nadiatims wrote:

Why not just practice output by speaking with people..? you know when there are people to speak with.

When you're actually speaking, there is not necessarily any one correct way of saying something and besides how can you predict what kinds of sentences you'll actually have the opportunity/need to say?

The best way to develop speaking skills is by conversing with people and having solid listening skills. This way you're forced to think on the spot and come up with unrehearsed sentences quickly and without reference to a "correct" answer.

I agree and I thought it's what the OP was implying before this became a discussion about how to use SRS even more.

I've been in Japan for a month now, and despite having more vocabulary and grammar than I'll ever be able to use (mostly thanks to SRS), I suck at some of the most simple elements of conversation, and it's got nothing to do with how well I can recall words, which is generally pretty well.

Speaking is a skill, and it's not a skill that reflects simple recall. You need to form good habits, adapt to the person(s) and situation, and be an interesting individual! Flashcards don't emulate this, and they take up a hell of a lot of time. I'm sure there's benefits somewhere, but I'd rather cut to the chase and spend the time trying to make some friends.

@kodorakun:
The Japanese Sensei App has a review mode called "Sentence Tester." Here's an example:

The public transport around here is good.
便、ここ、が、交通、よい、の、は

The user needs to select the words in the proper sequential order. Whether or not it helps with actual speaking is debatable, but it necessarily demands active thinking.

@nadiatims:
Aren't there situations that crop up frequently that involve certain words? Going to a restaurant, visiting someone at the hospital, etc. These different domains have their own associated vocabulary and set phrases. Knowing these stock phrases as well as word collocations can take you quite far, reducing cognitive load and conserving mental effort to only those cases where a new phrase needs to be created. How often do people create a truly unique sentence in a given day?

To return to the musical analogy, thinking on the spot is exactly what jazz musicians must do every time they perform. In order to do what they do, they have to practice extensively. Rather than create entirely new phrases, however, it's been shown that the great musicians used already-known small phrases as building blocks and re-arranged them to create unique solos.

Well I thought the whole blog post emphasized the point of having the production-like cards NOT be about the specific words or phrases as much as it was about the patterns and structure of the language itself. I do try to speak when I get the chance, and I will continue to do so, but this is just about "hacking" the process and trying to make it as efficient as possible. Anyways -- it would be nice to have a discussion on these methods and variations therein if anybody has tried out these methods.

I offered the idea that doing things like Tae Kim, Core 2k/6k and subs2srs is doing output. It seems reasonable that one would shadow what's said or even substitute a different set of words to apply to a new situation (like, instead of sister, brother, mother, you actually use names of people you know and maybe change the verb to something else). So, it's output, but in a very controlled manner with imediate feedback.

Better than the forced production without immediate correction you see in many college texts. For better feedback on production, probably lang-8 or whatever site is arranging language exchange will be the best resource.

kodorakun wrote:

Can anyone comment on their attempts to make "production" cards as discussed in the initial post of this thread?

As much as I hate the article (I pointed out some of the things that are terrible about it, not all), I'm not against production cards. In fact, as of lately, I switched to only adding such cards. I also did Rtk with production cards. (In between, I only did about 1500 sentences with input cards - this includes Tae Kim) However, my output cards are reading -> writing, not English - > Japanese (which, I found, is the least efficient, most frustrating way to do reviews):

1. I identify a word I want to get to use Anki to help me learn it (I identify it somewhere where I came across it by doing pure input - so I'm not just using Anki to learn it, I'm using Anki to help me learn it, while at the same time I'm using the text, song, video, etc. I found it in for the same purpose).

2. I find an example sentence for it (a sentence I understand without further trips to the dictionary or grammar book). I place (well, automatically import, if I can, copy paste otherwise) the sentence into the question field, the reading of the strange word and its translation + the translation of the sentence into the answer field.

3. I switch out the word in the question sentence with its reading, and set the actual Kanji word as the answer instead. But that's it: one word, reproduced on paper based on its reading. I DO write the word down.

