Adolescent suicide rates have risen over the past two decades, underscoring the need for improved risk detection strategies. Although natural language processing (NLP) tools are increasingly used to flag suicide-related content, little is known about how such approaches perform on adolescents' smartphone communications. Addressing this gap, this study leverages passively collected smartphone data to identify suicide-related language in adolescents' keyboard usage via NLP. We developed a lexicon of suicide-related adolescent language and validated it with labeled data (N = 171,468 text entries; e.g., messages, web searches), demonstrating higher performance in identifying suicide-related text than few-shot prediction with large language models (LLMs) and lexicons not designed for youth. Across two independent cohorts at elevated suicide risk (Ns = 208 & 257; >6 million text entries), lifetime suicidal thoughts and behaviors (STB) and current suicidal ideation were associated with increased frequency of smartphone suicide-related language. Human coding indicated varied language, including authentic first-person current suicidal ideation (14.5%) and jokes or hyperbole (20.2%). Compared with the lexicon alone, human coding of suicide-related entries with first-person language showed stronger associations with STB history. These findings highlight that effective NLP-based tools for suicide prevention will require more nuanced and context-specific approaches to better distinguish suicidal intent.

Adolescent smartphone language provides a lens into negative self-referential thinking, which is central to major depressive disorder (MDD). Prior studies have linked language features, including negative sentiment and first-person pronouns, to mood and depression, suggesting that naturalistic language may identify who is at risk and when that risk is greatest. However, studies of adolescent smartphone social communication have typically relied on rule-based models not validated for heterogeneous, context-sensitive language. To address this gap, we determined whether transformer models, including large language models, optimized detection of depression risk in extensive adolescent smartphone text data. In this study, 223 adolescents (Mage = 16.43 years, current MDD = 37, remitted MDD = 103, healthy controls = 83) installed a smartphone app, which prompted participants to provide mood ratings once a day and acquired all keyboard inputs over 12 months (mean text entries per participant = 17,683). Ten thousand text entries were double-coded for entry-level sentiment (positive, neutral, negative) and self-reference for training and testing traditional rule-based approaches (VADER, pronoun counts) and transformer approaches, including GPT-4-mini. In the full data set, between-participant associations with depressive symptoms and within-participant relations to daily mood and depressive episodes were examined. Fine-tuned transformer models best aligned with human-coded sentiment labels (F1GPT4-MINI = .84, F1VADER = .59) and accurately detected self-reference (F1T5-BASE = .97). Adolescents with current and remitted MDD exhibited more transformer-based negative self-referential language than healthy controls (ORs = 1.38, 1.26). Increased negative self-referential language predicted worse next-day mood (β = -.033, p < .001) and a higher likelihood of next-week depressive episodes (OR = 1.66, 95% confidence interval [1.06, 2.62], p = .028), although the association with depressive episodes was not robust to sensitivity analyses. Transformer models may be integrated into digital mental health care to detect when youth are at risk for depression. (PsycInfo Database Record (c) 2026 APA, all rights reserved).

Adolescent smartphone language provides a lens into negative self-referential thinking, which is central to major depressive disorders (MDD). Prior studies have linked language features, including negative sentiment and first-person pronouns, to mood and depression, suggesting that naturalistic language may identify who is at risk and when that risk is greatest. However, studies of adolescent smartphone social communication have typically relied on rule-based models not validated for heterogeneous, context-sensitive language. To address this gap, we determined whether transformer models including LLMs optimized detection of depression risk in extensive adolescent smartphone text data. In this study, 223 adolescents (AgeMean=16.43 years, current MDD=37, remitted MDD=103, healthy controls=83) installed a smartphone app, which prompted participants to provide mood ratings once a day and acquired all keyboard inputs over 12 months (Mean text entries per participant=17,683). Ten thousand text entries were double-coded for entry-level sentiment (positive, neutral, negative) and self-reference for training and testing traditional rule-based approaches (VADER, pronoun counts) and transformer approaches, including GPT-4-mini. In the full dataset, between-participant associations with depressive symptoms and within-participant relations to daily mood and depressive episodes were examined. Fine-tuned transformer models best aligned with human-coded sentiment labels (F1GPT4-MINI =0.84, F1VADER=0.59) and accurately detected self-reference (F1T5-BASE =0.97). Adolescents with current and remitted MDD exhibited more transformer-based negative self-referential language than healthy controls (ORs=1.38, 1.26). Increased negative self-referential language predicted worse next-day mood (β=-0.033, p&lt;0.001) and higher likelihood of next-week depressive episodes (OR=1.67, 95% CI = [1.06, 2.61], p=0.028), although the association with depressive episodes was not robust to sensitivity analyses. Transformer models may be integrated into digital mental health care to detect when youth are at risk for depression.

There is growing interest in identifying brain function underlying adolescent cognition, personality, and psychopathology. One promising approach is Precision Functional Mapping (PFM) of MRI functional connectivity, a data-intensive method for characterizing individualized brain networks. Foundational studies suggest that PFM can detect stable, task-responsive, and clinically relevant networks. Studies demonstrate that both functional connectivity reliability and network stability improve with increasing data quantity, although benchmark estimates vary across populations, preprocessing pipelines, and MRI acquisition approaches. Accordingly, it is important to understand how PFM performs in adolescent populations and with multi-echo fMRI acquisition. In a case study of eight youth (ages 10-17), we applied PFM to 80 minutes of combined resting-state and task-based fMRI. The resulting networks were highly modular, consistent with adult templates, and without evidence of structural registration artifacts. Functional connectivity reliability compared favorably to prior single-echo studies, with multivariate similarity and ICC estimates showing early stabilization around 10-15 min despite continued improvement with additional data. Trait-like stability increased gradually with acquisition time, and a Bayesian algorithm (MS-HBM) demonstrated higher stability than Infomap. Across algorithms, stability was greatest in the somatosensory, auditory, visual, and parietal networks. Furthermore, when evaluating task-based responses to threat and attention paradigms, only the auditory network consistently benefited from individualized mapping over group template networks. These findings suggest that, with constrained scanning time, PFM is especially effective for characterizing sensory and perceptual networks in adolescents. Bridging the methodological divide between deeply sampled individual cases and large-scale developmental studies will require further innovation and validation.

Mind‑body practices, such as meditation and yoga, involve paying attention to breathing sensations. During these practices, individuals report “interoceptive lapses,” moments when attention drifts away from internal bodily sensations. While lapses in attention to the external world have been widely studied, little is known about the physiological and neural mechanisms of interoceptive lapses. Interoceptive lapses may share markers with exteroceptive lapses—such as reaction time variability and default-mode network (DMN) connectivity—but may also depend on distinct brain systems and breathing physiology. We examined behavioral, physiological, and neural signals preceding lapses in a sample of 93 adolescents enriched for GAD and depression symptoms. Participants performed a 20-min breath counting task in the fMRI scanner with simultaneous breath recordings. Lapses were defined as moments when counting errors occurred. The sample was split into training and validation sets to test machine learning models predicting attentional lapses. The strongest predictors were timing and variability of button responses (AUCs > 0.75). Breathing variability and breathing–behavior synchronization showed smaller but generalizable predictive value (AUCs < 0.65). Whole-brain connectivity models also predicted lapses (AUC ≈ 0.65), incorporating the DMN, dorsal and ventral attention, and somatomotor networks. Furthermore, models that included brain connectivity marginally outperformed behavior-only models. Comparisons to previous exteroceptive findings indicate some common markers (e.g., reaction time variability) and some unique markers (e.g., selective perceptual coupling with attentional networks). Although limited by the clinical sample and lack of a control task, these results highlight brain–body markers of interoceptive attention that may inform real-time monitoring during mind-body interventions.

Major depressive disorder is a leading cause of disability among adolescents. Perseverative negative self-referential thoughts are a promising treatment target. Mindfulness-based real-time fMRI neurofeedback (mbNF), which guides mindfulness practice with feedback to train the downregulation of the default mode network (DMN), is an intervention targeting such negative self-referential thoughts. This study builds on a registered NIMH-supported trial testing the optimal dosing of mbNF on downregulating DMN activation among depressed adolescents. Adolescents (N=90), ages 13-18-years-old, with major depressive disorder will be randomized to receive either a 15- or 30-minute mbNF session. Before and after mbNF, participants will complete a self-referential encoding fMRI task, wherein they categorize whether trait adjectives describe themselves or a friend. It is hypothesized that a 30-minute versus 15-minute mbNF dose will relate to: (1) larger decreases in behavioral negative self-referential biases and (2) larger decreases in DMN activation during self-referential processing.

