SRS + "Brief Wakeful Resting"

Quantified Self Learning & Cognition
1

I don’t have easy access to a treadmill.

As Jeremy Howard mentioned in his talk, SRSing (is that a word?) is exhausting. Like him, after a period of about 20 minutes, I often reach a level of fatigue that makes it difficult to continue studying.

I 1st read about the “treadmill method” on Seth Robert’s blog & found it highly effective. Like Mr. Howard, I could study for hours without becoming bored; unlike Mr. Howard I didn’t track my error rate, so unfortunately I can’t offer more evidence for that effect. As I mentioned, the only problem here is that I don’t have easy access to a treadmill. My gym is quite far & it is impractical to go there every day, while I desire to SRS every day.

Fatefully, I stumbled across this study:

Brief Wakeful Resting Boosts New Memories Over the Long Term.

The gist of it is that subjects who spent 10 minutes relaxing after learning had improved long term memory vs. subjects who went on to perform some other cognitively demanding task.

Since reading the study, I have altered my SRS review method. I now study for about 20 or 30 minutes in a stretch, then spend 10 minutes lying on my back practicing “wakeful resting”. Unfortunately, I only recently discovered & became interested in QS, so I haven’t been recording data before & after. All I have so far are my subjective judgments.

My feeling is that this method is as or more useful than walking on a treadmill. For me, it is much more convenient. It has the added benefit of maybe enhancing long term recall, which is my end-goal with SRSing. It has the same effect of extending one’s ability to study for long stretches of time. It also leaves one feeling refreshed & ready for the next task. In fact, I often find that after 10 minutes of wakeful resting, I’m literally jumping to my feet when my phone’s alarm goes off.

I would love it if someone with a good data collection with SRS could adopt this method & track changes in error rates & so forth, as well as all of the quantitative elements I’ve mentioned.

If not, at least I’ve shared an effective method that will hopefully be useful to someone here.

2

Interesting approach. You know, it can be very easy to do this sort of experiment. If you have X new SRS items that you want to learn, then split them randomly into two decks of X/2 cards each. Do (at least) two SRS sessions a day of equivalent length, one with a wakeful rest after it on one of the decks, and one without a wakeful rest on the other deck. Alternate which deck you do first each day.

You can see how long it takes to get to the end of each deck, and you can measure your error rate on each deck. With enough cards, you could see if the wakeful resting method improves either total learning time or retention rate.

3

This is a great suggestion, but I have no idea how to randomly sort a large number of cards into 2 decks. It should be fairly easy to automate, but I have no programming background. Splitting up a deck of world countries & capitals into 2 decks would probably serve as a decent trial.

4

Sounds like you should learn or try programming.

What software are you using for SRS?

5

[quote=“kiba, post:4, topic:417”]Sounds like you should learn or try programming.

What software are you using for SRS?
[/quote]

I actually would love to learn programming. Although it is not technically programming, I used to do some amateur web design in HTML, and recently I’ve started looking at JavaScript. There’s a site that teaches you that I found really smart: http://eloquentjavascript.net/chapter1.html

I use Anki on my iPhone for SRS.

6

Anki is open source, and it’s available on desktop. If you have enough programming experience, you should be able to create your own plugin or hack the source code to make it do whatever you wanted.

7

Splitting an existing deck from within Anki could work, but would require learning much more programming than exporting it, splitting it outside of Anki, and re-importing it.

You could, for example, export the world countries and capitals deck into a spreadsheet, then randomize the order of the spreadsheet (Googleable), then import the first half into one deck and the second half into another.

8

We would really need someone to write Anki QS plugins, that make the work easier :wink:

9

They already graph thing for you, but there’s no way to export the information into CSV or JSON.

10

[quote]They already graph thing for you, but there’s no way to export the information into CSV or JSON.
[/quote]I know the graph thing that exists. It however lacks some interesting metrics.
For QS purposes “Median time per mature card” should be good.

Some days I can reach an average time per card of 3.5 seconds. Some days it’s 4.2 seconds.
Those are big daily changes. It would be good to have better tool to study them.

I would also like the ability to see how good a given day was at strengthen memories. Are cards that are marked correctly at that day more or less likely to be marked correct the next time they come up?

11

Now that I know, I want that data too! Anyway, the difference between the days probably have to do with the flashcard content.

Speaking of additional stats, I would like to see how many cards I add everyday as well if I reach a certain goal in adding cards. It would also be nice to know how many cards are on each deck at a glance.

12

[quote]Now that I know, I want that data too! Anyway, the difference between the days probably have to do with the flashcard content.[/quote]I think I have good reason to believe that isn’t true in my experience. Unfortunately I have no data to make the proof :wink:

But while we are at it: [(Time spent at card on day)*attempts at the card]/(Average time on card) would probably be better than going straight to the time.

13

I don’t entirely follow. Why do you need to split decks? Why not simply randomize whether one does ‘brief wakeful resting’ on each day, then test whether BWR increases one’s grades that day; or alternately, whether each flashcard reviewed that day has a higher followup grade? (Downside: requires a very long lag after the set of intervention days, enough time for many of the cards to have come up again for review.)