Why do the words in "long term memory" not stay there

Community-Created Courses
2

Long term memory is probably a misnomer for Memrise to use. It’s more like it’s in the "Memorized’ zone. As time passes, facts slowly slip from “Memorized” to “Forgotten”. The Spaced Repetition System (SRS) then reviews these facts individually to see which you forgotten. The facts you remember, it’ll test those at longer and longer intervals each successive time. The facts you forgot, you technically relearn them and they have shorter intervals again.

So, it can be Learn -> 4 hours -> 12 hours -> 24 hours -> 72 hours -> 1 week -> 2 weeks -> 1.5 month -> 3 months -> 6 months. To “Learn”, you were tested on an item 6 times in less than five minutes. After that, you are tested 9 times over the next year to verify you kept this fact in “Memorized” area of your brain.

Imagine learning those 100 facts. If after 4 hours you forgot 10% of them, ok, the remaining 90 are pushed further successively. Now imagine after learning those 100 facts and test them 6 months later. I’d imagine you’d forgotten 90% of the material. By spacing the reviews smartly, you retain near 100% of all facts without investing much time per fact.

6 Likes
3

oh okay that makes sense.
I just started, so I am still at the 4/12 hour range, whereas a long term user is farther up that chain?
thanks!!

1 Like
4

You don’t HAVE to review them five times a day, though :slight_smile:

Memrise doesn’t force you to do that; you chose to do that.

Some of the words I originally learnt three years ago, say something like “review in 167 days”. It is only at the beginning of the learning process that regular review is necessary.

3 Likes
5

Furthermore, the flash card system of memrise, you see a word, you learn, you review at digressive interval, is an introduction to a word.

You need to use that word (or image or info) to make sentences (for example), for yourself, in your learning process. This will further process the information and put it in context.

I have heard, some studies imply that if you work 7 bits of info (a sentence, a word, a concept) for 20 minutes, you can improve the data memorizing. So if you chose to learn in chunks of 7, you have a “scientifically” superior method of learning. I would encourage working sentences of the course (or create your own course with levels of 7 sentences of your own choosing)

2 Likes
6

By the way, there are two states: one where the review button is white and one where it’s blue. When the review button is blue, it means that you have words that need to be reviewed. When the review button is white, it means that you don’t have any words that need to be reviewed. You can still click the review button when you don’t have words that need to be reviewed.

1 Like
7

Oh, and another thing I like to do is to learn a new batch of words - say 100 in total, a few levels of a course - and just keep on going without reviewing.

And then, when the number of reviewed words gets really high - say, when 999 appears, then my tolerance level has truly been reached, because the system can’t record a number higher than 999, so that means I have 1,000+ words to review - I start to systematically review all the words in that course.

I am learning Swedish, primarily using a four-part course containing 8,000+ words in all. The number of words to be reviewed from the first part goes up very very slowly because I really do know most of those words now and the memrise algorithms seem to know that.

Words I don’t know very well will appear and re-appear more quickly, so, with the later parts of this course, the number of words to be reviewed goes up much more quickly than with part one. Does that make any sense?

I find it a good challenge to let the to-be-reviewed words accumulate and then test myself. When I review words (I always work with the web version), I often copy and paste the list of reviewed words (I set it to 50) and note the level of accuracy I achieved. I am at about 80 - 90% accuracy on the first part of my Swedish course now, whereas at the later levels I get between 50 - 80% accuracy.

I also note which words have been reviewed more often because that might simply mean that the definition of the word is not very good (I am also a so-called course contributor for these courses, so I can edit the definitions).

I am just passing on how I have worked with memrise so far to give you - perhaps - some new ways of approaching the system and to give you confirmation that using memrise to learn vocabulary really does work!!!

I can now read Swedish crime fiction with ease, which I certainly couldn’t do when I started just under three years ago!

Keep on keeping on, memrise works!

6 Likes
8

Yup, that is exactly how it works!

I have words to review in 169 days and other words to review NOW, others in just a few hours.

9

The brain has evolved to forget things, if you think about the sheer amount of information it processes in a day, most of it has to be discarded soon after. Short term memory is very short term, 5 minutes or so.
Remembering something for 30 minutes, it may well be gone if it’s not significant to you the next morning, yet reviewing it will tell the mind, it matters through re-exposure.

Reviewing takes very little time, I simply choose the “Learn new words” button, before review, then review items. Then I leave it a while, so I’m not doing all the repetitions of the word are at the same session. This worked very well for me and Memrise really pushed me into learning very many words at a rapid clip. The totally new ones were sticking to, not just the ones I’d seen or used a little, but not committed to memory ie gender etc

10

I don’t get the point of this, a review of a phrase in Memrise takes so little time if you know them well. There are supposed to be algorithms tuning the process to you, how can they work if you give them no data?

Most of my errors are due to using a phone, where accidental clicks or hasty pushes on screen are all too easy. Even the claimed 10-20% error rate on the stuff known well, seems crazy. If you use Memrise as intended you’ld be up close to 100% on the whole 8,000 words and I don’t believe it would take longer, as fresh things are rapid to type, nevermind when you hear or do multiple choice.

11

Thanks for responding to my post!

Can you explaining what you mean by “this” in your first sentence?

For your information, when I review 50 words on the first part of the 8K Swedish course, it takes me about 2-3 minutes if I go really fast and my accuracy rates are usually between 90 - 100%.

I also don’t understand what you mean by “giving them no data” in your last sentence. The algorithms get data in due course :wink: And it’s up to me how I choose to learn and, so far, I see no reason to change my strategy as it has worked really well. I am now reading Dutch crime fiction.