Trait mindfulness refers to one's disposition or tendency to pay attention to their experiences in the present moment, in a non-judgmental and accepting way. Trait mindfulness has been robustly associated with positive mental health outcomes, but its neural underpinnings are poorly understood. Prior resting-state fMRI studies have associated trait mindfulness with within- and between-network connectivity of the default-mode (DMN), fronto-parietal (FPN), and salience networks. However, it is unclear how generalizable the findings are, how they relate to different components of trait mindfulness, and how other networks and brain areas may be involved. To address these gaps, we conducted the largest resting-state fMRI study of trait mindfulness to-date, consisting of a pre-registered connectome-based predictive modeling analysis in 367 meditation-naïve adults across three samples collected at different sites. In the model-training dataset, we did not find connections that predicted overall trait mindfulness, but we identified neural models of two mindfulness subscales, Acting with Awareness and Non-judging. Models included both positive networks (sets of pairwise connections that positively predicted mindfulness with increasing connectivity) and negative networks, which showed the inverse relationship. The Acting with Awareness and Non-judging positive network models showed distinct network representations involving FPN and DMN, respectively. The negative network models, which overlapped significantly across subscales, involved connections across the whole brain with prominent involvement of somatomotor, visual and DMN networks. Only the negative networks generalized to predict subscale scores out-of-sample, and not across both test datasets. Predictions from both models were also negatively correlated with predictions from a well-established mind-wandering connectome model. We present preliminary neural evidence for a generalizable connectivity models of trait mindfulness based on specific affective and cognitive facets. However, the incomplete generalization of the models across all sites and scanners, limited stability of the models, as well as the substantial overlap between the models, underscores the difficulty of finding robust brain markers of mindfulness facets.

Mindfulness meditation is a systematic training in equanimity, sensory clarity, and concentration rooted in ancient contemplative traditions. Here, we synthesized cognitive-behavioral outcomes in long-term meditators (LTMs) resulting from diverse, prolonged meditation practices. Preliminary evidence suggests that LTMs exhibit increased cognitive-sensory integration and decoupling of affective processes, demonstrated by enhanced interoceptive awareness, reduced negative affective pain perception, and more rational decision making. Additionally, LTMs may experience more emotional neutrality, malleable self-boundaries, and altered self-awareness. Neuroimaging findings included increased bottom-up activation, particularly within the salience network (interoception, pain, affect), and reduced connectivity between the executive (dorsolateral prefrontal cortex) and salience (dorsal anterior cingulate cortex) networks (reduced pain). The literature also suggests reduced fear and amygdala activation (mitigated negative affect), increased temporoparietal junction activation (pre-reflective experiential processes, empathy), and altered midline default-mode network activation, which was associated with emotional neutrality and non-ordinary states of consciousness. Methodological limitations restricted the interpretation of trait effects, emphasizing the need for a unified framework to systematically investigate advanced meditation's states, stages, and endpoints using neurophenomenology. In summary, LTMs display a distinct neurophenomenological gestalt of mindfulness, wherein meditative expertise is reflected in enhanced cognitive flexibility and integration, self-regulation, and non-dual awareness-signifying a potentially important form of embodied cognition.

Adolescent suicide rates have risen over the past two decades, underscoring the need for improved strategies to detect risk. This study leverages passively collected smartphone data to identify suicide-related language in adolescents’ keyboard usage using natural language processing. We developed a youth suicide lexicon for adolescent language and validated it with labeled data (N=121,515 entries), demonstrating higher sensitivity and precision than lexicons not designed for youth. Across two independent cohorts at elevated suicide risk (Ns=208 and 211; >6 million text entries), both lifetime suicidal thoughts and behaviors and current suicidal ideation were associated with increased frequency of smartphone suicide-related language. Human coding indicated varied language—e.g., serious expressions of active suicidal ideation, jokes, hyperbole, and expressing support for others. Most suicide-related entries did not express serious current first-person suicidal ideation, underscoring the need for improved approaches to distinguish intent. Findings highlight both the promise and limitations of NLP approaches for suicide prevention.

Objective measurement of mindfulness could help us understand the mechanisms of meditation interventions and how individuals vary in their disposition to be mindful. One proposed measure is the breath-counting task (BCT), which measures how accurately one can count cycles of their breath. Breath counting, which involves sustained attention, meta-awareness, and an internal locus of attention, has been shown in adults to be related to measures of mindfulness even when controlling for established attentional measures. In this study, we test the psychometrics of the BCT in a convenience sample of 78 adolescents with elevated rumination. In preregistered analyses, we related breath-counting measures, including novel objective respiration measures, to a suite of self-report measures as well as the sustained attention to response task (SART). While breath-counting performance showed fair split-half reliability and similar distributions to studies in adults, it did not show the expected positive associations with self-reported mindfulness measures (neither trait nor EMA). Surprisingly, breath-counting accuracy showed negative correlations with a subscale measuring observing of emotions and body sensations, negative correlations with nonreactivity, and performance decrements were larger for individuals scoring more highly on mindfulness in general. The SART showed a small negative correlation with breath-counting resets (an index of mind-wandering). Finally, breath-counting performance was not related to other theoretically relevant clinical, personality, and executive functioning criteria. Our results suggest that, at least in ruminative adolescents, breath-counting may measure a very narrow, contextual form of sustained attention, may not capture other qualities of mindfulness, and may lack predictive validity.

Advanced meditation research investigates states and stages of practice that unfold with increasing mastery and time, which may include altered states of consciousness such as a diminished sense of self. In the current study, we examined a 7-T fMRI case study of jh & amacr;na, an advanced concentrative absorptive meditation (ACAM-J). Specifically, we examined the temporal properties of dynamic connectivity brain states that could reflect mental states and phenomena during ACAM-J. We identified two brain states that were more prevalent during ACAM-J than control conditions. One state, involving default-mode network anticorrelations with the rest of the brain, increased across ACAM-J. Another state, involving hyperconnectivity across many cortical networks, was correlated with reports of narrow attention and greater sensory awareness, as well as diminished across ACAM-J.

Mindfulness meditation training may cultivate interoceptive awareness and provide therapeutic benefit when implemented within mental and physical health interventions. This pre-registered meta-analysis evaluated the impact of mindfulness interventions on self-reported interoception measures and associated relationships with psychological outcomes. Twenty-nine randomized controlled trials with 2,191 participants (77.8% female, mean age 32.8 years) were meta-analyzed using correlated and hierarchical effects models. Interventions included mindfulness-based programs (k = 15), body-based approaches (incorporating elements like massage, k = 8), and other variations (k = 6). Five SIMs were tested; the Multidimensional Assessment of Interoceptive Awareness was the most common (22 studies). Results showed a small-to-medium positive effect on interoception measures across all studies (g = 0.31, p < 0.001, 95% CI [0.21, 0.42]) with low-to-moderate heterogeneity (τ = 0.16). Mindfulness-based programs demonstrated the largest effects (g = 0.41). No evidence of publication bias was found. No other moderators, such as practice dosage or clinical sample, were significant. Improvements in self-reported interoception were similar in size to improvements in self-reported mindfulness and were related to improvements in psychological distress. These meta-analytic findings provide evidence that mindfulness-based interventions lead to adaptive changes in the subjective experience of interoception, perhaps contributing to improved mental wellbeing.

Rumination, or perseverative negative self-referential thinking, is a hallmark of depression. In adults, a dynamic resting-state fMRI model of trait rumination was recently identified through predictive modelling. In adolescents, a development period during which rumination and depression increase, the neurobiological correlates of ruminative thinking are less clear. In the current preregistered study, we examine dynamic connectivity correlates of self-reported rumination in the largest sample of adolescents to date (n = 443, containing clinical and non-clinical individuals). Notably, the adult model failed to generalize to our sample. In addition, linear models trained on default-mode network (DMN) connectivity, as well as whole-brain connectome models, failed to generalize to held-out data. In an exploratory random forest analysis, we found significant prediction performance of a model where increased variability between DMN-cerebellum, DMN-dorsal attention network, and DMN-DMN connections was nominally associated with higher rumination. However, the model did not generalize to an external sample with lower rumination scores and a distinct scanner protocol. Our findings illustrate the difficulty of characterizing the neurodevelopment of risk factors for depression.

This systematic narrative review examines neuroimaging studies that investigated the neural correlates of mindfulness-based interventions in youth (ages 0–18). We extracted 13 studies with a total of 467 participants aged 5–18 years from the MEDLINE database on February 21st, 2024. These studies included both typically developing youth and those at risk of developing or recovering from neuropsychiatric disorders. Most studies (76.9

In-person mindfulness-based interventions (MBIs) have been shown to decrease symptoms of anxiety and stress in autistic adults, who often report high levels of these symptoms. Little is known about the effectiveness of remote MBIs for this population, which may be particularly useful given the common barriers autistic adults face in accessing in-person treatment. This study examined the feasibility and effectiveness of an app-based mindfulness intervention for autistic adults. This randomized controlled trial (RCT) examined whether a 6-week remote intervention, using a customized version of the Healthy Minds Program app, reduced symptoms of anxiety and perceived stress in 89 autistic adults. Participants were randomly assigned to either the mindfulness intervention or a wait-list control (WLC) group. The WLC group received the intervention after the RCT. Self-report measures of anxiety, perceived stress, positive and negative affect, and trait mindfulness were administered at several timepoints. The mindfulness group showed significant decreases in anxiety symptoms and perceived stress relative to the control group, with medium to large between-groups effect sizes (ηp2 0.07 to 0.14). These benefits, as well as significant decreases in negative affect and increases in trait mindfulness, were replicated when the WLC group subsequently received the intervention, and were retained in both groups 6 weeks after conclusion of the intervention. Results demonstrate both the feasibility and effectiveness of a remote mindfulness self-guided intervention for reducing perceived stress and anxiety symptoms in autistic adults. Future research can investigate the specific processes of how such an intervention exerts its effects. ClinicalTrials.gov TRN: NCT05880498, 5/30/23, retrospectively registered.

Attention regulation is a core mechanism of mindfulness meditation and has been proposed to underlie many of its health-related benefits. Here, we review and synthesize behavioral findings on attentional outcomes in long-term meditators, integrating neurocognitive evidence within a meditative development framework. Key findings indicate trait-level improvements across attentional functions—executive attention, sustained attention, hierarchical and general orienting—and attentional phenomena, such as the attentional blink. Preliminary evidence also identifies trait enhancements in response inhibition, alertness, and reduced mind-wandering. Interaction effects were found for response inhibition, sustained attention, reduced mind-wandering, and alertness, with alertness benefiting most strongly from long-term and intensive acute practice. As expected, attention-based outperformed non-attention-based techniques, while observe-and-release techniques facilitated attentional orienting and detection of closely spaced or unexpected stimuli during sustained attention tasks. These findings suggest that long-term meditation may enhance attention regulation in accordance with training specificity principles; the cognitive functions most directly targeted are the most likely to improve. Nevertheless, broader findings indicate that meditative development may depend on the balanced cultivation of multiple faculties over time, highlighting the non-linear and multidimensional nature of long-term meditative change. Consistent with traditional goals of cultivating mental faculties, the present findings may reflect attentional adaptations that support the development of advanced meditative states. Despite considerable consistency in empirical results, methodological limitations—including heterogeneous study designs and insufficient differentiation between states and traits—complicate interpretations. Future research should prioritize operationalizing and measuring contemplative constructs within integrative frameworks and using rigorous factorial designs to clarify state-trait interactions and meditation predictors.

