Not all screen time is sedentary: evidence from aligned smartphone sensing and accelerometry data in adults

Abstract

Background

Screen time is frequently used as a proxy for sedentary behaviour in physical activity and public health research, an assumption largely derived from television viewing. However, because smartphones are portable, screentime can occur across diverse physical contexts. This raises uncertainty about whether it reliably reflects sedentary behaviour at the momentary level.

Methods

This opportunistic analysis used data from an ongoing cohort study of adult caregivers of preschool-aged children. Smartphone screen time was objectively assessed using the Android-based passive sensing app Chronicle, and movement behaviour was assessed using wrist-worn accelerometers. Accelerometer data were processed using GGIR and the SedUp algorithm to classify epochs as sedentary or non-sedentary. Smartphone and movement data were aligned at 5-s epochs during waking wear time. Primary outcomes included the proportion of smartphone screen-use time occurring during sedentary versus non-sedentary behaviour. Secondary analyses examined inactive versus active behaviour during non-sedentary screen use, and app-level differences.

Results

The analytic sample included 63 participants contributing 785 person-days (3.3-million 5-s epochs). At the participant level, the median proportion of smartphone use occurring during non-sedentary behaviour was 25.9%, with substantial inter-individual variability (range: 5.6–73.7%). Among non-sedentary screen-use epochs, 52.1% occurred during inactive behaviour and 47.9% occurred during active behaviour. App-level analyses revealed that social media applications were predominantly sedentary, whereas apps such as Phone, Maps, Camera, and YouTube showed higher proportions of non-sedentary use.

Conclusions

Smartphone screen time fails as a momentary proxy for sedentary behaviour because a substantial, variable, and app-dependent proportion occurs during non-sedentary behaviour. These findings may have implications for both the measurement and development of interventions aimed at reducing sedentary behaviour and/or smartphone use.

Supplementary Information

The online version contains supplementary material available at https://doi.org/10.1186/s44167-026-00109-y.

Introduction

Sedentary behaviour is a well-established risk factor for adverse health outcomes, and reducing sedentary time is a public health priority [1, 2]. In both research and policy, screen time has often been treated interchangeably with sedentary time [3]. For example, the Physical Activity Guidelines for Americans operationalise sedentary behaviour to include “television (TV) viewing or screen time” [4]. This framing likely reflects the historical origins of screen time research, which focused primarily on television viewing, a behaviour that typically occurs in seated, stationary contexts [5, 6]. Within that historical context, treating screen time as a proxy for sedentary time was a reasonable assumption.

Screen time has frequently been assessed as a proxy for sedentary time in epidemiological research [3, 7]. However, much of the evidence this is founded on relies on self-reported measures of screen time, which are known to substantially diverge from objectively logged use [8–10]. Even when objective measures are used, associations between screen time and sedentary behaviour have been inconsistent. For example, while some studies have reported positive associations between smartphone use and sedentary time [11, 12], a recent study using passive sensing and thigh-worn posture assessment (activPAL) found no association between daily sedentary behaviour and daily smartphone use [13]. Importantly, these studies examine associations at the aggregate (e.g. person- or day-) level, asking whether individuals who use their smartphones more are also more sedentary overall. Such analyses do not determine whether smartphone use itself occurs during sedentary behaviour at the moment it happens. In this context, we use the term momentary proxy to refer to whether screen time validly represents sedentary behaviour at the same epoch in which it occurs (i.e., whether the behaviours co-occur at the time of use).

The distinction between screen time and sedentary behaviour has become increasingly important with the widespread adoption of smartphones [14], particularly given the amount of time adults now spend on their phones (one estimate was an average 6 h per day) [15]. Unlike traditional screen-based media, smartphones are portable and can be used across a wide range of physical contexts. Individuals can use their smartphones while seated for prolonged periods, but they can also use them while standing, walking, exercising, or engaging in other non-sedentary activities, both in leisure and occupational contexts [13]. Consequently, as smartphone use now constitutes a substantial proportion of adults’ screen-based media use [15], “screen time” may no longer inherently reflect sedentary behaviour. Sedentary behaviour is explicitly defined by postural and physiological characteristics independent of device use [16]. Inferring momentary sedentary behaviour from total screen time introduces the risk of ecological fallacy.

