Print this page

Repetition without Repetition: Bernsteinian Perspectives on Motor Learning for Musicians

October 1, 2011

Probably few musicians today are familiar with the motor control theories of Ivan Petrovich Pavlov (1849-1936). But when musicians think about motor behavior, their thinking is often implicitly Pavlovian; as we shall see, the influence of those theories persists, even though the theories themselves are largely forgotten. But the theories didn’t simply fade away; rather, they were decisively refuted by the pioneering neurophysiologist Nikolai Aleksandrovitsch Bernstein (1896-1966), who advanced a fundamentally different understanding of human movement. Bernstein never found the recognition he deserved, for his scientific career—and the publication of one of his most important works—were halted in the 1950’s, when the Soviet Union established a particularly rigid form of Pavlovian orthodoxy as “an inherent part of the scientific foundation of the Marxism-Leninism ideology.”1 Within the last twenty years, however, his work has experienced a renaissance, with a number of his works being published for the first time and/or republished in translation; commentaries discussing the relevance of his work to contemporary motor control science have often described his ideas as being twenty to fifty years ahead of their time.2

This paper introduces Bernsteinian motor control theory to the musical community, presenting both a general overview of his thought and practical ideas about how musicians’ practice habits might be informed by his theories.  Bernstein’s work should be of interest to musicians for two reasons. First, it counters a number of widely-held assumptions, and it does so in ways that are intuitively clear, persuasive, and full of practical implications. Second, Bernstein developed a clear, big-picture conception, and, while it is not up to date in its details, it is more broadly and straightforwardly applicable than much contemporary work in human movement science. The motor control literature is not currently a ready source of broad generalizations with clear implications for practical musicianship; this is because the field is currently dealing with a major theoretical conflict, and because many of the complex issues of motor behavior demand tightly-focused studies addressing very small parts of larger questions.3

Bernstein and Pavlov on Motor Control

During much of Bernstein’s career the dominant paradigms in neurophysiology—especially in Russia—were Pavlovian. In the Pavlovian view, motor learning uses repetition to strengthen the neural pathways that carry the neural impulses to the effectors. The Pavlovians believed in a very tight connection between nervous stimulation and action, so that, as summarized by Bernstein, "a central impulse a always produces movement A, while impulse b always produces movement B, which naturally leads to viewing the cortical motor area as a distribution panel with push-buttons.”4 For the Pavlovians, if you want movement A, just send nervous impulse a from the brain.

Many studio teachers think about human movement in strongly Pavlovian terms.  For example, it is not uncommon to hear an appeal for extensive practice grounded in neuroscientific research about the number of repetitions needed to establish a strong neural connection.  Similarly, some teachers, in trying to help students to prepare for orchestral auditions, will encourage them to limit their repertoire only to the pieces on the auditions, so that they constantly practice the same few pieces for years. There seems to be an implicitly Pavlovian conception behind that practice: that immense numbers of repetitions of the same task will make its execution both more precise and more stable.  

For Bernstein, this simple correlation between nervous stimulation and action was untenable; the problem was that it rested on an oversimplified view of what might count as “the same” action.  Even with an apparently simple action like reaching for a glass of water, varying factors both inside and outside of the body could make it necessary to use quite different neural impulses in order to produce the same end result.

To start with the body itself, there can be no simple correlation between nervous stimulation and desired motion, because the results of the nerve impulses will depend on the initial conditions of the body. That is, if body parts are in different states at the start of the motion (e.g. different positions or different states of muscular tension), then the movements that result from identical nervous stimulation will not be the same. At the simplest level, reaching for a glass of water will require different stimuli from the central nervous system depending on where my hand is when I start the motion (e.g. resting on the table versus resting in my lap). More subtly, the contractile force exerted by a muscle depends not only on the degree of nervous stimulation it receives, but also on its length and on the rate of change of its length. That is, the same nervous stimulus sent to the same muscle will not always result in the muscle exerting the same degree of force.  Rather, the degree of force that results will depend upon the muscle’s state at the time at which it receives the stimulus. Suppose we simplify the act of lifting the glass so that it is accomplished purely by the biceps, and imagine two ways in which the glass lifting might occur.  In the first, I simply extend my arm and pick up the glass. In the second, after extending my arm and before lifting the glass, I lean forward, so that my elbow joint closes somewhat, adjusting my posture so that my forearm stays in approximately the same position. With the elbow joint less extended in the second case, the biceps is less stretched and more relaxed. Because of this difference in the state of the muscle, the same degree of nervous stimulation sent to the biceps will result in a different contractile force.

