Do you have an exit plan for isolated vocabulary study?

I gave up on isolated vocabulary study with Anki a couple of years ago. It was too much work, it was painfully boring, and the near-synonyms were driving me nuts. Fortunately, I have pretty good tools for capturing vocabulary directly from native electronic materials and adding definitions as needed. So my Anki decks are really just an ultra-condensed version of my extensive reading and watching. And since I throw cards out whenever they bore or annoy me, my deck is usually pretty pleasant.

One of the problems with pure extensive reading is that takes forever to learn the weird, rare and obscure words. If you take a look at this chart from an American-Brazilian research project, you'll notice that it takes a long time for natives to get from 20,000 headwords to 30,000 dictionary headwords (using their particular methods of counting vocabulary size):


In this particular study, the median native speaker crossed 20K words at age 16, and 3oK words around age 40, despite living in full-time immersion. Now, the absolute numbers will vary from study to study, because there are a lot of different ways to count vocabulary. But the relative numbers are suggestive: native speakers need over two decades to finish filling in their adult vocabularies.

So once or twice a year, when I start getting annoyed with my vocabulary holes, I'll pick up another 500 words or so using Anki sentence cards. This allows me to deal with those funky words like tergiverser and ausculter that just don't want to "stick" from reading alone.

I find that this isn't a huge problem. About 80% of the time, knowing a word in a single context will allow me to recognize it in other contexts, and then I can rely on extensive reading to fill in the details normally. The other 20% of the time, I need more help to learn the word. Fortunately, I can just make another card the next time I run into an interesting usage of the word, and that will usually suffice. I think I've occasionally gone as high as three cards created over the course of a couple months for some especially stubborn words, but it's very rare.

I notice that you don't say it was ineffective in teaching a lot of vocabulary relatively fast.

Re synonyms and words with slightly different but close spellings, I do a couple things. First I transfer one of those words to a different sub-deck. Second, regardless of how hard/easy a word actually is, I try to make sure 2 confusing words are not on the identical review schedule by choosing a different option for one of them. Finally, assuming I can remember the other word (often I am vaguely aware a similarly spelled word is causing problems but can't quite remember it), then I stop and review both words outside Anki at the same time. Kato Lomb recommended this in her book, and called it something like making them report for duty.

Linguamor and Dr. Arguelles used to say that after 4-5 lookups a word encountered during ER would stick. But I remember reading a couple studies that said it took far more exposures, which of course will take a long time to occur for very low frequency words.

This study is in accordance with others, that state that for English, a typical college educated native speaker finishes college with around 34K words, and then adds 1K+ a year through normal reading and occasional lookups, peaking in the late 50s or early 60s in the 50-60K range. ER does work, just very slowly after acquiring the high frequency and perhaps mid-frequency words. But as I mentioned earlier, the advocates for same seemed to have been able to spend hours a day in ER for longish periods. Obviously it has other advantages, and also is more likely to be more efficient once knowing a related language to the one being read.

I remember a post of yours a couple years ago saying that you wanted to take your French vocab from around 5K to 15K over the following several years. So you obviously know how many words you need to hit the lexical threshold. And as well how much time you have available for studying and what your personal tolerance for boredom and tedium is. There are only a handful of other HTLAL regulars who are as diligent as your are, and as generous in sharing the results of their experiences. I wish that I could wave a magic wand and boost your vocabulary by 10K.

I guess that I underestimate greatly the ability of others to suffer through such boredom and tedium (which Anki is), and confess that my own willingness to do so is probably the result of an obsessive nature, which Anki fits perfectly. I suppose that I would have made a model teacher a hundred years ago in inflicting the grammar translate method on students in Latin and other languages, never giving a thought to "fun" and "exciting" or learning styles.

Which is pretty much my point: even given a very specific example or set of examples, we'll naturally generalise beyond it. The mythological part of the fable is that restricted context prevents overgeneralisation.

For me, the most efficient method is to capture text during extensive reading and make half-word cloze cards, where each card hides half the unknown word. I can lean 500 words in 15 to 25 hours of card marking and Anki reviews and get pretty decent active recall. But this probably requires specialized tools to do efficiently. And again, I don't do this for every word. It's more of a periodic "mop up" operation for words that are proving tricky to learn via extensive reading.

