The Neural Mechanism underlying Music Perception: A Meta-analysis of fMRI Studies

Abstract

Music is part of the human nature. Music perception involves a series of hierarchical processing levels, including auditory feature extraction, gestalt formation, interval analysis, and structure analysis. Previous functional magnetic resonance imaging (fMRI) studies have identified multiple cortical regions involved in different levels of music perception, but these studies yielded inconsistent findings possibly because different research paradigms and data analysis methods were used. Therefore, the neural mechanism underlying music perception is unclear. In this study we performed a meta-analysis to identify brain regions stably recruited by music perception. Specially, we focused on two processing levels specific to music perception, which were interval analysis and structure analysis. In addition, to explore the hierarchical structure of music perception, we examined the possible overlapping and dissociation of the regions involved in interval analysis and structure analysis. We used meta-analysis approach to re-analyze results from fMRI studies on interval analysis and structure analysis. There were eight studies on interval analysis which included 15 contrasts and 63 peaks, and ten studies on structure analysis which included 19 contrasts and 217 peaks. The coordinates of peak voxels reported in these studies were projected onto a brain template to visualize the distributions of activations recruited by interval analysis and structure analysis respectively. To identify brain regions stably activated by music perception, peaks of each analysis level were segregated into a number of spatially distinct clusters, using a hierarchical classification algorithm that minimized the spatial extent within each cluster while maximizing the peak-to-peak distance between clusters. Then, we calculated mean coordinates for each cluster in the MNI space and used FSL View procedure to identify the anatomical label of each cluster. Finally, to explore whether clusters belonging to the two analysis levels were spatially overlapped or dissociated, we examined the pairs of clusters with distances less than 7mm (about 3 voxels) by K-means clustering. The meta-analysis identified 12 clusters for interval analysis and 29 clusters for structure analysis. Clusters for interval analysis primarily localized in the superior temporal gyrus (STG) with the peaks distribution rate (PDR)being 43%, while clusters for structure analysis mainly localized in the prefrontal cortex with a PDR of 34%. Although both interval analysis and structure analysis involved frontal, temporal, parietal and insular areas, their activations overlapped only in the posterior portion of the superior temporal gyrus (STGp). In sum, the meta-analysis suggests that interval analysis and structure analysis are two separate processing levels in the hierarchical structure of music perception, with largely dissociated neural activations. In addition, the two analysis levels intersected only in STGp, which might play a role in information exchanges between interval analysis and structure analysis. Thus, our study provides clues for future researches on neural basis underlying hierarchical structure of music perception.