|
 |
Crazy long gaps b/w recall boost learning | ||
Tags: Memory | Spaced Repetition (SRS) | ||
Share with:
| ||
 Language Learning Forum : Learning Techniques, Methods & Strategies |  | |
| patrickwilken Senior Member Germany radiant-flux.net Joined 4329 days ago 1546 posts - 3200 votes  Studies: German | Message 1 of 14 13 November 2014 at 2:47pm | IP Logged | ||
|
Just read this crazy memory experiment that took nine years to run.
Subjects were given a stack of cards with foreign words to learn (L2 > L1). Subjects go through the stack and as they learn them put them aside until all the cards in the stack have been recalled correctly (basically what Anki does). Then the procedure is repeated multiple times at one of three different intervals (14, 28, of 56 days). What is amazing is that the cards recalled after 56 days are much better retained for years after the study period has ended - and this is particularly true for words that were originally categorised as the most difficult. I am not sure what practical significance this has for Anki, but I find it amazing that words can be learnt like this with such long delays between retest periods. Original paper: MAINTENANCE OF FOREIGN LANGUAGE VOCABULARY AND THE SPACING EFFECT by Harry P. Bahrick, Lorraine E. Bahrick, Audrey S. Bahrick, and Phyllis E. Bahrick (1993). Edited by patrickwilken on 13 November 2014 at 2:51pm 3 persons have voted this message useful | |||
 emk Diglot Moderator United States Joined 5328 days ago 2615 posts - 8806 votes Speaks: English*, FrenchB2 Studies: Spanish, Ancient Egyptian Personal Language Map | Message 2 of 14 13 November 2014 at 3:03pm | IP Logged | ||
This doesn't surprise me: The harder you have to work to claw something out of your memory, the better you'll remember it next time. The downside is that working with such long intervals is actually pretty painful—you spend a long time searching your memory trying to recover the answer, and you spend more time relearning when you forget. This is why I prefer the usual SRS algorithm, where I review something at gradually increasing intervals. This way, each review is generally quick and painless, but I still get the advantage of remembering things for the long term. Khatzumoto has also mentioned this tradeoff, if I recall correctly, and his argument was that, sure, maybe longer intervals provide a bigger memory boost, but the process eventually becomes so unpleasant that almost nobody's actually going to keep doing it voluntarily. 3 persons have voted this message useful | |||
| patrickwilken Senior Member Germany radiant-flux.net Joined 4329 days ago 1546 posts - 3200 votes Studies: German | Message 3 of 14 13 November 2014 at 3:11pm | IP Logged | ||
I need to think more about the paper, but I think they showed this isn't true. There was a bit of a cost to the 56 vs 28 interval, but it was pretty minor, and of course you have far less repetitions overall. 1 person has voted this message useful | |||
| Lemberg1963 Bilingual Diglot Groupie United States zamishka.blogspot.coRegistered users can see my Skype Name Joined 4035 days ago 41 posts - 82 votes Speaks: English*, Ukrainian* Studies: French, German, Spanish, Polish | Message 4 of 14 13 November 2014 at 8:38pm | IP Logged | ||
Anki takes this into account. Bahrick shows that at early stages retention is better with frequent training sessions, while at late stages (after high number of learning sessions) retention is better with less frequent training sessions. This is why Anki intervals quickly go from short (1 day to 3 days to 7 days) to long (2 weeks to 1 month to 3 months to 9 months). From a theoretical perspective what we're seeing is what the Bjorks call the theory of disuse. The theory says that memories are never forgotten, instead they have values for Retrieval Strength (RS) and Storage Strength (SS). RS is how likely we are to know the fact right now, and SS is how long we are going to know that fact before we "forget"it and need to have it retaught. RS increases to 100% the second you review a fact (Addis Ababa is the capital of Ethiopia) and then declines over time. SS is the speed at which RS declines. The way SS increases (ie slows down RS decline) is through forgetting. The brain is constantly throwing away information because neurons are energetically expensive and only when we forget and have to retrieve something does the brain say "oh crap, this neuron might be important, i should increase the amount of energy i devote to keeping it accessible". In addition, the increase in SS seems to be greatest when the RS is lowest (ie 0% chance of recall aka "forgotten"). In a vacuum, the greatest increase in storage strength would happen when the interval between review sessions is billions of years apart. Of course, nobody lives long enough to do this.From this perspective, Anki isn't a tool that helps you remember facts, but actually a tool that helps you optimally forget things. 7 persons have voted this message useful | |||
| luke Diglot Senior Member United States Joined 7001 days ago 3133 posts - 4351 votes Speaks: English*, Spanish Studies: Esperanto, French | Message 5 of 14 14 November 2014 at 12:24am | IP Logged | ||
That's exactly the information I have been looking for since I began using Anki a couple of months ago. 1 person has voted this message useful | |||
| patrickwilken Senior Member Germany radiant-flux.net Joined 4329 days ago 1546 posts - 3200 votes Studies: German | Message 6 of 14 14 November 2014 at 9:06am | IP Logged | ||
|
If I change the "graduating interval" from the default of one day to a longer interval (say seven days), would that have the effect of: (1) making the cards I remember after seven days better encoded; and (2) increasing the lapse rate? In other words, making the easier cards, even more strongly encoded (quickly learnt), and also increasing the chances the weaker cards being forgotten?
