Musical Memory in Alzheimer’s disease: Underlying mechanisms and therapeutic implications Marin C. Beims 11633514 Literature Thesis University of Amsterdam Master of Brain and Cognitive Sciences
Track: Behavioural Neuroscience
Evgenia Salta 31.03.2021
Contents
Introduction ... 5
Music as a Physiological Substrate ... 5
Alzheimer’s Disease: A Highly Complex, Incurable Disorder ... 5
Pathophysiology of AD... 5
Current Landscape of Alzheimer Disease Therapeutic Strategies ... 6
Alzheimer’s Disease as Multifactorial and Heterogenous Disease ... 7
Music in The Brain ... 8
Music Perception: From Decoding to Processing ... 8
Music Processing: From Computation to Meaning ... 9
Functional Magnetic Resonance Imaging (FMRI) Evidence ... 9
Brain Lesion Studies ... 10
Musical Pleasure ... 12
Music as Therapy ... 14
Musical Memory Systems ... 15
Different Musical Memory Systems Are Encoded in Distinct Neural Networks ... 16
Musical Memory in Alzheimer’s Disease ... 17
Neural Networks of Musical Memory in Alzheimer’s Disease ... 17
Preservation of Musical Memory: Anatomical Evidence ... 18
Anatomical Differentiation Hypotheses ... 18
Network Distribution Hypotheses ... 21
Preservation of Musical Memory: Behavioural Evidence ... 23
Semantic Musical Memory ... 23
Semantic Musical Memory is Distinct from Semantic Verbal Memory ... 24
Episodic Musical Memory ... 25
Procedural Musical Memory ... 25
Protective Effect of Music on AD ... 28
Structural Plasticity Induced by Musical Practice ... 29
Functional Plasticity Induced by Musical Practice ... 30
Music as an Enhancer of Brain and Cognitive Reserve in AD ... 32
Music as a Therapeutic Application in Alzheimer’s Disease ... 33
Music as Life Enhancer ... 33
Underlying Neurobiological Mechanisms of Music Therapy ... 39
Dopamine ... 39
Default Mode Network ... 40
Conclusions ... 42
Challenges and Limitations ... 42
Abstract
The increasingly ageing population demands the development of novel treatment and prevention programs for age-related neurodegenerative disorders, such as Alzheimer’s disease (AD). Music has a unique effect on the brain by engaging a wide network of brain structures related to auditory, memory, attentional, emotional, motor and sensory processing. Current literature demonstrates that various forms of musical memory are preserved in AD patients. Further, musical training is found to increase brain plasticity and heighten network activity yielding a protective effect in the brain against age-related neurodegeneration. Based on the current evidence this review highlights the important implications musical memory preservation and protection have for enhancing the quality of life of patients. Emerging evidence proposing a musical therapy approach to be adopted in the treatment of AD is critically discussed.
Introduction Music as a Physiological Substrate
Music is a ubiquitous and universal feature unique to humans. Music has been found to be practiced and enjoyed by societies all over the world throughout history. In the past, music has often been studied as a cultural invention, produced and refined by culture-specific exposure (Schönberg, 1984). The notion of music as a biological function, underlying physiological substrates in the brain is relatively recent and therefore far from established (Peretz, 2006). However, initial research to unravel the neural substrates of music processing reveals that music has the ability to activate a vast amount of structures in the human brain, including auditory systems, motor systems, limbic and paralimbic structures, as well as frontal cortical areas. Further, for a musical experience to take place, a complex integration of the information generated by these structures, is required, making music a highly complex and active cognitive process. The discovery of music’s unique role in the brain, has led research to using music to investigate neurological functions in healthy, as well as diseased brains. In this review, music will be discussed as a biological function with a possible role in Alzheimer’s disease (AD).
Alzheimer’s Disease: A Highly Complex, Incurable Disorder
AD is a progressive age-related neurodegenerative disease, and the most common cause of dementia. According to the World Health Organization, approximately 50 million people worldwide are currently living with some form of dementia, of which AD accounts for 60-70% of cases (WHO, 2020). However, with an increasingly ageing population, the prevalence of people affected by age-related diseases is surging. Consequently, disease prevalence is projected to rise to 82 million in 2030 and 152 million in 2050. There is currently no cure for AD, with only relatively few symptomatic treatments existing to date.
Pathophysiology of AD
The pathophysiology and neuropathology of AD is characterised by a spatiotemporal spread of amyloid beta plaques, neurofibrillary TAU tangles, neuronal degeneration and vascular
amyloidopathy accumulating in the brain (Lau et al., 2013). In the past, AD pathology was thought to be initiated by the pathological cleaving of the Amyloid Precursor Protein (APP) forming β-amyloid peptides, which consequently formed neurotoxic β-amyloid oligomers (Yiannopoulou & Papageorgiou, 2020). This process was meant to initiate local inflammation, oxidation, excitoxicity (excessive glutamate), and TAU hyperphosphorylation, leading to the clinical syndrome of AD.
Progressive neuronal destruction in AD leads to shortages and imbalance between various neurotransmitters, such as acetylcholine, dopamine and serotonin consequently causing cognitive deficiencies in the patient (Yiannopoulou & Papageorgiou, 2020). Therefore, most of the established pharmacological interventions for AD aim to counterbalance this imbalance of neurotransmitters.
Current Landscape of Alzheimer Disease Therapeutic Strategies
There are currently only four FDA approved AD medications available. Three of them are acetylocholinesterase inhibitors (AChEI) reducing the breakdown of acetylcholine in the brain, in turn increasing cholinergic neurotransmission (Yiannopoulou & Papageorgiou, 2020). The medication often mitigates the decline of cognitive dysfunction in the beginning of the disease. However, even temporary discontinuation results in an even more rapid decline of previously gained cognitive function. Further, adverse effects such as nausea, diarrhoea as well as the increased risk to develop bradycardia and syncope have been reported. Alternatively, either in combination with AChEI’s or in isolation, NMDA receptor open-channel blockers, which affect glutamatergic transmission, are being administered.
The disease progression of AD is marked by continuous deterioration of the central nervous system, detrimentally affecting nearly all aspects of a patient’s life. Clinically, AD is expressed in memory loss, impairment of executive functions, language deficits, and general cognitive decline. The most common symptoms of AD include agitation dysregulated mood, disturbed thought and perception, as well as aggression, psychosis, restless wandering, impulsive behaviour and loss of language, sleep and appetite (Vinoo et al., 2017; Schroeder et al., 2018). These symptoms are often comprised in clinical trials as behavioural and psychological symptoms associated with dementia (BPSD) (Thomas et al., 2017). 80 % of the people suffering from AD or other dementia exhibit BPSD’s, which have significantly adverse effects on their quality of life (Vinoo et al., 2017). In
treating BPSD’s, most often antipsychotic and antidepressant medication is administered by caregivers (Yiannopoulou & Papageorgiou, 2020). Serotonin reuptake inhibitors are often used to treat depression and anxiety, however much controversial evidence has been reported on antipsychotic, anticonvulsant as well as anticholinergics and sedative medication, making many caregivers advise against administration of these drugs (Ballard & Waite, 2006; Corbett, Burns & Ballard, 2014; Yiannopoulou & Papageorgiou, 2020).
