Neuroplastic changes in functional connectivity associated with problematic smartphone use in a data-driven rs-fMRI study

Summary

Previous studies revealed altered functional connections in problematic smartphone use (PSU), but most of the studies used self-reported data to assess PSU. Therefore, in the current study, we investigated neuroplastic changes in functional connectivity (FC) associated with PSU and objectively measured smartphone screen time (OMST) using data-driven multivariate pattern analysis. Based on our results, OMST had no direct effect on FC, but the FC between the somatomotor and the frontoparietal networks was negatively, while the connection between the somatomotor and limbic networks was positively associated with PSU. In addition, the FC between the default mode network and the limbic system was negatively, while connectivity between the default mode and the somatomotor networks was also negatively associated with PSU. In light of our findings, PSU may impair higher level brain networks, although future studies are highly needed to determine which specific objectively measured screen time activities might affect brain functions.

Graphical abstract

Highlights

• •Objectively measured screen time (OMST) and problematic smartphone use (PSU) were assessed
• •Data-driven multivariate pattern analysis was used to assess functional brain networks
• •OMST was not associated with brain connectivity
• •PSU showed correlations with the frontoparietal, the limbic, and the default mode networks

Health sciences; Medicine; Medical specialty; Internal medicine; Neurology

Introduction

Smartphones are present in all aspects of our lives, as we use them for socializing, working, shopping, and recreation. This significant impact is also well illustrated by the fact that in 2023 almost 70 percent of the world’s population were active smartphone users, and this has been on the rise ever since.¹ Nevertheless, the excessive and uncontrolled use of smartphones could lead to the development of a new form of behavioral disorder, such as problematic smartphone use (PSU),² which is characterized as an addictive disorder.^2,3 Several behavioral and magnetic resonance imaging (MRI) studies highlighted PSU as a potential threat to mental health, as these investigations found altered social-cognitive,⁴ emotion-regulatory⁵ and cognitive functions (e.g., altered executive and attentional control, decision-making, and language processing).^6,7

Resting-state functional MRI (rs-fMRI) is one of the most widely used techniques to advance our understanding of the brain’s complex functional organization and neurophysiological processes.⁸ This method identifies functional connections at rest between brain regions, so that it reflects the functional integrative networks of the brain from different brain areas.⁹ Several networks were found to be altered in PSU. Functional connectivity (FC) between the salience network (SN)—which is responsible for the allocation of attentional and cognitive resources—and the self-referential default mode network (DMN) was shown to be related to PSU.^10,11 Further findings include reduced FCs between the behavioral-regulatory central executive network and the SN¹⁰ and altered connectivity between the regions of the default mode and the somatomotor networks.¹¹ Depressive symptoms were associated with altered parahippocampal-middle temporal FC,¹² while impaired frontostriatal^13,14 and increased nucleus accumbens-cingulate gyrus FCs¹³ were associated with enhanced reward sensitivity in PSU. Finally, altered within-network connectivity between the language and frontoparietal networks (FPNs)¹⁵ was also revealed in PSU.

These findings suggest the diverse functional neural underpinnings of PSU, but the heterogeneity in data acquisition and analysis methods prevents us from comparing and generalizing these findings.^16,17 Previous PSU investigations mainly implemented model-based rs-fMRI analyses using a priori seeds,^{10,12,13,14,18} and only one preceding investigation used model-free independent component analysis.¹⁵

Additionally, the use of different self-reported questionnaires to assess PSU severity also makes it difficult to compare rs-fMRI findings,¹⁹ as the features of the assessment tools and the arbitrariness of cut-off scores might have a significant impact on the prevalence and the nature of PSU.^20,21 Different inventories often focus on distinct aspects of PSU, such as the key symptoms, or the impact of PSU on overall daily functioning.^22,23 Moreover, some rs-fMRI studies investigated general smartphone usage time based on self-reported smartphone screen time.^14,24 This could also be problematic, as participants are generally unable to accurately estimate their screen time activity,^25,26 leading to inaccuracies in the PSU-associated fMRI findings and the conclusions drawn from them. As suggested by Montag and Becker in their review,¹⁹ it would be fruitful to add another perspective to the neuroscientific investigation of PSU by conducting more studies that examine associations between objectively tracked smartphone data and brain functions. However, to date, no such rs-fMRI studies have been conducted.

