Mnemonics for Pronouncing Chinese Characters (the Marilyn Method)


Update: This is the Marilyn Method for memorizing the pronunciation of Chinese characters. The method has generated enough interest that I decided it needs a name of its own.

Update update: The method has now been adapted to Polish and German, as well as adapting the target language to Cantonese.



Photo by Felix_Nine

For me a new mental gimmick is the best kind of toy there isbetter than a new electronic gadget. My latest toy is a technique for memorizing the pronunciation of Chinese charactersmy own invention, I am proud to say. This might end up being my greatest contribution to civilization, even better than my killer nacho recipe.

Lots of attention has been given to the problem of remembering how to write Chinese characters (see, for example, my review of Heisig and Richardson's book). By comparison, remembering how to pronounce the characters has been the neglected well-behaved sibling. But pronouncing the characters is by no means trivialespecially since the Chinese (Mandarin) language uses important sound contrasts that just don't exist in English, including of course tone of voice.

With my system I can link just two concepts (as is done in most memory systems) to represent the pronunciation of any Chinese character, including the tone. If you are using Heisig and Richardson or something similar to remember how to write characters, this is the perfect companion. You can also use the system to memorize spoken Mandarin words even if you don't care about writing them (not what I would recommend, but some people like it). It will take a bit of time and effort up front to memorize the basic correspondences, but this will be paid back with interest pretty quickly (within a matter of days in my own case). This post gives a complete description of the methodso bear with the length. You might also learn a little about Mandarin phonology.

The fact that this is possible at all is thanks to the Chinese "one character = one syllable" principle and also thanks to the rather constrained Chinese sound system, which makes it possible to list all Mandarin syllables in a rather modest table. My system uses a similar table, but pushed and pulled around a bit. It also has much in common with the "Bopomofo" system, which represents every possible Chinese syllable with one to three symbols taken from an alphabet of 37 symbols. I won't use the rather exotic-looking symbols themselves here, but their pinyin equivalents.

21 out of the 37 symbols can appear only at the beginning of a syllable:

b-, p-, m-, f-, d-, t-, n-, l-, g-, k-, h-, j-, q-, x-, zh-, ch-, sh-, r-, z-, c-, s-

Another 12 can appear only at the end of a syllable:

-a, -o, -e, , -ai, -ei, -ao, -ou, -an, -(e)n, -ang, -(e)ng

(...although certain syllables in both cases consist of a single symbol such as s ["si" in Pinyin] or ai, so that the beginning of the syllable is also the end.)

Three more symbols, representing -i-, -u-, or -ü-, can appear at the beginning, middle, or end. The -i- and -u- sound essentially like they would in Spanish, whereas -ü- sounds like a French "u" or a German "ü".

 Examples:

zh+ang = zhang in Pinyin
u + ang = wang in Pinyin
zh + u + ang = zhuang in Pinyin
zh + (e)n = zhen in Pinyin
i = yi in Pinyin
j + i + (e)n = jin in Pinyin

If you have been paying attention, you should have noticed I have only listed 36 of the 37 symbols. The remaining symbol, representing er, never combines with another and represents a syllable in itself.

All Mandarin syllables are made up of a limited number of such combinations. (Compare this with the rough-and-ready situation in English, where a single syllable might be "strength" or "boxed" or "splashed". Good luck finding a table of English syllables.)

The Basic Approach

The first step in the method is to use a mental picture of a given individual (real or fictional) to represent each of the initial symbols. We're going to call these representatives personages. I have made the following assignments, but you can use others if you like them better:

Babe (Ruth) for b-
Peter (Pan) for p-
Mark (Twain) for m-
Frazier (Crane) for f-
Darth (Vader) for d-
Tarzan for t-
Noriyuki (Morita) for n-
Leonardo (da Vinci) for l-
Groucho (Marx) for g-
Karl (Marx) for k-
Herman (Munster) for h-
George (Washington) for zh-
Charlie (Chaplin) for ch-
Sherlock (Holmes) for sh-
Ringo (Starr) for r-
Zach (Galafianakis) for z-
(Christopher) Columbus for c-
Sam (Spade) for s-

I tried to pick these to be as visually distinctive as possible. I put the last names in parentheses because I recommend you get on a first-name basis with your personages. You'll be spending a lot of time with them. Notice everyone on the list is male; also I left out j-, q-, and x-. The reasons for this will become apparent in due time. The names are a clue to the associated sound: Babe for b-, Groucho for g-, and so on. This "clue" will actually be important only in the early stages of useeventually (pretty soon, really) you will learn the correspondences by heart anyway.

We also need to introduce a new "null" initial symbol, which I will denote Ø-. So, for example (b- + -ao) gives you bao but (Ø- + -ao) gives you plain ao. This null symbol also gets an associated individual:

Ø- Albert (Einstein)

We might as well go ahead and introduce the "null" final symbol, denoted -Ø. So (s- + -ao) gives you sao but (s- + -Ø) gives you plain s (which is, however, written "si" in Pinyin).

(Now I have to interrupt myself to emphasize an important point: the "i" in Pinyin "si" is nothing like the "i" in "yi" or "ji". "Si" is pronounced more like "sz" than English "sea". Same for "zhi", "chi", "shi", "ri", "ci", "zi". If you don't understand this point then go back and ask your Chinese teacher about it, because it will wreak havoc with your Chinese pronunciation. The Pinyin symbol "si" is written with just the "s-" symbol in Bopomofo. To help keep this point in mind, I'm going to use parentheses to write it like this: "s(i)", "zh(i)", etc.)

Now the do-it-yourself part: for each of the twelve final symbols -a-o-e-ai-ei-ao-ou-an-(e)n-ang-(e)ng, and  -Ø, choose a familiar placea house, a restaurant, or a hotel. You will have to exert the modest effort to remember which places go with which symbols. I can't do this part for you because they need to be places familiar to you. (This is a point about the "memory palace" that many people seem to miss.) But for the sake of explanation we're going to pretend with a few examples:

-Ø your own house
-a the Plaza hotel in New York City
-e Maxim's restaurant in Paris
and so on....

(We lump -e and  together because these just happen never to appear in the same situation.) Now each syllable can be visualized as an image of a particular personage in a particular place. For example:

ba = b- + -a = Babe Ruth at the Plaza Hotel.
zha = zh- + -a = George Washington at the Plaza Hotel.
zh(i) = zh--Ø = George Washington at your house.
zhe = zh- + -e = George Washington at Maxim's.
ta = t- + -a = Tarzan at the Plaza Hotel.
a-Ø + a = Albert Einstein at the Plaza Hotel.

We can summarize all the combinations (so far) of initials and finals in a table:


-a -o -e -ai -ei -ao -ou -an -(e)n -ang -(e)ng
AlbertØ- er a o e ai ei ao ou an en ang eng
Babe b- ba bo bai bei bao ban ben bang beng
Peter p- pa po pai pei pao pou pan pen pang peng
Mark m- ma mo me mai mei mao mou man men mang meng
Frazier f- fa fo fei fou fan fen fang feng
Darth d- da de dai dei dao dou dan den dang deng
Tarzan t- ta te tai tei tao tou tan tang teng
Noriyuki n- na ne nai nei nao nou nan nen nang neng
Leonardo l- la lo le lai lei lao lou lan lang leng
Groucho g- ga ge gai gei gao gou gan gen gang geng
Karl k- ka ke kai kao kou kan ken kang keng
Herman h- ha he hai hei hao hou han hen hang heng
George zh- zhi zha zhe zhai zhei zhao zhou zhan zhen zhang zheng
Charlie ch- chi cha che chai chao chou chan chen chang cheng
Sherlock sh- shi sha she shai shei shao shou shan shen shang sheng
Ringo r- ri re rao rou ran ren rang reng
Zach z- zi za ze zai zei zao zou zan zen zang zeng
Columbus c- ci ca ce cai cao cou can cen cang ceng
Sam s- si sa se sai sao sou san sen sang seng


Notice that certain of the possible combinations, for example *be, just don't occur. This would mean, for example that Babe Ruth will never visit Maxim's. Notice also that we have handled the oddball syllable er by arbitrarily representing it as (Ø- + -Ø), or Albert Einstein at your house. This saves us from having to create an entire new column of the table just to handle this one case.  

This is the basic idea of the system. We still have some unanswered questions, though:

(1) What about those pesky tones?
(2) What about the intermediate symbols  -i--u-, and -ü-?
(3) What about the initials j-q-x-?

All to be answered in the fullness of time.

Handling Tones

First, question (1): we're going to expand the image to include the tone for the syllable by refining the location:

1st tone: In front,
2nd tone: Just inside the entrance,
3rd tone: Any other location inside,
4th tone: In the bathroom.

So, for example George Washington at the reception desk of Maxim's is zhe (2nd tone) whereas George Washington sitting at a table in Maxim's is zhe (3rd tone).

More Enhancements

Now, question (2), but first we focus just on the intermediate -i-. Standard Bopomofo practice would be to represent the syllable liao, for example, as l- + -i- + -ao. But we don't want to juggle three parts for the syllable, so our approach is to merge l- and -i- into a new initial li-. And since we need to distinguish li- from l-, we introduce a new personage Lauren (Bacall) to represent li- as opposed to Leonardo (l-). Similarly, we introduce:

Indira (Gandhi) for y- 
Brigitte (Bardot) for bi-
Paris (Hilton) for pi- 
Marilyn (Monroe) for mi-
Dorothy (Gale) for di-
(Mother) Teresa for ti-
Nancy (Pelosi) for ni-
Lauren (Bacall) for li-

Some points worth noticing: (i) This list is much shorter than the first, because f-, for example, never appears before -i-; (ii) These personages are all female, which will help you keep track of which initials have -i- and which don't.

