Assessing the feasibility of a wearable consumer device for capturing newborns’ sleep patterns in naturalistic environments

Abstract

Newborn sleep plays a vital role in early neurological and physiological development and may have lasting consequences for infants’ developmental trajectories. However, few methods capture high-density sleep data in naturalistic settings during the newborn period. We address this by assessing the utility of the Owlet Dream Sock for capturing sleep patterns, timing, and duration in the home. Thirty-nine newborns (51.30% female, M age = 9.38 days old, SD = 6.40) wore the Owlet for fourteen days during all sleep periods. Newborns exhibited two common types of sleep bouts, light-only bouts (awake → light sleep → awake sequences) and full sleep cycles (awake → light sleep → deep sleep → light sleep → awake sequences). On average, sleep bouts began around 11:48 AM, and newborns spent approximately 172 monitoring minutes awake, 536 in light sleep, and 54 in deep sleep per day. The duration of time spent awake and in light sleep did not change over the assessment period; however, both the duration of deep sleep and the number of full sleep cycles increased each week. While nighttime sleep minutes did not change, daytime sleep minutes decreased. The Owlet Dream Sock shows acceptable feasibility for capturing high-density newborn sleep data to capture developmental changes in sleep patterns over the first weeks of life.

Adequate newborn sleep plays a critical role in growth, learning, and neurodevelopmental maturation, whereas early-life sleep disturbances can adversely affect these domains of health and development (De Beritto, 2020; Fifer et al., 2010; Freudigman & Thoman, 1993; Tikotzky et al., 2010). Sleep develops rapidly and consolidates across the newborn period, with increases in quiet sleep and decreases in active sleep occurring in the first five weeks of life (Wielek et al., 2019). Given the importance of sleep for infant development, many methods exist to capture sleep metrics, including parent reports, polysomnography, and actigraphy. However, each of these methods has limitations, and difficulties with the collection of high-density sleep data on newborns (e.g., the inclusion of thousands of minute-by-minute sleep data points) have made research assessing newborn sleep in naturalistic real-world contexts challenging. In this study, we assess the feasibility of the Owlet Dream Sock, a wearable consumer health monitor, for research use in the home setting to capture newborns’ minute-by-minute sleep states in the first weeks of life. If shown to be feasible, the Owlet Dream Sock could provide researchers with a new high-density sleep methodology that can inform efforts to understand and promote healthy newborn sleep.

Accurately characterizing newborn sleep is important because early sleep behaviors, such as sleep bout duration, sleep timing, and the number of sleep-wake transitions, may reflect individual differences in neuroregulatory functioning that are observable shortly after birth (Freudigman & Thoman, 1993). The National Sleep Foundation recommends that newborns sleep fourteen to seventeen hours each day (Hirshkowitz et al., 2015), which aligns with parents’ reports of newborns’ sleep duration (Parmelee Jr et al., 1961). Additionally, findings from wearable actigraphy and parents’ reports suggest that infants experience comparable amounts of sleep during the day and night (Rosen, 2008; Schoch et al., 2020), though the consistency of sleep may vary each day depending on individual and environmental factors (Rosen, 2008; Sadeh et al., 2009). Among older infants and toddlers, the number of nighttime wakings and overall quality of sleep, captured via wearable actigraphy and parents’ reports, are associated with cognitive, motor, attentional, behavioral, and socioemotional outcomes (Hysing et al., 2016; Sadeh et al., 2015; Zhu et al., 2025). However, due to the unique burdens of the newborn period, few studies have examined whether newborn sleep contributes to these outcomes. One rare exception, which used a motility monitoring system in a hospital setting, found that sleep bout duration and the number of sleep–wake transitions on the first day of life both predicted cognitive and motor outcomes at six-months (Freudigman & Thoman, 1993); however, the environmental constraints of a hospital setting limit the ecological validity of these findings (Bueno & Menna-Barreto, 2016).

Capturing sleep in the home is necessary to understand how different aspects of sleep naturally evolve during this period of rapid developmental change. While many methods have been used to capture sleep in infancy, each has limitations for data collection among newborns. For instance, parent-report measures have been used to capture sleep routines and broad sleep metrics across infancy (Galland et al., 2012; Sadeh, 2004; Sadeh et al., 2009; Touchette et al., 2005). However, this method is limited because parents’ expectations, awareness of sleep, and their own sleep disruptions may bias the way they report their infants’ sleep (Quante et al., 2021; Schoch et al., 2020). Parents often under-estimate the frequency of their infants’ nighttime wakings (Galland et al., 2012), which may be especially problematic for research on newborns, as nighttime wakings are frequent during this developmental period (De Beritto, 2020). Parents may also find surveys burdensome to complete during the newborn period due to the substantial caregiving demands experienced during this time.