Accordingly, the primary objective of the present study was to overcome the methodological inaccuracies described previously. In line with Montag and Becker’s recommendation,¹⁹ we used objectively measured screen time data to enhance our understanding of the effect of smartphone usage on FC. In addition, PSU severity was assessed by a composite score from two different self-reported questionnaires covering the symptom characteristics (e.g., craving and withdrawal) and PSU’s impact on general daily functioning. Finally, to evaluate our rs-fMRI data, we used data-driven, model-free analysis (multi-voxel pattern analysis, MVPA) instead of model-based analysis methods. This method is capable of searching for reproducible spatial patterns of neural fluctuations and decoding the neural background of different conditions such as PSU.²⁷ Data-driven rs-fMRI approaches provide an important complement to hypothesis-driven seed-based analyses because they do not require a priori specification of seed regions or network boundaries. This approach is particularly advantageous because the neural mechanisms of PSU are not yet well established, and previous studies have used heterogeneous assessment methods. Since data-driven methods (including MVPA) can capture whole-brain connectivity patterns and identify multiple resting-state networks simultaneously, they may reveal unexpected or spatially extended network alterations that could be missed by analyses restricted to predefined ROIs. Nevertheless, data-driven findings require cautious interpretation and independent validation, as results may depend on choices related to dimensionality reduction and other analytic parameters.

Results

Descriptive statistics

Descriptives of demographic variables and questionnaires are shown in Table 1. As expected, PSU (that is a composite score) showed a strong positive association with the original tools (Brief Addiction to Smartphone Scale [BASS] and Hungarian Smartphone Deprivation Inventory [HSDI]); the Beck Depression Inventory (BDI) was also positively correlated with all questionnaire scores, while the OMST (Figure 1) showed no associations with the other variables.

Table 1.

Demographics and questionnaire data of the sample

Variable	n	Mean	SD	1	2	3	4
Sex (males)	63 (23)
Age		24.3	3.8
1. Beck Depression Inventory		6.3	5.4
2. Brief addiction to smartphone scale		17.8	4.9	0.259∗
3. Smartphone deprivation inventory		16.4	8.2	0.481∗∗	0.357∗∗
4. Problematic smartphone use		17.1	5.6	0.456∗∗	0.684∗∗	0.899∗∗
5. Objectively measured screen time (min/day)		266.9	119.8	0.077	0.159	0.001	0.036

Spearman’s rank correlation coefficients (rho) are presented.

∗p < 0.05, ∗∗p < 0.01.

Figure 1.

Objective measurements of weekly smartphone screen time

An example of the Android Digital Wellbeing (A) and iOS Screen Time (B) applications which objectively measure total weekly smartphone screen time usage (n = 63)

Both the Android and iOS applications provide information on the total number of hours and minutes spent on smartphones over the last 7 days. Users’ total weekly screen time data are framed in red.

The relationship between PSU and functional brain connectivity

The MVPA analysis revealed four significant clusters, including the right hippocampus and amygdala (MVPA1), the left hippocampus and amygdala (MVPA2), the right precentral gyrus (MVPA3), and the left postcentral gyrus (MVPA4; see Figure 2 and Table 2). The areas in MVPA1 and MVPA2 are considered parts of the limbic network, while the brain regions in MVPA3 and MVPA4 were identified as parts of the somatomotor network. All clusters were associated with PSU.

Figure 2.

Functional connectivity patterns that are associated with problematic smartphone use and survived Bonferroni correction (n = 63)

Multi-voxel pattern analysis seed regions are shown in green. Red-yellow color bars indicate positive associations, while blue-light blue color bars indicate negative associations between MVPA seed regions and resting brain voxels. Axial slices are shown in radiological convention. p values were FDR-corrected at cluster level p < 0.001 for MVPA analyses and p < 0.01 for the seed-based analyses.

MVPA, multi-variate pattern analysis; FCs, functional connections.

Table 2.

Seed regions derived from the multi-variate pattern analysis that are associated with problematic smartphone use

Cluster name	Location (AAL)	Network	k	p (FDR-corrected)	Cluster coordinates (MNI)			Network function
MVPA1	hippocampus/amygdala right	Limbic	130	<0.000001	+26	−14	−20	affective and motivational processing, reward valuation, emotion-cognition interaction, and memory-related processing.28,29
MVPA2	hippocampus/amygdala left	Limbic	59	0.000243	−22	−16	−20
MVPA3	precentral gyrus right	SMN	139	<0.000001	+04	−20	+76	sensorimotor integration, somatosensory processing, and motor planning/execution.28
MVPA4	postcentral gyrus left	SMN	46	0.000081	−54	−10	+28

MVPA, multi-variate pattern analysis; AAL atlas, Automated Anatomical Labeling; k, cluster size in voxels; MNI, Montreal Neurological Institute; SMN, somatomotor network.

p-value (FDR-corrected at cluster level p < 0.001, adjusted for Beck Depression Inventory and objectively measured screen time).

