What the researchers wanted to know
We hear two common health messages: “get enough sleep” and “be physically active.” But it is not obvious whether people can realistically do both, day after day—and whether sleep and activity truly influence each other in everyday life (not just in short experiments). (Nature)
What they did
The authors analyzed objective, long-term data from 70,963 adults using two consumer devices: an under-mattress sleep sensor and a wrist-worn activity tracker. The dataset covered January 2020 to September 2023, across 244 geographic regions, totaling about 28 million person-days/nights. (Nature)
They asked two main questions:
How many people routinely hit commonly cited thresholds for both sleep and activity?
On a day-to-day basis, does sleep predict next-day step count, and does step count predict next-night sleep (and are those relationships non-linear)? (Nature)
What they found
1) Very few people consistently achieve both targets
Using common benchmarks (roughly 7–9 hours of sleep and >8,000 steps/day), only 12.9% of participants routinely met both. At the other end, 16.5% combined short sleep (<7h) with a sedentary step profile (<5,000 steps/day). (Nature)
2) Sleep was a stronger predictor of next-day activity than the reverse
In the day-to-day modeling, sleep characteristics were meaningfully associated with next-day steps, while steps had little association with the following night’s sleep.
Key patterns:
Sleep duration vs. next-day steps: In the unadjusted analyses, the curve peaked around ~6 hours, but when the authors accounted for time awake (a critical confounder—if you sleep less, you simply have more waking hours to accumulate steps), the peak shifted closer to ~7 hours and the differences got smaller. (Nature)
Sleep efficiency (how much time in bed is actually spent asleep): Higher sleep efficiency predicted more next-day steps (e.g., ~25th to 75th percentile difference corresponded to ~+282 steps/day, smaller after adjusting for awake time). (Nature)
Sleep onset latency (how long it takes to fall asleep): Taking longer to fall asleep predicted fewer next-day steps (again, somewhat smaller after adjusting for awake time). (Nature)
Steps → next-night sleep: Associations existed in places, but were small and often largely disappeared after adjusting for wake time—leading the authors to argue that the practical impact of “more steps today = better sleep tonight” was limited in these data. (Nature)
How to interpret this (without over-reading it)
The headline result is not “sleep 6 hours to be more active.” The paper’s own adjustment shows that much of the apparent “6-hour advantage” is explained by having more hours awake to accumulate steps; once you control for that, the peak moves toward ~7 hours and the effect size shrinks. (Nature)
A more defensible takeaway is: sleep quality (especially efficient sleep and falling asleep more easily) is linked with being more active the next day, while simply piling on steps may not strongly shift your sleep that night (at least as captured by these devices and metrics). (Nature)
Important limitations to keep in mind
This is a large, impressive dataset—but it’s not a perfectly representative slice of humanity.
Participants were self-selected consumers who bought Withings devices; the sample skewed toward developed regions and likely higher socioeconomic status, and in this dataset participants were predominantly male. (Nature)
The sleep sensor tends to overestimate sleep duration and time-to-fall-asleep compared with gold-standard lab sleep studies; if anything, the authors suggest this could mean the share truly meeting sleep recommendations may be even lower than reported. (Nature)
“Activity” was measured as steps, which misses other meaningful movement (e.g., cycling, swimming, resistance training). (Nature)
