Problematic patterns of social media use increase with spontaneous cortical activity and transdiagnostic mental health symptoms in youth Authors: Nathan M Petro, Lauren K Webert, Kellen M McDonald, Jake J Son, Thomas W Ward, Danielle L Rice, Grace C Ende, Anna T Coutant, Erica L Steiner, Cooper L Livermore, Mikki Schantell, Abraham D Killanin, Giorgia Picci, Tony W Wilson Published: 2026 ABSTRACT AbstractAdolescence is a period of profound biopsychosocial development, with peer interactions playing a critical role. Over the past decade, the dramatic increase in social media access among youth has fundamentally altered such interactions, enabling highly rewarding peer-to-peer social interactions to occur at almost any time and thereby escalate into problematic behaviors. Whether such a seemingly urgent need for interaction through social media constitutes problematic behavior remains unclear, but evidence suggests that social media use can become obsessive and lead to neglect in other areas of life. Unfortunately, the neurobiological underpinnings of such problematic social media use, and their unique relationship with mental health symptoms, are poorly understood. Herein, we assessed mental health symptoms, spontaneous cortical activity based on magnetoencephalography, and self-reported problematic social media use patterns (e.g., “I use social media to reduce bad or upsetting feelings”) in 80 typical youths (8–16 years). Controlling for age, sex, and overall social media use, we found that problematic social media use increased in parallel with spontaneous delta activity in parietal and occipitotemporal cortices, and spontaneous theta in prefrontal cortices. In addition, such social media behaviors were associated with increased internalizing symptoms, inattention/hyperactivity levels, and emotional symptoms, and reduced personal adjustment. In summary, problematic social media behaviors in youth are associated with altered spontaneous cortical activity and may portend increased risk for transdiagnostic mental health symptoms. Importantly, these effects were observed after controlling for overall use of social media, suggesting deleterious effects may be present even with minimal usage. These results should be interpreted with caution given that, among other reasons, social media use was completely self-reported and that these cross-sectional results cannot determine the causal direction among the variables.Keywords: Resting state, MEG, Social media, Addiction, Adolescence, Mental health Introduction Adolescence is a period of profound biopsychosocial development, with the deepening of peer relations and self-appraisal being critical to long-term outcomes (Blakemore and Mills, (2014); Crone and Dahl, (2012); Pfeifer and Allen, (2021)). The dramatic increase in social media access over the past decade has fundamentally changed the context in which adolescents interact (Nesi et al., 2018), with peer relationships historically being confined to specific times and environments (e.g., school, youth groups) now being available with much greater frequency through social media. Indeed, 90% of adolescents use social media multiple times per day, and ∼45% report being online “near constantly” (Anderson and Jiang, 2018, Council on Communications and Media, 2016, Faverio and Sidoti, 2024), with an average of 7.7 h per day outside of school (Nagata, Cortez, et al., 2022). Moreover, such rates have intensified since the COVID-19 pandemic (Environmental Influences on Child Health Outcomes Program, 2023, Nagata et al., 2022). There is an urgent need to understand the impact of social media use on adolescents given their unique sensitivity to developing problematic behaviors and the importance of peer interactions on socioemotional development. Adolescents appear to be particularly susceptible to risky behaviors toward social media (Andreassen et al., 2013, Haug et al., 2015, Nagata et al., 2025), which involve problematic use that interferes with daily life, creates conflict, and perhaps contains elements of withdrawal and relapse (Nagata, Singh, et al., 2022). These patterns of problematic social media use confer increased susceptibility to develop mental health disorders (Xiao et al., 2025), which in adolescents may be driven by the increased reward experienced from peer interactions compared to that in adults and younger children (Blakemore, 2008, Somerville et al., 2010, Steinberg and Morris, 2001). Moreover, adolescents are in general more likely to partake in “risky” and “sensation seeking” behaviors with their peers (Crews et al., 2007, Gardner and Steinberg, 2005), with short-term rewards being valued over long-term negative consequences (Romer, 2010, Steinberg, 2008). Neurobiologically, this is thought to reflect earlier development of limbic regions during adolescence relative to the protracted development of executive regions in prefrontal cortices (Crone and Dahl, 2012, Luna et al., 2015, Somerville et al., 2010, Steinberg, 2007). These neurodevelopmental factors are thought to heighten the perceived reward from peer interactions on social media during adolescence and thereby increase the susceptibility of developing problematic behaviors toward social media (Cheng et al., 2021, Dayan et al., 2010, Derevensky, 2019), including declines in other areas of life such as job/school performance and other relationships (Andreassen, 2015, Christakis and Hale, 2025, Wegmann and Brand, 2019). Importantly, problematic social media behaviors have been linked to multiple mental health disorders (Andreassen et al., 2016; C. T. Barry et al., 2017; Bozzola et al., 2022; Cataldo et al., 2021; Fors and Barch, 2019; Ko et al., 2012; Neophytou et al., 2021), while typical use patterns and overall screentime are known to be weak predictors (Derevensky, 2019, Eirich et al., 2022; C. Huang, 2017; Nagata et al., 2024; Przybylski and Weinstein, 2017; Torre et al., 2020; Valkenburg et al., 2022). In fact, a recent study found that risky and problematic behaviors, and not overall screen time, was related to increased suicidal