how i learn chinese characters (hanzi / 汉字)

This is a post from my other blog that i wrote last September. I thought i’d reprint it here because it’s more relevant to the topic at hand

I thought it’d be helpful for some people if i wrote out more details of my methods, so here we go. Firstly, there are many different ways to learn 汉字­, and there isn’t just 1 single way to do it. I actually like to use a combination of methods, and i hop between them just to give some variety. What i’d like to do here is demonstrate some things for people to try, but keep an eye out for other people’s methods too.

Just a bit of history first; I previously took some immersion courses at university, where we did about 20 hours of class time per week. Overall this was really helpful in many ways, but i think i currently learn more per hour of my time using my new methods. In those courses, we generally just wrote the character out on paper 20 times. This all happened at once over the span of an hour of homework time after class, and then we thought that we’d remember it (but typically didn’t). I’ll explain later why this doesn’t work that well.

The biggest single change has been my use of an SRS (spaced repetition system). My favourite is Anki, but other people like mnemosyne, supermemo, or others. Basically, this software is like flashcards, but the flashcards are strategically shown to you at the proper time in order to build your memory. It’s based on some psychological principles about memory that are well known amongst psychologists, but almost completely unknown to educators (in my experience, anyway).

Flashcards are blank until you fill them (obviously), so what should you put on them? This gets discussed a lot on language web sites, but here’s what i’m currently doing. In anki, i make two “cards” for every “fact” that i enter. A “fact” consists of a single chinese character, possible meanings for the character, pronunciations for the character, and any extra trivia that i want to throw in, like example words or proverbs. The “cards” that are generated from this are “production” and “recognition” cards. In a production card, i get the pronunciation and meaning (in english) and i have to think up the character (and practice writing it down). “Recognition” cards have just the character showing, and i have to come up with the correct pronunciation and the meaning. If i get either pronunciation (including proper tone) or the meaning wrong, then i fail the card.

For each card, i rank how hard it was (and whether i passed or failed), and Anki will reschedule my next viewing of it. Easy cards get a bigger gap between viewings, which saves me a lot of time. The way this works is that each thing in your memory has an exponentially dropping curve of how well you remember it, your “forgetting curve” as it has been refered to. For new things, the curve is steep, which means that after only a short time you’ve almost certainly forgotten it…after perhaps an hour or two you’ve forgotten he majority of the new things you just saw. For things that you are really familiar with, you could go for months or years without seeing it and you’d still have a high chance of remembering it…the curve is shallow for these. The more times something is repeated to you, the more the curve levels off (meaning you can remember it for longer periods between reviews).

Using the information you give it about difficulty, Anki is trying to schedule the cards so they show up when you have about a 90% chance of remembering it. Once you’ve been reminded, that memory jumps back up to 100% and starts ticking down again, but a bit slower than last time. (see the graph in the Wired article if this sounds confusing). This is why the gap increases between viewings….it takes longer and longer to get down to the 90% level because you’re getting more familiar with the card. At the end, you’ll continue to have all of your thousands of cards recognized with about a 90% success rate.

But the trick here is that there’s more to memory than just raw repetition. The other big component to memory besides repetition, is what i’d call “storage”. Storage is how well you anchor that thing into your brain. You anchor it by connecting it to other ideas, feelings, stories, colors, smells, situations you experienced…any concept in your brain already, and the more the better. Reinforcement is just what i described Anki doing…showing you something over and over. These two concepts combine together; If you have better “storage” then the forgetting curve starts off less steep…and then each repetition has a much bigger effect, and you create long term memories much more easily. If you store the idea more elaborately, then you get more bang for your buck with repetitions.

If i just show you some random chinese character for a few seconds, and then ask you a day later to draw it, you’ll probably look at me like i’m nuts. But if i show you a character and tell you the meanings of all the little pieces that make it up, and then we make up a crazy story that includes all of those meanings, then it’s quite likely that you’ll remember it the next day just by going over that story in your mind.

Even things that are entirely unrelated will help. Sometimes when i have trouble with a character, i’ll leave the room and go do something different that i can associate with that character. the other day i wanted to remember 抖, which is pronounced “dou3”, which is similar to the chinese word for beans (“dou4” / 豆). So i went out into the garden and watered my bean plants and flicked off some aphids. “抖” means to shiver or tremble, so while i was pouring cold water on the bean plants i watched the leaves shake, and imagined them being cold.

