Thursday, October 23, 2014

Kanji Town + image-concept pairing = easy recall across many languages?

I was able to memorize about 3,000 Japanese characters and the character readings (ON yomi, the reading based on Chinese. Japanese characters can be read multiple ways, and the ON yomi is the only reading that is consistent and can be grouped). To memorize the reading, each character was "stored" in a location that signaled the reading of the character. For example, everything pronounced "kaku" is stored in Shifting Sand Land from Mario 64. I applied this to every possible reading, creating a fairly robust list of location-reading pairings. This was resoundingly successful at learning Japanese.

When I began learning Chinese I created the same system for that language. At least one hundred locations for at least one hundred possible readings. This time, I had a harder time "pinning" the characters to each location. I toyed with things, but while my overall recall was good, but less than ideal when compared to Japanese, which was almost instantaneous.

Things changed drastically when I tried a different method. I decided to give a few Chinese characters a specific person (Real or fictitious) that became the "mascot" for that character. For example, Tony the Tiger became the mascot for "Great," Anthony Weiner is the mascot for "Remorse" and Method Man is the mascot for "Method." Once the assigned mascots were "stuck" to the desired character (Using initially short-interval flashcards), to remember that the character for "Remorse" is pronounced "ao (4th tone)," I simply imagine Anthony Weiner (Remorse) hanging out in the location for "AO" (The opening level of Banjo Kazooie). To remember "Method" is pronounced "fa" (3rd tone), I simply imagine Method Man in the tech demo I saw for the game "Heavenly Sword."

Creating (a) the link between character "concept" and the desired "mascot image" and (b) the link between a location and the pronunciation took quite a bit of time. But the ease with which I can remember a character, it's reading and concept is astonishingly fast. I speculate that it is because the concept "method" is somewhat abstract, but "Method Man" is very distinct, and will not likely get confused with concepts like "option" or "way." It is also much easier to store a concrete image in a mental location when compared to an abstract one. After my spaced repetition program sent one of my character pronunciation flashcards 20 or 30 days into the future (And it was reviewed 3-5 times), I had very little problems recalling the word. I can almost "feel" when a word I am learning goes from a short term memory to a slightly longer term memory.

Now that I had adequately memorized this concept's pronunciation in Chinese, I was curious to see if I could remember "method" in Greek (A language I am not really learning... yet). In Greek the word is "τρόπος" and pronounced "tropos." After creating a location to correspond to "tro" (ElecTROplankton level from Smash Brothers), I tried to memorize "method (man)" in Greek, using the location to remind me of the beginning syllable; Method Man was hanging out in the Electroplankton level. After a few days and a few flashcard repetitions, I was able to remember it with as much ease as I did "method" in Chinese. Because I knew the Chinese word very well, there was no overlap when I tried to recall the Greek word. I have tried this with other words, and it has worked very well.

Here is what I am doing: Right now I am creating a list of words and images that correspond to each of them, just like "Method Man" corresponds to the concept "method." Over the last month I have been able to link about 1,400 images to 1,400 concepts, most of them based on the Chinese character keyword assigned to them from the book "Remembering the Hanzi." Many Chinese character readings I already know, and there is no reason to re-memorize them using this updated method, but with these characters I have begun memorizing the same concept in other languages (Russian, Turkish, Hebrew, Spanish, etc.). It takes about 10-20 days for the new reading to become automatic, and then I can move on to another language. Over the past month or so I have learned to say "cat" in 6 new languages, for example.

For long-term implementation, I think this method could be used to gobble up important vocabulary words for languages one desires to learn. My long term goal is to speak as many languages as possible (20+), and while this method might be cumbersome for just learning ONE new language, for learning vocabulary over multiple languages over time, this seems to be very effective.

(I was listening to a Let's Play that featured a Russian character that said "harasho!" which I memorized was "good" in Russian a few days before. That felt good.)

Has anyone tried anything similar to this? It is working SUPER well for me, I can learn words as quickly as I can create flashcards for them (At least that's how it feels).


  1. "After my spaced repetition program sent one of my character pronunciation flashcards 20 or 30 days into the future (And it was reviewed 3-5 times)"

    Did you mean that you reviewed 3-5 times in the first day? (e. g. with the short-term schedule?)

    1. No, this is all long term. After about 3-5 reps in SuperMemo, most of my flashcards are at 20-30 day intervals. I know that I have the word "locked in" when I get it correct after a few weeks at least.

    2. Interesting. When I learned some German with the Pimsleur course, I added items with the audio files for the words and phrases. Most of those items got 20-30 days for the first repetition and, surprisingly, I got they all (or nearly all, at least) right. I'm pretty sure that the Pimsleur algorithm helped to establish a more solid memory and, hence, a longer interval, but I'm not so sure if the cost-benefit really pays off.

      Besides, I had an idea similar to yours. I was willing to try to create a peg for every possible combination of four letters, with some phonetic overlap ("Gato" and "Gatu" would have the same peg, for example). I thought that with this kind of system, any new word in any language could be easily learned. Your experiment seems to indicates that indeed it would have worked well.

    3. Neat, so was it only for German? How far did you get? My memory peg locations are only made on a "need to use" basis, so construction is ongoing. As far as I can tell, this seems to be a consistent way of getting words locked into the short term memory. This is roughly similar to how memory contestants memorize randomly shuffled decks of cards, and as far as I know, there hasn't been a uniform mnemonic system that does the job better than "image with location" for specially remembering abstract information...

    4. I used it only for German and for just a couple lessons (after that, I stopped learning the language). I have ~113 items for German and about 60 are from Pimsleur. So it's a small sample, but it may indicate something.

      That's right, I got my idea from exactly that! I used to memorize numbers, binary, cards, etc. I learned simple major system (00-99) for numbers. After expanding to a millenium major-category system (000-999), it seemed natural to expand to words. Much of the work was already done: all I need was to transform the images I already had into the phonetic combinations and complement the rest. Nevertheless, even the classical major was useful in memorizing words. You just need to transform the word in the number and... voi là. You have the peg. Of course, the problem is obvious: with just 100 images, it's bound to produce interference.

      A made-on-demand system seems to be more sensible. My word system was aptly called "The Never-Ending System" more or less because of that: I expected that it would be quite some time to complete it :)