Recent advances in objective measurement now allow the question of whether smartphone use occurs only during sedentary time to be addressed directly. Passive smartphone sensing can capture device- and app-level use with high temporal resolution [17–19], while accelerometry enables continuous assessment of movement behaviour and posture [20]. Aligning these data streams permits direct estimation of whether smartphone use coincides with sedentary or physically active states, moving beyond aggregate associations and treating smartphone use as a heterogeneous behaviour rather than assuming it functions as a momentary proxy for sedentary behaviour. In addition, understanding the behavioural context in which smartphone use occurs may also inform more targeted intervention approaches.

The purpose of this study was to examine when smartphone use coincided with sedentary versus non-sedentary behaviour using objective passive smartphone sensing and wrist-worn accelerometry in a sample of caregivers of preschool-aged children. Specifically, this study aimed to: (1) quantify the proportion of smartphone use occurring during sedentary versus non-sedentary time; (2) characterise inactive vs active behaviour during non-sedentary smartphone use; and (3) examine differences in commonly used smartphone apps during sedentary and non-sedentary use.

Methods

Participants and study design

This secondary analysis used baseline data from two observational studies involving adult caregivers of preschool-aged children: a pilot feasibility study [21] and an ongoing longitudinal cohort study (R01HD112311) [22]. Both studies were approved by the University of South Carolina Institutional Review Board (pilot: Pro00092634, August 2020; cohort: Pro00133829, December 2023). Participants were recruited across the Eastern United States through multiple channels, including social media, community flyers, local events, childcare providers, and research participant databases.

Eligibility criteria for both studies included: (1) being a primary caregiver of a child aged 3–5 years, (2) owning a smartphone, and (3) able to read and speak English. Study protocols were conducted remotely and included surveys, ecological momentary assessment, wearing a wrist-worn accelerometer, and passive smartphone sensing. The measurement period was 30 days for the pilot study and 14 days for the cohort study; however, analyses used the first 14 complete days of data, with a minimum of 4 valid days required for inclusion.

Measures

Passive smartphone sensing (Chronicle)

Smartphone use was assessed objectively using Chronicle, a passive sensing application available for Android devices [23, 24]. Chronicle records timestamped app-usage events via the Android UsageEvents API. Equivalent passive sensing data are currently unavailable on iOS devices due to operating system restrictions imposed by Apple.

Raw Chronicle data were processed using [25] publicly available software (version 0.1.4; https://github.com/uzaira0/chronicle-android-raw-data-preprocessing-app/releases/tag/v0.1.4) and a publicly available cleaning protocol to address background processes, implausibly long events, and gaps in coverage (version 1.0; https://github.com/joshculverhouse/chronicle-preprocessed-cleaning). App usage events, including home-screen events, were aggregated into 5-s epochs. Epochs were classified as “screen use” if > 50% of the epoch contained logged app activity.

For the purposes of this manuscript, smartphone use refers to periods during which the smartphone screen was active (i.e. on), as detected by Chronicle, and does not necessarily imply active engagement, viewing, or interaction by the user. It does not include use when the screen is inactive (e.g. listening to music while the screen is off), or locked (e.g. previewing messages on the lock screen).

Smartphone use was assigned to specific apps at the epoch level, such that each 5-s epoch of screen-on time was attributed to a single app based on Chronicle usage logs. App-level analyses were restricted to apps used for at least 30 min in total by at least 10 participants to ensure stable estimates and to avoid undue influence from apps with minimal or idiosyncratic use. Apps were grouped into broader categories based on Google Play Store genreId classifications (Supplemental 1). App and category-level estimates were calculated by aggregating app- and category-specific smartphone use.

Sedentary behaviour and physical activity classification

Participants were instructed to wear a wrist-worn Axivity AX3 accelerometer continuously on the non-dominant wrist for the full measurement period. Raw acceleration data were processed using GGIR (version 3.3.0) [26], with all default parameters (including imputation), and summarized as Euclidean Norm Minus One (ENMO). GGIR’s automated calibration, sleep detection, and non-wear identification were used. Days with < 16 h of valid wear time were excluded based on non-wear detection.

GGIR Part 5 epoch output (5-s) was used for all analyses. During valid waking wear epochs (invalidepoch =  = 0 & SleepPeriodTime =  = 0), GGIR-derived ENMO was used to classify epochs as inactive or active, using the GGIR default of 40 mg.