Another method used to assess infant sleep is polysomnography, which combines simultaneous observations of infant sleep with moment-to-moment cardiovascular, respiratory, and electroencephalography (EEG) inputs to more accurately capture infant sleep states (e.g., Grigg-Damberger, 2016; Wielek et al., 2019). Polysomnography has been used on newborns, with prior work finding that sleep-wake cycling can be identified via EEG in typically developing newborns as early as six hours after birth (Korotchikova et al., 2009). However, this method is limited because researchers typically collect polysomnography data in hospital or laboratory environments (Daftary et al., 2019; Rioualen et al., 2015), which may reduce ecological validity by disrupting newborns’ emerging circadian rhythms (Bueno & Menna-Barreto, 2016). Polysomnography is also resource-intensive, leading to reduced sample sizes and opportunities for repeated measurement (Rioualen et al., 2015; Wielek et al., 2019), which limits the types of computational models that can be conducted.

Actigraphy, a wearable device that continuously records participants’ motor activity, addresses some of these concerns and provides high-density, moment-to-moment estimates of sleep-wake patterns across infancy (Adams et al., 2019; Hall et al., 2015; Quante et al., 2021; Schoch et al., 2020). Results from actigraphy studies show decreases in daytime sleep, increases in nighttime sleep, and reductions in day-to-day variability across the first months of life (Adams et al., 2019; Davis et al., 2004; Schoch et al., 2020). A strength of these devices is that they can be worn continuously for several days, allowing for longitudinal assessments of infant sleep (Adams et al., 2019). However, most actigraphy research focuses on infants older than one month (e.g., Adams et al., 2019; Quante et al., 2021; So et al., 2005; So et al., 2007), and evidence regarding the reliability of actigraphy in the early newborn period is mixed, with some prior work suggesting limited validity for capturing newborn sleep (Rioualen et al., 2015; Unno et al., 2022). Even so, actigraphy devices illustrate the potential of wearable sensing for characterizing infant behavior as it unfolds in daily life using high-density longitudinal approaches.

Studies integrating multiple measurement approaches have identified discrepancies among reported sleep duration, number of wakings, and sleep timing in infants across measurement methods (Camerota et al., 2018; Hall et al., 2015; Quante et al., 2021; Werner et al., 2008). These inconsistencies highlight the lack of reliable and ecologically valid tools for capturing sleep patterns in newborns specifically. Together, these limitations underscore the need for novel wearable monitors capable of capturing high-density newborn sleep data in naturalistic home environments.

One new wearable technology that has the potential to address this gap is the Owlet Dream Sock, an FDA-approved wearable sleep monitor that captures infants’ minute-level sleep states, and second-level heart rate, SpO₂, and movement levels (Owlet Baby Care, Inc., Lehi, UT, USA). Unique strengths of the Owlet Dream Sock are that it is designed with newborns in mind and that it bridges the gap between researchers and participants by producing data that is accessible and beneficial for all parties. For researchers, the high-density continuous, multimodal data produced by the Owlet device may be used to understand how developmental changes in newborn sleep naturally operate in the home environment. Prior work supports this, and the physiological data collected from an earlier Owlet model, the Owlet Sleep Sock, were consistent with expected patterns among young infants (Mean heart rate = 136 BPM, SD = 9; Mean SpO₂ = 96%, SD = 1.7%; Dangerfield et al., 2017). For participating parents, the high-density continuous data are also available via the Owlet Dream mobile application, and parents may benefit from using these data to understand their infant’s sleep. Supporting this, in prior work, parents who use the monitor on their infants report feelings of decreased anxiety and improved sleep quality (Dangerfield et al., 2017). The Owlet device is also designed to be comfortable and secure for participating infants and is most used by consumers during the first two months of life, suggesting that it will be well-tolerated and unlikely to cause discomfort among newborn research participants (Dangerfield et al., 2017). Given that the Owlet Dream Sock is designed for newborns and provides mutual benefits for both researchers and participants, it has the potential to be a feasible and impactful tool for newborn sleep research. However, despite these strengths, there are no peer-reviewed studies assessing the feasibility of the device for research use, and no studies situating the descriptive sleep data produced by the Owlet Dream Sock within the broader infant sleep literature.

In the current study, we address this gap by assessing parents’ adherence to the sleep assessment procedure and by describing the minute-level sleep data collected from the Owlet Dream Sock. Feasibility is assessed by the number of days parents adhered to the protocol. Descriptive data include time spent in each sleep state, the time of day sleep occurred, the type of sleep patterns observed, and the device’s ability to capture developmental changes in these sleep metrics across the first weeks of life. The primary aim of the current study is to evaluate the feasibility of the Owlet Dream Sock to assess its suitability for research use. The secondary aim is to report the descriptive data produced by the device and situate these findings within the broader infant sleep literature. Establishing feasibility would enable the use of this wearable technology to model dynamic, longitudinal changes in newborn sleep and clarify potential associations with later developmental and clinical outcomes. By providing this foundational evaluation, the present study advances the use of wearable sensing to characterize infant sleep as a dynamic developmental process unfolding in daily life and to support future high-dimensional and computational analyses.

Method

Participants

This study draws data from an ongoing longitudinal investigation that aims to explore the role of early sleep difficulties as a mechanism underlying associations between newborn neurodevelopment and toddler attention and behavioral dysregulation. The recruitment goal is 300 families, and as of January 2026, n = 76 pregnant people and n = 56 partners have enrolled, and n = 51 participating infants have been born. See Figure 1 for the reasons for exclusion at each time point. Pregnant participants and partners, when available, are recruited during the second trimester of pregnancy and followed until their infants reach approximately 12 months of age. Potential participants and their partners are being recruited through community and hospital flyers, word of mouth, and targeted online ads through BuildClinical and social media platforms. Pregnant individuals are eligible to participate if they are between 18 and 45 years old, in their second trimester of pregnancy at the time of enrollment, fluent in English, carrying a singleton fetus, and have scores on a screening measure of emotion dysregulation that allow us to enroll participants with a wide range of distress (see below). Partners are eligible to participate if they are over 18 years old and sleep in the same home as the pregnant parent for four or more days of the week. Pregnant individuals and partners with significant health complications or who are in intensive mental health treatments are excluded to minimize participant burden. At all data collection timepoints, pregnant individuals and their partners are treated as independent participants and are compensated separately.

Figure 1.

Flowchart of included newborns and reason for exclusion at each time point

As part of an overarching aim of the parent grant, participants with high scores on the Difficulties in Emotion Regulation Scale (DERS; Gratz & Roemer, 2004) are being intentionally oversampled to achieve a more uniform distribution of emotion dysregulation. In the current sample, most emotion dysregulation scores are in the low to moderate range, similar to the general population; however, given ongoing recruitment efforts, a broad range of scores are represented (Pregnant individuals in our sample, M = 66.49; SD = 21.18; range of 40–121). Recruitment efforts also emphasize racial and economic diversity, which means that the current sample includes a greater proportion of non-White and Hispanic or Latino identifying participants than the local census. Our current participants have a higher median household income represented ($100,000 - $149,999) compared to $85,220 in the state where recruitment occurred (U.S. Census Bureau, 2024). See sample demographics in Table 1.). See sample demographics in Table 1.

Table 1,.

Demographic characteristics of participating newborns

Measure	Mean or Percentage
Mean age at assessment (SD)	9.38 days (6.40 days)
Mean gestational age at birth (SD)	39 weeks, 1 day (11.68 days)
Mean birth weight (SD)	7 lbs 10 oz (1 lbs 5 oz)
Delivery method (n)
Vaginal	51.30 (20)
Planned cesarean	28.20 (11)
Unplanned or emergency cesarean	15.40 (6)
Missing or declined to answer	5.1 (2)
Sex assigned at birth (n)
Female	51.30 (20)
Male	48.70 (19)
Race/Ethnicity (n)
White, not Hispanic or Latino	48.70 (19)
White, Hispanic or Latino	20.50 (8)
Multiracial, not Hispanic or Latino	10.30 (4)
Multiracial, Hispanic or Latino	10.30 (4)
Black, Hispanic or Latino	2.60 (1)
American Indian or Alaskan Native, not Hispanic or Latino	2.60 (1)
Another parent-described race, Hispanic or Latino	2.60 (1)
Declined to answer	2.60 (1)
Annual household income (n)
$5,000 – $7,499	2.60 (1)
$25,000 – $29,999	2.60 (1)
$30,000 – $34,999	2.60 (1)
$40,000 – $49,999	2.60 (1)
$50,000 – $59,999	10.50 (4)
$60,000 – $74,999	15.80 (6)
$75,000 – $99,999	7.90 (3)
$100,000 – $149,999	31.60 (12)
$150,000 – $199,999	18.40 (7)
Missing or declined to answer	5.30 (3)

This study presents analyses based on data collected from the first 43 families who completed the newborn assessment. Newborns were included in the analytic sample if any sleep data from the Owlet Dream Sock were available, yielding a final sample of 39 newborns. Four newborns were excluded from analyses due to a lack of Owlet Dream Sock data during the newborn period. With 39 participants, this study had 70% power to detect effects of r = .35 or higher (Funder & Ozer, 2019).

Procedure

As part of ongoing data collection, which begins in the second trimester of pregnancy and continues through the end of the first year of life, adult participants complete online questionnaires at eight timepoints, provide sleep data at eight timepoints, participate in three in-person assessments, and two virtual assessments. Participants receive incremental compensation for each study phase. Compensation includes up to $100 for completing second trimester procedures ($20 for daily surveys, $40 for the longer survey, $20 for assessments, and $20 for sleep tracking), $80 for completing third-trimester procedures ($20 for daily surveys, $40 for the longer survey, and $20 for sleep tracking), and $60 for completing the newborn procedures ($10 for surveys, $20 for assessments, and $30 for sleep tracking). Total possible compensation for each participant is $590 ($1,180 for the family unit) across the length of the study. All procedures have been approved by the University of Oregon's Institutional Review Board (IRB).

When birthing participants were in their second trimester of pregnancy, families were provided an Owlet Dream Sock (Owlet Baby Care, Inc., Lehi, UT, USA), which was used during subsequent postnatal assessments to assess their newborns’ sleep. At the third-trimester assessment, study staff assisted families with the setup of the Owlet Dream Sock and the Owlet Dream mobile application. Families created their own account with their own username and password, and one account was created for each newborn. Participants were instructed to use their de-identified participant ID in place of their name, so their newborn’s data could be anonymously connected to them. Study staff ensured that the Owlet Dream Sock was enabled to find devices and connect to the participant’s home internet. If internet connectivity issues occurred during the setup, study staff were equipped with a master sheet of troubleshooting instructions for every internet provider in the recruitment area and were able to resolve issues as needed. After the device setup was complete, participants were instructed to begin using the monitor as soon as possible after their newborn’s birth.

