It seems that during memorization, one must rehearse the content at earlier positions in order to not forget it, and the frequency with which this happens decays over time. Say you see and encode some information at a locus, perhaps next time you remember it is in 30 seconds, then perhaps 3 minutes, then maybe 20 or so, etc. Has anyone investigated the precise timings that are best for keeping memories stable? This seems like a very important piece of how memory works that has not been systematized. You could think of it as manually doing spaced repetition.
The problem is that time isn’t the only variable. The strength of the image, the circumstance in which you are while memorising, the mindset you have (“this is so interesting” vs “I have to know that for an exam”), the tools/techniques you’ve already trained with, the language you’re using, etc. All these things are going to have a big impact and are different all the time.
I mean I’m sure you could do a systematic study but then the results would only apply when conditions X, Y and Z are met… so not very useful.
It’s for that reason that good spaced repetition systems have very complex algorithms that adapt the review intervals every time you give an answer (depending if you remembered easily, well, not well or not at all).
1 Like
I am well aware of modern SRS algorithms, but in toy domains like memorizing numbers and such, this may be more straightforwardly applicable.
It may also be the case in more natural domains, as adding a systematic approximation may be better than using only natural memory. I am unsure! There are plenty of cases where tacking on mnemonic methods to natural memory gives better results than just doing the natural thing alone, this may also be the case here, even if the method isn’t capable of taking into account all of the details.
I’m curious about extensions that do also try to approximate all of the relevant parameters. It seems especially useful in hierarchically organized domains like language and mathematics.
Probably the thing to do there would be something like setting up a general mental timer first. You could then estimate using whatever you have available how long it will be till you have a 90% (or whatever) chance of recall, then using the timer you remember to remember in the future at the time where you have the 90% shot. Would also probably be a great way to get better calibrated about how your memory works in general.
I do find the question interesting, I just feel you’d spend such a long time planning out and calculating review intervals… (Not to mention getting distracted because your brain will need to be busy thinking about the intervals instead of learning the material). And during all this time you could be learning things or engaging with the material in other ways.
If you enjoy the process of calculating intervals then by all means study the question I’ll be interested in reading your results. I’m just not interested in investing time for this myself. Anki does a good job at telling me when I should review what, I don’t really want to reinvent the wheel so to speak.
1 Like
I think calculation would not be too bad once you got a lot of practice with it, I think the main difficulty is that there is no systematic method with which to practice! It wouldn’t need to be crunched out with arithmetic or anything, it really could just be building in intuition mainly.
It’s fairly easy IME to get a good rough estimate of how long I will remember any given bit of information. I will try a more formal practice here and report back.
I would also argue that any general improvements to the mechanics of memory are quite likely to pay for themselves over the long run. Up-front investment of course, but perhaps more efficient long-term.
Also, at the absolute minimum, civilizational investment in this has been far below what would be ideal, so I will pay the cost and check it out 
@EthanKKuntz here there’s a… proposed “repetition scheme”
We can only recommend an approximate temporary repetition scheme that goes as
follows.
The first remembering should be done after 40-60 minutes following the memorization.
It is during this time that the connections created once are destroyed in electric
memory.
The second remembering should be done after about three hours from the previous.
The third is done after about 6 hours and the fourth takes place the next morning.
This amount of repetitions is obligatory. Generally, the number of repetitions depends
on the complexity and volume of information. The more often you remember the data,
the better it is fixed in your brain.
Anyway, any new data must be remembered intensively within the first three-four days
after the primary memorization.
After such connection fixation it can be stored for approximately 6 weeks in your brain
even if you do not return to the information at all. After 6 weeks the data will gradually
start to erase.
This means that for the lifetime storage of information it should be remembered at least
once in 6 weeks.
If you memorize the necessary information that is regularly used in studying process or
at work, the information use will automatically keep it in memory.
If you memorized potentially necessary, but seldom used information, such data needs
to be regularly remembered in order to ensure its storage in the brain.
1 Like