Hi Michael – thank you for taking the time to participate in my PhD research. I am trying to understand and describe the discourse around sleep-tracking, and how the language and vocabulary around ‘sleep’ is changing when looking at it through sleep-tracker data. You published so many articles on sleep science. You are possibly the most published academic in that field and you are also working as a consultant for companies developing fitness-trackers and thus sleep-trackers. How did you get ^interested in working as a behavioural sleep scientist in the first place?
I always thought sleep was interesting. When I was a kid I thought sleep and dreams were cool to think about and read about and learn about. In college I found we had a sleep lab on campus that was doing sleep research. For me, that was the coolest thing ever and the guy who was running the lab [Michael Perlis] taught an undergraduate course that I signed up for. I loved it and volunteered to work in the lab and turned that into an independent study project and then an honours thesis and then, you know, I went to a clinical psychology programme that had a sleep research programme in it and from there I did a postdoc in a lab that was sleep focused.
AN Sleep-tracking is currently, maybe since the past 10 years, very present in conversations about health and wellbeing, where do you see sleep sitting within that?
MG Sleep I believe sits at an important point whereby the environment that we live in – our experiences in our life and who we are and where we come from – interface with our biology, our psychology, our health, our immune system, our thinking – all of the things that exist within us. Diet plays a similar role, but the difference between sleep and diet is that you know when you’re eating and you know what you’re eating when you’re eating it. When you’re asleep you’re not conscious of it and you don’t really understand what’s happening when you’re doing it so you’re not quite sure whether your sleep is good or not. Sleep tracking can be a useful tool to help open up our eyes to what’s going on under the covers. Sleep trackers tell you something about yourself that you might not have known otherwise.
AN What effects do digital media more broadly have on sleep?
MG Lots! I mean, sleep is very much an embodiment of the world around you. The most obvious is electronic media at night. First, the light itself interferes with circadian processes that try to get you prepared for night time. Your body sees daytime when it’s expecting night time and has a hard time getting into that nighttime mode. Second, the mental activation and stress that come from experiencing things like social media that are meant to get your attention, are playing on your emotions, and strong emotional responses are not really conducive to winding down. And the third thing is time displacement where people lose sense of the passage of time when they’re lost in a media that’s going at its own pace. Those are three ways that electronic devices can compete on sleep. That’s different from sleep tracking technology and how that might impact on sleep. Sleep tracking technology you’re usually not using at night, you’re usually looking at it in the morning as to what your sleep was the night before, or you’re trying to use that information to prepare for sleep. I see sleep trackers as measurement tools, not as intervention tools. They’re not the change agents in and of themselves
AN Do you think behaviour-change models feed into how sleep trackers are developed?
MG Not really at the moment, but I think that’s where the future is going. To be honest, at first computer scientists and engineers, not behaviour people, developed a sleep-measurement tool; they made it as good as they could and as scalable as possible. The question of how to determine its accuracy and whether it’s actually useful came a little bit later, that’s where we are now. There’s great technological innovation, but it’s just catching up to 1960s behaviour theory. Part of the issue was that you had people who could build it but didn’t really understand what it was that they were building. They understood the machinery of it, the mechanics of it, but not necessarily how it’s used or what it’s for. As a field we designed tools to measure sleep, assuming that if we measured sleep fairly well we would change the world because now people would know. Knowledge may be power, but knowledge is insufficient sometimes. People need to know what to do with that knowledge.
AN Who do you have in mind as a user, for whom sleep trackers are being developed?
MG The very first wrist-based sleep tracking device data was published in 1972, and the original use for wrist-based sleep tracking was for psychiatric patients who were in the hospital, to see what psychiatric inpatients did all day and how mobile they were. I don’t think that that’s the use case today, but I think it’s important to know the history. It started as a way to understand what people’s 24-hour movement patterns were. As the technology got better in the 80s it started broadening out beyond psychiatric patients into the general population, and more so sleep disorders patients. It was sleep and circadian researchers who were using it. From the 80s into the 90s it really transitioned into a way to understand what real world sleep was like outside of the laboratory in terms of overall patterns. Within the next decade or two it really started transitioning into more of a health metric of sleep. Now we know that sleep is an important health measure. It’s not just a psychiatric symptom, it’s not just a part of sleep disorders, it’s a part of overall health. That’s not a very new idea, it’s an ancient idea, but in terms of one that people took seriously and really devoted effort to, that’s a ten-year-old idea.
People now using sleep trackers want to know if they had a good night. They don’t know but they have a feeling that they sort of did or that they didn’t, but they’re not quite sure and they don’t really know how to quantify that. What the tracker does is, it gives them a language to use, it gives them a vocabulary to put their experience in.
I have tried really hard with any company I’ve worked with to try and make sure that the data is presented in a way that balances scientific accuracy and integrity without being either obscure or difficult to understand. For example, a tracker is going to record lots of awakenings during the night. That’s normal, people wake up way more times during the night than they remember. You can hide that data, which a lot of trackers do because they say it freaks people out if they don’t remember those awakenings. Some trackers show it all and then people don’t know what to make of the data. I would argue that you should find a way to show the data, to be honest, but then help orient people to what it means and help show them that it is okay what they’re seeing.
AN So how is waking up in those terms defined? If my Fitbit tells me I woke up 32 times I can usually only remember one or two of them.