I think this actually compares fairly well to "bare vocabulary" cards, especially since L1<->L2 cards don't work as well for me as they do for many people.

If I make Anki sentence cards from books that I'm reading, I need 1 or 2 occurrences of a word to internalize a rough, passive meaning. From there, I can fill in the details via extensive reading. Basically all Anki does is artificially boost the frequency of unknown vocabulary by repeating the relevant passages on a regular basis. Again, this probably works most efficiently for a closely related language like French—it's nice having lots of easy cognates. And again, tools don't hurt—I'm far too lazy to ever type sentences.

Aw, thanks. I'm pretty much past any kind of lexical threshold for French, thanks to the Super Challenge. By the end, I was seeing an average of 1 genuinely unknown word every 2 or 3 pages, and various online estimators suggest that I'm above 15K words now. Since I have about 1,500 French Anki cards, that means that I learned ~90% of my vocabulary via extensive reading and other exposure. Mostly when I encounter unknown words these days, they come in clumps. I'll be reading some high-brow article somewhere and suddenly I'll start feeling really dumb.

I can use boring techniques, but I also know how to make fairly efficient use of my "fun" time, and that allows me to save more of my willpower for other things. Of course, many of the things that I find fun would seem strange to other people, so it all balances out.

I don't do cards in both directions either. Card creation is quite the timesuck, but your estimate seems rather high. For my own cards I take a word, copy and paste into wordreference.com and dict.cc, copy and paste into Anki, with some adjustments, so my card creation rate per hour is probably not impressive either. But reviews only take seconds which add up to minutes long term. Neither of us uses bare vocab cards, but they have to be a lot faster long term. The time-leveraging factor for me in card creation is that I mostly have inputted a new list into Anki, and then work on the definitions while listening to German radio or watching German television in another window.

I've just started an experiment with a hybrid Iversen-Anki method. His name deserves pride of place since if successful, the Iversen method will teach 85% of the words, while Anki will mop up the other 15% and serve as a control for the rest. I'm not saving on card creation time, but hope to long term in respect of needing far fewer Anki reviews.

Obviously a word is internalized better when Anki is used in support of course texts or just regular texts. That is why pure ER doesn't seem too efficient, even with the quick use of popup lookups. Although I normally just copy/paste sentences from dictionaries if they look "real" instead of merely grammatically correct substitutions, I have typed sentences from one of those 750 verb books when I was still studying Spanish. It was very painful, and the act of typing them in did not seem to help recall. Wonderful product at the end, but inefficient for the effort.

It's very nice to hear that you reached that goal in a couple years instead of just years and years. If more learners on HTLAL would just work in spurts of 500 words like you do, they would be pleasantly surprised by the compounding of results. There is no reason someone cannot reach the lexical threshold in a Romance or Germanic language in two years, at the price of an hour a day for card creation and an hour a day for reviews, plus some grammar study and extensive reading and listening. As I have said before, the same 40-60 minutes of review time that once maintained a couple thousand words, then a few thousand words, now maintains 11K+, supported of course by reading and listening.

I can read and understand a lot now, and with nearly complete understanding, primarily news and non-fiction stuff. Literature, drama and everyday spoken stuff not as well. Normally I qualify "learned" with "or partially learned", but forget to sometimes. Obviously the most recent learned or reviewed cards are not as well known as most older ones that have "matured" according to Anki parlance.

No I don't feel the need to make that statement every time you make that assertion, as once is enough for it . But "25%" is a fairly exact figure as opposed to "less", "more", etc. Is exact and unprovable better than vague ?

I used the term "slow and fun crowd" because my, admittedly subjective, impression of many learners on HTLAL who advocate "fun", also are taking longer periods of time to progress in the language based on their self-assessments.

