key: cord-0336085-p3um3qzs authors: Criscuolo, Antonio; Pando-Naude, Victor; Bonetti, Leonardo; Vuust, Peter; Brattico, Elvira title: An ALE meta-analytic review of musical expertise date: 2021-11-12 journal: bioRxiv DOI: 10.1101/2021.03.12.434473 sha: 041ec857227c3df4ec227a4d50ce690e15c4529a doc_id: 336085 cord_uid: p3um3qzs Through long-term training, music experts acquire complex and specialized sensorimotor skills, which are paralleled by continuous neuro-anatomical and -functional adaptations. The underlying neuroplasticity mechanisms have been extensively explored in decades of research in music, cognitive, and translational neuroscience. However, the absence of a comprehensive review and quantitative meta-analysis prevents the plethora of variegated findings to ultimately converge into a unified picture of the neuroanatomy of musical expertise. Here, we performed the first neuroimaging meta-analysis of publications investigating neuro-anatomical and -functional differences between musicians (M) and non-musicians (NM). Eighty-four studies were included in the qualitative synthesis. From these, 58 publications were included in coordinate-based meta-analyses using the anatomic/activation likelihood estimation (ALE) method. This comprehensive approach delivers a coherent cortico-subcortical network encompassing sensorimotor and limbic regions bilaterally. Particularly, M exhibited higher volume/activity in auditory, sensorimotor, interoceptive, and limbic brain areas and lower volume/activity in parietal areas as opposed to NM. Notably, we reveal topographical (dis-)similarities between the identified functional and anatomical networks and characterize their link to various cognitive functions by means of meta-analytic connectivity modelling. Overall, we effectively synthesized decades of research in the field and provide a consistent and controversies-free picture of the neuroanatomy of musical expertise. Decades of research in psychology, cognitive and translational neuroscience have attempted to deepen our understanding of the cognitive and neural processes which allow individuals to reach high levels of mastery within a given domain. For instance, expert musicians (M) develop, through longterm training, complex and specialized auditory and sensorimotor skills, which enables them to achieve the highest levels of performance in playing a musical instrument (thus, the label of "virtuoso"). Notably, the acquisition of such fascinating skills is paralleled by continuous neuroanatomical and -functional changes, representing neuroplastic adaptations to environmental demands. Neuroplasticity generally refers to brain's ability to modulate its anatomical and functional features during maturation, learning, skill acquisition, environmental challenges, or pathology. Interestingly, these effects seem to be salient enough to be also observed macroscopically with magnetic resonance imaging (MRI). However, whole-brain analyses in humans often fail to convey a link between complex behaviour and neuroplasticity mechanisms. Such difficulty might emerge due to methodological and sample differences. For instance, background variables such as genome, psychological, socioeconomical characteristics are not always considered. On the other hand, the adoption of crosssectional (rather than longitudinal) designs and the lack of causal evidence represents another obstacle. Thus, while various research fields have highlighted a link between neuro-anatomical andfunctional changes and the length of specialized trainings (in various populations from athletes, chessplayers, golfers 1 ), any link between expertise and neuroplasticity should be interpreted with caution. Indeed, an alternative explanation is still considered possible, namely that brain differences precede the training and predispose subjects to dive into specific experiences, like musical training 2 . Musical training represents a stimulating experience which engages highly-specialized perceptual, motor, emotional and higher-order cognitive abilities 3 , ranging from multimodal (auditory, visual, motor) sensory perception, integration and predictions as well as fine movement control 4, 5 . Furthermore, it stimulates mnemonic processes associated with the acquisition of long and complex bimanual finger sequences, as well as fine-grained auditory perception (absolute pitch) 6 . The acquisition of such complex and specialized skills is the perfect scenario to investigate neuroplasticity mechanisms in humans and to monitor the continuous underlying neuro-anatomical and -functional adaptations. For instance, audio-motor, parietal and occipital brain structures (the dorsal stream) have been associated to the ability to play an instrument with automatic and accurate associations between motor sequences and auditory events leading to multimodal predictions 4 , the simultaneous integration of multimodal auditory, visual and motor information, fine-grained skills in auditory perception, kinaesthetic control, visual perception and pattern recognition 5 . Dorsolateral prefrontal structures, basal ganglia and mesial temporal structures have further been related to musicians' ability to memorize long and complex bimanual finger sequences and to translate musical symbols into motor sequences 6 . However, other results are often controversial. For instance, while some works reported an overall increase of GM volume when comparing musicians to non-musicians 7, 8 , other research mainly highlighted negative correlations between GM volumes and musical expertise 