A year-long study published in Frontiers in Physiology has produced one of the more detailed pictures yet of how training and sleep interact in recreational runners, tracking a cohort through Garmin wearables over twelve months to capture daily training volume, sleep structure and heart-rate variability. The central finding is one that will feel intuitive to most runners but has been difficult to demonstrate rigorously: training and sleep do not simply run in parallel, they actively shape each other, and the effect runs in both directions.
The researchers used a two-directional, one-day lagged linear mixed model to separate the effect of training on the following night's sleep from the effect of the previous night's sleep on the next day's training capacity. That methodological choice matters, because it allowed the team to show that intense sessions completed late in the evening carried a measurably worse effect on sleep quality than equivalent sessions completed earlier in the day, a distinction that simpler models tend to miss.
The data showed that periods of elevated training load were consistently associated with shorter sleep duration and lower sleep efficiency, while periods of reduced training saw sleep duration rise alongside a corresponding uptick in heart-rate variability. The researchers describe a supercompensation window, typically emerging several days to two weeks into a lighter training phase, in which improved sleep recovery and rising HRV appear to set up favourable conditions for a subsequent jump in performance.
The practical case the authors make is for wearable data to be built into dynamic monitoring systems capable of flagging early warning signs in recreational runners, rather than serving purely as a post-hoc record of training completed. That argument lands alongside other 2026 findings on the same theme, including survey data showing recreational runners averaging just 6.61 hours of sleep a night, well short of the seven-to-nine-hour range generally recommended for athletes carrying meaningful training volume.
As with most wearable-based cohort studies, the sample is a self-selected group of technology-using recreational runners in a single country, and the authors are careful to flag that the bidirectional pattern needs replication across different populations and ability levels before it can be treated as universal. Even with that caveat, the study adds useful texture to a growing body of research suggesting that sleep should be treated as a training variable in its own right, not an afterthought to be managed once volume and intensity have already been decided.