I am currently generating semi-automatically lots and lots of cards from reading newspapers. The default set-up in Anki is to maximally learn most of what you put in, but I am seeing many more unknown words per day than I can deal with. Perhaps a solution is to increase the default "graduating interval" as well as lower the suspend rate, and so filter out the hard cards, and encode the easier cards quicker. I can always go back and attempt to learn the suspended cards later. This might have the advantage that those words that are somehow better fitted to the pre-existing semantic German web that I have already developed in my head (the easier words) are automatically given priority in learning. The difference between an 'easy' word and a 'hard' word is perhaps that the 'easy' word slips in easily to the pre-existing semantic network, whereas the hard word requires a major modification of the network to graft it on. By learning the easier words first you grow the network till it reaches a point where it can more easily incorporate what were 'harder' words. Edited by patrickwilken on 14 November 2014 at 10:25am 1 person has voted this message useful | |||
 Iversen Super Polyglot Moderator Denmark berejst.dk Joined 6499 days ago 9078 posts - 16473 votes Speaks: Danish*, French, English, German, Italian, Spanish, Portuguese, Dutch, Swedish, Esperanto, Romanian, Catalan Studies: Afrikaans, Greek, Norwegian, Russian, Serbian, Icelandic, Latin, Irish, Lowland Scots, Indonesian, Polish, Croatian Personal Language Map | Message 7 of 14 14 November 2014 at 10:01am | IP Logged | ||
|
I have experimented with wordlists rather than with Anki, but the basic principles are not different.
My early experiments with word memorization told me that immediate repetition after a minimal break worked better than simple repetition without breaks. The 'group' feature of my wordlist system provides these short breaks. Next step: I found out that I needed to do one repetition round somewhat later, but still at a time where I remembered which 'memory hooks' (associations, intralinguistic relationships, external factors) I had used while I memorized the words. The repetition was actually more a question of reminding myself of the hooks than of the original words. And finally I tested whether a second and third repetition round helped me, but I found out that they added very little to my retention rate. This year I have done a new series of experiments with Serbian which confirmed these observations BUT they added something I had overlooked, namely that the words I miss at a late second repetition round have a quite low overlap with the words I miss during the first repetition round. And I think this could be an argument for doing a simple repetition round for instance a week or two later. And inspired by this last observation I have come up with the hypothesis that the blocking that prevents us from remembering words isn't simply a function of the individual word. If I don't work actively with a language for some time my ability to recall words - ANY words - deteriorates. And inversely: I have concrete evidence in my experiments for a beneficial effect of working with word memorization, namely that I also remember other words which didn't make it to the wordlists better than if I didn't do wordlists. Concretely this shows up as a boost in my wordcounts after a 'hot' wordlist section, even for words which I didn't really try to memorize through the lists, but just saw in the dictionary. Edited by Iversen on 25 November 2014 at 12:02am 4 persons have voted this message useful | |||
| Lemberg1963 Bilingual Diglot Groupie United States zamishka.blogspot.coRegistered users can see my Skype Name Joined 4035 days ago 41 posts - 82 votes Speaks: English*, Ukrainian* Studies: French, German, Spanish, Polish | Message 8 of 14 14 November 2014 at 7:15pm | IP Logged | ||
I haven't tried messing around with the time settings in Anki. Simplifying the phrase or changing the image/audio on the card fixes troublesome cards for me, which suggests the issue is with the encoding rather than retrieval. 1 person has voted this message useful |
|  | 
Printable version
|
| Copyright 2024 FX Micheloud - All rights reserved No part of this website may be copied by any means without my written authorization. |
