Frontoparietal network topology as a neural marker of musical perceptual abilities

The study addresses why individuals vary in musical perceptual abilities and whether network-level properties of domain-general brain systems underpin these differences. Musical competence—perceiving, remembering, and discriminating musical structures—varies widely in the general population and depends on domain-general processes such as attention and working memory (WM). Prior research has often focused on isolated regions (e.g., auditory cortex) or pairwise connectivity, leaving open how large-scale network topology contributes to musical perception. The authors hypothesize that the topology of the frontoparietal working memory network (FPN), specifically greater integration efficiency and centrality in prefrontal and posterior parietal nodes, supports superior musical competence. They further predict functional FPN organization relates to musical ability via WM, while structural FPN organization contributes through sensory integration processes.

Existing objective measures of music perception include the Musical Ear Test (MET), Advanced Measures of Music Audiation, Profile of Music Perception Skills, and Swedish Musical Discrimination Test. The MET is widely validated across cultures, brief (~20 minutes), and robustly correlates with musical sophistication while being relatively independent of demographic variables. Neuroscience work has linked musical perceptual performance to spontaneous auditory cortex activity, interhemispheric auditory connectivity, and integrity of white matter tracts (e.g., callosal pathways, superior longitudinal fasciculus). Network neuroscience conceptualizes the brain as a graph and uses metrics such as integration, segregation, and centrality to relate brain topology to cognition. Prior studies show associations between functional network efficiency and general cognitive abilities (intelligence, WM, cognitive control) and suggest that functional connectivity is more strongly linked to high-level cognition, whereas structural connectivity often relates to sensory processes. WM depends on a distributed network with core dorsolateral prefrontal and posterior parietal nodes; higher WM capacity is associated with more integrative networks. Music training has been linked to enhanced executive functions and frontoparietal activity. These findings motivate a network-level investigation of how FPN topology relates to musical competence measured with the MET, and how WM may mediate these relationships.

Participants: 300 healthy adults completed MRI and Gold-MSI; 241 completed MET; after outlier exclusion and network construction QC, analyses included 232 (functional) and 225 (structural) participants for brain–behavior correlations; 201 had WAIS-IV WM scores (functional analyses n=201; structural n=195). Most were non-musicians (60% no lessons; 36% ≤5 years training; 8 participants >6 years). Ethical approval obtained; recruitment at Aarhus University/Hospital. Behavioral/Cognitive measures: Musical Ear Test (104 trials: 52 melody, 52 rhythm), scored as percentage correct; primary outcome was percentage MET total score. Gold-MSI questionnaire provided subscales (Active Engagement, Perceptual Abilities, Musical Training, Emotions, Singing, General factor). WM assessed using WAIS-IV Digit Span (forward, backward, sequencing) and Arithmetic; scores scaled and combined into the Working Memory Index (WMI). MRI acquisition: 3T Siemens Prisma-fit. rs-fMRI: TR=700 ms, TE=30 ms, 2.5 mm isotropic, 1500 volumes (two runs). Quantitative MPM for synthetic T1w (FLASH-based), and HARDI with multi-shell acquisition (b up to 2500 s/mm², multiple directions; AP and PA b=0 for distortion correction). Preprocessing: Structural images processed with fMRIPrep (bias correction, skull stripping, segmentation, FreeSurfer recon-all, normalization). rs-fMRI processed with CONN toolbox: realignment/unwarping, ART outlier detection, coregistration, surface mapping, smoothing, denoising (CompCor, motion, outliers, trends), bandpass 0.008–0.09 Hz. Network construction: Cortex parcellated using Destrieux atlas (148 cortical nodes). Functional connectivity: Fisher-transformed bivariate correlations between node time series (148×148). Structural connectivity: dMRI preprocessed (denoising, Gibbs removal, motion/eddy/topup corrections). MSMT-CSD estimated fiber orientation distributions; ACT probabilistic tractography (target 10 million streamlines per connectome), SIFT2 weighting; matrices built using Destrieux cortex and FSL FIRST subcortical segmentations; scaled by SIFT μ. Networks of interest: Two 16-node subnetworks were defined: (1) Frontoparietal network (FPN): bilateral dorsolateral prefrontal (middle/superior frontal gyri and sulci) and posterior parietal regions (inferior/superior parietal lobules, intraparietal sulci). (2) Occipital control network. Graph analysis: Connectivity matrices thresholded using proportional costs k=0.15–0.30 in 0.01 steps; binarized, undirected adjacency matrices created. Node-level graph metrics computed with Brain Connectivity Toolbox: global efficiency (integration), betweenness centrality (centrality), clustering coefficient and local efficiency (segregation). Metrics aggregated across thresholds to reduce dependency on a single threshold. Statistical analysis: Second-level GLMs regressed node metrics on percentage MET total score, controlling for age, sex, and Musical Training Index (Gold-MSI). Node-level p-values FDR-corrected (q<0.05, two-tailed) per metric. Additional GLMs assessed associations between WMI and node metrics (same covariates). Mediation analysis used SEM (lavaan in R) testing rMFG global efficiency → WMI → MET, with bootstrap (1000) for bias-corrected CIs; reported standardized path coefficients and significance. Occipital network served as control in parallel analyses.

