Better default (re)learning steps

Currently, the default learning steps for new presets are “1m 10m”.

However, the 1m step is too small in my opinion. It is well established that memory stabilization is greater when recalling the information requires effort.

Given that many users don’t bother changing the defaults at all, I think that Anki should use better defaults. So, I recommend using a slightly larger first learning step, probably 10m or 15m.

The following guides also suggest longer first learning steps:

Also, I think that Anki can show a warning when setting too small (re)learning steps.

In addition, Wozniak says that one successful recall is enough and repeated reviewing on the first day doesn’t have any effect on long-term memory.
Ref: Myth: Review your material on the first day several times - supermemo.guru

So, I think that Anki should also display a warning when setting more than 1 (re)learning steps shorter than 1d.

Just to be clear, I wanted to mention that the above suggestions are for both SM-2 and FSRS.

1 Like

Maybe there are better steps, but I am afraid that might depend on the content you are learning.

It seems all the resources you linked are geared toward conceptual learning and medical students mainly. For those that study vocabulary in a foreign language, for example, or drill audio comprehension, I think reviewing after a short period might be more effective. I don’t know of any research about that, though, so it is just a guess. But if anybody knows of any studies about this I’d love to know about it.

If this is actually the case, and different types of contents have clear-cut better learning steps, then I would suggest adding info about the different numbers in the steps infobox (like: for language learning this is recommended, for medical students…). But I suspect even within the same sphere, different decks (and even different people) would profit from different numbers.

2 Likes

This myth busting on Supermemo wiki lacks any science research reference. The source is “trust me bro”.
I am using Anki for language learning and in my opinion and observation I retain new information better if I repeat it 2 or 3 times in the first day.

3 Likes

that’s entire Woz stick. He even has an essay written why he dosen’t use evidence

Maybe there are better steps, but I am afraid that might depend on the content you are learning.
For those that study vocabulary in a foreign language, for example, or drill audio comprehension, I think reviewing after a short period might be more effective.

It might be possible that the optimal learning steps depend upon the content of the cards. But, I still think that 1m is too small. Maybe 5m is a good compromise?

Also, What Anki Learning Steps To Use (eshapard.github.io) says:

I’m using much longer learning steps now. They’re equivalent to 15 minutes, 1.5 days, 7 days, and 30 days.
[…]
These intervals came from the book How We Learn, by Benedict Carey. Apparently, these were the intervals recommended by researchers. I doubt they’re the most optimal intervals possible, but they were actually pretty close to the algorithm-generated intervals of some of my decks, so I just went with them.

So, the intervals seem to be backed by research. Interested people may want to check it out.

2 Likes

Thanks for the reference. I’ll try to look it up and see where the data comes from.

Maybe @L.M.Sherlock has some data on this from the FRSR research?

I haven’t built a model for the short-term memory.

2 Likes

@L.M.Sherlock
Out of the curiosity: Are you planning development of short term memory model? Getting learning steps advice from a proper model may be interesting.

1 Like

I think there is more noise in the short term memory. I have no plan to research it. Recently I started to improve the post-lapse stability prediction of FSRS.

3 Likes

The 1 minute step is an attempt to keep the the working set somewhere close to 7. Working memory - Wikipedia. Once cards have been successfully remembered for that period, the 10 minute delay helps confirm the memory is strong enough that it has a decent chance of lasting for 1+ days.

Users who are diligent about creating mnemonics or taking some time to rememorize the words they forget before rating again may find they can skip the 1 minute delay, but I think as a lot of Anki users will hit again as soon as they realise they’ve forgotten, the 1 minute delay is a more useful default.

5 Likes

“more noise” well said, I am going to have to borrow that phrase.