Hey!

Currently I'm reviewing vocab cards the following way:

I'm identifying every kanji on the card by its keyword and then try recalling the word. If it doesn't come to mind in a few seconds, I press wrong.

I wonder if it would be more effective to simply look at the word and try recalling it immediately without spending time identifying the kanji, since that's what I would need to do anyway if I was to read a text fluently.

Does anyone have some experience with this?

Despite having done 1300 kanji in a RTK-like fashion, when I read, I almost never use the kanji to read words that I know. Sometimes I can guess the meaning of words I don't know, if there's only 1 kanji in them, but usually it's nearly useless for meaning. And of course, absolutely useless for pronunciation.

Thanks to the way the human brain processes the written word, the actual individual characters matter a lot less than the overall shape of the word.

In short, I wouldn't bother with each kanji, and I'd just go for the word. However, I'm not sure I'd toss it immediately. I'd probably give it a few seconds first.

@wccrawford - Just the opposite, re: word shape, as we don't read by whole word shape. ;p

Related: Forgot to post this a few weeks ago when rambling about mental orthographic representations and the visual word form area: http://www.unicog.org/publications/quiao...e_2010.pdf (Or was it this one? http://www.sciencedirect.com/science/art...1311000738 - Don't mind me, you know how disorganized I am.)

I see Dehaene has a more accessible bit on this in his book: http://books.google.com/books?id=NlYsTqt...22&f=false

wccrawford Wrote:Thanks to the way the human brain processes the written word, the actual individual characters matter a lot less than the overall shape of the word.

This is an outdated theory that is more or less considered incorrect these days.

Humans read english by recognising each letter individually, but multiple letters are recognised at a time in parallel based on the letter shape and geometric features in the letter. These are then used to determine what word you're looking at. At the same time, the candidate words feed back information to the letter recognition part to strengthen or eliminate possibilities because they would or wouldn't form a valid word. This gives some robustness to missing letters: you can still read words with some letters obscured.

This is a pretty approachable article:
http://www.microsoft.com/typography/ctfo...ition.aspx

Unfortunately, i've never seen the results of any studies on how japanese is read. It's conceivable that it's the same as english, with kanji in place of letters, and contextual information like the use of hiragana vs kanji playing the role spaces do in english to help you decide the next saccade.

As for guessing meanings, i can do that with a reasonable accuracy except where there isn't a clear tie to the keyword or other secondary meanings. It's possible that you can't because you didn't RTK enough kanji and have forgotten what you did study. If you learn just by reading words and don't pay attention to the detailed differences in the kanji, it's likely that your brain is just parsing similar looking kanji as being the same character (as if they were in a different font, perhaps). Maybe, for example, you're ignoring the radical and only reading the part with phonetic clues. Until you get a clash (words that are only differentiated by the character detail you're ignoring), this will work for reading... but make it hard to learn to write.

Not clearly distinguishing each character would also lead to an inability to see meanings in the characters. If you're more or less considering 郊, 効 and 交 to just be "こう", and not really focusing on the differences, then you won't be able to guess meanings in new words as those kanji have radically different meanings.

zigmonty Wrote:

wccrawford Wrote:Thanks to the way the human brain processes the written word, the actual individual characters matter a lot less than the overall shape of the word.

This is an outdated theory that is more or less considered incorrect these days.

Humans read english by recognising each letter individually, but multiple letters are recognised at a time in parallel based on the letter shape and geometric features in the letter. These are then used to determine what word you're looking at. At the same time, the candidate words feed back information to the letter recognition part to strengthen or eliminate possibilities because they would or wouldn't form a valid word. This gives some robustness to missing letters: you can still read words with some letters obscured.

This is a pretty approachable article:
http://www.microsoft.com/typography/ctfo...ition.aspx

Unfortunately, i've never seen the results of any studies on how japanese is read. It's conceivable that it's the same as english, with kanji in place of letters, and contextual information like the use of hiragana vs kanji playing the role spaces do in english to help you decide the next saccade.

Kana relate to letters, as phonograms; for morphograms there's another stage of buildup from the combinations of component parts into iconic wholes, processed differently in the brain from kana/letters.

In Japanese you've got a wider FOV, shorter fixations, and longer saccades, using kanji as the figure, in a manner of speaking, and kana as ground (and it helps that content words and function words are typically kanji and kana, respectively).