person who designed those particular mnemonics. Designing your own mnemonic for a given item is a helpful part of the memorization process.

    For an understanding of the craziness behind mnemonic processes, I highly recommend The Mind of a Mnemonist, by A. R. Luria.

    I have a book similar to Heisig's, but that gives real etymological explanations, called A Guide to Remembering Japanese Characters (Kenneth G. Henshall).

    If you're going to make a mnemonic story or picture for a character you want to learn, why not incorporate the real etymology?

    Some other wonderful books that could aid in doing this are Tracing the Roots of Chinese Characters: 500 Cases, and Evolutionary Illustration of Chinese Characters, both by Li Leyi.

    @Brendan: I'm really surprised that you say you learn characters by writing them over and over. I constantly tell my students that doing so is almost useless — especially when they try to use that way to memorize English words. I guess what's best is really a very personal thing. I'm sure there are hundreds of characters that I know how to write, but have never actually written. From now on, I guess I'll start telling my students to only write characters over and over if they really feel it helps them personally.

  35. John said,

    October 15, 2009 @ 8:55 am

    I'm with John Lawler and Randy Alexander. Actually, in Remembering the Kanji, Heisig himself says that his mnemonics work best when devised by the learner, and thus doesn't give stories for the characters in the final section of the book.

    The real "problem" is that many learners are not fascinated by Chinese character etymology. Should they have to be in order to learn Chinese characters?

  36. Ken Brown said,

    October 15, 2009 @ 10:26 am

    Gregory Karber said: "I memorized the US state capitals like this, and never once labored under the assumption that the reason Bismark was the capital of North Dakota was because it was named after the bees marking the Norse duck quota."

    But… but…

    Mnemonics good, yes, obviously. But why not make them real ones?

    If you had learned that the railway company renamed it after Bismarck to try to sell land to German immigrants that is *just* as memorable as the bees and the ducks. In fact *more* so because it fits in with other things you might know. Like who Bismarck was. (Which might help get the spelling right…) Or that the town was (according to Wikipedia) once called "Missouri Crossing". Or that Bismarck became chancellor of united Germany after the 1870 Franco-Prussian War which locates the renaming of the town in the 1870s…

    Then you can remember that episode of the West Wing when Donna went to Bismarck to hear some Democrat party members make a (fictional?) pitch for changing the state name to attract more tourists. So they are still at it? Or the story that Erik the Red (or was it Leif Eriksson?) named Greenland "Greenland" as a selling point to attract immigrants, after "Iceland" hadn't gone down too well. And that Leif went to Vinland and started a long traditions of various nordic settlers in North America… including places like North Dakota.

    Then instead of having to remember a list of 8 or 80 or 800 isolated little stories you remember things that form a connected system. And your mind and it becomes easier to remember and learn more things as you do on. If you've got a connected system, a structure of thought, then the bigger and more interconnected it gets the easier it is to learn new things that fit in with it. But if you just have a list of disconnected facts or stories or images then the bigger it gets the harder it is to add to it.

  37. Nick said,

    October 15, 2009 @ 10:29 am

    Good mnemonics can be useful. If nothing else, using mnemonics when you're starting out will teach you common components very thoroughly. After you know the components, your mnemonics can be as simple as "Doctor Hit Sauce". I will totally remember that. I couldn't remember 碧 ever until I started thinking of it as "King Whiterock."

    I think the takeaway of the article is that the mnemonics provided in this book are bad mnemonics. Even if you ignore phono-semantic breakdowns, you can at least get the components right or take advantage of pictographic characters.

  38. John said,

    October 15, 2009 @ 10:38 am


    What you say is true. The problem is that real character etymologies are incredibly complex, full of red herrings, and not even entirely known. Etymologists take delight in the labyrinthine web of associations, but the average student just finds it all frustrating.

  39. Chris said,

    October 15, 2009 @ 11:57 am

    I'm no linguist (but I am a teacher) and it seems to me that learning *real* explanations like Ken Brown's only really works if the learner has enough background to understand the real explanations. If you are eight years old, you probably don't know anything about where railway land sales fit into US history and you've almost certainly never heard of the Franco-Prussian War. So bees and Norse ducks might be more memorable, however wrong.

