Anki vs. Gold Lists vs. Iversen-style Wordlists

Bob asked me in a another thread on Anki, my opinion of Gold Lists. I have never used either the Gold List or the Iversen Wordlist methods, but have read many posts here and on HTLAL about them with interest. I thought it might be useful to compare and contrast the methods. But first I would like to throw out for discussion some metrics, both objective and subjective, to compare and contrast them.

In no particular order I would propose the following metrics, and hopefully others will chime in, after which perhaps we can then in fact compare/contrast the methods:

-initial learning of words and time spent on same
-how time is spent long term (i.e. time per word mostly spent up front vs. longer-term in smaller intervals)
-practical if not theoretical limits to how many words can be studied per day
-format: physical analog vs. digital
-long term rates of sticking to methods by users of same (if known or can be estimated)
-flexibility: i.e. can method be used by different learners in different ways
-learner success stories with each method (attrition rates are probably fairly high for any method)
-simplicity & difficulty of methods
-scale and context: can example phrases/sentences be used
-portability: can be used in many places with either physical/digital media
-scientific basis for either overall method or components of them
-overall time efficiency

I realize that as an Anki fanatic, my own biases are going to surface in such a list of metrics, so hopefully others will comment on metrics here. Also note that all methods are based on the assumption by most users that they are not used in isolation, but are part of an overall learning plan with the vocabulary method supporting course work and other extensive/intensive activities. To be comprehensive, extensive reading should probably also be included in such a comparison, when it is the primary method of vocabulary acquisition after initial courses.

yes, I was holding myself back until you said you were looking for an alternative. Makes sense to make this a new topic though.

Goldlist

-initial time: Initially .8 minutes per word. It's probably less than this because I rarely use the full recommended 20 minutes. If it's a cognate (and not a false friend, or some other oddity) I don't even waste my time. Or words that mean the same thing and are spelled almost the same.

-long term: There's some weird math here, but I'd say less than 8 minutes per word on average to complete the whole process. (including the initial time)

-practical : If I have the time, 150 new words per day, and close to that for review.

-format: physical (although I'd be open to electronic where you have to type it out, an audio version might be cool too)

-long term sticking rates: reading a passage with words I've Goldlisted on the average usually take 2 read throughs to get it to stick.

-flexibility: Hmmm, wait a least 2 weeks to review, work no longer than 20 minutes at a time. Pretty much everything else can be fiddled with. I might even fiddle with the 20 minutes.

-attrition rates: I have no idea. I don't think allot of people have tried it. I saw some that tried one list to see if it worked, could not remember enough words pat (which are not the instructions) and then declared Goldlist useless.

-simplicity: the difficulty is deciding what to do with the instructions "choose which 8 words you remember the best." I think this really confuses people. I may have found a way around this.

-scale and context: can you use sentences? yes! can you use paragraphs? yes (the creator did this for Japanese Kanji, though this feels quite different)

-portability: Get a good notebook and it's anywhere. I've even used this in power outages.

-scientific basis: kind of loose. the name Ebbinghouse comes up, but the details of how this works have been by experiment and observation.

-overall time efficiency: not sure what to say here. Once you have put in all the words you want, assuming your schedule does not change, this will be done in 3 months. I would think a year would be need for a busy person to list enough words for "advanced fluency".

Thanks for the reply Bob. I am not really actively looking for an alternative, merely open to one, which includes a hybrid one (like Iversen to initially learn and then Anki/GL), or a tweak of an existing one. However when I restart Spanish soon, I may try something else, at least initially.

I have used my own wordlist setup since 2007 (with minor changes, mostly concerning the repetitions), and I have not felt any need to supplement it with other systematic vocabulary learning methods. However as mentioned by Peregrinus no memorizing method is worth anything in itself - you ALSO need to get a reasonable amount of input (written and oral) AND you need to use the language activively.

Before responding to the questions I would like to refer to Ebbinghaus' famous curve for forgetting things: if you learn something and then don't do anything it will disappear over time, and there is a curve that shows this gradual decay. You can counteract this decay by regularly 'revisiting' your word stores, and both Anki and goldlists work on the principle that you don't do much at first except writing words down or entering them in some kind of software, but then there is a process that takes care of the repetitions which in the long run will make those words stick in your longterm memory. There is however one thing which I find problematic: when you declare that you know a word it will be eliminated from the repetition cycle. But up to the Berlin gathering I did some studies using my own system, where I contrary to normal practice used three repetition rounds, and it turned out that that the 'overlap' between each round was fairly small - or in other words: the words I had forgotten in round two and three were typically NOT the words I had problems with in round one. No, precisely because I focused on forgotten words I learnt most of them so I mostly remembered them in the following phases. The words I had trouble with in those phases were typically among those I had accepted as known in the first round .. and if I had done Anki or goldlist precisely those words would have been eliminated!