8,9 . Inconsistencies are also reported for fMRI studies, where musicianship was exclusively associated with stronger activity in, e.g., either premotor cortices 10 , right auditory cortex 11 , or prefrontal cortex 12 , ultimately failing to converge into a common functional network for music expertise. Despite a growing interest in the topic, we here highlight that there has never been a quantitative meta-analytic attempt to summarize existing findings and provide a unified picture of the neuroanatomy of musical expertise. To address this limitation in the field, we conducted the first comprehensive coordinate-based meta-analysis (CBMA) of (f)MRI studies, using the anatomic/activation likelihood estimation (ALE) method 13 , to investigate the neuroanatomical and neurofunctional changes associated with musical training on healthy humans. Specifically, we first provide a detailed overview of the studies included and their methods, paradigms, sample details and backgrounds so to guide the reader into a critical consideration of the results. Then, we characterize the topographical (dis)similarities between the identified functional and anatomical networks and link them to various cognitive functions by means of sub-group analyses and meta-analytic connectivity modelling. A total of 1169 records was identified through database searching, and 679 records were initially screened by title and abstract after removing duplicates. Next, 145 articles were assessed for eligibility in the full-text screening stage. From these, 84 studies fulfilled criteria for eligibility and were thus included in the qualitative synthesis. Finally, 58 studies reported results in stereotactic coordinates, either Talairach or Montreal Neurological Institute (MNI) three-dimensional-coordinate system and were therefore included in the quantitative synthesis (ALE meta-analyses) (Supplementary Figure 1 ). Details of the studies included in our work are provided in Table 1 . Eighty-four publications met inclusion criteria and were included in the qualitative synthesis which comprised of 3005 participants, with 1581 musicians (M) and 1424 non-musicians (NM). Eighteen studies (21%) included amateur musicians, and only 7 studies (8.3%) reported absolute pitch possessors (n = 97). Musical instruments were reported in most of the studies (81%): piano or keyboard (62%), string instruments (41%), wind instruments (26%), percussion instruments (17%), voice (8%), and 19% studies failed to report musicians' instrument. Years of education was described only in 8% of the included studies. Years of musical training was reported in 63% of the studies, with a mean of 15.6 ± 5.9 years. The age of onset of musical training was reported in 49% of the studies, with a mean of 7.4 ± 2.3 years old. Weekly hours of training were reported in 32% of the studies, with a mean of 16.7 ± 8.9 hours per week. Insert Table 1 . MRI quality of the included studies in the meta-analysis was assessed following a set of guidelines for the standardized reporting of MRI studies 14, 15 . All studies reported their MRI design, software package and image acquisition, pre-processing, and analyses. Overall, all the included studies followed good MRI practices. Neuroimaging data was acquired in either 1.5 T (39%), or 3 T (56%) scanners, while 5% of studies did not report the magnetic field strength. MRI scanners included Siemens (40%), General Electric (25%), Philips (21%), Bruker (5%), while 7% did not report it. Analysis methods included fMRI (52%), VBM (29%), DTI (18%), and CT (10%). Finally, 84% of studies included GM analyses, while 23% included WM analyses (Supplementary Tables 1, 2 , and 3). The quantitative synthesis of the primary outcome included 58 publications with 675 foci, 79 experiments and a total of 2,780 participants. Separate ALE meta-analyses were conducted for structural and functional foci, focusing on the comparison between M and NM. The structural ALE meta-analysis included 33 experiments and 1515 participants. The contrast M > NM in GM resulted in significant peak clusters in the bilateral superior temporal gyrus (primary auditory cortex), including the bilateral Heschl's gyrus and planum temporale, and the postcentral gyrus (somatosensory cortex, SI), including area 4a of the primary motor cortex (M1 4a). Conversely, the comparison NM > M in resulted in a significant peak cluster located in the right precentral gyrus (primary motor cortex, M1). In WM, musicians showed larger tracts of the internal capsule bundle (extending to the thalamus) and corticospinal tract. No significant clusters were identified in the comparison NM>M for WM (Figure 1, Table 2 ). The functional ALE meta-analysis included 46 experiments and 1265 participants. The contrast M > NM resulted in large and significant peak clusters of the bilateral superior temporal gyrus (the bilateral primary auditory cortices) extending to the left insula, the left inferior frontal gyrus, and the left precentral gyrus (primary motor cortex, M1). The comparison NM > N resulted in smaller peak clusters of the left inferior parietal lobule and the left precentral gyrus (Figure 1, Table 2 ). Insert Figure 1 Insert Table 2 2.4. Meta-analytic connectivity modelling (MACM). MACM was performed to functionally segregate the behavioural contribution and the patterns of coactivation of each music-related region-of-interest (ROI) resulted from the primary outcome (n=10). The ROIs were imported into the BrainMap database separately, to identify studies reporting activation within each