Join 3,512 readers in helping fund MetaFilter (Hide)


Optimum intervals for spaced repetitions. I remember, I remember, I forget.
May 24, 2012 2:53 AM   Subscribe

The science of learning: my Google Scholar-fu is failing me. Do we know the optimum intervals for super-effective spaced repetition learning?

I use Anki and Memrise to learn a range of topics (foreign language vocabulary, geography, random facts, etc.). Together they have vastly improved my knowledge of these topics, but I can still see space for improvement: I still find myself forgetting seemingly simple words or facts even after seeing them multiple times (i.e. they have become 'mature' words).

This has led me on a quest to try and discover the optimal intervals for effective spaced repetition learning. My time researching this topic on Google Scholar has so far proved fruitless.

I remembered reading in John Medina's Brain Rules that the optimum intervals were not known. This was back in 2008: is this still the case now?
posted by fakelvis to Education (9 answers total) 8 users marked this as a favorite
 
From my understanding,
the optimum intervals depend on the person, and the information being learned.

E.g. You could tell me something once, and I might remember it for years, but possibly not my entire life.


More accurately, the optimum interval for repetition is - 'right before you would have forgotten the information'.

That's why a number of the electronic versions, customise the space interval of repetition of facts, to the rate that you tend to forget items. And if you forget something you'd reviewed in the past, it goes back on the repetition list.
posted by Elysum at 3:43 AM on May 24, 2012


My experience is that you will forget things that are merely "trivia" fairly quickly. Especially if you are learning these facts for no other reason than to just know them. But if you use the knowledge in some way, it will get integrated into your memory much easier.

The brain is about patterns and connections. You are implanting the patterns, but they aren't connected to anything except the context of memorizing these facts.
posted by gjc at 5:51 AM on May 24, 2012


Some resources on this topic are on the bottom of this page (from the SuperMemo site).
posted by mark7570 at 6:04 AM on May 24, 2012


Elysum: "More accurately, the optimum interval for repetition is - 'right before you would have forgotten the information'."
So this is likely why my Memrise intervals are so specific (specific, not necessarily accurate), while Anki's intervals are less accurate: Memrise is trying to guess exactly when I'll forget a word, while Anki only gives me a few options on when to renew a word and must rely on my judgment.

Maybe an Anki plugin that allows me to be more specific on my estimated 'time of forgetting' could help (more options, etc.).

gjc: Exactly. This is why the Dutch words I learn and then hear or see in my daily life are remembered much easier. I guess I need to force myself even more to use what I learn between repetitions.

mark7570: Thanks. The SuperMemo site is a treasure trove of information; I'll find something useful there, no doubt. I completely forgot/ignored the site after discovering that it's Windows-only. :(
posted by fakelvis at 6:46 AM on May 24, 2012


A coworker of mine is working on writing up a paper on repetition effects in free recall. I sent her a link to this question, and she said, roughly:
The suggestion of the possibility that there's a definitive answer to something so vague is interesting to me. I'm guessing that the asker would be a little suspicious if someone simply answered something like: "Always, exactly ten minutes."
Basically, it depends on so much more than the time between repetitions, like the item's preexisting strength in memory (i.e., prior knowledge), salience, form of learning... the list goes on and on. Forgetting functions have been extensively studied in the literature, literally since the beginning of the study of memory with Ebbinghaus. My PI has an unpublished, math- and theory-heavy paper on the "power law" of forgetting (PDF; self-link, sort of).
The decline in memory performance with time or intervening events is well fit by a power function. This simple functional relationship accounts for a great deal of accumulated data. In this note, we consider a simple yet general memory model in which all items decay monotonically in strength, but at different rates.
It's the "different rates" part of that which implies no single optimal repetition interval.
posted by supercres at 8:53 AM on May 24, 2012


I'm an ex-teacher and there are several different theories on how much and how often a learner needs something repeated in order to learn it. What I think matters in your case is to figure out your learning style (try drilling yourself on something a set number of times across a set number of days and then test yourself to see how well you've learned it.) Also, learning is often seen as a continuum in education. At the lowest level, the student can repeat back exactly what you've told them. At the highest level, they can evaluate that knowledge. (This is known as Bloom's Taxonomy.)
posted by emilynoa at 9:29 AM on May 24, 2012


Ditto supercres- exact numbers are not known or generalizable (I am a memory researcher although not focused on this precise topic). But if I were working this out for my own purposes, I would try out this study schedule as a first pass: time 0 (t0), take a nap, later that day, the next night (t0+1 day), t0+4 days, t0+9 days, t0+16 days etc til well-known. This combines a rough power function with the knowledge that sleep might be contributing to some of the effects of spaced learning/ consolidation. But again: this schedule is an informed but imprecise starting point - the exact optimum will vary by the individual (you) and the materials.
posted by dino might at 9:52 PM on May 24, 2012


There's a considerable amount of uncertainty about intervals, for instance it hasn't even been reliably confirmed whether expanding intervals are more effective than fixed intervals. In any case, if you are at all interested in the research behind spaced repetition, this is the best place to start:

http://www.gwern.net/Spaced%20repetition#literature-review

For some reason it doesn't come up immediately in search results; I only found it after a few attempts over several days. Good luck!
posted by gt76 at 4:24 AM on May 27, 2012


Here you go! (research study on this)

There's no cut-and-dry answer because the answer depends upon how long you want to know the information, and noone has done the work for multiple repetitions (as opposed to when to do ONE repetition).

If you're going to do a *single* repetition, the best interval tends to be 10-20% of the testing interval (so if you want to know something for 6 months, it's around 1 month, and if you want to know something for 1 year, it's around 2 months.)

The problem is that you're looking at doing multiple repetitions, and you're repeating each time until you know everything. A very long interval (4-6 months) between each repetition would make each session very frustrating, and while you may progress "optimally," you would have a consistent feeling of not knowing *anything*. The expanding intervals of memrise and Anki are not an ideal solution, but they're a pretty good balance of comfort and effectiveness, since they push you to a large interval as quickly as is comfortable, and if you want to remember something for the extreme long term, larger intervals are better.
posted by sdis at 6:44 AM on October 16, 2012


« Older The quick and dirty: I'm seein...   |  How do I understand and cope w... Newer »
This thread is closed to new comments.