Two fMRI studies, which employed a short-duration piano training inside an MRI scanner, presented somewhat conflicting results (Chen et al., 2012; Brown and Penhune, 2018). In the first one (Chen et al., 2012) the activation of the left dorsal and ventral premotor cortices decreased as the training progressed, as did the activation of the bilateral superior temporal gyrus during a listening task in late compared to early training, while the activation of right retrosplenial cortex, left orbitofrontal cortex and frontal pole showed increased activation in the late vs early phase of training. On the other hand, the study by Brown and Penhune (2018), reported an increased activation of the occipital, temporal, premotor cortex and the cerebellum for late compared to early training. These discrepancies might be attributed to substantial differences in training and testing protocols and contrasts used in the two studies. Nevertheless, the involvement of the premotor regions and the superior temporal gyrus in learning a musical piece is not surprising, as they are parts of the dorsal auditory streams, where sound and movement are integrated into the context of music and speech (Zatorre et al., 2007). The dorsal auditory streams are responsible for sensorimotor integration and spatial processing (Rauschecker and Tian, 2000), which is necessary for the auditory-motor transformations that allow humans to play music. Chen et al. (2012) argued that the reduced activation in brain regions belonging to the dorsal auditory stream suggests increased efficiency in neural processing of a learned stimulus. However, Brown and Penhune (2018) interpret an increase in the activation of the premotor areas across learning trials as a probable reflection of the auditory-motor associations. In other studies, decreased activation of the superior temporal gyrus was associated with perceptual pitch training (Jäncke et al., 2001) and with micromelody discrimination training (Zatorre et al., 2012). Additionally, in the experiment of Brown and Penhune (2018), other notable changes associated with training included decreased activation during training tasks in the parietal cortex, which might facilitate the translation of pitch patterns into movement (Gnadt and Andersen, 1988; Mattingley et al., 1998), and the anterior cingulate cortex, involved in error detection and correction during learning. Interestingly, recent evidence suggests the anterior cingulate cortex activation during listening to dissonant compared to consonant music stimuli (Bravo et al., 2020) and its involvement in singing and playing the cello by professional cellists (Segado et al., 2018).

Lately, Tavor et al. (2020) used DTI approach to study short-term changes in neuroplasticity: immediately before and after each of the two piano training sessions. After the first training session of 45 min, a significant reduction of 2–3% in mean diffusivity (suggesting an increase in white matter organization) was found in the left premotor cortex, the left middle temporal gyrus, and bilaterally in the superior part of the cerebellum. Additional mean diffusivity reduction in the lingual gyrus was found after the second training session of 40 min. These timescales are consistent with astrocyte remodeling (Blumenfeld-Katzir et al., 2011; Sagi et al., 2012; Assaf, 2018), which might be indicative of long-term potentiation, a primary cellular mechanism of neuroplasticity. Previously, structural differences within the white matter of the cerebellum, a known structure facilitating motor learning and movement coordination, were identified in cross-sectional DTI studies comparing musicians and non-musicians (Schmithorst and Wilke, 2002; Abdul-Kareem et al., 2011). Moreover, a larger volume of the lingual gyrus, involved in musical score reading, was found in musicians compared to non-musicians (Bermudez et al., 2009; Sato et al., 2015; Vaquero et al., 2016). As the experiment by Tavor et al. (2020) involved no actual reading of the musical score, and the participants learned to play by viewing and hearing a sequence played on a virtual keyboard, it is more probable that the engagement of the lingual gyrus was related to the processing of complex visual stimuli. However, the authors argue that the plasticity in the lingual gyrus might reflect its direct involvement in the motor or auditory aspects of the tasks (Janata, 2002; Müller et al., 2002; Schmithorst and Holland, 2003; Parsons et al., 2005; Bengtsson et al., 2009). Crucially, the results by Tavor et al. highlight that the brain is an extremely dynamic structure with tissue plasticity occurring even over short timescales.

In recent years, several experiments measured changes on a longer timescale of weeks to months. A study on drums training (Amad et al., 2016), investigated changes in resting-state functional connectivity before and after 8 weeks of learning. Multivariate pattern analysis of the whole-brain connectome identified that drum training was associated with increased functional connectivity between the posterior part of bilateral superior temporal gyrus and the rest of the brain. The bilateral superior temporal gyrus was previously recognized in fMRI studies as involved in action-representation of sounds in musical tasks such as drumming (Tsai et al., 2012) and learning to play the piano in the abovementioned work (Chen et al., 2012; Brown and Penhune, 2018). Follow-up seed-to-voxel analysis revealed increased functional connectivity between the posterior and middle parts of the superior temporal gyrus (seed) and the motor and premotor brain regions and the right parietal lobe. The activation of premotor cortices is coupled with that of the auditory cortex in the music context, showing co-activation during listening to a melody previously learned on a keyboard, but not other melodies (D’Ausilio et al., 2006; Lahav et al., 2007). This co-activation was also observed in professional pianists who listened to musical pieces they knew how to play, and played them without auditory feedback (Zatorre et al., 2007). The coupling between the auditory and the (pre)motor brain areas seems to be involved in music production. The increase in functional connectivity between the posterior superior temporal gyri seed and the right parietal lobe cluster, which belongs to the dorsal attention network, was correlated with the improvement in drum performance. The authors postulate this might be related to the improvement of the integration of sensory and motor functions due to drum training. Moreover, the superior parietal lobule is involved in motor learning, sensorimotor and somatosensory integration, among other sensory and cognitive processes (Zatorre et al., 2007; Wang et al., 2015). It was suggested to play a role in the coordination of spatial and timing components of musical performance (Langheim, 2002) and somatic perception of bimanual object interaction (Naito et al., 2008).