MG Right, it all depends on what it means to wake up, what does that even mean if you aren’t conscious of it. Does it count? This goes back to our gold standard sleep measure, which is polysomnography. When you’re looking at brain wave activity we look at it in 30-second intervals and there you have the sleep stages. Even though the science is sort of moving past stages, and we’re getting into more detailed analysis of waveforms, we still use these stages as in what stage were you in every 30 seconds. One step better would have been probabilistic but the field hasn’t figured this out yet. As it is, you have to be in one stage or another, which makes things very tricky because some 30 seconds are easier to score than others. But also, why is it even 30 seconds? The only reason we use 30 second bins is because in the old days the old EEG machines were pens on actual paper, dot matrix paper, that was folded every 30 seconds. So each sheet was 30 seconds that they looked at, they scored them one page at a time, by hand. That’s why we have 30 second epochs in sleep, not because anything biological happens. It’s just convenience, historical.
Sleep stages were originally described in the 1930s and we’ve been stuck to them. We’ve modified them over the years, but this idea of what stage are you in each 30 seconds stuck. The device measures the biology and then we try to understand it by putting it in these bins. You look at it and if you could score it as a sleep stage you scored it as a sleep stage, but if you see movement, or if the brain starts showing waveforms that look more like wakefulness than sleep, you scored it as awake. Did the person actually wake up? Who knows? The definition from an EEG from a polysomnographer of awake is wake-like waveforms in the absence of sleep-like waveforms for at least the majority of a particular 30 second epoch. Those happen many times during the night.
You’re going to talk to people who are going to tell you that polysomnography is the gold standard and so that’s the ground truth of sleep measurement. But actually you’re not measuring sleep directly with polysomnography, you’re measuring something downstream of something downstream of something downstream of something downstream from sleep. As if sleep was even a single unitary thing – you’re measuring an estimate of an estimate and applying arbitrary rules to it that aren’t biologically accurate. Nature didn’t create binary sleep stages, we did. It’s only the gold standard because it’s the agreed upon standard.
AN What problems and needs do users of such devices have?
MG I think some people want to learn about their sleep, but I honestly think that the number one driver of people engaging in any kind of sleep tracking behaviour is because it’s impacting their daytime functioning. They have a hypothesis that it’s being driven at least partially by their sleep and so now they’re trying to figure out how to fix this now, they’re trying to diagnose the problem and because they can’t observe their own sleep directly they go to a device to try and see what’s wrong with their sleep.
AN How is your own personal sleep routine?
MG My own personal sleep routine is great, but before I knew all this stuff I had terrible sleep. I was a typical college student staying up too late, a bit groggy in the morning and not being able to show up to stuff in the morning on time. Through working in sleep I learned that sleep is actually way more modifiable than I ever thought and than most people realise. There are many things in my life that I’m not good at, there are many areas of my life where I could be making better and healthier choices, but sleep is one area that I don’t worry about. It doesn’t mean that I sleep great every night, nobody does, sometimes I’m sick or I have to get up at 4am for a flight or something and there are things that get in the way of sleep that are out of control, but at least I know how to deal with them and keep them from becoming a problem.
AN What role do dreams play in relation to sleep data, if any?
MG Dreams are inherently interesting and they’re fun to talk about. Actually the science of dreaming is picking back up – it was very big in the 60s and 70s and then sort of died out because we hit a wall with technology – but now it’s starting to come back, with dreams as a model for brain plasticity and networked information in a way that that we haven’t been able to measure before.
The kind of dreams that people remember occur more commonly in REM sleep which is more towards the end of the night, so the people who are getting more sleep are probably going to remember more dreams, but there’s no technology that exists to my knowledge that can tell for certain if someone is in a dream or not.
AN How do you think users interpret their sleep data and then, what are actionable and not so actionable data points?
MG Yeah I think the sleep stages data is not actionable. Largely people think it’s much more actionable than it is. Don’t worry about not getting enough deep sleep for example – you don’t even know if that’s the case and there’s no real way to improve that anyway reasonably. The one score that’s probably the most actionable is whether you are giving yourself enough time to sleep, or too much time to sleep. If you’re waking up a lot during the night, maybe you’re spending too much time in bed, maybe you need to consolidate your sleep a little bit, it’ll feel better. You know how much you slept and that’s probably also the most accurate thing that they can do is whether you were awake or asleep and how much sleep you got. The sleep stages are not so accurate.
AN Great thank you and is there anything else you would like to tell me before we end here?
MG The history and context of this is really important because sleep trackers in the public’s view have been around for like 10 years, but for the sleep field they’ve been around for a generation and there’s a lot of history and a lot of context, which brings about a lot of misunderstandings about what they are and what they aren’t.
The first three rules of science are: know your measures, know your measures and know your measures. We often don’t know our measures, we assume them. What does it actually mean, what are we measuring? And if we’re validating it against the gold standard of polysomnography, what does that even measure, and why should they be related, and how should they and how shouldn’t they be related?
So that’s what I want to leave with. I’m constantly trying to push the field to not forget that this question is still an open question, and that there’s lots of assumptions in statements that people make about sleep tracking, like that it’s accurate. Accurate relative to what, measured where, in who, under what conditions, and how many of those things will change your answer?
I can’t wait till people in the technology field start knowing just enough about sleep to dramatically innovate – who can bring us out of this. We’re still in 1930s technology with polysomnography and sleep staging, isn’t that just crazy? It’s crazy to me that we’re clinging to it because we’re comfortable with it. There’s so much more potential.