Study uses wearables to show that physical activity lengthens REM latency | Hacker News Reader

Study uses wearables to show that physical activity lengthens REM latency

news.utexas.edu

151 pointsgmays96 comments

This doesn't surprise me. A personal anecdote: a couple of years ago I traveled to Japan for the first time. During my time there, I walked much, much more than I usually do. As a result, I was physically exhausted at the end of each day, and had some of the best sleep I've had in years.

If you’re missing that now - perhaps a great opportunity.

yes lets leave our 9-5 to walk instead

You joke(?) but I have had tons of success in asking people if they’d like to go on walking meetings with me, both virtual and in person. I have sometimes clocked 20k steps at work!

Well, "Walk/Talk" is one of the most effective method. If the parties are abled, a walk/talk will give enough time even for non-walkers to talk and bring up topics/points. If they get tired, then the talk had extended beyond its need. I tend to have a few keywords to organize my calendar entries with the likes "TBD: foobar", "Plan: LoremIpsum", and one of them is "Walk/Talk: Awesome Person".

> one of the most effective method.

...well, for you, maybe. I'm of the kind that gets very easily distracted when walking outside. Also it's harder to take notes.

Transcription apps are your friend.

Do you have some good ones that you would recommend? I often type copious notes and I think there may be some value in the very act of the typing, but if something works really well, maybe I can change my approach.

On one "prepackaged" end the scale, MS Teams has an option to transcribe meetings. On the "Local" side, I've had success using Whisper.cpp to transcribe my own voice recordings.

I find it very acceptable and I appreciate when someone carries a small notepad (not a phone/tablet) with a physical pen to take notes. I do a lot of the times. However, here is the trick that works for me when you have none.

At the end of the meeting, the key thing is the actionable item (todo) -- say that out aloud. "So, my to-dos are this, that and this one. And you will be handling the other, and another." Say it a few times or even a few more times while adding to your calendar/notes (digital or otherwise) after the meeting.

Most of the times, you never needed to take the entire meeting's notes.

isn't that called golf for some execs?

The trick is to live somewhere where walking makes sense, and also add small opportunities to walk during the day. For example, if parking, park a bit further from where you're going. Or if you have calls to make, call people/businesses while walking. Or if hanging out with a friend, go for a walk or do something physical.

Everyone will have different constraints but it's generally possible to move more than you currently do.

Get a flexible job and spend a couple of those daylight hours active.

I do it all the time, don’t see why not

I actually get about 13,000 steps a day with a 9-5 job. I get two 15 minute breaks and an hour lunch, so I take three 15 minute walks per day. I work at a large campus, so it's not hard to get steps in.

The body gets used to it though

You can always try and walk faster

I've noticed the same phenomenon myself. On days when I don't go to the gym or engage in some otherwise exhausting activity in the evening, I get much lighter sleep. It also fits my personal wacko hypothesis as to why so many retired old people have so much trouble sleeping, which I continue to recklessly propagate despite a complete lack of solid evidence: they just don't work hard enough.

That’s definitely an intriguing idea, and imo would be worth studying. (Then we could ignore the results and instead demand pills that don’t help or have pyrrhic side effects if they do.)

My own anecdote is the opposite: I can't sleep well if I don't regularly exercise. I think we all kinda know this, but it's good to see more data to refine our intuitions.

Hey, that's the contrapositive, not the opposite.

Or idiomatically: “on the other side of the same coin”

Same experience here. That's my 'Convention Attendee' / 'Theme park '''vacation'''' exercise and get the heck to sleep program.

Sadly a Disneyland Paris season pass is more expensive than a home gym over the long term.

Not disagreeing, but another contributing factor is the exposure to outdoor sunlight which helps regulate circadian rhythms. People who sit at the beach all day sleep pretty well. I think both outdoor sunlight and exercise improve sleep quality.

Wrist watches can't really detect sleep stages. That requires at least an EEG to do reliably. The best they can do is maybe distinguish between sleeping vs awake. The only consumer device with an EEG is the Zeo, unfortunately they went bankrupt and people are unable to distinguish the flukes from the real ones.

https://pubmed.ncbi.nlm.nih.gov/25991187/

There was the Dreem 2 as well but they have also ended production.

