October 3, 2013

Sleep Cycle App: Precise, or Placebo?

Thanks to the Internet, it's the age of self-diagnosis. People like to learn about (and treat) themselves through technology.

Especially when pretty graphs are involved (see fancy screenshot at right).

As a sleep researcher, I was interested in my friends' use of sleep-tracking apps, and I received a pretty positive response when I prompted them for their thoughts:

"I'm a believer."

"When I use it right, I feel less groggy."

The website and smartphone apps like Sleep Cycle use the average human's sleep pattern to determine the best window of time that you should wake up. The idea is that interrupting the "wrong" sleep cycle stage, such as slow-wave ("deep") sleep or REM (rapid eye movement, when dreaming occurs), results in grogginess upon awakening, as many of us can attest. Sleep researchers call this phenomenon "sleep inertia."

It's such a big deal that, in the sleep laboratory, we as techs are instructed not to wake participants if they're in REM, even if the experimental recording time is over.

So when a friend told me that he only feels refreshed after (according to his sleep-tracking app) eight REM cycles, I got a little skeptical, given the average person will only experience four or five REM periods per night.

What's the verdict on sleep-tracking apps? How do they work, and how accurate are they? Is it all a big scam, or perhaps the placebo effect at work?

How is sleep measured in the first place? To understand the premise of these apps, we must first understand how scientists quantify sleep: polysomnography.

Polysomnography (poly="many," somnus="sleep," graphein="to write") comprises a series of electrodes (the electroencephalogram, or EEG) designed to measure cortical brain activity. These electrodes pick up voltage fluctuations in ionic current flow across the particular area of neurons below which each electrode is strategically placed. These fluctuations appear like waves on the computer monitor. In addition to the EEG, various sensors are also placed on the body to detect breathing and movement throughout the night.

The American Academy of Sleep Medicine currently recognizes four (formerly five) stages of sleep: three non-rapid eye movement (NREM) stages (N1, N2, N3), and REM. As an individual proceeds throughout the night, several characteristic features define each stage on the polysomnogram.

Notably, from N1 to N3, the waves we see on the computer become larger in amplitude and lower in frequency, eventually resulting in that "slow-wave sleep" (deep sleep) from which we so want to avoid a jarring wake-up call.

About every 90 minutes, we slip into REM, which is characterized by EEG waves that resemble wake, whole-body muscle paralysis, and—yes—rapid movement of the eyes.

In the end, a pretty little roadmap of one's night's sleep, called a hypnogram, can be generated. It looks a little something like this:
...which is exactly what apps like Sleep Cycle and Sleep Time attempt to create for you.

Except that, without precise electrophysiological measures, the apps' premise is fundamentally flawed.

Here are a few reasons why:

1. Rarely does anyone's night resemble the beautiful, textbook-perfect image above. In reality, a night's sleep is peppered with brief wakes and arousals, especially in individuals with apnea, insomnia, restless legs, or even just a bed partner who chooses 3 a.m. to relieve themselves before stumbling back into bed. More often than not, it looks a little more like this:

For most of us, it's easy to fall back to sleep despite these brief arousals, as the hypnogram shows. But your phone's motion sensor doesn't know that.

2. Sleep cycles are not perfect 90-minute periods. A 1979 study by Feinberg and Floyd reported that, for the majority of subjects (around 70%), sleep cycles were up to 20 minutes longer or shorter than average. For the remaining 30% of subjects, the extremes were even greater.

Furthermore, it's been well-established that slow-wave sleep dominates the first half of the night, while REM tends to be longer the second half. No two sleep cycles are the same length.

3. How you sleep tonight depends on how you slept last night. Sleep deprivation studies, by my lab and others, have consistently demonstrated that slow-wave sleep (N3) is compensatory.

That is, if you typically have about 100 minutes of slow-wave sleep on an average night but then completely miss a night of sleep, you'll almost certainly have around 200 minutes of slow-wave sleep on your recovery night. It's actually kind of funny how perfect it is.

4. Sleep latency (the time between lights-out and when you actually fall asleep) changes. Every night. When you enter your wake time on, it reminds you that "the average human adult takes fourteen minutes to fall asleep, so plan accordingly!"

But think about the last time you had an exam or stressful event the next day and how long you were up ruminating about it. Or that time you went to bed two hours earlier because you had to get up early the next morning, and you just couldn't fall asleep. Or that other time when, after a long day, you can't even remember your head hitting the pillow before you were out.

Best advice? Don't worry about getting the "perfect" or optimal number of sleep cycles, per se. Don't even worry about waking up during the "right" stage of sleep. Research by Jewett and colleagues at Harvard suggests that the stage of sleep we've been woken from, despite what we perceive as sleep inertia, does not actually have an impact on cognitive performance.

Rather, focus on keeping a regular schedule—your body likes predictability, which explains those creepy mornings you'll find yourself waking up just minutes before your alarm is scheduled to go off.

So while sleep-tracking apps are certainly not the most accurate way to assess your sleep, the upside is that people are becoming more attuned to their own circadian rhythms and the effect of sleep (or lack thereof) on their bodies. This is great news, as sleep is arguably one of the most (if not the most) important health factor that can be altered.

Of course, I may be biased.

Do you use a sleep-tracking app? How accurate do you think it is? If you use the alarm feature, do you wake up more refreshed using the app? If you've never tried it before, are you more or less willing to try an app after reading this post?


Czeisler, CA et al. Human sleep: Its duration and organization depend on its circadian phase. Science 210(4475): 1264-1267 (1980).

Feinberg, I and Floyd, TC. Systematic trends across the night in human sleep cycles. Psychophys 16(3): 283-291 (1979).

Jewett, ME et al. Time course of sleep inertia dissipation in human performance and alertness. J Sleep Res 8(1): 1-8 (1999).

Tilley A, Donohoe F, & Hensby S (1987). Homeostatic changes in slow wave sleep during recovery sleep following restricted nocturnal sleep and partial slow wave sleep recovery during an afternoon nap. Sleep, 10 (6), 600-5 PMID: 3432860

Image credit: Sleep Cycle, My Aspergers Child, NascarEd, RazerM, Alessandro Zangrilli (via Wikimedia Commons), Karyn O'Keeffe (via Photobucket).


  1. This is a very interesting post because I occasionally use a sleep tracking app when I feel that I'll need the extra help waking the next morning. Finding out that the sleep cycle usually varies by up to 20 minutes does raise questions. Generally speaking though, planning to wake close to that 90 minute mark will result in a better chance of missing the Slow Wave Sleep. The second graph shows a repeated hour long sleep cycle. This is an uncommon sleep cycle according to your sources, which stated, 'sleep cycles were up to 20 minutes longer or shorter than average.' Even with a uncommon sleep cycle, the 90 minute sleep cycle has a 50% chance to land in a sleep cycle point preferred for waking. I think that though the sleep planners cannot read a person's sleep cycles, that it still poses a useful tool in managing one's sleep cycles.

    1. Absolutely agree that the tool can be useful, though I think it's mostly in the sense that it brings about greater awareness; awareness to not only how sleep works, but also how your body feels after each particular and varied night of sleep.

  2. We've just finished talking about sleep in my psychology class. These apps can't possibly work as well as people believe. I agree with you, everyone's sleep cycle is a little different. Circadian rhythm isn't the same for everyone. It also takes a different amount of time to fall asleep each night, depending on what's running through your mind. The best way to get a good night's sleep is to have a before bed schedule that you follow strictly. This will allow your brain to start shutting down before you lay down at night.

    1. Thanks for your comment, Elizabeth. Google "cognitive behavioral therapy" (CBT) if you'd like some further reading; that's what you describe! It can work for many (but not all) insomniacs, especially those with a psychological basis.

  3. Interesting article. I haven't tried these apps, but have paid attention to my sleep cycles for a number of years. Mine is approximately an hour and 50 minutes (couldn't tell from the article if that was unusually long or not). In my experience, I tended to do OK as long as I got at least 3 sleep cycles as a minimum, and 4 for normal days, or even 5 or 6 for catching up on sleep.

    So I would aim to wake up at a certain interval - 5:30, 7:20, 9:10 or 11:00 (h/m) later. That usually worked much better than setting the alarm for, say, 6:30 (h/m) later. I would feel much better if I just got five and a half hours of sleep and woke up at the end of the third cycle, than trying to add an hour and wake up midway through the fourth.

    So I don't agree with the advice to not worry about sleep cycles. Of course, if you can establish a regular sleep pattern, everything else might take care of itself, but for many of us, that is very difficult to maintain.

  4. Thanks for writing this article. I eyed the Zeo sleep clock for a while, and was on the verge of buying one when the company went belly-up. It seems that the only consumer-grade sleep trackers on the market now rely on actigraphy, which I understand to be quite a bit less accurate than what Zeo was doing with their headbands.

    Any thoughts on this? I'm not sure which is the bigger waste: money spent on an actigraphy-based sleep tracker, or time spent waiting for a real successor to the Zeo.

    1. Hi, thanks for your comment and for reading! What kind of information do you want to glean by tracking your sleep? Are you trying to keep a regular schedule, or avoid waking during dreaming or deep sleep periods?

    2. Thanks for the quick reply – sorry for my delayed response!

      I'm interested in the data as an end in itself, but also as a tool for optimizing my sleep. This includes all of the things you mentioned in your reply above, as well as a desire to determine the effects of things like meal times, late-night activity, and schedule on my sleep patterns.

      While I'm sure that simply establishing a routine would go some way toward improving my sleep, I guess I'm curious about the nuts and bolts of it. There are so many variables involved that it'd be nice to have some hard feedback on what's happening on any given night, and see whether it relates to other behaviors/patterns over which I have control.

    3. Your best bet would be to keep a journal where you describe what you ate, what you did before bed, and how you feel you slept, and see if you notice any patterns after you review a few weeks' worth of notes. An app may not be able to do that yet, but a pen and paper can! If you identify any specific issues that are concerning to you and you'd like to be seen by a sleep specialist, that is always an option down the road.

  5. Thanks for the interesting article. It doesn't seem to really get to the crux of the issue, though.

    Here is the question that seems unanswered: Is my Sleep Cycle app even measuring anything relevant to my stages of sleep? The website says it measures motion by sound. It seems highly implausible that it could measure my sleep stages by measuring my motion from my bedside table, even using sound. Setting aside whether the sensor actually measures movement (let's assume it does), for the app to derive my sleep stages there would have to be some kind of direct correlation between my movements in bed and my sleep stages. Your article doesn't address whether there is any such correlation.

    You're right that the graphs are pretty, and convincing. The variability of the graphs from night to night even suggests that the app may be able to accommodate your second, third, and fourth objections. But what about movement and sleep stages? Is there any such connection? If there is, then your objections above seem irrelevant. If there isn't, then the app seem like non-starter, and you should just get right to that point.


I appreciate all comments and strongly encourage discussion on the topics I write about! Please be respectful to me and other readers. Spam comments will not be published.