ROI boundary (Supplementary Table 4 ). The functional characterization of each ROI is detailed in Supplementary Table 5 and include the behavioural domains of action, perception, emotion, cognition and interoception. and emotional processing; and experimental paradigms including finger tapping and visuospatial attention. The right internal capsule ROI (a5), including the right thalamus as the nearest grey matter, showed co-activation with left thalamus, right medial frontal gyrus, right insula, and left cerebellum. Relevant behavioural domains within its boundaries include execution, speech, attention, memory, reasoning, emotion, reward, and auditory perception; and experimental paradigms including emotion induction, finger tapping, passive listening, reward, and tone discrimination. The left inferior frontal gyrus ROI (b1) showed co-activation with right inferior frontal gyrus, left inferior and superior parietal lobules, left medial frontal gyrus, left fusiform gyrus, and right caudate. Relevant behavioural domains within its boundaries include execution, speech, attention, language, memory, music, reasoning, social cognition, emotion, and auditory perception; and experimental paradigms including encoding, finger tapping, music comprehension and production, passive listening, reasoning/problem solving, reward, and phonological, semantic, tactile, and tone discrimination. The right superior temporal gyrus ROI (b2) showed co-activation with bilateral inferior frontal gyrus, bilateral inferior parietal lobule, left medial frontal gyrus, left fusiform gyrus, and caudate. Relevant behavioural domains within its boundaries include execution, speech, attention, language, memory, music, reasoning, social cognition, emotions, and auditory perception; and experimental paradigms including finger tapping, music comprehension and production, passive listening, reasoning/problem solving, reward, theory of mind, and phonological, semantic, tactile, and tone discrimination. The left superior temporal gyrus ROI (b3) showed co-activation with right superior temporal gyrus, right middle temporal gyrus, right claustrum, left insula, and left medial frontal gyrus. Relevant behavioural domains within its boundaries include execution, speech, attention, language, memory, reasoning, social cognition, emotion, and auditory perception; and experimental paradigms including divided auditory attention, emotion induction, emotional body language perception, encoding, finger tapping, music comprehension and production, reasoning/problem solving, theory of mind, visuospatial attention, and oddball, phonological, pitch, semantic, and tone discrimination. The left inferior parietal lobule ROI (b4) showed co-activation with left medial frontal gyrus, right inferior frontal gyrus, and left precentral gyrus. Relevant behavioural domains within its boundaries include execution, speech, motor learning, attention, memory, music, reasoning, social cognition, emotion, and auditory perception; and experimental paradigms including emotion induction, finger tapping, motor learning, reasoning/problem solving, reward, visuospatial attention, and phonological, semantic, tactile and tone discrimination. The left precentral gyrus ROI (b5) showed co-activation with left precuneus, left superior frontal gyrus, right inferior frontal gyrus, right inferior parietal lobule, right claustrum, left fusiform gyrus, left thalamus, and right middle frontal gyrus. Relevant behavioural domains within its boundaries include execution, speech, motor learning, attention, language, memory, music, reasoning, social cognition, temporal processing, emotion, sleep, and auditory perception; and experimental paradigms including divided auditory attention, emotion induction, encoding, finger tapping, music comprehension, reasoning/problem solving, reward, theory of mind, visuospatial attention, and oddball, phonological, pitch, semantic, tactile, and tone discrimination. The link between musical expertise and humans' cognitive functions has been explored since the times of Pythagoras. Recent years reveal a more that vivid interest in the topic, as reflected in the rising number of research in the past half-century. Indeed, decades of investigations in psychology, cognitive and translational neuroscience have attempted to foster our understanding of the cognitive and neural processes underlying musical expertise. Thus, long-term musical training has been associated with neuro-anatomical and -functional specializations in brain regions engaged in multimodal (auditory, visual, motor) sensory perception, integration and predictions as well as fine movement control 4, 5 . However, this rapid growing field of research is also characterized by methodological and sample differences, discrepancies in the results and controversial interpretations of the findings. Such limitations, alongside with the absence of a comprehensive meta-analysis, prevent the plethora of variegated findings to ultimately converge into a unified picture of the neuroanatomy of musical expertise. To address this lack in the literature, we performed the first existing comprehensive and quantitative meta-analysis of neuro-anatomical and -functional studies investigating brain changes associated with musical training. Our coordinate-based anatomic/activation likelihood estimation (ALE) meta-analysis effectively summarizes decades of research in the field and provides, for the first time, a consistent and controversies-free picture of the neuroanatomy of musical expertise. Notably, we