Self-compassion, self-referential caudate circuitry, and adolescent suicide ideation

The study addresses why and how self-compassion may protect against suicide ideation in adolescents, particularly those with depression. Prior work shows depressed adolescents exhibit distorted self-referential processing and heightened responses to negative self-relevant stimuli, with striatal (including caudate) abnormalities linked to suicide ideation. Self-compassion has been associated behaviorally with lower suicide ideation and neurally with reduced cortical midline structure hyperactivity and increased caudate engagement during positive self-referential processing. The authors hypothesized that during self-appraisal, self-compassion would relate to stronger caudate functional connectivity with regions implicated in self and social cognition (e.g., medial prefrontal cortex, anterior cingulate, precuneus, pSTS/TPJ), and that such connectivity would mediate the inverse relation between self-compassion and suicide ideation, beyond established risk factors.

Existing literature indicates: (a) adolescent depression is a strong predictor of suicidality; (b) negative self-referential processing confers suicide risk; (c) general self-referential processing recruits cortical midline structures (medial prefrontal cortex, anterior cingulate, precuneus); (d) positive self-referential stimuli typically engage reward-related regions (e.g., striatum) more than negative stimuli in healthy individuals, whereas depressed youth with suicide ideation show greater caudate responses to negative vs. positive self-referential stimuli; (e) depressed individuals with suicide risk show aberrant caudate connectivity at rest and altered task-based connectivity between salience and task-relevant networks; (f) self-compassion is consistently inversely related to suicide ideation and corresponds to lower CMS hyperactivity in depressed adolescents during self-processing, and to heightened caudate engagement during positive self-referential processing. However, functional connectivity correlates of self-compassion had not been delineated, motivating examination of caudate circuitry as a mediator linking self-compassion to reduced suicide ideation.

Design: Cross-sectional fMRI study examining associations among self-compassion, caudate functional connectivity during self-appraisal (positive vs. negative), and suicide ideation, including mediation analyses while controlling for key confounds. Participants: Adolescents aged 11.0–17.8 years (M±SD=14.82±1.63) and caregivers recruited from psychiatric clinics at the Universities of Minnesota and Pittsburgh. Exclusions: IQ<70, primary diagnosis other than depression, left-handedness. Diagnoses via K-SADS-PL; interviews video-scored with 98% interrater agreement, discrepancies resolved by senior clinician. Final sample: 79 depressed youth (62 Minneapolis; 17 Pittsburgh) and 36 healthy controls. Measures: Self-compassion assessed by Self-Compassion Reactions Inventory (SCR; 0–16; higher scores = greater self-compassion). Suicide ideation quantified by averaging standardized suicide ideation items across K-SADS-PL, CDRS-R, and SIQ. Depression severity indexed by total CDRS-R score. Anxiety severity assessed with Beck Anxiety Inventory for Children, 2nd Ed. Non-suicidal self-injury (NSSI) assessed via K-SADS (any tissue-damaging self-injury without suicidal intent within past year). Self-processing task: In-scanner self-appraisal task where participants evaluated whether heard phrases were self-descriptive from multiple perspectives (self, mother, classmate, best friend) across domains (emotion, social), with positive and negative valence conditions. MRI acquisition: Two 3T Siemens Trio scanners (Minnesota and Pittsburgh). Structural MPRAGE: TE=3.31 ms, TR=2100 ms, TI=1050 ms, flip angle=8°, FOV=256×200 mm, matrix=256×200, 176 slices, 1 mm. BOLD EPI: 60 oblique axial slices, 2 mm thick; TR/TE=3340/30 ms; FOV=200×200 mm; matrix=80×80; flip angle=90°. Preprocessing: SPM12 pipeline: realignment, co-registration to anatomy, segmentation, MNI normalization, 7 mm FWHM smoothing. Artifact detection (ART): volumes with motion >2 mm, rotation >0.578°, or global signal outliers >9 excluded from first-level analyses. First-level modeling: GLM with factors 2 Valence (positive/negative) × 4 Perspectives (self, mother, classmate, best friend) × 2 Domains (emotion, social) resulting in 16 conditions; six motion parameters as nuisance regressors; t-maps per condition. PPI analyses: Seed-based psychophysiological interaction using individually defined left and right caudate seeds (7 mm spheres; participant-specific peaks within caudate mask). PPI contrast: positive vs. negative self-appraisals. Second-level seed-wise regressions included self-compassion, diagnostic group, and interaction; covariates included age and gender; additional models controlled for depression severity (non-suicide ideation), anxiety severity, and NSSI. Whole-brain thresholding: voxelwise p_uncorr<0.001 with cluster-level correction p_corr<0.10 (reporting trends), with supplementary stricter cluster-level p<0.05 where noted. Mediation: Correlated identified caudate connectivity clusters with suicide ideation (controlling for depression severity, anxiety severity, NSSI). Proceeded to mediation using PROCESS (self-compassion as predictor, suicide ideation as outcome, caudate connectivity as mediator), controlling for the same covariates. Subgroup analyses included only depressed participants and only non-medicated depressed participants.