When using MVPA1 as a seed, the seed-to-voxel analysis showed a positive association with PSU at the left and right postcentral gyri, bilateral precentral gyri, and the right superior parietal lobule—all of which are part of the somatomotor network. Moreover, a negative association was found at the left angular and supramarginal gyrus, areas considered part of the DMN (Table 3).

Table 3.

Functional connectivity results that are associated with problematic smartphone use for seed regions (reported in Table 2)

Peak coordinates (MNI)			Location (AAL)	Network	k	p (FWE-corrected)		Network function
Seed: MVPA1 Hippocampus/Amygdala right (Limbic)
−38	−34	+64	Postcentral gyrus left Precentral gyrus bilat	SMN	735	<0.000001a	+	Sensorimotor integration, somatosensory processing, and motor planning/execution.28
+34	−36	+70	Postcentral gyrus right Superior Parietal Lobule right	SMN	635	<0.000001a	+
−46	−54	+42	Angular gyrus left Supramarginal gyrus post div left	DMN	98	0.000066a	–	Self-referential and internally directed cognition, autobiographical memory, future simulation, and mind-wandering.30
+40	−66	−42	Cerebellum Crus2 right	Cerebellar	67	0.000902a	–	Motor coordination and timing, with additional contributions to cognitive and affective regulation through cerebro-cerebellar circuits.31
Seed: MVPA2 Hippocampus/Amygdala left (Limbic)
+32	−40	+72	Postcentral gyrus right Superior Parietal Lobule right	SMN	123	0.000006a	+	Sensorimotor integration, somatosensory processing, and motor planning/execution.28
−56	−08	+14	Postcentral gyrus left Central Opercular cortex left Precentral Gyrus left	SMN	76	0.000288a	+
−44	−64	+06	Lateral Occipital Cortex inf div left Middle Temporal Gyrus left (temporooccipital part)	Visual	53	0.002569	+	Basic and higher-order visual perception, visuospatial processing, and object/scene analysis.28,32
Seed: MVPA3 Precentral gyrus right (SMN)
+64	−42	+38	Supramarginal Gyrus post div right Supramarginal Gyrus ant div right	FPN	213	<0.000001a	–	Flexible cognitive control, working memory, goal-directed attention, and task-set maintenance.28,33
+22	−12	−22	Hippocampus right Parahippocampal Gyrus ant div right	Limbic	133	0.000008a	+	Affective and motivational processing, reward valuation, emotion–cognition interaction, and memory-related processing.28,29
+00	−20	+28	Cingulate Gyrus post div	FPN	94	0.000134a	–	Flexible cognitive control, working memory, goal-directed attention, and task-set maintenance.28,33
+36	+32	+38	Frontal Pole right Middle Frontal Gyrus right	FPN	83	0.000319a	–
−10	−68	+38	Precuneus Cortex	FPN	70	0.000937a	–
−56	−34	+38	Supramarginal Gyrus ant div left	VAN	57	0.002945	–	Detection of behaviorally relevant or unexpected stimuli and stimulus-driven attentional reorienting.34
+42	−44	−34	Cerebellum Crus1 right Cerebellum 6 right	Cerebellar	57	0.002945	–	Motor coordination and timing, with additional contributions to cognitive and affective regulation through cerebro-cerebellar circuits.31
−44	−30	+64	Postcentral Gyrus left	SMN	57	0.002945	+	Sensorimotor integration, somatosensory processing, and motor planning/execution.28
Seed: MVPA4 Postcentral gyrus left (SMN)
−38	−66	−40	Cerebellum Crus2 left	Cerebellar	64	0.001285a	–	Motor coordination and timing, with additional contributions to cognitive and affective regulation through cerebro-cerebellar circuits.31
−16	+54	+20	Frontal pole left	DMN	54	0.003252	–	Self-referential and internally directed cognition, autobiographical memory, future simulation, and mind-wandering.30

AAL atlas: Automated Anatomical Labeling.

k: cluster size in voxels.

MNI: Montreal Neurological Institute.

MVPA: multi-variate pattern analysis.

SMN: somatomotor network, DMN: default mode network, FPN: frontoparietal network, VAN: ventral attentional network.

p-value (FWE-corrected at cluster level p < 0.01, adjusted for Beck Depression Inventory and objectively measured screen time).

+/− indicate positive and inverse associations between problematic smartphone use and functional connectivity.

^asurvived Bonferroni correction (p < 0.01/4).