My own system (with three columns in round one and groups of 5-7 words) is based on the assumption that the really brutal loss happens within a minut after you have seen a word, and whatever happens after that phase is less important. So therefore I jot unknown down on the spot somewhere and later use the words I have jotted down in a wordlist, where I can memorize them properly - using associations and all the other tricks in the book.

I have normally just done one repetition (using a variety of methods and layouts), and the idea behind this is that my vocabulary grows faster by adding new words to the process and doing one repetition afterwards than it would by sticking to the same list and repeating it two or three times. But during the experiment I mentioned (with Serbian words) I used a fairly simple repetition layout: for the first repetition I simply copied the foreign words from the original list, and if I was in doubt about the meaning of one of them I added a translation, else I didn't, so it was easy to count how many words I had lost. And for the second repetition round I did simple added a dot if a forgotten word already had been marked, otherwise I added a translation. So using this method it was easy to calculate not only my loss, but also the overlap. For the third repetition round I copied the words again from the repetition sheet to a new sheet - nb. all the words, contrary to the goldlist method - and marked the forgotten ones with a dot or a translation.

Normally I say that the first repetition should be done about one day after the creation of the original list, and in this case I did a second round two days later. However the third round was done at least a week later. And the results were quite interesting: in repetition round one I typically had doubts about or simply forgotten approx. 20% of the words. But in the following round I was down to 12% dubious/forgotten words (with 3% overlap), and in round three I had 13% (with a 4% overlap to the two previous rounds). It should however be added that I did something similar after Berlin with the middle section of my Serbian-English dictionary, and here my loss in repetition round one was 17% and 16% in round 2 (no third round), maybe because I was less focused on the results of round one).

And now for the questions:

-initial learning of words and time spent on same:
Jotting down takes very little time, and writing and memorizing three columns with maybe thirty words will typically take around 10-30 minutes, depending on the time I spend reading the dictionary articles and checking my memory (I actually don't think much about my time expenditure so I had to do a test here and now to check these indications!)

-how time is spent long term (i.e. time per word mostly spent up front vs. longer-term in smaller intervals)
With the simple repetition method I described above each column of 30-40 words shouldn't take more than 10 minutes.

-practical if not theoretical limits to how many words can be studied per day
I regularly do one half sheet (sometimes two) in a given target language - then I want to do something different with another language. But in some cases I have taken a dictionary and moved through a number of pages, and then the 'production rate' grows. In that situation it is totally feasible to do several hundred words in an evening.

-format: physical analog vs. digital
I always write my lists by hand on paper, but doing wordlists using a keyboard may be more natural for younger persons

-long term rates of sticking to methods by users of same (if known or can be estimated)
See the figures above, which are relevant because I didn't do other activities in Serbian in the time right before Berlin. If you do the recommended thing - namely reading, listening, maybe even speaking your target language - there may be some confusion about the growth of your vocabulary. Is the progress you measure due to the other activities, to doing a wordlist or to doing both?

-flexibility: i.e. can method be used by different learners in different ways
I have seen some people suggest extra columns for Japanese or Chinese wordlists, but the format for the three columns is fairly stable. However the repetition can be done in different ways. I have experimented with two columns (own + target language) and for a time I tried going back to the sources for textbased lists, but I now prefer the simple method outlined above: one column with the foreign words plus translations of dubious/forgotten words - and only them. The simpler format makes it less likely that I will 'forget' doing the repetition, which is absolutely necessary for longterm retention.

-learner success stories with each method (attrition rates are probably fairly high for any method)
A few, but I don't have a long list of users to refer to

-simplicity & difficulty of methods
Of course it is easier just to write a bunch of words down with translations (as in a typical textbook) - but I have tested this and it didn't work very well. So I prefer my slightly more complicated layout.

-scale and context: can example phrases/sentences be used
I wouldn't use it for complete sentences, but expressions can be accomodated simply by using wider columns. However mixing words and expressions is problematic. I mostly stick to the words and expect to learn phrases from my reading, but if I wanted to include expressions I think I would jot both single words and expressions down while reading, but keep them separate in my wordlists.

-portability: can be used in many places with either physical/digital media
Wordlists can be written both on paper and on a gadget of some kind, but in practice I have only used a sheet of paper and 2 pens in different colours. You can't make them while walking or standing up in a crowded place, but I wouldn't even want to do them in such conditions.