And now we can introduce the remaining initials:

Gidget for ji-
Katharine (Hepburn) for qi-
Hattie (McDaniel) for xi-

There was no point in introducing these earlier because (just as f- never appears before -i-) j-, q-, and x- never appear except before -i- (or -ü-).

You may wonder why I chose Gidget to represent ji- instead of someone whose name begins with "J". This is to avoid confusion between zh- and j-, which are quite distinct in Chinese, although both come close to "j" than any other English sound. Confusing j- with g- is not really an issue because g- never appears before -i- or -ü-, exactly the opposite of j-.

(Representing j- by "G", q- by "K", etc. also has a solid historical basis, because Mandarin used to have a *gi syllable, but the g- took on a softer sound before -i [thus becoming ji], just as the English "g" in "giraffe" did, or the Italian "c" in "ciao". Likewise for *ki and *hi. Why do you think "Beijing" used to be spelled "Peking"? But once you learn the system by heart it won't matter anyway.)

So we can update our table with the new syllables (we're not finished yet, though):


-a-o-e-ai-ao-ou-an-(e)n-ang-(e)ng
Indiray-yiyayoyeyaiyaoyouyanyinyangying
Brigittebi-bi
bie
biaobianbin
bing
Parispi-pi
pie
piao
pianpin
ping
Marilynmi-mi
mie
miaomiumianmin
ming
Dorothydi-didiadie
diaodiudian

ding
Teresati-titie
tiao
tian
ting
Nancyni-ninie
niaoniunianninniangning
Laurenli-lilialie
liaoliulianlinliangling
Gidgetji-jijiajie
jiaojiujianjinjiangjing
Katharineqi-qiqiaqie
qiaoqiuqianqinqiangqing
Hattiexi-xixiaxie
xiaoxiuxianxinxiangxing


The columns of the table are represented with the same places you already identified (all though you won't need all of them for this table).

Final Pieces of the Puzzle

Next step is to handle -u-. Just as with -i-, we fold this into the various initial sounds to make a new series of initials bu-, pu-, mu-, etc. And just as before, we assign personages for each of the new initial symbols:

Wonder (Woman) for w-
Bugs (Bunny) for bu-
Popeye for pu-
Mickey (Mouse) for mu-
Fred (Flintstone) for fu-
Dick (Tracy) for du-
Totoro for tu-
Ninja (Turtles) for nu-
Lisa (Simpson) for lu-
Garfield for gu-
Kang (& Kotos) for ku-
Huey (Donald Duck's nephew) for hu-
the Joker for zhu-
Charlie (Brown) for chu-
Shrek for shu-
Richie (Rich) for ru-
Ziggy for zu-
Casper (the ghost) for cu-
Superman for su-


This time I used characters from either comics or cartoons, to help keep this group distinct from the others. And here is the corresponding part of the table:


-a-o-ai-ei-an-(e)n -ang-(e)ng 
Wonderw-wuwawowaiweiwanwenwangweng
Bugsbu-bu
Popeyepu-pu
Mickeymu-mu
Fredfu-fu
Dickdu-du
duo
duiduandun
dong
Totorotu-tu
tuo
tuituantun
tong
Ninjanu-nu
nuo

nuan

nong
Lisalu-lu
luo

luanlun
long
Garfieldgu-guguaguoguaiguiguangunguanggong
Kangku-kukuakuokuaikuikuankunkuangkong
Hueyhu-huhuahuohuaihuihuanhunhuanghong
Jokerzhu-zhuzhuazhuozhuaizhuizhuanzhunzhuangzhong
Charlie (Brown)chu-chuchuachuochuaichuichuanchunchuangchong
Shrekshu-shushuashuoshuaishuishuanshunshuang
Richieru-ruruaruoruiruanrun
rong
Ziggyzu-zu
zuo
zuizuanzun
zong
Caspercu-cu
cuo
cuicuancun
cong
Supermansu-su
suo
suisuansun
song


You will need to keep Charlie Brown distinct from Charlie Chaplin, but if you think this is going to be a problem for you, you can always change one name or the other. You will notice some spelling oddities: for example, zhu- + (e)ng becomes zhong instead of *zhung, but these merely reflect the way Pinyin works.

And then, finally, we create a series of initials with -ü- and their associated representatives:

Aphrodite for yu-
Neptune for nü-
Luna for lü-
Geryon for ju-
Cupid for qu-
Hermes for xu-

This time we use names from Greco-Roman mythology, to distinguish this series from the preceding three. You will notice that the Pinyin system drops the little dots over the -ü- in most cases. Once again, this is done only when there is no possibility of confusion. We keep the dots in nü-, because n- can precede both -u- and -ü-. We drop the dots in ju-, because j- can precede -ü- but not -u- (so you know the dots should be there even when not written). And Pinyin uses yu- to represent initial -ü- by itself.

Here is the final part of the syllable table:


-e-an-(e)n-(e)ng
Aphroditeyu-yuyueyuanyunyong
Neptunenü-nünüe
Lunalü-lülüe
Geryonju-jujuejuanjunjiong
Cupidqu-ququequanqunqiong
Hermesxu-xuxuexuanxunxiong


The final column of this table (yong, jiong, qiong, xiong) springs one last gimmick. The combination (ju- + -(e)ng) does not represent *jung (that doesn't exist) but rather jiong. Strictly speaking, jiong would be composed of (j- + -i- + -u- + -(e)ng). For our purposes, this gimmick saves us from having to come up with another four personages to represent, for example, (j- + -i- + -u-). I wish I could take credit (or blame) for this idea, but it comes from the Bopomofo system. My guess is the creators of Bopomofo did not want to deal with sequences as long as four symbols.

Putting it all together

Let's close with an example. I'm going to take the case of this character:



Heisig and Richardson assign this character the keyword "seedling". It is composed of two simpler characters:  ("flower") and  ("rice field"). To memorize the writing of the character, the normal approach would be to make up a little story combining the elements of "seedling", "flower", and "rice field". But we're going to go one better, by working the pronunciation of the character into the same story. This is pronounced miao with a 2nd tone. (Note  is pronounced hua and  is pronounced tian, so there's no help there.) We break miao up into mi- and -ao. The personage for mi- is Marilyn (Monroe) and my location for -ao is the Keio Plaza Hotel in Shinjuku (you would of course pick your own location). So we could imagine something like this:

The lobby (2nd tone) of the Keio Plaza hotel (-ao) has inexplicably been converted into a rice field (). Strange seedlings sprout and rapidly grow to a great height, developing large flowers (). Each flower blooms to reveal a figure of Marilyn Monroe (mi-). It's important to visualize this, like a movie, rather than just read the words. This image will stick with you, and it contains all the basic facts about .

76 comments:

Steven Correy said...

Hey - I found your post through Google. This is exactly what I was looking for, I'm also reading the books you talked about to memorize Chinese characters. Thanks for the post!

kujiranoai said...

Hello - I was just about to work on the same type of mnemonic system for memorising the Chinese characters myself when I found this.

I already speak Japanese and after studying Chinese for about a year now know around 2000 Chinese Kanji but sometimes stumble on the pronunciation which doesnt stick so well.

I am going to give it a go and will let you know how I get on but wanted to thank you for putting together such a great start.

Serge Gorodish said...

Thanks. I hope it helps. It has certainly helped me, and by now it's second nature.

I've thought about expanding the system to cover Japanese on-yomi (as well as Cantonese, for when I get around to it). The first step was to create a syllable table for on-yomi along the lines of the syllable tables for Mandarin you see all over the place. This might be the first-ever syllable table for Sino-Japanese--at least I couldn't find an example on the Web. I came up with a system of 24 initials and 30 finals (no tones, of course). It would remain to come up with 24 individuals and 30 locations and sub-locations to flesh out the system.

I have given some thought to whether it is a good idea to re-use the same individuals or locations from language to language, and I think it best to come up with a new set for each language to avoid confusion.

For the record, the 24 initials for Sino-Japanese are: (null), B, T, D, H, K, G, M, N, R, S, Z, W, Y, CH, J, SH, KY, GY, BY, HY, MY, NY, and RY. I was figuring to use female individuals to distinguish the embedded "Y" sound--for example maybe Kim Kardashian for "KY" but Kurt Vonnegut for plain "K".

The 30 finals are: A, ACHI, AI, AKU, AN, ATSU, E, ECHI, EI, EKI, EN, ETSU, I, ICHI, IKI, IKU, IN, ITSU, O, OCHI, OKU, ON, OTSU, OU, U, UI, UKU, UN, UTSU, UU. Again, one could use sub-location to indicate the (first) vowel (A, I, U, E, O) and location to represent the ending. So, for example, a certain restaurant could represent "TSU" and depending on location within the restaurant you get ATSU, ITSU, UTSU, OTSU, or ETSU. Or another place could be "I" and sublocation indicates "AI", "EI", or "UI".

Thus one needs really eight locations to represent -(null), -CHI, -I, -KI, -KU, -N, -TSU, and -U.

Miguel said...

Thank you a lot for this. I've just started with Remembering Simplified Hanzi (Spanish version) last week, having finished Remembering The Kanji back in 2005. Having a background in Japanese Kanji, I found that, for me, the hardest part of the Chinese Hanzi was not the recognition -which is almost automatic- but the pronunciation. This article has spared me from lots of frustration. Thank you again.

Katie J said...

Please do continue with the Sino-Japanese version. This one will be helpful when I work on Mandarin, though! Thank you! :D

Katie J said...

