You are currently browsing the tag archive for the ‘method’ tag.

Earlier I spoke of a method for maintaining a more-or-less constant number of daily SRS reviews. That method did prevent an excessive number of reviews (my goal at the time) while maintaining a minimum daily progress. But as my business interests get more complex and diversified, I’m finding that I have less and less time time available for SRS reviews, and I’m also becoming more and more distracted as I do my reviews.

My solution is to move back to how jMemorize does its reviews: a fixed time per day on a stopwatch timer. Number of reviews does not matter. Size of expired deck and number of added cards does not matter. Just X-minutes of completely focused time split into 15- or 20-minute chunks. If you run out of time with unfinished cards, forget it and move on. If you run out of cards with the stopwatch still running, start adding new cards. Progress will be made just as consistently as before, but the smaller, fixed goals will eliminate distraction and allow for effective time management.

This is an experiment in progress (just a few days old), but already I am pleased with the results.


The introduction to Kanji ABC outlines two methods for using the book: (a) alternate between learning a dozen or so graphemes and then the characters which are made with those elements, or (b) learn all 500 or so graphemes first, then approach the kanji in whatever order you desire. I have tried both of these paths and would like to relate my experiences:

For the first 300 or so kanji I took path (a). I felt that this more closely matched the approach of Heisig, and at the time I didn’t see any reason to deviate from that path. But then I began to get impatient about starting Chinese characters and my 10,000 Japanese sentences, both of which would require learning some of the frequently used, but complex characters found at the end of Kanji ABC [a bad idea, but for reasons I’ll explain some other time]. So I put my Kanji input on hold and started adding just graphemes to my SRS.

This worked… for a while. But the more I advanced in my grapheme-only studies, the more trouble I had learning and retaining that knowledge. By the very end I had come up against a wall. I was forgetting the writing of graphemes just as quickly as I was adding them. In retrospect the reason was clear: unlike kanji, the graphemes have no mnemonic components, so frequent use is the only reliable way to remember them. And by neglecting my kanji studies I was never using the graphemes. In essence I learned the 500 graphemes following the traditional rote method for learning the kanji, in less than a month. That itself is pretty incredible, but hardly efficient (I could have learned twice as many kanji in the same period of time at the rate I was progressing). Taking this approach was extremely aggravating, and probably added 2 weeks to my kanji studies. I would not recommend this approach.

Conclusion: study graphemes and kanji simultaneously. It is more efficient, less frustrating, and will speed you towards your goal of complete mastery of the kanji.

Earlier I said that I would wait until I was further into my kanji studies before I gave details of my own tips and techniques. My reasoning may have been justifiable from a results-oriented point of view (I want to thoroughly test any advice I give), but I failed to factor in just how many tips and tricks I would develop. It’s far too many to keep track of in my head until I’ve mastered the kanji, so I’m going to start relating the most important tips right away, starting with this post, and just come back to make corrections if I need to in the future. So take these methods with a grain of salt; they are not yet fully tested.


When using any SRS, it’s essential that you review expired cards daily, and fully review the entire queue of expired cards. For the sake of progress it is important that you consistently add a certain number of facts to the SRS every day. You will find that the day-to-day workload will steadily increase with time, even if only the same number of cards are added each day. (There’s an exponential/logarithmic relationship here that I do not have the time or patience to derive.)

When I first started on the kanji I settled in on adding 50 cards/day to my SRS. I could do this in about 1.5 hours or so (a reasonable amount of time for me). However within a few weeks I was up to 100 reviews/day and reviewing was taking up so much time that on some days I was unable to add new facts. And when I went a day without reviewing…the backlog of cards took a full week to clear out.

To the point I am getting. Since the number of reviews were getting out of control, I decided to cut in half the number of facts I add each day. Now I am adding only 25 cards/day, and over the last week the per-day workload has been steadily decreasing. On the morning when I open Anki to find less than 50 cards up for review, I will resume adding 50 cards/day until the workload again reaches 100 reviews/day and I cycle back to 25 new cards/day. This is a sustainable system; a method that I can keep up indefinitely, and a method which will ensure a steady progress of 50-100 reviews and 25-50 new cards on any given day.

I think I’ve mentioned that prior to using Anki, my experience with SRS software was limited to jMemorize and Reviewing the Kanji. Both systems use a simple pass/fail scoring mechanism, where either you remembered the fact or you didn’t. Anki, on the other hand, comes from the line of SRS software the has a wide spectrum for scoring yourself on how well you remembered, or failed to remember a fact. This added input enables Anki to better and more efficiently schedule cards for review. Khatzumoto has a method for scoring himself that’s worked well for me (after adapting the 0-5 system of Mnemosyne to Anki’s 0-4).

Just how much of an effect that has isn’t apparent unless you get lazy (like I did a week ago) and start answering all cards as “1 – Made a mistake” or “3 – All right”. I guess I figured it was no big deal since having to stop and evaluate myself on every card was a real pain in the ass and was slowing me down. Boy was that a mistake. The few extra minutes it would have taken me to do a proper evaluation is nothing compared to the hours I spent in the last few days reviewing cards that should have been marked “5 – Easy” and relearning cards that I waited too long to review because they should have been marked “2 – Difficult” or “0 – Completely forgot”.