Sedentary behaviour was classified using a hybrid approach integrating a validated sedentary behaviour detection algorithm SedUp [27], with GGIR processing. Briefly, SedUp uses a logistic regression classifier based on wrist elevation and motion variability from raw wrist-worn accelerometer data to distinguish sedentary postures, defined as sitting or lying, from standing. Sedup output was aligned to waking wear epochs identified by GGIR and replaced inactive epochs where SedUp classified them as sedentary. Non-sedentary epochs, therefore, included inactive epochs not classified as sedentary (e.g. low-intensity upright behaviours such as standing), as well as active epochs. Specific sedentary, or active, behaviours were not observed.

Smartphone and accelerometer data were aligned at the 5-s epoch level using Chronicle timestamps. Epochs were then restricted to valid accelerometer wear periods and waking time identified by GGIR.

Statistical analysis

Analyses included only epochs where concurrent smartphone sensing and accelerometer data were available during valid waking wear time. Participants with 4–14 valid days were included. The primary outcome was the proportion of all of an individual’s smartphone use occurring during sedentary versus non-sedentary time. Secondary outcomes included inactive versus active behaviour during non-sedentary screen use and app-level movement behaviour compositions.

All proportions of smartphone use were calculated at the participant level and summarized as the medians and interquartile ranges to reflect skewed distributions, while ranges are reported to further characterise between-person variability. Calculation at the participant level thereby ensured equal weighting of participants despite differences in the number of valid days and total smartphone use time. To provide context for comparison with prior literature, associations between mean daily smartphone use and mean daily sedentary time were examined using participant-level averages. Pearson correlations were used to assess linear associations, with Spearman rank correlations used as a sensitivity analysis to account for potential skewness.

Results

Sample

Of 89 participants with Chronicle data from the two studies, 63 had at least 4 days of valid overlapping Chronicle and Axivity data. This final analytic sample contributed a total of 785 valid person-days of concurrent smartphone sensing and accelerometer data. Participants provided a median of 14 (range: 4 to 14) valid days each. Across all valid days, a total of 3.3 million 5-s epochs with smartphone use during waking wear time were included in analyses. Descriptive characteristics of the caregiver sample are presented in Table 1.

Table 1.

Participant demographics of n = 63 caregivers

Characteristics
Age (years), mean (SD)	35.7 (6.7)
Sex, n (%)
Female	59 (93.7%)
Race, n (%)
Black	11 (17.5%)
White	48 (76.2%)
Asian, More than 1 race, or Other	4 (6.3%)
Hispanic ethnicity, n (%)	2 (3.2%)
Education, n (%)
High school or less	8 (12.7%)
Some college or 2-year degree	24 (38.1%)
College graduate or higher	31 (49.2%)
Employment, n (%)
Employed	33 (52.4%)
Student	21 (33.3%)
Unemployed	4 (6.3%)
Other	5 (7.9%)
Income, n (%)
< $20 k	5 (7.9%)
$20 k–$49 k	17 (27%)
$50 k–$99 k	25 (39.7%)
$100 k+	16 (25.4%)
Movement behaviours (hours/day), mean (SD)
Sedentary behaviour	9.2 (2.2)
Inactivity	3.2 (1.0)
Physical activity	4.1 (1.8)
Smartphone use (hours/day), mean (SD)	5.7 (2.5)

Proportion of smartphone use occurring during sedentary and non-sedentary time

At the participant level, the median proportion of smartphone use occurring during non-sedentary behaviour was 25.9% [IQR 19.1–40.7%]; however, substantial between-person variability was observed (range: 5.7–73.9%; Fig. 1).

Fig. 1.

Participant-level distributions of the percentage of smartphone use occurring during sedentary time across valid days (orange, left). Horizontal lines denote medians and diamonds denote means; boxes represent interquartile ranges and whiskers indicate the 5th and 95th percentiles. The red boxplot (right) summarizes the distribution of participant-level medians at the group level

When participants used their smartphones during non-sedentary time, an average of 52.1% of that use occurred during inactive behaviour and 47.9% occurred during active behaviour. Although non-sedentary smartphone use was slightly more common during inactive than active behaviour at the group level, substantial between-person variability was observed in the relative contributions of these behavioural contexts (Fig. 2). Participant-level daily distributions for each non-sedentary behavioural contexts are presented in Supplementary 2.