    The big problem with mnemonics is, however, that no matter HOW many times or HOW plainly you say that the mnemonics you're offering have *nothing to do* with real history, some people will take them as such anyway. Witness how many people learn their "history" from historical movies…. {sigh}

  40. Rick S said,

    October 15, 2009 @ 12:14 pm

    Ken: "If you've got a connected system […] then the bigger and more interconnected it gets the easier it is to learn new things that fit in with it."

    That's how I've always learned. But the flip side of that coin is, if you haven't built up that "structure of thought" through the habits of a lifetime of curiosity (and many people who might need to learn Chinese will not have), you'll have little for new facts to hook up with. In that case, there is an advantage in absurdities: For some reason they're more memorable–maybe because there's a playful element to them–than more mundane facts.

    Not that I would recommend anybody over the age of 8 adopt a lifelong Weltanschauung based on silly stories, mind you. But if your job is transferring you to Shanghai in three months, this might be the way to go.

  41. Rick S said,

    October 15, 2009 @ 12:22 pm

    Whoops! The articulate Chris beat me to the punch!

  42. Jim said,

    October 15, 2009 @ 1:10 pm

    "See, David and Jim both claim to know the best way, and this being Language Log, a place for scientists who love Science, I wonder if they can show me what studies or data or evidence they have to back up their claims. "

    I only meant best for me. I take it for granted people know that learning styles vary. Mnemonics may very well work for other peole. For me they don't even work for English.

    "However, they are far from ideal for Chinese, as most of the rebuses are not based on any modern Chinese language. Your example betrays this — do you really expect most people to link 奇 with 口 (I can't even see a 古 in there). I don't, but I know it's part of 奇怪 and I remember it associated with that word most of the time."

    I boo-booed – qi2 is derived from he2, not from gu3. As for the rebuses, qi2 is not such a stretch from kou3. Zhao Yuan Ren said that when it came to palatals, the native intuition was that they related to velars in Mandarin. In fact in some areas of the north,velars athat are palatized in Beijing are stiil velar – Dalian is one of these places, I think.

    When it comes ot phonetic resemblances that may be a strecth for antive or even foreign speakers, it just seems a lot more economical and a lot less effort to teach some general phonetic principles and to point out those similarities, which happen to have the advantage of a factual historical basis, than to teach a bunch of arbitrary, unrelated and superficial mnemonics. It's more a coment on teaching strategy than anything else.

  43. Brendan said,

    October 15, 2009 @ 1:10 pm

    @Randy — the nice thing about character practice is that it doesn't require much at all in the way of attention — you can do it while you watch TV or listen to the radio. 字帖 character tracing workbooks are particularly good for this — it's not like using flashcards (which I've never had much use for, though plenty of very talented language learners I know swear by them); in my experience, all it requires is semi-attentive repetition to really lock stuff down.
    I wouldn't advocate rote learning for students picking up English vocabulary simply because it's a completely different ballgame: a Chinese student who needs to write in English can just sound out a word, or at the very least rely on a spellchecker; not so for the student of Chinese.

    In general, while there's nothing wrong with character mnemonics per se, I tend to think that anyone relying on them past the first two hundred characters of Chinese (or Japanese) is really going about things the wrong way: they're great when you're just starting out and can't relate characters to other characters, but once a student has a hundred or two hanzi/kanji under their belt, they should already have begun noticing patterns — radicals, shared components, common phonetics — even if their teacher hasn't pointed them out. (Which is the case shockingly often, and which is why John DeFrancis' books should be required reading before students even begin to deal with characters.)

    But mnemonics are fun, and apparently lucrative, judging by the number of books available for foreign learners in the Wangfujing Foreign Languages Bookstore. In that vein, I humbly submit:

    天: HEAVEN only knows how this MAN 人 is still standing when he's got TWO RAILROAD TIES 二 running through him.

    閒: "Me and the boys was just RELAXIN' next to the SWINGING DOORS 門 at the FULL MOON 月 Saloon when the Sheriff come in and bust up the place."

    無: The DANCER IN THE SWISHY GOWN HAS NO coordination, which would probably make shorter robes a better choice.

    計: The idiot savant ACCOUNTANT keeps TALKING 言 about train schedules and multiples of TEN 十.

    讓: You use your WORDS 言 to tell people to move the BOXES 口 containing CLOTHING 衣 out from under the COVERING 亠, but they are so utterly confused at the appearance of a GIGANTIC POUND SIGN 井 that they just stand around until you ALLOW them to leave.

    I could go on for days.

  44. J. Marshall Unger said,

    October 15, 2009 @ 1:54 pm

    Please see Chapter 5 of my book _Ideogram: Chinese characters and the myth of disembodied meaning_ for a fuller discussion of mnemonic systems.

  45. dr pepper said,

    October 15, 2009 @ 2:09 pm

    Ok, if the ideogram concept is a myth, what about the claim that people with mutually incomprehensible spoken languages can read what each other writes?

  46. Kragen Javier Sitaker said,

    October 15, 2009 @ 4:47 pm

    dr pepper asks, "what about the claim that people with mutually incomprehensible spoken languages can read what each other writes?"

    Well, sure, that's a widespread phenomenon. Scholars in Europe before 1800 all wrote in Latin so they could understand each other, even though their spoken languages are mutually incomprehensible. Spanish and Portuguese, modern dialects of Iberian vulgar Latin, are mutually incomprehensible when spoken (although if you live in Uruguay or southern Brazil you probably understand them both, especially when people speak slowly) but the divergences in grammar and vocabulary are much smaller than the divergences in pronunciation, so it's pretty easy to puzzle out a text in one if you can speak the other.

    One issue is that reading is more forgiving of lower levels of competency than listening is. If you read 25% slower than an expert reader, it's barely noticeable. But if you listen 25% slower, you miss one out of every four words when people are speaking at native speed, preventing you from following anything more than the vaguest outlines of the conversation.

  47. David said,

    October 15, 2009 @ 4:56 pm


    Actually you the phonetic elements can be useful for Japanese also (but of course only for on-yomi): For example 同、銅、胴 all have on-yomi "dou".

    This kind of thing is also a useful way to learn kanji, just not as useful as the mnemonic method. But of course the best thing is to use both and whatever other method you can think of.

  48. Fresh Sawdust said,

    October 15, 2009 @ 6:39 pm

    What may be one person's mnemonic slog may well become just another person's "structural" (radico-graphetic) bare-bones breakdown of a character – witness Brendan's jocular mnemonic story for 醫 ("The 医 doctor started 殳 hitting the 酉 sauce pretty hard, so he was promoted to an administrative position") , as opposed to Nick then simply firing off the 'names' almost of the clear components thus: 'Doctor – Hit – Sauce' (which are all radicals, in fact – well, 'sauce' is rather 'wine', but that's not to detract from the joke). Which to my mind is basically the difference between the likes of Heisig (or – shudder – the Matthewses' work) and a gem like Foerster & Tamura's Kanji ABC (essentially a list of radicals, along with appreciably more components intermediate between the level of radical and full character that the authors have termed 'graphemes'), which worked pretty well for me when I was cramming some Chinese.

    Incidentally, isn't 醫 simplified (still) to 医 in Chinese at least? (The reason I ask is not to nitpick, but to raise the boring old question of which would be easier to "learn" – recall so as to manually write, as opposed to just recognize and read? – within a semantico-mnemonic approach).

    Generally, I can see the value and use of mnemonics when learning "ideographic" (that is, graphically quite complex) orthographies like Chinese, but they seem much less useful when it comes to languages written in alphabets; certainly, the "keyword" or "pegword" technique really does start to seem lame busy work when, due to the supposed difficulty of learning that 'katana' means 'sword', one is presented with images of samurai kittens waving swords (that's CAT-a-um-er-GEDDIT!? Genius, right? RIGHT? :). (Can't remember exactly where I read this example, but it was in an otherwise solid and convincing book on lexis in the context of ELT).