Re slow on production, I don't buy that output is as important as input in learning a language. I have read a lot on both sides of the issue, and know that many accomplished polyglots do believe it important to give at least some focus to production from the beginning. Is it contested that free output is more difficult than input? Sure I could have worked on stringing together a bunch of declarative sentences in various topical areas to a shallow degree this past year, but to achieve what? Only now is spontaneous thought in German starting to happen, and while I don't write much and have to proof it, I am able to write more naturally with a much larger vocabulary. Iversen always emphasizes thinking before speaking and writing, and that is what I am going to work on first. Also, remember that although some Assimil users like Luca like to begin the active wave near the beginning, the creator of Assimil chose to put it halfway through for a reason, namely that it would be easier.

No matter how many times I might be able to visit a German speaking country in the future, the simple fact is that the vast majority of my language usage of it will be reading and listening. There simply are not many German speakers in America unless one counts the Amish and Mennonites, both of which populations mostly know English and also speak a different dialect.

As to "balance", I look at that as with other concepts in the long term, and not most applicable to right now. Obviously learners with differing goals, which include more desire to speak and write, should give more priority to output at some point. But ALL their goals will be easier met with more vocabulary.

Re vocabulary counts, despite the fact it plays into my more obsessive side, I don't view it as artificial any more than than the lexical threshold is. The lexical threshold sets my vocabulary goal, and my vocabulary count shows my current progress on that quest. I am trying to maximize it as quickly as possible, so that I can get on with actually using the language better for comprehension, and also be able to then concentrate on finer details of usage and some production.

There is a small group of words which are very common, very versatile and quite often irregular. These words form so say the connective tissue of a language and you need them to formulate any thought in the language - even if you just think. But the goodies are in the rest of the vocabulary, and most of these words are NOT very common. And if you study two samples of text the lexical overlap at this frequency level is surprisingly small. So basically you have willy-nilly to learn a lot of words to deal with genuine texts or genuine speech, unless you are satisfied with simplified textbook excerpts or smalltalk specifically tailored to discussions with foreigners who against all odds insist on speaking the local language. So getting a large passive vocabulary is a necessity if you want to deal with those texts or recordings which are worth reading/listening to - and much of this vocabulary consists of rare or specialized words. The treshold can be lowered by using bilingual texts or popup dictionaries on the internet, but the acquisition of vocabulary may be the most time consuming task for a beginner or mediocre language learner, and it can't be avoided so willy-nilly every language learner has to figure out how to do it in an efficient way. For some of us ANKI works, for others - like me -wordlists work, and for still others any formalized study technique is taboo so they have to learn from reading or listening. But I have to admit that I have problems seeing how that functions at the stage where you can't yet open a newspaper or an essay collection or listen to the news on the telly without getting stuck on unknown vocabulary.

At what level can you use your language actively? Well, you can survive simple conversations with the 200-300 most common words plus a sprinkling of rarer words which for some reason are relevant to you (like the name of your country and your profession), and with time you can let more content words seep into this fragile web of function words. But it is quite understandable if native speakers don't find it interesting to have discussions at this level, and then you are down to asking for bread or the way to the toilets. And you may choke on even simple answer - like if you ask for directions to the loo and you don't know the words for "behind" and "tree" yet. Besides home learners rarely have native speakers around them so in that situation thinking in a language is a practical solution, and nobody will object to banal content like "that is a green tree" or content with holes like "that is a [image of 'green'] tree". We all have to start somewhere.

The idea that learning single words is wrong because language by nature is idiomatically loaded is in my opinion wrong. I do work with contexts when I study texts, but my chances of remembering and reproducing idiomatic expression is far better if I know the constituent words - just as it is easier to remember long compound words if you know the coonstituent parts, like prefixes or endings. So I would definitely not say you shouldn't learn expressions at an early stage, but personally I wouldn't dream of memorizing a long expression unless I also knew all the words in it - if necessary by looking them up one by one. Expressions may be unpredictable, but they are rarely totally devoid of logic - though that logic may not be visible if you use your own language as a filter.

