My new project to learn the kanji

So, I've been using SRS with Anki to study languages since late 2009, almost three years. Mostly with Japanese, but French and other languages as well. It's been interesting.

I got suckered into the sentence-mining AJATT championed then, but apparently that site has abandoned that method in favor of something called "MCDs" or whatever. Sentence mining is interesting, but it's fairly labor-intensive to make new cards, which tends to lead to wide variation in study times. I think it's helped some with my particle use, but it's too hard for the reward I get.

I suspect that the best way to use Anki is to focus exclusively on reading comprehension, whether it's comprehension of the kanji or vocab more generally. Furthermore, I think it's best to use Anki to build on existing body of knowledge, rather than begin by introducing a whole new body of knowledge. And to introduce things slowly, rather than blasting through dozens of new flashcards a day.

So August 1, I started a new project to simply recognize individual kanji with basic definitions and meanings (in Japanese, not Heisig's method). I'll work up a deck and simply introduce 3 new kanji every day, including the 300-400 I already know. If I do three a day for a year, I'd be able to recognize 1000 or so by the end of next summer.

Rather than using the order of the kyōiku kanji, in which Japanese children learn them, or by levels on the JLPT, I'm crafting a more idiosyncratic order: first, all kanji by stroke count, up to six strokes; then, the rest of the kanji used as radicals (which can have up to 13 strokes). After that, it's either more kanji by stroke count or instead kanji by frequency of use. The Japanese Wikipedia Kanji Frequency List, discussed further in this thread, is a fascinating resource: 173 kanji make up 50% all kanji in Wikipedia, 454 kanji cover 75% of all kanji, and 874 kanji cover 90%. Here's a list of the 2,500 most common kanji in Japanese newspapers, used by jisho.org.

I've wanted to put together a learner's semantic dictionary of kanji compounds, which gives examples of kanji compound words in each entry that contain only kanji prviously introduced. But that's a huge amount of work, a years-long project, and I'd never get started if I waited to do that. =/

Bilingualism, brain health, and mental training

In America, we live in one of the few countries where most of the population is exclusively monolingual. Bilingualism and multilingualism are more the human norm. Jared Diamond writes, in The Benefits of Multilingualism (Science, 330:6002 (15 October 2010), 332-333.):

Multilingualism—the ability to understand and speak several languages—is exceptional in the United States but common elsewhere, especially in small-scale traditional societies. For instance, once while I was camped with some New Guinea Highlanders conversing simultaneously in several local languages, I asked each man to name each language in which he could converse. It turned out that everyone present spoke at least 5 languages, and the champion was a man who spoke 15. What are the cognitive effects of such multilingualism? Recent studies (1–5) show that children raised bilingually develop a specific type of cognitive benefit during infancy, and that bilingualism offers some protection against symptoms of Alzheimer's dementia in old people.

...Our minds are assaulted by varied sights, sounds, and other external sensory inputs, plus thoughts and proprioreceptive sensations (which make us aware of the relative positions of our own body parts) (see the figure). To succeed in doing anything at all, we must temporarily inhibit 99% of those inputs and attend to just 1% of them, and the appropriate choice varies with the circumstances. That selective attention involves a set of processes, termed executive function, that reside in the prefrontal cortex and develop especially over the first 5 years of life.

Multilingual people have a special challenge involving executive function. Monolinguals hearing a word need only compare it with their single stock of arbitrary phoneme (sound) and meaning rules, and when uttering a word they draw it from that single stock. ...Multilinguals participating in a multilingual conversation...switch frequently and unpredictably between their stocks of phoneme/meaning rules. As a result, multilinguals have constant unconscious practice in using the executive function system.

Such pervasive multilingualism can cause enormous cross-pollination and "covergent evolution" in languages. Bill Foley says, in More on Papuan Languages:

Somewhere close to a quarter of the total of the world's languages are spoken in the New Guinea region-about 1100 languages. ...The small size of many Papuan speaking speech communities has often led to persistent multilingualism in the language of adjoining communities, the development of trade jargons for interlanguage communication or the shifting of language allegiance to languages of more powerful or economically advantaged neighbours. In such a complex, fragmented linguistic situation, Papuan languages not unexpectedly exhibit a pattern of enormous cross-influence in all areas. All types of linguistic features, basic vocabulary, pronouns, grammatical patterns, discourse styles can be and have been borrowed from one language into another. This makes the establishment of genetic links among Papuan languages doubly difficult: with no documentation for the vast majority of them older than 50 years, it is problematic indeed to sift what is true genetically inherited material from what is borrowed from other languages, especially borrowing from genetically related contiguous languages or borrowings centuries old from now deceased languages.

Diamond cites the research of Ellen Bialystok; LanguageLog and languagehat point us to The Bilingual Advantage, a Q&A with Bialystok at the NYT:

Q. One of your most startling recent findings is that bilingualism helps forestall the symptoms of Alzheimer’s disease. How did you come to learn this?

A. We did two kinds of studies. In the first, published in 2004, we found that normally aging bilinguals had better cognitive functioning than normally aging monolinguals. Bilingual older adults performed better than monolingual older adults on executive control tasks. That was very impressive because it didn’t have to be that way. It could have turned out that everybody just lost function equally as they got older.