This way, I am doing production as far as the writing of the word. I am doing this because I found that it's the only way to use Anki to properly learn the writing of a word (at least using sentences). Input (within Anki!!!) doesn't work well, because you guess the word from context, before you have to bother with identifying the Kanji and figuring out which word it is, based on that (as you sometimes have to, when reading a text).

Note that this doesn't say anything about learning these words by reading, as Krashen suggests. That does work just fine. When reading, the context changes every time you encounter the word, so the only way to identify it is by looking at the word itself. You in fact are performing the minimum effort necessary to gain the ability to recognize that word. So it works just fine.

I'm guessing the studies which attempt to prove that it doesn't work are using a different definition of "learned" than the ability to recognize it in context. Yes, if you have students read a bunch of text, and then ask them to write down the words in it (or recognize them out of any context), they will have problems with that. Naturally. But that's a ridiculous criteria to use, to prove that something doesn't work. There's no need to "fully understand" a word, to understand text that uses it. Needing context to understand it is fine, because in the real world words are always in context. The intended purpose of Krashen's method isn't full mastery of the language, just from input (at least I'm yet to come across him making such absurd claims). Its purpose is to PROGRESS with the language enough that output then becomes a natural, enjoyable activity. And it works. Once you can understand and read a language reasonably well, output becomes fun. Someone who has the ability to (barely) read a difficult English novel, will in fact be able to have fun on a forum like this one, producing at a massive rate no textbook/Anki deck would've allowed him to produce at.

So, to recap my overall point:
1. You need Anki (input and output, but output kept to the minimum necessary, and done in a way that keeps individual card answers simple, as per the philosophy of SRS-ing explained on the Supermemo site), as well as reading practice, audio immersion, and just plain watching media (with subtitles, imho, is fine, so long as it's done in conjunction with the active, daily study of the language - I have never watched media without subs as a beginner, and I progressed just fine) to get you to an intermediate level: a point where real world input is enjoyable to you.

2. Using enjoyable, but challenging (comprehensible, i+1) input (books, audio, and media without subs) to get you to an advanced level, where real world output is enjoyable to you.

3. Using real world output to perfect your mastery of the language.

P.S. I do like the idea of micro-blogging. I'm not gonna do it, because my friends would all throw things at me if I started tweeting them in Japanese, but if you have someone to do it for, it seems like a good idea (so long as you do it carefully - having someone on the other end to correct you would be ideal; if I was a language teacher, monolingual twitter accounts would be mandatory, and I would follow every student and correct all their tweets).

Last edited by Stansfield123 (2013 March 07, 4:04 am)

Tzadeck wrote:

Yeah, already posted about how I worded that badly.  I'm not sure why you wouldn't responding taking into account the most recent post I made about it?

Because I read you first post, and then responded to it right away. It's how I roll. You could've added a note at the end, stating that you re-worded it, you know.

Anyway, I apologize for my confusion.

We all can get obsessed with numbers, and in our attempt to cover a lot of ground, end up diluting our learning across a lot of material, at the expense of true mastery of the essentials.

Of course, external events like exams might necessitate learning a defined set of material fast, but excluding those cases, we may benefit by slowing down, learning less, and focussing more (deliberate practice, anyone?).

Thanks, Aspiring. I just realized, however, that I made a major omission: I forgot to point out that Japanese is a special case, and that cards which go Kanji Sentence ->Reading/Meaning are in fact not pure input cards either. You are in fact producing the reading. ( I just realized it as I was reviewing my now dead Core2k deck, and struggling to remember if 屋根 is yame or yane).

Same can apply to reading: by reading out the full text (be it out loud or just for oneself - which I at least always do, simply because it helps comprehension), the reader is producing each word's reading. This makes it impossible to simply apply studies done about the effectiveness of "incidental learning" by reading, in other languages (still have trouble buying those claims, because the method worked very well for me with other languages too - and I haven't read the hundreds of books that site claims it would take to learn only 1000 words).

