nadiatims Wrote:

corry Wrote:Maybe we should just increase the Interval Modifier in Anki? I am going to try 400%.

if interval modifier does what it sounds like it does, I'd recommend setting it to something like 2000%.

Sorry but this wouldn't work – unless you work with a pre-made deck (what you don't recommend at all, as I recall), you still would be wasting time on creating new cards.

nadiatims Wrote:

corry Wrote:Maybe we should just increase the Interval Modifier in Anki? I am going to try 400%.

if interval modifier does what it sounds like it does, I'd recommend setting it to something like 2000%.

What benefit does that have?

Quote:Sorry but this wouldn't work – unless you work with a pre-made deck (what you don't recommend at all, as I recall), you still would be wasting time on creating new cards.

So it would work with a premade deck?

Inny Jan Wrote:

nadiatims Wrote:

corry Wrote:Maybe we should just increase the Interval Modifier in Anki? I am going to try 400%.

if interval modifier does what it sounds like it does, I'd recommend setting it to something like 2000%.

Sorry but this wouldn't work – unless you work with a pre-made deck (what you don't recommend at all, as I recall), you still would be wasting time on creating new cards.

My reasons for suggesting this have little to do with card creation. Btw for people who do use anki, I'd suggest using rikaichan to export words and definitions to a text file at at the click of a button. This is what i used to do when I was still using anki. Cards don't take that long to create.

My reason for suggesting boosting the interval modifier to 2000% are to eliminate the short interval reviews. The idea being that difficult cards need time to sink in. It doesn't necessarily help to repeatedly bludgeon yourself over the head with same repeatedly failed cards. Instead ignore them for a while and let them become words that your brain gets used to over longer stretches of time (weeks/months, not hours/days).

Why not just..suspend them?

because that takes time, and you'd also have to unsuspend them at some point otherwise what would be the point in making the cards in the first place?

What time are you talking about? It takes 2 seconds to (un)suspend a card, and you can more accurately specify which words you don't think deserve being repeatedly reviewed, instead of making all words like that.

Well 2 seconds is 2 seconds.
But really, the whole point of anki is that you don't have to micro-manage which cards to review and when you review them. Boosting the interval modifier to something more sensible is something you can do once and then continue as usual (and you can always change it back). You don't have to go back through the card browser and decide which cards to unsuspend.

Anyway even the cards you're getting correct can afford having their intervals boosted.
Really though I just think the SRS idea of pushing things from short to long term memory via increasing spacing is a flawed concept.

My idea was to reduce the number of reviews you had to do so you would have more time to learn new cards but I think I got the math wrong. I dont know maybe it is better to just delete easy cards. Anyway doing reviews when you know 90% of the cards seems like a waste of time compared to learning new ones.

Why did Anki remove the "very easy" butten in 2.0? Setting that to 20d would have been a nice way to avoid repeating easy cards more than necessary.

Is this post concerned with cloze delete or only pure sentence mining?

The reason I ask is that it's pretty easy to amass thousands of cards with cloze delete, but in doing so you can have lots of cards for just one sentence. Would you count the total number of cards, or do all the cards for one sentence only count as one sentence?

In the end it doesn't matter as there is no magical number that leads to fluency, but I am curious as that is basically what the title of this post claims.

Well, it says 10 000 words, not 10 000 cards.

nadiatims Wrote:Btw for people who do use anki, I'd suggest using rikaichan to export words and definitions to a text file at at the click of a button. This is what i used to do when I was still using anki. Cards don't take that long to create.

This is what I do.
1) Type words in text file (1 per line)
2) Copy into editpad.org
3) Use Rikaichan to create 2nd text file with expression and meaning fields
4) Import into Anki
5) Activate "generate readings" function in the browser (part of "Japanese Support" plugin)
6) Flag cards with mistakes (wrong expression, definition or reading) on first review and correct

nadiatims Wrote:Really though I just think the SRS idea of pushing things from short to long term memory via increasing spacing is a flawed concept.