When using MVPA2 as a seed area, the seed-to-voxel analysis revealed positive associations with PSU at the left and right postcentral gyri, left central opercular cortex, left precentral gyrus, and the right superior parietal lobule, all of which are considered part of the somatomotor network.

The seed-to-voxel analysis with MVPA3 as the seed showed one positive relationship with PSU at the right hippocampus and right parahippocampal gyrus, which are part of the limbic network. Moreover, negative associations were found at the right supramarginal gyrus, posterior cingulate gyrus, right frontal pole, right middle frontal gyrus, and the precuneus, all considered parts of the FPN.

Using MVPA4 as a seed revealed a negative association in the cerebellum. Other brain areas did not survive Bonferroni correction.

OMST showed no significant relationship with any of the MVPA seeds.

Discussion

In this study, we investigated neuroplastic changes in FC associated with PSU and OMST. In the absence of consistent, replicated a priori knowledge, we decided to use a data-driven multivariate pattern analysis method to explore potential seed areas and a seed-to-voxel analysis to reveal FC alterations that are correlated with PSU and OMST. Based on our results, OMST is not related to FC; however, PSU showed correlations with several between-network FC patterns. In general, we found that the connection between the somatomotor and the FPN was negatively, while the connection between the somatomotor and the limbic network was positively associated with PSU. Moreover, the connectivity between the DMN and the limbic network was negatively associated with PSU.

Generally, enhanced connectivity—reflected in increased shared variance of BOLD signal fluctuations—between functional networks may suggest improved communication between these networks. This, in turn, could support the integration of distinct psychological and neurobiological constructs into a cohesive system essential for everyday functioning.³⁵ Conversely, a loss of function is thought to be reflected in a loss of cofluctuations in dedicated brain systems. With these considerations in mind, there may be two major lines of argument for assigning physiological significance to the changes in FC, and cortical plasticity processes play a central role in both. Firstly, a well-replicated finding in task activation studies is that problematic smartphone users (regardless of the measurement method) show an altered recruitment of brain areas that are considered to be parts of the DMN,^{36,37,38,39,40} the somatomotor network,^39,40,41 the limbic system,^4,39 and the FPN.^36,42,43 These studies revealed activation patterns among functionally related brain regions. In line with the “coactivation” theory, our findings may represent residual signatures of such coactivations, with FC capturing the statistical history of coordinated activity within cortical circuits. According to Hebbian plasticity theory,⁴⁴ stable changes in synaptic strength are driven by simultaneous pre- and postsynaptic activity; thus, frequent coactivation of brain regions may strengthen their structural—and consequently functional—connections. Conversely, reduced between-network FC observed in problematic smartphone users may indicate diminished coactivation between specific brain areas under this framework.

The second line of interpretation draws on the understanding that increased FC may reflect greater neural effort in cortical processing as a response to structural damage. According to the “structural damage” theory, increased FC can serve as a compensatory mechanism that limits the consequences of structural neural decline and helps maintain a viable level of computational capacity. Several studies have investigated structural changes in PSU and have reported structural decreases in the volume of gray matter,^{45,46,47,48,49} brainstem substructures,⁵⁰ and subcortical areas including the limbic system,^47,51,52 as well as decreases in the complexity of cortical folding.⁵³ According to this theory, the positive association between PSU and FC connecting the limbic and the somatomotor network can be interpreted as compensation for structural decline to maintain mental and biological functions. The involvement of the limbic system is expected, since it was previously shown to be associated with PSU; however, the involvement of the somatomotor network across multiple dimensions might seem surprising. Indeed, we might claim that this network is a novel functional target in the context of identifying potential imaging biomarkers of the tendency toward behavioral dysfunctions. Although known as a key hub⁵⁴ of the functional brain, the potential role of the somatomotor network in mental disorders (such as depression and anxiety) is only beginning to be recognized.^55,56 It is involved in bodily self-consciousness and interoception, processes that are crucial for several neuropsychological functions.⁵⁷ At a coarse level, our results underscore the importance of subjective valuation and bodily states in the pathophysiology of mental dysfunctions such as behavioral addictions.

The FC between the somatomotor network and the FPN was found to be negatively associated with PSU, suggesting disrupted communication between these functional networks in problematic smartphone users and indicating impaired sensorimotor integration in PSU. This reduced connectivity could impair the coordination between thoughts and movement,⁵⁸ leading to impaired motor inhibition. Motor inhibition is usually assessed by cognitive tasks (e.g., Stroop, Go/No-go, and Simon tasks), and it has been found to be decreased in PSU⁴² and in other forms of internet-related addictions, such as general internet addiction,⁵⁹ social media addiction,⁶⁰ and binge watching.⁶¹ Moreover, motor inhibition is a major form of self-regulation, the deficits of which have also been identified as key features in PSU.^62,63 The decreased FC between the somatomotor network and FPN suggests a disturbance in attention regulation and psychomotor activity, which could also be linked to the dysregulation observed in motor behavior.⁶⁴

FC between the DMN and the limbic system was negatively correlated with PSU. This connectivity suggests a role for the DMN in emotional processing, although there is still limited consensus about its exact function during affective processes. A recent model proposes that the DMN plays a constitutive role in generating discrete emotional experiences (e.g., fear or anger), with its nodes supporting the perception of specific physiological sensations as expressions of broader emotional categories. This process—often referred to as “conceptualization” or “meaning making”—occurs at the psychological level of analysis.⁶⁵ In addition, loss of DMN integrity may impair the experience of emotions.⁶⁵ From this perspective, our results were not surprising at all. PSU (especially social media engagement) has previously been linked to higher levels of stress, anxiety, and depression.⁶⁶ Reduced emotional regulation has also been associated with smartphone use,⁶⁷ and in our recent paper, we revealed decreased emotional face expression recognition in PSU, with corticolimbic areas playing a crucial role in this association.⁴

The present finding that PSU was negatively associated with FC between the DMN and the limbic network is broadly consistent with previous rs-fMRI studies suggesting that PSU is related to altered interactions among large-scale networks involved in internally directed cognition, salience processing, affective regulation, and reward-related processing. Previous studies have reported altered connectivity involving the SN, DMN, attentional networks, and frontostriatal circuits in PSU, although the exact direction and localization of effects have varied across studies. For example, Ahn et al.¹⁰ found altered FC between the salience network and DMN and between the salience and central executive networks, while Kwon et al. reported altered dACC connectivity with attentional network and the DMN in PSU.⁶⁸ Other work has emphasized reduced frontostriatal connectivity related to cognitive control and reward processing in excessive smartphone use.¹³ Our finding extends this literature by suggesting that PSU may also be associated with weaker integration between DMN-related self-referential/internal mentation processes and limbic affective-reward systems. However, differences between our results and previous studies may also reflect heterogeneity in PSU assessment, sample characteristics, and analytic strategies. Therefore, rather than pointing to a single localized neural substrate, the available evidence suggests that PSU is associated with distributed alterations in large-scale functional brain networks.

Lastly, according to our results, there was no association between OMST and FC. The lack of results may seem unexpected at first sight, since previous studies on screen time revealed several altered FCs, including the connections between the visual word form area and Brodmann areas, such as 13, 20, 24, 25, 39, and 47,⁶⁹ the connections between the somatomotor network and the DMN, attention network, and auditory network,⁷⁰ and the connections between several subcortical areas (such as the amygdala, putamen, and the accumbens).⁷¹ However, one important difference between the previous and the present study must be noted: these studies all used self-report methods to estimate screen time, whereas we measured the real usage time through applications (Digital Wellbeing app on Android and Screen Time app on iOS). This difference is highly important, since previous studies have highlighted that self-reports of total smartphone use have only poor or moderate agreement with objective measures.^72,73 Moreover, OMST may be a relatively crude indicator of smartphone-related problems, as it captures the amount of use but not the psychological quality, context, or motivational aspects of use. In contrast, PSU scores reflect maladaptive features, such as impaired control, preoccupation, and withdrawal-like experiences, which may be more directly related to brain connectivity alterations than total usage duration alone. Furthermore, OMST likely includes both problematic and non-problematic activities, such as work, communication, navigation, or administrative tasks, which may dilute its association with neural outcomes. Obviously, based on the lack of results in this study, we cannot declare with certainty that screen time per se has no influence on functional neural connections. Future studies with enhanced sample sizes and a focus on specific areas of use—such as social media, gaming, and shopping—are highly needed.

In conclusion, we investigated neuroplastic changes in FC associated with PSU and OMST. OMST had no effect on FC; however PSU showed associations with several network alterations including the somatomotor, the frontoparietal, the limbic, and the DMNs. Our results confirm that PSU might significantly alter brain functions, but future studies are needed to reveal which specific areas of smartphone use are behind these changes.

Limitations of the study

This study has some limitations. The first limitation is the correlational nature of the study. Therefore, we cannot determine the causality between the detected brain alterations and PSU. Longitudinal studies are highly needed to overcome this problem. In addition, it would be beneficial to validate our results on specific cohorts, such as people with smartphone addiction. The second limitation is the relatively small sample size for a data-driven rs-fMRI study, particularly when examining complex behavioral correlations. This may have limited our ability to detect smaller effect sizes. The sample size of the present study should be considered modest for whole-brain MVPA analyses. Although MVPA is well suited for detecting distributed multivariate connectivity patterns, brain-wide brain–behavior association analyses generally require larger samples to obtain stable and replicable effect estimates. Therefore, the present analysis may have been primarily sensitive to medium-to-large PSU-related connectivity-pattern effects, whereas smaller or more spatially distributed associations may have remained undetected. Future studies with larger independent samples are needed to replicate the present results and to assess the generalizability of the identified connectivity patterns. The third limitation is the lack of information on specific domains of smartphone use when assessing OMST. Unfortunately, during the pilot phase of the study, we found substantial differences in application categorization methods across operating systems, and manual categorization was considered highly subjective and unreliable. Future studies focusing on specific categories of smartphone use would therefore be important and may reveal category-specific brain alterations.

Resource availability

Lead contact

Requests for further information and resources should be directed to and will be fulfilled by the lead contact, Gergely Darnai (darnai.gergely@pte.hu).

Materials availability

This study did not generate new unique reagents.

Data and code availability

• •Data reported in this paper will be shared by the lead contact upon request.
• •This paper does not report original code.
• •Any additional information required to reanalyze the data reported in this paper are available from the lead contact upon request.

Acknowledgments

This study was supported by the Hungarian Brain Research Program 2017-1.2.1-NKP−2017-00002 government-based fund. Our research was partly financed by the Higher Education Institutional Excellence Program of the Ministry of Human Capacities Hungary, within the framework of the 5th thematic program of the University of Pécs, Hungary (20765/3/2018/FEKUSTRAT); the National Research, Development and Innovation Fund of Hungary’s Thematic Excellence Programs (TKP2021-EGA-16 and TKP2021-EGA-13); and the National Laboratory of Translational Neuroscience, within the program of “Adult Nervous System Disorders-RRF-2.3.1-21-2022-00011” framework. Á.A. and S.A.N. were supported by the University Research Scholarship Program (EKÖP-25-4-I-PTE-434 and EKÖP-25-4-II-PTE-536) by National Research, Development and Innovation Fund of Hungary. G.D. and S.A.N. were supported by the János Bolyai Research Scholarship of the Hungarian Academy of Sciences. J.J. was supported by National Research, Development and Innovation Fund of Hungary (NKFIH-K142479).

Author contributions

Á.A. and G.D. were responsible for writing the manuscript and supervising the whole process. S.A.N., G.P., and G.O. were responsible for MRI measurements and the statistical analysis. G.D., A.T.S., H.A.A., and B.D. were responsible for the manuscript review and critique. Á.A. and G.D. contributed equally to the conception, organization, and execution of the research project.

Declaration of interests

The authors declare no competing interests.

Declaration of generative AI and AI-assisted technologies in the manuscript preparation process

During the preparation of this work, the authors used ChatGPT 5.6 for grammatical correction and language editing. After using this tool, the authors reviewed and edited the content and take full responsibility for the content of the publication.

STAR★Methods

Key resources table

REAGENT or RESOURCE	SOURCE	IDENTIFIER
Software and algorithms
Presentation	Neurobehavioral Systems Inc.	Version 17.1
CONN functional connectivity toolbox	Whitfield-Gabrieli and Nieto-Castanon74	Version 21.a; Whitfield-Gabrieli and Nieto-Castanon74
IBM SPSS Statistics	IBM Corp.	Version 28
CONN project file	This paper	Supplementary file
Android Digital Wellbeing	Google	Android Pie or later
iOS Screen Time	Apple	iOS 12 or later
Other
Edinburgh Handedness Inventory	Oldfield75	Oldfield75
Brief Addiction to Smartphone Scale BASS	Csibi et al.22	Csibi et al.22
Hungarian Smartphone Deprivation Inventory HSDI	Csibi et al.22	Csibi et al.22
Beck Depression Inventory BDI	Beck et al.76	Beck et al.76

Experimental model and study participant details

Sixty-three healthy young adults participated in the study, including 23 males (based on self-report), with a mean age of 24.3 ± 3.8 years (age range: 18–29). All participants were right-handed according to the Edinburgh Handedness Inventory.⁷⁵ Participants with chronic illnesses or neurological or psychiatric disorders were not included. Participants were instructed to abstain from alcohol-containing substances for 24 h before the experiment. As participants were not assigned to separate groups or experimental conditions, randomization was not applicable to the present study.

All participants received information about the study and provided written informed consent before participation. Each participant received 5,000 HUF (approximately 14 USD) as compensation for their time and effort. The study was approved by the National Medical Research Council (registration number NNGYK/36085-6/2025).

Method details

Questionnaires

Since there are no widely accepted diagnostic criteria or cutoff points for screen-related addictions (such as PSU), we used questionnaire scores as continuous variables.

To assess problematic smartphone use, the Brief Addiction to Smartphone Scale (BASS)²² was applied, which is a self-report questionnaire, created to screen for the risks of smartphone application-based addiction. It contains six items that refer to the core criteria for addictions. The items are rated using a 6-point Likert-type scale ranging from 1 (strongly disagree) to 6 (strongly agree). Higher scores on the BASS indicate a greater risk of developing addiction to smartphone application use.

The Hungarian Smartphone Deprivation Inventory (HSDI)²² was also used in this study for assessing problematic smartphone use. This scale focuses on recognizing persistent emotional states and assessing withdrawal symptoms that occur after stopping the behavior. The symptoms of the nine emotional states are as follows: emotional-psychological symptoms (agitation, restlessness, nervousness, discouragement, sadness) and psychophysiological complaints (insomnia, tension, stiffness, exhaustion). The 9 items are rated on a seven-point scale, ranging from 1 (not characteristic at all) to 7 (highly characteristic).

To make the interpretation of our results easier, a data-driven composite score was calculated based on the total scores of the two questionnaires. Principal Component Analysis with a fixed one-factor solution was conducted using the regression method to derive factor scores based on the total scores from the BASS and HSDI. The principal component explained 70.7% of the variance and had high loadings on both measures (0.841 for both scales). The regression-based factor scores were saved and used as a composite measure of PSU in subsequent analyses.

The Beck Depression Inventory (BDI) was also administered, which is a widely used self-report questionnaire designed to assess the severity of depressive symptoms.⁷⁶ Participants rate their experiences on a four-point scale (0–3), with higher scores indicating greater levels of depression. The BDI has demonstrated strong psychometric properties, making it a reliable tool in both clinical and research settings.

Screen time measurement

As previous studies on resting-state brain functions only investigated the subjectively estimated smartphone screen time (participants were merely asked to estimate the time of use subjectively and retrospectively (e.g.,: “How much time do you usually spend using your smartphone each day?”), in this study, we focused instead on objectively measured screen time (OMST).

Regarding our OMST investigation, recent versions of Android (since August 2018, Android Pie or later) and iOS (since September 2018, iOS 12 or later) smartphone operating systems allowed users to access their detailed smartphone usage time through software-related applications (Digital Wellbeing app on Android; Screen Time app on iOS). Both the Digital Wellbeing and Screen Time apps provide information on the total number of hours and minutes spent on smartphones over the last 7 days (see Figure 1 for details). In contrast to other third-party screen time trackers, the Android- and iOS-based applications have a major advantage: they record activity whenever the smartphone is unlocked, even in airplane mode. To access and create weekly OMST reports, users need to enable the feature on either operating system.

Approximately one week before the experiment, a member of the research team called the participant to

• 1.ask the participant to download the Android Digital Wellbeing or iOS Screen Time app if it was not previously installed.
• 2.ask the participant to start collecting data if the app was installed on the phone but not active.
• 3.ask the participant to check that data collection was working properly if the app was already active.

On the day of the MRI measurement, we asked participants to provide their total screen time for the past week based on the Android Digital Wellbeing or iOS Screen Time report. Based on a previous behavioral study,⁷⁷ we used average daily OMST data in our analyses, calculated by dividing the total weekly usage by seven.

Resting-state FMRI paradigm

In the scanner, the participants were instructed to rest and keep their eyes open for 10 min, while focusing on the white fixation cross overlaid on a black background. Participants were asked to stay awake and try not to think about anything. The paradigm was implemented using the Presentation software (Version 17.1, Neurobehavioral Systems Inc., Berkeley, CA, USA) and the stimuli were presented using an MRI-compatible LCD screen (BOLDscreen 24 LCD for fMRI, Cambridge Research Systems Ltd., Rochester, United Kingdom).

MRI data acquisition

Participants were imaged using the same 3T MRI scanner (MAGNETOM Prismafit, Siemens Healthcare, Erlangen, Germany) with a 64-channel Head/Neck coil.

Functional images were obtained using a 2D single-shot gradient-echo echo planar imaging (EPI) sequence with the following parameters: repetition time (TR)/echo time (TE) = 3000/30 ms; flip angle (FA) = 83°; field of view (FOV) = 210 × 210 mm2; 70 × 70 matrix; 44 axial slices with a thickness of 3 mm; 2040 Hz/pixel receiver bandwidth and interleaved slice order.

After the fMRI measurement, a field mapping sequence (TR/TE1/TE2 = 480/4.92/7.38 ms; FA = 60°; 44 axial slices; slice thickness = 3 mm; FOV = 210 × 210 mm2; 70 × 70 matrix; receiver bandwidth = 290 Hz/pixel) with the same orientation and adjustment parameters as the fMRI scan was acquired.

Anatomical images were acquired using a T1-weighted 3D MPRAGE sequence (TR/TE/inversion time = 2530/3.4/1100 ms; FA = 7°; FOV = 256 × 256 mm2; 256 × 256 matrix; slice thickness = 1 mm; 176 sagittal slices; 200 Hz/pixel receiver bandwidth).

MRI data preprocessing

The images were processed using CONN functional connectivity toolbox, version 21.a.⁷⁴ Image preprocessing including the following steps: functional realignment and unwarping (by integrating fieldmaps for susceptibility distortion correction), slice-timing correction, structural segmentation and normalization, functional indirect normalization, outlier detection, and smoothing (with a 6 mm full width at half maximum Gaussian kernel). Potential outlier scans were identified using the Artifact Detection Tools scrubbing procedure (implemented in the CONN toolbox): acquisitions with framewise displacement above 0.9 mm or global BOLD signal changes above 5 standard deviations were flagged as outliers. Since the number of volumes flagged as outliers divided by the total number of volumes was not greater than 25% in any of the subjects, no subjects were excluded. After segmentation and normalization, all segmented tissues were visually checked. The CONN project file (S1) is attached to the manuscript as Supplementary Material.

Outlying volumes (7 principal components, PCs) were used for scrubbing during the denoising step. In addition, the following confounding effects were also regressed out from the BOLD signal: 5 PCs from white matter, 5 PCs from cerebrospinal fluid, 24 PCs from subject-motion parameters and 2 PCs for effects of rest with linear detrending. To minimize the effects of head motion, physiological noise, and other noise sources, a band-pass filter was applied and set to 0.008–0.09 Hz.

Multi-voxel pattern analysis (MVPA)

MVPA was used to estimate whole-brain resting-state FC. This data-driven approach computes the FC map between each individual seed voxel and the rest of the brain, making it possible to determine whether the FC pattern for the whole brain was related to PSU or OMST. Because the connection between each voxel and the rest of the brain is analyzed to discover maximal intersubject heterogeneity, dimensionality reduction is needed; the dimensionality was set to 64 in this study (the default setting). During the group-level analysis, the number of eigenpattern scores (which characterizes the FC of each voxel) was five. This number of eigenpattern scores seemed to be sensitive enough to detect large effects of FC, but it is capable of finding more marginal effects as well. The resulting MVPA components were used as seeds in the subsequent seed-to-voxel connectivity analysis.

To identify potential seeds, statistical analyses were performed on each of the 5 components using general linear models (GLMs), where the PSU scores (with BDI and OMST as covariates) and OMST (with age and PSU scores as covariates) were used as regressors of interest, separately. The results were thresholded using a combination of a cluster-forming p < 0.001 voxel-level threshold, and a false discovery rate (FDR)–corrected p < 0.001 cluster-size threshold. Significant clusters were used in seed-to-voxel analysis. When evaluating the results, the 7-network Yeo atlas was used to define which network the significant brain regions belonged.²⁸

Seed-to-voxel analysis

Seed-to-voxel analyses were performed using the MVPA cluster results as seeds. Since the MVPA analysis of OMST revealed no significant results, all clusters used as seeds were derived from PSU-related MVPA analyses. Pearson’s bivariate correlation analyses were performed to calculate the connections from the seeds to other voxels in the whole brain. Similarly to MVPA, GLMs were conducted to examine the associations in FC, where MVPA clusters served as seeds and PSU scores (with BDI and OMST as covariates) were used as the regressor of interest. The results were thresholded using a combination of a cluster-forming p < 0.0001 voxel-level threshold, and an FWE–corrected p < 0.01 cluster-size threshold. The Bonferroni correction was used to adjust the p-value to compensate for the family-wise error rate in multiple comparisons (the initial α value was divided by the number of significant clusters resulting from MVPA, in our case: 0.01/4). Again, the results were interpreted based on the 7-network Yeo atlas.²⁸

Quantification and statistical analysis

Spearman correlation (two-sided) analysis was performed to find the interrelationships between the questionnaire scores and the OMST using IBM SPSS Statistics (Version 28). The alpha-level threshold was set to 0.05 for correlations. p values were FDR-corrected at cluster level p < 0.001 for MVPA analyses and p < 0.01 for the seed-based analyses.