Another factor that can complicate the control of motion is the presence of forces that originate outside the body, forces that vary greatly in their degree of predictability. Whether we consider running with arm weights or doing Tai Chi in a stiff and gusty wind, external forces can necessitate significant changes in the muscle actions and nervous stimuli required to produce a given bodily motion. Example 1 reproduces Figure 2.1 from Bernstein’s paper “The Problem of Interrelations between Coordination and Localization.”5  It vividly depicts the influence of time-varying external forces.  Curve B represents two periods of a repetitive pattern of motion.  Curve A represents the influence of external forces; it lacks the periodic behavior of curve B.  Region C represents the muscular forces needed to produce curve B as the summed result of muscular and external forces. This figure shows that in order to produce an apparently simple, repetitive motion a person may need to use a highly irregular pattern of muscular forces. If I reach for my glass of water in the cabin of a small boat on a rough sea, I will be very lucky if the nerve impulses that brought my hand to the glass on land now have the same result.

Example 1. Figure 2.1 from Bernstein (2001)


Bernstein saw that Pavlovian motor control theory rested on the stability of two correlations: nervous stimulation must correlate with a muscle’s contractile force, and the muscular forces exerted must correlate with the motion of the body.  Bernstein also saw that neither correlation is valid. Nervous stimulation and contractile force correlate only if the muscle starts from the same state, and muscular forces correlate with motions only if the body is acted upon by no external forces. Because few real-world situations involve the same motions made from the same initial positions with no external forces involved, Bernstein rejected the Pavlovian theory.

At this point some musicians—or at least classical musicians—may be wondering if the Pavlovian theory might not still apply to their activities. Jazz musicians face those kinds of task variations when improvising, but classical musicians prepare their performances quite exactly, and they rarely give those performances on ships in stormy seas.

The problem with this line of thinking is that Pavlovian theory describes classical musicians only in the practice room—and then only because they are often so obsessive about creating idealized conditions.

Let us focus on factors that disrupt a correlation between muscular forces and bodily motion. Sometimes these factors come from elsewhere in the body: nervous shakes in the torso may affect the results of standard contractions of finger muscles; if hands are hot and sweaty or cold and clammy, the coefficient of friction between hands and instrument will be altered, so that the customary forces will result in motions that stop too soon or go too far; and of course the variability of a singer’s vocal apparatus is well known. Disruption can also originate outside of the body: a familiar example is the action of a piano—exact reproduction of a sequence of motion from an outstanding performance will be of limited value if the pianist is playing an instrument with a significantly different action.  In each of these examples, making the usual motions with great precision will not be sufficient to assure a successful outcome.

One of the things that characterizes outstanding performers is the ability to adapt to these kinds of varying conditions.  For example, in the late 1980's it was part of string-playing lore that Yitzhak Perlman attributed his superior intonation not to greater accuracy in placing his fingers but to quicker corrections of position after placing them.  I had the opportunity to witness this in person during those years, when Perlman performed the Dvořák Violin Concerto with the Chicago Symphony Orchestra.  In addition to being deeply satisfying musically, his performance was, to my ear at the time, perfectly in tune, including the extended passage in octaves near the end of the concerto.  Before playing an encore, he started to retune his violin, and the instrument was so out of tune that the audience burst into laughter. Clearly his excellent intonation at the end of the concerto did not result from knowing exactly where his fingers should normally go and putting them in the same places that they had gone ten thousand times in the practice room.

Variations in the acoustics of the hall can also create the need for adjustments to the usual motions.  In another example from the same time and place, I frequently heard concerts at a recital hall that suffered from overly live acoustics. On one occasion I heard Cyrus Forough give a violin recital there, and he was the only musician I could remember who did not sound muddy. It seems safe to assume that Forough did not usually have a dry tone and play with short and sharp articulations, so that when put in a wet hall everything was just right—a sort of acoustical Goldilocks effect. Presumably he would have sounded just as muddy as everyone else had he played the same way (made the same motions) as he would have in a more normal hall.

For a variety of reasons, then, musicians—even classical musicians—will often find themselves in the unenviable position of needing to produce the sorts of irregular forces that are represented by curve C in Example 1. Pavlovian motor control theory cannot explain how musicians spontaneously make the needed adjustments; if we want to understand it, we will need a more broadly-applicable theory.

Bernstein and Pavlov on Motor Learning

Bernstein made a second critique of Pavlovian theory that was specifically concerned with motor learning. The Pavlovians’ theories of motor learning flowed directly out of their theories of motor control. If motor control was a matter of sending neural signal a in order to produce action A, motor learning was about the strengthening of neural connections based on repetition of the motion.  Examples of this kind of thinking in the literature on motor control have been found by Reed and Bril dating as recently as 1984, and, as I have suggested at the beginning of this essay, it is still widespread among musicians.6

We have seen that Bernstein attacked Pavlovian motor control theory by demonstrating that producing the same motion under different conditions would require different neural signals.  He attacked Pavlovian motor learning by pointing out that in many tasks, a novice, amidst his many repetitions, executes the motion correctly only rarely. The Pavlovian theory simply cannot account for the commonly observed phenomenon of inept and highly variable novice behavior giving way to consistent expert behavior through a learning process. For neural connections to be strengthened by repetition, the neural commands would have to be unchanging. This in turn suggests the motion being performed correctly from the first. The Pavlovian account deals well with a situation in which the performance of a simple task is made more automatic through practice, but it cannot accommodate situations in which a complex task is mastered through a process of gradual improvement of the form of the motion.