Please do continue with the Sino-Japanese version. This one will be helpful when I work on Mandarin, though! Thank you! :D

Anonymous said...

Brilliant! So very helpful!

Anonymous said...

It's a very interesting idea. I'm going to give it a try.

One note, you can double check, but I believe your statement "艹 is pronounced hua" is incorrect.

Serge Gorodish said...

I think you're right--not that 艹 seems to have much life as an independent character. I might have dug that pronunciation out of some dictionary, but now I can find both cao and ji.

Erik Blomqvist said...

Lovely, Steve! Found your post through Benny the Irish polyglot. I've been using mnemonic techniques for a looong time, remembering all sort of stuff, but never had the thought of bringing it to learning words in Mandarin. Thank you so much!

By the way, when you have longer words, with my than one syllable, do you make a story with them combined or what's your technique?

Best

Neil Keleher said...

Hey there, thanks for this.
Re 艹 , it's a modification of 艸 which is grass, cao.

Anonymous said...

I started implementing your technique today and was a little confused to read "a = -Ø + a = Albert Einstein at your house", and "er = Ø + Ø = Albert Einstein at your house". First one must be "Albert Einstein at the Plaza", following your example. BTW I choose Ø = space station, because I moved often recently and my home is a little less exciting than a space station anyway.

Serge Gorodish said...

Easy explanation--I made a mistake. Thanks for catching it; it's fixed now.

For words with more than one character, I use the Heisig "keyword" technique to remember the sequence of characters. This method then tells how to pronounce each character on its own. Chinese characters are fairly (not completely, alas) stable in terms of pronunciation from one word to another.

Antti said...

Hi,

Seems like a very thorough system. Thanks for sharing!

Have you come across 'Learning Chinese Characters' by Alison and Laurence Matthews? They have a similar system for remembering pronunciation and the character. It's just a bit simpler: tones are persons (e.g. third tone is a teddy bear) and syllables are remembered as a single word soundbite/keyword (e.g. ye -> Yeti).
You can see the basic system in Amazon's preview mode (the first pages):
http://www.amazon.com/Tuttle-Learning-Chinese-Characters-Vol/dp/080483816X

I'm curious if you considered representing each syllable with just one memory symbol (like the Matthews method) but then discarded the idea and decided to split the syllables and use two symbols?

Serge Gorodish said...

No, I haven't seen the Matthews and Matthews book--thanks for mentioning it. I never considered using just one "keyword" for each syllable and the reason is you will need roughly 400 different keywords to represent all the different syllables, plus 4 for the four tones. So you have about 400 things to memorize altogether to put the system in mind. And it still takes two ideas for any character (pronunciation + tone).

My system requires memorizing 51 personages for the initials and 12 places for the finals--much less overall. And it still takes just two ideas to represent any character (pronunciation + tone).

The insight that made this doable for me was to break a character's total pronunciation (initial+medial+final+tone) not into (initial+medial+final)+tone [which is the most intuitive] nor into initial+(medial+final)+tone [which is traditional], but into (initial + medial)+(final+tone) [which is the best from the standpoint of arithmetic].

My impression from a look at Matthews' book is that the system is intended to be more intuitive--using English words that sound kind of like the Chinese syllables, so you don't have to commit the whole system to memory up front. But what about something like "xuan", or distinguishing "che" from "chi"? I found that I need the most help remembering the syllables that are the hardest to distinguish based on English pronunciation.

Antti said...

The Matthews book is rather limited and sadly it doesn't have a complete list of keywords for all syllables. In fact it covers just about 800 characters. And some of those 800 don't have keywords at all, i.e. for some syllables you have to come up with a keyword yourself. For example 'xuan' (which you mentioned) has no keyword in the book. There are no similar sounding words in English for 'xuan'. One solution is to user other languages. Or you could find a person or character whose name has that syllable (http://en.wikipedia.org/wiki/Xuan) and use that.

On the other hand some syllables can have more than one keyword because there are many similar sounding words in English for the syllable. For example 'dai' can have 'dice', 'diamond', 'diary'. Having many keywords doesn't tax memory because the similarity of the sounds is a memory aid itself.

But as said some syllables don't have similar sounding English words so you have to make keywords that are not based on sounds. There are probably around 50-100 of these cases. And some syllables sound so similar to the English speaker that it's hard to find distinctive English keywords for them. Like 'che' and 'chi' which you mentioned. In the book 'che' has at least two keywords 'church' and 'cheroot' while 'chi' is given just one: 'chapati'. Then you just have to know that 'chapati' corresponds to 'chi' even though it sounds similar to 'che'. But this is not that hard, i.e. it does not hinder the recall process.

However in some cases the Matthews book handles the lack of good keywords in rather confusing ways. For example one keyword can be used for two different syllables. An example of this is the keyword 'powder' which can be used for both 'pao' and 'piao'. Then how can you know which syllable it refers to? Well the difference is marked by doubling the memory symbol which represents the tone. For example the pronunciation mnemonic for '跑' ('pao3') could be "...TEDDY BEAR is marking his runs with POWDER" and for 票 ('piao4') "...*TWO* DWARVES are snorting POWDER from the ticket". So if there is just one DWARF in the story the pronunciation would be 'pao4' but because there are two it is 'piao4'.
This is not a very good solution. Marking the pronunciation this way makes the recall process very slow. I have to make an effort to remember if it's pronounced 'pao' or 'piao' by consciously checking if there are two individuals in the story or not. Because this solution was so taxing I replaced 'piao' like syllables with my own keywords.

There are some other tweaks in the system too (like how to remember if a character has multiple pronunciations and tones) but this reply is getting too long already so I won't go into them.

In general the Matthews system feels a bit unfinished. Your system sounds much better - at least on paper ;)
Some questions:
1) How many characters have you learned with your system?
2) How fast is the recall process? My main concerns with your system are: dividing the syllable to many different memory symbols, the symbols do not always appear in a logical sequence e.g. the inital can appear at the end of the story, and the fact that the individual's names don't resemble the sounds that much. I've used systems like this before and I'm certain you can remember everything with it but how fast is it? Do you have to consciously construct the syllable by adding the symbols together or does it just come instantly?

Thanks!

Serge Gorodish said...

I'm currently at around 2400 characters and adding new ones at around three per day. It would not be fair to say I learned all of these "using" the system, as I knew a good many before I invented the system. But undeniably I have found the system a tremendous help.

I have also found now that the system works sort of backwards to help me recall the writing of characters. Chinese characters often have their own internal hints as to how they are pronounced. Unlike my system, these hints are often ambiguous. But I find that once I recall the pronunciation of a character, that combined with the native Chinese phonetic hints embedded in the character often brings the writing to mind.

As a rule I don't expect this nor any mnemonic system to bring facts to mind "instantly"--and certainly not fast enough for fluent speech. I think of it more as a dictionary you carry around in your head and can consult at will. Over time--with repeated practice--the mnemonic code fades and is replaced by natural memory, which is instant. I use Anki, about which I will have more to say in a future post.

The "stories" I use are actually more in the way of "images" combining several elements. These sometimes include motion but almost never duration, if that makes any sense. Thus placement of an idea at the "beginning" or "end" is not an issue. Again, in practice, recalling the pronunciation of a character generally seems to be a combination of natural memory and mnemonic code. Often I end up relying on the code for a specific, easy-to-confuse detail, such as which is the tone, or does it begin with "p" versus "b"? This crucial detail tends to pop into mind pretty instantly. It doesn't take me any longer now to recognize, for example, that Mark Twain stands for "M" than it does to recognize that the letter "M" stands for the sound.

This comparison points up an inherent trade-off between the initial effort required to use a system and its efficiency in the long run. People tend to gravitate towards immediate gratification. I recall a competition on some late-night talk show between 90-something-year-old Morse-code operators and kids text-messaging on their cell phones. The old codgers blew the kids away. But Morse code takes effort to learn, and a lot of practice to do well.

Anonymous said...

Really excellent system!

This is useful if anyone wants to choose some new names:
http://en.wiktionary.org/wiki/Appendix:Fictional_characters

Really looking forward to trying this out!

-Rob

Anonymous said...

Interesting system. How would you adapt to handle multiple-parts?

ex. chu1 zu1 che1 = taxi or

For chu1 zu1 che1= could be Charlie Brown and Ziggy and Charlie Chaplin in a taxi outside of Location 1

Or at different locations, i.e.,

huo3 che1 zhan4 = train station

Serge Gorodish said...

This is where the Heisig system comes into play, wherein each character is assigned a unique English "keyword" related to the meaning. For example, in simplified characters) 出 is "exit", 租 is "rent", and 车 is "car." So you need to remember "taxi" is "exit-rent-car" or "出租车". You can use my system to recall that 出 is pronounced "chu1", 租 is pronounced "zu1", and 车 is pronounced "che1".

Loyal-Job Jones said...

I have been looking for something like this for so long. Most of the time when I would search for Chinese mnemonics, I would just find them for characters or people would simply say "is none". Thanks so much!

Stefan said...

Hi Serge,

very interesting system, I'm currently trying to adapt it to Cantonese. You mention in the comments that you planned to do that too, did you get around to it? :)

cheers!

s.

Serge Gorodish said...

I haven't really got into Cantonese but I've done some preliminary work. Cantonese seems to have more vowels, more tones. One thing that made the Mandarin system manageable was splitting the syllable into four components:

initial consonant + initial vowel + final vowel or consonant + tone

That way there are not too many possibilities for each component. Then I use a person to represent the first two components (name + category) and a place to represent the last two (location + sublocation).