Following childbirth (M = 9 days postpartum, SD = 6 days), participants began the newborn assessment period. Daily surveys were omitted from this time point to reduce participant burden. Parents were given a step-by-step visual guide on how to place the Owlet Dream Sock on their newborns, and they were instructed to use the monitor during all sleep periods for 14 consecutive days, including naps. If the newborn woke up and was moving around, parents were instructed to remove the monitor. When newborns were born preterm (i.e., born before 37 weeks of gestational age), the newborn assessment period was delayed until the newborn reached at least 38 weeks but up to 40 weeks of gestational age. One newborn in the current sample was born preterm at 34 weeks and 3 days gestation, and the family was instructed to wait 35 days before completing the newborn assessments to ensure data were collected when the newborn was between 38–40 weeks of gestational age.

During the fourteen-day assessment period, the Owlet Baby Care Inc. data management team identified study participants using their de-identified participant ID, exported participant data in a .csv format, and uploaded them to a secure University Dropbox account at the end of each business day. Throughout the newborn assessment period, study staff reviewed the Owlet data files to ensure that the monitor was being used and was working properly. If families missed two full days of monitoring, study staff texted the birthing parent participant, reminding the participant of the importance of these data and checking in again to see if there were any technological issues or sleep irregularities affecting data collection. If families were having issues with the device, study staff provided technological support over the phone and via Zoom as needed. If technological issues could not be resolved over the phone or via Zoom, study staff provided participants with replacement sensors as needed, which typically resolved any remaining issues. If families missed five days of monitoring, study staff called both adult participants for a more in-depth discussion about protocol adherence to determine why they were not using the device and to provide further technological support if technological difficulties were the cause of the missing data. If, by the end of the two-week window, participants missed four or more days of Owlet Dream Sock data collection and technological limitations were not the cause of the missing data, then participating parents received slightly less compensation for that time point.

Measures

Owlet Dream Sock.

The Owlet Dream Sock is an FDA-approved wearable sleep monitor that tracks newborns’ sleep states, pulse rate, SpO₂ levels, and movement during sleep among infants from birth to 18 months. The Owlet Dream Sock generated second-by-second pulse rate, SpO₂, and movement data, and minute-level sleep state. Sleep states were automatically coded by the monitor as 2 = awake, 1 = light sleep/REM, and 0 = deep sleep/NREM. For data collection, each participant received a package that included four sock sizes, one photoplethysmograph sensor, one wireless charging base, and a charging cable. Participants also had access to the Owlet Dream mobile application, which provided real-time physiological insights and sleep trends over time.

Feasibility Outcomes.

Feasibility was operationalized as parent adherence to the newborn sleep assessment protocol, defined by the number of 24-hour days that the newborn wore the device. Adherence was categorized as especially low (≤7 days of wear; ≤50% of the assessment period) or especially high (≥10 days of wear; ≥71% of the assessment period). These thresholds were informed by prior research indicating that at least seven days of recording are needed to obtain reliable sleep estimates among infants and children (Acebo et al., 1999; Sadeh, 1996; Schoch et al., 2020) and that due to illness, technological difficulties, and participant non-compliance, data loss of up to 28% is common in actigraphy studies of young children (Acebo et al., 1999). Accordingly, the Owlet Dream Sock was considered feasible for research use if most participants (i.e., >50% of the sample) provided at least seven days of data.

Sleep Metrics.

To further evaluate the device’s suitability for research, descriptive data from the Owlet Dream Sock were used to characterize three sleep domains commonly examined in infancy research: sleep state composition, sleep timing, and sleep bout characteristics (Schoch et al., 2020). Sleep state composition was assessed as the proportion of monitoring time spent in each sleep state. Sleep timing captured the distribution of sleep across the 24-hour day, with daytime sleep defined as occurring between 7:00 AM and 7:00 PM and nighttime sleep between 7:00 PM and 7:00 AM, consistent with prior work (Adams et al., 2019). Sleep bout characteristics described the number and duration of discrete sleep bouts per day.

Missing data.

Attrition analyses were conducted to compare the 39 participants used in the analyses to the full sample. No differences in demographic data, emotion dysregulation, or parity were detected between the full and analytic samples.

Analytic strategy.

All analyses were conducted in R (version 4.5.2; R Core Team, 2025). Sleep metrics were derived from minute-level data for each sleep bout, defined as transitions from wake to any sleep state (i.e., light or deep sleep) and ending at the next wake epoch. Two predominant patterns were identified (see Table 2). Pattern 1 consisted of light sleep-only bouts containing only light sleep epochs (i.e., awake → light sleep → awake sequences). Pattern 2 consisted of longer sequences of light sleep → deep sleep → light sleep, which were operationalized as full sleep cycles. In addition to these distinguishable patterns, we also identified four anomalous patterns in our data that involve direct transitions from deep sleep to awake states.

