A quite long while ago now I posted what was meant to be an introduction to this post; and this post was meant to take place two or three weeks after that one; and, moreover, this post was meant to be a definitive guide to incremental reading for language learning.
Some of that didn’t happen, and the rest won’t. It’s become clear that developing this method will take considerable time, so instead of attempting to nail down something this fluid I’ll just dispense spoonfuls as I discover them. Herewith then, some of these, not really in any specific order.
Perspective first: I’m mostly talking here about the Alice in Wonderland deck. These cards are anywhere from 400 to over 3000 characters long, and have a parallel English text along with the reading on the back of the card. I’m also using a four-panel manga deck, of which more later.
One of the first things I noticed was that words that got skipped over the first or second or third time would eventually assert themselves and become clear. This happened again and again with every card. It felt quite remarkable really. It was a bit like those shows where they take the terribly blurry surveillance camera pictures and hit a magical “enhance” button repeatedly. Any photographer would kill for such a tool, but the language student has one available.
A problem I’ve encountered often whilst reviewing sentences is that the sentence gets memorized. The first couple of words come into view and the brain rattles off the rest by itself, without really reading the sentence at all. This does not happen with this deck – the cards are much too long. There is no way to gloss over the text in this manner; no choice but to actively engage with it.
The text used for the Alice deck actually has very few kanji; so few in fact, that adding rather a lot more would make it easier to read. Some parts are a little denser though. The more kanji with unknown readings in a passage there are, the shorter the card must be. Looking up words and readings is totally fine of course, that’s half the point, but once you have to look up as much as you know it just gets tiresome. The tadoku principle of sticking to texts that you understand at least 95% of is still very much in force here. You can get away with a little more, but it’s best not to push it. Some words – 公爵夫人 was a good example – gave me a lot of trouble because there were too many unfamiliar kanji in a row. This sort of deck is not well suited to learning such words. (Eventually though it did stick.) Sometimes a card will have a brief passage that is at a much higher level than the rest of the card. The dodo’s speech was a good example (the Eaglet even interrupted him with “日本語をしゃべれ!”). This can be frustrating because you keep having to come back to that card even though the large majority of it is perfectly clear. I still haven’t confirmed a good solution for this but I think I do have one, more later. Words that are reasonably clear from context stick like magic. After the card interval is over a month or so I hardly even remember that there were ten or twenty words in it that I didn’t know at all.
The parallel text has proved completely unnecessary. I never look at it. I suspect this would be the case whenever a text is within the proper 95% to 99% comprehension range.
Perhaps it goes without saying (I’m saying it anyway), but you had better really, really love the work you decide to review in this way. It’s pretty easy to get tired of it when you are rereading the same book piece by piece over and over. Choose carefully. It could be that a book of short stories, or a set of poems, or lyrics, would lend itself a little better to incremental reading.
Also about avoiding burnout – I found it best to limit myself to one review daily. This usually amounts to anywhere from ten minutes to half an hour. Because of this, cards often don’t get reviewed exactly on the day Anki decrees, so set the program to display cards from the shortest interval to the longest. That way the fragile young cards don’t get put off too long. Also, if there is a review due I don’t add a new card. As you can imagine this has stretched out the process of getting through the book by quite a lot.
Grammar has started to feel natural, to the point that I very seldom think about it. This should be the case with normal extensive reading, too, but I think there is a benefit to sticking to one author’s (translator’s, in this case) style.
Some of these cards I have made far too long. Over 3000 characters – that’s too much at once. To make them I took the line numbers in Notepad++ and took an approximately equal division of three for each chapter. This did not work well. Some passages with lots of conversation ended up being quite short, and others with more narrative were a great deal longer. Better would have been to count characters and make cards of perhaps 400-800 characters. Also, there’s no real reason to make the entire deck at once as long as you have the text file. Just make the next cards as you need them.
My intention at first was to extract sentences as I went along, for my sentence deck. My thinking was that once a card was over a certain interval, if there were still passages that were problematic, I’d isolate those. I haven’t done this even once, though. Instead, I’ve started a simple vocabulary deck. As I begin each card, every new word gets put into a text file, and once I’m done the words get added to the vocab deck. The idea is that by the next time I see the reading card, the new words will already have a review or two behind them. This is something I’ve only started very recently, but it seems to be working well. Usually when I see the word I instantly think of the context in which I found it. This way, I am pretty sure the issue I mentioned of thorny passages in the middle of easy cards will be resolved.
About that manga deck: it’s a sort of hybrid of ordinary sentence decks and incremental reading decks. Each card has a four-panel strip on it, but they’re short enough to treat them exactly like sentence cards. They’re just a lot more entertaining! and of course images make things easier. I highly recommend this, but it isn’t really incremental reading like I thought it would be.
That, it looks like, is the end of my notes for this time; I’ll post again once a few more have accumulated. As you can see, the method is far from polished yet, but the current rough version is working well enough that I definitely intend to carry on.
Lan'dorien's Mysterious Journey 
Very interesting insights. I think the one review per day is a good idea. I find it interesting that you seem to be quite easily learning the vocabulary just from the reading. I wonder, is the vocabulary deck really even necessary then? I found it necessary to use something else (in my case MCD’s) to push the material beyond familiarity into memorized but that was probably because I use much shorter IR cards. And not coming from a single author, or even source. But from everything you’ve said, the vocabulary is learning itself just fine. Are you just speeding up the process a bit?
I really do think this method lends itself well to follow-up with MCD’s, but perhaps in your case that’s essentially what your doing, just splitting up the massive context and the “cloze deletion” into two seperate decks. Anyways, a bit out of it here. if i think of anything else to say on the subject i’ll drop by and comment again.
Well, some yes some no. The words that are pretty clear from context, and that I have at least some notion of the kanji readings for, or perhaps are kana words, those get absorbed and stay there, but trickier ones need a little help it seems.
As time goes on more and more words will fall into the first category.
I do want to experiment with shorter cards – I think my next project will be a deck with lyrics. This will also have the benefit of boosting listening comprehension, as I listen to the songs. Although a book with an audiobook would do the same thing.
I still haven’t adopted your grading method, by the way. I keep thinking about it. Maybe this deck isn’t the best suited.