[ Question ]

What metrics should I track when experimenting with sleep?

by [anonymous]1 min read8th Feb 202113 comments

10

Self ImprovementSelf ExperimentationSleep
Frontpage

I used to sleep very little (6 hours or so) and function well with extremely good academic achievement. Instead of sleeping 8 hours now, I would love those extra 2 hours each day again but I'm not sure it won't affect my cognitive ability in the short term. I don't trust my impressions and would love to have some quantitative data.

  • What are areas that might be affected the most (e.g., long-term memory, working memory, speed of arithmetic calculations, vocabulary, reasoning, mood, etc.)?
  • What are some tasks that would measure my performance in those areas? Ideally those tasks should be:
    • quick to perform
    • easy to implement (I could write a quick program to run those but if there is a web app somewhere that does that already, that would be ideal)
    • fine with the subject knowing what is being measured (I'm going to apply them to myself, after all)
    • produce similar results regardless of how much experience I have with them (I don't want to wonder whether improvement/regression is due to the changes in my sleep or getting better/tired with the task.) It's ok if the skill builds over a week or so and then stays constant.

 I'm not attempting to prevent long-term negative consequences since AFAICT nobody knows what those might be. Besides, there are significant changes in my lifestyle every month or so and so there would be too much noise in the data.

New Answer
Ask Related Question
New Comment

4 Answers

The thing I've found most interesting by far to track is reaction time.  

There's a lot of research showing that reaction time correlates with intelligence.  Unfortunately, most of this is on the scale of individuals rather than individual-days; but that is of course in part because it's hard to give someone an intelligence test and a reaction-time test every day for a while, and (relatively) easy to just give someone an intelligence test and a reaction-time test just once

I track simple reaction time (how fast I can hit a button after a visual stimulus) and also recently began tracking choice-involved reaction time (how fast I can correctly indicate one of two alternatives for a small parity problem).  Research shows the choice-involved reaction times seems to correlate better with intelligence.

Altogether the tests take < 5 minutes, it could be brought down to < 2 if you were ok with smaller sample sizes / smaller stimulus windows.

(Both improve for ~2 weeks as you start, but afterwards seem to even off pretty quickly.)

I've gotten interesting correlational results with nootropics, meditation, etc., while tracking this.  I haven't been tracking my sleep quality, but it also frequently matches my perceived quality.

(I've also tried DNB.  I'm less certain it's useful.   Overall, I think this is a really interesting problem; mentally, I designated it as the Taravangain Problem, after the fictional king in the Stormlight Archive, cursed with variable intelligence day-over-day, who takes a test administered by trusted stewards every morning to determine what level of authority those stewards will grant him.)

 

[-][anonymous]2mo 1

Brilliant! Sounds like exactly what I was looking for, thanks!

For the general usecase of tracking cognitive performance during self tracking Quantified Self folks build http://www.quantified-mind.com/

[-][anonymous]2mo 1

Thanks! I've set a reminder for myself to explore the website over the weekend. Looks like it will be extremely useful and I wish I had know about this resource long time ago.

[-][anonymous]2mo 1

Well, that was a short exploration. They require you to have a Google account and only work with Google Chrome. Thanks for the links anyway!

2ChristianKl2moWhy not just register a Google account for this one usecase and use some Chrome clone if you don't want to install Google Chrome directly?
1[anonymous]2moPure laziness. More specifically, it's a high upfront cost just to access a new site when I don't know whether the site is worth it. And these requirements suggest poor engineering and little attention to user experience which increases the chance that the site is not actually worth it.
2ChristianKl2moI think here it suggests that the website is designed for people with technical abilities for whom registering a throw away Google account wouldn't be an issue. It's an non-for-profit project where people are more interested in the actual intervention then they are interested in working on login forms and they want to use all javascript features when programming.

Besides cognitive measures, measures of stress such as the lowest night time heart rate and heart rate variance are good metrics to track.

[-][anonymous]2mo 1

Thanks but that requires better model of how humans function than I have. It's not obvious to me how to interpret those metrics.

2ChristianKl2moThey are metrics of whether people are stressed. If your lowest nightly heart rate goes up that's a sign that you are doing something that stresses your body. People who get a COVID-19 vaccine for example usually report that their lowest rate rate goes up in the night. The same goes for getting ill or being stressed at work. For heart rate variance higher numbers are better and indicate that the body is less stressed. People with burnout have very low heart rate variance. If you would observe that additional sleep increases your lowest nightly heart rate and reduces your heart rate variance that's a sign that your body recovers less well in the night.

You might find some of the writings on supermemo.guru regarding sleep interesting. 

SuperMemo is a spaced repetition app (it was the very first in fact) that has a thing called sleepchart built in. It lets you both track sleep for finding interesting patterns alongside getting altertness/cogniti8ve data from correlation of say hour of day vs. average grades:

There are a bunch of confounders (they could be eliminated were I more rigorous) but I think it's interesting and even without SRS data I get some useful data:

Basically red line is how long sleep episodes are, blue line is how frequent they are around that time. You can see there's a peak around 6 hours from waking and around 16/17 hours from waking. The green line shows breakeven (if I sleep at that time, for the amount of hours on the left, it'll add up to 24).

 

It's probably really complicated looking but if you want to try it let me know and i'm happy to show you how. Doesn't take long to learn and takes very little time to add track in a new sleep episode. 

(this is from my own data)

1 comments, sorted by Highlighting new comments since Today at 1:49 AM

Whatever your results, it may also be a good idea to control and/or experiment with all the things that can affect sleep quality, not just quantity. Lighting, blue light, temperature, humidity type of mattress and pillow, congestion, allergies, dust, air quality/filtration, and so on. That can easily affect how much sleep I need by quite a bit, and how well I function with less.