Fig. 2.

Participant-level distribution of smartphone use across behavioural contexts. Each bar represents 100% of an individual’s smartphone use, partitioned by the proportion occurring during sedentary behaviour, inactive non-sedentary behaviour, and active behaviour. Mean daily sedentary time (Sed. Time) and mean daily smartphone use (hours/day) are provided for each participant. The group bar displays the median proportion for each behavioural context; segment widths are renormalised to sum to 100% for visual alignment, while segment labels reflect the unnormalised median values and therefore do not sum to 100%

Mean daily smartphone use was positively correlated with mean daily sedentary time at the participant level (r = 0.28, p = 0.026; Spearman ρ = 0.29, p = 0.023), indicating a modest positive association between smartphone use and overall sedentary time. However, a meaningful proportion, typically around one-quarter, of participants’ smartphone use occurred at times when they were non-sedentary.

App-level differences in sedentary and non-sedentary screen use

A total of 20 apps met the inclusion criterion of being used for at least 30 min by at least 10 participants and were included in app-level analyses. Substantial differences were observed across apps in the proportion of screen-use time occurring during sedentary versus non-sedentary behaviour. Social media applications were some of the most sedentary apps, on average 80% of individuals’ (n = 30) Instagram use was sedentary followed by TikTok (n = 16; 80%), and Facebook (n = 53; 75%). The least sedentary apps were Phone (n = 19; 46%), Camera (n = 23; 49%), Maps (n = 37; 51%), and YouTube (n = 36; 54%). However, sedentary behaviour still accounted for more than half of screen use for most apps (Fig. 3A).

Fig. 3.

Behavioural context of smartphone use by app (A) and app category (B). Bars represent 100% of screen-use time for each app/category and are stacked by the average proportion of time accumulated during sedentary behaviour, inactive non-sedentary behaviour, and active behaviour across participants who used the app for at least 30 min. Apps are ordered by the proportion of screen-use time occurring during sedentary behaviour. For each app, the number of users, mean daily sedentary time, and mean daily app-specific screen use (hours/day) are shown

When grouping individual apps into categories, the most sedentary categories were Games (n = 30; 73%), Internet Browsers (n = 63; 71%), Lifestyle Apps (n = 54; 70%), and Social and Communication Apps (n = 63; 70%). The least sedentary categories were Travel & Local (n = 39; 52%), Camera & Photos (n = 30; 54%), and Video Players (n = 37; 55%). (Fig. 3B).

Across apps, users showed similar average daily sedentary time despite large differences in the proportion of sedentary use within each app. This suggests that variation at the app level mainly reflects differences in usage context rather than differences in users’ overall sedentary behaviour.

Discussion

This study examined when smartphone use coincided with sedentary versus non-sedentary behaviour by aligning objective passive smartphone sensing and wrist-worn accelerometry in adults. Most smartphone use coincided with sedentary behaviour. However, approximately one-quarter of smartphone use occurred during non-sedentary behaviour. This directly challenges the assumption that screen time functions as a momentary proxy for sedentary behaviour.

Much of the existing literature treats screen time as sedentary behaviour for two reasons. First, traditional screen-based activities, such as television viewing, are predominantly sedentary [5, 6]. Second, screen use is positively associated with sedentary time when aggregated at the person level [11, 12]. However, smartphone-use has also been documented in non-sedentary contexts. For example, one observational study reported that 29.7% of observed pedestrians were walking while using a smartphone [28]. This finding supports the premise that smartphone use can occur outside sedentary behaviour, although prior observational evidence has not quantified the proportion of smartphone screen use occurring during sedentary versus non-sedentary behaviour at the momentary level.

In the present study, mean daily smartphone use was similarly correlated with mean daily sedentary time (r = 0.28), indicating that individuals who used their smartphones more tended to be more sedentary overall. However, at the epoch-level, a substantial proportion of smartphone use occurred at times when participants were non-sedentary.

Mean daily smartphone use was positively correlated with mean daily sedentary time in the present study, however, this association does not imply that smartphone use itself occurred during sedentary periods. Visual inspection of participant-level patterns suggests that lower proportions of sedentary smartphone use tended to coincide with lower overall sedentary time for some participants, but substantial overlap was observed across the remainder of the distribution. Together, these findings indicate that aggregate associations capture broad tendencies but do not reliably reflect the momentary co-occurrence of smartphone use and sedentary behaviour.