Additionally, functional connectivity decreased between the superior temporal gyrus seed and the cerebellum, what might indicate an increase in efficiency of the cerebellar regions in response to motor learning (Petrini et al., 2011; Vahdat et al., 2011) during drum training.

In another fMRI study (Herholz et al., 2016), participants underwent six weeks of piano training. The activation of the left dorsal premotor cortex increased after training while listening to a familiar melody compared to a random tone sequence. This replicates previous results on the activation of premotor regions when listening to musical pieces one can perform (D’Ausilio et al., 2006; Lahav et al., 2007). Moreover, imagining and listening to trained compared to untrained melodies elicited increased activation in the premotor and dorsolateral prefrontal cortices, posterior parietal cortex bilaterally, and the cerebellum after training. The dorsolateral prefrontal cortex, together with parietal regions, is thought to store representations of learned sequences (Doyon and Benali, 2005; Penhune and Steele, 2012). Increased activation in the parietal lobe stands in opposition to the findings of Brown and Penhune (2018), however, it is hard to interpret this difference given the different paradigms and timescales of these experiments.

More recently, Wollman et al. (2018) investigated brain activation and functional connectivity changes related to 4 weeks of cello training. After the training, increased activation in the supplementary motor areas (pre-SMA & SMA), the right dorsal premotor cortex and the left posterior parietal cortex was observed during a listening task. This functional reorganization, in general, follows the patterns discussed previously, namely the altered response of the premotor and parietal regions of the dorsal stream, occurring after training while listening to trained melodies. Additionally, supplementary motor areas were involved as well, which also belong to the dorsal auditory-to-motor cortical pathway. Similar structures were activated while playing in the scanner as in the listening task, providing further evidence for the close coupling between those areas for musical tasks. Importantly, cello playing and singing by expert cellists evokes overlapping activations in the dorsal premotor cortices, the supplementary and primary motor areas, and the intraparietal sulcus (Segado et al., 2018). Further, resting-state connectivity between the superior parietal lobe and bilateral auditory cortex as well as between the superior parietal lobe and the SMA is positively correlated with performance on musical tasks (Segado et al., 2018). These findings corroborate the outcomes of the drum training experiment (Amad et al., 2016) showing the increased co-activation between premotor and parietal regions, even though widely different instruments were used. In contrast to the studies on piano-training, the activation changes in the frontoparietal network occurred for non-trained melodies but were not seen in the task-related functional connectivity. This difference might stem from the fact that piano is an equal-tempered instrument, and cello is a non-tempered one, where a single pitch can be played in multiple ways. Although no differences were found in brain responses while playing the cello between the first and the fourth week of training, the authors argue that such changes could have occurred during the very first week of practice.

In the longest reported study so far, participants were scanned with resting-state fMRI (rs-fMRI) and DTI before and after a 24-week intensive piano training (Li et al., 2018, 2019). In case of the rs-fMRI, a significant interaction with a control group over time was found in areas of the bilateral postcentral, left superior and right inferior parietal gyrus as well as right precentral gyrus, and right superior frontal gyrus. Additionally, musical training was associated with increased functional connectivity between the right postcentral and the right precentral gyri, and between the auditory and the motor networks. Importantly, functional connectivity within the sensorimotor network and structural connectivity of the auditory-motor network were found to be positively correlated with practice time. These outcomes corroborate previous findings on the changes in the auditory-motor-parietal network related to musical training. Moreover, fractional anisotropy of the corticospinal tract, the superior longitudinal fasciculus, and the corpus callosum after training was increased in the training group compared to controls. Increased fractional anisotropy of the corticospinal tract and the corpus callosum in musicians compared to non-musicians was identified in cross-sectional studies (Schmithorst and Wilke, 2002; Steele et al., 2013; Giacosa et al., 2016; Schlaffke et al., 2020). The dynamics of the resting-state functional connectivity was also analyzed using graph approach, where network flexibility was used as a measure of changes in the local properties of individual elements of a neural network (Li et al., 2019). This analysis yielded similar outcomes. This was the only study on musical training so far to include a follow-up period, which lasted for 3 months after the training was completed. The trainees’ increased connectivity and flexibility related to the 24 weeks of piano training returned to baseline at 12 weeks after the cessation of training (Li et al., 2019, 2018) suggesting that any training-induced alterations are quickly lost if the skills are not used. Although there are no other studies on musical training which included a follow-up period, recent studies on tactile reading (Matuszewski et al., 2021), second language acquisition (Kuper et al., 2021) and sign language learning (Banaszkiewicz et al., 2021) also show a decrease in myelination, but not necessarily in gray matter volume or brain activation, after the cessation of trainings.

