Sleep Regularity Index: Formula, Range, and Wearable Use
The sleep regularity index scores whether the same minute is asleep or awake 24 hours later. Phillips et al. 2017 rescale that probability onto 0 to 100.
Short answer: The sleep regularity index (SRI) asks a simple question of a sleep/wake series. Pick a minute. Look at that same minute tomorrow. Were you in the same state both times, asleep or awake? Average that yes-or-no across the record, then stretch the result so a perfectly repeated schedule scores 100 and a random schedule scores about 0.
This page explains a published index. It does not diagnose a sleep disorder or tell you how to treat one.
The 2017 definition
Phillips, Clerx, O'Brien, Sano, Barger, Picard, Lockley, Klerman, and Czeisler introduced SRI in Scientific Reports in 2017 (7:3216, doi:10.1038/s41598-017-03171-4). They followed 61 undergraduates with sleep diaries for about 30 days. The index does not need one "main" sleep episode. Naps and split sleep count, because every epoch has a state.
Their methods compute SRI as the likelihood that any two time points, minute by minute, 24 hours apart, were the same sleep/wake state, across all days. That likelihood can run from 0 to 1. They rescale it with
so the reported index runs from −100 to 100. The same clock time asleep or awake each day scores 100. A pattern that matches only by chance scores 0. They write that ordinary schedules land between 0 and 100, and that values below 0 are still possible in theory (sleep for 24 hours, then wake for 24 hours) and very unlikely in practice.
The paper does not print this as a numbered display equation. The sums below are the same likelihood written out, so the rescaling has something to attach to. If each day has epochs and the record has days, and when epoch on day matches epoch on the next day:
For minute-by-minute diaries, .
A worked sketch, not a row from the paper
Two days, four blocks of six hours, are enough to see the arithmetic. This is a teaching sketch. It is not a subject in Phillips et al.
Scroll horizontally if needed
| Block | Day 1 | Day 2 | Match |
|---|---|---|---|
| 00:00–06:00 | sleep | sleep | yes |
| 06:00–12:00 | sleep | wake | no |
| 12:00–18:00 | wake | wake | yes |
| 18:00–24:00 | wake | wake | yes |
Three of four pairs match, so and . Identical days give and SRI 100. A day of sleep followed by a day of wake gives and SRI −100. A pattern that matches half the time sits near 0.
What that student sample showed
In the 61 diaries, SRI ran from 38 to 86. The mean was 73 and the standard deviation was 11, and the distribution was negatively skewed. Sleep duration ran from 5.7 to 9.3 hours, mean 7.4 ± 0.7 hours. Among the 59 students who reported a grade-point average, SRI and GPA correlated at r = 0.37 (p < 0.004). The paper also compared circadian phase between more regular and more irregular quintiles. Those are associations inside one college cohort. They are not a diagnostic threshold, and they do not show that changing the index would change grades.
What a later simulation paper adds
Fischer, Klerman, and Phillips compared five regularity metrics in Sleep in 2021 (volume 44, issue 10, doi:10.1093/sleep/zsab103). Andrew J. K. Phillips is the last author. The full text is paywalled; the points here are from the Crossref abstract, which describes large-scale simulations rather than a wearable validation.
The five metrics are the intra-individual standard deviation (StDev), interdaily stability (IS), social jet lag (SJL), composite phase deviation (CPD), and SRI. Scrambling the order of days left StDev and IS alone and moved CPD and SRI, so the abstract treats CPD as a multi-day metric and SRI as a day-to-day metric. SJL picked up weekly changes and missed daily ones. StDev and CPD missed sleep fragmentation. IS and SRI responded in a similar way to naps and awakenings, and they diverged for all-nighters. StDev and IS needed more than a week of data for unbiased estimates. CPD and SRI needed larger samples to detect group differences.
Using it on a wearable series
A watch or ring can emit a sleep/wake series, and SRI can be calculated from any series that has a state in each epoch. That arithmetic does not, by itself, tell you whether the wearable's sleep detector matches a diary or polysomnography. Staging and overnight optical accuracy are covered in the PPG sleep tracking guide, the sleep-staging algorithms note, and the wearable accuracy during sleep review. Read those before treating a watch's "regularity" tile as the Phillips index.
SRI also answers a different question from a night of heart-rate variability. RMSSD compresses beat-to-beat intervals inside a window. Pulse rate variability versus HRV is about whether that window came from ECG or from PPG. The Garmin Fenix 6 nocturnal comparison is one published RMSSD check during sleep. None of those three is a regularity score across days.
Frequently Asked Questions
- What is the sleep regularity index?
- SRI is the probability that any two minutes 24 hours apart are in the same sleep or wake state, averaged across the days you have, then rescaled. Phillips et al. (2017, doi:10.1038/s41598-017-03171-4) map a raw probability x with SRI = 200(x − 1/2). The same state every day scores 100. A coin-flip pattern scores about 0.
- Can the sleep regularity index be negative?
- The rescaled index can theoretically run from −100 to 100. Phillips et al. say people in ordinary life land between 0 (random) and 100 (the same pattern each day). A score below 0 would require an inverted pattern, such as 24 hours asleep followed by 24 hours awake, which they call very unlikely.
- What SRI did the 2017 college sample actually have?
- In 61 undergraduates keeping sleep diaries for about 30 days, SRI ran from 38 to 86, with mean 73 and standard deviation 11. Sleep duration in that sample ran from 5.7 to 9.3 hours, mean 7.4 ± 0.7 hours. Those are diary results in one student group, not a clinical reference range.
- Does a wearable sleep stage automatically give a valid SRI?
- SRI needs a sleep/wake series, and a watch can supply one. This page does not cite a study that checks a named watch's SRI against the diary method in Phillips et al. Staging accuracy is a separate question, covered at /blog/ppg-sleep-tracking and /blog/ppg-sleep-staging-algorithms. SRI is also a different measurement from overnight RMSSD (/learn/rmssd).
- Is a low sleep regularity index a diagnosis?
- No. Phillips et al. reported an association with grade-point average in their student sample (r = 0.37, p < 0.004, grades from 59 of 61 students). That is an association in that group. It is not a cutoff, a cause, or a medical label.
Core guide: What Is Actigraphy?
Related articles
PPG Sleep Tracking SOTA: Staging Kappa, AHI/REI Cutoffs, and What Wearables Actually Measure
State of the art for PPG and wearable sleep tracking in 2026: four-class staging Cohen's kappa (SleepPPG-Net2, wav2sleep), AHI/REI screening cutoffs (≥5 / ≥15 / ≥30), ODI and hypoxic burden, plus public datasets and leaderboards.
Overnight Oximetry vs Sleep Study: What Oximetry Can Show, What It Misses, and Why False Reassurance Happens
Overnight oximetry can be useful for screening and triage, but it is not a full sleep study. This guide explains what oximetry measures, where it fits clinically, and why normal results can still miss meaningful sleep apnea.
Home Sleep Apnea Test vs Polysomnography: Where Wearables Fit and What They Cannot Replace
HSAT and polysomnography answer different clinical questions. This guide explains when each test is appropriate, what wearables and PPG can support, and what they still cannot replace in sleep apnea care.
PPG Wearable Accuracy During Sleep: Heart Rate and HRV Validation
How accurate are wearable PPG devices for sleep heart rate and HRV? Review of validation studies for Oura, Apple Watch, Fitbit, and Garmin during overnight monitoring.
Wearable AHI Estimation: Can PPG Approximate the Apnea-Hypopnea Index?
Can wearable PPG estimate AHI? See where ODI-based severity tracking helps, where calibration fails, and what validation wearables need before clinical use.