    @Dr Pepper: (Following DeFrancis's use of 'dialect' only for the mutually intelligible varieties of the widespread northern regionalect that is Mandarin) I haven't studied any other regionalect of Chinese than Mandarin, but I was rather under the impression that the southern regionalects have, in order to represent certain items of their speech, had to develop or use certain characters that are not (or at least no longer) recognized as part of the national written standard. Wouldn't that suggest that certainly the regionalects are not only speaking but in those certain respects also writing a mutually unintelligible (or at least mutually exclusive or "excluded") language in spite of the standard?

  49. Jay Casey said,

    October 15, 2009 @ 8:44 pm

    Actually, while I haven't seen this particular book, the method you describe works quite well for most people. It's a well-known mneomic technique that uses absurd mind pictures to remember things that would otherwise be hard to remember – like number sequences and lists. Look up "Roman room" technique.

  50. fs said,

    October 15, 2009 @ 10:15 pm

    @Nathalie: Sure, 電 kinda looks like a kite in the rain. Hilariously, the PRC's simplified character is just 电, removing the rain altogether and making the exercise pointless! :)

  51. Bob Violence said,

    October 16, 2009 @ 5:36 am

    I looked it up. 義 (simplified variant 义) does in fact mean "righteous" (according to nciku and it does look a bit like 羊 (sheep) on top of 我 (first-person pronoun), though I'm not clear enough on different forms of radicals to know whether that is really a 羊 or another similar character.

    Dunno about 羊, but I suspect 我 is just a phonetic — its reconstructed Old Chinese pronunciation is *ŋˤajʔ, as compared with *ŋ[r]aj-s for 義. Of course this hasn't stopped others from coming up with clever all-semantic etymologies like this and this.

  52. Fergal Daly said,

    October 16, 2009 @ 6:05 am

    @Fresh Sawdust

    What is the difference between Kanji ABC and Heisig? After a quick skim of Kanji ABC it seems very similar to Heisig but without the mnemonic technique and possibly with more historically accurate names for the graphemes (Heisig calls them "primitives").

  53. stevelaudig said,

    October 16, 2009 @ 7:44 am

    oh dear, I bought the book. it was delivered yesterday. I give it a whirl.

  54. stripey_cat said,

    October 16, 2009 @ 8:29 am

    I personally hate mnemonics: you have to learn the wretched mnemonic, which is usually more complicated and just as arbitrary (and as unmemorable!) as the original word (or phrase, or formula, or whatever). Plus they're actually often very easy to subvert or mis-remember (after I taught an entire lower class "in fourteen-hundred and ninety three, Columbus sailed the deep blue sea" my primary school teachers were ready to kill me); I've also done that kind of thing accidentally to myself a few times! I can just about see it working for a few dozen items, but once you're trying to learn hundreds, good old repetition is the way to go in my experience.

  55. Wun Cheng said,

    October 16, 2009 @ 8:38 am

    I have just brought the book, i've been using mnemonics to remember all sorts of things including chinese characters. making them up takes time and effort and this book should speed up my learning process.

  56. mike said,

    October 16, 2009 @ 11:30 am

    As someone who has been learning Chinese characters for more then three years now, my sympathies are with anyone who makes the attempt, although to be clear I find it a sometimes fun and always rewarding process. I think mnemonics are often very helpful. In fact, many Chinese characters etymologies are natural mnemonics, my recent favorite being ”漏“ meaning "leak" and in addition to three-dots-water on the left, is the character for rain under the character for residence.

    That said, THE MNEMONICS IN THIS BOOK ARE TERRIBLE. Really, really awful. I call your attention to ˘旦“ for dawn, which is obviously the sun coming over the horizon. But because the mnemonic for one "一" is "unicorn's horn" (that's right, this book was a mnemonic for "one") the story is "Picture the sun rising and glinting on the unicorn's horn at dawn."

    They get worse from there.