Table 2.

Newborn sleep patterns observed during the 14-day assessment period

Pattern Label	Pattern Description	Observed Count
Light Sleep Only	Awake → Light Sleep → Awake	7,751
	Awake → Light Sleep → Deep Sleep → Light
Full Sleep Cycle	Sleep → Awake	3,758
Anomalous or Incomplete
Cycles
	Awake → Deep Sleep → Light Sleep → Awake	17
	Awake → Deep Sleep → Awake	9
	Awake → Light Sleep → Deep Sleep → Awake	3
	Awake → Deep Sleep → Light Sleep → Deep
	Sleep → Awake	1

Note: Each sleep bout may contain more than one full sleep cycle. For example, a sleep bout consisting of Awake → Light Sleep → Deep Sleep → Light Sleep → Deep Sleep → Light Sleep → Awake is coded as two full sleep cycles. This pattern description is simplified in the above table for maximum clarity.

Descriptive statistics were computed for device usage and sleep metrics at both the daily and minute levels. To examine temporal trends, we fit linear mixed-effects models (LMEs) using the lme4 and lmerTest packages (Bates et al., 2015). To assess potential linear changes in newborn sleep metrics across all newborns, each model included assessment day (i.e., 1–14) as a continuous fixed effect, and participant ID as a random intercept and a random slope for day. Only days for which data were available were included in analyses. Models were fit using restricted maximum likelihood (REML). For each primary outcome (i.e., total monitoring time; minutes spent in wake, light sleep, and deep sleep; total daytime sleep minutes; total nighttime sleep minutes; total number of light sleep bouts; and total number of full sleep cycles), assessment day was centered at 7.5 (the midpoint of the 14-day assessment period) and scaled in weeks by dividing by 7. This parameterization allows fixed and random effects to be interpreted as expected values at the midpoint of the assessment period and as change per one-week increase in time, while improving numerical stability and interpretability of model estimates. We report results from unstandardized models, including fixed-effect estimates (B), standard errors (SE), and p-values. Statistical significance was evaluated at an alpha level of 0.05.

Results

Protocol Adherence.

Of the 39 newborns included in the analyses, 19 provided data for all fourteen days, whereas 20 had at least one full day of missing data (M = 3.15 days, SD = 4.56; see Table 3). Consistent with the predefined feasibility criterion (>50% of participants providing ≥7 days of data), the device met criteria for feasibility, with 82% of participants contributing at least seven days of data. Thirty-two families showed especially high adherence (≥ 10 days of wear), whereas seven families showed especially low adherence (≤7 days of wear).

Table 3.

Distribution of newborn monitoring adherence during the 14-day assessment period

Total Monitoring Days	Percentage of Analytic Sample (n)
1	5.13 (2)
2	2.56 (1)
3	7.69 (3)
7	2.56 (1)
10	5.13 (2)
11	7.69 (3)
12	10.26 (4)
13	10.26 (4)
14	48.72 (19)

We followed up with these families to determine the reasons for low adherence, and several contributing factors were identified. Some parents reported increased anxiety from the notifications provided by the Owlet mobile application, and some parents reported that the Owlet Dream Sock caused discomfort for their newborn. Exploratory comparison analyses were conducted to assess whether parents’ emotion dysregulation, parity, or newborn age predicted low adherence. Results indicated that emotion dysregulation and newborn age were not associated with missingness; however, first-time parents had more missing data (M missing = 4.82 days, SD = 5.26) than experienced parents (M missing = 1.40 days, SD = 2.44), t(33.24) = 2.71, p = .01. Notably, all seven of the especially low-adherence families were first-time parents.

For subsequent analyses, day-level missingness was handled using pairwise deletion due to the lack of auxiliary variables necessary for imputation.

All newborns were missing at least some minute-level sleep data. On average, newborns had 95 minutes (SD = 163 minutes) of minute-level missing data each day they wore the Owlet Dream Sock, resulting in 1130 minutes of missing data on average (SD = 1143 minutes) during the assessment period. Missingness on the minute-level is likely due to multiple reasons, including the previously mentioned adherence reasons as well as potential technological limitations, such as internet connectivity and newborn movement affecting the sensor placement.

For subsequent analyses, minute-level missing sleep-state data were addressed using a two-step procedure for intervals lasting five minutes or fewer: (1) Same-state anchors: When a missing interval was flanked by the same sleep state (e.g., awake → missing → awake), the entire interval was assumed to match the surrounding state, and coded accordingly. (2) Different-state anchors: When the states surrounding a missing interval differed (e.g., awake → missing → light sleep), a transitional sliding-window rule was applied. Minutes closest to each anchor were coded to match the adjacent anchor’s sleep state; for intervals with an odd number of missing minutes, the middle-most minute was assigned the preceding sleep state. Missing intervals longer than five minutes were handled using pairwise deletion, given a lack of auxiliary indicators necessary for imputation.

Sleep state composition.