If we were to try the same approach with Cantonese, we might take a syllable such as taam6 and split it into:

t + aa + m + 6

Then t would come from the first letter of the person's name, aa from the category of the person, m from the location, and 6 from the sub-location.

At a first pass, it seems we need initial letters for:

-, b, p, m, f, d, t, n, l, g, k, ng, h, gw, kw, w, z, c, s, j

Use some gimmick to represent ng, for example, since I don't know anyone whose name starts with ng (like maybe use "sh" instead?)

Initial vowels are:

aa, a, e, o, oe, eo, u, yu

So we need eight categories of persons (maybe singers, athletes, politicians, ...?).

A few combinations (like b+oe) are not used.

Finals are:

i, u, m, n, ng, p, t, k

So we need eight locations.

And then 5 or 6 sublocations for the tones. (Some people distinguish a "high falling" tone and some don't.) Some say there are 8 or 9 tones, but the last few can be identified by the finals p, t, k, so they don't need a separate code.

And then finally some gimmick for the syllables m and ng. Maybe code them as "mi" and "ngi", which don't exist.

Stefan said...

Thanks a lot for the Cantonese details! I think you missed "i" as an initial vowel, but OTOH, "eo" and "oe" seem to be kind of the same (at least never appear in the same situation, like -e and -ê in your Mandarin method), so you still end up with eight categories of people.

The "high falling" tone is mentioned as "nearly extinct" on a 1970's Cantonese course (FSI), and seems to be gone as of today, but there are still six tones left: http://en.wikipedia.org/wiki/Jyutping#Tones

Also, there should probably be a "-"-final,too?

I tried to map it out, and I arrive with 146 people to memorize: 11 for -/b/p/m/f/etc., 20 for baa/paa/etc., 20 for -a-, 16 for -e-, and so on. If I got it correctly, Cantonese should have 635 syllables, so it would still be a time saver in comparison to the brute force approach; but when looking at the Cantonese syllable table, I get the impression that with all these holes, there should be a more efficient method, including significantly less than the roughly 150 items needed with the modified Mandarin approach.

Any ideas if there's significant space for improvement? I'm really bad with this kind of stuff (is it combinatorics?) unfortunately :P, but will try my best to do the work on this system that I'm capable of.

For example, using just initials (20) and complete finals (such as -aang, -eng, ..., 56 in all) would end up with roughly half the items to memorize, but then the familiar place approach won't work that well (I have heavy orientation problems myself, coming up with more than ten places i'm familiar with is a daunting task :P). One idea I'm having now is to use people for the full finals (but then it would be harder to associate the names in the beginning; how to connect a famous person and "-ei", "-aak", "-aat" etc.?). I could then use the location idea for the 20 initials, which seems more manageable, and attach the tone to it, like in your system.

If you find any errors in my thoughts, or have ideas on improvement, I'd be very glad to hear about it :)

cheers!

s.

Serge Gorodish said...

I have to disclaim here: I was kind of winging it, and my knowledge of Cantonese is pretty rudimentary anyway. So I am not surprised you caught some slip-ups in my previous comment.

That said, I think your suggestions are extremely useful.

Six tones rather than five: You are correct. Jyutping doesn't seem to provide a way to identify the "high-falling" tone, and from my limited listening experience, it seems to be on the way out. I have taken to labeling this "Tone 0", with the understanding that it is generally indistinguishable from Tone 1 in Jyutping.

Need a "-" final: You are right.

Need "i" as an initial vowel: You are right.

You clearly have a natural flair for this kind of thing, and did some useful legwork here. I think using places for the initials and persons for the complete finals has a lot of promise. It cuts down on the number of items and is fundamentally very straightforward--always an advantage.

Maybe you could reverse the final to get the person's name? E.g. "-aak" turns into "Kaa-" or "Kate", "-im" turns into "Mi-" or "Mike."

Keep in mind that you can use a different letter, as long as you are systematic about it. Like use "F-" to represent "-u", since Cantonese never has "-f". Use "H-" to represent "-ng", based on the fun fact that in English "ng" never comes at the start of a syllable but "h" never comes at the end.

Or use the first letter of the name to represent the final consonant and "category" to represent the vowel.

There is some room for savings by mashing together rows of the table. One of the worst offenders is "-ep", which apparently appears only in "gep". So one could code it as "gek" (which doesn't occur) and remember that "gek" is really "gep". Each time you do something like this, you cut out one personage from the system but give yourself an additional loophole to remember. It's not clear that the trade-off is beneficial.

The Cantonese symbol table has a lot of voids in it, but somehow these seem to be less systematic than in Mandarin--at least to my eyes thus far--and thus harder to exploit. But there's nothing wrong with keeping it straightforward.

For the places in my Mandarin system, I used houses of people in my family, as well as restaurants, etc. You could also use your school or office. Personally, I wouldn't have a problem coming up with 20 new locations. Go visit 20 new restaurants if you have to.

I would love to hear details if you come up with something. I'm more than happy to give suggestions, but I think you've got great ideas of your own.

Stefan said...

Thank you so much for your input! Those few suggestions made all the difference of letting me complete the system for Cantonese.

I use U/V/F/W for "U", as U was written V in Roman times anyway :) H for NG, as both are hard to pronounce for some speakers. L for not having a final, as it never occurs at the end. P can also be B, K can also be G, T can also be D. I can also be J.

So I come up with eight categories for the finals:
-aa- -a- -e- -i- -o- -ö- (aka oe/eo) -u- -ü- (aka yu)
In order not to have to come up with eight categories, I partition them into male/females:

-aa-: male actors
-- : Leonardo di caprio
-i : sam Jackson
-u : schWarzenegger
-m : steve Mcqueen
-n : jack Nicholson
-ng : Harpo marx
-p : al Pacino
-t : quentin Tarantino
-k : ben Kingsley

-a- : female actors
-i : angelina Jolie (as Tombraider)
-u : sigourney Weaver (in Alien)
-m : marilyn Monroe
-n : anna Nicole smith
-ng : selma Hayek (in Dust Til Dawn)
-p : natalie Portman (in Leon)
-t : liz Taylor
-k : woopie Goldberg

-e- : male musicians
-- : franz Liszt
-i : michael Jackson
-u : stevie Wonder
-m : wa Mozart
-ng : Heino
-p : elvis Presley
-k : Cat stevens (as Yussuf Islam)

-i- : female musicians
-- : Lisa Lopes (as Left Eye)
-u : nelly Furtado
-m : Madonna (with cone bra)
-n : Nena (with balloon)
-ng : whitney Houston
-p : Bjork
-t : Tatu (kissing)
-k : shaKira

-o- : male fictional characters
-- : Leatherface (with chainsaw)
-i : Yoda
-u : Frankenstein
-n : Neo
-ng : sherlock Holmes
-t : Tarzan
-k : King Kong

-ö- : female fictional characters
-- : eLastigirl
-ng : Helena (riding a wooden horse)
-k : Catwoman
-i : Jeannie (with crossed arms)
-n : Nike (wearing Nike shoes)
-t : Daisy Duck

-u- : male historical politicians
-- : Louis XIV (with wig and fat)
-i : Yeltsin (with vodka bottle)
-n : Napoleon
-ng : Hitler
-t : Trotski (with ice pick in neck)
-k : helmut Kohl (alternatively Qadafi)

-ü- : female historical politicians
-- : rosa Luxemburg
-n : Nofretete
-t : margaret Thatcher

leaves two: m and ng, that both (although the Cantonese syllable table above shows it) don't occur with initials. M is thus Mao (as -um doesn't exist), ng is Hillary (Clinton) (as -üng doesn't exist).
It was amazing how hard it was to find female counterparts to many of these categories (not that I found it easy to find the males)! I cheated a little in some parts (ignoring the initial sh in schWarzenegger and shaKira), but don't think that will be too confusing further down the road. Also, I used last names most of the time, but made some exceptions, and hope these won't confuse me. We'll see if Harpo Marx ends up in my mind where Sherlock Holmes ought to be :P

For the six tones, I use:
1 (highest tone): on the roof
2 (rising to high): on the stairs/elevator
3 (middle): anywhere inside
4 (low falling): in the cellar
5 (rising to middle): on the toilet
6 (low): in front of the location

One benefit of your Mandarin system lacking in my Cantonese one is that as you associated the initials with persons, you immediately have the initial letter from the name, whereas it's more abstract to arrive at the letter from the location (with is my initials). We'll see how that works out.

I found 20 locations, and will see in the next days how the system works for me. Feel free to adapt/change/criticize :) I'll probably write it up more cleanly on my blog in the next days/weeks.

Serge Gorodish said...

This is awesome. I have nothing to suggest at this point.

This definitely deserves a self-contained write-up. Please send me a link when you do so.

Stefan said...

Here's my post on the Cantonese variant.

Time to sleep :)

s.

Felix said...

Very interesting post. I think this will be useful for me, when i'm starting to learn chinese characters with Heisig.
But I have one question:

How are you handling the fifth neutral tone?

Serge Gorodish said...

Almost none of the characters have the fifth tone defined as their "native" tone. For example the character 了, commonly pronounced "le" (neutral) in its citation form is "liao3". I have found there are many tone shifts when a character is incorporated into a word and haven't decided on a systematic way to remember these.

However, if you want to incorporate a fifth tone into the system, all you need is a fifth sublocation for each location (the back door? on the roof?).

Anonymous said...

Thanks for the post.

I have designed a mnemonic system for memorizing the consonant and vowel simultaneously. So rather than t + a as two images. I just have one ta=tar(the black sticky stuff on the road). To which I add the tone using a background happening. For example the 4th tone must be to with water.