Alzheimer’s Disease as Multifactorial and Heterogenous Disease
AD is marked by its heterogeneous and multifactorial nature of disease pathology and progression. Neurodegeneration in AD results from the complex interaction of multiple factors, such as the development of the above-mentioned histological and biochemical pathology. In addition, genetic predispositions influence AD pathology, such as polygenetic inheritance, allelic and locus heterogeneity and epistasis (Veitch et al., 2019). Therefore, heterogenous findings of symptoms and pathology in patients have been reported, clinically expressed through certain cognitive functions being more highly preserved and other areas more significantly impaired than others (Vanstone & Cuddy, 2010). This makes the hypothesis of a linear disease progression cascade following amyloid deposition highly unlikely. Rather, positive and negative feedback loops, as well as genetic and epigenetic factors have been associated with the existence of multiple different pathways that interact to influence AD progression and contribute to the observed heterogeneity of AD pathology (Li et al., 2019; Veitch et al., 2019).
This complex nature of AD must be reflected in current treatment management of the disease. In regarding AD as a multifactorial and heterogenous disease, a “one-size-fits-all” treatment approach must be disregarded. Rather, therapies need to consider the multifactorial nature in targeting multiple disease-related factors and treatment plans need to be individually tailored in biomarker-guided targeted therapies (Hampel et al., 2019). As pharmacological interventions are progressing at a slow and unsuccessful pace, research has begun to turn towards the development of innovative therapeutic approaches focusing on the enhancement of preserved abilities of AD patients, rather than their deficits, to increase quality of life (Groussard et al., 2019). Music processing and the ability to enjoy and remember music is uniquely preserved in the AD brain.
Therefore, music has been considered a promising therapeutic tool, in providing an anchor for care in patients who have lost most abilities fundamental to their normal life.
The aim of this review is threefold; first, I will critically examine current literature reporting musical memory preservation in AD patients. Secondly, I will investigate possible underlying mechanisms of such preserved memory findings based on current evidence. Finally, I will present important implications that a preserved musical memory system can have as a therapeutic application in AD.
Music in The Brain
Music is one of the most ancient cognitive traits, with evidence for music making dating back to 35.000 years ago (Zatorre & Salimpoor, 2013). Making music recruits a multitude of complex cognitive processes: Perception, multimodal integration, learning, memory, action, social cognition, syntactic processing and processing of meaning information. The ability of music to activate such a large network of different brain structures, makes music a valuable tool for the investigation of the healthy and diseased brain (Koelsch et al., 2014). In the following segment music’s underlying neural circuitry as well as encoding processes and reward responses will be discussed. Further, the existing evidence for therapeutic applications of music in mental disorders will be explored.
Music Perception: From Decoding to Processing
The decoding process of acoustic information starts when a sound is taken up by a human and translated into neural activity in the basilar membrane of the cochlea. It then travels further into the brainstem, to regions of the olivary complex and inferior colliculus, where it is investigated for sound properties such as timber, sound intensity, intramural disparities and frequency information (pitch) (Koelsch & Siebel, 2005). The medial geniculate nucleus in the thalamus subsequently projects auditory information to the auditory cortex. In the auditory cortices, more fine grained and specific analysis of acoustic features takes place.
Interestingly, such sophisticated decoding of musical information is observed in individuals without any form of musical training (Tillmann et al., 2000), as well as in newborns (Nakata & Trehub, 2004). The latter already display increased attention towards the mother when she is singing, as compared to when she is talking (Nakata & Trehub, 2004). Further, imaging evidence shows increased activation in music and language processing centres during hummed as compared to flattened speech (Perani et al. 2011). This means that infants must already have the capacity to distinguish between pitch contours, as well as preferentially process phonological information (prosodic and phonemic) rather than lexical or syntactic information. This claim is supported by imaging data, which demonstrate that the inferior fronto-occipital fasciculus (IFOF), a brain network strongly associated with music processing, is already present at birth (Sihvonen et al., 2017a, Perani et al., 2011). This evidence serves as a proof for a degree of innateness of music in the human brain.
Music Processing: From Computation to Meaning Functional Magnetic Resonance Imaging (FMRI) Evidence
While feature extraction occurs in auditory cortices, acoustic signals are being transformed into a set of perceptual auditory representations, grouped according to melodic, rhythmic, timbral and spatial structures (Peretz et al., 2009). These representations make contact with working memory structures which temporarily store auditory units, compare and connect sound patterns with other representations of previous musical experiences one has been exposed to (Peretz et al., 2009). This process is termed the musical lexicon (Peretz et al., 2009) or auditory scene analysis (Koelsch et al., 2011). This process underlies musical meaning formation, allowing us to recognize familiar or unfamiliar melodies.
Neuroimaging data finds a key neural structure involved in this comparative analysis to be the superior temporal gyrus (STG) and middle temporal gyrus (MTG) (Peretz et al., 2009; Koelsch et al., 2011; Groussard et al., 2010). In a familiar versus unfamiliar music condition, contrast Blood-oxygen-level-dependent (BOLD) fMRI shows significant activation in the right superior temporal sulcus (STS) (Peretz et al., 2009). Interestingly, an unfamiliar tune was even associated with a deactivation of STS. These findings support the existence of a musical lexicon structure being
involved in a selection stage. After a melody is being determined unfamiliar, the lexical search is terminated, and a deactivation takes place.
The comparison process of the musical lexicon further feeds and receives information to and from cortical structures along both the ventral and dorsal streams, such as the inferior frontal areas, the planum temporale (PT) and supplementary motor areas (Peretz et al., 2009). Regions of the associative cortex are also involved, as the melody of a song might immediately trigger the recollection of the spoken lyrics of a song. These interactions between auditory and frontal cortex allow representations of structural regularities of music to be stored, and retrieved, which is essential for forming expectations of musical melodies (Herholz et al., 2012). Such expectations play an important role in defining to which extent a musical experience is enjoyed (Zatorre & Salimpoor, 2013)
Given the variety of processes involved in how the brain perceives and processes music, it is important to control for acoustical and perceptual attributes when investigating neural substrates of music processing (Peretz et al., 2009). A technique applied in music research is using mere exposure rather than explicit tasks asking for familiarity, in order to avoid activation of associative memories. In a study by Peretz et al. (2009), mere exposure had a stronger effect of activating the musical lexicon structures than explicit tasks.
Brain Lesion Studies
Considering the difficulty to isolate neural substrates of music processing due to the vast network of regions involved, a stream of research focuses on lesion studies to investigate neural substrates of music. Brain damage causing impairment of functional musical processing, creates a unique opportunity to investigate critical neural architecture of musical processing in the brain. Lesion studies have therefore provided further evidence for the autonomy of music processing and its distinct processing modules in the brain, supporting previous fMRI data (Peretz, 2006, Sihvonen et al., 2017a). Amusia is the disability to perceive music, and therefore often used in lesion studies investigating music. One differentiates between congenital amusia, which has a neurodevelopment origin, and acquired amusia, which occurs after brain tissue damage that occurs, for example, after a stroke (Sihvonen et al., 2017c). Often, acquired amusia, wherein a healthy brain able to perceive and process music, is suddenly stripped away from that ability, is preferred for isolating neural
substrates of music. This allows addressing causality and thereby drawing important conclusions about the direct effect of such lesions on music processing.
Voxel based lesion symptom mapping (VLSM) is a technique using advanced MRI, that investigates the relationship between focal brain damage and behavioural data on a voxel-by-voxel basis (Sihvonen et al., 2017c). In a study investigating the neural basis of acquired amusia, MRI data was collected at the acute and 6-month post-stroke stages. Recovery of music perception was further behaviourally assessed at acute 3-month and 6-month post-stroke stages.
At the acute stage it was found, that amusics show reduced activation in the right temporal STG and MTG, leading to grey matter volume (GMV) decrease in these areas. Further, lesions in relationship to impaired music processing are found in subcortical regions (striatum & globus pallidus), the right inferior gyrus, right insula and hippocampus.
Longitudinal fMRI data finds, that after a 6-month post-stroke period, better recovery was linked to increased activation in the right frontal and parietal areas as well as increased functional connectivity in the bilateral frontoparietal (attention) networks, correlating to increased behaviourally displayed attention to instrumental music (Sihvonen et al., 2017c). Poor recovery (at 6 months post-stroke) correlated to white matter volume (WMV) decreases in the right MTG, IFG, striatum and hippocampus, suggesting a disruption of neural networks leading to the observed behavioural deficits (Sihvonen et al., 2017c). These results suggest that acquired amusia seems to stem from brain damage to the right temporal region, which leads to the reported activation deficits in the lesioned areas (Sihvonen et al., 2017c). The authors also suggest mechanisms of good amusia recovery to be related to spared white matter pathways, interconnecting crucial regions underlying music processing. This interconnecting nature of recovery is supported by the fact, that lesion size seems unchangeable in amusia recovery, not differing between recovered versus non-recovered amusics (Sihvonen et al., 2017c). Therefore, amusia recovery instead seems to be connected to a dynamic functional shift of music processing regions originating in right temporal areas and recruiting bilateral frontotemporal and parietal regions at later stages of acquired amusia to recover musical processing abilities.
Interestingly, amusics showed fewer activation deficits during vocal music conditions as compared to instrumental music, suggesting partially preserved vocal music processing, supporting literature of distinct networks of music processing in the brain (Sihvonen et al., 2017c).
One important limitation of VLSM studies is the lack of pre-injury data of the subjects. Therefore, age-matched controls without a stroke history, are important to compare activation patterns to music in non-music patients. Further, the presented results focused on cognitive (pitch and rhythm) processing aspects of music. However, it would be interesting to investigate hedonic aspects of music, targeting the pleasurable experience of music as reward.
Taken together, lesion studies provide valuable evidence supporting existing fMRI data investigating music processing in healthy individuals. In order to firmly conclude about the neural substrates underlying music processing in the brain, further research is required to assess whether the same regions are consistently identified in lesion as well as healthy music studies.
Musical Pleasure
In the previous chapter, evidence for the neural processes that seem to underly our perception and experience of music was outlined. Our musical experience can be led back to interactions between primitive brain regions, decoding, comparing and forming patterns based on syntactic structures of pitch, timbre, melody and temporal components. However, how can representations of sounds and sound patterns that have no intrinsic reward value, become so pleasurable?
A highly adaptive human ability is to predict future events based on past regularity patterns. Interactions between auditory and frontal cortex in the form of functional loops along both the ventral and dorsal streams make it possible to integrate auditory information with other modalities, such as motor systems or memory centres (Zatorre & Salimpoor, 2013).
The strong affective reaction to music seems to be based on temporal expectancies, such as delay, anticipation and surprise (Salimpoor et al., 2013). Anticipation and expectancies can arise from explicit knowledge about musical structures, but can also occur implicitly. Listeners seem to develop implicit rules about musical structure that they acquired from previous experiences (Zatorre & Salimpoor, 2013). This leads them to expect certain continuations of tones and tonal structure with greater likelihood than others. Based on these expectancies, associated predictions are made. Musical reward seems to be experienced in line with the prediction error model (Niv et al., 2012). Namely, when set predictions are fulfilled, reward value is generated, mediated via a dopaminergic response (Salimpoor et al., 2013).
Neural substrates of musical expectancies are based on auditory change detection (Schönwiesner et al., 2017). Electroencephalon (EEG) data show that infrequent, deviant sounds in musical pieces evoke a frontal negative reflection in the auditory event-related potential, termed mismatch negativity (MMN). Based on fMRI data, the anatomical loci of these activation maxima were in the medial and lateral Heschl’s gyri (HG), anteriolateral and medial portions of the left PT and portions of the STG and STS, as well as the mid-ventrolateral prefrontal cortex (Schönwiesner et al., 2017).
During peak emotional response to music measured by chills, a well-established marker correlated with pleasure, Positron emission tomography (PET) data revealed increased endogenous dopamine transmission bilaterally in both dorsal and ventral striatum (Salimpoor & Zatorre, 2011).
Furthermore, fMRI data of another study reveals that the right nucleus accumbens, a key region associated with the positive prediction error (Niv et al., 2012), showed increased functional connectivity with large portions of the STG, while subjects listened to music they enjoyed (Salimpoor et al., 2013). These findings link the intense pleasure experienced when listening to music to up-regulated dopamine activity in the mesolimbic reward system. Such increases in the dopamine reward system as a result of music might have important applications for therapeutic applications in subjects with impaired dopamine functioning.
Other areas also showed increased functional connectivity with the nucleus accumbens as music value increased (Salimpoor et al., 2013). For one, the ventromedial prefrontal cortex (VMPFC) and the orbifrontal cortex (OFC), prefrontal areas involved in loops important for retrieval of musical information (Herholz et al., 2012). Further activations were observed in the amygdala, hippocampus, right IFG, anterior cingulate cortex and clusters in somatosensory and motor areas. Importantly, these areas exclusively demonstrated increased functional connectivity when sounds gained reward value.
Auditory cortical areas are thought to store templates of previously heard sounds, likely having a key role in forming musical expectancies (Peretz et al., 2009). This musical lexicon is thought to provide feedback information to other brain areas. The above-mentioned findings suggest that the nucleus accumbens receives information about temporal predictions from the STG, integrating this information to create a rewarding or non-rewarding musical experience. These findings further support the notion of a musical template centre (musical lexicon) to be located in the STG, storing individualised auditory information (Zatorre & Salimpoor, 2013).
Taken together, musical pleasure seems to underly temporal expectancies that stem from an implicit understanding of rules of musical structure based on past occurrences, and the positive prediction error that results from these expectancies. This means that the reward value of music is abstract; it does not involve an inherent, tangible substance of pleasure, but rather is the result of a combined sensory and cognitive experience that can influence one’s affective state (Salimpoor et al., 2013).