Heh my SO just showed me her Garmin-something tracker log the other day, and it showed her having an hour of deep sleep when she'd been lying on the couch watching TV.

It's also showing various sleep stages whens he knows she's been awake, trying to sleep.

> It's also showing various sleep stages whens he knows she's been awake, trying to sleep.

To be fair, people actually aren't that great at detecting when they (or others) are really sleeping. Especially for one's self, it's common to believe that you spent more of the night tossing and turning than you really did, when in reality you actually got some light sleep sprinkled in those periods.

That said, consumer-grade wearable devices are demonstrably not great at this either.

I love Garmin watches for many reasons, but their sleep tracking is horrible. Even my ancient MS Band did a better job than my Garmin Fenix 6

It's pretty awful. Though they seem to be capable of detecting terrible sleep, at least. Gotta love a 40 sleep score night.

Agreed. But it can't tell if I'm jet-lagged and I wake up 3 hours into my sleep at 2:30 am and lay there still, yet unable to get back to sleep. That is sometimes an 80 sleep score night, when really it should be a 40.

I got a epix 2 pro which has their new heart rate tech. It tracks my sleep very reliably.

> it showed her having an hour of deep sleep when she'd been lying on the couch watching TV

The Garmin may be right considering the quality of TV... even if she had her eyes open...

Same here. Sleep tracking is mediocre at best. Some nights I’ve spent hours wide awake but lying down, the watch thinks I was awake for 20 min. Or when I wake up in the morning and read a book or scroll the phone to not wake anyone else up, the watch will usually log it as REM.

But it never showed any naps as actual sleeping time. Until I saw your comment I thought that the garmin would only log sleep overnight and never during the day. Did you have to enable something for it to log naps?

That study is nearly ten years old and almost certainly not relevant anymore.

The Basis Peak did a good job at tracking sleep but they got bought up and killed off by Intel for god knows what reason. The Peak also did automatic activity detection and it was quite good at it.

Basis also had hands-down the best visualization of activity and sleep trends, and was the only watch at the time to do full time HRM.

They claimed the Peak watches were a fire risk and recalled them, probably to have a reason to get rid of all the back end infrastructure.

Garmin still can't reliably autodetect biking and its sleep tracking is pretty terrible, too.

I miss the Zeo. The app would still work locally long after they shut the backend down, until there just was no more aftermarket / eBay supply of their unique silver coated contact padded headbands that actually did the brainwave monitoring. I do find the Apple Watch Ultra seems to be close to at least being in the ballpark in terms of stage and duration accuracy as its output is comparable to what I last got with Zeo, albeit 10 years ago.

This review channel begs to differ: https://www.youtube.com/@TheQuantifiedScientist/videos

He very throughly compares countless models with a Polar H10 strap. Some come close enough.

> Wrist watches can't really detect sleep stages.

As someone else noted, the paper you linked to is from 2015. Apple Watch couldn't do it until watchOS 9 (2022, 7 years later), but can now do this pretty reliably.

You might find the "Algorithm Development and Validation" section of this whitepaper interesting: https://www.apple.com/healthcare/docs/site/Estimating_Sleep_...

Thanks for linking this. I always did wonder how they did it. From a cursory read, it looks like the basic approach is to collect accelerometer data and train it to predict ground truth labels generated via a consumer EEG device. Not sure if each epochs (30s) were considered iids. I'd assume that the overall sequence of activity would be much more informative than a single 30s activity considered independently.

Sometimes at work I’m concentrating so hard my Apple Watch logs a sleep session…

Definitely not 100% accurate but as someone who checks their sleep log (and time in each stage using the Pillow app) daily, it seems pretty accurate and I can correlate how I feel on a given day back to that day’s sleep report with pretty good correlation.

Definitely not perfect but better than nothing.

If you’ve ever been on medication that suppresses REM sleep, Apple Watch detects REM rebound very clearly.

I won't be surprised. The accelerometers seem to be pretty sensitive. Enough to measure any variations in heart rates and breathing patterns which are likely highly correlated with sleep patterns.

I didn't think it tracked sleep unless you put it into Sleep mode explicitly.