-scientific basis for either overall method or components of them
None

-overall time efficiency
Check the figures above. And add a few more: I did a word count exercise during my Serbian pre-Berlin adventure, using pages from the 'studied' part of the alphabet and from the 'unstudied' part. Some words from the first letters were culled from an Serbian-Italian dictionary, but a statistics based only on my Serbian-English dictionary (with around 12.000 headwords) gave no less than 67% known words from the studied part and only 33% from the unstudied part of the alphabet. Given that I only included approx. a third of the words in my wordlists (1724 words from А to И including those from the Serbian-Italian dictionary) it is pretty clear that even working diligently with a dictionary can have positive results that go beyond the words actually included in some kind of wordlist.

I would like to reply to some points of Iversen's excellent and detailed explanation above, but first I am going to use the criteria I gave above and evaluate Anki from my viewpoint.

-initial learning of words and time spent on same
Varies. One can either prelearn them some way as I do by using the preview option in the browser first to step forward and back over them, or just do it hard core and go through them as normal with Anki. Many on HTLAL maintain that Anki should not be used for initial learning, and probably they are right, because that is the most time-consuming step with repeated failures and reviews at the very beginning. I myself just prefer to use all digital means.

-how time is spent long term (i.e. time per word mostly spent up front vs. longer-term in smaller intervals)
I suspect apart from the first introduction of a word in Anki, that long-term time per word is not very much. Even if you failed a word every other day for a month, those reviews take only seconds, which add up to only minutes. So I think it is efficient long term.

-practical if not theoretical limits to how many words can be studied per day
Some users have reported learning as many as 250+ words in a day, while for myself I find 30 a comfortable pace (may be 60 one day and 0 the next), but often do 40-50, and have done 100. Even 100 is not overly difficult *the first time*. But a pace like that will mushroom the daily reviews and time spent on them very fast. If you have literally hours a day to study vocabulary, then other methods, including extensive reading, may be better.

-format: physical analog vs. digital
Obviously all digital.

-long term rates of sticking to methods by users of same (if known or can be estimated)
Anki has a large installed base and one can find on the net reports of tons of people using it and sticking with it for longish periods. Nonetheless, the attrition rate has to be high just as it is with sticking to learning a language in general. For people who only want to learn 3000 words, perhaps the sticking rate is far higher.

-flexibility: i.e. can method be used by different learners in different ways
While the creator of Anki is fairly rigid in his views of how it ought to be used, there are ways to used it differently, as with regular notes vs cloze deletions, and being able to tweak the algorithm to some extent. And as I mentioned in the thread regarding minimal failing, one could do it that way. Where Anki is inflexible is the requirement to use it daily lest the review pile up (there is a way to deal with this slowly by only restarting so many cards per day and the program will give you the ones that came due first). In a way that is good as it motivates one to keep at it. But the Gold List and Iversen wordlists win out for scheduling flexibility.

-learner success stories with each method (attrition rates are probably fairly high for any method)
There are many reported, including myself. Starting around the beginning of June last year, I have learned or partially learned almost 11,000 words in German now. It is obvious though that orders of magnitude more people use Anki or similar SRS programs than use GL or Iversen lists, so naturally one finds more success stories.

-simplicity & difficulty of methods
Anki is as simple or as difficult as one wishes to make it. One can have either simple definitions and no example sentences, or very detailed cards as I create. Anki can automatically create a card in both directions, although I only use L2>L1. How difficult one finds it overall is probably related to how well one uses computer programs in general, and especially shareware/freeware which involves more technical ability in both setup and tweaking. There is plenty of help to be had online, but no one is going to hold your hand.

-scale and context: can example phrases/sentences be used
Yes, either added to individual word cards or as the main entry themselves.

-portability: can be used in many places with either physical/digital media
With Ankiweb one can use that app to access your decks which are stored online, so yes it can be portable. I myself only use it on the desktop, as I prefer to listen to L2 audio when away.

-scientific basis for either overall method or components of them
Ebbinghaus curve and the Supermemo algorithm used to implement it for SRS.

-overall time efficiency
Time involved is both for creation of decks, unless using a shared deck, and the review time itself. Having learned partially or fully almost 11K cards now, it still takes me in the 40-60 minute range depending on how many new cards I have learned recently. New card fails and reviews are the time hog, as with older cards when you fail them, you also recognize them much more quickly on the next review. If you stop learning new cards, review times go down quickly. Overall I view an hour a day for creation and an hour for review as very efficient, as long as one does not make that the sole learning activity, in which case Anki won't be reinforced by other, preferably native, materials. For learning the required amount of vocabulary to hit the lexical threshold (15-20K words), it has to be efficient for the amount of time involved supplemented by a modest amount of extensive reading, as opposed to hours and hours a day of ER only. Whether it is more time efficient than either GL or Iversen WL, is the question.

I didn't list excitement/boredom as a criterion, because I don't really care. An hour of Anki reviews is tedious, no question. But the results, accrued little by little over time, are in fact exciting.

