Pebble Sleep Data Demonstrates Biological Phenomena
Do you bounce out of bed happily with the morning sun, or do you hit the snooze button 5 times before zombie-walking to the shower? Asked another way… Do you think that the early bird gets the worm, or does the second mouse get the cheese? The correct answer is… It depends! It depends on your chronotype. So what is chronotype?
Chronotype, first described by Nathaniel Kleitman, refers to an individual’s predisposition to either go to bed late and wake up late (late chronotype, aka night owl), or go to sleep early and wake up early (early chronotype, aka early bird or lark). Everybody else falls somewhere in between, defined as the intermediate chronotype.
Chronotype is determined in part by genetics, which is why you’ll lay in bed staring at the ceiling if you try to go to bed at the “wrong” time. In addition to biological differences, evidence suggests that chronotypes demonstrate behavioral differences, with night owls more prone to extraversion, or overtly social behaviors (1–2). Finally, additional evidence suggests that there are gender associations, with women less likely than men to fall into the night owl group (3–4).
Given the established biological and behavioral differences between chronotypes, the data science team at Pebble was drawn to the following question… Do we see these trends in our data? Data in aggregate holds valuable potential for exposing meaningful insights around human biology and behavior. Thus, is it possible that we might be able to identify distinct chronotypes within the Pebble community?
In order to address this question, we fed the data from hundreds of thousands of sleep nights into a k-means clustering algorithm, a method that groups individual observations into distinct clusters based on the relative similarities of those observations. For example, if oranges and bananas were fed into a k-means clustering algorithm, they would be separated by their characteristics of orange/round vs. yellow/oblong.
In the case of our sleep study, we calculated average bedtime, wake time, total sleep, and deep sleep for each user. With these components as feature inputs, we ran the clustering algorithm on all users, and each individual was assigned to one of four groups.
As can be expected given that bedtime and wake time were input features of the model, our four groups included the following: one group went to bed relatively early; one group went to bed relatively late; and there were two groups that fell somewhere in between. The two groups that fell somewhere in between (intermediate chronotypes) were separated based on the third dimension of total sleep, with one group sleeping longer (~7.5 hours on average; seen in light green on the graph) than the other group (~6.5 hours on average; seen in light blue on the graph).
Cool! So our model identified what appears to be larks, night owls, and intermediate sleepers. But this begs the question… Do our groups actually represent those scientifically-established chronotypes that are driven by underlying genetics?
In order to dig deeper into this question, we looked at other characteristics of the individuals in each group. As mentioned earlier, evidence suggests that night owls tend to be more extraverted, so we assessed whether or not night owls received more social notifications when compared to others.
We found that, on average, night owls received more notifications on their watch relative to larks. Upon a deeper inspection, we found that night owls specifically received more notifications of the social variety (e.g. Facebook messenger, WhatsApp, Google Hangouts); but there were no group differences in the quantity of messages received from non-social notification sources (e.g. Google Maps, Gmail, etc).
The other chronotype correlate we tested was gender. Within our sleep groups, we found that there were significantly more females in the lark group relative to the night owl group, which is the directionality seen in prior research that suggested women are less evening-oriented than men.
Additionally, we found that women were more likely to be in the intermediate chronotype group that averaged ~7.5 hours of sleep (“long sleepers”), relative to the group that needed less sleep (“short sleepers”, with ~6.5 hours of sleep on average). This is fascinating, given that the National Sleep Foundation asserts that women need more sleep than men.
To summarize, not only did we identify distinct clusters that aligned with night owl and lark tendencies, respectively, but the behavioral and biological data also suggested that these groups represented biologically established chronotypes! It is incredibly exciting to replicate this phenomenon on humans in the wild, beyond the constraints of a science lab.
So what does this all mean for you? If you’ve ever made the claim, “I’m a night owl” or “I’m a morning person”… now you have the scientific data to back it up. And since knowledge is power, you’ll be able to make more informed decisions about your day. When are you most alert, creative, thinking constructively, and able to tackle difficult issues? Armed with a deeper understanding of this phenomenon, and a greater self-awareness of one’s personal chronotype, people should be empowered to choose when they participate in different activities, and when they protect time for sleep. This knowledge can inform everything from career choices to when you spend time with your kids, significant other, or simply the time you protect for sleep.
Beyond individual empowerment, this class of large-scale aggregate health research holds great potential for shaping our understanding of typical human behavior and biology. With greater depth of understanding around what is “normal” and normal variability, we will have a stronger baseline for identifying aberrations from the norm. Early identification of problematic aberrations will promote early intervention, thus leading to reduced costs of healthcare and greater wellness in general.
While the lark would say that the early bird gets the worm, the night owl can say that the early worm should have stayed in bed. Whatever your chronotype, we wish all a sweet slumber.
References
- Maestripieri D. Night owl women are similar to men in their relationship orientation, risk-taking propensities, and cortisol levels: Implications for the adaptive significance and evolution of eveningness. Evolutionary Psychology. 2014.
- Diaz-Morales J. Morning and evening-types: Exploring their personality styles. Personality and Individual Differences. 2007.
- Roenneberg T, Kuehnle T, Juda M, Kantermann T, Allebrandt K, Gordijn M, Merrow M. Epidemiology of the human circadian clock. Sleep Medicine Review. 2007.
- Adan A, Natale V. Gender differences in morningness-eveningness preference. Chronobiology International. 2002.
Data: Anonymous data from May-June 2016 was collected and analyzed in aggregate.