I agree with this. I'm no expert, but I've found that my long term memory is mostly filled things that happened very often, or that simply made a large impression on mind. To be honest, SRS is bogus, but it's free so I can't be mad at it.

Take, for instance, a simple greeting. We normally remember these because they happen all the time. They aren't spaced out..it's literally every day, sometimes several times a day. Words this frequent you will never forget as long as you breathe.

Then there are words that you learn within an unforgettable context. I remember my friend Atsushi loling at me because I used the word "kimochi" and it reminded him of naughty scenes from your typical adult video. To this day ill never forget the word, and I make sure to avoid using it whenever possible lol.

None of this really had to do with the artificial spacing of material. The true benefit of SRS, IMO, is that it gets ppl to actually regularly use flash cards. You could say the SRS itself is a placebo, and the regular attempts at studying are what is truly benefitting learners

amtrack Wrote:None of this really had to do with the artificial spacing of material. The true benefit of SRS, IMO, is that it gets ppl to actually regularly use flash cards. You could say the SRS itself is a placebo, and the regular attempts at studying are what is truly benefitting learners

So basically you're saying that the timing is irrelevant, and the regularity is the only thing that matters at all? That doesn't make any sense to me. Learning things is dependent on a whole bunch of different variables, and you can't just rule one of them out (spacing) in every single context just because it isn't relevant in some of them, like in your aforementioned example.

As with most other things, it's not a black-and-white situation. SRS has its uses, but also its alternatives.

SRS* is a defence against forgetting – nothing more, nothing less. The actual learning and language acquisition takes place outside of SRS.

*I.e. SRS that is used in a way it's meant to be used (Effective learning: Twenty rules of formulating knowledge). People who use sentences in SRS don't follow advice from Piotr Wozniak - their SRS shows them sentences and they effectively do reading within SRS. I don't think this the most effective way of spending your time, but you can't go wrong with it either.

Let me explain a little better why I think the reasoning behind SRSing is flawed (which isn't to say it's useless).

First of all, though I have used these terms quite often myself I don't think there really is such a thing as a clearly defined short and long term memory. Instead I think we have something more like a working memory and a massive cache (or history).

Say you encounter the sum: 1+1=? and try to calculate it. What will happen is your brain will search your cache for the idea of "one", "plus" and "equals" and will be able to access these memories almost instantaneously because they have already been linked to millions of times in countless different situations and contexts. Your brain can then quickly perform the calculation using it's working memory. Now suppose you see the sum: x+y+z=? (x=1, y=2, z=3). In this case your brain checks it's cache and realises that xyz are variables. The calculation is more mentally taxing, because some of the working memory's capacity is devoted to temporarily storing the temporary values for the variables x, y and z which do not yet (and probably never will) exist as deeply connected memories in the brain's history. After the calculation the temporary values are cleared from the working memory. The memory that x, y and z were on this occasion 1, 2 and 3 respectively is cached but accessing this memory will become nearly impossible over time as we have no reason to re-access this data and no further links are established, meanwhile new information
is constantly being added to the cache.

I think it's the same when we learn words. If we've just started learning and don't know any of the words in a sentence, you have to look them all up, keep their meanings in your working memory and figure out what the sentence means. It's mentally taxing. If you already know the words well, the data is quickly retrieved from your history and your working memory can be fully devoted to figuring out the meaning of the sentence.

The problem with SRS is that I don't really think that forcing yourself to encounter the same fact many many times over a short period of time in exactly the same context actually improves the interconnectedness of the memory within the cache which is what really leads to effortless retrievability and usability. You just retrieve some information, pass it through your working memory, do nothing with it and then discard it again. If you go and find new sentences for the word each time you see it in the SRS that's a different matter, but then you may as well just be reading. I think you only have to see a word once for some trace of it to be established in your cache. If you need to remind yourself of the word again because it has become irretrievable, that's fine and normal but I believe more robust secondary connections will be formed if you let some time pass before doing so because you'll be re-accessing a memory that has been pushed deeper into the cache and also if you establish further connections by seeing and using the word in different contexts.