Edit to add: One additional and to me substantial, benefit of Anki, is that it functions as my private dictionary. I tweak cards all the time to change the order of various elements of the definition, chiefly so that the most likely to be encountered comes first. And I add example sentences if I think they add value to what is already there. I use 2 or more online dictionaries as sources, and believe my synthesis of them is the most accurate (once having enough experience in the language to hopefully accurately discern same). Naturally, I would expect professionally edited printed dictionaries (and their usually pay-for-use digital versions) to be the most accurate. But my own dictionary that I create is as accurate as I can make it, and in the form most convenient to me to use.

Before I reply later to some points in Iversen's post, I want to introduce another method created by DaraghM on HTLAL, and which Iversen commented on favorably, because I want to reference it when I respond.

This method is The three dictionary technique. You take 3 dictionaries of different sizes, say 20K, 60K and 100K entries respectively.

Quoting DaraghM's instructions:

I am not sure this can be done on the computer efficiently unless one can get pdf files of various dictionaries. Using typical online dictionaries like wordreference.com, Collins or dict.cc, there is too great a waste of time in mouse clicks and scrolling. And for smaller languages, perhaps it might even be difficult to find 3 such printed dictionaries to give that spread. I find it an enticing method though, but not enough to quit Anki (at least for German).

With German the past year I have only been concentrating on passive skills. I have never seen much evidence that active skills teach you language (as in speaking/writing), but perhaps I have just not paid close enough attention. And the way you use your lists includes an active component, i.e. translating from your L1 back to L2. Many use Anki the same way, i.e. they translate both ways, but I never have. However the principle must be similar to an Assimil active wave and there is probably a lot to be said for translating L1>L2, at least after you have some experience with a language. I could generate reverse cards in Anki, or reverse the deck (technically note type which affects multiple decks), and perhaps I should.

True for GL, but not for Anki. A word will come up again, just far into the future according to the cycle prescribed by the algorithm and tweaked by the user. I did however propose in the thread on minimal failing with Anki to also move very easy cards to a different archive deck where they would be eliminated.

This is a key insight, the "brutal" loss in the first minute. It points to why with Anki one has so many fails and reviews of new words early in the cycle. It is the most time-consuming and frustrating part. While some words which previously were marked as known get forgotten some time later, they are quickly remembered again and one marks them easier than a newer card, thus shortening its new cycle.

Regarding memory associations and tricks, I don't use them, though perhaps I should. It does not come naturally to me, and I tend to look at vocabulary analytically, which is only helpful only once I know or recognize some base component in a new word.

So this is the key difference, i.e. spending more time up front learning a word well, or relying on much shorter periods of time over a longer span to gradually make those words known. I can't really accurately estimate how much total time gets invested on average per word with Anki in the course of say a year, though perhaps I could devise such a test.

Again though, you expend more time up front, so it is unclear whether your wordlist method or Anki vary significantly in total time on a word in the course of a year.

This is an interesting and objective test of your method. As you have commented on before, the overlap phenomena, i.e. that a perfect overlap does not exist, is intriguing, and also occurs with Anki, and presumably GL.

While your loss rate is impressively low, over time it has to add up with the goal we both have of 20K words. I am unwilling to winnow out those hard words, especially in higher frequency bands. However for lower frequency words, say over 6-10K, it probably does not matter since the relative frequency differences are not that great. And the strong point of your winnowing is precisely as you said, that it allows you to use the time instead for more new words. This indicates to me that perhaps I should be willing past a certain frequency threshold (if it can be estimated), to in fact winnow out cards, at least by moving them to a different deck that is not active.

So that's 20 minutes for doing 2 columns, about the middle of the 10-30 minute range you give. 20 minutes for 30-40 new words plus whatever time the next review takes the next day, is an efficient rate indeed, even with a 15% or so loss rate, when taking into account that you are then done with learning those words. Again for me at least, I might lack the ability to use memory associations and tricks to learn them that fast.

It would be easy to implement your lists on a spreadsheet, including just narrowing a column so that it can't be read temporarily. One could even use logical fields to denote finished and then resort a very long list of words to be learned. Based on my experience learning Mandarin decades ago, I believe physically writing characters to be extremely helpful for long term retention. Whether that is so with an alphabet language as well I am not sure, i.e. whether typing is as good as writing with your method.

With all the time and posts you have generously made on your method at HTLAL and now here, I don't remember very many learners in posts on the topic or in their logs, saying they stick with your wordlist method for a meaningful period, i.e. a year as opposed to a month or two, and the same for Gold List. Probably only the fact that Anki has shared decks explains higher sticking rates for it, and even then maybe not. There are just far more people who have at least tried Anki than your method or GL.