I just choose an approximate English sound as a hook, not as the exact way to pronounce. If there is no English sound, I use a word starting with same approximate sound.

It means I have to start off with a much bigger mnemonic infrastructure, but once learnt there are less components to the image.

P_ter said...

Your method is fine. But!(pardon me). It is how to backtrack. I see a Chinese word which expresses one idea. How do I get back to the pinyin?
For your information: I was impressed by an article on arxiv.org about learning Chinese from a group led by Jinshan Wu: see there article number 1303.1599
Peter

Serge Gorodish said...

I'm not sure I understand your question. If you successfully associate a personage and location with a given character, then the personage gives you the beginning of the Pinyin and the location gives you the end and the tone. For example, if you picture Tarzan (t) at the entrance to the Keio Plaza Hotel (ao, 1st tone), then you know the Pinyin is tao1.

Thanks for the pointer to the article. I haven't looked at it carefully yet, but it appears the authors have taken a systematic approach to something that Heisig and others do informally: which is to determine a study order for characters by balancing the character frequency with interrelationships. I myself have memorized a fair number of fairly obscure characters because, although rare, they appear as components of more common characters.

P_ter said...

With help of the memory palace a story leads to an object. That is fine for western language because objects are referred to in language. In "your" case it leads from pinyin to the meaning and with a little help to the character. Now I want to get back from the character to the pinyin. Also in a system. So from cùn to 寸 and then from 寸 back to cùn. In this case, in one way or another, it is my brain who has to backtrack this.
For your information the following:
See http://jinshanw.6579.idcice.net/indexE.php
In the middle you can put in a character and it will tell you its composition according to Jinshan's order.

Anonymous said...

Nice!

I use a different system for Japanese syllables.

I create a person-action pair for each of the Japanese syllables. So that means roughly 100 person-action pairs. Eg, あ= Einstein, か=Carmen Electra, etc.

(if you don't know what a person-action pair is, please see the "Dominic system" in mnemonic techniques websites).

Now I can remember any word by combining this to create the spelling.

To remember kanji, just form the word using person-actions, and apply that to the meaning of the Kanji.

This way, you have now memorised both the meaning and readings.

The only problem is that Japanese kanji usually have two readings, so it's a little annoying to create two stories for each character.

This has become second-nature for me by now and there is no difficulty!

Gerald Manley said...

Dear Sir
I have just started studying your mnemonic system and am quite impressed but I don't know if I have enough spare brain cells to make full use of it (I am 71). I hope you won't be offended if I point out a couple trivial typos. 1.
Fred fu- fu
Darth du- du duo
dui duan dun
dong
Totoro tu- tu tuo
tui tuan tun
tong

Should "Darth" be Dick (Tracy)?
2. Even more trivial---
Katherine (Hepburn) for qi- According to the google search, her name is spelled with 2 a's and one e ie Katharine.
Sincerely
Gerald Manley

Serge Gorodish said...

Thanks for the catch! I have made the corrections.

My father used to have a T-shirt that read "Age and treachery will overcome youth and skill!" A sentiment I try to live by, or perhaps more accurately "Age and efficiency will overcome youth and raw power." There is some decline in raw brain (and muscle) power with age--one of my projects is to see how far one can offset this with training, optimization, and discipline. As of my mid-fifties this is coming along quite well. The system here is an example of exactly this. I plan on tackling new languages well past 71.

Here's another great example.

Anonymous said...

Hi,

What a fantastic idea. Quick question though before I start making my own stories. I was wondering, if you are using these as an addition to normal Heisig stories and just review with a SRS, or if you have put all these images in a memory palace?

If you use a memory palace, won't it get really growded really quick? Lets say you desided to go through all 3000 characters in books 1&2 of which about 40% or 1200 characters are fourth tone (according to my google search). Even if equally divided between all 12 locations,that makes for very long toilet queues. How do you get past this problem? Or do you even have a problem? :)

Thanks

Alex

Serge Gorodish said...

This system is intended to be complementary to Heisig. Heisig's system is for associating the writing of the character with a rough idea of the meaning. Mine is for the pronunciation of the character.

I don't make stories for every character--just those that turn out to be hard to remember. The more characters you acquire the more hints you find as to the pronunciation. And as a rule I use separate stories for the writing and the pronunciation, although I key the pronunciation off the Heisig keyword. Some characters get one story, some the other, some both, and some neither.

The only "palace" I use for these consists of the 12 locations and 48 sublocations. Indeed each location gets used over and over again. This would be a problem if I ever wanted to sit down and list all the characters that end, for example, in -ao4 (which all refer to the same location)--but I never do that. But I can remember twenty different stories that take place in the same location (triggered off of twenty different keywords) with no problem.

Similarly the only reason I would create one giant palace to hold everything would be if I wanted to be able to list the characters in order. If I were going to do this (which I ain't) I probably would organize the palace according to Heisig number and associate the keyword with each number,

LB said...

Dear Serge,

Thanks so much for the excellent idea – am working on it as I try to gather the strength to tackle Mandarin. Not quite as old as one of your more recent correspondents, Gerald Manley (Hopkins?), but getting there.

Wonder if you’ve seen some recent postings, in a similar vein, on work in Sweden (at Uppsala University) by a Björn Liljeqvist. This seems to have spawned a 75-page report, “ Using mnemonics with distributed practice and practice testing to facilitate learning of Chinese characters” by Rickard Andersson and Jimmy Jonecrantz. The Swedish people go for a single table, rather like the audio-supported version at chinese.yabla.com/chinese-pinyin-chart.php. This ends up giving about 52% of the cells being vacant, which contrasts with your 4 sub-tables case, where the figure is less than half of this. At this stage I’m making the judgement that I prefer your case of a total of 55 characters and 12 locations, versus theirs of 24 people and 35 places – mainly because one needs to visualise 4 locations in each of these places. Should be easier with fewer places, given that the people are in clearly identified sub-groups. They do advocate sorting the locations into sub-groups based on the initial vowel, so 5 in their case and only 3 for you (plus ø), which seems useful.

I noticed some slight inconsistencies between your table and than of yabla.com, but these seem minor, although I know next to nothing about this language.

Now need to stop procrastinating and get started. Plan to use Anki after getting a better grip on the major radicals.

Thanks again for some great thoughts.

Lance Bode
(ex-mathematician and programmer, amateur linguist with emphasis on amateur)

Serge Gorodish said...

Thanks very much for the feedback, and for the reference. Indeed I had not heard of the PPM system, nor seen the article.

I would say both my system and the PPM system involve a significant time investment up front before being of use. Probably for either system, the initial time investment is ultimately well repaid.

A quick skim of the article shows that one question of interest is why some students use the mnemonic system and others don't. A possible factor, which I'm not certain is considered, is that individuals differ in their willingness to invest time in learning something for delayed reward.

Good luck with your Chinese study. It's a journey with no end (true for any language, but Chinese seems particularly rich). My reviews of Anki and Heisig's book are elsewhere on this blog. I highly recommend both, as well as the ChinesePod website.

LB said...

Cheers Serge.
If ever you are looking for richness in a language, try Arabic. A hard slog but rewarding.
LB

Serge Gorodish said...

Funny you should mention... I'm currently in a second-year Arabic class, and this is to be one of my areas of focus for 2015. I agree with you about the richness.

I have a theory that the difficulty of a language varies from individual to individual. The U.S. State Department rates Japanese as the most difficult language, but I find Japanese somewhat easier than German, for example. Sure, there are lots of characters to memorize but that's not difficult, it just takes time.

So me, personally, I find Arabic to be one of the hardest. Pronunciation was a difficult hurdle at first, but I think I'm largely over that now.

I recently settled on a mnemonic system for Arabic that I'm happy with. It's a little complicated, but as a mathematician, I'm sure you would get it. Maybe I should make this my next post?

LB said...

Would be interested to see what you have come up with regarding an Arabic mnemonic system. Have only recently discovered mnemonics via contributions such as yours (plus Fluent Forever & Moonwalking with Einstein), although I must have used them unconsciously for decades. However, I’ve decided to “park” Arabic for the moment and try my hand with Mandarin (pinyin + characters), hence the pressing need for mnemonics.

Had plenty of context and motivation for studying Arabic, as I recently lived in the Middle East for 4&1/2 years – but made the crucial error of not getting the phonology absolutely right from the start, when I was surrounded by native speakers, many of whom would have been willing to help, had I asked the right questions. It’s a big help to be confident with the sounds of ح, ع, and غ from the start, and to be able to distinguish س/ص, ت/ط, etc., in both production and comprehension. Am sure I would have benefitted from a mnemonic system that kept the roots and the structure in my head. Of course there’s always the issue of getting MSA right and then having to deal with one of the numerous dialects. But as a mathematician, you should love the way the language spreads in an almost programmatic manner from its triliteral root foundations.

Serge Gorodish said...

A strong suggestion that I would make for either Arabic or Mandarin (or any language with a phonology significantly different from English) would be to get a Pimsleur course (subject for a future post) and work through it. Speaking a language is a physical skill, like hitting a tennis ball, or roller skating, and the more practice you can get the better. One drawback of the language class is you spend a small percentage of the class time actually speaking. With Pimsleur you can focus on your pronunciation as you repeat over and over again, while painlessly acquiring some basic vocabulary and grammar.

I would never be able to speak Arabic comfortably without this kind of practice. To be able to produce all those different guttural sounds on the fly...

Pimsleur recently released an MSA course. Yes, there is the issue of MSA versus dialect, and it seems all the more difficult because the boundaries are apparently rather fuzzy.