So do yourself a favor: find a scoring mechanism that works for you and your SRS, and stick to it.

For a site that is titled “All Mandarin, All The Time” it must seem odd that the only posts of substance (so far) are on Japanese. Well, I’ve been focusing solely on Japanese since I started this blog. In a way the title was a promise to myself to stop studying Japanese exclusively and start on Mandarin, but that hasn’t happened yet. Partly because I’m having trouble tracking down the resource I’ve chosen for learning hanzi (step 1 of the AMATT method!): Cracking the Chinese Puzzles by T.K. Ann. It’s the closest thing I’ve found to Heisig’s method and ordering for the hanzi, but unfortunately it’s been out of print for more than a decade and I can’t seem to find any decently priced copies. It is also in part because I am not comfortable embarking on my studies until I’ve acquired a decent pronunciation of Mandarin phonetics, but my native speaker is away visiting family at the moment. Those were the reasons, but now I’ve discovered(/verified) a third: you can’t start learn Mandarin and Japanese at the same time with the All Input, All The Time method.

Khatzumoto writes about mixing up languages when learning more than one at the same time on his blog. I read this, but at the time I thought it was total hogwash. I’ve tried to learn many languages before, and I’ve experienced the symptoms: mixing up words and sentence patterns, such as using a Japanese word in a German sentence, or saying something-は to mark the topic in Tagalog. But I found that this was a temporary condition; after trying hard enough I no longer made these mistakes very often, and was able to study a great many languages at once. At my height I was learning Japanese, German, Tagalog, Romani, Russian, and Thai together with daily practice and without any confusion.

But that was before the AIATT method. A few days ago I sinoified my environment and spent a whole day listening to Mandarin podcasts and language tapes, just to get a feel for the spoken language. My plan was to alternate languages: Chinese, Japanese, Chinese, Japanese, etc. on either a daily or weekly basis. I made some progress with my Chinese, but when I opened up my SRS at the end of the day, I found that 24 hours of non-Japanese language practice was all it took to undo two or three days of effort. It was not a minor setback–it was a complete military retreat. So bad that I immediately canceled the experiment and I’m worried about ever doing this again. I know that making a decision on one data point is usually a poor choice… but I cannot even relate the extent to which my skills atrophied over such a short period of time, and what kind of effect that would have had if I’d let it continue.

What changed between this and my pre-AIATT experience? I’m not sure. I think it has something to do with the fact that I was listening to audio I did not fully comprehend, and reliant on my subconscious brain to process and make sense of it. Whatever the cause, I now implicitly trust Khatzumoto’s claim that languages must be learnt in series for the AIATT method to work. I’ve therefore reevaluated my strategy: I will continue to learn Japanese in isolation until I can make the switch to monolingual dictionaries (with the exception of hanzi, which I will start learning and report on as soon as I can find that damn book). Only then will I begin to collect my 10,000 Mandarin sentences and listen to Mandarin audio, using Japanese as the base language for my Chinese studies. But I will keep an even mix of Japanese and Mandarin audio at all times.

Enough of that. I’ll end this post here–I prefer to talk about what I’ve done that works, not theorize about what might be useful in the future. Results are all that matters. I’ll post more as these results come in.

I said in my last post that Heisig can lead you astray with his ill-chosen English keywords. To give a concrete example, the one I’ve been having trouble remembering recently is 召. Heisig assigns this kanji the keyword SEDUCE, which given the primitives involved (DAGGER and MOUTH) is an evocative image. I never, ever, had trouble with this kaniji before. Then I found out it was wrong. The keyword given by Kanji ABC is SUMMON, which is a very good word to sum up it’s actual usage (召集する: to call together a meeting, or 召し上がる: send for food/drink, for example). [How do I know? I looked it up in a dictionary. But more on that some other time.] For the life of me I cannot come up with a better image Heisig’s. When I see 召 I think “SEDUCE”, and when I review SUMMON I think “…wtf?” because whatever image I came up to associate 召 with SUMMON lost its battle against Heisig’s champion SEDUCE.

As nice as it may be to use Heisig and not encounter these problems (since 9 times out of 10 the keywords, if they differ from Kanji ABC, are easier to remember), it breaks the cardinal rule AIATT: no mistakes. Every kanji for which you learn a misleading or incorrect keyword you will pay for, with interest, later in your studies. I guarantee it.

So what did I do? …I took a page from Heisig. RtK, unlike Kanji ABC, assigns primitive meanings to many characters used as primitives–images which for whatever reason are easier to remember than the character’s keyword, and are only used when the character is used as a primitive in another kanji. Strangely, your mind (if it works anything like my own) rarely has trouble differentiating between the true meaning of a kanji and its primitive/grapheme meaning, as long as you keep that distinction clear from the start.

My modification to the method is the following: if more than once you completely blank on a grapheme, change it’s meaning (this does not apply to kanji keywords). Be sure to check the index to make sure that you’re not using a meaning that will be introduced later. Also go back and change every card in your SRS that references it. This is easiest if you’re using an SRS like Anki that lets you tag cards (which i use for tagging primitive meanings of components). That said, only do this sparingly, when it’s actually required.