BackgroundThere is burgeoning interest in the application of neuroscientific technology to facilitate meditation and lead to beneficial psychological outcomes. One popular approach is using consumer-grade neurofeedback devices to deliver feedback on brain targets during meditation (mindfulness-based neurofeedback). It is hypothesized that optimizing brain targets like alpha and theta band activity may allow meditators to experience deeper mindfulness and thus beneficial outcomes. ObjectiveThis study aimed to systematically review and meta-analyze the impacts of consumer-grade mindfulness-based neurofeedback compared with control conditions. Included studies involved mindfulness practice operationalized as open monitoring or focused attention meditation. This study was preregistered. MethodsA total of 16 randomized controlled training trials, as well as 5 randomized within-participant designs were included, encompassing 763 and 167 unique participants, respectively. Effects were categorized outcomes (ie, psychological distress, cognitive function, and physiological health) and process variables (ie, state mindfulness and brain measures). Study risk of bias, reporting bias, and publication bias were assessed. ResultsSamples were typically small (n=30-50), and the majority of studies used mindfulness apps as controls. To deliver neurofeedback, most studies used the Muse device (11/16 randomized controlled trials [RCTs]). There was a modest effect for decreases in psychological distress compared with controls (k=11, g=–0.16, P=.03), and heterogeneity was low (I2< 0.25). However, there was no evidence for improvements in cognition (k=7, g=0.07, P=.48), mindfulness (k=9, g=0.02, P=.83), and physiological health (k=7, g=0.11, P=.57) compared to controls. Mechanistic modulation of brain targets was not found in RCTs or within-participant designs. Sex (male or female), age, clinical status, study quality, active or passive controls, sample size, and neurofeedback duration did not moderate effects. There was some evidence for reporting bias, but no evidence of publication bias. Adverse effects were not assessed in 19 out of 21 studies and not found in the 2 studies that assessed them. ConclusionsAssertions that consumer-grade devices can allow participants to modulate their brains and deepen their meditations are not currently supported. It is possible that neurofeedback effects may rely on “neurosuggestion” (placebo effects of neurotechnology). Future research should examine more extensive calibration and individualization of devices, larger sample sizes, and gold-standard sham-controlled RCTs.

Breathing meditation typically consists of directing attention toward breathing and redirecting attention when the mind wanders. As yet, we do not have a full understanding of the neural mechanisms of breath attention, in particular, how large-scale network interactions may be different between breath attention and rest and how these interactions may be modulated during periods of on-task and off-task attention to the breath. One promising approach may be examining fMRI measures including static connectivity between brain regions as well as dynamic, time-varying brain states. In this study, we analyzed static and dynamic functional connectivity in 72 adolescents during a breath-counting task (BCT), leveraging physiological respiration data to detect objective on-task and off-task periods. During the BCT relative to rest, we identified increases in static connectivity within attention-direction and orienting networks and anticorrelations between attention networks and the DMN. Dynamic connectivity analysis revealed four distinct brain states, including a DMN-anticorrelated brain state, proportionally more present during the BCT than the rest. We found there were distinct brain state markers of (i) breathing tasks vs rest and (ii) momentary on-task vs off-task attention within the BCT, yet in this analysis, no identifiable brain states reflecting between-individual behavioral variability.

Mindfulness meditation is a form of mental training rooted in ancient wisdom traditions and is focused on cultivating a non-judgmental stance toward present-moment awareness. Here, we synthesize cognitive-behavioral effects in long-term meditators (LTMs) resulting from diverse and prolonged meditation practices. Preliminary evidence suggests that LTMs exhibit increased cognitive-sensory integration and decoupling of affective processes, as demonstrated in enhanced interoceptive awareness, reduced negative affective pain perception, and more rational decision-making. Additionally, LTMs may experience more emotional neutrality, self-boundary dissolution, and less normative self-awareness. Neuroimaging findings include increased bottom-up activation, particularly within the salience network (interoception, pain, affect), and reduced connectivity between the executive (dorsolateral prefrontal cortex) and salience (dorsal anterior cingulate cortex) networks (reduced pain). Research also displayed reduced amygdala activation to fear (reduced negative affect), increased temporoparietal junction activation (pre-reflective experiential processes, empathy), and altered midline default-mode network activation, which is associated with emotional neutrality and pre-reflective experiential processes, such as non-ordinary states of consciousness. Methodological limitations, specifically heterogeneous predictor variables, restrict the interpretation of trait effects, temporal dynamics in cognitive processing, and the unique influences of meditative activities. These limitations indicate the need for a unified research framework and a systematic neurophenomenological investigation of advanced meditation—through the study of unfolding states, stages, and endpoints in meditative development. In summary, LTMs display a distinct neurophenomenological gestalt of mindfulness, wherein meditative expertise is reflected in altered general brain processing, potentially enhanced cognitive integration, increased cognitive flexibility and self-regulation, and heightened non-dual awareness—signifying a potentially important form of embodied cognition.