After much careful listening, I believe that the distinction between س and ص, for example, has far more to do with the quality of the adjoining vowels than the consonants themselves. In English, likewise, we distinguish beat from bead largely because the latter has an extended vowel sound.

I struggled greatly with Mandarin pronunciation until I got an analytical description of the phonology (although this only confuses and frustrates many people). I think, for example, Dover's Modern Chinese: A Basic Course includes this.

I think making an effort to learn the tones and difficult distinctions from the start is important. Once you've learned a bad habit, it is much harder to dislodge.

Even though I can produce ح and ه, for example, pretty well, and distinguish them in listening, I'm still prone to confuse them in memory, because as a native English speaker I feel them to be variations on a single sound rather than fundamentally different. My mnemonic system addresses this.

I suggest you try also dabbling at least a little in Sanskrit. Sanskrit grammarians were remarkably analytical. And through the spread of Buddhism it influenced (and also via Pali) most Asian languages to some extent. Michael Coulson's book is an excellent introduction.

LB said...

Hi Serge,

Will be interested to read your future post on an Arabic mnemonic system. This got me thinking about how one might do this. Not wishing to necessarily pre-empt anything you might write, but it seems to me that, for the triliteral root system, one needs 3 orthogonal (discrete) axes. Your Mandarin system has two, the People and the Places. So, how to get a third. Well, in fact your system has a 3rd axis, that for tone, so in fact for Arabic one could use this, but this would require at least 28 spots in each location. Ugh! (But could kick myself for not trying to develop mnemonics at the time I was trying to get Arabic roots to stay in my head!)

Cue to reading about the Baker/baker paradox, and the relative ease with which we remember someone’s job vs their name. So, one could use profession for the 3rd letter in a Semitic root. More complex words originating from each root raise further issues, although one could just stick with a system for getting the (non-unique) meaning of a root into one’s head. Anyway, I look forward to seeing what you have constructed.

Brings me back to the issue of linking Pinyin (Person+Place+Tone) to the Chinese character. You seem to indicate you use existing knowledge of radicals to get started, then (I think) you link in with the keyword from Heisig (which I don’t own). What happens with multiple homophones? These are, of necessity, plentiful in Mandarin, given the small number of syllables used – see e.g. Figure 1 in “Chinese (Mandarin), Phonology of” (in Encyclopedia of Language and Linguistics, 2nd edition, Elsevier), by San Duanmu, University of Michigan, USA (Feb 2005), which I found online. From this I see one can find 106 occurrences of the syllable “ji”, so let’s say 25 per tone, with presumably quite a few of these in common usage. In your system each is simultaneously Gidget at one of the 4 locations in your house. How to distinguish these apparent clones? Would the professions idea help out here?

Serge Gorodish said...

As it happens my system (and I am now preparing a post on it, though I can’t promise it soon) is based on completely different principles, and in any case there is no territoriality here; borrow and improve on anything you like. But you have an interesting approach here, and I encourage you to pursue it.

A lot of competitive memorizers like to use a PAO system for playing cards. That means Person-Action-Object. So with one image of a person doing something to something they can encode three cards at once. This implies 52 images of each category, so what you are proposing (28 of each) is by no means unreasonable.

Actually it’s a PAOL system (Person-Action-Object-Location), with the location being the mental tag on which to hang the others. I think you might like to have 28 locations as well, which would give you flexibility as to encoding the vocabulary item.

For example, if the item is an object, you can use it as the Object of the image and use Person-Action-Location to code the root. If the item is an action, you can use Person-Object-Location, and so on.

As to your question about distinguishing homophones, this goes back to Heisig’s method. One essential aspect of the method is assigning a unique keyword in English to each character, which ideally is at least approximately related to the character’s Chinese meaning. As you note, the mapping from readings to characters is far from injective, but the mapping from characters is roughly injective. (There are many exceptions, but I only try to encode one reading for any character, and trust to experience to take care of the rest.) So I focus on deriving the reading from the keyword. And these days, if I know the reading, I find my background knowledge of radicals and phonetics often is about enough to recall the written character.

But if I’m in a situation where distinguishing one homophone from another is important, I rely on the keyword. For example, “jué” includes possibilities 訣 (trick of the trade), 決 (decide), 絕 (discontinue), 掘 (excavate), etc.

I think you might enjoy this also:
http://countryoftheblind.blogspot.com/2011/04/language-learning-tip-avoiding.html
The more you get used to it, the more opportunities you find for small-scale ad-hoc mnemonic codings.

Anonymous said...

Dear webmaster,
I'm mailing you to request a link exchange between our website and yours. We are now the world's premier online Chinese school.
I found that both of our website have similar theme, so I am interested in exchanging links with you.
My site link information is given below for your reference.
Site Title: Learn Chinese online
URL: http://www.hanbridgemandarin.com
Description: Learn Chinese online at Hanbridgemandarin.com. We offer Mandarin Online Course by live teachers from China and face-to-face mandarin courses in Shenzhen. Take a FREE trial lesson to learn Chinese now!

I will placed your site at here
http://www.hanbridgemandarin.com/exchange-links
Please forward me your site detail's with the following information:
Title :<< recipient-link-text>>
URL: <>
Description :<< link-description>>
If you agree to exchange links, I would like you to send me the exact URL where my link can be found.
I wish you the best with your site and believe this link exchange will be useful for both of our websites by getting more visitors and securing better search engine ranking. If you have any questions, do not hesitate to contact us.
Thanks,
Hanbridge Mandarin
marketing02@hanbridgemandarin.com

Luke Neale said...

Hi Serge,

I saw this post and applied its principles with the Heisig books one and two and now, almost a year later, I am still bale to recall the characters with relative ease. I came here to say a couple of things and ask a couple of questions.

First of all, thank you. This system has helped me greatly in the tough task of memorising hanzi. I have taught the technique to several friends and it has helped them also. I mention your post any time I hear of struggles in the realm of Chinese study (which is often!).

I'd also like to ask your permission to create a video series on my new website explaining this mnemonic technique step by step for anyone who might need it (not for personal profit, I assure you). I think that the information in your post doesn't seem to have the spread it deserves. I will of course cite you as the creator!

Great job and thanks for making a few peoples' lives easier!

Luke Neale said...

I found this post almost a year ago and the mnemonic devices you teach for learning Hanzi were incredibly effective. I learned all 3,000 Heisig characters in under 3 months and I can write and pronounce almost all of them to this day. Reviewing the characters regularly is still essential, of course, as it is for a native Chinese person, but the system is sound.

I came to your site to say thank you. Your technique helped me immensely and two of my friends who also read your post on my recommendation. I mention your post any time I come across somebody struggling with Chinese and they are always sceptical at first at its apparent complexity, but come around to realise in the end that it's way better than anything else they've tried.

I'd also like to ask your permission to create a video series detailing step by step how to go about this, and try to spread the word through a website I'm working on. I can assure you it wouldn't be for profit and I will of course cite you as the creator. What do you think?

Serge Gorodish said...

Wow. I'd be pleased for you to do this. Please let me know where I can find it.

BTW, while I would appreciate a pointer back to this site, I have no problem with you making a profit. I decided when I did the original post to put the idea out there for anyone to modify, re-interpret, or whatever (like Stefan's Cantonese version). Your video series will have value added beyond the original idea.

And I am most pleased to hear that the system has helped you and your friends.

Guillem Palau said...

Hi, your method is magnificent. I used heisig method to learn japanese kanji years ago. I can recall at least 900 of them even I gave up the active study of the language. I had learnt few pronunciations in Japanese.

Recently I had the idea to learn the Korean readings also called hanja, which are far less popular, using your method. At least the 1800 from the official list. Your workaround to turn initials, medials and finals to just two dimensional matrix is what I am trying to accomplish with hangul jamos.

There are 19 initials, 21 medials and 28 finals. Merging two dimensions into one, let's say initials and medials, will mean 19*21=399 characters, which kind of defeats the purpose to use this method for just 1800 cases. Alternatively, we can try to make 19 locations with 21 sub-locations. For me is way too many. Then, I think it is better to keep three dimensions.

I'll try the person-object-location method. I'm not using actions, they create too many interference with the mnemonics and systems I have already used and I need to avoid them. So basically, I use 28 fictional characters for finals, 21 locations for medials and 19 objects for initials. Therefore, each mnemonic have 4 items, those three plus the keyword I used to learn them as japanese kanji, the initial, the medial and the final.

Some exceptions are the called Gukja, that only exist in Korean language. Since they are meant to represent proper names, the keyword I'll use is the proper name itself.

The wikipedia explains clearly all this:
https://en.wikipedia.org/wiki/Hangul_consonant_and_vowel_tables

My aim is to memorize the persons, locations etc and start when the recall time is within a split second to recall. Any suggestions or guidance ?

Serge Gorodish said...

Thanks for the encouragement. Let me think about this and get back to you. I'll need to brush up on Korean....

Quick question, comment, whatever: Are you looking to handle all Korean vocabulary, or just to recall the pronunciation of hanja? Because if the latter, you might be looking at a much more limited (and manageable) set of syllables.

With Japanese, I just went through the index of a Japanese Kanji dictionary and wrote down all the GO-ON that were listed--then went about breaking them down into initials and finals.

Guillem Palau said...

Hi. Sorry for the late reply. I was looking to just recall the pronunciation of the hanja. I am aware there are quite a bit combinations that doesn't exist, but I have not enough knowladge to simplify the method beforehand. Just to clarify, often the pronunciation is different than the actual spelling on finals. So, strictly speaking my aim is to recall the spelling. I have tried this method with few hanja that are mature as kanji and works fine. The recall time takes longer than I would like nonetheless. I'll wait until the whole anki deck becomes mature.