further characterize the topographical (dis)similarities between the identified functional and anatomical networks and link them to various cognitive functions by means of meta-analytic connectivity modelling. The publications included in this systematic review and meta-analysis reported a clear research question, inclusion and exclusion criteria for participants, description of methods and explicit results. Most of the studies used state-of-the-art techniques and computational MRI-tools, important for the support of standardization and reproducibility of neuroimaging studies. However, some of the studies lacked important demographic data such as the years of education, age of musical training onset, and current time of musical practice, which may influence behavioural tasks and neuroimaging data. Thus, our research encourages to adopt in future studies standardized tools specifically designed and validated for assessing musical expertise 16 . Our results highlight that expert musicians exhibited higher GM volume in the bilateral superior temporal gyri and right postcentral gyrus and greater WM volume in the right internal capsule bundle and corticospinal tract, as compared to non-musicians. Additionally, musicians exhibited higher activity of the bilateral superior temporal gyri, left inferior frontal gyrus, left precentral gyrus, and left insula. On the other hand, musicians had lower GM volume in areas of the sensorimotor cortex and no WM structure was found to have larger volume in non-musicians as compared to musicians. Finally, musicians exhibited lower neurofunctional activation of the inferior parietal lobule and motor cortex during a variety of cognitive tasks. One of our main findings shows enlargement of GM volume in musicians located in medial and posterior superior temporal regions, with clusters extending into primary and secondary auditory cortices. These regions include neuronal assemblies dedicated to encoding of spectro-temporal features of sounds relevant to music 17 , such as the discrete pitches forming the Western chromatic scale and fine changes in pitch intervals 18 . More specifically, it seems that the posterior supratemporal regions are more involved in encoding the height of pitch, whereas the anterior regions are representing the chroma, that is the pitch category irrespectively of the octave 19 . Moreover, these areas participate in auditory imagery of melodies 20 and in the processing of the contour and Gestalt patterns of melodies, allowing for recognition and discrimination of mistakes 21 . Beyond music-related functions, superior temporal regions are recruited for phonological processing and multimodal integration of sensory information. Accumulating evidence has shown that STS and posterior STG, together with early auditory regions (HG), are involved for the processing of speech sounds, abstract representation of speech sounds and phonemes, and audio-visual integration mechanisms. Therefore, temporal regions seem to represent fundamental structures for both language and music processing 22 . The inferior frontal gyrus has been described as an important hub of both the dorsal and ventral auditory streams. The dorsal auditory stream connects the auditory cortex with the parietal lobe, which projects in turn to the inferior frontal gyrus pars opercularis (Brodmann area 44). This area has been related to the articulatory network, dedicated to specific functions of speech comprehension and production, and highly connected to premotor and insular cortices 23 . The ventral auditory stream connects the auditory cortex with the middle temporal gyrus and temporal pole, which in turn connects to the inferior frontal gyrus pars triangularis (Brodmann area 45). This area has been associated with semantic processing 24 . These two regions within the inferior frontal gyrus constitute Broca's area. The supramarginal gyrus is also a relay of the dorsal auditory stream involved in processing of complex sounds, including language and music 25 . As such, it is considered an integration hub of somatosensory input 26 . The parietal lobe has been also described as an integration area of sensory inputs. The superior parietal lobule includes Brodmann areas 5 and 7, which are involved in somatosensory processing and visuomotor coordination, respectively. The inferior parietal lobule includes Brodmann areas 39 and 40, the angular gyrus and supramarginal gyrus, respectively. The angular gyrus has been related to projection of visual information to Wernicke's area, memory retrieval and theory of mind 27 . The precentral and postcentral gyrus represent the primary motor and somatosensory cortex, respectively. These two areas are divided by the central sulcus, whose extension represent the sensation and motion of segregated body parts. Our findings show both convergent and divergent effect of musical training in these areas, suggesting a more complex picture than previously thought. For example, neuroadaptations in the sensorimotor system may vary depending on the musical instrument of use 28 . The basal ganglia are nuclei of neurons important for the initiation and suppression of movements. In the motor loop of the basal ganglia, inputs from motor cortices project to the dorsal striatum, composed by the putamen and caudate. In the presence of adequate dopaminergic signaling, the direct pathway works to facilitate movement, while the indirect pathway suppresses it. By effective disinhibition processes, the striatum transiently inhibits the pallidum, and in turn, the motor area