I think you do have several elements of your method that have a plausible scientific basis, even if unproven via a study. You do take into account the Ebbinghaus curve and seek to counter it by spending more time up front on new words. You use small groups of 5-7 words at a time in a session, which corresponds to studies showing one can only keep so many things in short-term working memory at a time. You right them down, which may involve muscle memory. And you have experimented and tinkered with your method over the years and noticed a corresponding increase in efficiency for yourself. While that is only a n=1 sample size, it is one by a very accomplished language learner, so it counts too.

The fact that you do measure your method by dictionary headword tests and the tests you noted above in your post, give an objective basis to your method that is lacking in the usual subjective anecdotal assertions for various methods/topics where the persons doing the asserting have a vested financial or professional interest in making those assertions (the 3000 = fluent crowd).

This is where I wanted to reference the 3 dictionary method I posted about above. Since you use dictionaries so much for new words, you get a lot of the same benefits that come from reading entire entries and example sentences. I wonder what your results would be if someone just handed you a 20K word list for a new language and you used only that. I suspect that they would not be as good. Using dictionaries so much may be a non-explicit but important part of your method.

While I look up all my own definitions for my Anki decks, using online dictionaries, I don't get the benefit of seeing neighboring words with connections to the word I am looking up. And you may be using better quality dictionaries.

I marked three points for comment in Peregrinus' answer above, namely that he only goes for passive knowledge of German, doesn't do much associating or similar tricks and finally, that he wonders what my results would be if I only were handed a 20K wordlist for a new language and only used that. Those three questions may seem unrelated, but they aren't.

If you only go for passive skills then you don't really have to care about the syntactic fabric below the words - typically represented by 'grammar words', more or less irregular forms of a few very common words and things like word order. You can get an awful lot of meaning out of a text just by guesswork based om relatively rare 'meaning words', and that kind of words are precisely the the kind of words you learn through wordlists or SRS. But you can't make the language in question active without also learning the 'grammar words' and the grammar that specifies their use. That's why I recommend that beginners shouldn't do wordlists etc. based on a dictionary - they should work with texts, preferable using bilingual versions (which you can produce yourself with the help of bilingual homepages or machine translation), and this will result in such an avalanche of unknown words that it won't be necessary to use other sources. I would nevevrtheles mention one: I start speaking late, but try to think in my target languages much earlier. And at least when I'm sitting in my armchair I often jot down words I missed while trying to think a certain thing, and then I look a whole bunch of such missing words up in one go. The result is of course destined to go directly into a wordlist - it is 100% certain that I also will need those words and expressions later on.

Another facet of working with texts is that I may want to use an expression with a certain word as a memory aid - for instance I want to remember a German substantive, and then I memorize it with an article and an adjective to hammer down which gender it is. I don't quote a whole sequence of words in my wordlists - at most I indicate the gender if it isn't clear from its form or a few simple rules - but while I study the original text and jot things down in the margin it doesn't cost anything to let 3 or 4 words pass through your mind. As for wordlists directly from dictionaries: well, at some point I get tired of looking words up all over the dictionary, and then I think that it would be much easier just to memorize a whole string of words from one page in it. And at that point I typically also have learnt enough words to make it possible to make useful associations to other words in the target language (not least words with the same root).

And from that it is just one step to discuss associations and other tricks. I have seen some reports that indicate that the use of associations actually helps people to remember words, but compared to the unanimous recommendations from memory artists the 'scientific' reports are rather vague. As far as I can see this is due to some glaring errors in the ideas some researchers have about associations. For instance I have seen several reports where the research team suggested the associations to be used, which is absolutely idiotic. In one case the purpose of a study was to test whether there is such a thing as visual learners, so the test team showed pictures of dogs and horses and other things with the words - and no, it didn't help, so the conclusion was that visual cues don't have an effect. But in reality they just asked people to remember two things instead of one, where the relevant would be to let the test persons connect a test word to an image they already had in their mind (which just needed an instruction to do so and not a lot of prefabricated images). I know from personal experience that it is the action of connecting something new to something already known that establishes relevant and useful associations - not putting two new things on a table instead of one. Other misguided researchers let people memorize nonsense syllables, which are notoriously difficult to memorize with any method, including associations. But language learning is NOT a question of memorizing nonsense - it is a game where you convert partly or fullblown nonsense into meaningful units.

There are basically two ways to use associations. Memory artist typically have memorized one complete system of entities (f.ex. a 'memory palace'), and then they 'attach' the things they are asked to remember to elements in that system. This system is excellent if you need to remember a number of known things in a fixed order, but outside morphological tables language learners don't have to learn things in a particular order, and their 'target objects' are a mixture of halfway known words and totally incomprehensible chunks. Here it is more logical to run through a number of things which might remind you of the target word - and the translation into your own language is of course one of these things, together with loose associations based on the sound or spelling. Sometimes it is possible to create a bizarre 'story' that covers the whole target, but in my experience it is enough to find memory hooks that references some of it - typically the beginning or a word root in the middle. If you spend too much effort on making 'silly stories' you'll get tired and drop the exercise before time - better go for something less ambitious and be able to carry on forever.