All of these experiences and thoughts are going to help me anchor the meaning and pronunciation of this character. Every time i see it in Anki now, i try to recall all the things i did that day, and the thoughts i had. I want to tie the things together in my mind. It’s sorta like tying a rope to something so it doesn’t fall out of your head. 😉

There are other things that help me learn characters, such as context. I usually remember a character quite well when i come across it in a sentence from a book. After i make a new Anki card for that character, i’ll add that sentence to the “extras” portion so that it shows up on the answer side of the card every time i see it. Characters that i’ve seen in a book are always easier to remember than random characters i just pulled out of some list somewhere.

It also helps me to learn other characters with similar pieces. An example from today is 蹄 (ti2, meaning “hoof”). I already know the character 帝 (di4, meaning “emperor”). Notice that the right-hand side of 蹄 is 帝. The left-hand side is 足 (slightly altered, but meaning “foot”). Knowing that this character has a meaning similar to “foot” and sounds like “di” from emperor makes it really easy for me to remember “ti2” for “hoof”.

Sometimes the real meanings of all those sub-pieces don’t do it for me, though, but i have no hesitation about making up completely wacky meanings for some of them. One example is 降, which can mean both “falling” and “surrender” (examples: 降落伞 [jiang4 luo4 san3] = parachute, and 投降 [tou2 xiang2] = surrender). If you look at the piece in the top right of 降, you’ll see it’s the same as the right-hand side of 致, 效, 攻, and many others. That piece isn’t a separate character on its own, so i found it hard to remember for a while. But then i decided that it looked vaguely like a standing person with a ponytail. Since ridiculous things are easier to remember, i decided that the ponytail person would henceforth be called “Steven Seagal“, the well-known bad-ass from countless martial arts movies.

Next, i decided that the piece on the bottom right of 降 looked like a telephone pole. So this character for “falling” is now associated with a mental picture of steven seagal parachuting down above a telephone pole, and every time i picture it i hope that he doesn’t land on the telephone pole with his crotch, because that would probably hurt quite a bit. The utter ridiculousness of this mnemonic means that i’ll never forget this character as long as i live. It doesn’t matter that the character has no historical association with either action movie stars or telephone poles. All that matters is you find a way to make it stick in your head. The ridiculous story will fall away by itself later.

Ok, so to tie up loose ends, why doesn’t it work that well if you just sit down and write a character 30 times? The reason is the forgetting curve. If you just write that character 30 times and then you don’t read or write it for a week, you probably only have a 20% chance of remembering it now (you know, just to pick a random percentage out of my ass). To really make it stick you have to be reminded a few times, with gaps between the reminders.

Instead of writing it 30 times in one day, i suggest the following. 3 times the first day, once the second day, once the fourth day, once the 8th day, once the 16th day. After this, you’ve only written it 7 times instead of 30, but you’ll remember it much much longer because you’ve become more familiar with it over time. It’d be logistically hard to try to do this by hand for every character, but that’s what you have SRS software like Anki for. It’ll calculate those intervals for you and show you all the right cards at the right times, based on the individual difficulty for that particular card and how long it’s been since the last time you saw it.

This is how an SRS will allow you to do less work but also learn more. It works with the way your brain works, so you can write that character 7 times instead of 30. It also spreads things out so that you can have thousands of cards in your virtual deck, but you only do a comparatively small number of reviews each day. It’s not feasible to do 4000 reviews of characters each day, but you totally don’t need to do that anyway. I do about 50 – 100 reviews per day, but that’s because i’m always adding new cards. If i just kept practicing the ones i’ve got in there and didn’t add any new ones, it would quickly go down to about 10 reviews per day for my 4000 cards. That’d be enough to keep me at a 90% recall rate for all of them.

What really inspired me to start doing this method was that i’d heard that there were people who learned 2000 characters in only 3 months. During my immersion classes at the university, i learned maybe 800 characters to high accuracy, and 1200 total if you count the ones i only recognized but couldn’t remember how to write. And that was after a total of 1 year of hard study in immersion classes. Now, after using Anki for a couple months and starting over again to review all the easy characters too, i’m up to 1800 characters with 90% accuracy, and i’m getting faster at learning new characters all the time. I think my current rate is about 200 new characters per week, but that tends to go up and down depending on how busy i am with work. I feel much more familiar with all the characters overall. I really wish someone had told me right at the beginning that this was possible….would have made things so much easier.

Well, that’s it for now. I have to go get ready for cycling tomorrow with a friend. Stay tuned for a future article on learning to read chinese…learning the characters is only part of the story, just like learning the alphabet doesn’t let you read english.