• Structural FPN: Higher MET scores associated with greater betweenness centrality in right superior frontal gyrus (SupFG; F=4.40, pFDR=0.0002) and right superior parietal lobule (SupPL; F=3.70, pFDR=0.002). In right SupPL, MET also positively related to global efficiency (F=3.16, pFDR=0.029). MET was negatively related to segregation metrics in right SupFG (local efficiency: F=-4.34, pFDR=0.0003; clustering coefficient: F=-4.78, pFDR=0.00004) and right SupPL (local efficiency: F=-4.21, pFDR=0.0003; clustering coefficient: F=-4.61, pFDR=0.00006). No significant MET associations in occipital control structural network.
• Functional FPN: MET positively correlated with global efficiency of the right middle frontal gyrus (rMFG; F=3.06, pFDR=0.043). No significant MET associations in occipital control functional network.
• WM and MET: WMI correlated with MET (r=0.21, pFDR<0.01).
• WM and FPN topology (functional): WMI positively related to global efficiency in right MFG (F=3.18, r=0.20, pFDR=0.02), right supramarginal gyrus (F=2.70, r=0.19, pFDR=0.03), and right superior frontal sulcus (F=2.71, r=0.17, pFDR=0.03). No significant WMI associations in occipital network or in structural networks.
• Mediation (SEM): rMFG global efficiency → WMI (standardized β=0.21, p=0.002); WMI → MET (standardized β=0.18, p=0.010); direct rMFG → MET (standardized β=0.17, p=0.013); indirect effect rMFG → MET via WMI significant (bootstrap 95% CI [1.180, 12.477]). Total effect significant.
• Emotion: Gold-MSI Emotions subscale correlated with MET, but not with WMI (r=-0.02, pFDR=0.74), and showed no significant associations with FPN/occipital network metrics.

Findings support the hypothesis that frontoparietal network topology marks individual differences in musical perception. Superior musical competence is linked to greater integration and centrality, and reduced segregation, in key right-hemisphere FPN nodes. Functionally, higher global efficiency of the right dorsolateral prefrontal cortex (rMFG) relates to both WM and MET performance, with WM partially mediating the rMFG–music link, suggesting that efficient prefrontal integration supports domain-general processes that facilitate music perception. Structurally, superior parietal lobule integration and centrality predict MET but not WM, aligning with the notion that structural pathways underpin sensory integration and attentional processes essential for music perception. The divergence between functional and structural results reflects their distinct roles: functional networks align with flexible, high-level cognition like WM, whereas structural networks support stable sensory processing routes. Lack of associations in the occipital control network underscores specificity of effects to the hypothesized FPN. The relationship between emotional aspects of musicality and FPN topology appears limited, suggesting partly distinct neural systems (e.g., limbic/reward) underlie affective components.

The study demonstrates that intrinsic communication efficiency and integration capacity within the frontoparietal network serve as neural markers of musical perceptual ability in the general population. Functional topology of right prefrontal regions (rMFG) supports musical competence directly and indirectly via working memory, while structural properties of superior parietal cortex relate to sensory/attentional processes contributing to music perception. These results clarify distinct contributions of functional and structural neurocognitive variability to musicality and provide a framework for future work, including dynamic task-based network analyses, cross-cultural validation, and exploration of broader components of musical ability.

• The study focuses on relatively static, resting-state and diffusion-based network properties; it does not capture dynamic reconfigurations during music listening. Task-based, time-resolved network analyses could reveal transient states linked to music perception.
• Generalizability is limited by a predominantly single-culture European sample; further cross-cultural validation of MET and replication in diverse populations are needed.
• The scope is limited to basic perceptual abilities (melody, rhythm) in listeners; broader aspects of musical ability (emotion processing, creativity, social and motor skills) were not directly tested and may involve distinct networks.
• While neuromodulation (e.g., tDCS) is discussed as a potential avenue, causal interventions were not tested here.