Inny Jan Wrote:SRS* is a defence against forgetting – nothing more, nothing less. The actual learning and language acquisition takes place outside of SRS.

*I.e. SRS that is used in a way it's meant to be used (Effective learning: Twenty rules of formulating knowledge). People who use sentences in SRS don't follow advice from Piotr Wozniak - their SRS shows them sentences and they effectively do reading within SRS. I don't think this the most effective way of spending your time, but you can't go wrong with it either.

Piotr Wozniak is using sentences in most of his examples in that link. Using sentences is not the same as learning sentences.

Sentences should be used to provide context and give concrete-ness for the word being learned (whether it's the reading or writing of the word, or, if you're clozed deleting, both). Personally, I think it should be either reading or writing, not both at the same time (meaning you shouldn't put (...) instead of the word, you should put either the kanji or hiragana).

The context provided by the sentence is important, because it helps the user associate the word with a concept (meaning) as well as a concrete (if possible), rather than an English word. The sentence should of course be a description of a situation in which the concept is concretized (i.e. if you're learning the word "chair", your sentence should be about a chair which may be in a kitchen, someone may be sitting on it, etc.). Your sentence shouldn't be about the concept chair, as in "A chair is a four legged wooden object used for sitting on".

Those who simply use words, are memorizing a dictionary, not learning a language. Obviously, memorizing a dictionary helps in later learning the language as well, but not as much as learning the names of concepts and concretes directly.

P.S. of course, the thing being learned could also be the structure of a sentence, in which case all the other components should be known (or at least explained) already.

Inny Jan Wrote:People who use sentences in SRS don't follow advice from Piotr Wozniak - their SRS shows them sentences and they effectively do reading within SRS.

A more accurate way to put it is that people who use sentences combine some of the benefits of reading (with limitations, of course, because you're not getting the rich context you would from reading, you're only getting some context) with the benefits of SRS-ing, without any drawbacks.

But, again, this assumes that the sentence is known, except for the one word in it, that's being learned. The drawback of using an entire sentence that you don't know (or worse, a sentence with a structure you don't know) is of course huge.

Stansfield123 Wrote:A more accurate way to put it is that people who use sentences combine some of the benefits of reading (with limitations, of course, because you're not getting the rich context you would from reading, you're only getting some context) with the benefits of SRS-ing, without any drawbacks.

I agree with nadiatims when he says:

nadiatims Wrote:I don't really think that forcing yourself to encounter the same fact many many times over a short period of time in exactly the same context actually improves the interconnectedness of the memory

You need context, and many of those, to develop understanding of a word and SRS in that regards is limiting. If you give it a try, I'm sure you will have little problems with remembering what 求める is and how it is used:
それから調理にやたら時間がかかる。食事に求められるものは多々あるが―――栄養バランスだの、しあわせ感だの、心の交流だの―――しかしそのなかでもっとも求められているものは、空腹を満たす、一刻もはやく満たす、ということのはずだが、男はこの一点を忘れがちである。
(料理・角田光代)

or
「悪魔よ。去れ。すべて地上のものは、わたしの求めるところではないのだ。」
(The Teaching of Buddha)

nadiatims Wrote:Let me explain a little better why I think the reasoning behind SRSing is flawed (which isn't to say it's useless).

First of all, though I have used these terms quite often myself I don't think there really is such a thing as a clearly defined short and long term memory. Instead I think we have something more like a working memory and a massive cache (or history).

