Performance of any skilful activity is regulated by a highly refined and integrated system that includes motor planning, sensorimotor integration, execution and adaptation, following either disruption or improvisation. A complex, integrated feedback loop is created by receiving information from sights and sounds, and by interacting with the external environment (the music, the orchestra/other musicians, the conductor, and the instrument). The ability to harmonize a physiological response to changes in environmental alterations relies on the fluid integration of multisensory stimulus and appropriate physical adjustments in motor control.

Motor control is a learned skill that is continually developed throughout one’s life. The success of a planned motor response will be determined by a number of factors. These factors include motivation and goal-identification, experience of the same or similar tasks, success and rewards. Underpinning these cognitive processes is a complex omnidirectional network. This comprises multisensory inputs via afferent feedback from the environment, for example visual, auditory and somatosensory (both tactile and kinaesthetic, see Demain et al. (2013) for further information). Sensory information is then integrated depending on the intention of the task, for example integration of the visual and proprioceptive (joint positions) senses to estimate a movement trajectory toward a target (piano keys). Sensory information then enters the cerebral cortex, which includes the motor cortices; premotor, motor and supplementary cortices (Enoka, 2008; McMillan and Carin-Levy, 2012).

Motor activity originates in the motor cortices, basal ganglia and cerebellum. Voluntary and automatic movements are initiated in the motor cortex and basal ganglia respectively. The cerebellum integrates vestibular, visual, proprioceptive and tactile sensory information, and by using this integrated information, adjustments can be made to cortical output to modify the amplitude and trajectory of movement, for example when you react to a perturbation in the environment (Enoka, 2008). Adjustments to the movement are auctioned via descending neural pathways, including the corticospinal tract. In addition, the vestibulospinal and reticulospinal tracts, arising in the brain stem, ensure the appropriate postural tone in the trunk and shoulder girdle, thus stabilizing the upper limb and allowing flexible control of the wrist and fine, dexterous movement of the fingers.

Signals then travel via the upper motor neurones, which terminate on the anterior horn cells in the spinal cord. Action potentials generated in the anterior horn cells are conducted via lower motor neurons, which synapse on the muscles and produce a muscle contraction (McMillan and Carin-Levy, 2012). The strength of the muscle contraction depends on both the number (spatial summation) and rate (temporal summation) of lower motor neurone activation. Finally, further control of anterior horn cells output is achieved through direct sensory input at spinal cord level, such as what occurs in a tendon reflex.

The cognitive, sensorimotor and neurophysiological elements involved in the preparation, control and execution of a functional movement have been well defined in the literature, and therefore to review these in depth would be supplementary to the primary focus of this review. Therefore, for more information on these areas, refer to Jeannerod (1997), Peretz and Zatorre (2003), Jones and Lederman (2006), and McMillan and Carin-Levy (2012).

With specific reference to measuring musical performance, imaging studies have been used to identify cortical activity. During magnetic resonance imaging (MRI) and magnetoencephalographic (MEG) studies, professional musicians (compared to amateur and non-musicians) showed a greater volume of gray matter and/or enlarged cortical representation in the primary motor, somatosensory and premotor areas, the cerebellum, and the anterior superior parietal areas, thus illustrating honed ability in sensorimotor control, audio-visual and spatial processing (Elbert et al., 1995; Schlaug, 2001; Münte et al., 2002; Gaser and Schlaug, 2003; Hutchinson et al., 2003). In addition, Bengtsson et al. (2005) used diffusion tensor imaging to study the how piano practice affects white matter between three different age groups (children, adolescents and adults). The results showed that within each age group, practice was positively correlated with fiber tract organization in different regions. These results are interesting because they indicate structural differences in the musical brain: a powerful motivation for increasing the potential for intellectual endeavor. With reference to the study by Bengtsson et al. (2005), the results also suggest that the training undertaken by a practicing pianist while the fiber tracts are maturing induce white matter plasticity. Thus an increase in activity, initiated by piano practice, causes increased myelination (the increase of myelin around axons of neurons). For a detailed review on neuroimaging of structural plasticity see Zatorre et al. (2012).

It is at the level of cortical organization that the relationship between complex hand dexterity and bimanual instrumental practice can be observed. Elbert et al. (1995) showed increased cortical representation of the left hand (fingering hand) in string players versus the right hand (bowing hand); indicating the musician’s choice of instrument influences changes at a cortical level, and thus makes findings less generalizable to “musicians” per se. This review will specifically concentrate on piano performance, where bilateral fine motor control is compulsory. For more information on imaging techniques and reviews on neuroplastic changes underpinning motor control and learning, see Münte et al. (2002), Watson (2006), and Furuya and Altenmüller (2013).

Despite this interest, our understanding of complex hand dexterity is still in its infancy. Fundamental questions still remain; particularly surrounding how function is honed in various activities for virtuoso expertise to manifest. The multidisciplinary nature of this area brings together neuroscience, motor learning and control, physiology, biomechanics, anthropology, behavioral, and cognitive sciences. This narrative review will focus on the relatively new use of biomechanical methods and their emerging contribution to this subject, given the complex nature of analysis required. This review presents a comprehensive background section giving an overview of complex hand dexterity (See The Human Hand in Performance). This is followed by an historical overview of the pioneering techniques developed for capturing movement (See Historical Methods for Capturing and Measuring Movements) and current biomechanical methods, derived from those techniques, to measure complex hand dexterity (See Current Biomechanical Methods of Measuring Complex Hand Dexterity During Musical Performance).

The versatility of possible hand movements available to complete any given task has made standardization of activity difficult, therefore hindering comparative investigation (Metcalf, 2009). In addition, the complexity of movements and the smaller anatomy of the hand have historically limited our ability to measure all the composite movements of the wrist, hand, fingers and thumb. Hand movements provide a conduit for examining how skill manifests at a physical and actioned level. It is the result of intention.

Hand dexterity can be thought of as a spectrum (see Figure 1); from impaired, affected function through to highly skilled, virtuoso performance. Complex hand dexterity describes the skill(s) that are required for increasingly higher levels of function. Thus a virtuoso performer will exhibit the highest levels of complex hand dexterity in their given activity.

Musical expertise is extraordinary. The speed, fluency and tempo of elite instrumentalists can be mesmerizing to witness and often audiences will favor seats where they can observe feats of seemingly impossible dexterity and control during a performance. Musical expertise is arguably the ultimate example of elite performance in complex hand dexterity. Aspects of motor control and learning, precision, timing, strategies for compensation, adaptation and coordination can be observed during playing. Analysis of instrumentalists allows us to study complex hand dexterity just as the analysis of elite sports allows us to study whole-body dynamics.

Glazier et al. (2006, p. 50), stated “movement variability is an essential feature of human motor behavior that affords the sensorimotor system the necessary flexibility and adaptability to operate proficiently in a variety of performance, development and learning contexts” (Bernstein and Popova, 1930). The literature widely supports this statement and, particularly with reference to the upper limb, is an essential factor in allowing the versatility of movements required to perform functional activities. This versatility has also been referred to as the “degrees of freedom problem” by Bernstein (1967b). He commented on the redundant degrees of freedom (DOF) observed in the upper limb, and how this facilitates the complex range of solutions to everyday activities. Bernstein also commented on the human capacity to hone the multiple and redundant DOF in the upper limb to facilitate highly skilled activities, such as musical expertise (Bernstein and Popova, 1930).

Manifestation of skill can therefore be observed and quantified by the analysis of movement. Studying movement allows analysis of the functional strategies adopted by an individual to achieve a goal, whether that goal is playing a piano to produce “musically beautiful sounds” (Bernstein, 1967a), holding a teacup or using a tool.

The following section will review the historical techniques for capturing and measuring movements during the late nineteenth/early twentieth century and provide the precursors for current biomechanical methods of measuring complex hand dexterity during musical performance.

Modern methods of optical motion capture can trace its origins to the late nineteenth century and the pioneering work of individuals such as Marey (1873), Muybridge (1887), and Braune and Fischer (1895). The enthusiastic adoption of the newly developed field of photography, and particularly sequence photography, allowed human and animal movement to be not only captured, but analyzed for the first time.

In 1878, a photographer, Eadweard Muybridge, devised an experiment whereby a horse would gallop through a track lined with still photographic cameras. These individual cameras were triggered in succession as a horse galloped down the track by threads placed along the width of the track and attached to the camera shutters; capturing the movements of the gallop. Muybridge then “played” the silhouette of each image in sequence using a Zoopraxiscope. This method was groundbreaking and considered to be the first movie projector.

By 1884, Etienne-Jules Marey, a physiologist, had designed a black suit with white tape down the arms and legs, which was to be the first method of marker-based motion capture. Marey also revolutionized the ability to capture motion; developing systems capable of capturing 12 frames per second, and then later 60 frames per second (Klette and Tee, 2008).

With the formation of the motion picture industry during the early 20th century, motion analysis became a useful tool that was adopted in various ways from scientific to artistic pursuits. Concurrently, in 1927 Nikolai Bernstein, a neurophysiologist, brought a new scientific approach to studying human movement and based his hypotheses in neurophysiology and motor control. Münte et al. (2002) and Furuya and Altenmüller (2013) have also adopted this approach much later. Bernstein et al. also developed a new way of capturing motion, the Kymocyclograph, which has been heralded as “probably unsurpassed until the recent advent of optoelectronic techniques” (Kay et al., 2003). The Kymocyclograph used film and a camera shutter lens at high speed to capture images of light bulbs placed on the moving body (Kay et al., 2003). The Kymocyclograph was capable of capturing up to 600 images per second and presented a series of images of the light bulbs at positions throughout a movement. Bernstein et al. used these images to measure joint movement (Gelf and Latash, 2002). The combination of this new technology, and rigorous methods of analysis, forged a new area: the field of biomechanics.

Bernstein was a proponent of biomechanics and advocated it as a tool for understanding complex interactions of sensorimotor system using robust, accurate and objective measures. One application area investigated by Bernstein and his team was the “biodynamics of piano strike” (Bernstein and Popova, 1930). In one of the earliest examples of studying musical technique using biomechanical methods, Bernstein used his Kymocyclograph to answer the questions: (1) “Do changes in the tempo of a movement influence its construction and dynamics?” and (2) “To what extent does the weight of the extremity contribute to the studied exercise?” The study compared shoulder, elbow and wrist movement of pianists and selected music that constrained the movements of the individual to eliminate any opportunity for “artistic performance.” The results of this early study showed inter-joint coordination increases with increased tempo, where faster tempi show the movements that contribute to key strike become more continuous and fluid, and slow tempi appear segmented. This had interesting implications in piano pedagogy by identifying that coordinated movement sequences are fundamentally different when played at different tempo. This indicates that from a motor control perspective, practicing a piece of music at a slower pace does not help learning because as speed increases, old movement sequences are un-learned and new sequences learned. It is therefore beneficial to practice at the required tempo from an early stage of learning each piece. However, this is not the only perspective and from a pianist’s perspective, note learning and memorization may require slow practice.

This study also showed that at faster tempi, wrist movement seemed “forced” by elbow movement, thus illustrating for the first time that these adjoining segments are highly coupled. This has interesting implications for motor learning and automatic sequence actuation. Bernstein and Popova (1930) also clarified that the weight of the extremity does not contribute to an exercise, as active muscle forces are present during movement illustrating independence.

The following section will provide an overview of the advances in motion analysis and where appropriate, will provide examples of their application within musical performance.

Within biomechanics, arguably the most common and widely accepted methods of motion capture are camera-based optoelectronic systems. As this technology has advanced, so has the ability to study movements in greater detail. Throughout the latter part of the 20th century the majority of motion capture was applied to analysis of gait. Walking could be simplified into a two-dimensional (2D) perspective. Providing information from the sagittal plane allowed study of the primary components of gait, i.e., hip and knee flexion/extension and foot dorsiflexion/plantar flexion. The assumption that gait can be simplified into a 2D perspective obviously affects the accuracy of the measurement. In simplifying to two-dimensions, any third-dimensional rotations are lost; rotations are often affected during injury and impairment and omitting these characteristics limits useful measurement.

However, due to the complex anatomy and vast movement potential within the upper limb (Bernstein’s redundancy), methods of motion capture are required to be comprehensive, three-dimensional (3D) and versatile. Until recently, camera-based motion analysis of the distal upper limb, namely the wrist and hands, were limited (Metcalf, 2008); thus giving rise to alternative methods of motion capture, with inherent benefits and limitations.

Glove-based systems allow analysis of movements that may otherwise be occluded using camera-based technology. However the accuracy of glove-based systems has been shown to vary greatly. Overall average errors of 11°; have been reported when compared to computed tomography (CT; Buffi et al., 2011), which are higher than figures shown for goniometry of the hand; the gold standard for clinical-based joint range of motion measurement (7–9°; Ellis et al., 1997; Ellis and Bruton, 2002). Errors at specific joints varied from 1 to 23°, generally increasing in more distal joints, specifically the distal interphalangeal (DIP) joint (9–23° error), which limits interpretation of any results (Buffi et al., 2011). The technical rationale behind such errors is probably due to the technology that underpins these systems. The resistive flex sensors are placed in the glove on the dorsal aspect of the hand. Movement of the glove material, and therefore the sensors, is independent of the underlying joints, therefore increasing the potential for measurement error. In optical marker-based motion capture, there is always an error due to skin movement. In glove-based systems, this error is compounded due to the additional movement of the glove material. In addition to being less accurate than optical motion capture systems, glove-based systems can, by their very nature, impede movement of the joints, just as when wearing a normal glove.

The type of glove used will also have an impact on the accuracy of measurement. Such as the CyberGlove II, this uses two sensors per finger, therefore omitting movement of each DIP joint. In addition, many glove systems cover the fingertips, thus affecting tactile feedback from the piano keys. An important aspect of motor learning is sensory input to an integrated sensorimotor system. Tactile feedback from keys plays an important role in feedback for learning and memory recall. Goebl and Palmer (2008) showed that tactile feedback from key press enhances timing accuracy through practice. Sensorimotor feedback should not therefore be underestimated. For a comprehensive overview of the anatomy and physiology of the tactile sensory system, see Demain et al. (2013).

Glove-based systems, however, have been used for measurement of hand kinematics during instrumental performance (Furuya et al., 2011a). In this study, the authors used Principal Component Analysis to identify patterns of coordination and synergistic responses between individual movements of the hand. The effect of tempo on finger movements was also investigated in a later analysis of this study (Furuya and Soechting, 2012), and will be discussed later in Section “Movements at Faster Tempi.”

Other methods of motion capture were also trialed in the analysis of pianist movements, namely inertial measurement devices, such as accelerometers. Sensor topology conforming to the underlying anatomy resulted in individual units being applied to each phalanx of the finger (Rahman et al., 2011; Kortier et al., 2012). The large size of these devices provides a distracting and impeding obstruction to “natural” performance, which may contribute to these sensors not being widely adopted. In addition, these devices are tethered to provide power and data transfer, which can also hinder performance. Tethering can also be observed in some optical motion capture systems, such as the CODA system (Rahman et al., 2011).

As technology progressed, so did the incentive to investigate hand and wrist movements. An important consideration when using any measurement system is that you do not change the natural process of what you are trying to measure. This is particularly relevant when assessing the complicated, intricate movements involved in complex hand dexterity. While optical motion capture was still predominantly applied in gait analysis, pointing, reaching and grasping analysis did not begin in earnest until the 1980s (Jones and Lederman, 2006). However, advances in optical camera technology, from approximately 2000, were integral to methods of hand movement analysis being developed and adopted; enabling research on complex hand dexterity. Analysis of complex hand dexterity will further our understanding in areas of motor learning, skill acquisition, intention and virtuoso performance; facilitating an approach to understanding motor control.

These analyses however require a much higher level of precision. This poses the problem of balancing the need for optimal data collection through system set up and any imposed physical restrictions from associated equipment, whilst allowing for functional, natural movements. In order to investigate any kind of movement performance using biomechanical techniques, natural movements must be allowed in order to analyse true performance.

In contrast to other systems, marker-based motion capture systems are highly accurate but also have potential limitations, such as marker placement error and accuracy due to marker occlusion (Metcalf, 2009). This occlusion can be overcome, to some extent, by the use of algorithms that utilize various methods of interpolation.

An alternative to marker-based systems is markerless systems: systems that rely on tracking algorithms and landmark definition to identify the hand and its segmental anatomy. Topical interest in adapting gaming technology to capture and measure body movement is increasing. Researchers are now concentrating on capture and measurement of hand and finger movements (Metcalf et al., 2013). These systems are versatile in their potential application and do not require any additional setup equipment, in contrast to their marker-based and glove-based counterparts, however they are limited by their reduced accuracy.

The following section will review how various current methods have been used to quantify complex hand dexterity during musical performance. Each section will highlight studies and, where necessary, describe results that provide important information for this emerging field.

The following section will outline the adoption of various biomechanical techniques to further our understanding of complex hand dexterity and how they have been adopted within the study of musical performance. A comprehensive approach is required to provide a neurophysiological roadmap of complex hand dexterity. However many studies only focus on particular aspects of this problem, such as particular joint movements, muscle activity, motor cortex activation through movement initiation, etc., probably due to the inherent complexity of the activity and the volume of data involved. To date, there is no comprehensive study that takes into account all these neurophysiological parameters.

Research shows that healthy individuals, given an everyday task, will repeat movements with little kinematic variability (Palmer, 1997; Parlitz et al., 1998; Shan and Visentin, 2003; Metcalf, 2008). In addition, individuals with neurological dysfunction, but with functional ability, aim to minimize variability between repeated hand and wrist movements, compared with lower functioning individuals. However, as a group, impaired individuals adopt different strategies from each other (Metcalf, 2008). These observations were also shown in a healthy group (Ibid.). The ability of individuals to hone particular movements is therefore indicative of higher functioning, and thus indicative of developing expertise; results that have also been shown in coordination and performance in elite sports (Wagner et al., 2012). For the first time, it is possible to test performance strategies, such as minimizing variability, of higher functioning, musical virtuosos, using the most accurate and comprehensive performance metrics. Learning a strategy helps us cope with the redundant DOF in the upper limb (Davids et al., 2007). Bernstein (1967a) stated improved motor skills are synonymous with increasing the number of joints employed; utilizing more joints and less overall muscle activity, an individual becomes increasingly economical at a physiological level. Highly skilled individuals, such as musicians, can exploit the redundancy in DOF of the movements in the upper limb; reversing Bernstein’s “problem.” Redundant DOF not only provide multiple solution strategies to completing a functional task, but can also be utilized by experts to develop and hone complex hand dexterity and refine a musical piece. It can also be used for compensation during learning and adaptation following injury, pain, or impairment.

Hand function can be thought of as a spectrum of acquired skill, initially ranging from novice to expert. Individuals can be placed at either end of the spectrum; either impaired function prior to novice, or virtuoso exceeding the skill of experts. Previous research has shown that expert pianists reduce muscle activity by employing more joints proximally (Furuya and Kinoshita, 2007). It is also widely shown that brain structures and function differ between trained musicians and non-musicians as demonstrated in imaging and EEG activity studies (Rüsseler et al., 2001; Gaser and Schlaug, 2003; Limb, 2006). These factors, as well as those identified above, contribute to skill acquisition in expert musical performance.

Analysis of expertise has been the focus of many studies on musical performance. These studies identified characteristics that highlight key factors associated with, and contributing to, complex hand dexterity. The bulleted list below summarizes the findings of biomechanical studies investigating complex hand dexterity in musical performance. This list specifically summarizes the factors attributable to expert playing, and has been produced for ease of identification and further investigation.

Integrated and highly refined sensorimotor feedback is pivotal to achieving excellence in musical performance. Adopting a bottom-up approach using biomechanical methods, the functional mechanisms that contribute to the effectiveness of the sensorimotor system can be measured. Bernstein and Popova (1930), and later reignited by Münte et al. (2002) and Furuya and Altenmüller (2013), proposed that musical performance provides a unique model for assessment of motor control and learning. Furthermore, dysfunctional mechanisms, such as those following injury or disease, can also be assessed using this paradigm, thus highlighting the importance of skill acquisition as a spectrum of complex hand dexterity. There may also be additional principles that can be applied to the general population, given the high prevalence of work-related musculoskeletal disorders, such as repetitive strain, that are increasing with the use of computer keyboards (Kryger et al., 2003), and other physical interfaces, such as gaming consoles, cellular, and smartphones.

Within elite performers, such as instrumentalists, there is a higher likelihood of injury following the elements of practice and performance that place increased demand on the neuromusculoskeletal system. More commonly, these could give rise to pathophysiological changes and, particularly, playing-related musculoskeletal disorders (PRMD’s).

The exact prevalence of PRMD’s in pianists is unclear; the majority of studies suggesting 40–65%, with an overall range of 26–93% (Bragge et al., 2006). This variability in range however, should be interpreted with caution. The definition of PRMD relating to pianists (and indeed any other instrumentalists) is not clear or consistent in the literature. Added to this, the varying methodological quality of the studies, it is not possible to clearly identify the prevalence of PRMD’s in pianists.

However, some considerable effort has been made to synthesize conflicting results in the literature. Bragge et al. (2006) performed a systematic review of prevalence and risk factors associated with PRMD’s in pianists. From a clinical perspective, Bragge, et al. highlighted that from the 12 studies that met their inclusion criteria, only four provided any operational definition of “injury” to identify cases of PRMD. These ranged from “Hand pain solely from playing-related overuse,” through to “Any problems caused by playing the piano which prevented piano playing for a period of 48 h or longer.” These descriptors serve to maximize the inclusion criteria of their respective studies, but do not provide consistent information to the prevalence of PRMD between each other. It is worth noting that the latter definition of 48 h or longer, provided in a study by Shields and Dockrell (2000), reported 25.8% suffered with symptoms of PRMD during the playing lifetime, or career, of the pianist. In a more recent study, Allsop and Ackland (2010) surveyed 505 pianists in Australia; 42.4% reported PRMD’s at the time when participating in the study (including the duration and location of symptoms), which aligned with the majority of studies reported by Bragge et al. (2006).

The highest ranked paper in the review by Bragge et al. (2006) was a survey and video study of 33 pianists (Yee et al., 2002). While their results indicate that 91% of participants had “a history of musculoskeletal symptoms,” all participants were measured with the clinical outcome SF-36 and all respondents were within normal ranges for that measure. Therefore any influence of PRMD at the time of measurement was retrospective. This study also highlighted the limited applicability of using broad inclusion criteria and the effect this has on skewing results. Thus, the prevalence of PRMD must be interpreted as “within the career” of a professional pianist, rather than at any one time.

The most prevalent pathologies that contribute to PRMD’s have been shown to be (Sakai, 2002):

These pathologies are commonly thought to relate generally to risk factors including excessive overuse, misuse, repetition of movements, and playing conditions (Lippman, 1991; Winspur, 2003; Furuya et al., 2006), which are often indicative of the high levels of practice and performance required of expert musicians.

Sakai (2002) investigated 200 consecutive overuse injuries related to piano performance and found pathologies divided into six main areas: tenosynovitis or tendinitis (56 cases), enthesopathy (49 cases), muscle pain (38 cases), neurological disturbance (28 cases), joint pain (24 cases), and neck or scapular pain (5 cases). Kjelland (2000) suggested that wrist movements during performance can fatigue forearm muscles, and therefore professional musicians can suffer from wrist injury. In the study by Sakai (2002), 35% reported the onset of symptoms, whilst playing an octave or chord, related to positions involving hyper abduction of the thumb and little finger in order to span the keys. However, a pianists’ hand span has not been shown to be a significant factor in symptoms of PRMD (Furuya et al., 2006).

The instrument played may also have an impact on the symptoms and locality of the PRMD. For example, pianists may exhibit symptoms relating almost exclusively to their wrists and hands (focal hand dystonia), whereas, due to required posture, other instrumentalists, such as flautists, are susceptible to neural or vascular impingement between the neck and the axilla (Sheibani-Rad et al., 2013). Violin and viola players may also experience nerve impingement in the neck and shoulder girdle (Berque and Gray, 2002). Level of expertise may also be an important factor in acquiring injury, for example expert pianists utilizing proximal joints to create passive forces at the wrist (Furuya and Kinoshita, 2008a), therefore minimizing the effort required at the distal finger joints.

In summary, common factors associated with impacts on health in performance science are: