Performing Music on Stage: The Role of the Hippocampus in Expert Memory and Culture

Abstract

This overview chapter discusses memory functions from the viewpoint of the performing arts. ‘Playing music by heart’ is taken as an example to illustrate the role of the hippocampus in acquiring and expressing expert memory. Many more aspects depend on hippocampal mechanisms beyond declarative memory, for example, motor sequence learning, phrase boundary processing, and time-precise sequence recall. In consequence, changes in size and/or functional activity also occur in the hippocampus, known as hippocampal plasticity. Whenever the to-be-remembered items have to be stabilized even further, certain mnemonic strategies are effective, of which the oldest is the (hippocampal-based) method of loci, using visuospatial imagery. Mnemonic techniques also play a role in ethnomusicology. For example, North Indian tabla players combine drum patterns with certain onomatopoeic syllables to keep on track when performances last over hours. The value of memory processes is also discussed from a sociocultural perspective. Since priests, teachers, heads of tribes, and many others are explicit carriers of internalized knowledge, they help preserve oral traditions and culture. A special emphasis is on the accurate memorization of the Quran in Arabic, revealing that internalized sacred knowledge acquired through learning by rote can serve as a moral compass for the individual.

1. Introduction

Soloists sometimes produce 1800 tones per minute when performing live on stage. Note-perfect, virtuoso recitals have their roots in the nineteenth century, in the works by Franz Liszt, Paganini, and Frédéric Chopin. By convention, and for better expressiveness, skillful pieces are played by heart. For that, expert memory [1] is a precondition manifesting itself in time-precise and highly accurate sequence recall. This way every structural detail (and nuance) of the piece can be retrieved in a stable manner, also when performances last 20 min or longer (Figure 1). Expert memory, therefore, differs in many respects from self-paced (episodic) recall, that is, from the standard type of sequence memory as in storytelling.

Figure 1.

Expert memory is a precondition for expressive performance without the score.

It seems reasonable to assume that hippocampal mechanisms are involved, both during memorizing and recall. However, in the field of “Music and Neuroscience” (also termed “Neuromusicology”) studies on the relationship between (active) long-term memory processes and hippocampal structures are scarce. The most crucial question in this respect is if the human hippocampus is sensitive to temporal details at all, in particular as regards tone lengths, micro-pauses, and the sequence order of elements. A preliminary proof has already been given in the visual domain: Barnett et al. [2] could show in a functional MRI study, using a match/mismatch detection paradigm and picture series as stimuli, that both hippocampi were strongly activated when the temporal information was held constant, that is, when event lengths and inter-stimulus durations within the picture series were exactly the same at encoding and the test phase. Note that this is an example of retrieval success (see, e.g., [3]). However, closer to a musician’s reality is an fMRI study from the field of psychoacoustics [4] in which hippocampal sensitivity to temporal details was put to the test once again: In this study [4], a series of monotonous but otherwise regularly spaced pure tones (offset-onset distance: 150 ms) was occasionally interspersed with pure tones occurring earlier, that is, with offset-onset intervals of 142-, 130-, and 100-ms duration, resulting in a sort of “stumbling” listening impression. (Note that detecting these too-early-appearing deviants was more or less automatic since participants were instructed to ignore the sounds while watching a silent movie.) Sections of temporal irregularity activated the left anterior hippocampus in a group of professional musicians, which was not the case for the group with nonmusician controls. (However, for both subject groups, an enhanced BOLD response could be observed in the right planum temporale.) Note that in [4], hippocampal activity was induced by “novelty in temporal structure,” whereas in the previous study with picture series [2] the hippocampus reacted to sameness, since all temporal details were kept constant. This inconsistency in the results might be attributed to the fact that [4] had all the hallmarks of a classical mismatch design, whereas Barnett et al. [2] did not. Some scientists also point out that the cell types within the hippocampus may have different properties, with one type of neurons showing reactions to novelty, whereas the other showing reactions to items remembered as old. So, all in all, the human hippocampus indeed seems to be sensitive to fine-grained temporal aspects, which is a precondition for building exact representations of musical sequences in mind and making the sound results stable. (To some degree this holds also for actors learning their roles).

However, with these findings a basic problem becomes obvious: Gaps in research, especially missing functional MRI data on how the hippocampus might react to musical features other than those related to time, make it unavoidable to sometimes argue by analogy. In other words, conclusions occasionally have to be drawn from results obtained in other domains, outside the field of neuromusicology. In addition, it would be more trustworthy if results would be less contradictory as is the case with [4] compared with [2]. Thus, to achieve homogeneity more studies are needed in the field of Music and Neuroscience with special focus on hippocampal activity (replication studies included).

On the other hand, a considerable number of hippocampal studies deal with episodic memory, that is, when certain personal memories are built or retrieved. The respective key terms are: “binding,” “linking,” or “relating” persons and/or disparate objects with each other. Furthermore, to achieve coherence persons and/or objects have to be set into the situational context, that is, events must be adjusted to space and time (e.g., [5, 6]). In this regard, Maguire and Mullally [7] recently put emphasis on the words “scene” and “scene construction” underlining that episodic memory is holistic (or: pattern-like) in nature. Accordingly, episodic recall is completely different and has to be distinguished from time-precise sequence recall needed to play musical pieces by heart. Thus, drawing parallels between both memory types has to be considered with caution.

This book chapter discusses memory functions from the viewpoints of the performing arts and culture. Wherever possible, the focus is on hippocampal activity in terms of processing music. The text is divided into three parts. The first gives an overview of studies from the field of Music and Neuroscience, in particular, what has been found out about hippocampal plasticity and hippocampal involvement regarding recognition memory. The second part starts from the observation that soloists work hard to make the recall of musical passages during performance more stable. It will become evident that a distinguishing feature of expert memory is the use of certain mnemonic strategies, including a grouping mechanism called “chunking.” Maguire et al. [8] found out by questionnaire that superior memorizers instinctively (but consciously) use mnemonic strategies when asked to memorize some simple sequences, whereas matched control subjects do not. A special focus is on the so-called method of loci, an old mnemonic strategy already known by the ancient Greeks and Romans. Interestingly, a special case of mnemonic technique has been discovered in the field of ethnomusicology: While playing drum patterns North Indian tabla players think in mnemonic syllables and say them aloud to keep on track when performances last over hours. Finally, part three goes beyond individual memory, showing the role of mnemonic devices in preserving oral traditions and culture. A special emphasis is on the accurate memorization of the Quran.