of the thalamus is disinhibited and free to project back to the motor cortex and initiate a motor program down the corticospinal tract. Similarly, the subthalamic nucleus in the indirect pathway is transiently inhibited when suppressing movement, increasing the inhibition of the pallidum over the thalamus, therefore blocking the motor cortex activity 27 . Our findings show neuroadaptive processes in the putamen and caudate of musicians, presumably resulting in enhancement of such disinhibition mechanisms reflected in fine movement control. The cerebellum has been shown to play a crucial role in multiple cognitive processes such as sensory discrimination, rhythmic perception and production, working memory, language, and cognition 28 . Previous fMRI studies in humans suggest that the cerebellum shows segregated activations for motor and cognitive tasks. Motor tasks seem to activate lobules IV-VI in the superior parts of the anterior cerebellum. In contrast, attentional or working memory tasks activate posterior cerebellar hemispheres, namely lobule VIIA, which is divided to crus I and crus II, as well as lobule VIIB 29 . Musicians and non-musicians show GM volume differences in the cerebellum, specifically in area Crus I. In our study, this area did not survive correction for multiple comparisons, however MACM revealed that the cerebellum is functionally connected to auditory cortices, somatosensory cortices, and the thalamus. It has been demonstrated that the activity in crus I/II has a specific relationship with cognitive performance and is linked with lateral prefrontal areas activated by cognitive load increase 30 . In other words, the crus I/II seems to optimize the response time when the cognitive load increases. Additionally, it has been suggested that crus I/II is associated with beat discrimination thresholds. Thus, there is a positive correlation between GM volume in crus I and beat discrimination performance, evidenced by enhanced ability in musicians 31 . It has been proposed that the insula and the anterior cingulate cortex (ACC) are part of the salience network, and coordinate interactions between the default-mode network and the central executive network 34 . The ACC has been related to cognitive and emotional processing. The cognitive component projects to prefrontal, motor, and parietal areas to process top-down and bottom-up stimuli. The emotional component features connections from the insula to amygdala, nucleus accumbens, hypothalamus and hippocampus, with the scope to assess the salience of emotional and motivational information 35 . Moreover, the insula integrates information from the internal physiological state, and projects to the ACC, ventral striatum and prefrontal cortex to initiate adaptive responses 36 . Thus, enhanced function of these areas after musical training may be associated with a more efficient coordination between interoceptive, emotional, salience and central executive networks. M exhibited larger clusters of WM as compared to NM in the internal capsule, corpus callosum, longitudinal fasciculus, and thalamic radiations. The internal capsule is a WM structure which connects basal ganglia regions and carries information from and to surrounding cerebral cortex. Connecting fibres in basal ganglia might be thickened by musical expertise because of their involvement in motor control, rhythmic processing, sequence learning, reinforcement learning and memory processes 37 . In general, basal ganglia structures are recruited during working memory processing for musical motifs 38 and the most ventral regions are a core structure of the reward circuit. Interestingly, they are found to be more active in musicians as compared to non-musicians while listening to expressive music 39 . Lastly, increases in the CC volume may be related to the facilitation of necessary communications between hemispheres and functional networks underlying the coordination of complex sequential bimanual motor sequences and the recall of stored motor programs 38 . This comprehensive review and meta-analysis had the scope to summarize decades of research investigating neuro-anatomical and -functional changes associated with musical training. Our qualitative review highlights that previous studies in this field are characterized by heterogeneity of methods, paradigms, and sample backgrounds, as well as relevant missing information. While arguing that the field will benefit from more clarity (e.g., thorough description of methods) and consistency, we also delineate limitations for our meta-analysis. For example, we set a contrast based on the comparison M vs NM with the aim to narrow down the heterogeneity of the sample and methods in use. However, by doing so we relied on two assumptions: (1) the data we pool is based on best research practices; (2) the validity of the GingerALE method. Indeed, to conduct the ALE meta-analysis, we pooled peak coordinates derived from the included studies, rather than using original raw structural MRI images. Thus, the accuracy of our findings relies on the result of a statistical estimation of coordinate-based anatomic foci (input), treated as spatial probability distributions centred at the given coordinates. The heterogeneity of the methods in use in previous studies (ranging from preprocessing software, smoothing, statistical thresholds and participants' characteristics) are not under our control and represent potential confounders for the results. Perhaps a regression-based assessment of the influence of those heterogenous factors on the