As for the discussion between Big_dog and Peregrinus about the amount of words you can permit yourself to learn 'without a context' (I hope I have understood that right!), I think it is too complicated to pin down as a percentage. For me there is a lot of context in a good dictionary, and much of that context is of the kind I need to know in order to use a given word later - like its morphological class or the most important idiomatic expressions with the word. Granted, you may learn something about both from occurences in real texts, but often the morphological class can't be deduced when the form used, or you don't realize that you actually have an idiomatical usage of the word in front of you. So even the most extreme way of using wordlists - i.e. basing them directly on directionaries - isn't devoid of context, and the kind of context you see in a random passage may not be enough for you to really get to know a word. But there is one thing you can do with a real text or speech sample: you can study a theme through texts or documentaries, and then you may learn words which you didn't even know in your own language with the explanation that makes that information stick. But beware: reading or listening at this level is only possible if you already know a helluvalot of words.

If I add up all the time I spend reviewing sentence cards, it works out to about 3 minutes per card (until the intervals space out to 3+ years, at which point it's moot). Each card contains 1 to 3 items of vocabulary (even at higher levels, it's weird how unknown words tend to cluster). This seems reasonable to me—the last time Iversen and I tried to compare how much total time we spent on a single word, neither of us had an obvious advantage.

I do spent a relatively large amount of time on card creation, but almost none of that time is wasted on flipping between windows, typing or other inessential details. Rather, 90%+ of my card creation time is spent trying to figure out the specialized vocabulary in the sentences that I highlighted, which is not always easy. For example, a card from Le Déchronologue:

Je pressentais, hélas, que la colère de François Le Vasseur n’était pas à la veille de s’éteindre, et ces algarades visant mes matelots me faisaient craindre de plus sérieuses anicroches si nous restions ici plus longtemps.

(not on the card, just for readers here) I foresaw, alas, that the anger of François Le Vasseur was not on the eve of its extinction, and these provocations targeting my seamen were making me fear more serious complications if we remained longer.​

Here, we have a useful phrase that I already knew (n’être pas à la veille de), plus the rather obscure algarade "row, dispute, insult" and the uncommon-but-useful anichroche "hitch, snag". There's also matelot "seaman, sailor (of ordinary rank)", but I had learned that a couple of books ago. Personally, it takes me a minute or two to look up all these words, to figure out which definitions apply, and to make sense of the sentence. But in exchange for that effort, I get to reinforce n’être pas à la veille de and matelot, and I pick up algarade and anicroche. Plus I can reinforce a bunch of other useful-but-obvious vocabulary like pressentir.

Personally, I can't imagine learning vocabulary like algarade from word lists or from L1<->L2 cards. I'm sure that some people can learn obscure vocabulary like that off a list, but as previously mentioned, I find that single-word L2->L1 cards do bad things to my listening comprehension, because they encourage me to translate reflexively. (I strongly suspect that only a minority of learners are actually affected by this difficulty, but it matters for me.) This is why my main vocabulary strategy is to do lots of extensive reading, and to "distill" interesting passages for more frequent review. For me, Anki is nothing but a temporary amplifier for lower-frequency and tricky vocabulary.

At least among the French students, I have the general impression that the HTLAL "fun" crowd actually makes pretty rapid progress, at least for receptive skills, in the area between A2 and C1. Below A2, it's hard to use native materials (except for parallel texts, and even those must often be graded), and above C1, well, there's just so much obscure stuff to learn that I can't imagine anyone mastering it without sustained exposure. But I've actually Skyped with another member of the "fun" crowd, and his French was a very solid B2, both passive and active, after about two years. Of course, doing this required a ridiculous amount of reading, watching and Skyping, but for the right personality, that can be a pleasant way to reach a serviceable level.

One tool that I find especially useful at lower levels is parallel L1/L2 (or L2/L3) texts. This reduces the problem to one of "aligning" two sentences in my head. For Spanish, this is almost trivial. For Egyptian, I needed ~50 Assimil lessons worth of graded L2/L3 texts, but now I can usually mentally align real parallel texts. And honestly, Egyptian is nearly a worst-case scenario: it has only a tiny handful of cognates, non-IE grammar and an unfamiliar writing system. For more information, see my article on how I use Anki for Egyptian.

I absolutely believe your methods work very well, of course. How could I doubt it? But it's interesting to see how many different ways there are to tackle these problems.