Say you encounter the sum: 1+1=? and try to calculate it. What will happen is your brain will search your cache for the idea of "one", "plus" and "equals" and will be able to access these memories almost instantaneously because they have already been linked to millions of times in countless different situations and contexts. Your brain can then quickly perform the calculation using it's working memory. Now suppose you see the sum: x+y+z=? (x=1, y=2, z=3). In this case your brain checks it's cache and realises that xyz are variables. The calculation is more mentally taxing, because some of the working memory's capacity is devoted to temporarily storing the temporary values for the variables x, y and z which do not yet (and probably never will) exist as deeply connected memories in the brain's history. After the calculation the temporary values are cleared from the working memory. The memory that x, y and z were on this occasion 1, 2 and 3 respectively is cached but accessing this memory will become nearly impossible over time as we have no reason to re-access this data and no further links are established, meanwhile new information
is constantly being added to the cache.

I think it's the same when we learn words. If we've just started learning and don't know any of the words in a sentence, you have to look them all up, keep their meanings in your working memory and figure out what the sentence means. It's mentally taxing. If you already know the words well, the data is quickly retrieved from your history and your working memory can be fully devoted to figuring out the meaning of the sentence.

The problem with SRS is that I don't really think that forcing yourself to encounter the same fact many many times over a short period of time in exactly the same context actually improves the interconnectedness of the memory within the cache which is what really leads to effortless retrievability and usability. You just retrieve some information, pass it through your working memory, do nothing with it and then discard it again. If you go and find new sentences for the word each time you see it in the SRS that's a different matter, but then you may as well just be reading. I think you only have to see a word once for some trace of it to be established in your cache. If you need to remind yourself of the word again because it has become irretrievable, that's fine and normal but I believe more robust secondary connections will be formed if you let some time pass before doing so because you'll be re-accessing a memory that has been pushed deeper into the cache and also if you establish further connections by seeing and using the word in different contexts.

I think that would be a decent metaphor for acquiring knowledge (as in, filling a newborn's brain, a blank slate, with knowledge of the world around it). It's nowhere near complete, but it's certainly better than the alternative of a "linear storage unit".

However, that's not what language learning is. Learning a new language, for the most part, doesn't fill any blank slates, it just adds one extra label to a bunch of already existing concepts within a person's hierarchy of knowledge. Sure, there are some new concepts, and there is grammar, but, for the most part, it's just new symbols (be it sounds or written symbols) for already existing concepts.

So, vocab learning (again, for the most part, not always) is in fact an exercise in memorizing random sounds and written symbols, rather than the far more complex mechanism of forming new concepts and connecting them to one's hierarchy of knowledge (taking the time to form the many neural connections required to do that). It is, like I said in the above post, important to make sure we connect these symbols directly to the concepts, rather than to the English equivalents of the symbols. In other words, it's important to note that the word "chair" is not the same thing as the abstraction "chair" and all it subsumes (all the concrete chairs that there are), and connect isu to the abstraction directly rather than the English word (this would probably be better illustrated with more abstract concepts than chair, but I'm too lazy to think of any).

nadiatims Wrote:Say you encounter the sum: 1+1=? and try to calculate it. What will happen is your brain will search your cache for the idea of "one", "plus" and "equals" and will be able to access these memories almost instantaneously because they have already been linked to millions of times in countless different situations and contexts.

That is a terrible example, to illustrate language learning, because 1+1=? is the same exact thing in Japanese. So you need to do exactly nothing, to be able to do this in Japanese.

But let's change it to one + one = ?, instead. Now do it in German. Guess what, you know how to answer ein + ein = ? just fine, without doing any of the things you are describing, simply by forming one connection: that ein refers to the same concept as one.