Interestingly, music evoking activity in reward circuitry seems to be unique in humans (Salimpoor et al., 2013). A possible reason for that is the pivotal role of the working memory in storing and providing information on past templates of musical structure. Research has hypothesised that the reason why monkeys do not have such evolved musical systems is because they cannot dynamically store auditory information in working memory (Scott et al., 2012). Although they have excellent visual systems their combinatorial abilities for auditory information seem to be limited. Further, an “overwriting” effect was observed in monkeys of not being able to retain auditory information longer than a few seconds, as compared to visual information (15-20 minutes) (Scott et al., 2012). These findings highlight, the essential role for such combinatorial and integrative processes as well as a highly developed working memory in musical perception, processing and reward. However, further investigation into working memory capacities of non-human animal models, such as rodents, would allow more direct dissection of mechanistic underpinnings and interactions.
Music as Therapy
The previous findings have firmly demonstrated that music can evoke activity changes in numerous brain regions, such as motor systems, limbic and paralimbic structures, as well as frontal cortical areas. In addition, music has been shown to be a ubiquitous feature in human society, recognised and enjoyed by every individual, with infants already showing responses to musical sounds. These features render music a powerful tool to gain insights into the human brain. Considering music’s powerful ability to access and activate a such vast network of brain regions, therapeutic applications of music in cases of dysfunctional or aberrant brain conditions have been widely explored.
Depression is a mental disorder often studied in combination of music therapy. Depression is characterised by dysfunction in limbic structures, such as a hyperactivation in the amygdala (Koelsch, 2009). In addition, patients suffering from major depressive disorder (MDD) show reduced activity in lateral prefrontal cortical areas during emotional experiences (Rive et al., 2013). Hyperactivation in the amygdala combined with hypoactivation in frontal areas seem to make depressive patients incapable of regulating their emotions. Hsu & Lai (2004) found that after 2 weeks of music therapy of listening to their own choice of music for 30 minutes daily, subjects in the music condition reported significantly better depression scores compared to controls. Also, another study found that after 3 months and 6 months of music therapy in combination with standard care, patients showed greater improvement in depressive symptoms, anxiety symptoms and general functioning than patients receiving only the standard care (Erkillä et al., 2011). Music therapy has also been applied in schizophrenic patients. Ulrich et al., 2007, found that group music therapy sessions improved the self-reported psychosocial orientation in patients, meaning their ability to interact better in groups. Therapeutic sessions like these, improving social cohesion, could help patients adapt to the social environment after discharge from mental institutions. However, other studies reported controversial or non-significant results for music therapy in schizophrenia, calling for more robust evidence prior to making firm claims about the effectiveness of the treatment (Talwar et al., 2006).
Overall, it appears that music therapy is considered a useful and innovative tool to treat mental health conditions and promote well-being in patients. Music appears to be beneficial both for the individual per se, as well as for promoting social cohesion. Music has the unique ability to engage multiple aspects of the self. Further, musical processing abilities and the capacity to enjoy music seem to be encoded in every human being. These features make it a powerful tool to explore therapeutic interventions in a multitude of neurological and psychiatric disorders.
Musical Memory Systems
Music seems to have a unique role in the brain, having strong effects in the healthy as well as pathological brain. In investigating literature of music related activations, insights into the neural correlates of musical processing and musical memory can be gained. Musical experience involves the interaction of rhythmic, temporal, melodic, but also spatial, verbal motor and even social
information. Music making and learning is marked by such multimodal integration, requiring the involvement of multiple systems. It is therefore comprehensible that research has found different types of musical memory to be differentially engaged as well as impaired in different forms of disease (Baird & Samson, 2009).
Prior to discussing the pertinent literature on diseased musical memory, it is necessary to outline a clear conceptualisation of the different existing memory systems of music. A general definition of musical memory systems cannot be found in the existing literature, however, a widely agreed upon conceptualisation will be presented here that can be used to compare, evaluate and discuss the empirical evidence presented in this segment.
First, one should differentiate between implicit and explicit long-term memory (Baird & Samson, 2009). In implicit memory, information is encoded in an incidental, passive manner, whilst the encoding of explicit memory requires effort and deliberate encoding (Baird & Samson, 2009). Under implicit memory, falls procedural memory, which is the “knowing how” of fluid motor sequences, often encoded by repeated exposure and without awareness.
There are two distinctions to be made belonging to explicit memory. Semantic memory encodes the general knowledge we have about a concept or entity. Its information is accessed by a sense of familiarity of a melodic progression (Groussard et al., 2019). Episodic memory on the other hand encodes the spatiotemporal context, which the remembered music is imbedded in. Importantly, episodic memory related to a specific experience in time and space having at its core a musical excerpt. It is therefore vulnerable to change over time, with a changing context, and might even become semantic memory (Groussard et al., 2019). The existence of multiple human memory systems is thought to make up the autobiographical record of an individual, essential to their sense of identity and life’s meaning (Clark & Warren, 2015, Groussard et al., 2019). This multi-systems nature of memory, including musical memory, emphasises the complexity that memory and music have in the brain.
Different Musical Memory Systems Are Encoded in Distinct Neural Networks
Studies have provided empirical evidence for the different memory systems to be involved in distinct neural networks in the brain. (Platel et al., 2007; Groussard et al., 2010; Slattery et al., 2019). A PET study finds a neural distinction between semantic and episodic memory (Platel et
al., 2007). Semantic memory, being assessed by asking participants for familiarity of a musical piece, was associated with activations in bilateral medial and orbital cortex region as well as the left anterior middle temporal gyrus, the left inferior frontal gyrus and the left angular gyrus (Platel et al., 2007). Episodic memory, following the semantic task, was assessed by subjecting participants to a number of melodies and asking which of them they had heard during the semantic task (then acting as distractors). During this task, activations in the rostral part of the medial frontal cortex and the precuneus were found, as well as the right middle frontal gyrus and right superior frontal gyrus. These findings suggest, the precuneus to have a unique role in episodic versus semantic memory, as well as semantic being associated more with left hemispheric temporal structures while episodic memory more with right hemispheric temporal structures. Also explicit and implicit memory have been anatomically distinguished with procedural memory, associated to regions of the ventral pre-supplementary motor areas in contrast to explicit memory (Jacobsen et al., 2015). The emerging dissociation between musical memory systems and their neural correlates has important implications for therapeutic applications making it imperative to support this research with more evidence.
Musical Memory in Alzheimer’s Disease Neural Networks of Musical Memory in Alzheimer’s Disease
Musical Memory has been found to be partially preserved in AD patients. Numerous studies find striking evidence of the powerful effect of music in patients suffering from the disease. Of note, musical memory is not uniformly preserved in AD (Jacobsen et al., 2015; Baird & Samson, 2009). This is in line with the functional and anatomical differentiation of the different memory system presented previously. Some forms of musical memory are impaired, whilst others are spared. This is an important observation when considering the presented evidence for the use of therapeutic applications.
Early AD pathology is marked by relatively focal atrophy predominately affecting the temporal lobes (Baird & Samson, 2009). The temporal and frontal regions have been found to be involved in long-term explicit musical memory, both semantic and episodic. Due to the severe neurodegeneration of these regions in early AD, one would expect explicit memory to be impaired
already in early stages of AD (Kerer et al., 2013). In contrast, procedural memory is associated with regions of the ventral pre-supplementary motor area, which are relatively spared by AD pathology until late disease stages, suggesting procedural, and other forms of musical memory that are not dependent on temporal lobes, to be partially preserved in patients (Baird & Samson, 2019). The majority of evidence indeed finds episodic musical memory to be early and consistently impaired in AD (Slattery et al., 2019; Baird & Samson, 2015; Groussard et al., 2010). However, variable results have been reported of semantic and procedural memory to be preserved in AD, with learning of new procedural information even being observed (Jacobsen et al., 2015; Cowles et al., 2003). The following segment aims to assess and critically evaluate the current literature, finding musical memory preservation in AD. Various studies, reporting anatomical and behavioural evidence of complete, partial or non-sparing of musical memory in different stages of AD will be presented. Further, in investigating the underlying mechanisms of musical memory preservation, two streams of hypotheses have emerged. First, musical memory preservation is attributed to an anatomical differentiation of music in the brain, i.e. an intact musical memory function is regulated by anatomically distinct musical memory regions in the brain which are spared in AD. Secondly, musical memory preservation is associated to the wide network distribution of music in the brain, allowing for functional alterations to take place that compensate atrophied regions. The latter hypothesis postulates that musical memory is preserved because even if some regions are impaired by pathology, others will be recruited to compensate and “fill the gap”.
Preservation of Musical Memory: Anatomical Evidence Anatomical Differentiation Hypotheses
In a recent study investigating the anatomical differentiation between long-term music memory and brain areas affected by AD pathology (Jacobsen et al., 2015), fMRI was used to localise brain regions involved in long-term musical memory in healthy adults. A long versus recent known contrast found the caudal anterior cingulate gyrus and the ventral pre-supplementary motor area to be crucial. Interestingly, in the long versus unknown contrast additional brain regions were recruited, namely the bilateral frontal pole, temporal pole and the insular cortex, suggesting additional brain regions involved in first time exposure to music, but not significant in long term
musical memory. In addition, the study analysed three essential biomarkers of AD: amyloid beta deposition, hypometabolism and structural atrophy (grey matter volume reduction) in AD patients and compared the location of highest biomarker accumulation to previously found long-term musical memory regions. Importantly, the study used a control group, in which no evidence of AD biomarkers was found. The study finds that cortical degeneration and hypometabolism in AD patients show nearly no overlap with the network reported to be involved in long-term musical memory. Highest atrophy values were observed in the temporal cortex and the inferior parietal cortex and the precuneus and highest hypometabolism observed in precuneus, posterior cingulate gyrus and temporal and parietal cortices. Musical memory regions even displayed among the lowest levels of grey matter reduction and hypometabolism in the entire brain of AD patients (Jacobsen et al., 2015). Interestingly, for the amyloid beta marker, musical memory structures were not located in regions with significantly lower levels of deposition. In AD patients, amyloid beta deposition was predominantly found in medial and orbital prefrontal cortex, precuneus and posterior cingulate. Based on these findings, the authors suggest that temporal areas might be essential in encoding explicit musical memory, as opposed to their role in long-term musical memory processing. Further, they connect the caudal anterior cingulate gyrus and ventral pre-SMA to automatic semantic musical memory retrieval, meaning the degree of familiarity an individual experiences when hearing a musical piece. These areas have been associated with complex planning and content evaluation supporting the previous conceptualisation of music as a rhythmic multimodal sequence of structured sounds (Koelsch et al., 2014).
Finally, considering the finding that regional differences between AD biomarker concentration and long-term musical memory structures could be found for grey matter atrophy and hypometabolism but not for amyloid beta deposition. The authors claim, this supports the amyloid cascade hypothesis about the anatomical disease progression of AD (Jacobsen et al., 2015). According to this hypothesis, amyloid beta deposition is thought to precede hypometabolism and atrophy in the brain. Regions found to be involved in long term musical memory are known to degenerate last in patients. Musical memory regions showed nearly no overlap with regions displaying high levels of hypometabolism or atrophy, however, amyloid beta deposition was found within musical memory structures in AD patients, suggesting that these regions are preserved as they are still in very early stages of degeneration.
Another study finds evidence for an anatomical dissociation between semantic verbal and semantic musical memory (Johnson et al., 2011). Using voxel-based morphometry, a difficulty in naming songs was associated with the volume of bilateral temporal lobes (including inferior and middle temporal gyro and temporal pole) and inferior frontal gyrus. A difficulty in detecting pitch errors on the other hand correlated primarily with right temporal lobe structures. Structures recruited by semantic musical memory, seem to be later reached by AD pathology. A crucial structure found by Jacobsen et al., 2015, involved in long term musical memory, was the anterior cingulate gyrus. A study comparing metabolic alteration ratios in form of hypometabolism, to structural alteration ratios in form of atrophy, finds that in moderate stage AD patients, the cingulate gyrus displays greater hypometabolism compared to atrophy (Chetelat et al., 2008). On the other hand, the hippocampus, one of the regions affected first in AD (Jones et al., 2006; Jacobsen et al., 2015), was one of very few regions, that displays metabolic as well as structural alterations of similar magnitude, namely it did not display greater hypometabolism than atrophy. The hippocampus is among the earliest sites in the brain to be targeted by disease pathology. These findings might provide a potential underlying mechanism explaining preserved musical memory in AD patients, from an anatomical perspective, namely by demonstrating that essential structures are only later targeted.
The cingulate gyrus per se has also been found to be differentially affected in AD pathology (Jones et al., 2006). A study finds that the anterior region seems to be significantly less affected than posterior regions, which displayed the greatest atrophied volume reduction. Interestingly, in both AD and control subjects, cingulate regions were larger on the right side. The only statistically significant difference though was found in the rostral region of the anterior cingulate cortex. Nevertheless, a tendency towards a greater volume of the right cingulate gyrus was reported. Further, the severity in volume loss did not have a hemispheric asymmetry in the cingulate regions, i.e., no difference in reduction could be observed between the left and the right side. Based on these findings, even after disease onset and volume reduction, the right anterior cingulate cortex would have a tendency of greater total volume compared to the left size.
Considering the important role of the right hemisphere and the role of the cingulate cortex in musical memory (Jacobsen et al., 2015), evidence suggesting a greater total volume of this structure, than its hemispheric counterpart, even after onset of neurodegeneration, carries
important implications for this brain structure and its role in the preservation of musical memory, necessitating further functional studies.
Network Distribution Hypotheses
Network connectivity, particularly with respect to the anterior cingulate cortex is also key to musical memory preservation in AD. Zhou et al. (2010) report that the anterior cingulate cortex bilaterally displays increased connectivity in AD patients as compared to controls. Specifically, patients with AD showed reduced Default Mode Network (DMN) connectivity compared to controls in the left retrosplenial cortex, left posterior hippocampus, left cuneus and midbrain tegmentum. However, increased salience network connectivity compared to controls was observed in bilateral anterior cingulate cortex and left ventral striatum. Interestingly, the right anterior cingulate cortex showed a significant inverse relationship in DMN connectivity with the precuneus and ventromedial prefrontal cortex in AD patients. This means that as those two structures lose connectivity due to atrophy in AD pathology, the anterior cingulate cortex seems to gain connectivity, indicating active functional alterations in these neural networks. Further, the anterior cingulate cortex stood out as the only region exhibiting increased grey matter volume in AD patients.
These findings introduce a second stream hypothesis within the field of long-term musical memory research. Namely, musical memory preservation occurring due to a network distribution effect. This line of research emphasizes the importance of measuring changes in musical memory systems in AD, in order to systematically explore the key role music plays in the disease. It also underlines the potential of using music as a window to investigate neural network changes in AD (Slattery et al., 2019).
A recent study contrasts music processing in AD patients to healthy controls and finds functional neuroanatomical alterations in musical memory of patients (Slattery et al., 2019). The study reported that healthy controls showed activation in the precuneus of the posteromedial cortical region, in hearing previously unfamiliar melodies. This activation seemed to be lost in AD patients. This is consistent with previous findings of AD pathology targeting the hippocampus and temporo-parieteal circuits associated to the decoding of musical novelty (Kafkas et al., 2014). More specifically, a dysfunction in the precuneus is thought to impair appropriate responses to external
sensory events and the encoding of those events into memory, processes that are crucial for episodic memory (Slattery et al., 2019). Further, for musical semantic processing (familiar versus unfamiliar distinction) a bilateral activation of supplementary motor cortices and a bilateral activation of anterior superior temporal cortices was observed in healthy controls. In AD patients, bilateral activation of supplementary motor area was still observed, however, only the left anterior superior temporal cortex showed activation.
In summary, for semantic processing, the dorsal medial prefrontal cortex seemed to be relatively resistant to the effects of AD pathology, providing a possible substrate for the partial preservation of musical semantic memory in AD. Altered activation of posteromedial cortical regions (precuneus) during episodic processing is in line with behavioural data reporting impaired episodic memory (Campanelli et al 2016; Samson et al., 2012).
Möller et al., 2013 report interesting findings on differences in atrophy patterns between late versus early onset AD. Specifically, they find more pronounced atrophy in the precuneus in early onset AD as compared to late onset, whereas medial temporal lobe structures and the cerebellum were more deteriorated in late onset AD. Linking these findings back to Slattery et al., a dysfunctional precuneus was associated with impaired episodic musical memory (2019). Considering the network distribution hypothesis, it would be of interest to investigate differences in networks activated by music in early versus late onset AD patients, to see if a different set of structures might compensate due to differential deterioration. Studies like these, reporting structural differences based on disease onset should furthermore be carefully considered for therapeutic applications. The authors of this study go so far as to suggest that in younger AD patients, posterior parts of the brain, including the precuneus, might provide valuable information (Möller et al., 2013). In addition to disease onset, the study investigates the apolipoprotein E (APOE) in its role to modulate AD disease pathology. APOE has been associated with more severe brain atrophy in regions usually affected by the disease such as temporal lobes (Honea et al, 2009). However more research needs to be conducted in this field.
Finally, other studies make an interesting point in suggesting the preservation of musical memory to be due to the larger, more distributed network it engages as compared to other memory networks such as verbal semantic memory (Peretz et al., 2009; Groussard et al., 2010).
Taken together, the above-mentioned findings present important evidence for the existence of functional network changes in AD and the potential music can have in capturing these disease-associated effects.
Preservation of Musical Memory: Behavioural Evidence
In the previous part, anatomical evidence for a preservation of musical memory was presented. However, musical memory is also often assessed in terms of behavioural cues. In assessing behavioral data some important aspects should be considered. Studies behaviourally investigating musical memory mostly rely on explicit verbal or written responses. However, with disease progression also other cues are used to identify musical memory performance such as facial expressions, quality and duration of attention or general responsiveness of the patients (Groussard et al., 2019). There are generally three methods for assessing musical memory behaviourally, most often employed by studies. Although small differences might occur between studies, a general consensus towards these methods is observed. To test procedural (implicit) memory, researchers often use free recall, which means that a subject can independently retrieve a melody and reproduce it by humming or playing along (Groussard et al., 2019).
To test semantic musical memory recognition tasks are employed, which determine whether a subject recognises familiar or unfamiliar songs as familiar. Importantly, in referring to this task, only musical semantic memory is assessed, not verbal information related to the musical piece. Finally, recollection tasks are used to test episodic musical memory, which include the recognition of a familiar melody (semantic) but in addition also the spatiotemporal context of that familiar piece. The most common way to test this would be to present the subject with novel melodies during a learning phase, and then presenting the subject with a distraction task. The episodic memory is then tested by having to identify the previously heard melodies (targets) among a random assortment of both targets heard and novel (distractor) melodies (Groussard et al., 2019).
Semantic Musical Memory
Musical Memory seems to be partially preserved in AD patients. However, as AD pathology is very heterogenous and musical memory highly complex, different findings have been reported over the years as to which musical memory domains are preserved, to what extent they are
preserved, and in which stage of the disease. Although no certain generalisable conclusions about musical memory in AD have been made to this point, the here presented evidence can be used to find common themes, implications and future avenues for research.
To start, the most agreed upon memory preserved in studies is retrograde semantic musical memory (Vanstone et al., 2012; Johnson et al., 2011; Bartlett et al., 1995; Samson et al., 2012; Vanstone et al., 2009). Vanstone et al., 2012 find that although younger controls outperformed AD patients and older controls in both procedural and episodic musical memory tasks, no difference in performances was observed in the semantic memory task of recognising traditional melodies over their lifespan. In two other independent studies, the authors investigated semantic memory in severe cases of AD (Vanstone et al., 2009; Vanstone et al., 2010). In a 85-year old female patient of AD a partial preservation of semantic musical memory was observed (Vanstone et al., 2009). Although severely cognitively impaired, the patient exhibited fully preserved discrimination of familiar songs and even somewhat discrimination of familiar lyrics and singing of familiar lyrics. In post-mortem analysis, severe atrophy was found in frontal and medial temporal lobes. The findings provide interesting evidence for a partial preservation of musical semantic memory throughout AD progression.
Cuddy & Duffin (2004) assessed an 84-year-old woman suffering from severe AD. The patient responded to familiar melodies by singing or humming along and to distorted familiar melodies with surprised facial expressions. She never exhibited responses to unfamiliar melodies. These results suggest a partially preserved semantic musical memory of melody recognition. Interestingly, the authors bring forth the hypothesis that a preservation of this memory system might be due to music enhancing a general level of activation in the brain, due to its large and widely distributed network, linking these behavioural findings back to the network distribution hypothesis presented above. The authors postulate, that in AD, weakened and atrophied components might be supported and reinforced through this general level activation, prompting motor activity and memory recall. Anatomical evidence of network distribution research, which would further support the author’s conclusions, was, however, missing.
Other studies tested behaviourally the dissociation between verbal semantic memory and musical semantic memory (Johnson et al., 2011; Kerer et al. 2013). AD patients performed significantly worse in verbal memory, tested by having to name titles of played songs, than heathy controls. However surprising to the authors, AD patients even outperformed controls in judging pitch distortions in songs, or altered melodies reflecting preserved semantic musical memory (Kerer et al., 2013). Also Johnson et al. (2011) report no group differences in neither pitch discrimination nor melody discrimination tasks between AD and controls, however, in familiar title recall, AD patients performed significantly worse than controls. These findings support the notion of a specialised memory system for music, distinct form other memory domains, such as verbal or visual memory. Further, they provide behavioural evidence in addition to the anatomical evidence already presented above, of an anatomical dissociation between verbal and musical semantic memory (Johnson et al., 2011).
Episodic Musical Memory
Another consistently reported finding in the literature is the impairment of episodic musical memory across AD stages. Campanelli et al., 2016 find episodic musical memory significantly impaired in AD patients, assessed through an incidental recollection task where subjects had to distinguish which of the presented melodies they had heard during a learning phase. Also Samson et al., 2012 find in moderate stage AD patients, impaired explicit learning as well as recollection of familiar and unfamiliar melodies compared to controls. However, the patients did display successful access to music recognition units, suggesting spared semantic musical memory. Taken together, these findings suggest that throughout AD disease progression,a partial preservation of semantic musical memory exists, contrasting consistent impairment of episodic musical memories. Further, behavioural evidence supports anatomical evidence of a dissociation between musical memory systems and other memory domains, as verbal semantic memory is consistently impaired, while musical semantic memory can still be accessed.
Next, procedural memory preservation has been largely investigated. As for procedural memory assessment, often no verbal or written responses are needed, but rather behavioural cues such as the playing or humming of a memory, procedural memory is often assessed in severe AD patients. In a case of an 82-year-old musician suffering from AD, preserved procedural musical memory was reported, while the patient exhibited other procedural memory defects (Crystal et al., 1989). The patient was reported to be unable to recall the composed of any of 6 titles played to him, however, when brought to the piano the patient could in 13 out of 15 cases play musical pieces as they were originally written. Interestingly, the patient was even observed to be able to learn patterns of tapping, showing capacity for anterograde procedural learning. When asked about the newly learned sound patterns though, the patient showed no declarative memory for newly learned information.
A similar finding was reported in which repeated exposure could lead to partial anterograde semantic musical memory (Quoniam et al., 2003). This study tested exposure effects of unfamiliar melodies in AD and depressed subjects. They found that AD patients showed learning effects with an increase in repetitions of presented sound stimuli. With increasing exposure, AD patients rated the melodies as more pleasant. Depressed subjects failed to show such a learning effect. Even with increasing exposure, pleasantness ratings did not change due to severe emotional dysfunctions (Quoniam et al., 2003). These findings have important implications for the role that emotion plays in automatic processing of musical stimuli, which will be explored later in this review. Also, in severe AD patients preserved procedural memory has been reported. One study examined a 79-year-old AD patient, who learned to play the piano as a child (Beatty et al., 1999). This study is of particular value not just because it considered a severe case of the disease, but also because it studied the patient over a time period of three years, assessing the progression of AD both behaviourally as well as through neuroimaging techniques. The patient displayed unusual deterioration in language, visuospatial function, attention and even recognition of familiar songs. However, motoric learning of playing familiar songs on the piano remained intact throughout the entire study. The authors link their findings to preserved circuits found to be involved in procedural memory: the basal ganglia, cerebellum, and motor areas of the thalamus and cerebral cortex, suggesting an anatomical differentiation of musical memory as one underlying mechanism of preservation.
What is distinctive about these findings is the severity of the dementia of the patient and the consistency of preserved piano playing skills contrasting progressive decline of other cognitive abilities (Beatty et al., 1999). Further, the study finds that although motoric learning of familiar songs is flexible enough to switch from a piano to a xylophone, the patient could not learn new songs, even implicitly. Further supporting the anatomical reasoning, the authors suggest that the above mentioned circuits involving the basal ganglia, cerebellum and motor areas, are capable of controlling and executing the performance of pre-morbidly acquired songs, however, seem to be incapable to allow new motor learning. The authors discuss, that this might be because of the inability to encode information of this complexity, or it might reflect disturbances of other cognitive capacities such as sustained attention (Beatty et al., 1999). Interestingly, this study reports no evidence of spared right hemispheric structures as compared to the left. This hemispheric differentiation has often been reported in other studies (Jones et al., 2006).
In contrast to these findings, another study did indeed find partial procedural memory learning for novel songs, in addition to preserved retrograde procedural musical memory (Cowles et al., 2003). The patient was able to learn a song on the violin, which was brought out after the onset of his probable AD. He displayed modest retention of the song, after removal of the sheet music. Although exhibiting profound impairments in the retention of any other anterograde memory tests involving words, stories or environmental sounds as well as lost episodic musical memory, procedural musical memory, both anterograde and retrograde seemed to be preserved. Again, the findings of this study suggest that the learning of a new song does not depend on usual structures of explicit episodic musical memory such as the hippocampus or the limbic system, as those were heavily deteriorated based on the patients MRI scans.
As the above-mentioned findings of preserved procedural memory often are reports from musicians suffering from AD, it is important to consider evidence from non-musicians. A recent study reports intact pitch and rhythm recognition of familiar melodies, but importantly also the ability to learn a new song in form of humming along to it in a non-musician (Baird et al., 2017). The 91-year-old woman, displayed impaired cognitive functioning in line with severe dementia, not being able to recall three words after a 2-minute delay. However, next to intact semantic musical memory, she showed the ability to learn a new song in a first session with recalling it by means of prompts and help from the experimenter at first but later could produce it spontaneously, unprompted, humming along. The authors note, that the free recall of the patient in humming the
song reflects not just pitch matching of prompted songs, but genuine recall of memory (Baird et al., 2017).
It is important to note, that such remarkable findings in this patient can by no means be generalised to all individuals suffering from AD. As the disease progresses, cognitive abilities become more heterogenous, making it important to consider findings with caution (Groussard et al., 2019). Taken together, different studies report relatively consistent results in the preservation of long-term procedural musical memory as well as the preservation of semantic musical memory and the impairment of episodic musical memory. They differ, however, in their findings of preserved anterograde procedural and also semantic memory, i.e., learning abilities in AD patients. Such findings, although to be treated with caution, might carry highly valuable implications. If in late stages of AD it is still possible to encode new representations, store and retrieve them from long-term memories, which would be the neural substrates of such new representations? And what would the therapeutic applications be for making use of this unique ability?
Protective Effect of Music on AD
Musical memory seems to be partially preserved in AD. I have presented anatomical as well as behavioural evidence of musical memory systems being completely or partially spared or disrupted only later by AD pathology. Here, I will discuss scientific findings that support an active role for memory in rendering the brain more resistant to the neuropathological effects of ageing.
The human brain is known to be plastic, meaning it can be modified in terms of structure and function, based on the environment and the stimulation it receives (Rogenmoser et al., 2018). Plasticity effects in the brain have been shown to underly the specific experience or behaviour one is engaged in, with effect size related to the intensity or frequency of experience or behaviour. The brain therefore has a certain capacity to adapt to the environment to optimise required functioning of activities.
Music playing is marked by its need for practice, meaning a continued repetition of cognitively taxing behaviour is required for developing and increasing skill (Sutcliffe et al., 2020; Gaser & Schlaug, 2003). Studies have shown, that due to its nature, music making can have significant and long-term effects on the anatomy and functionality of the human brain. Further, music learning is multimodal, demanding communication between multiple brain regions, engaging auditory,
visual, motor, attention, emotion and reward networks. Therefore, music making has the capacity to result in structural changes in all these regions (Sutcliffe et al. 2020).
Structural and functional brain differences, in terms of grey matter volume and increased activity between neural networks, have been reported in several studies comparing musicians to non-musicians. In the following, evidence of music-associated neural plasticity will be presented. The links of these findings to AD and to a potential protective effect over the neurodegenerative process will be discussed.
Structural Plasticity Induced by Musical Practice
A systematic review investigating 84 music intervention studies finds that musicians exhibited greater grey matter volume (GMV) in the bilateral superior temporal gyri, post central gyrus, medial frontal gyrus, supramarginal gyrus, hippocampus, insula, thalamus and left cerebellum (Criscuolo et al., 2021). Evidence of greater white matter volume were found in bilateral internal capsule bundle, anterior corpus callosum, longitudinal fascilus and anterior thalamic radiations in musicians compared to non-musicians.
Such structural changes in musicians compared to non-musicians seem to underly musical practice rather than genetic differences (Hyde et al., 2009). A study, assessing brain plasticity in children found that after 15 months of learning to play a musical instrument, children showed changes in anatomical features in music-related brain areas. Specifically, larger voxel sizes were found in the right precentral gyrus, related to the motor hand area, corpus callosum and right primary auditory region in instrumental children compared to controls. The structural changes were positively correlated with improved performance changes on motor and auditor musical behavioural tests. Prior to the study, the children showed no difference in brain structure or behavioural performance. Although this study provides interesting evidence for a nurture view on musical plasticity, it did not control for genetic predispositions, which might still have an effect on musical training abilities in individual children.
An important factor in how music can affect neural plasticity is the time spent practicing (Bergman Nutley et al., 2014). A longitudinal study assessed musicians versus non musicians, aged 6 to 25, at three points in time, with 2-year intervals. A beneficial effect of music on working memory, processing speed and general cognition was found. The change in working memory over time was
proportional to the time spent practicing the instrument, suggesting a dose-dependent effect of music to be present. Supporting the importance of a training factor in music plasticity, a voxel-by-voxel morphometric (VBVM) study, finds significant positive correlations between musician status, ranging from non-musicians to amateur musicians to professional musicians, and grey matter volume increases in primary motor and somatosensory areas, premotor areas, anterior superior parietal areas and inferior temporal gyrus bilaterally (Gaser & Schlaug, 2003). Additional positive correlations, yet non-significant, with musician status were seen in the left cerebellum, left Heschl’s gyrus and left inferior frontal gyrus.
Finally, structural changes in grey matter volume seem to appear gradually, having regional selectivity (Groussard et al., 2014). In line with other studies, the authors reported an overall positive correlation between duration of musical practice, and changes in grey matter volumes. However, structural changes in brain volume seem to appear gradually, depending on the brain regions and the years of musical practice. Specifically, increases in grey matter volume in the right middle and superior frontal regions and left hippocampus were already reported in novice musicians, and grew larger with time. However, structural changes in left superior temporal, posterior cingulate and right supplementary motor areas were observed to be less linear and appeared only after several years of music practice.
A constant finding of studies is the enlargement of the superior temporal gyrus in musicians compared to non-musicians, an area previously connected with the musical lexicon (Groussard et al., 2014; Criscuolo et al., 2021; Hyde et al., 2009). The authors make an interesting contribution by connecting the volume increase in this structure with a conceptual enrichment of the musical lexicon, based on musical expertise. Further, they link the increased volume in the hippocampus, to training-related neurogenesis. However, more empirical evidence is needed to validate these claims.
Functional Plasticity Induced by Musical Practice
Functional connectivity in different brain networks as a result of musical training has been reported in several studies. In a number of auditory tasks, a recent systematic review found heightened activity in inferior frontal gyrus and superior temporal gyrus (Criscuolo et al., 2021). However, lower activation was found in parietal lobes and cerebellum in auditory tasks.
An interesting study found that fibre tracts (white matter) were susceptible to training-induced plasticity, while still under maturation (Bengtsson et al., 2005). Fractional anisotropy (FA) was used to indicate the integrity of the fibre system, meaning its level of activity. Significant relationships between FA and musical practicing were found among all age groups, however for different brain regions, namely those that were still under maturation. In children under 11, fibre tracts in the internal capsule showed heightened FA, while in adolescents, fibre tracts in the corpus callosum displayed greater activity. In adults, the acute fascilus, involved in the corticocortical tract, displayed increased FA. Both corticospinal tracts and corpus callosum continue maturation throughout childhood. Based on these findings, musical training can induce white matter plasticity if it occurs in a period in which involved fibre tracts are still under maturation.
Finally, AD patients listening to preferred musical pieces, have been found to show temporary increases in functional connectivity in the corticocortical and corticocerebellar tracts, up to 10 minutes after the task (King et al., 2019). Both of these tracts are involved in sensory and attentional networks. Interestingly, functional decreases in the brain’s Default Mode Network can be observed already in early AD pathology. However, the salience network involving structures such as the insula and the anterior cingulate cortex (Criscuolo et al., 2021) is somewhat preserved until later stages of AD. This might give music the ability to facilitate attention, reward and motivation through the engagement of this network (King et al., 2019). However, an important limitation of this study is that it only investigates short-term functional connectivity changes. More longitudinal studies are needed to make more robust statements about long-term effects and generalisability of music’s effect on functional brain connectivity.
The above-mentioned anatomical findings have been found to have important behavioural implications. A recent study finds musicians to have improved memory functioning compared to non-musicians (Diaz Abraham et al., 2019). Memory performance was manipulated in musicians versus non musicians by means of musical improvisation versus imitation. Participants (either musicians or non-musicians) had to either improvise a rhythmic pattern upon hearing a musical piece or imitate the experimenter while being shown emotional versus neutral images. They then had to describe in a test phase as many of the images as they could remember. The study found that memory performance was best for the musical improvisation group of musicians. This effect was even stronger in the conditions in which emotional images compared to neutral images were shown. Based on these findings an interaction between musical improvisation and visual memory