findings would sharpen the results. However, meta-regression analysis is not compatible with GingerALE. When assessing publication bias using the Fail Safe-N analysis, we found adequate robustness of our results, with only 2 ROIs showing an FSN below of the minimum imposed in each CBMA, thus, indicating a robust convergence of foci in these regions (further information is reported in Supplementary Table 6) . Lastly, on a more theoretical perspective, our results contribute but do not solve the long-standing "nature vs nurture" debate. Indeed, based on evidence that musical training stimulates highercognitive functions, auditory-motor integration, attention, memory and engages reward networks, some have suggested that it may be particularly effective in driving neuroplastic mechanisms 41 . However, we are indeed blind to whether the highlighted differences emerging when comparing M Vs NM are training-dependent or due to innate predispositions. Altogether, the most reasonable conclusion is that the observed neuro-anatomical and -functional changes may be attributed to the interaction between brain maturation processes and the environmental demands of musical training 13, 44, 50 . Notably, multiple studies demonstrated a strong correlational link between the length of musical training and neuro-functional-anatomical changes. For instance, the study conducted by Gaser & Schlaug 44 reported that amateur musicians showed an intermediate increase in gray matter volume when compared to NM and M, supporting the idea of use-dependent structural changes. The same pattern was found when comparing cognitive abilities, with amateurs showing higher cognitive abilities than NM, but lower than M 45 . To be noted, however, this research field suffers of the paucity of longitudinal (f)MRI studies conducted with children, which thus far amount only to seven [46] [47] [48] [49] [50] [51] [52] . Longitudinal studies are the only ones promising to better elucidate on the causal link between musical training and neural adaptations. The musician's brain has been repeatedly suggested as an ideal example of neuroplasticity mechanisms. Yet, decades of research in cognitive neuroscience have provided a scattered and partially controversial series of findings. The present coordinate-based meta-analysis represents the first comprehensive and quantitative attempt to finally summarize existing literature and provide a unified picture of the neuroanatomy of musical expertise. We show that music experts exhibit corticosubcortical and bilateral neuro-anatomical and functional differences as compared to laypersons, challenging the simplistic view of left-lateralization for music-related brain functions. This comprehensive work strengthens the view that musical training represents a beneficial and stimulating multisensory experience which engages a wide variety of brain structures and associated cognitive functions. This systematic review and meta-analysis followed procedures from the Cochrane Handbook for Systematic Reviews 53 and from the Centre for Reviews and Dissemination (Centre for Reviews and Dissemination, 2014). The review protocol was registered with PROSPERO No. [CRD42017060365]. This review was carried in accordance with the PRISMA statement 54 . Systematic search was performed using PubMed, PsycInfo and Scopus, of publications that reported brain structural or functional differences between M and NM. The search (March 2021) included MeSH terms ("music", "education", "brain", "motor skills", "magnetic resonance imaging") and key words ("musical training", "musician"). No years or places of publication were imposed. For qualitative synthesis, studies were included if they met the following criteria: (1) studies comparing brain structure and function between musicians and non-musicians, (2) in adult population, (3) by means of magnetic resonance imaging, in either structural modality (e.g., voxel-based morphometry [VBM]) or functional modality (e.g., functional magnetic resonance imaging[fMRI]). For quantitative synthesis (meta-analysis), studies were included if results were reported in stereotactic coordinates either Talairach or Montreal Neurological Institute (MNI) three-dimensional-coordinate system. Studies were excluded using the following criteria: (1) review articles with no original experimental data, (2) neuroimaging data from non-MRI studies (e.g., PET), (3) pathological population, (4) longitudinal designs, (5) functional connectivity analyses, and (6) analyses based on region-of-interest (ROI) rather than whole-brain (only quantitative synthesis). Two reviewers (AC and VP) independently screened by title and abstract and selected articles for fulltext review and performed full-text reviews. Screening and data extraction were performed using the Covidence tool 55 . Any disagreements that arose between the reviewers were resolved through discussion or by a third and/or fourth reviewer (LB, EB). From each study, the following variables were extracted: first author, year of publication, population of interest, number of participants, age, sex, absolute pitch, musical feature, years of education, years of musical training, age of musical training onset, weekly training, musical instrument, MRI-system, MRI-model, head-coil, image acquisition parameters of T1, T2* and DWI sequences, repetition time (TR), echo time (TE), voxel size, analysis method and software. The main outcome to extract was any difference in structure or function, in stereotactic coordinates, comparing a musician group and a