Newborns wore the Owlet Dream Sock for an average of 11 hours, 20 minutes each day of the assessment period (SD = 4 hours, 26 minutes). During periods when the monitor was worn, newborns spent an average of 172 minutes each day awake (approximately 20% of monitoring time; SD = 121 minutes). They spent an average of 536 minutes each day in light sleep (approximately 65% of monitoring time; SD = 208 minutes) and 54 minutes each day in deep sleep (approximately 7% of monitoring time; SD = 40 minutes). The distribution of percentages of monitoring time spent in each sleep state is shown in Figure 2. The percentage of monitoring time spent in each sleep state for three representative newborns is shown in Figure 3.

Figure 2.

Distribution of percentages of monitoring time spent in each sleep state

Figure 3.

Distribution of sleep states across 24 hours for three representative newborns

Note. Each datapoint represents a minute-level epoch classified as awake, light sleep, or deep sleep, plotted by time of day. Figures illustrate the device’s sensitivity to detect sleep states and between-participant variability in sleep–wake patterning. Data are shown for a single assessment day per newborn

Sleep timing.

Across all participants and assessment days, sleep bouts began throughout the 24-hour day, with a modal onset time of 12:00 PM (SD = 5 hours, 44 minutes; see Figure 4). Sleep bout onset times were widely distributed, indicating substantial variability in sleep timing across the day and night, further depicted in Figure 3. Aligning with prior literature, newborns’ daytime sleep duration was assessed as total sleep between 7:00 AM and 7:00 PM, and newborns’ nighttime sleep duration was assessed as total sleep between 7:00 PM and 7:00 AM (Adams et al., 2019). On each assessment day, newborns slept on average 430 minutes during the daytime hours (SD = 123 minutes) and 187 minutes during the nighttime hours (SD =117 minutes).

Figure 4.

Distribution of sleep bout start times across 24 hours

Note. Sleep bouts were defined as transitions from wake to any sleep state (i.e., light or deep sleep) and ending at the next wake epoch. Sleep bouts were pooled across all newborns (N = 39) and all 14 assessment days. The distribution of sleep bout onset times throughout the 24-hour day is shown using 1-hour histogram bins; bar heights represent the frequency of sleep bout starts within each hour.

Sleep bout characteristics.

Across the full 14-day period, newborns experienced an average of 28.77 sleep bouts (SD = 12.46), with each bout lasting approximately 26 minutes (SD = 48 minutes). Bouts characterized by only light sleep averaged 19 minutes (SD = 28 minutes), and newborns experienced an average of 18.00 such bouts per day (SD = 11.08). Full sleep cycles, defined as light sleep → deep sleep → light sleep sequences, averaged 28 minutes (SD = 30 minutes), with newborns completing an average of 5.95 full sleep cycles each day (SD = 3.28).

Changes in sleep metrics over time.

Linear mixed-effects models were estimated to examine weekly changes in sleep metrics across the 14-day newborn assessment period. Estimates reflect fixed effects representing change per week. Total monitoring time did not significantly change across weeks. Similarly, neither daily minutes spent awake nor total minutes in light sleep showed significant linear trends over time. However, the duration of newborn deep sleep increased significantly over the course of the assessment period. Newborns experienced approximately ten additional minutes of deep sleep for each week of the assessment period (B = 9.52, SE = 3.59, p = .01; see Figure 5).

Figure 5.

Linear changes in newborns’ deep sleep across the fourteen-day assessment period

Note. Individual lines represent newborn-level trajectories, and the overlaid line represents the fixed-effect model-estimated trend. Deep sleep was defined as the sum of minutes classified as deep sleep by the Owlet Dream Sock for each newborn on each assessment day. For visualization, deep sleep is depicted as the total percentage of monitoring time spent in deep sleep. Data include all newborns contributing usable sleep recordings (N = 39) across all available assessment days.

Newborns’ daytime and nighttime sleep were also assessed, and while total nighttime sleep did not significantly change across the assessment period, total daytime sleep minutes significantly decreased. Newborns experienced approximately twenty-five fewer minutes of daytime sleep each week (B = −25.03, SE = 10.86, p = .03; see Figure 6).

Figure 6.

Linear changes in newborns’ daytime sleep across the fourteen-day assessment period

Note. Individual lines represent newborn-level trajectories, and the overlaid line represents the fixed-effect model-estimated trend. Daytime sleep was defined as the sum of minutes classified as either light sleep or deep sleep by the Owlet Dream Sock occurring between 07:00 and 19:00 for each newborn on each assessment day. For visualization, daytime sleep is depicted as the total percentage of sleep slept during daytime hours. Data include all newborns who had at least one minute of daytime sleep (N = 37) across all available assessment days.

With respect to sleep-bout patterns, while the number of light-only sleep bouts did not significantly change, the number of full sleep cycles showed a small but significant increase over time (B = 0.65, SE = 0.31, p = .04; see Figure 7).

Figure 7.

