yogert909 Wrote:In case anybody is interested, I just ran across this on the supermemo website. (Anki is heavily based on the SM2 algorithm.)
http://supermemo.com/articles/theory.htm
Quote:In a long-term process, for the forgetting index equal to 10%, the average rate of learning for generic material can be approximated to 200-300 items/year/min, i.e. one minute of learning per day results in acquisition of 200-300 items per year. Users of SuperMemo usually report the average rate of learning from 50-2000 items/year/min
Yes, but SuperMemo defines items in a way that is very different from how most of us here do. They are a very small, singular piece of information. For instance, they sell an Advanced English deck where the pronunciation, spelling and meaning of a word are three separate items.
On the other hand, when a student of Japanese is faced with a word like 携帯 for the first time, that's a lot of information, and it can only be learned if the theoretical requirement that each card should be an item is thrown out the window. For one, even if you finished RtK already, you're still not likely to recognize the two Kanji easily. So just struggling to recognize them amounts to working through two items. For the reading of the word, one has two options:
1. remember the readings of the two Kanji (two more items)
2. remember what meaning these two Kanji have when put together (an item) and remember how to say that in Japanese (another item); most of us follow this second path;
So, just reading this one word, you've processed four items, not one. But it doesn't end there, since Supermemo also recommends the use of context. While in their Advanced English deck that context would be information the student is plenty familiar with (and English sentence made up of simple words the student's seen a million times), for us that context is a Japanese sentence, which often contains at least one or two more "items" that we would struggle to process.
All that makes reviewing Japanese vocab at least five times slower than the estimate you quoted. So it's 40-50 cards/year/min. In other words, if you add 50 cards that contain a compound word you don't know, who's Kanji you have to recall a mnemonic for, to recognize, who's readings you don't know automatically, and an example sentence that has another Kanji or word you have to exert some effort for, count on having to spend 350 minutes reviewing those 50 cards, in the future.
And it's not like we can choose to make things easier. Making a separate reading and recognition card seems pointless (since in both cases, you have to recognize the word, to read it...you don't know how the individual Kanji are read, just from looking at them...they are read differently in different contexts, not to mention most of us don't pay attention to individual Kanji readings anyway).
So this is the material we HAVE, to review. This is what Japanese is made of. We can't study single items, as defined by Supermemo, because the items Japanese is made up of are mostly compound words with two or more Kanji in them. The only thing we can do is choose good mnemonics in RtK (to make sure we remember 99% of the Kanji we encounter), and do lots of immersion, to help remember the words. And, believe me, it takes a lot of both, to make sure one's failure rate in reviewing Core sentences is only 10%, as assumed by the math you posted.
Edited: 2014-07-11, 8:15 am