ideation in adolescence (Xiao et al., 2025). Thus, understanding the neurobiological bases of such social media use patterns is important to curbing the deleterious outcomes of social media use. Along these lines, the current neuroimaging work regarding problematic social media behaviors, although limited (Montag et al., 2023), involves the dysfunction of reward and executive regions (Wadsley and Ihssen, 2023). Receiving feedback on social media activates regions typically associated with social reward, such as the striatum (Meshi et al., 2013, Sherman et al., 2018) and the anterior cingulate and prefrontal cortices (Sherman et al., 2016, Sherman et al., 2018, Wikman et al., 2022). Accordingly, excessive social media use is related to differences in striatal and amygdala volume (He et al., 2017, Montag et al., 2017), as well as frontal cortical volume (D. Lee et al., 2019; Loh and Kanai, 2014; Montag et al., 2018) and function (Moisala et al., 2016). Of note, other types of problematic internet use have been related to differences in frontal cortical thickness (Hong, Kim, et al., 2013), function (Ko et al., 2014; B. Li et al., 2014), and striatal connectivity (Hong, Zalesky, et al., 2013; B. Li et al., 2014; Lin et al., 2015). These fronto-striatal brain circuits are consistent with those implicated in risky “sensation seeking” (Crone and Dahl, 2012, Somerville et al., 2010, Steinberg, 2007) and impulsive behaviors toward immediate and short term rewards (Bechara, 2005, Brand et al., 2014) during adolescence. Most neuroimaging studies of social media use to date have used structural and/or functional MRI to examine differences in neuroanatomy (e.g., volume, cortical thickness) and the activation of neural regions, respectively. However, another important dimension of brain function is broadly referred to as the neural dynamics, which can be measured with magnetoencephalography (MEG). The temporal resolution of MEG, like other electrophysiological recording technologies, is on the order of 1 ms, while the spatial precision ranges from ∼3–5 mm. Such spatiotemporal precision allows the power of cortical rhythms in local neural populations to be quantified, which reflect summated spontaneous neuronal discharges and dendritic current fluctuations (Buzsáki et al., 2012, Mountcastle, 1998). The frequency spectra of these cortical rhythms vary across brain areas, partially reflecting local cortical organization (Bollimunta et al., 2008, Steriade et al., 1990), neurotransmitter distribution (Edden et al., 2009, Flint and Connors, 1996), and thalamic inputs (Contreras et al., 1997, Lemieux and Chen, 2014). Thus, cortical rhythmic activity is a key component of the neural computations necessary for perception, motor control, and cognition (Cohen, 2017, Fries, 2009, Heinrichs-Graham et al., 2018, Schantell et al., 2024, Springer et al., 2023), and may provide important data on the impact of problematic social media behavior on brain function. Studies linking rhythmic cortical activity with problematic internet use are limited (Burleigh et al., 2020, Kashif et al., 2021), but suggest that delta and theta activity are consistently stronger in individuals with an internet gaming disorder compared to controls (Kim et al., 2017, Son et al., 2015). Stronger delta and theta has also been observed in individuals with problematic internet use during behavioral inhibition tasks, including Stroop (Rostami et al., 2024) and Go/No-Go paradigms (Balconi et al., 2017, Tseng et al., 2024), where stronger neural activity related to poorer behavioral inhibition in these individuals (Qi et al., 2022). In addition, delta and theta have been related to similar problematic eating behaviors (De Ridder et al., 2016, Imperatori et al., 2015), substance use (Castelblanco et al., 2024, McDonald et al., 2024, Springer et al., 2023, Weyrich et al., 2023), and a range of psychiatric disorders where executive function is known to be impaired (R. J. Barry et al., 2003; Newson and Thiagarajan, 2019; Perera et al., 2019, Perera et al., 2023). Moreover, delta and theta activity have been linked to prefrontal cortical function (Knyazev, 2007, Knyazev, 2012, Murphy et al., 2009, Picci et al., 2023) which, as discussed above, is thought to be central to dopaminergic pathways involved in habit formation and even addiction. However, no study to date has tied rhythmic cortical activity specifically to problematic social media use behaviors, particularly with respect to the developing brain. In the current study, we examined the neurobiological underpinnings of such social media behaviors in a sample of youth using resting-state MEG. Because surveys pertaining to risky or problematic social media use are limited and not yet validated, we assessed social media use behaviors using a redesigned 9-item survey that closely resembled a well-known and validated questionnaire that was originally designed to assess problematic patterns of substance use (e.g., preoccupation, conflict over use, and interference with daily life). It should be noted that this social media use survey has not been used previously and its ability to identify problematic social media use akin to substance use has not been validated, although that is a future goal. We predicted that problematic social media behaviors would be significantly related to the strength of spontaneous neural activity in the “slow” delta and theta spectral bands, as observed in other problematic internet-related behaviors (Burleigh et al., 2020, Kashif et al., 2021). Further, we hypothesized that these effects would be strongest in the frontal cortices, based on patterns observed in related literature (Knyazev, 2007, Knyazev, 2012, Murphy et al., 2009, Picci et al., 2023). Finally, we predicted that more problematic social media behaviors would be associated with increased transdiagnostic mental health symptoms, as measured by the Behavior Assessment System for Children (third edition; BASC-3; Reynolds and Kamphaus, 2015). Methods Participants The current cross-sectional sample consisted of 92 participants who