I had to abstract Iversen's post above because he covered so many important points, so that I could respond more appropriately. A couple of items, like thinking in the target language and the sources of vocabulary for any of the 3 methods discussed here, are a little beyond the scope of this thread and deserving of their own detailed discussion. However as he points out, working with texts allows one to use the context of words found as part of the mnemonic association. This is why I try to add example phrases and sentences to my Anki cards, to provide some context, even if not the context in which I actually found a word.

If I understood correctly, the connection between my focus on passive skills only for the time being, and mnemonic associations and tricks, is that those associations and tricks are more useful for activation of vocabulary, i.e. for speaking and writing, than they are for passive skills, because passive activities coupled with non-grammar words learned from word lists, GL or Anki, allow one to in fact ignore much of the grammar associated words, which is what I have done (although in classical Latin poetry and prose, which lacks the greater use of prepositions centuries later, a looser word order often requires more knowledge of grammar just for reading - and the same with any highly inflected language in literary works). While I started with a brief grammar overview, I am only now making an effort to make a more thorough revision of grammar. This too I think is the subject for another thread, so I will comment on some other aspects of Iversen's post in more detail.

Mnemonic associations and tricks is a subject of great interest to me, and among other books on the topic, I have read Spence's The Memory Palace of Matteo Ricci, and Yate's The Art of Memory, the latter of which is an excellent history of mnemonic techniques of earlier times. I have made experiments with the approaches Iversen mentions of using a memory palace, and of creating bizarre images. And despite my interest in mnemonic techniques, as Iversen concludes, those techniques of memory palaces and bizarre visual associations are time wasting and tiring. Part of the reason that earlier times favored such techniques can possibly be assigned to the factors of the expense of paper and books, and the lack of our digital resources.

This then leaves other types of associations. As I said earlier, I like to approach language and language learning as analytically as possible, and I too look for morphological associations to use. And I look at the example sentences I have added to most cards. But as Iversen says, one must first have a basic stock of vocabulary and morphological knowledge (the reason Kato Lomb started with a dictionary to learn same though advising against its overuse later for lookups). With German the hardest class of vocabulary for me by far, is verbs. There is ein-this and and ein-that. This-setzen and that-setzen. While a knowledge of prefixes can often help in learning a related form of a verb once the base form is known, such is not the case enough of the time to matter a lot. Additionally, many verbs, just as in other languages, often have multiple meanings which don't seem to be connected in any way (though I try only to hook one meaning at first). Germans of course learn einsetzen and ansetzen and eingehen and angehen as lexical chunks, whereas I can't help but try to always decompose the verb into a prefix which is used for hundreds of verbs, and a base form which has anywhere from a few to dozens of related prefixed forms. So if the prefix is not overly helpful, and the meaning lacks a clear connection to the base form, or the base form is less common, then a word becomes tough to learn.

Naturally as I have progressed in German I learn one word in a word family, and can relate it to others. Some word families contains lots of related words, as in verbs, substantives, adverbs/adjectives, past participles used independently, etc. Often the case is that one or two of those forms in a word family have a far higher frequency than the others. But some word families are both fairly small, and have no members that occur very often. So you won't even come across such a family's most frequent member in extensive reading as often as needed to help cement its meaning well enough to assist when coming across the second most frequent member.

I learned 59 new words yesterday, harvested from the novel I am reading as discussed in my German log. Many of the words are easy, as I can make morphological associations. And in fact I did look them up quickly as I read. But many words, often synonyms for more commonly used words, I have never seen before in any related form. I made more of an effort when previewing the group, to just visualize its meaning, such as kläffen, which means to yap, yelp or bark. I naturally visualized a dog yapping, and it helped to learn it. But as Iversen noted, that really is just putting two things on the table to memorize. With his word lists, his loss rate, while low, probably represents over time a substantial loss of such tougher words. And the same with Gold List where some portion of the 70% of words in a list that are rated as better known, in fact end up being forgotten, but without a chance to relearn them, since they are eliminated completely. Anki however, unless one deletes a word, never forgets, and never lets you forget. The algorithm only pushes better known words ever further into the future, but it never totally eliminates them. One can argue that such is in fact the case once one gets past a year or some such period, but under 6 months for sure, Anki presents even better known words many, many times.