  57. Fresh Sawdust said,

    October 16, 2009 @ 12:56 pm

    @Fergal Daly: I haven't actually used Heisig's kanji books, but I have read e.g. Unger's description of the mnemonic techniques employed (see the plug Unger gives his 'Ideogram' in the comments above, completely justified as it's well worth reading!). [I guess I'd probably prefer to get Henshall's book for (Sino-)Japanese study, but for Chinese study I'm still happy to make use of stuff like Karlgren, Wieger, Wilder & Ingram (even though they all employ Wade-Giles, not that that's an insurmountable problem), rather than splash out on Heisig's more recent Remembering Trad/Simp Hanzi (and certainly not on the Matthewses' book, at least not for more serious study)]. Anyway, your take on the difference between Heisig's Japanese works and Kanji ABC sounds about right to me!

    Perhaps the most obvious advantage to Kanji ABC is financial – that it is "just" the one volume, whereas you'd need to buy the first two of Heisig to cover as many kanji (roughly 2000 i.e. the Joyo kanji). (Going by the back cover to volume 3 of Heisig available on Google Book Search).

  58. Fergal Daly said,

    October 16, 2009 @ 1:50 pm

    @Fresh Sawdust

    Hesig Vol 1 covers 2042 Kanji writing only. Vol 2 presents a similar technique to learn pronunciation but many people say that this is not worth it. Vol 3 covers a further 1000 but by the end of Vol 2 you should be well able to add new Kanji with Heisig's help.

    I get the impression that he created Vol 2 afterwards as an application of the same technique to a different problem. I think learning in context and natural memory works pretty well for pronunciation whereas writing needs far more application of something – be that mnemonics, repetition or whatever you like youself.

  59. Fresh Sawdust said,

    October 16, 2009 @ 2:16 pm

    Ah, OK – thanks for that, Fergal! If I weren't so lazy (or glued to my PC and LL) I'd've had a quick re-read of Unger before saying anything about Heisig, but what you've said is helping ring the right sort of bells now – that vol 2 covers more the pronunciation of the same 2000-odd characters.

    I'm not entirely sure how serious the problem of multiple readings is for kanji (assuming that is at least part of the reason for Heisig's decision to do a two-volume job of "the essentials") because my Japanese simply isn't yet good enough to be speaking let alone reading much, but a totally separate volume (by any author) for the pronunciation certainly of core Chinese hanzi would seem slight overkill.

    Like you say, 'learning in context and natural memory works pretty well for pronunciation' at least, but the majority consensus on this thread seems to be that more complex orthographies do require that little bit more (which again in your words could be 'mnemonics, repetition or whatever you like youself'). :)

  60. Fresh Sawdust said,

    October 16, 2009 @ 2:32 pm

    Actually, has anyone used Heisig's more recent books for learning Chinese hanzi? Does he maintain the division, into separate volumes, of graphic form "versus" pronunciation that there seems to be in his kanji books? (I'll try to see if enough info can be gleaned from GBS previews, but just thought I'd post the question here as it occured to me. Sorry for "replying to myself" with the two posts in quick succession).

  61. perspectivehere said,

    October 16, 2009 @ 8:02 pm

    Memory Palaces of Matteo Ricci by Jonathan Spence shows how Ricci memorized prodigious amounts of characters using memory techniques similar to (but not the same as) the one described here.

  62. Fresh Sawdust said,

    October 16, 2009 @ 8:17 pm

    Apologies for yet another post, but just wanted to quickly say that I missed but have now followed adjusting's link (about a third of the way down the thread here), and have also read reviews of Heisig's RTH on, all of which has pretty much given me the answer my previous question! 8)

  63. Ken Brown said,

    October 18, 2009 @ 10:13 pm

    Perspectivehere: "Memory Palaces of Matteo Ricci by Jonathan Spence shows how Ricci memorized prodigious amounts of characters using memory techniques similar to (but not the same as) the one described here."

    Yes, but Ricci was already trained in the Renaissance Art of Memory. In effect you put in the work of building your connected systems of thought beforehand – traditionally by remembering or imagining places or buildings, the "Method of Loci" that goes back to ancient times – and then fit what you want to learn into your carefully pre-prepared system. So you have a sort of ready-made catalog or index in your memory into which you can slot anything you want to remember.