nonmusician group. If any of these points were not reported in the original article, authors were contacted to retrieve this information. Six authors were contacted, with 2 positive answers. Criteria for MRI quality reporting was selected from a set of guidelines for the standardized reporting of MRI studies 14, 15 . Such guidelines dictate a more consistent and coherent policy for the reporting of MRI methods to ensure that methods can be understood and replicated. To test the convergence of findings from the neuroimaging studies, we used the anatomic/activation likelihood estimation (ALE) method implemented in the GingerALE software v3.0.2 13 , a widely used technique for coordinate-based meta-analysis of neuroimaging data. Statistically significant foci from between-group contrasts were extracted and recorded for each study. If necessary, coordinates were converted from Talairach coordinates to MNI space using the Lancaster transform (icbm2tal) incorporated in GingerALE 56, 57 . The ALE method uses activation foci (input) not as single points, but as spatial probability distributions centred at the given coordinates. Therefore, the algorithm tests to what extent the spatial locations of the foci correlate across independently conducted MRI studies investigating the same construct and assesses them against a null distribution of random spatial association between experiments 46 . Statistical significance of the ALE scores was determined by a permutation test using cluster-level inference at p < 0.05 (FWE), with a cluster-forming threshold set at p < 0.001. The primary outcome was to identify brain structural and functional differences measured by MRI between musicians (M) and non-musicians (NM), to examine comprehensively the neural signatures of musical expertise. To test the directionality of the primary outcome, foci were pooled reporting higher volume/activity in musicians (M > NM) and lower volume/activity in musicians (NM > M) for both structural and functional studies, independently. Meta-analytic connectivity modelling (MACM) was performed to analyse co-activation patterns of regions-of-interest (ROI) resulting from the primary outcome, aiming to functionally segregate each region's putative contribution to behavioural domains 58, 59 . Co-activation analyses were performed using Sleuth 60 and GingerALE from the BrainMap platform. To identify regions of significant convergence, an ALE meta-analysis was performed over all foci retrieved after searching Sleuth by each music-related ROI independently and included the experiment level search criteria of "context: normal mapping" and "activations: activation only". Music-related ROIs were created in Mango 61 with a 5mm-radius sphere. The functional characterization of music-related ROIs was based on the "Behavioural Domain" (BD) meta-data categories available for each neuroimaging study in the BrianMap database which include action, perception, emotion, cognition and interoception. All meta-analytic results (ALE maps) were visualized using Mango on the MNI152 1mm standard brain, and resulting coordinates were cross-referenced to the Harvard-Oxford Cortical and Subcortical Atlas and the Juelich Histological Atlas via NeuroVault 62 and FSLeyes 63 , respectively. Finally, as all meta-analyses, coordinate-based meta-analyses such as ALE can be subject to different forms of publication bias which may impact results and invalidate findings (e.g., the "file drawer problem"). Thus, the Fail-Safe N analysis (FSN) 64 was performed as a measure of robustness against potential publication bias. It refers to the amount of contra-evidence that can be added to a metaanalysis before the results change and can be obtained for each cluster that survives thresholding in an ALE meta-analysis. For normal human brain mapping, it is estimated that a 95% confidence interval for the number of studies that report no local maxima varies from 5 to 30 per 100 published studies. Using the upper bound and the fact that the CBMA's consist of 33 structural studies and 46 functional studies, an estimate for the number of unpublished studies is 10 and 14, respectively. Therefore, the minimum FSN was defined as 10 for structural studies and 14 for functional studies. A higher FSN indicates more stable results and hence a higher robustness. AC, VPN and EB designed the meta-analysis. VPN guided AC in the initial steps of the meta-analysis after which AC and VPN conducted the screening of the studies, and EB and LB controlled the final selection. AC performed the initial analyses, wrote the first draft of the manuscript, and prepared the initial versions of the tables and figures. VPN performed all analyses, prepared the figures and tables and edited paragraphs in the Methods, Results and Discussion sections. All authors contributed to the final version of the manuscript. PV contributed to financially support the study. The data supporting the findings of this study is freely available at the Open Science Framework (OSF) website https://osf.io/5ekqr/?view_only=4416037a1b164e6287d95e7f24dd0a0a ---------Pro Y 5 Angulo-P 2014 fMRI timbre 28 25 -29 24 53 a,b 28 7 29 9 ----------Pro Y 6 Bailey 2014 VBM+DBM structure 30 20 --- Y 63 Petrini 2011 fMRI audiovisual+motor 11 11 -22 0 22 d 35 12 35 11 ----24 11 ----Pro Y 64 Rüber 2015 DTI structure 20 10 -16 14 ------4 ---Pro Y 67 Schlaffke 2020 fMRI+DTI motor+structure 20 24 -44 0 44 d --------17 6 --10.5 8 Pro N 68 Schlaug 1995a VBM structure 30 30 -44 16 ---------Ama N 72 Schmithorst 2003 fMRI melody+harmony 7 8 -11 4 15 ---------------Ama N 73 Schmithorst 2004 fMRI math 7 8 -11 4 15 ---------------Ama Y 75 Schneider 2002 VBM structure 16 8 -16 8 24 a ------------- Changes in grey, matter induced by training How musical training affects cognitive development: rhythm, reward and other modulating variables Plasticity of the human auditory cortex related to musical training Art and science: how musical training shapes the brain The brain of musicians. A model for functional and structural adaptation Structural neuroplasticity in expert pianists depends on the age of musical training onset Musical training intensity yields opposite effects on grey matter density in cognitive versus sensorimotor networks Regional cerebellar volumes are related to early musical training and finger tapping performance A network for audio-motor coordination in skilled pianists and nonmusicians Subcortical and cortical correlates of pitch discrimination: Evidence for two levels of neuroplasticity in musicians Moving on time: Brain network for auditory-motor synchronization is modulated by rhythm complexity and musical training Coordinate-Based Activation Likelihood Estimation Meta-Analysis of Neuroimaging Data: A Random-Effects Approach Based on Empirical Estimates of Spatial Uncertainty Guidelines for reporting an fMRI study Best practices in data analysis and sharing in neuroimaging using MRI Measuring the facets of musicality: The Goldsmiths Musical Sophistication Index (Gold-MSI) Structural and functional asymmetry of lateral Heschl's gyrus reflects pitch perception preference Musical scale properties are automatically processed in the human auditory cortex Analyzing Pitch Chroma and Pitch Height in the Human Brain There's more to auditory cortex than meets the ear Evidence for the role of the right auditory cortex in fine pitch resolution Neural overlap in processing music and speech Music and language side by side in the brain: a PET study of the generation of melodies and sentences Phoneme and word recognition in the auditory ventral stream Mechanisms and streams for processing of Phonological decisions require both the left and right supramarginal gyri The angular gyrus: Multiple functions and multiple subdivisions Structural, functional, and perceptual differences in Heschl's gyrus and musical instrument preference Activation of a cerebellar output nucleus during cognitive processing Cerebellar Transcranial Magnetic Stimulation Impairs Verbal Working Memory Cognitive and Motor Loops of the Human Cerebro-cerebellar System Rhythm and Beat Perception in Motor Areas of the Brain Saliency, switching, attention and control: a network model of insula function Cognitive and emotional influences in anterior cingulate cortex Passive music listening spontaneously engages limbic and paralimbic systems Distinct basal ganglia territories are engaged in early and advanced motor sequence learning Functional architecture of verbal and tonal working memory: An fMRI study. Hum. Brain Mapp It's sad but I like it: The neural dissociation between musical emotions and liking in experts and laypersons Action in perception: Prominent visuo-motor functional symmetry in musicians during music listening Early musical training and white-matter plasticity in the corpus callosum: evidence for a sensitive period Cortical thickness maturation and duration of music training: Healthpromoting activities shape brain development Early Musical Training is lnked to gray matter Structure in the Ventral Premotor Cortex and Auditory-motor Rhythm Sybcrhonization performance Brain Structures Differ between Musicians and Non-Musicians On the Association Between Musical Training, Intelligence and Executive Functions in Adulthood Can you hear a difference? Neuronal correlates of melodic deviance processing in children Increased engagement of the cognitive control network associated with music training in children during an fMRI Stroop task Effects of Music Training on Inhibitory Control and Associated Neural Networks in School-Aged Children: A Longitudinal Study Neural Dynamics of Improved Bimodal Attention and Working Memory in Musically Trained Children Music training and child development: a review of recent findings from a longitudinal study Childhood Music Training Induces Change in Micro and Macroscopic Brain Structure: Results from a Longitudinal Study The effects of musical training on structural brain development: A longitudinal study Cochrane Handbook for Systematic Reviews of Interventions Preferred Reporting Items for Systematic Reviews and Meta-Analyses: The PRISMA Statement Available at www.covidence.org Activation likelihood estimation meta-analysis revisited Comparison of the disparity between Talairach and MNI coordinates in functional neuroimaging data: Validation of the Lancaster transform Networks of task co-activations Metaanalytic Connectivity Modeling: Delineating the Functional Connectivity of the Human Amygdala. Hum. Brain Mapp The Social Evolution of a Human Brain Mapping Database Automated regional behavioral analysis for human brain images org: a web-based repository for collecting and sharing unthresholded statistical maps of the human brain Assessing robustness against potential publication bias in Activation Likelihood Estimation (ALE) meta-analyses for fMRI The Center for Music in the Brain (MIB) is supported by the Danish National Research Foundation (grant number DNRF 117). The authors wish to thank Hella Kastbjerg for assistance with language check and proof reading. The authors declare no conflict of interest for this work.