Obviously, there are counter-examples to what I'm saying. There are in fact many Japanese words which refer to concepts that aren't equally matched by an English equivalent. Those are (while often similar to old concepts) new concepts (Inny Jam already seems to be providing an example). You are in fact right, when it comes to those. Same is true for some grammar (not all, though, especially for someone who speaks more than one language already). But the bulk of Japanese does in fact only differ from other languages in the symbols it uses to label the same concepts that we already know about (and I don't just mean basic terms like chair, but also the vast majority of technical and scientific vocabulary - in science, there's a carefully maintained one to one parity among most major world languages, precisely to make cooperation easy - with the exception of the headstrong English/Americans and their units of measurement, of course). So these counter-examples don't contradict me. They just mean that SRS is useful to learn most of a language, but reading is a requirement to learn the nuances.

I was thinking about the title and i think it's not true. I was thinking of the listening aspect. I don't think one year is enough time. I've listened to so much japanese (definitely more than one year's worth) but I still cannot catch as much as japanese as a Japanese person just because they have way more years on me or maybe it's because they can only speak/understand Japanese. I'll ask on chiebukuro what did they say @ this minute/second mark of a drama or anime or talk/variety show and usually somebody answers me. on a few occasions they're also mentioned or say the person in the video said this but they talked really unclearly or the speaker fumbled with the words that it's hard for native japanese speakers to even understand. I feel like most of the time what i wasn't able to catch, a japanese speaker can but on rare occasions there's such poor speaking or whatever that even native speakers can barely catch what they say or they only have a guess as to what they said. So from my experiences of asking on chiebukuro I feel like my listening skill isn't as good as native speaker obviously for obvious reasons. I'm not sure if I'll ever reach their level... I feel like somebody who only knows/listened to japanese is going to be better than somebody who can listen/understand english and japanese at catching japanese no matter how much japanese the english+japanese listening person listens to japanese? I feel like there's a limit...

Stansfield123 Wrote:That is a terrible example, to illustrate language learning, because 1+1=? is the same exact thing in Japanese. So you need to do exactly nothing, to be able to do this in Japanese.

But let's change it to one + one = ?, instead. Now do it in German. Guess what, you know how to answer ein + ein = ? just fine, without doing any of the things you are describing, simply by forming one connection: that ein refers to the same concept as one.

I think you're misunderstanding my point. Maybe I didn't make it clearly enough. I was using maths to explain my idea of working memory and cache, nothing to do with Japanese.

Nevertheless, supposing I didn't know that ein=one, just showing myself the front/back flashcard ein/one isn't going to get it into my brain in the way I want it be so that I can instantly and effortlessly recognise and use it. Something like ein=one or neko=cat (most vocabulary) are indeed extraordinarily simple connections but there is more to knowing a word than just having a barely retrievable trace memory of its meaning in your native tongue. When you truly know a word, you can't help but understand it the instant you hear it, you'll understand the word before it even registers that you're listening to a foreign language. If i learned ein=one while reading a book, and then I encounter it again in another book a month later (maybe even having to look it up again), then hear it on TV, and one day even order ein loaf of bread at a bakery, I'm a lot further along in knowing the word than if I had just been reminded of it 16 times (because i failed it a bunch of times) in the SRS.

Stansfield123 Wrote:But the bulk of Japanese does in fact only differ from other languages in the symbols it uses to label the same concepts that we already know about (and I don't just mean basic terms like chair, but also the vast majority of technical and scientific vocabulary - in science, there's a carefully maintained one to one parity among most major world languages, precisely to make cooperation easy - with the exception of the headstrong English/Americans and their units of measurement, of course).

It's not about the complexity or interconnectedness of a given word within the language. It's a matter of the interconnectedness of the word with other memories in your brain making it instantly and effortlessly retrievable. You could memorise thousands of words in the SRS only to discover that when you read a book, recall of the meaning of many of those words is just out of your reach and that you need to check them again. I'm not saying SRS is useless. I think it's one of many ways to give yourself mass exposure to and start building trace memories of new vocabulary but I don't think SRS really pushes the vocabulary much further than that. And once you recognise that it's reading/hearing/using vocabulary that ultimately gets it deeply rooted in your brain and instantly retrievable, then you realise the limits of SRS. People frequently complain that they get overwhelmed by SRS reviews and I think that is evidence that SRSing isn't all that fun, that SRS alone isn't deeply rooting items into people's memories, and at the end of the day that frequent reviewing of the same material (to "space" it into the long term memory) isn't as great as it has been made out to be.