    As described by Frances Yates in her books Giordano Bruno and the Hermetic Tradition and especially The Art of Memory. I first heard of it in John Crowley's wonderful novel Little, Big where it is a major part of the plot. Crowley put me on to the Yates books (and also The Architecture of Country Houses by Andrew J. Downing which I didn't really realise was a real book until I found a Dover facsimile in a shop)

    The techniques were used both for memorising and for training the mind and for generating new connections between things. And they were used by all sorts of people such as Thomas Aquinas, Ramon Lull, Giordano Bruno, Michel de Montaigne, Theresa de Avila, and many others. And as Crowley said it was made obsolete by the invention of the card index. It exists today mainly as a party trick by extreme memorisers and stage "magicians". Or as a way for prisoners in solitary confinement to occupy their minds by building and maintaining dream castles of the mind. Or in such fun but really not very scholarly pursuits as the analysis of dreams (or attempts to control ones own dreams)

    But its hard. Its not easy at all to build and remember a set of imaginary loci and situations rich enough to map the set of Chinese characters. Unless you had a very good reason to put all that work in beforehand, it would probably make more sense just to learn the characters!

  64. ziggler said,

    January 8, 2010 @ 4:54 am

    Studying Mandarin (with simplified characters) full time, I've used the Matthews book and continued using the strategy of the Matthews book to learn about 1400 characters in 3 1/2 months. Combining this book with flashcards and grammar books, I can now read/figure out many basic Chinese documents. I've also used the McNaughton book, which has etymology-based tips for remembering characters, but I don't use this nearly as much. The reason is that a very large number of Chinese characters that etymology does a very poor job of explaining. I too find that mnemonic devices require substantial effort, but I find rote memorization coupled with tidbits of etymological clues doesn't work as well for most characters. However, as others have said, you don't need to stick to strictly one method; I occasionally do memorize characters based on their etymology too.

  65. ziggler said,

    January 8, 2010 @ 5:37 am

    – Even when characters have a sound etymological basis, I find mneumonics are still helpful for learning characters' pronounciation.
    – I agree that the Matthews book does have a number of mneumonic stories that aren't necessarily 'great'. However, often these stories form the basis of a better mnemonic story I'll make up on my own, sticking with the general Matthews system. Often, I find the best mnemonic stories feature proper nouns or 'rated R' material that isn't appropriate for publication purposes.

  66. Fresh Sawdust said,

    January 10, 2010 @ 7:47 am

    Here's another one for the big Matthews'-fuelled bonfire roaring away nicely here on Language Log: 'Chinese Characters: Learn & Remember 2,178 Characters and Their Meanings', by Alan Hoenig PhD. I was having a nosebleed by the time I got to character 6 in the Amazon preview, fits by character 9, and convulsions by character 11.

    But seriously, although it may well be an interesting and even quite worthy book (to those who've actually bought it), I think that anyone needing such "detail" right from the very get-go it seems must have some sort of attention deficit and trouble therefore even noticing let alone remembering anything.

    I mean, when I was beginning Chinese I seem to recall that it wasn't honestly that hard to learn and appreciate (by simply looking, and GASP a bit of good ol' rote) the visual difference between say the 'scholar' and 'earth' radical-characters, and the materials that I depended on at the time didn't encourage me to think and make so much that I risked having a breakdown about it all (see 'nosebleed', 'fits', and 'convulsions' above).

    I discovered this book by the way from browsing for something or other!

  67. Fresh Sawdust said,

    January 10, 2010 @ 8:04 am

    Oops, a little correction: "and the materials that I depended on at the time didn't encourage me to *make so much of everything* that I risked having a breakdown about it all"

    Bonus thought (though probably expressed somewhere in the above posts): these sorts of mnemonics may seem like busy work when applied to the more basic/frequent/easier/simpler/straightforward/whatever characters, but in a reasonably succinct and well-designed approach, they would appear to ("must"?) start coming into their own as one reaches the "thousands rather than only hundreds of characters known"-level.