First, I will provide some background on the anatomical facts and findings that are relevant for this chapter*: The anatomical core regions necessary to bring thoughts, facts, and events back to mind are the hippocampus proper and its adjacent structures located in the medial temporal lobe (MTL), the so-called perirhinal, parahippocampal, and entorhinal cortices. On the whole, highly pre-processed information reaches the hippocampus proper, which is often described as a “convergence zone” (e.g., [6, 9], see also [10]). Today, the conventional way of assigning memory functions to the left and the right hippocampus has been given up. According to this oversimplified view, the left hippocampus is involved in the storage of personal events (or: episodic memories), whereas the right is involved in spatial navigation and the storage of places (for this see, e.g., [11]). At present, the focus has shifted towards the long axis of the hippocampus, and specific functions are associated with its anterior and posterior subregions [12]. Note that some studies use both reference systems, long-axis as well as laterality (e.g., [13]). There is strong agreement that spatial information is processed in the posterior subregions bilaterally, whereas semantic aspects mainly activate the left anterior part. (The latter points to the substantial role the hippocampus has in terms of both the semantic and the episodic types of declarative memory.) Moreover, based on progress in measurement technology, the focus has shifted further towards anatomical and functional connectivity, which is investigated with diffusion tensor imaging (DTI, e.g., [14], see also [9]). Accordingly, the hippocampus is no longer considered as an isolated structure but is rather seen in conjunction with neocortical and subcortical brain regions, to those it is linked via strong reciprocal pathways. Thus, hippocampal research now pursues a more dynamic approach, and this has been extended even further: Ritchey et al. [15] propose the existence of two large-scale systems interacting with each other, a posterior medial system (PM) and an anterior-temporal one (AT) subsumed under the acronym “PMAT framework.” Due to its strong connectivity with both systems the hippocampus plays dual roles in PMAT and serves also as an integration zone.

Reciprocal (or: bidirectional) connections between the hippocampus and the neocortex also play a decisive role in regard to memory consolidation, which is often described as “resistance to interference [16].” It is common knowledge that the hippocampus merely acts as a temporary store. That is, even though the memorized content initially depends on hippocampal processes to make it resistant to interference on the molecular level, it is passed on to neocortical regions to be stored there in a distributed but permanent manner. Although there is strong consensus on these hippocampo-cortical mechanisms regarding “the way to get there,” that is, while building stable memory traces, researchers disagree on the reverse, that is, on the role the hippocampus has when stored content is re-activated during retrieval. Some researchers are convinced that long-term information is retrieved independently of hippocampal activity, since, approximately 24 h after encoding, cortico-cortical connectivity increases, while hippocampo-cortical connectivity decreases (see, e.g., [17] for details). In contrast to that, Teyler et al. [18, 19] proposed the so-called “hippocampal memory indexing theory” in which, to the contrary, the hippocampus proper plays a decisive role during retrieval. Teyler’s theory assumes that for each complex event stored in the neocortex in a distributed manner a certain index (or: addressing mechanism) exists in the hippocampus that can be re-activated. By this, the hippocampus is able to point to the respective neocortical sites which, for their part, can quickly be re-activated to restore the whole memory pattern.

Finally, the work of the so-called Hippocampal Subfields Group has to be mentioned [20]. This is a sort of working group or league, currently consisting of 200 scientists from 18 countries. By using submillimetric high-field MRI the group has set itself the goal to standardize segmentation methods and newly define the anatomical boundaries of the MTL regions. A special focus is on the hippocampus proper and its subfields along the anterior-posterior axis, termed cornu ammonis (CA) fields 1, 2, 3, and 4, dentate gyrus (DG), parasubiculum, presubiculum, and subiculum (Figure 2). Interestingly, some new approaches are taken to build research on these subfields. For example, Cremona et al. [21] found a link between CA1 volume and the outcome of a learning test using word lists; in other words, subfield-level volumetry revealed a correlation between CA1 volume and the scores achieved by free recall, but not between the scores and the entire hippocampal volume.

Figure 2.

Hippocampal subfields as displayed through a transverse section of the hippocampus. CA—cornu ammonis. (Kannappan et al. PLoS ONE 17(7), 2022).

2. The hippocampus in musical contexts

2.1 Studies on hippocampal plasticity

Neuroplasticity is one of the most widely studied issues in the field of “Music and Neuroscience.” The term describes experience-driven reorganization in the human brain, mainly caused by hard exercise and extensive training (e.g., by repeating certain finger movements hundred times). Thus, the brain of practicing musicians is oftentimes regarded as a prototype model of neuroplasticity [22]. Changes in size and/or functional activity have mainly been studied on the macroscopic level, (but see [23] for a discussion of the underlying cellular and molecular mechanisms). For example in professional string players, sensorimotor areas are enlarged in the right half of the brain, because the trained left hand, that is, the fingering hand, has its cortical representations there [24]. Pianists, by contrast, reveal a larger anterior corpus callosum since piano playing requires precise coordination of the hands due to bimanual movements on the keyboard [25].

In a prominent MRI study on neuroplasticity, Maguire et al. [26] could show that the hippocampus is subjected to structural change too. However this time, plastic changes were caused by map-like knowledge and tremendous experience in navigation, that is, by spatial memory. What was examined there were the brains of a group of licensed London taxi drivers. In detail, the posterior parts of their right and left hippocampi revealed a significant increase in gray matter volume (in comparison with age-matched controls). (Note that activity in the taxi drivers’ left hippocampus was interpreted as being caused by the innumerous social actions with which taxi driving is connected).