We may now be veering out of the territorial waters of vocabulary acquistion methods, but my way to survive aligning in complicated cases has always been to make hyperliteral translations. Some call them derogatively word-for-word translations and warn against them, but when I had to get a hold on things like ownership in Russian (where you don't 'have' things, but just somehow find them in your vicinity) or subordinate clauses instead of infinitival constructions in Romanian and Greek, it helped me enormously to equate such things with translations like "by me is" and "I will that I-do" instead of more idiomatic formulas like "I have" or "I wanna do" (or their Danish equivalents). In a wordlist I might write "у меня" with the translation "[to have]" (in square parentheses to make that there is something fishy going on here). I wouldn't write whole sentences or even longer excerpts in a wordlist, but even a short formula like "[to have]" would remind me of the kind of context where I found "у меня". Similarly I wouldn't quote a a long and complicated passage from a poem if I wanted to indicate the construction possibilities of a verb or the gender of a substantive. I would use a short, more standardized notation ("to do st. to sby.").. BUT I would still repeat the original passage to myself while jotting the word in question down on my copy sheet, and with a bit of luck I might even remember it when I did the real wordlist later. Which is one good reason for compiling a wordlist shortly after working with the original text.

Your second statement does not jive with your first to my mind, as it implies taking an awful lot of time which that learner happens to be able to put in daily. Obviously there is a difference in the same time period as say a year, with two learners who reach a similar level, but one doing it only by dint of putting in far more hours with his/her method(s). This implies to me that "fun" is not as efficient time-wise. However it may be what motivates such learners to persevere instead of quitting. But I am comparing passive learning time to passive learning time, and his time includes active methods.

Howso? You were the one criticizing lack of output. Did I misunderstand that to mean you were implying it should be equal when you really only meant 25%?

However, allow me to assert the reverse. It is my belief that input alone suffices for quite a long time in learning a language, and that afterwards, output can come relatively quickly. A Krashen kind of position, but one I am not dogmatic on. I don't really know how fast I can activate my German yet, but I haven't seen many dispute that one can activate relatively quickly after a longer period of only input.

You neglected to quote the part of my reply where I said I was comparing passive apples to passive apples but that their study, and thus study time, also included active oranges, i.e. implicitly admitting it would be difficult to judge.

They are using whatever methods to learn language. I am using Anki plus other (passive) activities, to learn language. Both of us are learning language. They are using presumably fun and exciting methods to learn vocabulary, just relatively slowly, while I am using Anki to learn vocabulary, just relatively more quickly.

Speaking of strawmen, you seem to keep implying that Anki is all that I do, when it is not, by comparing my use of Anki (thus isolating it) to others' use of a variety of activities.

Personally I find that passive acquisition of basic to intermediate vocabulary mostly takes care of itself through exposure and usage. The thing I've found Anki most useful for is "activating" words, expressions, and grammar points (which preposition to use in a certain place, etc.): helping me to use them as opposed to just understand them. For this I tend to use the sort of sentence cards with half-word cloze deletions that emk described, with a "clue" in English, the target language, or another related language I know - whichever is quickest and most direct.

It's certainly not the core of my language learning, it's just an accessory that I spend 5-10 minutes per day on and that I find keeps things moving along nicely. A while ago I hit a big plateau in French because I was trying to just learn extensively with input and conversations, and adding Anki back into the mix was what got me progressing again. (Of course, maybe simply more input and/or more conversations would have also had the same effect, but spending a few minutes a day making and reviewing cards is far quicker than spending another hour a day watching TV or reading and far easier than trying to find French people to talk to!)

I also find it useful for learning the obscure/advanced vocabulary that doesn't come up often enough in "normal" contexts to stick, again like emk said, and I think that especially if you're not in an immersion situation then there's nothing wrong with a bit of "artificial" help... but I think that many learners would benefit more from getting more solid at using the basic/intermediate stuff well rather than trying to learn every last obscure word. I realise it doesn't have to be one extreme or the other and it depends on your goals, but I'm happy to focus on using day-to-day language more effectively at the expense of some fancy vocabulary.