Linear changes in the number of newborns’ full sleep cycles across the fourteen-day assessment period

Note. Individual lines represent newborn-level trajectories, and the overlaid line represents the fixed-effect model-estimated trend. Full sleep cycles were defined as sequences within a sleep bout beginning with an awake state and including the ordered transition light sleep → deep sleep → light sleep. Sleep bouts that did not contain at least one full sleep cycle were excluded from this analysis, and a single sleep bout could contain multiple full sleep cycles (e.g., Awake → Light Sleep → Deep Sleep → Light Sleep → Deep Sleep → Light Sleep → Awake was coded as two full sleep cycles). Data include all newborns who experienced at least one sleep cycle (N = 36) across all assessment days.

Discussion

Newborn sleep is important for growth, learning, and neurodevelopmental maturation (De Beritto, 2020; Fifer et al., 2010; Freudigman & Thoman, 1993; Tikotzky et al., 2010), yet challenges with collecting high-density sleep data on newborns have limited research assessing sleep patterns in naturalistic, real-world contexts. In the current study, we address these limitations by assessing the feasibility of the Owlet Dream Sock for research use and situating the descriptive data produced by the device within the broader infant sleep literature. Findings indicate that the parents generally adhered to the 14-day assessment protocol, and the Owlet Dream Sock produced interpretable minute-level data across multiple sleep metrics. Together, these findings suggest that the Owlet Dream Sock may be feasible for newborn sleep data collection, and illustrate how wearable sensors can capture newborn sleep as it unfolds in daily life.

Families demonstrated generally high adherence with the Owlet Dream Sock protocol across the assessment period. Nearly half of families provided data every day, and three-quarters missed no more than three days, indicating that most families were able and willing to use the monitor as instructed, despite the substantial parenting demands of the newborn period. However, seven families showed substantially lower adherence relative to the rest of the sample. We followed up with these parents, and they had two recurring concerns with the device: (1) notifications from the Owlet Dream mobile application elevated parents’ anxiety regarding their newborn’s well-being during sleep, and (2) the Owlet Dream Sock caused their newborns discomfort. These concerns are consistent with prior reports on infant wearable technologies (Beňo et al., 2024; Dangerfield et al., 2017; Dosso et al., 2025), with comparable day-level missingness reported in other newborn actigraphy studies (e.g., Marthinsen et al., 2022). Exploratory comparisons indicated that first-time parent status predicted lower adherence, and all seven low-adherence families were first-time parents. The demands of the newborn period are likely more overwhelming for first-time parents relative to experienced parents, and this may be contributing to reduced adherence. Future research should continue to assess participant characteristics that affect participants’ experiences of the Owlet Dream Sock and their willingness to use the device.

In addition to demonstrating feasibility, the Owlet Dream Sock captured high-density sleep data that were broadly consistent with expectations for newborn sleep. Newborns wore the monitor for an average of 11 hours, 20 minutes per day (SD = 4 hours, 26 minutes). Given that newborns from birth to two months of age spend up to 70% of their time sleeping (Wielek et al., 2019; Yasova Barbeau & Weiss, 2017), the average monitoring time is somewhat lower than expected but remains close to anticipated values, particularly in light of the substantial variability in sleep duration reported in prior work (Rosen, 2008). The variability in monitoring times observed in our sample may reflect meaningful individual and environmental differences in newborn sleep, suggesting that wearable monitors such as the Owlet Dream Sock are effective at capturing this variability.

While wearing the monitor, newborns spent on average approximately 65% of their time in light sleep, 20% awake, 7% in deep sleep, and 10% in missing or uncertain states. The light and deep sleep states captured by the Owlet Dream Sock map onto polysomnography-defined active (REM) and quiet (NREM) sleep, respectively, and the overall pattern of greater time spent in light sleep relative to deep sleep aligns with expectations for newborn sleep (Grigg-Damberger, 2016; Wielek et al., 2019). However, while the overall pattern of results aligned with developmental expectations, the proportions of time spent in each sleep state did not align perfectly with prior estimates derived from polysomnography. Newborns in our sample spent a slightly higher percentage of time in light sleep and a slightly lower percentage of time in deep sleep than reported in other samples. In prior work, REM sleep accounts for 23–60.6% of total recorded sleeping time, and NREM accounts for 11.5–36% of total recorded sleeping time in newborns (Korotchikova et al., 2009; Rioualen et al., 2015; Wielek et al., 2019). However, newborn age in these studies ranged from ten hours old to two weeks old. This variability in infant age across studies may contribute to these observed differences, as sleep organization develops rapidly during the newborn period (Grigg-Damberger, 2016; Wielek et al., 2019).

Several other factors may also contribute to the differences in sleep state estimates. Prior studies assessing the percentage of time in each sleep state use polysomnography, and newborn sleep is assessed for only a few hours at a time (e.g., Korotchikova et al., 2009; Rioualen et al., 2015; Wielek et al., 2019), whereas in our study, the Owlet Dream Sock provided data for several hours a day across multiple days. Because the Owlet Dream Sock provided more data points, it may be that the estimates in our study more accurately reflect newborn sleep patterns. However, it may also be that the minute-level missingness in the Owlet Dream Sock data, which is likely due to poor internet connectivity and infant movement levels, may have affected the observed percentages. Prior studies typically rely on data collected in hospital or laboratory environments, which may address internet connectivity concerns; however, this also reduces ecological validity, which may further contribute to differences in sleep-state estimates (Bueno & Menna-Barreto, 2016). Future work combining multiple measurement approaches will be important for clarifying how methodological, ecological, and developmental factors shape sleep-state duration as captured by the Owlet Dream Sock.