Yeah likewise. If I fall asleep during normal wake hours, such as an unexpected nap, it will vibrate and remind me to stand up because I’ve been inactive for too long.

There should be a toggle for this.

There is. But it’s easy to forget because it’s an unexpected nap

i have purchased 2 amazfits and they work fine for sleep records, they use the zepp app. I've had 2 other watches that had sleep tracking, but the amazfit seems better, then again, it is much newer.

The data quality of wearables to detect sleep stages is in the 90%+ accuracy range now. Which is near what polysomnographers score at - nobody completely agrees on what stage is what, there is always deviation. Wearables are definitely good enough at this stage.

Zeo is not the only consumer EEG either, though all the current headbands (including what we are building at https://affectablesleep.com) have some sort of sleep stimulation at it's core.

Frenzband I believe is shipping, I haven't tried one, and don't know anyone who has. We have a Somnee at the office, and it is unbearably painful to wear. I'm highly skeptical of the "science" of either of these devices as well.

90+ accuracy seems very optimistic.

Still, I'd bet that whatever accuracy wearables get only work on a population, not individuals. I'm not surprised they are decent (though they can barely detect when you are sleeping...) but I don't put any faith at all on an individual basis.

Given the data they can collect that is the best anyone can hope for anyway.

So if you want to diagnose your own sleep, which you presumably would want to do if you are suspecting an issue, I wouldn't trust them one bit.

> The data quality of wearables to detect sleep stages is in the 90%+ accuracy range now.

Citation needed.

Personal anecdote: when strength training, I find that I have the best sleeps on the night and second night after I do a session of heavy farmer walks/carries. They are one of the most exhausting exercises for me and great sleep is a good reward.

REM latency is good because? I find I feel more rested with more REM cycles which generally means REM starts earlier.

> I find I feel more rested with more REM cycles which generally means REM starts earlier.

This is probably the case because your NREM / slow-wave sleep need was low to begin with. If you go to bed relatively rested, it's easy to be fully rested when you wake up. REM is typically delayed when you are exhausted.

Anecdotally, ever since I started wearing the whoop, I have been very cognizant of my sleep and recovery. My sleep quality has improved merely by following their recommendations. A great burden lifted tbh.

My health watch both adds to and decreases my mental burden. I am healthier. But I often think of the person who said, "My Garmin has made me less fun." My health watch has increased the amount of time I spend feeling guilt.

I tried wearing my Garmin to bed for a while, but the information just wasn't that interesting. I can see where science gets value out of a population study but for the N=1 population consisting of yourself, I just wasn't seeing the point.

I enjoyed getting the metrics (not sure how accurate they were) but my garmin is so clunky it's impossible to sleep in. Great for everything else.

I wear the women's model of the smallest one they make for this exact reason. Most of the watches Garmin is making are just ridiculously big.

On the contrary, I’ve been wearing my Apple Watch nearly 24 hours a day for a few years and find the sleep info extremely useful.

Really, the primary benefit I get is twofold:

1. I know approximately how long I slept for and remind myself of that all day via a widget. A short sleep is common for me, but I end up feeling foggy. Being reminded that it’s OK, I had a poor sleep, helps ease my mind.

2. I see approximately when I fell asleep and when I woke up. I end up having to add extra time if I sleep past my alarm, but seeing that I didn’t really fall asleep until late into the night helps me manage the pattern of late nights.

Both of those are easily covered by even semi-diligently tracking your sleep on paper, but having my watch do it for me is what I like!

> I end up having to add extra time if I sleep past my alarm

I find this one a painful design decision on the Apple team's part. The end time of the normal desired sleep schedule does not mean the literal end of sleep every day. I wish they wouldn't clip the data at the end of the sleep schedule. Maybe it's because they're using heuristics and so all the data is guesstimates?

I find the Autosleep app to be much more accurate than the Watch's native sleep tracking.

I have been using Autosleep now with an apple watch 6 - for 3 years daily.

The learnings and the gamification are huge for me. I got my deep sleep up from 1.5h to consistently over 2h.