12

Hmmm, seems the forum loses the context or I just am used to different quoting system and nesting.

The this refers to the post where review is ignored until over due words due hit 999. Waiting until a thousand words are due for review starves a program of data and speed of review response assuming rate of learning of even 100 words a day. I felt from the data provided that the method sounds sub-optimal.

Using the Memrise language courses, which have video clips, waiting for 1,000 words would mean no review until tackling 80% of level 3, then reviewing the whole of level 1, 2 and almost all of 3 at the same time.

Whilst it’s great that it does work for you, it’s a public forum and perhaps I am wrong, but there seems to be flaws in the approach of ignoring word reviews; whilst I repeated data provided in your post and used reply, I was in fact writing in an impersonal way, though I now see “you” is ambiguous, it was a general you. I could have written more formerly “how can it work optimally if one gives them no data?”. I would not have posted a response at all, if I thought it would look like personal criticism.

There are implications of this Review Procrastination strategy and after all the OP was suggesting that review stops him learning new words. This old thread was found searching for discussion on a topic about “form” ie. variation in learning performance and recall.

As I understand it, an SRS system works by exposing the learner to repeated exposure to solidify the retention of words/phrases. Traditionally people would do this by leaving notes around the house and regular review of vocab lists. The whole point is frequent regular exposure and with scientific learning many students achieve scores of 100% or very close in tests.

13

Hi,
my language school “forces” me to have full LTM and grade students with memrise info. They check into students’ memrise accounts at unexpected, random times and grade us using the LTM word count percentage at the “random checking time”. Because of this, I have to repeatedly login to memrise everyday and review words that are not yet in my LTM. Because of that, I am struggling to find time for my other homework and activities. Is there a way to make memrise easier?

Loads of thanks to any kind soul who’s going to help me

14

I didn’t know that you can actually find out how many words are in LTM. And particularly research into sb. else’s profile and find out!? Can you elaborate?

1 Like
15

I don’t know your schedule and what the school requires, might be unreasonable amounts of words a day. Otherwise, with the mobile app you could use any spare time, commuting, waiting in line, ‘smoke breaks’ …

16

This may be an issue relating to the ‘group’ feature. Different stats are available to teachers:

https://memrise.helpshift.com/a/memrise-learn-a-new-language/?p=web&s=groups&f=groups-leaderboards-explained&l=en

4 Likes
17

Thanks, Allan, that does make sense!

18

@TianXiangRong That sounds rough :confused: I wonder if your teacher understands how the system works. 加油 :slight_smile:

My understanding is that any words that are waiting to be reviewed are counted as not being in Long Term Memory (LTM), and any words that are not waiting to be reviewed are counted as being in LTM.

The way that Memrise works is that the more times in a row that you correctly answer while doing reviews (if the Review button is dark blue), the longer the word will be marked as being in long term memory.

If you get a word wrong while reviewing, then this will reset things.

Therefore, in order to increase your score:

  1. Track when your next batch of words are due to be reviewed. This can be done with the Memrise Monitor Chrome browser extension (let us know if you need help with it).

  2. Assuming that you get the word right every time that you review it, I think the review intervals are 4 hours, 12 hours, 1 day, 6 days, 12 days, 24 days, 48 days, 96 days, and finally 180 days. Using this information, you might be able to do plan when you learn a word. For example, if you learned a word at 5PM, then you’d need to review it again at 9PM, and then the next day at 9AM. This starts to break down if you get words wrong while reviewing, but it could still be helpful.

Hope that’s helpful :slight_smile:

2 Likes
19

Not EVERY few hours…it really depends if you got them correct or wrong.

You already got a good detail explanation from others how the “classic review” (blue button with due numbers printed on) and over-watering (grey button, random words which are not due) works.

I want to add:

Because there is the “forgetting curve” so they test you on your RECALLING abilities in spaced (longer and longer) intervals.

However, with the fixed implementation on Memrise there are also some problems as pointed out here:

It does not really fit everyone and not all scenarios (easy words vs (more) harder words first SR interval, single words vs phrases / longer sentences, multiple-choice with given hints vs “all typing”, very difficult you get all the time wrong,…) when there is no dynamic adjustment in the algorithm.

Maybe you confuse the 6 planting steps of learning NEW words with the first 4-5h / 12h SR intervals and “classic reviews” (blue REVIEW button)?

This is the “good” thing about Memrise.

You are free in your decision to press the “Learn new words” button any time, either in the course or for a specific course level.

You are allowed to do this even if your backlog queue is FULL.

If you want to do only the first 3 learning steps for the first 10 words per level you have to use the mobile app.
My older Android version was based on that implementation that you had to use two “learning sessions” to fully plant words.

I used that concept to “half-plant” the first 10 words of multiple lessons, but not to fully plant them:
So I could avoid that these “seen” words were added to the “full” queue and I did not have to review them in the next 4h / 12h SR intervals.

Sorry, I am not aware of any Tampermonkey / Violentmonkey userscript for the Memrise web portal to FORCE this (6 vs 3 learning steps) mode.

On the Memrise web portal you are allowed to learn 20 NEW words (per session)…my older Android app only has a setting for 10.

1 Like
20

Thank you so much for your help! I’m using my phone to write this message and I think this might be sent to Memrise/Decks. If so, please help me convey my gratitude to user Neoncube on Memrise! Once again, a million thanks!

1 Like
21

@TianXiangRong Glad we could help :slight_smile: