Abstract
BACKGROUND: Social anxiety (SA) is highly prevalent among adolescents and is closely linked to psychological wellbeing and social adjustment. Physical activity (PA) has been identified as an important protective factor; however, most existing studies rely on cross-sectional designs and focus on unidirectional effects, overlooking potential bidirectional dynamics between PA and SA. OBJECTIVE: This study aimed to examine whether PA and SA exhibit stable bidirectional longitudinal associations in adolescents and to test their temporal stability and measurement invariance across three waves. METHODS: A three-wave longitudinal study was conducted over 12 months with 989 adolescents (aged 10-14 years) from 3 junior middle schools in Jiangsu Province, China. PA was assessed using the PARS-3, and SA was measured using the Short Form Social Interaction Anxiety Scale and the Short Form Social Phobia Scale. Descriptive statistics and correlations were computed in SPSS 26.0. Measurement invariance and cross-lagged panel models (CLPMs) were estimated using Mplus 8.3. RESULTS: (1) Both PA and SA demonstrated configural, metric, and scalar invariance across the three waves, supporting comparability over time. (2) PA and SA were significantly negatively correlated at all time points, and both variables showed substantial temporal stability. (3) CLPM results indicated that PA significantly negatively predicted subsequent SA (β = -0.259 to -0.273), while SA also significantly negatively predicted subsequent PA (β = -0.175 to -0.373), revealing a robust bidirectional negative longitudinal relationship. CONCLUSION: PA and SA influence each other dynamically over time: lower PA exacerbates SA, whereas higher SA reduces PA participation, forming a negative feedback loop. Future interventions hould simultaneously promote PA and reduce SA to break this maladaptive cycle and foster adolescents' physical and psychological development.