A key methodological implication is that researchers should avoid treating smartphone screen time as a direct proxy for sedentary behaviour when the research question concerns sedentary posture, low energy expenditure, or related health mechanisms. In these cases, screen time and sedentary behaviour should be measured separately, or aligned at the momentary level, rather than assuming that one behaviour represents the other.

Substantial heterogeneity was observed across individuals. Some participants used their smartphones almost exclusively during sedentary periods, whereas others routinely engaged in smartphone use during non-sedentary behaviour. This variability suggests that smartphone use is a flexible behaviour that adapts to situational context. It is not necessarily a fixed sedentary habit. These findings underscore the limitations of inferring behavioural states from aggregate screen time measures. This interpretation is consistent with prior work using objective mobile sensing work showing that smartphone use varies systematically across real-world spatial contexts. Specifically, app use has been shown to differ across habitually travelled routes and visited destinations, with patterns also varying by app type [29]. The present findings extend this context-dependent view of smartphone use by showing that app use also differs by movement context, including whether screen use occurs during sedentary or non-sedentary behaviour [29].

A substantial proportion of smartphone use occurred during non-sedentary behaviour, although this use was slightly more common during inactive than active behaviour. This pattern suggests that screen-on time frequently co-occurred with standing or other low-movement upright behaviours, which may help explain why smartphone use remains correlated with sedentary time at the aggregate level despite frequent non-sedentary co-occurrence. However, Chronicle captures screen-on time and does not distinguish between active handheld use and passive exposure (e.g. a video playing while the device is propped up). Accordingly, non-sedentary screen use does not necessarily imply concurrent manual interaction with the device.

Prior experimental work indicates that active smartphone use during upright and walking-related tasks can impair motor performance [30], suggesting that when interaction does occur during movement, it may not reflect functionally optimal activity. From a physical activity and health promotion perspective, these findings suggest that smartphone use during non-sedentary time should not automatically be interpreted as beneficial or detrimental. Smartphones may support movement through functions such as navigation, activity tracking, communication, or prompts, but active interaction with the device during movement may also introduce distraction or reduce the quality and safety of activity [28].

A novel contribution of this study is the identification of substantial variation in the behavioural context of smartphone use across apps. Social media applications were among the most sedentary, whereas apps such as Phone, Maps, Camera, and YouTube exhibited comparatively higher proportions of non-sedentary use. These patterns were observed despite similar average daily sedentary time across users of different apps. In other words, whether smartphone use occurred during sedentary or non-sedentary behaviour appears to depend more on what people were doing on their phones than on how sedentary the users were overall. This app-specific patterning is consistent with prior work suggesting that social media use is more strongly associated with sitting than other forms of digital media use, even when overall sedentary time does not differ [31]. The app-specific patterns observed here further suggest that interventions or guidelines that broadly target “screen time” may overlook important differences in behavioural contexts. For example, social media use was predominantly sedentary, whereas other commonly used apps were frequently used during non-sedentary periods. Future studies could examine whether targeting specific types of screen use, rather than overall screen time, yields greater impacts on sedentary behaviour and physical activity.

Together, these findings suggest that smartphone use is, at best, an imprecise momentary proxy for sedentary behaviour at the momentary level. Treating smartphone screen time as interchangeable with sedentary behaviour may obscure meaningful differences in when and how digital media use intersects with movement. This distinction is particularly relevant as adults accumulate substantial daily smartphone use [15]. It also has implications for the interpretation of physical activity and 24-h movement behaviour guidelines that include recommendations to limit screen time. When screen time recommendations are embedded within broader sedentary behaviour guidance, they may be interpreted as implying that reducing screen time will also reduce sedentary time. Our findings suggest that this assumption may not always hold for smartphone use. Guideline messaging may therefore benefit from distinguishing screen time as a digital media exposure from sedentary behaviour as a posture- and energy-expenditure-based behaviour.