George Mark said...

Really great stuff you have shared. It's really helpful for learners. Any one wants to learn chinese may follow Learn mandarin in shanghai

Tori said...

Serge, thanks a lot for describing in detail this approach. I'm a starter in learning Chinese and I have looked a long time for some system I do not have to learn the characters in parallel.
I follow your advice not to learn each and everything using your system, but only the words which are really hard to get persistent in my brain.
But one general problem popped up - for regular objects (house, spice, train, etc.) or actions (run, walk, play, etc.) it is no problem to provide abstraction stories. But what about words like: "前天" (qiántian) - the day before yesterday? It's quite hard to remember for me as I mix it constantly up with "昨天" (zuotian), thus I would prefer to have it in my memory palace.
Following your system "qián tian" would be:

Katharine (Hepburn) (qi) in the entrance (2nd tone) of The Angle Store (-an)
(Mother) Teresa (ti) in front of (1st tone) of The Angle Store (-an)

Okay, two persons, luckily at the same place.
But how should I now compose the meaning of "day before yesterday".
I'm really stuck with these situations representing abstracts.

I do appreciate every help to overcome this knot in my brain. How do you do it?

Serge Gorodish said...

Thanks for commenting!

So the system here is something I use just for single characters. In this case we have three single characters involved (with their Heisig keywords):

前 "in front of"
昨 "yesterday"
天 "heaven"

So you make up a story involving Katherine Hepburn, in front of, and the entrance to the Angle Store. Maybe she can't get out because you are standing in front of (前) her and she keeps bumping into you.

And then a separate story with Mother Teresa sitting and playing a harp in Heaven (天), which for some reason has been relocated to the front of the Angle Store.

And yet a third story with Ziggy (zu-) in the bathroom of whatever location singing Yesterday (昨). It's a sad song and tears are running down his face.

(Those abstract concepts are tricky to work with but come up with a gimmick if you must. Work in a word that sounds similar, or whatever.)

Now the multisyllable words are a totally separate layer, and I approach these with different tools. As with the single syllables, in most cases I can get by with nothing special but then there are those stubborn cases.

The first step of escalation I use is just to make a point of remembering the Heisig keyword sequence for the word:

"Yesterday" is yesterday-heaven
"Day before yesterday" is in front of-heaven

Sometimes this is enough to keep the word in mind, even though some of the keyword sequences are pretty random (such as "persona" 形象 shape-elephant). In this case, since yesterday (昨) is part of "yesterday" (昨天) the odds look pretty good.

But suppose I'm still having trouble. I make up a story to deal with the issue. Like "The day before yesterday comes in front of yesterday." Or even "Q comes before Z (in the alphabet) so of course qiantian comes before zuotian." The story can be totally logical, totally random, or anywhere in between.

I hope this helps.

Are you using Anki? I plan a future post on handling leeches (Anki's term for those stubborn words and phrases) so keep watching this space!

Tori said...

Hi Seth, thanks a lot for your verbose explanation. In fact that's the missing point for me: "the system here is something I use just for single characters". I haven't split these combined words into their atomics, thus I end up in quite long words with lots of details in explanation.
That means for particular multisyllable words I will have to separate the different syllables and extract their meaning to compose the meaning of the story. That's I assume the missing link to combine Persons, Locations and Actions to result in the desired meaning.
For sure, after trying to grab the essence of your text several times, the rather elegant "Q" before "Z" approach creates the stickness I need.

Thank you.

And yes, I'm using Anki, and I will be interested in contributions for the Anki-app.

Serge Gorodish said...

Yes, you see my meaning. There is a principle of economy here. No need, for example, to memorize that 人 is pronounced “rén” more than once. You can use the keyword for any character to remember the pronunciation of that syllable in any word. (Except when you can’t—but we’ll come back to that later).

Some seemingly random compounds can be explained by a traditional Chinese story. For example “contradiction” 矛盾 is formed from 矛 “spear” with 盾 “shield”. It refers to this story:

There was a man in Chu who sold shields and spears. Praising the shields he said: 'My shields are so strong that nothing can pierce them.' And praising the spears he said: 'My spears are so strong that there is nothing that they cannot pierce.' A person asked: 'What if someone pierces your shields with your spears?' He could not respond.

I found this by searching for “矛盾 etymology.” If you find no story, then you have to make up your own.

Chinese is rich in literary allusion and metaphor. Complicates learning, but we just have to embrace it. Four-character aphorisms are a particular favorite.

Now back to the awkward situation of a character with quite different pronunciations. For example, 行, which is pronounced “háng” in 银行 yínháng “bank” but more often “xíng” as in 行为 xíngwéi “conduct”. Now it seems to be a safe rule of thumb that the “hang” reading is used only in business-related words, but a more hard-core approach to the situation would be…

You could think of 行 “hang” and 行 “xíng” as “identical twin” characters—different meanings, different readings, but just happen to look identical. And give them distinct keywords. You could then used the Marilyn system if necessary to recall how each is pronounced and remember which pronunciation is used in a given compound by the appropriate choice of keyword.

Marek P said...

Thank you for your system.
I have modified names for Polish users (that is, for myself). Instead characters from either comics or cartoons (group -u-), I used characters from the novels of Henryk Sienkiewicz, a famous Nobel-prized Polish author. Now I have for these (I reduced Polish characters for your convenience):

Wolodyjowski (Michal) for w-
Boguslaw (daemonic prince Radziwill) for bu-
Petroniusz for pu-
Macko (z Bogdanca) for mu-
Fulko (de Lorche z Lotaryngii) for fu-
Danusia (Jurandowna) for du-
Tuhaj-bej for tu-
Nel for nu-
Longinus (Podbipieta) for lu-
Glaukus (medicine doctor) for gu-
Kmicic for ku-
Horpyna for hu-
Jagienka for zhu-
Chamis for chu-
Skrzetuski for shu-
Roch (Kowalski) for ru-
Zagloba for zu-
Carolus Gustavus (latin form required) for cu-
Sanderus for su-

I have mentioned that some persons have -u- in their names, but it is only accidentally (there is not a great number of characters and almost no possibility to choose; I am happy finding all the set).
Other personages are not with ‘u’ as a second letter, and of course not from Sienkiewicz. Also a group with intermediate -i- is all with names starting with ‘-i’, like Nimrod (whom I prefer for Nancy Pelosi). Indira, Brigiitte are acceptable (the last of course prettier than Bismarck), so as Marylin Monroe (very important person, because her name is the name of the method itself!), the rest are strictly ‘i-starting’: Pitagoras, Timur, Linneus and so forth, including Polish personages like Kiepura (very good for qi-). I also deleted ‘-i’ forms from other characters: Rambo is better than Ringo Starr. Also, according to cultural differentiation, I don’t know who exactly is Babe Ruth, but Beethoven is at hand.

Serge Gorodish said...

Wonderful, and very interesting! I put a call-out to your system at the top of the post above. Let me know if you decide to post it somewhere and I'll make a link.

Marek P said...

Then, I would simply send you a full description of method...

Marylin Method in Polish:

I have modified names for Polish users (that is, for myself). Instead characters from either comics or cartoons (group -u-), I used characters from the novels of Henryk Sienkiewicz, a famous Nobel-prized Polish author. The novels (all adapted for the screen) are: Sienkiewicz's Trilogy (TRI, containing: With Fire and Sword, The Deluge, Fire in the Steppe), The Teutonic Knights (TK), In Desert and Wilderness (DW), and of course Quo Vadis (QV). Abbreviations are my own :-)

I use Polish forms of names, and for convenience eliminate Polish letters. Further explanations in brackets [], and in notes below.

Basic modifications are:
- no pi-, bi- etc. forms except a proper group; - only pi-, bi- etc. forms in a proper group,
- Henryk Sienkiewicz’s personages, in and only in ‘-u-‘ group,
- mythology ,
- only personages who I can recognize :)


MARYLIN METHOD:

a) basic group:
Albert (Einstein) for Ø-
Beethoven (Ludwig van) for b-
Pasteur (Ludwik) [note 1] for p-
Magellan (Ferdynand) for m-
Fellini (Federico) for f-
Dante for d-
Tarzan for t-
Napoleon (Bonaparte) for n-
Leonardo (da Vinci) for l-
Galileusz for g-
Konfucjusz [Confucius, I need at last one Chinese person!] for k-
Hammurabi for h-
Zeromski (Stefan, great Polish writer, probably freemasonry master) for zh-
Charlie (Chaplin) for ch-
Sherlock (Holmes) for sh-
Rambo [note 2] for r-
Zapolska (Gabriela, she was popular Polish novelist and playwright) for z-
Cwietajewa (Marina) [Tsvetaeva, Russian and Soviet poet, see note 3] for c-
Sokrates for s-

b) initials with –i-:
Indira (Gandhi) for y-
Brigitte (Bardot) [Bismarck can be even better] for bi-
Pitagoras for pi-
Marilyn (Monroe) [note 4] for mi-
Dioklecjan (ancient Roman emperor) for di-
Timur (Tamerlan) for ti-
Nimrod (Bible: a great hunter) for ni-
Linneusz (Karol) for li-
Dzyngis-chan [Genghis Khan, note 5] for ji-
Kiepura (Jan, a famous Polish singer) [note 6] for qi-
Hipokrates [note 7] for xi-

c) initials with -u- (Sienkiewicz):
Wolodyjowski (Michal, TRI) for w-
Boguslaw (daemonic prince Radziwill, TRI) for bu-
Petroniusz QV for pu-
Macko (z Bogdanca, TK) for mu-
Fulko (de Lorche z Lotaryngii, TK) for fu-
Danusia (Jurandowna, TK) for du-
Tuhaj-bej TRI for tu-
Nel (Rawlison, DW) [note 8] for nu-
Longinus (Podbipieta, TRI) for lu-
Glaukus (medicine doctor, QV) for gu-
Kmicic (Andrzej, TRI) for ku-
Horpyna TRI for hu-
Jagienka TK for zhu-
Chamis DW for chu-
Skrzetuski (Jan, TRI) for shu-
Roch (Kowalski, TRI) for ru-
Zagloba (Onufry, TRI) for zu-
Carolus (Gustavus, latin form required, TRI) [note 9] for cu-
Sanderus TK for su-

d) initials with -ü- (Greco-Roman mythology):
Ariadna for yu-
Neptun for nü-
Luna for lü-
Gerion for ju-
Kirke [note 10] for qu-
Hermes for xu-

Marek P said...