To further explore the processes of hippocampal reorganization, memory tasks in music are suitable too, both in regard to functional plasticity and structural plasticity. For example, Groussard et al. [27] studied semantic memory, that is, retrieval of musical facts such as style and/or composer. As stimuli, excerpts from the classical and modern repertoires were taken. Musicians and non-musician controls had to judge how familiar these excerpts were, using a four-point Likert-type scale for rating. (However, in order to study semantic memory more properly, an explicit naming task [e.g., which musical piece has just been heard] would be better.) Functional MRI revealed that musicians (in comparison with non-musician controls) had higher gray matter density in the left anterior hippocampus (also called hippocampal head) as well as increased functional activity in both the left and right hippocampi (Figure 3). Note that many studies testing semantic memory with authentic pre-existing musical stimuli have to deal (or struggle) with certain confounds. Because oftentimes, not only semantic content is remembered, but personal memories and emotional impressions are re-experienced too. In other words, semantic and episodic aspects cannot be properly disentangled. This, however, is less a methodological problem (or a design problem), but rather an inherent property of music itself and the effects it has on the listeners. Groussard et al. reported in regard to this point that the chosen “musical excerpts evoked personal memories in 85% of musicians, but only in 30% of non-musicians [27]”.

Figure 3.

When listening to familiar melodies both, semantic content and personal memories may come to mind. Plastic changes in the hippocampus are higher in musicians than in non-musicians. (Groussard et al. [27]).

A second point in neuroplasticity research is the time course, that is, the time frame in which structural reorganization occurs. To my knowledge, this issue was first addressed in a DC-EEG study conducted by [28]. In this study, Bangert and Altenmüller found co-activity in the auditory and sensorimotor areas already after 20 min of practice, that is, when beginners had had their first piano lesson. Jäncke [29] writes the following: “Memory consolidation is time-dependent since the biochemical processes modulating synaptic processes need some time (at least 25 minutes) to develop and to install the new and altered synaptic contacts in the memory networks, including the release of various hormones into the bloodstream (i.e., epinephrine, norepinephrine, and cortisol) [29].”

Several other studies used diffusion tensor imaging (DTI) to study reorganization on the cellular level. This time, also hippocampal and parahippocampal structures were explicitly examined (e.g. [30]). Most interesting in this respect is a study by Jacobacci et al. [31]. It is a sort of extended replication using DTI and fMRI since the paradigm itself has been developed by Bönstrup et al. [16] before. In both studies, participants learned to tap a four-finger movement sequence (4-1-3-2-4), which had to be executed as quickly and precisely as possible. In short, beginners trained their motor skills. To see how dexterity improved on the micro-level the focus was on the average tapping speed within the test blocks and also on processes during the short rest periods in between; in other words, on improvements micro-online and micro-offline (see [16, 31] for further details). Interestingly, the hippocampus and precuneus were strongly activated during rest, that is, not during task execution, and this functional activity was immediately followed by plastic changes in the same anatomical regions. In contrast to that, cortico-cerebellar and cortico-striatal circuits were most active during task execution, that is, within the test blocks. So, two aspects are important here: First, rapid types of functional and structural reorganization begin about 20 or 30 min after the very first training, which confirms that consolidation is initially dependent on biochemical processes in the hippocampus [29]. Second, most crucial is the point that the hippocampus not only plays a decisive role in the consolidation of declarative memories but also in that of procedural content, that is, when motor sequences are learned and made more stable. In consequence, this enables musicians to recall finger movements automatically and in the correct serial order. Jacobacci et al. [31], therefore, suggest a common hippocampal-based mechanism for both the formation of declarative and non-declarative memories.

2.2 Recognition memory for melodies and hippocampal involvements

Every time when listening to a musical piece one has come across before, two processes can be distinguished from each other—a pure “feeling of familiarity” on the one hand and the “remembering” of the musical details on the other hand. So, recognition memory consists of two components. Interestingly, both recognition types depend on the hippocampus, which has to be seen as a part of a larger cortico-subcortical network this time. Thus, melody recognition is another example revealing how the hippocampus reacts in the context of music. In detail, the feeling of familiarity is defined as a vague impression of somehow knowing a musical piece, and for estimating how familiar it sounds a 7- or 10-point Likert-type rating scale can be useful. Remembering, by contrast, is often tested with an old/new auditory recognition task to prove how well the details of musical structure have been kept in mind. In these auditory recognition tasks, musical motifs, or longer excerpts, are played again, either in the original version or with slight variations, and these target stimuli have to be compared with the imprints stored in memory. (Note that for reasons of clarity, brain regions other than the hippocampus are not described in this section here.)

Interestingly, a similar differentiation has been made in regard to episodic memory, that is, when personal events are retrieved with certain vividness. In episodic memory, the term “recollection” is used to describe the process of recall (and re-experiencing) as many contextual details as possible. So, distinctions are made between the “feeling of familiarity” on the one hand and “recollection” on the other hand. There is a strong consensus that the latter, recollection, is hippocampal-dependent; however, in terms of the vague feeling that some episodic memories seem familiar it is not clear if this is hippocampal-dependent or rather based on activity in the perirhinal cortex (see, e.g., [32] for further details).

Back to the feeling of familiarity in music, Plailly et al. [33] found evidence for the hypothesis that all types of vague impressions have a common neural source, based on the idea that the feeling of familiarity is quite similar across all sensory modalities. They tested this common source hypothesis with fMRI and two modalities, music and odors. Participants had to judge whether musical excerpts (instrumental) and odors were familiar to them or not. An underlying common brain network could indeed be identified, whereby brain reactions to familiar sounding music were stronger than those to familiar smelling odors, for example those of herbs or flowers. Interestingly, a distinct left-hemispheric tendency could be observed that also referred to the left hippocampus and the left parahippocampus. So, for this bimodal type of stimulation a large overlap in (left-hemispheric) neural substrates has been found (including the hippocampal formation), which suggests that the feeling of familiarity is multimodal indeed, or, one could also say: modality-independent.

In contrast to that, Watanabe et al. [34] conducted an fMRI experiment to find out how well the details of musical structure could be remembered. So, in Ref. [34] “remembering,” that is, the second hippocampal-dependent type of recognition memory, was examined, and old/new-decisions had to be made. To my knowledge, this fMRI study was the first in which co-activation of personal memories (“this reminds me of …”) was strictly minimized, since new musical stimuli had been generated with auto-composing software, by which any type of additional memory or association was kept at a minimum. During encoding, participants listened to 20 of these pieces (each of 3 s length, repeated three times). Then, 10 novel examples were added, and the subjects had to decide which stimulus was old (i.e., previously heard), and which was new. Brain activation was found in the right hippocampus for having correctly recognized pieces as old, so thus again retrieval success could be proven (cf. e.g., [3]).

Retrieval success was also the topic of a recent MEG/MRI study conducted by Bonetti and colleagues [35]. Again, old-new decisions had to be made; however, this time, the task was more demanding: During learning (or: encoding) participants listened four times to the right-hand part of a Bach Prelude (BWV 846) and also four times to its atonal counterpart. However, during retrieval, only short “snippets” (5-tone motifs) were presented. Half of these motifs were taken from the memorized prelude, whereas the other half was newly invented. Thus again, old examples had to be distinguished from novel ones, and this was done separately for the tonal and atonal versions. Overall, strong brain activity was observed in left MTL regions, including the left hippocampus and the left parahippocampus. Interestingly, this was only the case when the motifs were tonal and recognized as old, that is, as a part of the original prelude. Memorized atonal motifs, by contrast, evoked activity in right-hemispheric auditory regions, that is, without any significant functional change in the right hippocampus. In this study, retrieval was also examined in more detail in that event-related fields were analyzed in two frequency bands, a slow band (0.1–1 Hz) and a faster one (2–8 Hz). In the slow frequency band, voxel number was high for memorized tonal motifs from the third tone on, whereas for memorized atonal types voxel number was already high at motif beginning. These details suggest that in regard to tonal motifs, the hippocampus reacts to the whole, the entity or: Gestalt, in accordance with the observation that the hippocampus is engaged in the processing of patterns or “scenes” [cf. [7]]. Atonal motifs, by contrast, were processed tone-by-tone by activating large parts of the auditory network. Obviously, participants use different listening strategies during motif recognition: holistic whenever the replayed motifs are tonal (activating the hippocampus), and punctual whenever the motifs are atonal (activating the auditory network, including the insula).

Both studies on recognition memory [34, 35] reveal that the time windows between learning and recognition were very short. In other words, the test phase was immediately after encoding, with a delay of 15 min at the most (cf. [34]). This calls into question that the hippocampus is exclusively involved in long-term memory processes. Some researchers (e.g., [36, 37], also [5]) argue in favor of a more flexible approach. They understand hippocampal functioning beyond strictly set time frames through which by definition (hippocampal-independent) working memory and (hippocampal-dependent) long-term memory processes can be distinguished from each other. For example, Jeneson and Squire [36] point out (note that this is from a working memory perspective) that some task requirements; for example, greater memory load or maintenance of some complex information may occasionally exceed the capacity of working memory, and in such cases, the hippocampus is needed to support performance. Accordingly, in regard of working memory processes two new terms are suggested [36]: “subspan memory” and “supraspan memory”—the first for all those cases in which the storage capacity of short-term memory is sufficient. And the latter, supraspan memory, for the opposite, that is, when also long-term memory is needed to solve the task for which activity can be observed in the hippocampus.

3. Expert memory: strategies for memorizing and recall

3.1 Long-term working memory (LT-WM), chunking, and memory subtypes

Interestingly, this flexible interaction between working memory and long-term memory (plus hippocampal activity as just described) (see Section 2.2) is one out of three fundamental mechanisms by which the effectiveness of expert memory can be explained when skillful pieces have to be played by heart. So, in this section, I will elaborate on these three points to describe in detail the mechanisms underlying expert memory. These points are: (a) the use of a so-called retrieval scheme (which is based on flexible long-term/working memory interaction processes), (b) a principle called chunking as well as (c) different memory subsystems needed for expert performance.

1. Today, several studies (or: authentic reports) exist in which professional musicians describe their steps when preparing a new recital program. A successful example is the report series by Roger Chaffin (US-American psychologist) and Gabriela Imreh (US-pianist, of Romanian origin, e.g., [38], also [39]). According to them, a large part of daily practice consists in acquiring a so-called “retrieval scheme,” which is part of conceptual (or: declarative) memory. For this, a musical piece is first partitioned into several segments (or: sound portions), and for each of these segments, a self-set landmark (i.e., a cue or an opener) is consciously memorized too. (Cues can be, for example, a structural detail, a sound, or a tricky fingering.) During recall, that is, during performance on stage, soloists just proceed from cue to cue, present in working memory, whereby each sound portion, stored in long-term memory, is directly accessed too. So, all in all, this mechanism makes fluent performance possible. For this linkage (or: interaction) between working memory and long-term memory, Ericsson and Kintsch coined the term “long-term working memory” (abbreviated “LT-WM” [1], see also [40]). However, to my knowledge, no study exists, in which LT-WM processes underlying time-precise sequence recall have been investigated with fMRI. At first glance, some striking parallels between LT-WM and Teyler’s so-called “hippocampal memory indexing theory” might come to mind (see [18, 19]). According to the indexing theory, certain pointers are located in the hippocampus, through which the neocortical sites of a distributed array are re-activated so that the whole memory pattern can be restored. However, in regard to this parallel, some counterarguments exist: First, Teyler’s theory is a neurophysiological one, based on the existence of reciprocal connections and synaptic physiology. LT-WM, by contrast, explains retrieval processes from a pure psychological point of view. A second point is that Teyler’s theory provides the basis for episodic memory, enabling a subject to retrieve certain personal memories with all contextual details [19]. In other words: Teyler’s theory might be suitable for a scene-like type of retrieval. However, to remember tone elements in serial order and precisely in time, as in performance situations, LT-WM seems to be the more suitable explanation: It describes a flexible, cue-triggered (WM-based) mechanism, making possible a rapid and reliable access to large amounts of domain-specific information, stored in long-term memory.

2. A second hallmark by which the efficiency of expert memory can be explained is a principle called “chunking,” more precisely: a principle of thinking and performing in chunks. Chunks are defined as meaningful units of information (e.g., [41]), and the level of argumentation is working memory. In musical contexts, chunks often are pattern-like. Typical examples are short segments of musical scales that can be found in various musical pieces. From this follows that melodies are not processed tone-by-tone, but rather as groups or distinct entities, namely as chunks. Note that chunks can easily be detected in the written musical texts, the scores, since almost every composer adds certain slurs to indicate that a bunch of notes has to be played coherently as an entity, namely as a chunk or a musical phrase (see Figure 4). So, the key unit in music is the “musical phrase” rather than the chunk, also implying that every type of musical information is stored in musical phrases. Interestingly, it has been found that certain working memory processes occur at the boundaries of the musical phrases, and these transitional processes are more important than generally assumed. Because for grasping a melody as a whole, a first phrase has to be maintained in working memory, while the attention focus is directed to the next phrase. In other words, while playing or listening to melodies certain memory-and-attention-related micro processes occur, in particular at the edges of the phrases, the phrase boundaries. Interestingly, these transition processes in melodies do activate the hippocampus. Knösche et al. [42] found activity in the posterior parts of both hippocampi and also in parahippocampal areas, when one musical phrase (or chunk) was closed, and the next phrase was opened (for source localization EEG and MEG data were used) (see also [43]).