Strengths and limitations

Several limitations should be considered when interpreting these findings. At the same time, this study has important methodological strengths, including the use of objective passive smartphone sensing, high-resolution accelerometry, and epoch-level alignment of screen-on time with movement behaviour. This approach moves beyond aggregate person-level associations and allows direct estimation of the momentary co-occurrence of smartphone use and sedentary behaviour. By examining not only the duration of smartphone use but also the behavioural state in which it occurred, this study provides a more nuanced understanding of how digital media use intersects with movement behaviour, which may inform the design of more context-sensitive interventions.

Sedentary behaviour was assessed using wrist-worn accelerometry combined with a validated sedentary behaviour detection algorithm. Although this hybrid approach improves classification compared with wrist accelerometry alone, inactive and active time were still derived using acceleration-based thresholds, which may introduce some misclassification. Thigh-worn devices such as the activPAL provide more direct assessment of posture and remain the criterion standard for sedentary behaviour measurement [32, 33]. Future studies could extend this work by aligning passive smartphone sensing with thigh-worn accelerometry to further refine estimates of momentary co-occurrence.

The study sample consisted of caregivers of preschool-aged children and was relatively small and demographically homogeneous (94% female; narrow age range). Smartphone use patterns and their behavioural context may differ in other populations, including children, adolescents, and adults without caregiving responsibilities [14]. The sample size also limited the ability to examine demographic moderators of smartphone use context. Future studies should examine whether these patterns generalise to other populations, including children, adolescents, adults without caregiving responsibilities, and more socio-demographically diverse samples, as well as to other screen modalities such as tablets and television.

This study focused on behavioural co-occurrence rather than temporal sequencing or causality. Although smartphone use and movement behaviour were aligned at the epoch level, the analyses do not address whether smartphone use precedes, follows, or influences changes in activity or sedentary behaviour.

A further limitation is that only app-usage events were used in this study, and other forms of smartphone exposure are not be captured, including brief lock-screen checks, notification previews, or other interactions that do not generate logged app-usage events. Conversely, logged screen activity does not necessarily indicate active viewing or engagement by the participant. Smartphone use in this study should therefore be interpreted as screen-active app-usage time detected by Chronicle, rather than a comprehensive measure of all smartphone interactions or attention. In addition, because sensing was conducted on the caregiver’s own smartphone, some logged use may reflect brief use by another person, such as a child or other household member.

Conclusion

For most participants, smartphone use occurred primarily during sedentary time, yet a meaningful proportion, typically around one-quarter, occurred during non-sedentary behaviour, with wide variation across individuals, days, and apps. These findings demonstrate that smartphone screen time is likely a poor momentary proxy for sedentary behaviour in adults, and that treating these constructs as interchangeable may obscure meaningful behavioural patterns. Therefore, when research questions concern sedentary posture, low energy expenditure, or related health mechanisms, smartphone screen time and sedentary behaviour should be measured separately or aligned at the momentary level. More precise measurement and context-aware approaches are needed to understand how digital device use intersects with movement behaviour and to inform interventions targeting sedentary behaviour and physical activity in the modern digital environment. Future studies should reconsider the use of smartphone screen time as a momentary proxy for sedentary behaviour and prioritize context-aware measurement approaches. Particularly as traditional sedentary screen behaviours such as television viewing are increasingly supplemented or replaced by mobile device use.

Supplementary Information

Below is the link to the electronic supplementary material.

Author contributions

JC and BA conceptualized the study. JC conducted the formal analysis. JC, BA, OF, and RGW prepared the original draft. TA and MR managed project administration. All authors (JC, OF, KB, SN, RG, JSR, TA, MR, JW, ZZ, BH, AH, KK, SB, ELA, RGW, MWB, BA) contributed to review and editing of the manuscript.

Funding

Supported by the National Institute of General Medical Sciences of the National Institutes of Health under award numbers P20GM130420 and T32GM081740-15, and by the Eunice Kennedy Shriver National Institute of Child Health and Human Development of the National Institutes of Health under award number R01HD112311.

Availability of data and materials

The datasets used and/or analysed during the current study are available from the corresponding author on reasonable request.

Declarations

Ethics approval and consent to participate

Both included studies were approved by the University of South Carolina Institutional Review Board (pilot: Pro00092634, August 2020; cohort: Pro00133829, December 2023). Participants provided informed consent prior to participating.

Consent for publication

Not applicable.

Competing interests

The authors declare that they have no competing interests.