After high and mid-frequency vocabulary, every word is relatively infrequent, and thus by itself not terribly important. But as a mass they are important to reach the lexical threshold and beyond. Other than one's specialized interests and truly archaic and technical words, it probably does not matter just which words comprise such a mass of lower frequency vocabulary for a learner. It is the process of exclusion they enable, i.e. as a mass they reduce the quantity of unknown words in a text. So while it may not be important to learn any one word, it is important to learn a great number of such low frequency words, at least up to the spoken and written lexical thresholds, regardless of the means used to acquire them, in order to use non-simplified native materials.

Well, for some people, it's hard to find two good hours for studying every day, when they have their wits about them and they're ready for sustained intellectual activity. On the other hand, many of the various "fun" activities can still be pursued even after a long and demanding day. I mean, how many brain cells do you need to watch a few episodes of Buffy or L'Avatar? And yet, a single episode of television can contain a huge amount of spoken language.

Honestly, I've hit the point of diminishing returns for passive vocabulary. I still run into plenty of words that I don't know, but they don't make much practical difference. The biggest payoff for me right now would come from either (1) improving my listening comprehension one more notch, or (2) lots of professional conversations in French. So I've been taking it easy on Anki since the start of the new year, and virtually all of my new cards are for Egyptian. I'll probably add some new French cards at some point, but there's no rush.

Well, I'm not quite sure where to draw the lexical threshold. I started reading native books around A2, and I was heavily watching TV series when I was a strong B1. And I learned a huge amount of vocabulary that way, and I still do. But there's also a residue of vocabulary that's hard to learn for one reason or another—perhaps it tends to occur in abstract contexts, or native speakers use it to provide color without providing any hints about the meaning. And for those words, SRS software can be very helpful.

Even though I'm not making any new cards at the moment, I'll still marking interesting vocabulary when I read. Here's what I've marked in the last 24 hours, copied from the Kindle highlights page:

Les dernières cendres (29 pages, not bad)

Persuadé en ses tréfonds qu'elle vaincrait la maladie et qu'il n'aurait pas de promesse à honorer.
This feels like one of those words that mostly appears in a handful of collocations.

-Ah ouais, j'en ai déjà vu. C'est un truc pour les musiciens. Chais plus le nom. -Un diapason, c'est un diapason, avait-il acquiescé. Mais celui-là est spécial.
I didn't know the French word for "tuning fork."

Il tendit l'index de la main gauche en l'air et frappa l'instrument sur la jointure des premières phalanges. Les lames vibrèrent mais Emilie n'entendait rien. L'homme la sonda du regard et elle trépigna, poussant un gémissement d'impatience. Alors il leva le diapason et l'approcha de son oreille.
"To stamp ones feet". How did I miss this one up to now? And this is almost a perfect sentence card: it provides strong internal context.

Nous voyons les choses séparées par du temps et de l'espace, mais c'est une astuce que la conscience se fabrique pour pouvoir se repérer.
In theory, I already knew this word, but it took me a moment to puzzle it out in this context.

Le Tabouret (13 pages, not especially worth reading)

Quand bien même, c'est ma vie. Donc, jusqu’à preuve du contraire, je reste le mieux informé pour juger de la nullité d’icelle.
A rather rare and interesting demonstrative pronoun.

Cette jolie métaphore paraît bien simpliste, mais elle permet de voir en un coup d’œil introspectif à quel point votre vie pourrait être bancale.
I haven't noticed this one before.

Sur le moment, j’avoue, j’eus des doutes sur le caractère non-professionnel d’un tel retournement de situation. Cependant, mon côté chevaleresque m’interdit de détourner les oreilles d’une demoiselle en détresse.
Completely transparent, but it seems like the sort of word it would be nice to have in my active vocabulary.

En dehors de quelques émissions de qualité médiocre diffusées à la télévision. Si tant est que cette activité débilitante puisse être considérée comme vecteur d’épanouissement personnel.
OK, that's a weird construction and I've never noticed it before.

Toujours est-il que si j’étais conscient du vide, ce n’était pas si dramatique à mes yeux avant l’arrivée du tabouret. Comme une petite démangeaison qu’on efface du bout des ongles pour l’oublier dans la foulée.
"Itch, urge". I know I've seen this before, but there's a lot of similar words that have interfered with it.

