Dear Mr. Heisig…

Dear Mr. Heisig…

Despite what I’m about to say, your books and your methods are by no means rubbish. They represent a pivotal step in students of Japanese attaining fluency. For Japanese, they rock. Nothing better on the market… fact.

HOWEVER, when it comes to Chinese, they’re a slice of bread short of a sandwich and I’d like to spend a few minutes explaining why you should go and stand in a corner, right now, and think about what you’ve done…

~ Eldon

Death of the Heisig Cards

A while ago I stopped doing Heisig-style Anki reps because the production cards were too hard and too draining.  It was fine while I was going through RTK, but I found character production based simply on the English word got harder and harder with time to the point of breaking my motivation to learn.

And so, reasoning that being able to write single Chinese characters based on English words was not worth the heavy investment of time and effort required[1], I transferred all Hanzi-learning activies to my L2IR (and later Cantopop) decks.

Above is a graph of my progress in learning Hanzi since the beginning of February[2].  It shows that my reading has improved consistently at a rate of almost 10 characters per day; each day I’ve been reviewing cards for ~25 minutes[3].  This way I’ve been learning word-usage and pronunication data about each character as well as writing, and I’ve been having a lot more fun in the process than doing English-Chinese production cards.

Does this mean that using Heisig’s books/methods is a waste of time though?  Not exactly…

Chinese ≠ Japanese

Something that Heisig ignores wholesale[4] is the fact that 90% of Chinese characters are composed of both a radical and a phonetic.  In Japanese, this is fine.  Characters are loaned for native Japanese words irrespective of pronunciation so the Heisig method is time-cheap.  The radical-phonetic link is not apparent until Chinese loan-words are encountered – which may not happen for a long time for some characters.  This makes it worthwhile to use sophisticated mnemononics to remember seemingly-randomly composed characters.

By contrast, in Chinese languages, the radical-phonetic link is very commonly encountered.  Take for example characters in which “我” (“ngo5”) is the phonetic.  哦,鵝and 蛾are all prononced “ngo4” – the meaning are “oh” (an exclamation, hence the mouth radical), goose (with a bird radical) and moth (complete with insect radical).  If you were following Heisig to the letter, you might spend 2-5 minutes coming up with images for each character; however, it takes virtually no effort (and hence no time) to remember each of these characters if you exploit the meaning-pronunciation information of each.

This is especially true if you’re using context- and repetition-heavy L2IR cards, since you’ll have a strong mental picture of what’s being discussed to attach new characters to.

The Recalcitrants

As you might expect, there exist characters that do not conform to the above rules.  Around 90% of characters were conceived as radical-phonetic pairs, which means that there are still 10% that need to be learnt some other way.  In addition, some of the radical-phonetics have changed pronunciation over the centuries to the point of the original link no longer being clear – 義 (ji6) and 蟻 (ngai5) being ready examples (although the Mandarin pronunciation is still ‘yi’ for both of them).

The Quick and Easy Road

And so the shortest path to character literacy is not to come up with Heisig-style mnemonics for characters willy-nilly.  It doesn’t matter how many characters we’re planning on learning – once you’ve learnt the radicals/primitives that you need and any irregular characters, it’s totally painless to absorb new ones.

So, for Chinese languages, if one could identify said building blocks/irregularities, one could streamline the Hanzi-learning process massively.  Rather than mnemonicking how to write 4500-odd characters (ignoring pronunciation and usage information completely and failing to remove English from our learning activities), we could perhaps reduce this number by up to 75%.

La crème de la crème

We can cherry-pick the very best bits of Heisig that make it so great for Japanese and streamline the method for Chinese.

Our list of characters to learn would come from a combination of Heisig’s primitive system (including 300 primitives) and characters that aren’t radical-primitive compounds. Single keywords would still be used because it makes life simple.

The list would not be bound by character frequency – rarer characters could be used if they were instructive and useful for the learning process – but it would nonetheless include a special ordering of the 1000 most common characters so that a solid foundation in character pronunciation could be quickly built up through context.  The only important thing would be to cover every radical/phonetic extant in modern written Chinese. 

A special extra section could be added at the end to focus on ‘broken’ radical-phonetic characters (whose pronunciation is not as a student would expect or where characters have multiple pronuncations) although this would clearly be different for different Chinese languages.  Oh, also, the Simplified/Traditional version would have to be separately imagined, for reasons I don’t want to get into right now[6].

That’s All, Folks!

So… yeah.  Thus concludes another “wouldn’t it be great if…” post.  I guess it kind of sums up how I’d go about studying if I could wind back time, but this is an experiment!  I expect things to go wrong!