The key strength of Iversen's method to me is the up front care taken to counteract what he called the "brutal forgetting" that occurs in the first minute after new word is studied. Neither Anki nor GL address this issue. Theoretically Anki could, but it would require that one review a word in multiple sessions the first day and following couple of days, provided that one as well could implement Iversen's working in small groups of 5-7 words at a time. I am going to strongly consider using Iversen's method at the beginning. Even with words already entered into my Anki queue, I can easily just open the deck in browser mode and write down small groups at a time to work with. Then when first learning them officially in Anki, just choose the longest interval (which I have set to 2 days), indicating it is then very easy. The next day I would review my Iversen style list, and the following day be presented with the first Anki review, which would act as a control to gauge how well using Iversen's method up front was working for me. If I could match his success rate, that would involve choosing the longest intervals for most words for the subsequent reviews, reducing both daily numbers of reviews and total review time, at the expense of the additional time spent up front.

As for associations, Iversen knows a large number of languages, many of which are related to each other, which allows him to form associations that I can not. While I have a very large English passive vocabulary which allows me from its Anglo-Saxon inputs to discern many German words, that only goes so far. So I suspect that in the associations and tricks department for remembering tough to remember words, I am always going to be in the corner with a dunce hat on, just relying on the dogged persistence of daily Anki use to overcome it, just not as efficiently. However I will in fact learn those words, since they won't be eliminated.

This is actually a problem in many languages. Ultimately you have to learn such words as individual entities rather than inflections of a root, where the meaning is clear to everyone. But remembering or even pronouncing long words is hard, and being able to cut such words into wellknown pieces is one way of making them more memorable - even when the meaning has to be learned by heart. And even where the meaning couldn't be guessed from the parts, there may be some logic in the combination which can be used as a memory hook. One interesting corollary of this is that the same rule applies to idiomatic expresions: we often hear that they should be learnt as complete expressions because their meaning can't be guessed from the parts alone, and also because you couldn't know beforehand that a certain combination would become popular. But it is always easier to remember a string of known elements than one immensely long string without any inner divisions, and even though the total meaning may be unexpected it is rarely totally devoid of logic.

I think you should imagine a dog whose barking sounds like "kläff, kläff, kläff" - that's better than a dog that says "bow wow" and the word "kläffen" side by side.

The combination of wordlist memorization and Anki has been suggested before, and if I already used Anki I would definitely try it out. The only problem is that you have to write the words in two places, but that could in itself be beneficial. When you speak about "browser mode" I assume that you then can choose to see newly added words first, and then it shouldn't be too much trouble to use such a screenful new words as your source. But only if it is possible to see the translations too, of course.

When we discuss mnemotechnic tricks including associations it should be clear that these can be used in any learning method, also in the case where some authority figure claims that they are unnecessary or superfluous. They are useful and very easy to use once you get into the habit, and you just have to keep the game at a level where you don't feel it as a chore to invent new and strikingly original associations all the time. Associations are even useful when you forget them because they make words 'ring a bell' the next time you see them, irrespective of whether you remember the original association or not.

Prefixes are more meaningful than they often appear, but by the time you know the language well enough to identify their meanings, you've already gone through the hard part. So it's really a job for a teacher or a book, but such books don't seem to exist. The same goes for "particles" with English phrasal verbs.

Gradint does this. I'm not a fan of the stupid big blank gaps it generates in the first few lessons (I always edit them out in Audacity) but overall, the approach is an interesting one.

An example: the Latin prefix re- has two main meanings, captured in the difference between returning and resending a letter. These meanings are almost diametrically opposed, because in Germanic English, they're "send back" and "send again". Do the same thing, vs do the opposite. But just because you can't predict whether "retickle" would mean to tickle some who'd just tickled you or to start tickling a person you'd previously tickled doesn't mean that the prefix isn't a meaningful morpheme.

But then someone might want to point out words like "reserve" that don't have a redoing or a returning involved. Except that to reserve something is to hold it back or to have it held back for you, so even if the underlying logic is obscure, it's still strong enough to have survived in both the Latin and Germanic sides of English.

And consider... why do we put a fire out in English? I din't understand the logic, but what I do know is that even the Romans put their fires out -- they extinguished them.

For this subtle logic to survive intact in different language families must mean that it's built on a pretty solid foundation, making it a fundamental component of the languages,. While native speakrsmay not be consciously aware of the logic, it is part of their internal model of the language, and if we are trying to reconstruct the same model, then that logic should be part of the blueprint -- ie the teaching.

That is a very nice association indeed. Part of the problem for me I guess is that I tend to think very concretely, i.e. very literally. While I rarely have problems discerning subtle humor, as in satire and irony, if an expression has two possible non-humorous meanings, I am going to assume the most literal of them, usually by focusing on only one detail rather than the whole. Obviously more concrete objects lend themselves to associations, but for abstract ideas, I am not sure that I am much good at it unless a morphological associations seems evident. Still I guess I just need practice, and as you suggest, at a level where it both goes fast enough to not become tedious, yet remains effective. So, Übung macht den Meister goes for the ability to easily make associations just as with other skills.