Few MEG studies investigated the influence of musical training on brain activation. In each of them, participants were divided into groups for auditory-sensorimotor training (playing the piano, experimental group) or auditory training (listening and judging the correctness of a heard performance, control group). In one of these studies, the training consisted of eight sessions of 25 min over two weeks. MEG recordings were performed before and after training (Lappe et al., 2008; Pantev et al., 2009). A greater increase in mismatch negativity amplitude in the auditory cortex was found after training in the piano training group compared to the control group indicating greater enhancement of auditory cortex’s response to musical stimuli. Similarly, in an experiment by Paraskevopoulos and colleagues (Paraskevopoulos et al., 2012), participants were divided into an auditory-visual-sensorimotor (experimental) and auditory-visual only (control) training groups. Trainees took part in eight sessions spread over 10 days, lasting 58 min each. The interaction with the control group over time showed increased amplitude of the mismatch negativity in the Brodmann’s area 22 (superior temporal gyrus) during an audio-visual task, with a right-hemisphere dominance. Those outcomes confirm the results of previous studies of increased engagement of the auditory association cortex to musical stimuli after musical training (Lappe et al., 2008; Pantev et al., 2009).

To our knowledge, only one longitudinal EEG study on musical training was conducted so far (Bangert and Altenmüller, 2003). In this study with 10 piano training sessions over 5 weeks, auditory-sensorimotor EEG co-activation was observed already after the first session of 20 min. This effect increased after the training was complete. Moreover, piano training was associated with sensory hand area activation during listening. The co-activation of the auditory-sensorimotor network regions while listening to learned musical stimuli has been identified in the previously discussed studies investigating musical training with fMRI (Amad et al., 2016; Li et al., 2018, 2019; Wollman et al., 2018).

Predispositions are factors which are positively associated with the rate at which a new skill is acquired or the skill level attained at the end of the training. Thus, they can strongly influence the time needed to achieve musical expertise or the potentially achievable expertise level, resulting in individual differences in musical skill acquisition. So far, very few studies examined how various predisposing factors affect the outcomes of the learning process in a short time-scale.

Two recent behavioral studies found attitude and intelligence might be indicative of learning success (Bianco et al., 2019; Burgoyne et al., 2019). In the first study (Bianco et al., 2019), as participants were learning to play melodies on a keyboard, the accuracy and asynchrony of keystrokes were taken as measures of motor performance. The better liked melodies were also learned more easily, showing less asynchrony in the learning trials. Moreover, participant’s motivation, reflected in the individual’s task-perceived competence, correlated with the performance on the not-liked melodies. These outcomes are in line with previous findings on the role of reward and pleasure in learning music performance (Schultz, 2000). Namely, participant’s positive attitude toward a certain piece of music before training positively influenced the ability to learn it and achieved performance proficiency. In the second study (Burgoyne et al., 2019), participants underwent a short session of learning how to play Happy Birthday on an electronic piano. General intelligence was found to account for 21.4% of the variance in skill acquisition, while the effects of mindset and aptitude were negligible and non-significant. This result points to the potential role which intelligence plays in the facilitation of early learning.

The topic of predicting participants’ performance after training was a subject of interest in two of the aforementioned longitudinal fMRI studies. In the research of Herholz et al. (2016), distinct neural structures were identified, where stronger recruitment during listening to a known melody before training predicted the slope of participant’s learning curve. The activation of these structures, including the right Heschl’s gyrus, left mid-premotor cortex, bilateral caudate nucleus, and right hippocampus, did not change in response to training; therefore, it can be understood as a specific predictor of learning success. In the context of musical training and auditory cognition, the activation of Heschl’s gyrus and hippocampus can be attributed to the encoding of stimuli, while the activation of premotor cortex and caudate nuclei to aspects of motor control. Heschl’s gyrus is a region commonly associated with auditory pitch processing and discrimination (Patterson et al., 2002), and has been previously identified as a predictor of auditory learning success (Wong et al., 2008; Zatorre et al., 2012). In the cello training study (Wollman et al., 2018), fMRI and DTI were used to identify neural substrates which could predict training success. The activation of the pre-supplementary motor area during passive listening of trained melodies and its functional connectivity with the auditory cortex before training were predictive of subsequent training success after four weeks.

Additionally, in a DTI study, microstructural organization of the right arcuate fasciculus (i.e., volume and fractional anisotropy) before training predicted the learning rate and learning speed in rhythm- and melody-learning tasks (Vaquero et al., 2018). Higher fractional anisotropy in the arcuate fasciculus was associated with musical training in previous studies (Halwani et al., 2011). Reduced volume of the arcuate fasciculus may be associated with congenital amusia (Loui et al., 2009) and predicts non-recovery from acquired amusia after a stroke (Sihvonen et al., 2017).