• Behavioral associations: Self-compassion correlated negatively with suicide ideation (r=-0.65, p<0.001), depression severity (r=-0.68, p<0.001), anxiety severity (r=-0.62, p<0.001), and NSSI (r=-0.45, p<0.001). The self-compassion–suicide ideation link remained significant controlling for depression severity, anxiety severity, and NSSI (partial r=-0.27, p=0.003).
• Group differences: Depressed youth (DEP) reported lower self-compassion than healthy controls (HC) (DEP: 6.82±4.44; HC: 13.25±3.42; t(113)=-8.49, p<0.001). Suicide ideation standardized scores were higher in DEP than HC (table: t(113)=12.97, p<0.001). Depression severity higher in DEP (63.47±14.63) than HC (20.35±5.83; t(113)=22.57, p<0.001).
• Regression on suicide ideation: Significant effects of self-compassion (B=-0.34, p<0.001), diagnostic group (B=0.47, p<0.001), and interaction (B=-0.18, p=0.034). Diagnostic group effects on suicide ideation were larger at lower self-compassion.
• fMRI connectivity correlates of self-compassion (positive>negative self-appraisals; left caudate seed): Greater connectivity with bilateral pSTS/TPJ (BA39/13/22; left cluster r≈0.53), left middle occipital gyrus (r≈0.47), left MTG (BA21; B≈0.46), and dorsomedial prefrontal cortex (trend). Results persisted and in some cases strengthened when controlling for depression severity, suicide severity, anxiety, and NSSI. No significant right caudate connectivity correlates and no diagnosis or interaction effects.
• Links to suicide ideation: Left caudate connectivity with left pSTS/TPJ and left MTG correlated negatively with suicide ideation after covariate control (pSTS/TPJ r=-0.22, p=0.021; MTG r=-0.33, p<0.001), whereas other clusters did not survive covariate control.
• Mediation: Only left caudate–left MTG connectivity significantly mediated the association between higher self-compassion and lower suicide ideation, controlling for depression severity, anxiety severity, and NSSI. The mediation held in depressed-only analyses; in non-medicated depressed youth, left caudate–pSTS/TPJ connectivity showed mediation in a smaller subsample.
• Lateralization: Associations were specific to the left caudate, possibly reflecting language-related processing demands of the verbal self-appraisal task.

Findings support the hypothesis that self-compassion relates to enhanced caudate-based circuitry favoring positive over negative self-referential processing, and that this circuitry links self-compassion to reduced suicide ideation in adolescents. The pattern of stronger left caudate connectivity with pSTS/TPJ, MTG, and middle occipital regions during positive appraisals suggests integration of reward/motivation (caudate) with social-cognitive (pSTS/TPJ) and language/semantic and perceptual systems (MTG/visual cortex) when processing positive self-descriptors. Crucially, left caudate–MTG connectivity, rather than pSTS/TPJ, statistically mediated the self-compassion–suicide ideation association even after controlling for key risk factors, highlighting early semantic/linguistic integration as a pathway through which self-compassion may shift self-processing toward positive valence and reduce suicidal thoughts. The left-hemisphere specificity is consistent with language-dominant processing of verbal self-appraisals. Clinically, these results suggest that interventions enhancing self-compassion or directly modulating caudate–MTG circuitry during positive self-referential processing may help mitigate suicide ideation in depressed youth, independent of medication effects.

Self-compassion is linked to greater left caudate connectivity with regions supporting social cognition and language/semantic processing during positive vs. negative self-appraisals in adolescents. Left caudate–left MTG connectivity mediates the inverse relationship between self-compassion and suicide ideation, including among depressed youth and after controlling for depression severity, anxiety, and NSSI. These findings advance mechanistic understanding of self-compassion’s protective role against suicidal ideation and point to left caudate–MTG circuitry during self-referential processing as a potential neural target for intervention, either through self-compassion training or neuromodulatory approaches. Future work should test causality via interventions and delineate which self-compassion components drive these neural and behavioral effects.

• Correlational design precludes causal inference among self-compassion, caudate connectivity, and suicide ideation; intervention studies are needed.
• Self-compassion assessed with a unidimensional measure (SCR); specific contributions of components (e.g., self-kindness, common humanity, mindfulness) are unknown; replication with multidimensional scales (e.g., Neff’s SCS) is warranted.
• Focused solely on caudate circuits; other striatal regions (e.g., putamen, ventral striatum) were not examined and could be relevant.
• Some subgroup mediation results (e.g., non-medicated depressed sample) are based on small Ns and require validation.
• Task did not include neutral, non-evaluative self-statements, limiting conclusions about valence specificity beyond positive vs. negative contrasts.