Yes browser mode allows one to see a table listing of records as with most databases, and all the fields that comprise a record, which means for both sides of a card.

There are two types of associations in mnemonics, i.e. permanent structures and temporary ones. The temporary ones, which Cainntear would probably refer to as "scaffolding", based on some of his blog entries, are supposed to just help one learn a word or concept until it becomes a natural part of one's long-term memory, after which it falls away.

To sum up what to me are the aspects of the Iversen method that are the basis of its effectiveness:

-spending more time up front to combat the forgetting that occurs in the first minute
-working in small groups of 5-7 words at a time
-double translation.

I think that before I begin trying to implement these aspects as part of a hybrid method with Anki, I need to spend some time practicing making associations on the fly at an effective yet non-tedious pace. Perhaps I will simply go through my Anki deck and select words which I have already studied for some time, yet remain habitually difficult and keep getting reviewed. I might as well practice with something worthwhile. I doubt it is possible to make a mess and cause such words to be even more difficult to remember.

While this is true, as I mentioned above, it is not so helpful a lot of the time for a fairly good proportion of words, especially verbs, at least to me. Many German prefixes and roots are more reliable than others. And if one consults a dictionary from the 19th century, one can see that many of the meanings of various words have shifted over time. I seem to recall as well some German verb root forms having fallen into disuse, while the prefixed forms remain active.

Something I have done over time with creating Anki cards, is to record verbs in a spreadsheet arranged by root forms. While I sometimes study parts of it, I have yet to make full use of it. I find that contrasting and comparing the various related forms of a base verb to be helpful in remembering them, when I can see them all at one glance.

Regarding such morphological approaches, there is a book which I don't have, Keller's German Word Family Dictionary: Together with English Equivalents, which seems useful but has mixed reviews. For just verbs, the only book that I have ever seen (I imagine others exist in German), is Kirschbaum's German Verbs: Primitives and their Compounds, published in 1906, and available free from Google Books. The latter book does what I am attempting to do with my spreadsheet, but has some drawbacks. On the one hand it does not give as many prefixed forms as are in active use for many root forms, and on the other hand many of the prefixed forms given have fallen into disuse. Also, its translations should be used with caution by modern learners, as many of them given there are now also obsolete.

There's something to consider... should you do vocab studies both directions? I don't think the method matters here. For me I only look at words in L2 and try to think about what they mean in L1. When I later understand some phrases in L2, and start using them myself, I feel I get the other direction for free.

I'm somewhat averse to translation, being in the camp that would prefer to understand meaning from the L2 side alone. Even being aware of the dangers of assuming a 1 to 1 correspondence of meaning between L1 and L2 lexical items, the danger still exists subconsciously I think. Learning L2 in context avoids that trap.

Nonetheless, double traslation is a tried and true method, especially for activation of language, as with Assimil's active wave. Although I don't use Anki in both directions as many do, I will be translating when I switch to a hybrid method as discussed above, since Iversen's method includes this in the review phase.

Perhaps later when I decide to concentrate on more active skills with German, I might be open to double translation as a means to help achieve that.

But are you in favour of translation when there is a near-perfect correspondence, eg car/voiture/Auto?

I don't know any better tool to activate a full concept like that than he word....

If I know a L2 word and enough of L2 to reliably make that determination, then it probably it already activated for me.

And in writing this reply, what was activated was a memory of a discussion from the HTLAL archives which I searched for and found. Readers might wish to visit that old thread and see the exchange between Linguamor and xtremelingo and Iversen among others, on the topic of one-to-one correspondence between lexical items in different languages. Strategy: Learn 600 words a week (p. 9 and following pages). Linguamor's point to me was that context shades the meaning of individual words, which is why I go for longer, more detailed, definitions on my cards. Note that Iversen is not advocating learning words in isolation and making such correspondences, though he and Linguamor might seem to have been talking at cross-purposes.

Ah, you have skilfully dodged my trap. I could ask "well, what if you were told by someone else?" but I won't bother.

The thing is, the process of learning the finer nuances of a word in any language is a process of approximation and continual refinement. There are two main ways to approach any refinement: subtractive and additive. You start with a "theory" of its meaning and either add new possibilities to that theory or take away incorrect possibilities.

None of us do pure subtractive or pure additive refinement - when we make an additive refinement, there's a good chance we'll over assume, and then need to subtract certain cases later.

The idea that learning in context will prevent overgeneralisation is unproven, and I don't believe it possible.

We will never get the whole thing right from the first exposure, so why be afraid of approximation and refinement.