Notes:
1. Peter (Pan) is not recommended, because in Polish translation it is Piotrus Pan, starting with Pi-.
2. Ringo (Starr) starts with Ri-, not recommended.
3. Marylin has ‘i’, although starts with ‘Ma’; despite this, it is fully acceptable, because the method’s name is from Marylin of course! Brigitte Bardot is more problematic (Bri-, not proper Bi-), but I decided to stay with her :-)
4. In Polish, the letter ‘c’ is pronounced ordinary like ‘ts’ or something. Only in foreign words we use pronunciation like K, or Ch as in Charlie (of course we have also ‘h’ in a ‘ch’ form). Despite this, there is ‘cz’ pronounced as in Charlie and a similar pronouciation has “ci” – for example, Czarniecki, present all around Trylogy. That is why there’s almost no family name (originally Polish or translated) beginning with ordinary ‘c’ prononciation! Cwietajewa (Marina Tsvetaeva) is a rare exception.
5. Kiepura is a specific name, connected with old invective ‘kiep’ (used frequently by Sienkiewicz’s Trylogy characters); the word ‘kiep’ is now archaic and neutral, but it’s variant, starting with ‘c’, makes a word that is very unpleasant also today. There is ethymological relations between ‘k’ and ‘c’ in this word, and this is why it’s good for pinyin ‘q’ (see also Kirke/Cyrce, note 10). Also a combination ‘ki’, good for this ‘q’.
6. There are two alternate Polish forms: Dzyngis-chan or Czyngis-chan, I prefer the first, because ‘Dz’ (I speak all the time of a Polish ‘dż’) is a tratitional transcription of ‘j’, and ‘Cz’ is like ‘Ch’ in Charlie.
7. If we treat Hippocrates as a personage of Greco-Roman mythology, we would have to seek another word (to use consequently h as a pinjin ‘x’ seems to be a good idea). But it is not me who recommends Hitler. What about Edmund Hillary, standing on Mount Everest or sitting at South Pole?
8. Albeit Nel from In Desert and Wilderness, we may use Neron (Roman emperor) from Quo Vadis. Both personages are quite characteristic and easy to imagine. But Neron is also historic person, and Nel merely fictional and avoids confusion. First of all, if we exclude Nel, Chamis remains the only one person from DW, not keeping proporthions.
9. Carolus Gustavus is a latin form pronounced like ‘K’, but written as ‘C’; Polish form Karol Gustaw has absolutely no relation to the letter ‘c’. Also, latin form is absolutely correct, because Henryk Sienkiewicz frequently uses it in original text of the novel. Carolus is better than his antagonist Czarniecki, because of possible confusions of ‘ch’ and even ‘j’ or ‘zh’ in pinyin. There are almost no family names with ‘C’ initial in Polish, see note 4. We have to keep even a slight connection with ‘c’ then!
10. We use a latin form Cyrce and a greek-based form Kirke as well; additional argument to connect it with pinjin ‘q’, in spite of combination ‘ki’.

Marek P said...

Please replace

Gerion for ju-

for

Junona for ju-

in Polish it is quite natural :)

Thank you

Marek P

Surendra Parihar said...

Thanks for giving suggestions and valuable tips for learning Chinese. Wants to learn to speak Chinese:Learn Chinese in Shanghai & Learn to Speak Chinese

Zorak said...

This is amazing!!!

Anonymous said...

This is a page with a Polish version of Marylin Method (in Polish):

http://zyskiwanieprzewagi.com/mnemotechnika-w-praktyce-cz-2-chinski-alfabet/
and if you want to see a table directly:
https://www.dropbox.com/s/i0h1acvzm6573jj/Tabela-chinski-alfabet.pdf?dl=0

Marek P.

Volodymyr Maksymchuk said...

Could you, please, explain the benefit of having less Places and more Characters?

Your approach: 55 Characters x 12 Places -> 67
Standard approach: 22 Characters x 37 Places ->59

I see that the advantage of the standard approach is that the place correlates to the pronunciation + no additional rules for the combinations (such as u+eng->ong).
However I clearly see that some places are used very rarely.

I would like to start Heisig in combination with pronunciation but due to lack of prior experience cannot make a decision which method is preferable. Maybe it's not so "painful" but I am afraid of "relearning" due to wrong decision as afterward it may become total mess.

Another question when selecting the names, what is more important - more accurate pronunciation or grouping (male/female/fictional/cartoon/historical?

Serge Gorodish said...

Short answer to your first question: It isn't necessarily better. One should feel free to modify and restructure at will.

Longer answer: For me it least, it is easier to come up with a large number of personalities and to sort them into categories, than to come up with an equivalent number of familiar locations (with sublocations). So more personalities and fewer locations works better, even if the total number of items is larger.

Note by the way that when Stefan adapted the method to Cantonese:

https://languageninja.wordpress.com/2013/02/06/cantonese-using-mnemonics/

he found it useful to reverse the roles of locations and personalities, I think for reasons related to this issue.

I understand your desire to get everything "right" at the start. I too am reluctant to tweak my methods once I get started--but nonetheless end up doing so sometimes. I think, however it is not so important to choose the absolute best method (if such a thing exists), as to find something that works pretty well for you.

As to your last question, accuracy of pronunciation is important, but has really nothing to do with how you choose your personalities or other mnemonic keys. Using "categories" and sounds as clues is useful at the beginning in committing the system to memory, but doesn't matter so much in the long run. In the same way that the shapes of the letters in most alphabets have almost no connection with their pronunciation (although there are exceptions), you could use any random collection of people to represent various sounds, although learning the system would be more difficult.

For example, "Hattie" really has little to do with the Chinese "X-" sound that it represents.

Kim Phuong Tran said...

The point is to lay a foundation in your brain (roughly 60 to 70 associations is fairly easy to learn). once it becomes your second nature, it hardly has to make sense why a certain personage is associated with a certain initials etc.

for example, why aren't the keys on keyboards laid out in alphabetical order? wouldn't that make more sense? however if you are a touchtyper, you can type without looking at the keyboard but if you'are asked to reconstruct the keyboard layout, i doubt you can do it. it's because the locations of the keys have been associated with your fingers movements, but the initial associations that help you memorize the those movements have faded away.

i think however you break it down doesn't matter as much as how comfortable you are with it.

i created a memrise course to learn the system myself with a few changes that i feel more comfortable with, and i learned it within two days, and it took me another two days to become reasonably "fluent". and since then, my character acquisition have accelerated. i've been learning Chinese for a month now and i've only learnt about 80 characters - an average of less than 3 words per day. now i'm learning about 15 to 20 characters a day and as i progress (meaning as i become more "fluent" in using the system) i feel that i can learn at a much faster speed.

check out the course at
http://www.memrise.com/course/1333874/mnemonic-system-to-learn-chinese-pronunciation/

if you prefer to learn the 22 character x 37 place system, someone had created that course too. check it out at http://www.memrise.com/course/952870/memory-system-for-learning-chinese-hanzi/

Serge Gorodish said...

All valid points. And thanks for your work!

Meng Chen said...

Good article!Thanks for your hard working! Chinese is a most difficult language, Recommended friends who are learning Mandarin Chinese a good learning
article after reading this article:
How to Learn Chinese Characters?
Hope all of us can make rapid progress!

Serge Gorodish said...

感谢你的支持!

Serge Gorodish said...

可能你学中文的朋友们也会喜欢Heisig的书:

http://countryoftheblind.blogspot.com/2011/10/product-review-remembering-traditional.html

谢谢!

Kohl Kimoto said...

What about remembering a word when two characters meanings don't really make sense together? How would you do that?

Serge Gorodish said...

Ah, now that is an interesting problem, quite different. A surprising number of Chinese words appear to be random combinations of characters. Take for example 矛盾 "contradiction", which is formed from 矛 "spear" + 盾 "shield."

I've been told that the origin of this word is a story about a merchant who was selling spears and shields. He claimed that his spears were so sharp they could penetrate anything. And the shields were impervious to any blow. A skeptical customer asked: "What happens if I hit one of your shields with one of your spears."

It's a cute story. It might be true, at least in the sense of explaining the origin of the word. But for the purpose of remembering the word it doesn't matter whether it's true or not.

So my approach to these combinations is to make up a story that ties the word together with the sequence of keywords for the individual characters.

Take for example "碟中谍", which is "Mission Impossible." Literally it means "saucer-middle-espionage." "Espionage", well, fair enough, but why "saucer"? I don't know why but I decided to pretend the reason is because in the latest movie Tom Cruise plunged down into the middle of that saucer-shaped cooling tank.

Hope this helps.