Alors elle a ce regard. Ce regard espiègle et malicieux. Celui qui suffit à exprimer mille mots inutiles. Le tabouret se volatilise sous mes fesses et je me retrouve allongé sur un sol de béton, ma déesse à califourchon sur mon bas-ventre.
*sigh* The narrator is dreaming, the narrator is an idiot, and this story is a waste of my time. But these two words are worth grabbing, especially since the context is very strong here.

— J'ai l'impression d'être invisible à vos yeux. Rien de moins qu'un... qu'un larbin !
"Servant, servile person." This one is definitely new.

Mais là, ce que tu me racontes, et ce que cela sous-entend… non !
This one I knew, but this particular usage was a bit surprising.​

So 42 pages gets me 12 words and one new grammatical construction. Of course, there were a few more unknown words, but they were either (1) completely obvious from the context, or (2) they were even less useful than "tuning fork."

So this is what one particular end-game for Anki might look like: after 10,000+ pages of extensive reading and 1,500–2,000 Anki cards, I can read 40 pages of randomly selected native materials and find a few good new examples of how to use medium-frequency vocabulary, and some fun low-frequency vocabulary. At some point in the future, I'll get bored and say, "It's time to learn another 500 rare words", and make cards from these highlights. At this point in my studies, I'm up against the law of diminishing returns, and the remaining unknown vocabulary is sufficiently infrequent that it's more efficient to capture it and use Anki to artificially boost the frequency. But an equally valid strategy would be to stop worrying about vocabulary altogether, and to focus on other weaknesses or even another language.

See links in this post re lexical threshold, although I imagine that you have seen them before. For English:

listening lexical threshold: 6000-7000 word families (10,200-11,900 words)
reading lexical threshold: 8000-9000 word families (13,600-15,300 words) - bottom range for newspapers & top range for novels.

Obviously it depends on what kind of vocab one learns. I am "only" at about 6,400 word families now but read not overly technical newspapers and non-fiction with ease. But I ascribe that to the fact that many of the lists I used to stick into Anki came from CEFR courses. What I definitely need to mop up is mid-range vocabulary, called highly available vocabulary in this interesting thread on HTLAL: Basic Vocabulary (De Mauro).

There is limited research for other languages, but it suggests to me that the reading threshold is higher for German, hence why my goal is 20K there, and less for Spanish and French. Not as conclusive as for English, but suggesting that the English figures can be tweaked +/- 20% to approximate for other languages.

I discussed my own rough mapping of vocabulary to CEFR levels in this thread: Vocabulary - Synergy Part 7.

[Source for 1.7 multiple used to translate word family numbers into words: Kusseling & Decoo (2010): Europe and language learning: The challenges of comparable assessment]

Because I don't see anyone claiming to learn 10K+ words per year (or even half that) using such other vocabulary learning methods, without also stating that they spend several hours a day with same, instead of 1-2 hours that I spend with Anki. If I have missed such claims, please direct me to them.

I call and raise you 3 more strawmen .

On average, I study 2-3 hours per day, but more on others. Since I have already created so many new cards for the queue, I don't need to spend time on that every day, and as I have said, often do same while listening/watching German radio/television. So Anki is more than 25% on the days I study less for sure, since it takes 40-60 minutes for reviews. You already know that I don't accept your 25% balance figure, so why measure my learning activities against it when you know it won't mean anything to me? Also as I have also said, the vast majority of my cards contain sentences, which is learning in context, even if I don't look at them for every card. I get a lot out of my Anki deck.

My current level of listening and reading comprehension is primarily due not to output, because I don't do it, nor to grammar study, because I don't do it as much as vocabulary study. Nor can I even ascribe same to extensive reading and listening which I do in fact do, because without the vocabulary I have learned, those would be much more intensive (and because those extensive activities function mostly to reinforce Anki-learned vocab). The amount of time I have spent on learning vocabulary via Anki is the primary reason for that level of comprehension. I know that it works, and that I don't see others claiming ER or other activities are more time efficient for learning vast amounts of vocabulary.