were previously enrolled in an ongoing longitudinal study. These participants were selected from the entire sample based on their age (i.e., participants younger than eight were excluded) and ownership of a mobile smartphone device. Each participant completed an online social media use questionnaire. Of those 92 participants, four did not complete the MEG resting state recording, six had unusable MEG resting state recordings (i.e., excessive noise), and two were excluded because they indicated that they did not use social media on our questionnaire. Thus, 80 youth (38 females, 8 – 16 years, Mage = 12.23, SD = 2.29) completed all aspects of this sub-study, including resting-state MEG, structural MRI, behavioral and puberty questionnaires, and the social media questionnaire. Prospective participants were excluded from the longitudinal parent study if, at the time of initial enrollment, they had been diagnosed with any psychiatric or neurological condition, had a history of significant head trauma, active drug or alcohol use, the inability to commit to a longitudinal study, current use of medications known to affect CNS function (e.g., antidepressants, antipsychotics, anticonvulsants, etc.), or the presence of any ferrous metal implant, including orthodonture. After a complete description of the study, written informed assent and consent were obtained from the child and the child's parent or legal guardian, respectively. All procedures were approved by the Institutional Review Board (IRB). Mental health symptoms questionnaire Participants completed the self-report version of the BASC-3 (Reynolds and Kamphaus, 2015), which is a validated measure that comprehensively assesses behaviors and emotions related to psychopathology and well-being in children and adolescents (Canivez et al., 2021). This assessment involves participants responding to statements describing how often they exhibit specific behaviors (i.e., 0 = never, 1 = sometimes, 2 = often, or 3 = almost always). We focused our analyses on the composite scores of the BASC-3, including Internalizing Symptoms, Inattention/Hyperactivity, Personal Adjustment, School Problems, and the Emotional Symptoms Index. Given that the current sample is typically developing and that we controlled for age and sex in the statistical models, raw scores, rather than T-scores, of such behavioral composite scores were summed among the subscales of each composite score. Because summed raw scores were used, the subscales of Sensation Seeking and Somatization were excluded from the School Problems and Internalizing Problems composite scores, respectively, given that these subscales are not completed by participants younger than 12 years-old, and thus, would lead to different composite values for younger and older participants. The BASC-3 self-report clinical and adaptive scales have good internal (α =.82 – 84) and test-retest reliability (α =.78 –.83). Social media questionnaire and data reduction Participants completed an online questionnaire, which assessed their social media use patterns and associated emotional responses. Specifically, participants reported on when they started using social media, which applications they typically use, and characteristics of daily use for specific apps (e.g., amount of time used per day, use during bedtime, number of app checks per day). In addition, participants completed a series of questions related to both active and passive social media use. Here, passive use was the sum of all consuming items (e.g., scrolling, viewing posts), and active use was the sum of all content creation items (e.g., communicating, sharing, and creating) from the questionnaire. All items from this survey are listed in the Supplemental Materials. To assess problematic behaviors on social media, participants completed a 9-item survey based on the National Institute on Drug Abuse (NIDA) Modified ASSIST assessment and contained questions pertaining to specific social media use behaviors (e.g. “During the past three weeks, how often have you failed to do what was normally expected of you (e.g., homework, chores, extracurricular activities) because of your use of social media?”). This scale was reliable in our sample (Cronbach’s α =.74). To obtain a single summarized score of the most consistent responses, the responses were submitted to a principal components analysis (PCA). Each participant’s score for the first component was computed using the regression method and used as a metric of social media behaviors for subsequent analyses. Because only the first component was considered, no rotation was applied. MEG data acquisition The MEG recordings were conducted in a two-layer magnetically-shielded VACOSHIELD room. Neuromagnetic responses were sampled continuously at 1 kHz, with an acquisition bandwidth of 0.1–330 Hz, using a MEGIN Triux Neo MEG system with 306 magnetic sensors (Helsinki, Finland). During six-minutes of data acquisition, participants were seated with their eyes-open and maintained gaze on a fixation, while being monitored via real-time audio-visual feeds from inside the shielded room. Structural MRI acquisition, processing, and MEG-MRI coregistration Individual structural MRI data for each participant were acquired using a Siemens Prisma 3 T scanner (Siemens Medical Solutions) with a 32-channel head coil and consisted of one T1-weighted 3D isotropic sequence with the following parameters: TR = 2.4 s, TE = 2.05 ms, flip angle = 8°, FOV = 256 mm, voxel size = 1 × 1×1 mm, slices = 192. All T1-weighted structural MRI data were segmented with the computational anatomy toolbox (CAT12 v12.6; Gaser et al., 2024) within SPM12. Segmented T1 images underwent noise reduction using a spatially-adaptive non-local means denoising filter (Manjón et al., 2010) and a classical Markov Random Field approach (Rajapakse et al., 1997). An affine registration and a local intensity transformation were then applied to the bias corrected images. These preprocessed images were segmented based on an adaptive maximum a posteriori technique (Ashburner and Friston, 2005) and