A final quick little note on my stated “to fluency by September” mission – it ain’t gonna happen, if only because it’s going to be another six months before I’m at the stage where I can just read out (not necessarily understand) 3500 characters.  But it’s cool.  No excuses – I just haven’t been working hard or fast or efficiently enough.  Add oil, eh?

[1] In total, I spent about 40 hours reviewing cards in my English-to-Hanzi deck and at the time of ending had around 1300 active cards out of a total of 3500.  There was a period of about 70 days in which I added new cards, after which I just reviewed then for about 100 days before stopping.  These stats are impossible to analyse though because I already knew a truck-load of Japanese characters – what was newly learnt and what was just revised is not at all clear.

[2] The data comes from the Hanzi-stats plugin and represents the total number of Hanzi on the question-side of cards that had been answered before at any given time.  I was surprised to see the consistent increase as a function of time – if I carry on at the same rate I’ll be able to read 3500 characters in six months or so – this generally being considered enough to read a newspaper in Chinese.  Four months thereafter I’ll be at 4500 or so characters, which I imagine is around what the average HK high-school graduate can read.

[3] Subject to a large margin of error because of the way Anki calculates time reviewed – it’s capped at 1 minute per card because it assumes that if you spend more than sixty seconds on a card, you’ve forgotten about it and gone off to do something else.

[4] He ignores it by not including pronunciation information in his “Remember the Writing” books.[5]

[5] There is a (possibly erroneous) working assumption here that the “Remember the Reading” ones are a waste of time.

[6] It’s complicated and the Interwebs have trolls, y’know?


8 thoughts on “Dear Mr. Heisig…

  1. Pingback: Mission II « Thousand Mile Journey

  2. Excellent idea. Per your suggestion, how might this be accomplished. Are there any websites giving detailed clues, codes for the phonetic part of Chinese characters? I am learning Traditional Chinese characters and was contemplating using Heisig, so I would very much like to know.

  3. You can try this one:

    7000 characters are arranged by components, which effectively means by sound.

    You can use Cantofish or peraperakun to parse the pinyin for each charater.

    Downsides: it doesn’t sort characters by usefulness and is in Chinese…

    I’m currently still working on something similar to the 1000 character-thing outlined above; if it ever gets to a publishable state, I’ll let you know =]

  4. This is a great explanation of why Heisig’s method doesn’t work for Chinese. Besides, knowing the keyword for each Chinese character wouldn’t help you read Chinese like it does in Japanese, because everything, including the grammatical words, are in hanzi. I know because I tried to learn Chinese that way, and I couldn’t read anything, so I memorized the readings for about 3,000 characters, which was easy because I knew them from Japanese. However, this doesn’t allow you to read Chinese, either. There is no shortcut. I would recommend somebody just learn to read the Chinese characters for the most important words as they come across them and not bother trying to memorize “all of them.” Eventually you will reach a saturation point where you know all the important ones even if you can’t read 刽子手 or 嵩栾 (NB: Neither can many Chinese people.) Also, learning to write characters by hand doesn’t help you get better at Chinese. It’s irrelevant.

    • Yup, I generally agree that the readings are less useful for Chinese than I imagine they are for Japanese. There are too many characters borrowed for sounds. (I’m not fluent in Japanese so I’m not qualified to give an opinion.)

      Here’s what I hope is a better example of how characters should be introduced.

      I think a method should teach these characters initially, with readings, because they’re all common, useful, and demonstrate how meaning-sound compounds are created:

      口 鳥 虫
      哦 鵝 蛾

      It should deliberately not teach these characters, because they’re less common and follow the normal meaning-sound character composition rules:

      娥 峨 俄

      This would allow students to know the most useful characters, understand typical character composition and be able to pick up new characters easily, without wasting too much time on the method itself. (In other words, they could get to FUNBUN material faster that Heisiging ever character.)

      The method should also teach common exceptions. If, hypothetically there were a character 睋 which was both common and had an unusual reading (say, jat6 in Cantonese), then that should be taught in the method.

      Put simply, the method should both give the student lots of common characters, and also enable him/her to learn new characters easily following typical composition patterns.

  5. (I think a key point is that just learning the most common characters as the learner comes across them is not a good strategy, because character composition patterns are invisible to him/her for too long.)

  6. Finally, I don’t think handwriting characters is pointless (I hypothesise that since they take longer to write than type, you have to pay more attention to them, thereby creating stronger memories), but I’m no expert on the subject, so…

Leave a Reply

Fill in your details below or click an icon to log in: Logo

You are commenting using your account. Log Out / Change )

Twitter picture

You are commenting using your Twitter account. Log Out / Change )

Facebook photo

You are commenting using your Facebook account. Log Out / Change )

Google+ photo

You are commenting using your Google+ account. Log Out / Change )

Connecting to %s