Ortega, S.; Lenz, A.; Lundgren, E. J.; Mizuno, A.; Poo Hernandez, S.; Nakagawa, S.; Lagisz, M.

Anxiety and depressive disorders impose a major global burden, prompting interest in non-pharmacological interventions that may influence affective processes. Music exposure has often been reported to affect anxiety- and depression-like behaviors, but preclinical findings remain heterogeneous and have not been quantitatively synthesized. Prior work has also focused almost entirely on mean behavioral responses, largely overlooking inter-individual variability as a biologically meaningful dimension. We conducted a preregistered systematic review and multilevel meta-analysis of experimental studies testing music exposure in laboratory rodents. Following PRISMA and PRISMA-EcoEvo guidelines, we synthesized 298 effect sizes from 20 studies using multilevel models to account for non-independence among effect sizes. We quantified effects on mean behavior with the log response ratio (lnRR) and effects on variability with the log variability ratio (ln VR). Overall, music exposure was associated with a statistically significant reduction in anxiety- and depression-like behaviors, corresponding to an average decrease of about 18% relative to controls. This mean effect was detected across outcome types and life stages despite substantial heterogeneity. By contrast, music exposure did not produce a statistically significant overall change in inter-individual behavioral variability. Instead, variability responses were context dependent: behavioral assay type and music meta-genre significantly moderated lnVR, with anxiety-like assays tending to show increased variability and depression-like assays tending to show reduced variability under music exposure. These results suggest that music exposure reliably shifts average affect-related behavior without uniformly changing behavioral stability across individuals. Because the evidence comes mainly from short-term exposures in young adult laboratory rodents, generalization beyond similar contexts should remain cautious.