  68. Colin McLarty said,

    June 11, 2010 @ 12:11 pm

    There are no lies in the Matthews's book. There are silly mnemonics using things like "unicorn's horns." The mnemonics are deliberately silly for two reasons: to make them instantly memorable to beginners, and to make them easily forgettable when you move on. They are obviously not true etymologies.

    At first I wished they would use more accurate etymologies when they could. That is the strategy of the heroic book Chinese Characters: A Genealogy and Dictionary by Rick Harbaugh. But it has a problem that Harbaugh knows very well. Serious, scholarly etymologies of Chinese characters from the great classic Shouwen to today are regularly refuted by later archaeological finds. We do not know reliable etymologies for many characters. If you aim for true etymologies you will learn a good bit of false information albeit unwillingly.

    The only serious question to ask about these books is: do they work? That will vary from one person to another, and from one time to another. After months of writing drill from conventional workbooks I turned to the Mathews and their book did me a lot of good. Harbaugh's book has probably done me more good. Lately I am reading books in the Chinese Breeze series with huge pleasure and huge benefit. But I probably need to go back to writing drill from workbooks soon.

    If I thought some method would teach me 1,400 characters in 3 1/2 months I'd use it for 2 months and then go score an A on the Hanyu Shuiping Kaoshi (Basic). But I doubt any method will do that for me.

    These books are all cheap enough that you should get a lot of them.

  69. Cralls said,

    November 12, 2012 @ 3:46 pm

    Views have been pretty much expressed at this point so I don't want to add a repetitive postand I read through all comments, but I don't THINK this was really said, or very clearly marked at least:

    I studied 3rd year chinese in an intensive summer course. In this course we memorized so many characters a day it was ridiculous (we did so by traditional write each characters until we had them down). After a few weeks, if I wrote most characters 5-10 times, I had it memorized for the whole day. It was great. The problem was RETAINING this information. Specifically for writing. It took great amounts of time (which became even more boring as I knew the character after 5-10 times) to memorize for long, and then if I didn't continue work on those characters every day, I'd forget it again (usually if I glanced at it I would remember it, but I couldn't conjure it up).

    Mnoemonics are a slow and clumsy way to read and remember characters at first, but they keep it in your head until you get to a point where you don't need to remember it. Same with other methods of breaking characters down and memorizing all the parts of it etc etc such as brought in by T. K. Ann's Cracking the Chinese Puzzles. My fiancee learned Chinese since she was young, but doesn't have the largest vocabulary. She took the summer course with me, and she would look at a character, take a few seconds to break down the components (she couldn't generally tell the basic sound/meaning immediately) and then could memorize it for good. Most those characters she'd still be able to write off the top of her head a good 3-4+ months later.

    Of course different people are better at learning different ways, but I think these more technical ways may seem slowly and more clumsy and cumbersome, but I think they are uncomparable to most other methods when it comes to truly memorizing characters.

  70. Tarik said,

    January 25, 2014 @ 3:37 pm

    That book (Tuttle) worked extremely well for me. I completely disagree with this post

  71. Victor Mair said,

    January 25, 2014 @ 11:55 pm

    To all those who swear by zany mnemonic stories about the characters as the best way to learn them, I would like to ask how well you read and write Chinese 5 or 10 years after you utilize these methods to "master" 200, 1,200, or 2,000 characters this way? Do you really comprehend what you read? Can you read a wide variety of materials in different fields? Can you express yourself clearly in writing without having constant recourse to dictionaries or translation software?

    My experience with people who try to rely on arcane mnemonics is that they seldom get beyond several hundred characters using these methods. Can you imagine keeping straight in your mind 2,000 or more weird stories about the characters? That would be harder than just memorizing the characters directly.

    I agree with those who recommend understanding the characters on their own terms (knowing their radicals, phonophores, components, and so forth). Above all, I strongly urge Chinese language learners to read, read, read as much as you can with the aid of phonetic annotation. That's how I learned to read and write Chinese, and it was relatively painless (actually quite fun most of the time simply because the materials I read were interesting and edifying) and very efficient.

    See "How to learn to read Chinese".

RSS feed for comments on this post