That evidence made us look at people who didn’t have normal cognitive function. In our next studies , we looked at the medical records of 400 Alzheimer’s patients. On average, the bilinguals showed Alzheimer’s symptoms five or six years later than those who spoke only one language. This didn’t mean that the bilinguals didn’t have Alzheimer’s. It meant that as the disease took root in their brains, they were able to continue functioning at a higher level. They could cope with the disease for longer.

Which reminds me somewhat of the Gizmodo post How to Make Yourself Smarter in 20 Days, which highlights the use of n-back mental training exercises for 20 minutes a day for 20 days to increase fluid intelligence for three months. Not that I've tried it, but Brain Workshop vaguely reminds me of electronic flashcards and spaced repetition software like Anki, for memorization. Executive control, fluid intelligence, and vocabulary memorization are all very different, but it seems like 20 minutes a day of abstract memory practice would get pretty boring, and something language-related may be more interesting.

EDIT
Brain Calisthenics for Abstract Ideas (NYT)

SRS after three weeks

So, three weeks in, and I still think this SRS stuff is pretty cool. Even though I started reviewing even the most basic Japanese, I've polished up a recollection of Japanese postpositions of motion that I hadn't realized had gotten so weak. Starting with Japanese vocab and sentences that are relatively easy, as well as stuff that needed review, studying the flashcards has been fun.

It's fairly easy to make flashcards that review front and back, so that (for example) you're testing your understanding of the phrase from Japanese to English and from English to Japanese. Creating the cards to do this is a little different from the introductory instructional screencast video, possibly because Anki is frequently updated. I'm using Anki 0.9.9.8.5 with the Japanese plugin. To create forward and reverse cards in the Basic model, choose "Cards: Forward, Reverse" before you create in the cards. In the Japanese model, make sure it's set to "Recognition, Recall".

My strategy is essentially based on sentence mining, studying grammar and vocabulary in full sentences (even if it is something as basic as "This is a chair"/"Iste es un sedia"/"これはいすです"), and say the phrase on each card aloud in the target language before I check and score it. For Interlingua, I've created Forward and Reverse cards focusing on vocab and sentences from "Curso de interlingua pro comenciantes anglophone"; I've gotten through the first four lessons in three weeks at the default settings (20 new cards a day). For Japanese, I've created Recognition and Recall cards for model sentences and vocab from "しんにほんごのきそ I" and "漢字マスター, Vol. 1: 4級漢字100, 100 Kanji in 10 Days, The Easy Way". In three weeks, I've gotten through two lessons of each book. At this rate, I'll finish the Interlingua text in about four weeks, the kanji text in 12 weeks, and the Japanese grammar book over a period of several more months.

The workload is pretty managable. As long as I study every day, reviewing the flashcards doesn't take more than about ten to 15 minutes. Making cards does take some time, but spending an hour or two once a week usually creates enough cards for the rest of the week. I'm working on two decks, which means I'm spending twice as much time reviewing cards.

So, for example, on December 1, I spent 13 minutes reviewing 86 cards, which was pretty typical. Then I spent a few days away from the computer, so when I came back on December 5, I had to spend 38 minutes reviewing 243 cards (with some modest distractions). On December 6, I spent 9 minutes reviewing 60 cards; on December 7, I spent 14 minutes reviewing 90 cards; and today I spent 5 minutes reviewing 39 cards, because I had run out of new cards.

I can see the need, apart from SRS, to create a language-immersion environment for yourself. This method doesn't teach you to think in the target language, or teach fluency or creativity. However, I always found memorization through flashcards to be rather a chore; this is more like a game. And making memorization of vocab relatively fun is both remarkable and useful.

Spaced Repetition Systems

One of my friends is a big fan of the kanji-learning method of James W. Heisig, in which you first memorize the meanings of all the characters, and then later go back and attempt to learn their readings. I'm pretty skeptical of this system of learning to read Japanese, for many of the reasons discussed by Nihongo PeraPera. It may be that those of us who first learned a few hundred kanji are a little more dubious of this strange method, and ascribe success students may have with it to enthusiasm and diligence.

But it has become popular on the Internet. Many people use it with specialized flashcard software, known as spaced repetition systems or SRS ( 1, 2, 3, 4 ). This is a component of the method of study championed by AJATT ( 1, 2 ) and others.

After reading about it on Omniglot the other day, I've been re-evaluating SRS. I've always read about is as part of learning the kanji Heisig's way. But it seems like a feasible way to memorize things, regardless of what you're trying to memorize.

There are a couple of open-source programs for SRS: Anki and Mnemosyne are the most notable ( 1, 2, 3 ). SuperMemo is Windows-only and reportedly very buggy, and my friend has used both smart.fm and Anki. I think I'd prefer to use Mnemosyne, but the latest client only runs on the Intel Macs. On first use, making a deck of flashcards is easy, but using it to organize a system of self-study is somewhat more complex and non-intuitive.

Rather than the Heisig method, I suspect sentence mining, perhaps something like Kanji in Context after starting with my trusty old "しんにほんごのきそ". I may try it with Interlingua or such first, to figure out how best to organize a deck.

I also suspect that while relying on flash cards may be an effective means of improving reading comprehension, it'd be a weaker method of developing fluency, listening comprehension, and other aspects of language proficiency. Of course with Japanese, where this method is most commonly used, vocabulary, literacy, and reading comprehension are very serious barriers to proficiency.