The high-density data collected by the Owlet Dream Sock also enabled the examination of sleep bout onset times and patterns across the newborn period. Sleep bout onset times were relatively normally distributed across the 24-hour day, aligning with prior work indicating that newborns do not yet exhibit a circadian rhythm (i.e., sleep occurs throughout both daytime and nighttime hours; Galland et al., 2012; Meltzer et al., 2021). Additionally, newborns slept just over two times as many minutes during daytime hours, compared to nighttime hours, aligning with prior literature (Sadeh et al., 1996). Using data from all sleep periods, two commonly occurring sleep bout patterns were identified: light sleep-only bouts and full sleep cycles involving transitions between light and deep sleep. These patterns are consistent with expectations, given that the majority of newborn sleep is spent in active or REM sleep (De Beritto, 2020). In addition to these commonly observed patterns, the Owlet Dream Sock also captured four infrequently occurring incomplete or anomalous sleep bout patterns involving direct transitions between deep sleep and awake states. Direct transitions from awake to deep sleep are not typical in newborn sleep but may reflect significant exhaustion associated with being overtired, whereas direct transitions from deep sleep to awake may reflect abrupt environmental disruptions. Although these patterns were rare in our sample, they may provide important information about individual differences in newborn sleep organization and environmental experiences.

In addition to providing overall estimates of newborn sleep, linear mixed-effects analyses further demonstrate the device’s sensitivity to the rapid developmental changes in sleep during the newborn stage. Both minutes spent in deep sleep and the number of full sleep cycles increased across the assessment period, which is consistent with prior work showing that quiet sleep duration increases across the first weeks of life (Wielek et al., 2019), despite the data being collected in naturalistic home environments rather than laboratory or hospital settings. Additionally, newborns’ daytime sleep duration significantly decreased over the assessment period, which is consistent with prior literature indicating that daytime sleep decreases throughout the newborn period (Adams et al., 2019; Sadeh et al., 2009). This change in daytime sleep duration is thought to reflect developmental maturation as sleep becomes more consolidated during the nighttime hours (De Beritto, 2020; Sadeh et al., 2009; Yasova Barbeau & Weiss, 2017).

To our knowledge, this study is the first to assess longitudinal sleep patterns in free-flowing, naturalistic home environments among newborns. Using minute-level data from the wearable sensing monitor, we were able to capture patterns that are not typically observable with standard actigraphy devices (Smith et al., 2018). These findings underscore the potential of the Owlet Dream Sock for capturing high-density newborn sleep data in ecologically valid contexts (i.e., home settings). Minimally intrusive, minute-level data collected in home environments can be difficult to achieve with traditional sleep assessment methods, and data collected with the Owlet Dream Sock show promise for use in high-dimensional analyses (i.e., analyses featuring a large number of variables per sample), modeling sleep dynamics, and the consolidation of sleep across the newborn period. High-dimensional approaches are particularly important during the newborn period, as sleep is known to be both highly variable (Rosen, 2008) and critically important for later development (De Beritto, 2020; Fifer et al., 2010; Freudigman & Thoman, 1993; Tikotzky et al., 2010).

While the study has many strengths, there are some limitations to consider. One limitation is the socioeconomic composition of the sample, with a median household income of $100,000 - $149,999 compared to $85,220 in the recruitment area. Recruitment is ongoing and aims to increase economic diversity. A second limitation of this study is that the accuracy of the Owlet Dream Sock for capturing newborn sleep states cannot be directly evaluated due to the lack of prior research using the monitor and because we did not validate the device against polysomnography or other established methods. However, the sleep patterns recorded by the Owlet Dream Sock align with developmental expectations, providing preliminary evidence that the monitor may be acceptable for research use. Future studies with access to larger samples and multiple assessment methods will be important for validating sleep-state estimates, addressing missingness, and leveraging minute-level data in dynamical analyses to model newborn sleep patterns over time. These are all aims of the overarching study for which these data are being collected.

Findings from the current study suggest that consumer wearable devices may address the methodological limitations associated with collecting newborn sleep data and may bridge the gap between researchers and participants. As part of the longer ongoing study, we anticipate using these data to identify what aspects of sleep are most influential for toddlers’ attentional and behavioral dysregulation. This and other work are expected to clarify the origins and consequences of individual differences in newborn sleep dynamics and inform the development of interventions aimed at promoting healthy sleep during early life.

Highlights.

• This study is the first to use the Owlet Dream Sock to capture newborn sleep.

• Parents had good protocol adherence, but first-time parents had lower adherence.

• Newborn deep sleep increased by approximately 9.5 minutes each week.

• Full sleep cycles increased modestly each week.

• Daytime sleep decreased by approximately 25 minutes each week.

Data Availability Statement

These data are part of an ongoing research study. The study data are available and periodically updated as part of the NIMH Data Archive: 10.15154/bn6r-2k47