Of course, this applies less and less, as one progresses and learns the readings of Kanji. But, at that point, other forms of output become easy to do. There's no need to make life difficult by creating facts with ambiguous answers and using SRS in a way it wasn't intended to be used (to study something other than SIMPLE facts, one at a time).

Last edited by Stansfield123 (2013 March 07, 10:42 am)

Tyreon wrote:

I personally believe that doing "forced" output, without tons of preceding input, leads to reinforcement of wrong patterns.

Just my 2 cents, but I've met foreigners living for decades in my country that make the same mistakes they made when they first arrived, because of forced output.

Even if it does lead to wrong patterns--I don't know if it does--for some people that's not all that important.  I try to be as correct as possible, but I know I make mistakes.  More importantly, I know that it's not all that important.

Learning a language, to me, is about having new experiences, making new friends, and learning new things.  Making mistakes doesn't seem to limit my ability to have new experiences or make new friends.  Waiting around to get tons of input and avoid the reinforcement of wrong patterns would have delayed these things, however, which seems worse to me than using wrong patterns sometimes.

it doesn't matter if you make mistakes when you speak. It's pretty much inevitable and they get less and less common (approach native level) as you improve at the language.

The question is, is it possible or more importantly worthwhile attempting to iron out these mistakes while studying at home with your flashcards. To me the answer is a resounding no, because doing and making production cards will be slower than just doing recognition ones, and when you actually engage in a real conversation there is no predicting what you or the other person will say anyway. Furthermore, natives will tend to meet you half way when your level is still low (ie. they will speak somewhat unnaturally).

When you have a conversation and you're trying to say things, you ultimately draw from your memories of how people say things. So the wider and the more accurate and well connected that web of inputted memories is the easier it will be to access and "produce" those patterns. So rather than rehearsing specific phrases via production cards, it would seem to make a lot more sense read and listen more to build a better understanding of how the words you know are used.

Some people probably do make progress by "production" because they enjoy going out and speaking with people and would otherwise be unmotivated to study. But this kind of production is totally different that production using flashcards imo.
Rather than attempting to converse with your SRS, why not just find language partners to chat with?

Last edited by nadiatims (2013 March 08, 1:31 am)

mmhorii wrote:

@kodorakun:
The Japanese Sensei App has a review mode called "Sentence Tester." Here's an example:

The public transport around here is good.
便、ここ、が、交通、よい、の、は

The user needs to select the words in the proper sequential order. Whether or not it helps with actual speaking is debatable, but it necessarily demands active thinking.

Practice like this can be nearly useless for improving speaking skills if you do it wrong. But it can be a decent, if not perfect, way to access your proficiency. The reason is simple.

There are various strategies to solve the kind of problem in which you sort scrambled words into a grammatically correct sentence. But I'll focus on the two most relevant ones. One is how a native speaker typically approaches when asked to solve this type of problem. The other is the typical strategy beginning learners of a foreign language tend to employ.

When a native speaker solves this type of sorting problem, what they rely on the most is collocations, typical patterns, overall structures that flow well with given words, and similar intuitions about naturalness based on their mental database developed through an extraordinary amount of exposure to the language. For instance, when I try to order the words "便、ここ、が、交通、よい、の、は," my brain automatically notices the very strong collocation between 便 and 交通. Recognizing this, what pops up on my mind is natural-sounding sentences and/or vague ideas about how the correct answer should look like. This naturally forces my brain to notice another collocation between set phrase "交通の便" and "よい." At this point, I know the most natural-sounding sentence structure with these collocational patters indeed gives a grammatically correct sentence. In this case, this strategy immediately gives the most natural and grammatically correct answer, which is most likely the answer the questioner had in mind.

Native speakers' strategy is very much neural network-ish. They rely on collocational proximities between given words to sort them in such a way that, in a sense, the resulting sentence achieves a local minimum of "naturalness entropy."

So, increasing grammatical complexity doesn't really make a problem difficult. If you want to create a problem that is hard for native speakers, a simple way is to use a very dry, highly contrived sentence that is grammatically correct but is not idiomatic and does not evoke any natural context it would be frequently used in. If you want a tricky question, you deliberately use, in an unexpected way, words that collocate very well in the correct answer.

