jeffers wrote:The other thing the OP doesn't mention is "how well". Good enough to recognize when reading? Good enough to use in writing? Good enough to recognize subtle differences in style and register? Or, like I did with 300 in a month, just to know the basic "standard" translation?
My wordlists don't leave space for long rants about any word - sometimes I write two lines because a word has two totally distinct meanings, but my philosophy is that you first learn a core meaning, and then the rest will be easy to add later - expressions that include the word in question will also have to be added later. The point is that wordlists (and resumably ANKI too) should give you a massive passive vocabulary, but they
only serve a purpose if you then read something (for reinforcement of the passive side)
AND use the language, which only will activate a fraction of the passive words you have learnt. And for me easiest way to use a language actively is to try to think in it. English is normally the only language where I regularly notice that I have switched to it without even noticing or taking a decision, though during travels in countries where I know the local language well this may also happen with other ones. But once I reach a level where I can write in it
with a dictionary I also start thinking - with lots of holes and guesses of course because I can't look things up at the spot, but correctness is not important, it's the action itself that counts.
As for percentages
einzelne wrote:Will all due respect to Iversen, I doubt the retention rate is great if you decide to do 50 words per day without subsequent reviewing after a 2-day session of doing wordlists.
I actually did a fairly thorough (though not scientifically tenable) analysis of this some years ago, when I decided to learn Serbian at a point where my only other experience with Slavic languages was some
very shaky Russian. The idea behind the experiment was that I went through a Cyrillic Serbian-English dictionary (for the first couple of letters also a Serbian-Italian one), and I did two repetition rounds and up to letter И also one check round two weeks later, I counted the percentages of words I didn't remember. I have still the spreadsheet data, and out of 1724 headwords I had lost 349 at the first repetition one day later, 217 at the second one two days later (not necessarily the same ones) and 247 at the check round two weeks later. So I think this shows that the words didn't disappear as rapidly as you might think. HOWEVER it should be said that I was my own judge as to how precisely I had to remember a word - and I accepted near synonyms and Danish equivalents for the translation in the dictionary. So don't trust my figures blindly please..
On top of that I did wordcounts before and after the point I had reached in the dictionary with my wordlists. With a wordcount I check
all words on random pages, not only those I have used in my wordlists, and the results were as follows:
before the wordlist phase I checked 812 words and estimated that I knew a
third, 267 words (including some that looked the same in Danish ansd Serbian, but not necessarily in English), and after the exercise I judged
two thirds (499 out of 750) to be known - and then there was a middle category I called 'guessable' where the words rang a bell, but not so loudly and clear that I would claim to understand them. That also went up, of course. And again: these estimates are only valid at the passive level, NOT the active one. So basically I had added a third of the words in the first half of the dictionary to my Serbian vocabulary in a few weeks, - and presumably also two thirds of those in the second half although I didn't do wordcounts on that part. With around 12.000 headwords in that particular dictionary (Сазвежћа) this would amount to around 4000 headwords in less than two months. Try to beat that with other methods.
However Serbian hasn't been a top priority for me later on, and on the two wordcounts I did in 2021 my percentages of definitely known words had shrunk to a mere 54% and 48% respectively (plus 13% resp. 8% *la-la* or guessable ones on top of that). Not as much as in 2014, but enough to read stuff in Serbian (and with some difficulty also its relatives, such as Croat).