It seems to me that lot of people think they need to "master" items in the SRS first and then "master" it again in the wild. I think this concept is a flawed idea, because it's only mastery in the wild that has any real value. Use SRS by all means to spam yourself with new vocabulary, but don't worry so much about retention rates or keeping up your reviews and so on.

I'm probably missing most of the subtext here, but it seems as though there is an underlying accepted hypothetical condition for these arguments to be able to transpire, and that it that an individual studying Japanese using an SRS system or not using an SRS system won't also involve an array of outside context for their words/concepts/grammar they are learning in the target language... almost as though there seems to be an arbitrary condition that for these arguments to even reflect reality, they have to be based on unrealistic conditions such than an individual learning a language will limit their exposure to their scheduled study.

Who learns a language to be able to study it better? I have never honestly came across anyone who seem to think they need to "master" something in any given SRS system before they can come across it outside of their SRS study. If someone is curious enough to self-study another language, surely they aren't the type which is able to fit into some hypothetical stereotypical pigeon hole of limiting their language exposure and study to a specific regiment.

It seems to fly in the face of the act itself.

:/

nadiatims Wrote:The problem with SRS is that I don't really think that forcing yourself to encounter the same fact many many times over a short period of time in exactly the same context actually improves the interconnectedness of the memory within the cache which is what really leads to effortless retrievability and usability.

On the contrary, the SRS appeals to a lot of people because it allows you to know what you know. &;I feel like this "cache" is officially called a schema.


It is possible to mimic interconnectedness through interleaving decks e.g. through Mighty Morphin Morphology's i+0 or i+1.

What I got from reading Kraschen is that if you are trying to pick up a language on your own then the way to do it is to make sense out of the Japanese you are reading and listening to. I dont think it really matters how you make sense out of it. But SRS'ing random bits and pieces as a way to prep yourself beforehand doesnt seem like the most direct way to me.

mourei Wrote:So basically you're saying that the timing is irrelevant, and the regularity is the only thing that matters at all? That doesn't make any sense to me. Learning things is dependent on a whole bunch of different variables, and you can't just rule one of them out (spacing) in every single context just because it isn't relevant in some of them, like in your aforementioned example.

As with most other things, it's not a black-and-white situation. SRS has its uses, but also its alternatives.

Learning is not overly complicated. If you do it enough, you'll eventually learn it. It is really that simple. How much you need to do it largely depends on how hard a concept it is for you to grasp. But once you "learn it," forgetting it is extremely difficult. All I'm saying is that the SRS algorithm itself does absolutely nothing for you. What helps you learn is that you "just do it". Does anyone honestly think that doing those reviews 3 days later gives them some kind of learning advantage? Of course not. That's just silly.

To be blunt, SRS is basically "made up BS," but it works in that people believe in it enough to stick with their flash card repetitions. Its nothing more than doing flash cards. Thats it. There's no magic in it beyond that. The only difference is that people think these flash cards are "magic" so they do them vigorously. If you did normal flashcards with as much zeal your results would basically be the same.

Artificially spacing things does no more for your memory than encountering the same fact tens of times a day over several days. This is easily found to be true when examining your own memory and thinking of how you have come to learn as much as you do. In this way you come to see that SRS is the usual "scientific hogwash" that is made up to popularize a method.

There is a reason SRS isnt more popular. There are very very intelligent people in charge of learning, and teaching at universities. They recognize that it is basically bologna, which is why its not widely published. Get the right amount of practice and you learn. Don't get the right amount of practice and you don't. There are thousands of ways to get in your practice, so anything works as long as you do it.