3. Finally, expert musicians rely on different memory types in parallel, by switching the attention back and forth between them. At least five subtypes can be distinguished from each other: Conceptual (or: declarative) memory in which the retrieval schemes and the details of music analyses are stored. Another subtype is auditory memory in which the entire musical piece is represented as a sound product. In addition, also motor, kinesthetic, and haptic subtypes exist, carrying memory details in terms of finger movements, muscle tension, and somatosensory impressions, for example, how it feels when strings are plucked. Some musicians also possess eidetic or photographic memory, which enables them to see the details of the score before the inner eye (see [38, 39] for further details). So, expert musicians memorize a musical piece in a multifaceted way. In other words, several backups exist in parallel, each with different pieces of information: auditory, conceptual, motor, haptic, kinesthetic, and/or visual. To my knowledge, studies, using fMRI and DTI, are still missing, in which the focus is set on how these memory subsystems interact with each other. However, in terms of music perception, that is, when subjects passively listen to music, some whole-brain connectivity studies do exist, and for this, all types of neural networks could be identified (e.g., see Alluri et al. [44]).

Figure 4.

A music example of high structural density, always played from memory during a recital: Frédéric Chopin Piano Sonata Nr 1, c minor, op. 4 (excerpt). Musical phrases, the chunks, are indicated by the larger slurs, whereas some of the smaller ones are signs for articulation. Some are also certain “ties,” indicating that two notes of same pitch are attached.

3.2 Mnemonic techniques: Method of loci

Whenever the to-be-remembered items have to be stabilized even further, people sometimes make use of a certain strategy of double encoding, commonly known as “mnemonic technique”, for which hippocampal activity can be observed once again. Across the mnemonic types, the principle is as follows: (1) During encoding a primary item is tied (or: connected) to a secondary item, and this linkage is consciously memorized (e.g., [45]). Often, the primary item is complex or abstract; however, it is “the real thing,” that is, the original piece of information that has to be remembered. The secondary item is a self-chosen adjunct, which, in most cases, is familiar to the subject and also pictorial in nature. (2) During retrieval, this additive, that is, the secondary item, can be easily remembered and by this, the primary item is activated too. In general, mnemonic techniques are used for maintaining sequence order, for example, when 100 words have to be recalled in the same order as they were presented. It becomes clear that connections are mostly one-to-one; that is, assignments are made between one primary and one imagined item [45]. Some types of mnemonic techniques exist in which both first and second items are merged, that is, integrated into a new unconventional entity (or: “composite” [46]), for which bizarre imagery is explicitly required [47]. It should also be mentioned that all types of mnemonic techniques are bi-modal; that is, both the primary item and the adjunct are taken from different domains (e.g., digits and pictures, faces and objects, words and places). The main point here is that these bi-modal connections (or: “between-domain associations” [48]) are produced not before reaching the hippocampus proper, that is, both items converge into the hippocampus, and bi-modal binding occurs exclusively there (see Mayes et al. [48] for further details). Obviously, this is not the case for “within-domain associations” (face-face, object-object, word-word), which “converge sufficiently in the perirhinal cortex to create a […] memory representation [48].” Zeineh et al. [49] could demonstrate with fMRI that during bi-modal binding different hippocampal subfields are active, depending on whether data acquisition is during encoding or during retrieval. In this study [49], face-name associations were built and retrieved; in other words, names had to be assigned to new faces. During encoding, a complex of adjacent regions was active, in detail, CA fields 2 and 3 and the dentate gyrus (in short: CA2, CA3, DG), whereas during retrieval, changes in activation were primarily found in the posterior subiculum. (Note that face-name pairing tasks are quite frequent in everyday life situations, e.g., when a teacher has to memorize the names of all schoolchildren of a new class after summer vacation).

The oldest mnemonic strategy is the famous “method of loci,” going back to Greek and Roman antiquity. For example, Marcus Tullius Cicero used this method for learning his long speeches by heart. The method of loci is based on visuospatial imagery and mental navigation, and the procedure is as follows: (1) During encoding, a familiar environment is chosen in mind (sometimes also physically), for example, rooms within a house or paths through the city or countryside. Anchoring (or: encoding) is in such a way that the primary items are put at distinct places along this path. (2) During retrieval, one simply imagines walking along the path, while, simultaneously, the anchored primary items come to mind in proper order. This method is useful when word lists and other types of sequential information have to be remembered. Cicero described the method of loci in detail in his De oratore, a textbook on rhetoric, which is one out of three sources informing about this “ars memoriae” in ancient times [50] (Note that the word mnemonic technique (or: mnemotechnic) was coined only in the nineteenth century.) It is said that Cicero used certain buildings of the Forum Romanum for this linkage which helped him remember the main points of his speeches reliably and in proper order when speaking to the audience [51] (Figure 5).

Figure 5.

Cicero described the “method of loci” in De oratore, a textbook on rhetoric. For anchoring, he himself used certain buildings of the Forum Romanum which helped him remember the main points of his speeches in proper order.

Even though the method of loci is mainly used for memorizing word lists or large amounts of text, musicians occasionally make use of it too (violinist, personal communication). While preparing for a concert she physically moved from corner to corner of her practicing room, each time ‘depositing’ a certain amount of sound information in one corner. During the concert, she simply imagined walking along the same path. Given that for this type of linkage a path through nature is chosen, an additional effect could be to feel mentally (and physically) relaxed. In other words, during recall, while performing a concert on stage, an imagined path through a natural environment could reduce inner tension. Since violinists play pieces from memory in upright position, this also raises the question whether the method of loci supports some egocentric (or: trunk-related) processes which, however, are located in the parietal lobe (for further details on egocentric frames and representations, see [11]). If this were the case, the retrosplenial cortex might play a role in mediating between the hippocampus and the superior parietal lobe (SPL). The latter, SPL, is a multimodal spatial reference system into which visual, auditive, vestibular, tactile, kinesthetic—and, possibly, geocentric spatial—inputs converge for purposes of proprioception, that is, to bring room coordinates and one’s own body position into balance (I do not want to elaborate on this further, since, to my knowledge, this lacks of empirical proof). What has been shown, however, is that after six weeks of daily mnemonic training with the method of loci (a web-based training platform) memory performance of naïve subjects increased significantly. More importantly, significant changes in network connectivity were found in the left hippocampus and bilateral retrosplenial cortex (cf. fMRI study by Dresler et al. [52]; see also [53] for similar results).

3.3 Mnemonic techniques: playing the North Indian tabla as a special case

In many non-Western cultures, especially India, mnemonic techniques are even more essential since in these countries, music does not exist in written form as scores, but rather it is transmitted aurally/orally [54]. Although mnemonic systems are quite similar across these non-Western cultures, the basic mechanism is entirely different from the principles on which the method of loci is based, namely spatial navigation and visual imagery. In non-Western cultures, connections are sound-to-syllable, that is, the sounds of a musical instrument are combined with certain syllables which are thought and/or spoken simultaneously while playing. At first glance, these syllables seem to be nonsense and arbitrary chosen. However, spectral analyses reveal that musical sounds and the syllables share common acoustic properties (see [55] for further details). Regarding non-Western traditions, the interesting point is the aspect of weighting or primacy, in other words, how sound-syllable connections should be understood. Which element is the predominant one—the musical sound or the speech syllable, that is, the first or the second item? This will be discussed in the following by taking as an example the North Indian tabla tradition. A tabla is a pair of hand drums of different size and pitch played in North India (Figure 6). Singers and sitar players are accompanied by this percussion instrument in small ensemble formations typical of the North Indian (Hindustani) classical music tradition.

Figure 6.

The tabla is the typical percussion instrument of North India: A pair of hand drums of different size and pitch is played by one musician seated on the ground. Different drum sounds (depending on where and how the drums are struck) are linked with spoken syllables, called bols, resulting in certain sound-to-syllable connections. Playing the tabla is a special case of using mnemonic techniques.

There are many different ways of striking the tabla, however each time, a single drum sound is linked with a certain spoken syllable, called bol (from bolna, a Hindi verb for “to speak” [56]). Example syllables are: dha, dhin, dhun, tin, tun, kat, ghe, tra, kra [55, 56], thus, appearing as a sort of onomatopoeic equivalent for the drum sounds. However, the important aspect in terms of weighting or primacy is that bol syllables should not be considered as mere adjuncts. Because in India, the human voice is regarded as the origin of all music, the instrumental and the vocal forms, by which cosmic energy flows through the chest and the lungs to be formed by the vocal cords, the mouth, and the lips. Manuel and Blum [57] put it this way: “Musical sound […] proceeds along a spiritual pathway from an unmanifested ideal form, through the navel, heart, throat, and finally the mouth. Vocal music, generated by vital breath […], was conceived as a sublime manifestation of nâda brahma, a sort of primordial and divinely animated substratum of cosmic sound. [57].” Accordingly, vocal utterances are of particular value, and this holds also for the bol sequences. In this context, Rowell [58] writes the following (though the focus seems to be more on the South Indian counterpart, a syllable system called solkattu, [59]): “These syllables are more than drum syllables - they are abstract phonetic patterns that can be recited, played, and danced. […] They are not medium-specific.” So, the question about weighting or primacy can be answered in that way that drum sounds and phonetic elements should be considered as equivalent.

Regarding this sound-syllable system the wide range of applications becomes evident when having a look at the non-Western aural/oral teaching practice. This will be briefly explained by taking as an example the music education in Japan. Both Hughes [60] and Shehan [54] report that beginners learning to play the nōkan (a traditional Japanese flute) are taught to—first of all—sing the syllables in pure form before being allowed to pick up the flute and play the melody. (Note that in the Japanese nohkan tradition, the mnemonic syllables “hya” and “hyo” are central.) Obviously, the same holds for children in North India learning to play the tabla. Farrell [56] describes this learning process as “saying bols, writing them down, translating these into finger movements,” meaning that the first step consists in building a mental representation of the sound syllables—enabling the novices to think the mnemonics fluently—before hand movements, that is, motor practice is added. Farrell continues: “The process of thinking, saying, and playing in tabla [sic] provides a fascinating example of how the building blocks of a music are conceptualized in thought and realized in sound [56].”

Another aspect worth considering is whether it is the sound of the tabla to which the spoken syllable is connected or the drum stroke itself (cf. [55]). In case of the latter, a close link would exist between the syllable and the sound-producing action. If so, the (chronological) order would be: “bol - action - sound” which is in line with the ancient Sanskrit wisdom that “the Word is translated into action [61].” From a neuroscience perspective, this might go into the direction of the human mirror neuron system: It is common knowledge that a special subset of audiovisual mirror neurons does not only respond when certain hand actions or gestures are observed, but also when the sounds resulting from these hand actions are heard [62, 63]. For listening to these action-related sounds, a left-hemispheric frontoparietal motor-related network has been identified [64], including Broca’s area and the premotor cortex. This might also be active when bol sequences are played on the tabla, that is, when drum strokes (as sound-producing actions) are combined with syllables (the speech sounds). Note that with “tonic sol fa,” or “solfège”, a similar sound-to-syllable system exists in the Western culture. The principle is that an easily singable syllable, for example, do, re, mi, fa, or sol, is assigned to each of the seven pitches of the diatonic scale. However, in contrast to non-Western systems, these pitch-syllable pairs do not have any acoustic similarities. The practice of this solfège system can be traced back to Guido d’Arezzo, a famous music theoretician in the first half of the eleventh century, and it is mainly used for intonation exercises in singing lessons and for training auditory imagery until the present day (see, e.g., Hughes [60] for more details on Western arbitrary versus non-Western non-arbitrary sound-to-syllable systems).

4. What is memory for? Some reflections from the perspective of culture

So far, it has been shown that the hippocampus is one of the most relevant brain regions in humans, since almost every memory process—declarative and non-declarative—depends on hippocampal activity. Furthermore, to illustrate how sensitive the hippocampal region in terms of time-related micro-processes is, expert memory in music served as an example, manifesting itself in pitch- and time-accurate sequence recall. In this section, this data-driven approach is put aside in favor of reflections on memory from the viewpoints of anthropology and culture. What is memory for? What is its value? While searching for answers it makes sense to distinguish between the individual and the collective level. Regarding the first, the individual level, at least two points should be mentioned here: First, memories shape our character. That is, each remembered experience enriches the self, making each human unique, that is, different from others. Personal memories also guide people during decision making and help them avoid making the same mistake over and over again. Accordingly, memories play a decisive role to give persons a sense of identity. Second, semantic memory, that is, facts, and also know-how are some sort of preconditions for being creative. In other words, a fund of knowledge is beneficial for creative problem solving, enabling persons to build associations and have new ideas. For instance, during his first stay in Italy (1769–1771), the young (13-year-old) Mozart seized the opportunity to study counterpoint with Padre Martini in Bologna, a famous music theoretician at that time. Thus, although knowledge is often domain-general, the domain-specific type is even more useful, for example, to improve one’s compositional style.

In the middle of the 1980s, human memory also became a topic in the socio-cultural sciences. Assmann [65] differentiates between memory as an “outer or: external phenomenon” (studied by cultural scientists) and memory as an “inner phenomenon” (studied by neuroscientists and psychologists). Moreover, den Boer [66] set both memory types, the internal and the external, in relation to each other. He compared the method of loci (the internal type, located in mind) with the so-called lieux de mémoire (the external type), that is, with places of memento in reality, such as certain monuments or holy places. (NB lieux de mémoire is a term coined by Pierre Nora, a French historian, in the late 1970s). In terms of inner and outer, den Boer [66] has the following opinion: “For the ancients, the loci memoriae were a necessary mnemotechnics in a society without modern media […] For Cicero and Quintilian the loci memoriae were practical mental tools, free of ideology. […] Nora’s lieux de mémoire are also mnemotechnical devices, but […] far from being neutral or free of value judgments. Most lieux de mémoire were created, invented, or reworked to serve the nation-state.”

First attempts to consider memory as something social were made in the 1920s by Maurice Halbwachs, a French sociologist. He coined the term “mémoire collective” (collective memory), which should, more or less, be understood as the common knowledge of society. In other words, specific memories are of concern to a group or the entire society, and, to some degree, they are also present in the mind of each individual. Several measures are taken to keep these group and nation memories alive; for example, minutes of silence or days of remembrance to honor those who were killed in war or by assassinations, but also national and religious holidays belong to this category. In short, the past is tied to the present; that is, collective memory is refreshed by repetition at regular intervals, mainly from year to year. Thus, joining festivities or commemorative ceremonies may strengthen a person’s cohesion with the group, in other words: his or her cultural and/or national identity.

Another important point is that, since ancient times and throughout all societies and ethnic groups, certain persons stand out from the crowd through calling myths, heroic epics, and/or group-related events back to mind by transmitting them orally, for instance, by telling these stories on market places. The aim is to strengthen the audience’s cultural identity. Assmann [65] explains this point by taking as an example the “griots,” a caste of wandering singers in Africa. But also shamans, bards, heads of tribes, and, in a broader sense, priests and teachers are good examples. All have in common to be explicit carriers of internalized knowledge (somehow serving as a medium), and with this, they help preserve oral traditions and culture. What they communicate are, for the most part, words of wisdom and insights, that is, facts about the past, what had happened to a tribe, combined with own experiences and some advice for the future. Walter J. Ong, an American media theorist [67], set out that griots, bards, and other carriers of wisdom make use of certain poetic devices (functioning as a sort of mnemonic device) to transmit their narratives properly from generation to generation. Such poetic devices are, for example, rhymes, alliterations, assonances, formulaic expressions as well as many repetitions and redundancies regarding content. In consequence, a story can be clearly remembered and exactly repeated by the storyteller, while, at the same time, this is catchy for the audience. The effectiveness of such poetic devices was verified in a study by Tillmann and Dowling [68]. They conducted an experiment on short-term memory, using two kinds of verbal material, prose and poetry. In this study, short-term memory for surface details declined significantly later in poems than in prose. This is because in poetry, the metrical structure and the rhyming at the end of the verse lines tie phrases together, resulting in a coherent whole. So, in this study subtle distinctions could therefore be easily detected in poems, for example, when after delays of either 4 s or 30 s certain lures were presented that had to be distinguished from the original verse lines (verbatim) (see [68] for further details).

Regarding oral traditions, a further point is of utmost importance: What is passed on by word of mouth? Is it the gist, that is, the storyline, in other words the central plots and aspects leaving space for improvisation? Or is it that all the details can be precisely remembered and recited even three or four generations later? The answer goes in both directions, since this depends on the type of narrative and on the efforts made during encoding. Anyhow, when holy wisdom has to be transmitted in religious contexts oftentimes the latter is the case. Rowell [61] describes the process of verbatim recall as follows (referring to collections of hymns from the Rgveda): “Young trainees were put through memory-building routines that boggle the mind: recitation of the text both with and without consciousness of its meaning, recitation both forward and backward, recitation both with phonetic junctures […] between words and with separations between words, metrical recitation and recitation in a continuous flow (except for the obligatory caesuras), and recitation of pairs of syllables in distorted […] sequences. […] The aim was to instill an automatic and total command of the text that would rule out even the slightest possibility of error.” Assmann puts it this way [65]: “The Vedas are not written down because Brahmans trust in written texts less than in memory. Each holy text is a kind of spoken temple, making the Holy present through the medium of voice.”

From the neuroscience point of view, an interesting result was obtained in a study by Robin and Moscovitch [69]. There, hippocampal activity could be observed for both—the gist and the details, in other words, for coarse-grained and fine-grained aspects. An example is: “party” (gist) vs. “cake at 10th birthday party” (details of a single episode). In terms of this, the hippocampus shows activity in different subregions along the long-axis. Robin and Moscovitch [69] propose the following: “Perceptually detailed, highly specific representations are mediated by the posterior hippocampus and neocortex, gist-like representations by the anterior hippocampus […] These representations can co-exist and the degree to which each is utilized is determined by its availability and by task demands.” Obviously, this seems to be valid for both processes, encoding and retrieval, showing once again, the dynamics of the hippocampus proper.

Another example of verbatim recall can be found in the religious practices of the Islam: The Quran is considered as the word of God, that is, as literal truth, and any type of misquoting, or passing it on in incorrect form is considered a sin [70]. Furthermore, the Quran is recited in in its original language, Classical Arabic, without exception. In other words: no translations are allowed which is in stark contrast to Christianity. Since best effects are achieved in aural/oral form, Muslims prefer reciting (and memorizing) to reading the Quran in silence. Thus, religious Muslims make learning the Quran their life’s work, starting at the age of four or five when attending the Islamic preschools and the Quranic schools [71]. This exemplifies that every religious Muslim is the carrier of (internalized) Quranic knowledge, guiding him (or her) through life as a sort of moral compass. Boyle [71] speaks of “Quranic memorization as a process of embodiment.” In this case, however, learning is mostly by rote, that is, listening to the Quran teachers (and/or reading the Quran), and then repeating and rehearsing the text word by word and line by line ([71], see Figure 7). Some Muslims also make use of their eidetic or photographic memory through encoding the Arabic calligraphy too. Saleem [70] interprets this as a mnemonic mechanism by which the pictorial calligraphic characters are the secondary items so that each sura is memorized in the form of sound-picture combinations or “letter-sound relationships [70],” to use his own words.

Figure 7.

Memorizing the suras of the entire Quran is a lifelong task, starting at the age of four or five when attending Islamic preschools and the Quranic schools. In principle, every religious Muslim is the carrier of (internalized) Quranic knowledge, guiding him (or her) through life as a sort of moral compass.

However, the interesting point is here that Quran memorization is purely through the sound of the language, and for this, the phonetic characteristics of Classical Arabic are crucial. That is, no semantic analysis or deeper understanding of the meaning is required [70, 71, 72]. This special, sound-based practice of memorizing has to be seen against the background of religious aesthetics. Navid Kermani, a cultural researcher of Persian origin, puts emphasis on the point that the Quran should be experienced in an aesthetic (and acoustic) manner, that is, sensually, in a poetic way through the attractiveness of sound. He has elaborated on this in his book God is beautiful. The aesthetic experience of the Quran ([73], see also [74]). Interestingly, a similar tendency prevailed in the Christian Occident during the Middle Ages. Yates [50] reported that in the time of Albertus Magnus and Thomas Aquinus (both thirteenth century) the mnemonic techniques moved over from rhetoric to ethics and have been regarded as part of virtue theory ever since.

5. Conclusions

In this book chapter, playing music by heart served as an example to illustrate that sequence recall is dependent on the hippocampus, a brain region within the medial temporal lobe. In detail, hippocampal activity was evoked by micro-processes at phrase boundaries [42], during motor sequence learning [31], and when detecting irregular offset-onset intervals [4]. Thus, emphasis was on expert memory, that is, on entirely different aspects than those in episodic memory, when certain unique experiences are remembered (e.g., [5, 6]). This performance example has also shown that hippocampal activity is needed every time when memory processes are required in the performing arts, that is, in contexts specific to humans. On the other hand, numerous cross-species studies exist, in which the hippocampi of birds, rodents, and monkeys have been studied in terms of anatomy, functioning, and connectivity, illustrating that memory is also of immense value for survival [cf. [11]]. For instance, sparrows or squirrels, remembering during wintertime where they have laid in their supply of nuts, are good examples to show how vividly important spatial memory is from a Darwinian point of view.

In this overview chapter, it has also been shown how dynamic and flexible the hippocampal brain region is, oftentimes, because different hippocampal subfields along the long-axis are active. Example processes, illustrating this flexibility, are: (1) changes between the gist and the details [68, 69], (2) flexible interactions between long-term and working memory [1, 36], (3) reactions to retrieval success and to the opposite, that is, novelty detection [4, 34, 35] as well as (4) reactions to the feeling of familiarity and the remembering of structural details [33, 34, 35]. In many studies, hippocampal reactions were also compared between groups, in particular between musicians and non-musician controls, showing that the hippocampus of musicians is not only a prototype example in regard to neuroplasticity but also in terms of many other respects. (Still, hippocampal plasticity is one of the most studied topics in the field of Music and Neuroscience (e.g., [27]).) The disadvantage is that, until now, there are many gaps in research; that is, functional MRI data are missing on how the hippocampus might react to musical aspects other than time, making it unavoidable to sometimes argue by analogy. In other words: many suggestions need verification.

Occasionally, musicians make use of the (hippocampal-based) method of loci, the oldest mnemonic strategy based on visuospatial imagery and mental navigation. For all mnemonic techniques, bi-modality is characteristic, meaning that both the primary item and the adjunct are taken from different domains to stabilize content by way of double encoding. Zeineh et al. [49] could demonstrate with fMRI that during bi-modal binding different hippocampal subfields are active, depending on whether encoding or retrieval processes are examined. Further research on hippocampal subfields (see, e.g., [21]) is a promising avenue to explore the functioning of the hippocampus in all its detail. Mnemonic techniques are even more essential in non-Western cultures, since there music is frequently transmitted in aural/oral form [54]. This has been exemplified by the North Indian tabla tradition. Whether and to what extent the hippocampus is involved when drum patterns are played and certain onomatopoeic syllables are thought or spoken simultaneously is a matter for future research, and suitable paradigms have to be developed to disentangle the process of drum sequence playing from that of syllable speaking.

Further points of interest are implications regarding culture, more precisely: which relationships exist between cognitive neuroscience of memory on the one hand and the socio-cultural sciences on the other hand. In this respect, the following points are important: (1) memories shape the self and make humans unique; (2) knowledge stored in semantic memory is beneficial for creative problem-solving; (3) collective memories strengthen a person’s cohesion with the group, that is, his or her cultural and/or national identity, (4) throughout all ethnic groups shamans, bards, priests, and teachers are explicit carriers of internalized knowledge, thus helping to preserve oral traditions and culture; (5) religious knowledge, when internalized, may serve as a moral compass for each individual, this is especially true for religious Muslims memorizing the Quran their whole life.

Thus, one can conclude that the hippocampus is indeed one of the most relevant brain regions in man, having effects on almost every aspect of human life from basic survival to moral behavior and the preservation of culture.

Conflict of interest

None. I am the single author.