Learnings:

- early to bed makes more deep sleep. Huge difference from 10pm or 12am

- magnesium doesn’t help at all

- sauna before does help, but hard to do daily

- deep sleep less than 1.5h makes for an awful day with angry mood

- cold room: biggest lever. Haven’t pulled the trigger on an 8sleep yet. But that wouldn’t be my next thing if sleep wouldn’t be so great.

- also, as weird as it sounds: having something exciting to work on improves sleep by a lot.

YMMV

From Apple Watch, I learned I get 11 minutes of deep sleep last night, and generally average well under half an hour (for a full 8 hour typical). I can fall asleep easily, but apparently not very well. Yours sounds great.

Mine was absolute crap at sleep detection. Typically it would report that I fell asleep before I went to bed, or I was asleep longer than I was in bed.

Heart rate detection and so on was generally fine through the night, as was Hrv. I sort of wonder if the actual sleep detection is done on the phone, and if the phone is not nearby, it just fails. (My phone doesn’t go to the bedroom. Makes it far easier to actually sleep)

(Also, Watch is dead now, and didn’t replace it, so I might be misremembering the exact bits, but the results were generally paradoxical)

I've been considering buying an apple watch just for the stats. I really don't like how my co-workers watches light up when the move, or when they receive a message, so I wonder if I can turn off everything except health stat collection, at which point the whole thing seems a bit absurd.

If that's your only showstopper you can change that in settings. Look up Raise to Wake[1] and Notifications[2] in Settings.

I'm usually using Theatre Mode which keeps the screen off until an alarm or I wake it up and have pretty much all notifications off. I still get heavy use of the watch itself for timers, reminders, calendar, tracking workouts and health stats and whatnot so I'm very happy with it. [1] - https://support.apple.com/en-gb/guide/watch/apd748b87e2a/wat...

[2] - https://support.apple.com/en-gb/guide/watch/apd9b833c9f3/wat...

I used to be someone who might have a single beer with dinner a few nights per week. When I got a Garmin and saw what it did to my stress levels and sleep, I essentially stopped drinking at all except for rare social situations.

In addition to improving the numbers, I noticed a qualitative improvement in my sleep, energy levels, and exercise recovery.

What are measurements for stress levels? Are they accurate enough?

I may be wrong, but I think it's just heart rate and heart rate variability.

I find that I can pretty consistently see an increase in resting heart rate by several beats per minute the day after drinking, an elevated skin temperature and a reduction in Deep- and REM-sleep. Particularly heavy drinking will manifest itself as a large drop in heart rate variability.

Garmin wearable devices display a proprietary stress metric from their Firstbeat Analytics subsidiary. This is basically just heart rate variability (HRV) normalized to the user's baseline.

https://www.firstbeatanalytics.com/en/features/all-day-stres...

As for accuracy, compared to what? Any definition of "stress" is inherently somewhat arbitrary and subjective. There seems to be a decent correlation to how I'm feeling. But the metric isn't really actionable for most people.

Whether the measurements are accurate enough is not clear, but they are certainly precise enough. I (and all my friends) can easily see huge differences in HRV on nights of drinking, it's pretty amazing.

I had the same experience. It puts things into circumspect when you can physically see how erratic things like heart rate and sleep stage times differ with just a single drink. The difference for my was so stark you'd think it were a bug if it weren't for the fact I could spot the pattern and test to see the truth of things.

OK but once you've learned this fact (that I think you could have learned from numerous sources without any technology), what is the ongoing value of the sleep tracker? I feel it is in the category of things that you learn from but then stop needing, including a cadence meter on a bicycle, bathroom scales, and similar devices.

You never really know when it might matter. One huge thing is when I have a serious illness I can see my heart rate remains about 80bpm all night long. I know my immune system has quieted down when my nighttime heart rate gets back down in the 50s. This doesn't even require accurate sleep tracking. (In fact the watch has a hard time even noticing I'm asleep when I'm that sick, but it doesn't really matter, it's very useful.)

I also notice similar things related to heavy exercise etc.

Do you not get fever dreams? Even if I have a slight fever (say 100) my dreams are uniquely different, so I always know when I have a fever at night. Hard to explain "fever dreams" if you don't get them...

I've seen my heart rate elevated when I don't have a fever, and it can take several days to fully recover. Also no, I don't get fever dreams.

It's like the speedometer on your car. Even once you reach the speed limit, you may need to check your speedometer again and again, both for safety and comfort. Consider that different people place a different value on their health. In a high stakes environment like a school zone, you would prefer to check your speedometer more often. Also consider that the accompanying apps are gamified, as in addiction forming.

You can dial in several things if you know how well you slept:

I can judge how performant I'll be in the gym and fitness competitions - ie, how well I can push myself and recover. I can work out whether what I ate before I slept had an impact (high cards vs low, or if certain foods are a trigger for poor sleep). I can understand if things like meditation, light sources, and noises outside, among others, actually impact me or not. I'm constantly learning something new from it month by month.

I find my Apple Watch information useful. I don't remember when I fell asleep but my apple watch does. This keeps me honest about going to be too late.

Additionally, when I first got it, I found that I was only getting 40 minutes of deep sleep which is why I was always somewhat tired. After adding magnesium to my diet I am getting around 70 minutes of deep sleep. I wouldn't have been able to pinpoint this without the data. The apple watch is the best sleep tracker [1] so your data might be too noisy to find results.

[1] https://youtu.be/LPqtfC70QTU?si=6tTayfNCkcJwDayW&t=12m30s

Does Watch accurately capture when you fell asleep, as opposed to when you lay down and stopped moving?

Not that well. You have to extrapolate from when the first first deep sleep occurs in comparison to other nights.

It is interesting but I find Garmin data completely unreliable. For example, it claims I’m awake for 2-3 times each night for 30 minutes or more, it claims I’m asleep when I know I was on Tik Tok because I couldn’t sleep.

I had the same experience when I first started sleep tracking before beginning our start-up https://affectablesleep.com.

My oura would tell me I didn't get enough deep sleep, or REM sleep, and I was like - so what? What do I do with that information?

We're using auditory stimulation to increase the efficiency of deep sleep during the night. The data after a night of sleep is less valuable than the real-time intervention.

what do you think of binaural beats (cool edit pro did them the best, you could even use arbitrary audio) for helping with sleep?

I know it's probably a big ask to wear headphones prior to sleep but when i was younger i used binaural beats for a few years to try hacking my brain a bit. When it "clicks" the sensation is pretty noticeable and intense. I described the default "noise" binaural beats "activating" like the sound of a helicopter circling around you, but quietly and quickly.

The only research I've seen says that binaural are no more effective then monarual across the population.

But also, our stimulation isn't music, it's sound. It is 50ms timed near the peak of a slow-wave, so completely different mechanisms.

When people ask what it sounds like, my response is "if I gave you a pill, you wouldn't ask what it tastes like".

So I'm not the person to ask about binaural beats.

IIRC, the qualitative (or proven quantitative) connection between exercise and improved sleep is most pronounced with morning exercise (HIIT/weight lifting).

This is interesting to me, do you know any articles about this? I feel when I put off exercise until later in the evening(after 7pm) that I often get poorer quality sleep, but I'd be surprised if working out at 6am vs. lunchtime would make that much a difference.

I used to have to work out late at night (like 8-10pm), and found that 30+ mins of low intensity cardio and a long period of relaxed stretching after the workout helped.

This is one of the great “no shit” type studies. Good to validate basic and ancient wisdom I guess.

First you validate the no shit, and then you can ask why!

lol, the why is also no shit too though in this case.... because you are tired.

Some sequence of studies are not meant to validate that an effect exists, but to quantify the size of the effect. And the size of the effect under varying circumstances.

What? Exercise decreasing the proportion of sleep spent in the REM phase is "basic and ancient wisdom"?

Exercise improving sleep quality, sure, but this study is saying something much more specific than that

From this study I don't think it's even clear why or under what conditions higher REM latency is better, or that increasing it in an individual can necessarily be considered an improvement.

(I also don't think that's what the study was for; I think they were mostly trying to validate the methods against established results.)

“65 participants (Mage = 21.40, SD = 3.21, N = 42 females) with an average of 35.20 (SD = 27.01) days of observation per participant.”

Reply on news.ycombinator.com