a partial volume estimation with a simplified mixed model of a maximum of two tissue types. Lastly, the segmented images were normalized to MNI template space and imported into Brainstorm for co-registration. Preceding MEG recording, five head position indicator (HPI) coils were attached to the participant’s head and localized, together with three fiducial points and at least 100 scalp surface points, with a 3D digitizer (Fastrak, Polhemus Navigator Sciences, Colchester, VT, USA). Once in the MEG, electrical currents with unique frequency labels (e.g., 322 Hz) were fed into each of the HPI coils, which induced measurable magnetic fields and allowed the position of the coils to be actively tracked relative to the MEG sensors throughout the recording. Since the HPI coil locations were also known in head coordinates, all MEG measurements could be transformed into a common coordinate system. With this coordinate system, participant-wise MEG data were coregistered with the participant’s high-resolution structural T1-weighted MRI data prior to source reconstruction in Brainstorm. MEG data processing Each MEG recording was individually corrected for head motion and underwent noise reduction using the signal space separation method with a temporal extension (tSSS; Taulu and Simola, 2006). The noise-reduced MEG data then underwent standard data preprocessing using the Brainstorm software. This process included a high pass filter of 1 Hz, a low pass filter of 200 Hz and a notch filter of 60 Hz and its harmonics to eliminate line noise. Cardiac and ocular artifacts were identified and removed using signal-space projection (SSP), which was subsequently accounted for during source reconstruction (Ille et al., 2002, Uusitalo and Ilmoniemi, 1997). Data were split into 4 s epochs for the detection and rejection of bad segments of MEG data. Amplitude and gradient metrics for each 4 s epoch were computed using custom Matlab code, and epochs that contained outliers were rejected using thresholds determined per participant, given the impact of individual head position on signal strength (i.e., magnetic field strength falls off exponentially with distance from the current source; Baillet, 2017; Wiesman and Wilson, 2020; Wilson, 2014; Wilson et al., 2016). Following this artifact rejection step, the mean number of accepted epochs was 87.81 (SD = 3.00, min = 77, max = 95). The total number of accepted epochs did not statistically covary as a function of age, sex, the social media behaviors component score, or passive or active social media use (all ps > .05). MEG source imaging and frequency power maps The procedures used to project the MEG data onto the neuroanatomy followed the analysis pipeline outlined by Wiesman and colleagues (2021) and also used in a recent large-scale lifespan study (Rempe et al., 2023). Using an overlapping-spheres model that is not constrained to the cortical surface (M. X. Huang et al., 1999), this approach uses a single sphere for each of the 204 gradiometers (minus any channels previously labeled as bad). A linearly constrained minimum variance (LCMV) beamformer (Van Veen et al., 1997) is then used to spatially filter the data based on the data covariance, computed from the resting-state recordings, and the noise covariance, computed from empty room recordings. Of note, except for removal of cardiac artifacts, the empty room recordings underwent preprocessing steps identical to the participant recordings. The resulting data were then used to estimate the power of cortical activity at each vertex in each of the canonical frequency bands (delta: 2–4 Hz, theta: 5–7 Hz, alpha: 8–12 Hz, beta: 15–29 Hz, low gamma: 30–59 Hz, high gamma: 60–90 Hz; Fig. 1). Using Welch’s method (Welch, 1967), power spectrum densities (PSD) were estimated for each 4 s epoch, with a 1 s sliding hamming window that overlapped at 50%. These raw PSD maps were averaged across epochs per participant, with units corresponding to the Pseudo Neural Activity Index (PNAI), which is a modified version of Van Veen’s Neural Activity Index (Robinson, 2004, Van Veen et al., 1997). Lastly, the norm of the three unconstrained orientations per location were then projected onto a common MNI ICBM152 brain template (Fonov et al., 2009) surface and a 3 mm FWHM smoothing kernel was applied before further statistical analysis. Average spontaneous neural activity spectra. The MEG signals were projected to the cortex using an LCMV beamformer. The frequency plot illustrates the spectra averaged across participants for seven bilateral cortical regions: lateral occipital cortex (green), middle temporal gyrus (cyan), temporal pole (blue), superior parietal cortex (purple), precentral gyrus (gold), rostral middle frontal cortex (pink), and the frontal pole (red). (right): Brain images illustrating the absolute power per vertex averaged across all participants for each canonical frequency band. Data analytic plan We computed descriptive statistics on demographics and all study variables, including social media use. Next, we tested whether the social media behaviors component score was related to the strength of brain activity using a vertex-wise multiple regression implemented in SPM12. Specifically, separately per vertex and frequency band, the spontaneous power was submitted as the outcome in a multiple regression across participants, with the social media behaviors component score as the predictor of interest, and age, sex, passive social media use, active social media use, and time between visits (i.e., days between social media questionnaire completion and MEG recording) as covariates of no interest. Since the primary goal was to identify neural patterns related to problematic social media behaviors, we controlled for passive and active social media use as well as age, which scales strongly with overall social media use. In other words, our analysis strategy directly targeted problematic social media behaviors rather than changes related to development and/or other social media behaviors. To correct for multiple comparisons, threshold free cluster