Leaners who haven't been exposed to the language enough can't employ this strategy. They can't tell which words collocate more strongly or where they appear more frequently in an idiomatic sentence. When given two grammatical correct sentences, they can't tell which is more natural or if native speakers would phrase the same ideas that way. So they have to resort to low level information such as grammar, syntax, semantics, and so on. If they use information that a pair of words form an idiom, it's not because the combination is more likely than others in a natural conversation or writing. It's because such a combination is more likely the subject of a question in a standardized test or an exam. The key information they take advantage of is mostly global rules such as "the subject, if not omitted, must be a noun or noun clause" and "an adjective precedes the target noun" and semantics deduced from translation.

Beginners' strategy is quite inefficient because it requires exponential time to reach a complete sentence. Sometimes you might be able to exploit certain information of higher level, but it wouldn't be a general polynomial time algorithm. So, no matter how much you improve your skill to sort words with this approach, in general you're still way too slow for spontaneous conversations. It's a fundamental limitation of this strategy due to computational complexity.

So, if you want to speak spontaneously at a natural speed, this type of exercise never improves the kind of skill you need. You'll see progress at first, and soon hit the unbreakable wall. You're trying to achieve something mathematically impossible.

Native and near-native speakers rarely use low level information such as grammar when speaking their own languages. They use a sub-optimal algorithm that finds a local max of naturalness quickly. It may not always give you a grammatically correct complete sentence. But each chunk of the resulting (possibly incomplete) sentence sounds natural.

Sorting problems may be used as a measure of progress for advanced learners. If you solved a problem the way native speakers would, and if you find problems easy or difficult the same way as native speakers, you know that you already familiarized yourself with tested words, phrases, and sentence structures through input. And if you had to use grammar extensively to sort words, most likely you need more exposure. But using this kind of problem as an exercise for output is nonsense.

Last edited by magamo (2013 March 08, 4:39 am)

@Inny Jan

Maybe my previous post wasn't clear enough. But what I'm saying is that the kind of learning process you described in the latter half of the first paragraph of your post is the one that is required to improve your output skills to a very high level. Like I said, you will make progress at first with the kind of practice you did at high school. But that's one of many things that help get you ready for real learning at best. It can be useful exercises for memorizing some rules, which can be beneficial at an early stage of learning. But it's not output exercises that eventually lead to speaking proficiency if you keep doing.

I'll retract my claim if you give evidence or a convincing argument that shows that you will eventually achieve high proficiency through what you did at high school or that the particular type of exercise I'm talking about is absolutely necessary. But if you can only say that it helps at a beginner/intermediate level, I don't think it counts as "exercises for output" for the reason I already described. It may be a kind of exercise that helps memorize rules though.

Edit: Sorry, but I don't follow your second paragraph. I have no idea what nadiatims said or what you mean by "you can't generalize."

Also, "X is nonsense" and "using X for Y is nonsense" are two different things. The type of exercise we're talking about may be of value for some purposes. But I don't think using it as exercises for output directly helps you achieve the intended goal.

Edit2: To make my point clearer, think of memorizing English translations of Japanese words. Translations given in your bilingual dictionary may be useful as an initial foothold for comprehension. I don't think it's crucial for foreign language learning, but I wouldn't say it's completely useless to anyone at any level. But it's obvious that cementing translations into your head won't get you very far in terms of proficiency in comprehension unless it is followed by actual learning of your target language.

Now, consider multiple choice exercises where you're asked to pick the correct English translation of a Japanese word. Memorizing the most typical translations of basic words may be helpful. But is this kind of exercise an "exercise for comprehension"? If you spend a lot of time doing this type of translation practice hoping you'll eventually develop advanced comprehension skills, you will surely be disappointed. So while this kind of exercise may have some value, it's nonsense to do it as an exercise for comprehension.

Last edited by magamo (2013 March 08, 7:35 am)