enhancement (TFCE; E = 1, H = 2; 5000 permutations; Smith and Nichols, 2009) was applied to each of the resulting statistical maps. Following permutations, these TFCE maps were assessed with a cluster-wise threshold of pFWE < .05 and a cluster forming threshold of k > 100 vertices. Data from the peak vertices were used to plot and interpret the effect of the component score on neural activity. We next tested whether there was a similar relationship between social media component scores and transdiagnostic symptoms of mental health. This was conducted in the same multiple regression framework implemented in the whole-brain analysis described above, where the social media use component scores served as the predictor of interest and age, sex, passive social media use, active social media use, and time between visits served as covariates of no interest. This model was computed separately with each of the five BASC-3 composite scores (Internalizing Problems, Inattention/hyperactivity, etc.; see Section 2.3) as the outcome variable for each separate model. The p-values were FDR corrected across the five statistical tests. Results Data reduction and descriptive statistics Of the 80 participants included in this substudy, one did not complete the BASC-3 survey and was excluded from the analyses of mental health symptoms. The BASC-3 composite T-scores for the full sample are provided in the Supplemental Materials (Figure S1). Correlations among assessment scores are also reported in the Supplemental Materials (Table S1). The analyses of social media use considered the scores from our questionnaires regarding problematic social media behaviors, active social media use, and passive social media use. All items and scoring strategies are listed in the supplemental materials. For active social media use, participant scores ranged from 14 to 46 out of a maximum possible score of 70 (mean: 24.03 (SD = 7.39)). For passive social media use, participant scores ranged from 4 to 20 out of a maximum possible score of 20 (mean: 11.46 (SD = 4.02)). For the problematic social media use behaviors questionnaire, participant scores ranged from 0 to 43 out of a maximum possible score of 57 (mean: 20.54 (SD = 11.28)). The responses to all items in the problematic social media use questionnaire were submitted to a principal component analysis to obtain a summary metric of such social media use behaviors. The Kaiser-Meyer-Olkin measure of sampling adequacy was.741 and the Bartlett’s Test of Sphericity was significant (χ236 = 161.68, p < .001), indicating sufficient factorability of these items (Hair, 1995). The first component showed an eigenvalue of 3.12 and accounted for 34.5% of variance. The component loadings for each of the nine assessment items are listed in Table 2. Table 1Participant demographics.Sample n = 80Age (in years)8 – 16; 12.23 (2.29)Sex (Male/Female)42/38Race (White/Black or AA/Asian/AIAN/Multiple)67/3/1/0/9Ethnicity (Not HL/HL/Unknown)76/3/1Active Social Media Use14 – 46; 24.03 (7.39)Passive Social Media Use4 – 20; 11.46 (4.02)Internalizing Problems (BASC)2 – 104; 37.41 (24.42)Inattention/Hyperactivity (BASC)0 – 34; 14.44 (8.33)School Problems (BASC)Emotional Symptoms Index (BASC)Personal Adjustment (BASC)0 – 33; 10.98 (7.73)3 – 102; 37.84 (22.84)43 – 100; 75.66 (12.15)Problematic Social Media Behaviors Raw Score0 – 43; 20.54 (11.28)Problematic Social Media Behaviors Component Score-1.77 – 1.99; 0 (1)Note. Ranges, means, and standard deviations are reported for continuous variables. Counts for categorical grouping variables are reported. AA – African American; AIAN – American Indian/Alaska Native; HL – Hispanic or Latino. Participant demographics. Note. Ranges, means, and standard deviations are reported for continuous variables. Counts for categorical grouping variables are reported. AA – African American; AIAN – American Indian/Alaska Native; HL – Hispanic or Latino. Table 2Component loadings of the social media survey items.ItemItem LoadingDuring the past three weeks, how often have you failed to do what was normally expected of you (e.g., homework, chores, extracurricular activities) because of your use of social media?0.697Has a friend or relative or anyone else ever expressed concern about your use of social media?0.679During the past three weeks, how often has your use of social media led to issues at school or issues with your peers and/or family members?0.675In the past three weeks, how often have you had a strong desire or urge to use social media?0.669Have you ever used social media to reduce bad or upsetting feelings (e.g., guilt, sadness, worry, loneliness)?0.610In the past three weeks, how often have you used social media?0.587Have you ever found it difficult or upsetting when you are not able to use your phone or use social media (e.g., at dinner, during school/church/extracurricular activities)?0.585Have you ever tried and failed to control, cut down, or stop using social media?0.380Has an adult ever taken away your social media account privileges?0.230 Component loadings of the social media survey items. Statistical analysis Spontaneous cortical activity Whole-brain vertex-wise regressions revealed a positive relationship between the problematic social media use component score and spontaneous neural activity in the delta and theta spectral bands, when controlling for age, sex, passive social media use, active social media use, and time between visits (Fig. 2). The relationship between the problematic social media behaviors and spontaneous cortical activity. Cortical surface maps (right) display the vertex-wise F-values representing the relationship between the social media component scores and spontaneous delta (A) and theta (B) cortical activity. Green dashed circles denote the vertices containing the strongest effect. Absolute power from these peak vertices is plotted in the scatterplots (left, y-axes) against the component scores (x-axes), after adjustment for all nuisance variables (i.e., age, sex, passive social media use, active social media use, and time between visits). All clusters were significant at pFWE < .05. In the delta range, higher component scores were related to stronger spontaneous delta activity in two bilateral clusters (peak-F1,73 = 9.60, pFWE <.05, β = 0.40, ηp2 <.001), which peaked in the left superior parietal cortex and extended into temporal and frontal cortices, with notable peaks in the temporal poles, left superior temporal cortex, posterior cingulate, and parahippocampal gyrus. While the effect in most regions was clearly bilateral, relationships tended to be stronger in left hemispheric regions. In the theta range, higher component scores were related to stronger spontaneous theta activity in one cluster (peak-F1,73 = 9.46, pFWE <.05, β = 0.44, ηp2 =.06), which peaked in the right medial prefrontal cortex and extended into the anterior cingulate and orbitofrontal cortices. The relationship between problematic social media behaviors and mental health Regression models were used to assess the relationship between the component scores and the BASC-3 composite scores representing transdiagnostic mental health symptoms (Fig. 3), while controlling for age, sex, passive social media use, active social media use, and time between visits. The problematic social media component scores were positively related to Internalizing Problems (F1,72 = 8.05, pFDR <.05, β = 0.42, ηp2 =.06), Inattention/Hyperactivity (F1,72 = 5.48, pFDR <.05, β = 0.35, ηp2 =.08), and Emotional Symptoms (F1,72 = 7.73, pFDR <.05, β = 0.41, ηp2 =.09) composites, and were negatively related to the Personal Adjustment Scale (F1,72 = 7.85, pFDR <.05, β = −0.41, ηp2 =.02). Note that higher scores on the Personal Adjustment Scale indicate better or more positive adjustment. The component scores were not related to School Problems (F1,72 = 1.01, pFDR >.05, β = 0.15, ηp2 =.03). The relationship between problematic social media use and mental health symptoms. Scatterplots illustrate the relationship between the problematic social media use component scores (x-axes) and the BASC-3 composite scores (y-axes). Measures have been adjusted for all predictors and nuisance variables (i.e., age, sex, and time between visits). In each plot, the trendline indicates the linear least-squares line and the surrounding gray error ribbon indicates the 95% confidence interval. All p-values are FDR corrected across the five separate statistical tests. Discussion We examined the neurobiological underpinnings of problematic social media use behaviors during the transition from late childhood to adolescence, a developmental window where the susceptibility to mental health disorders is elevated. More problematic social media use behaviors were related to stronger spontaneous delta activity in parietal, temporal, and frontal cortices, as well as stronger prefrontal theta activity. In addition, more problematic and risky social media behaviors were related to increased transdiagnostic symptoms of mental health disorders. Notably, these effects were present when controlling for overall use of social media, consistent with recent literature showing that problematic social media use, rather than simply using social media, leads to deleterious mental health outcomes (Nagata et al., 2024, Xiao et al., 2025). The interpretation of these results warrants caution, given that the study of social media use is in its early stages and the extent to which problematic social media use resembles well-established deleterious behaviors (e.g., substance use, gambling) is not clear. Overall, stronger spontaneous delta and theta activity were related to more problematic social media use behaviors. These results are consistent with previous EEG literature, which has linked delta and theta activity to problematic internet-related behaviors (Balconi et al., 2017, Burleigh et al., 2020, Kashif et al., 2021, Kim et al., 2017; J. Lee et al., 2014; Qi et al., 2022; Rostami et al., 2024; Son et al., 2015; Tseng et al., 2024), as well as previous studies tying neural activity in these spectral bands to other types of problematic behaviors (Castelblanco et al., 2024, Coutin-Churchman et al., 2003, De Ridder et al., 2016, Imperatori et al., 2015, Kamarajan et al., 2004, McDonald et al., 2024, Pandey et al., 2016, Reid et al., 2003, Springer et al., 2023, Weyrich et al., 2023). Delta and theta activity have together been broadly linked to executive functions supporting impulse control (Kamarajan et al., 2004, Knyazev, 2007, Pandey et al., 2016) and suppression of risky decision making (Massar et al., 2014, Schutter and Van Honk, 2005). Theta activity, in particular, has been related to a variety of higher-order functions (Cavanagh and Frank, 2014, Karakaş, 2020), including social cognition (Güntekin and Başar, 2014, Petro et al., 2025), semantic processing (Rempe et al., 2025), memory (Arif et al., 2024, Hall et al., 2024, Herweg et al., 2020, Killanin et al., 2024b), and visuospatial attention (Meehan et al., 2021, Wiesman et al., 2017, Wiesman et al., 2018, Wiesman and Wilson, 2019). Taken together, these results are consistent with prior EEG and MEG studies which point to poor impulse control as a key factor in risky or problematic behavioral patterns. Future work is needed, however, to establish the link between social media use and impulse control. The relationship between problematic social media use behaviors and theta activity was limited to a relatively focal right frontal cluster that spanned prefrontal, orbitofrontal, and anterior cingulate cortices. The location of this effect is consistent with previous reports linking frontal cortical differences with excessive social media use (D. Lee et al., 2019; Loh and Kanai, 2014; Moisala et al., 2016; Montag et al., 2018) and other problematic internet-related behaviors (Brand et al., 2014, Hong et al., 2013, Hong et al., 2013, Ko et al., 2014; B. Li et al., 2014; Lin et al., 2015). Consistent with the EEG findings noted above, these frontal cortex effects may at least partially reflect differences in impulse control (Hare et al., 2009; C. R. Li et al., 2009; Steinbeis et al., 2012), a hallmark feature of problematic or even addiction-like behavioral patterns (Bechara, 2005, Brand et al., 2014, Grant et al., 2010). More generally, the prefrontal cortex is critical for cognitive control and goal-directed behaviors (Friedman and Robbins, 2022, Heekeren et al., 2006, Kaller et al., 2011, Tsujimoto et al., 2011), which are extensively refined during adolescence (Crone and Dahl, 2012, Fuster, 2002, Romine and Reynolds, 2005) to help support decreases in risky decision making (Crews et al., 2007, Gardner and Steinberg, 2005). Again, this interpretation is only speculative, and future work is necessary to investigate the links between impulse control and social media use. Beyond the more focal theta effect, spontaneous delta activity was robustly related to problematic social media behaviors in extended areas of the bilateral superior parietal cortices, superior temporal, temporal pole, parahippocampal gyrus, and prefrontal regions. The superior parietal has long been associated with visual attention (Corbetta and Shulman, 2002, Mesulam, 1999, Petersen and Posner, 2012, Proskovec et al., 2018, Salmi et al., 2007), and may also mediate the relationship between perceptual and motor functions (Christensen et al., 2007, Sepulcre, 2014, Stern et al., 2007). The superior temporal cortex has also been implicated in visual attention (Pulliam et al., 2024, Ramezanpour and Fallah, 2022) as well as several other aspects of cognition (Cabeza and Nyberg, 2000), including semantic recognition (Bhaya-Grossman and Chang, 2022, Patterson et al., 2007, Pehrs et al., 2017, Rempe et al., 2022) and working memory (Heinrichs-Graham and Wilson, 2015, Killanin et al., 2024a, Park et al., 2011, Proskovec et al., 2018, Wiesman et al., 2016). The parahippocampal gyrus has similarly been linked to higher-tier visual processing (Aminoff et al., 2013) and memory (LaFlamme et al., 2021, Xie et al., 2025). Taken together, these results suggest that problematic social media use behaviors may be related to differential spontaneous activity in cortical regions implicated in visual attention, memory, and executive function. Interestingly, more problematic social media behaviors were related to higher transdiagnostic mental health symptoms in the domains of Internalizing Problems, Emotional Symptoms, Personal Adjustment, and Inattention/Hyperactivity. Critically, these effects were significant when controlling for overall social media use, and the descriptive statistics indicated no correlation between overall social media use and any mental health domain. Such results are consistent with recent trends in the literature showing that the relationship between mental health and social media use is specific to particular problematic patterns of social media engagement (Cataldo et al., 2021; C. Huang, 2017; Petro, Picci, et al., 2025; Przybylski and Weinstein, 2017; Valkenburg et al., 2022; Xiao et al., 2025) and not necessarily to typical social media use or screentime per se (Eirich et al., 2022; C. Huang, 2017; Nagata et al., 2024; Przybylski and Weinstein, 2017; Torre et al., 2020; Valkenburg et al., 2022). This is obviously an important distinction for future research and a major consideration for public policy changes directed at youths’ phone and social media use. In fact, it should be noted that beneficial effects of social media have been reported (Crone and Konijn, 2018, Krause et al., 2021) and peer interactions are undeniably necessary for typical social and brain development (Blakemore and Mills, 2014, Crone and Dahl, 2012, Pfeifer and Allen, 2021). Thus, these results should not be interpreted to mean that any type or frequency of social media use (including those involving peer interactions) would be associated with mental health symptoms, as only certain types of problematic use would be captured by our social media behaviors survey. The internalizing, emotional symptoms, and personal adjustment scales involve questions related to feelings of depression, anxiety, and decreased self-worth, as well as interpersonal relationships with peers and caregivers. One possibility, which should be explored in future studies, is that social media use may impact social bonds and subsequently lead to increased anxiety and depression symptoms. Such social bonds are known to be intimately linked to adolescent mental health, as peer relationships are foundational in sculpting behavior during this developmental period (Blakemore, 2008, Foulkes and Blakemore, 2021, Long et al., 2020, Steinberg and Morris, 2001). The items contributing to the inattention/hyperactivity composite probe classic symptoms of attention deficit hyperactivity disorder (ADHD), including those pertaining to impulse control (e.g., speaking out of turn). Past studies have linked reduced impulse control to excessive social media use (Chen et al., 2023, Lewin et al., 2023), although whether problematic social media use behaviors contributed to this relationship was not examined. Reduced impulse control increases behaviors that produce strong short-term rewards and are central to models of problematic or even addiction-like behaviors (American Psychiatric Association, 2022, Grant et al., 2010). Notably, the social media behaviors survey items with the strongest component loadings (see Table 2) pertained to academic performance and social bonds. These results point to a far-reaching and multifaceted relationship between problematic social media use and mental health involving internalizing problems, social connections, impulse control, and attention problems. Before closing several limitations should be noted. First, it must be acknowledged that the study of social media use is in its early years, and the concept of problematic behaviors toward social media is not settled. While we investigated the potential for such behaviors using a survey that was modeled on a substance use survey, social media use should not be treated as interchangeable with substance use, gambling, or other widely accepted forms of risky or problematic behaviors. Further work is needed to understand how behaviors toward social media should be conceptualized in the context of broader mental health. Second, while we based our social media behaviors survey on the widely used NIDA Modified ASSIST assessment to facilitate future validation work, we used a different scoring method in the current study that relied on PCA-based component scores instead of the standard total (summed) score approach. We used PCA instead of summing the scores because this enabled us to derive the most consistent variance across participants and avoid contaminating this metric with “noisy” responses (i.e., those where the NIDA survey’s language did not appear to closely track social media use). This departure from the standard scoring method appeared to have a negligible impact on our findings in follow-up studies, but future work should more closely examine this and other possible scoring approaches to ensure the most valid approaches are utilized. Third, this study was cross-sectional and does not establish causal links between the variables. Future studies should investigate how the different facets of mental health are affected by problematic social media use, namely those involving interpersonal relationships with peers and caregivers, and if the brain may mediate these effects. In addition, studies should identify the causes of problematic social media use. For example, adolescents may attend to social media to interact with peers, or perhaps to avoid unpleasant emotions from pre-existing mental health disorders (Herman and Duka, 2019). Fourth, social media use was self-reported in this study. Future work should use passive recordings from smart devices to provide more objective data regarding which social media applications are most likely to trigger risky and problematic behaviors. As noted above, while the current study focused on problematic social media use, there are also benefits of social media (Crone and Konijn, 2018, Krause et al., 2021) that should be explored. Fifth, participants who completed the study reported no current or history of previous psychiatric diagnosis. Future studies may test the extent to which these results generalize to individuals with diagnosed mental health disorders, as risky or problematic behaviors may manifest differently in this context (Szerman et al., 2022). Finally, the generalizability of resting brain activity to task-based activity is limited. Future studies should test if the current results extend to task-based neural responses, especially paradigms assessing impulse control and attention. Adolescence is a time when peer-relationships are highly rewarding and highly influential in social and cognitive development. Social media presents a new context for such interactions, which, recent trends suggest, allows the desire for peer interactions to escalate toward excessive and problematic levels that resemble other risky behaviors. The current study found that such social media behaviors were related to greater spontaneous delta and theta activity, and to multiple transdiagnostic mental health symptoms in youth. The brain effects were strongest in prefrontal, parietal, and temporal cortices, which underlie many processes ranging from perception to memory and executive function. Alterations in these cognitive faculties can lead to difficulties in impulse control, a hallmark feature of risky and problematic behaviors. Importantly, all findings were after controlling for overall social media use, suggesting that social media use per se is not detrimental. This nuance is consistent with the recent literature, which points to deleterious effects of specific problematic patterns of social media use. Thus, the nature of one’s engagement with social media may constitute an environmental risk factor for altered brain function and mental health. Funding acknowledgments This study was supported by the National Institutes of Health through grants R01-MH121101 (TWW), P20-GM144641 (TWW, GP), S10-OD028751 (TWW), F31-DA063403 (KMM), F31-DA056296 (MS), F30-MH134713 (JJS), and F30-MH130150 (ADK). The funders had no role in the study design, collection, analysis, or interpretation of data, nor did they influence writing the report or the decision to submit this work for publication. CRediT authorship contribution statement Danielle L. Rice: Investigation, Data curation. Grace C. Ende: Project administration, Data curation. Jake J. Son: Supervision, Project administration, Methodology, Conceptualization. Thomas W. Ward: Project administration, Methodology, Conceptualization. Giorgia Picci: Writing – review & editing, Visualization, Supervision, Project administration, Methodology, Conceptualization. Lauren K. Webert: Writing – review & editing, Visualization, Validation, Software, Methodology, Formal analysis, Data curation. Tony W. Wilson: Writing – review & editing, Visualization, Validation, Supervision, Resources, Project administration, Methodology, Funding acquisition, Conceptualization. Kellen M. McDonald: Writing – review & editing, Project administration, Methodology, Conceptualization. Mikki Schantell: Writing – review & editing, Supervision, Project administration, Methodology, Data curation, Conceptualization. Abraham D. Killanin: Project administration, Methodology, Conceptualization. Nathan M. Petro: Writing – review & editing, Writing – original draft, Visualization, Supervision, Software, Methodology, Formal analysis, Data curation, Conceptualization. Erica L. Steiner: Investigation, Data curation. Cooper L. Livermore: Methodology, Conceptualization. Anna T. Coutant: Investigation, Data curation. Declaration of Competing Interest All authors report no biomedical financial interests or potential conflicts of interest. Footnotes Supplementary data associated with this article can be found in the online version at doi:10.1016/j.dcn.2026.101735. Supplementary material Supplementary material Data availability Data will be made available on request. References Associated Data Supplementary Materials Supplementary material Data Availability Statement Data will be made available on request.