Pavlovian motor learning for musicians?

Bernstein rejected Pavlovian theories of motor learning because they could not accommodate phenomena such as the sudden transitions seen when small children are learning to ride bicycles, transitions from veering wildly back and forth to smooth, skilled riding. To fit Pavlov’s thinking, they would have to have been riding correctly from the start.  But many musicians go to great pains to avoid ever behaving like those wildly-veering children. They believe that practice should always be exactly correct, and that the way to master a task is by doing it extremely slowly, starting slowly enough that the passage can be played correctly and gradually moving to the desired tempo.  This may often be a successful strategy, but not because consistent neural signals are being developed in a Pavlovian manner. Fast and slow motions differ at a biomechanical level, and therefore they differ at the level of neural stimuli: if the task is slowed down enough, it is no longer the same task.  

This was demonstrated by Bernstein and Popowa in an early article that deserves to be better known among music psychologists.7 Over a period of a year they made measurements of the performance of fourteen famous virtuoso pianists using an early way of recording movement known as kymocyclographic recording. The pianists performed two basic tasks, each of which involved playing repeated octaves in notes of equal notated duration and unchanging pitch.  In the first task the pianists were asked to play an accelerando followed by a ritardando, starting Adagio, going to Prestissimo, and returning. They played at a medium dynamic intensity throughout (between mf and f) and reached rates of 500 attacks per minute. In the second task the pianists played at a constant, moderate tempo (200-250 attacks per minute) and played a crescendo followed by a diminuendo, starting pp, going to ff, and returning. The scientists found that in the crescendo test, pianists showed only quantitative changes over the course of the task, changing amplitudes of motion but not the basic biodynamic construction of the movement.  In the accelerando test, however, they found three distinct types of movement composition, corresponding to slow, medium, and fast tempi. When playing the octaves slowly the pianists employed isolated impulses with pauses in between. At a medium tempo these impulses united to form continuous chains, with active impulses coming from both the wrist and elbow and with hand-arm coordination suggesting the dynamics of coupled pendula. At fast tempi the wrist muscles abandoned the active impulses and instead simply produced static isotonic contractions, so that the hand underwent harmonic motion driven by the lower arm.

Bernstein and Popowa thus disproved the theory that slow practice was advantageous on the grounds that it allowed careful self-observation by the pianist.  Because the dynamic construction of the movement is radically altered by the change in tempo, the pianist could observe herself carefully, but not performing the same movements.8

Bernstein’s Theory of Motor Control

Bernstein developed devastating critiques of Pavlovian theories of motor control and motor learning—but scientific reputations rarely rest on critiques. Bernstein’s own theories of motor control and learning cast the issues in a completely different light, and they are, in my view, very illuminating for musical practice.

The interdependence of action and perception

We have seen that successful completion of motor tasks cannot be a matter of sending a static set of well-trained neural stimuli. If you are reaching for a glass of water on a storm-tossed boat or trying to play an out-of-tune violin, that approach will be frustrated. And if doing the usual thing will not lead to a successful result, it follows that you will need to make up something new, and doing this will require input from the senses. Perception—both of states of the body (proprioception) and of the exterior world (exteroception)—is therefore central to the control of motor behavior. When the motor task is under continuous control, it is possible for perception to guide action with real-time reactions to unexpected events. A flight attendant carrying a cup of hot coffee down the aisle of an airplane experiencing turbulence would be an example of this. When the motor task involves a temporal separation between the influence of the person and the desired outcome, adjustments to motor programs will need to be calculated (estimated) in advance. A golf player who uses sensory information in order to estimate the effects of wind is an example of this second kind of perceptually-guided motor correction. The example of Perlman playing the Dvořák concerto combines both of these approaches, as his intonation presumably resulted from predictive corrections based on adjustments from the recent past combined with corrections that zeroed in on the desired pitch after putting his finger down. However a given situation balances predictive and real-time adjustment, consciously-controlled motor behavior relies so strongly on sensory input that perception must be considered to be an integral part of action.

The centrality of goals and intentions

For Bernstein, the necessarily active role of perception brought with it a central role for the goals and intentions of the agent.  Pavlovian psychology had a decidedly behaviorist tendency, in that it tried to understand movement in terms of conditioned reflexes, with the main agents of conditioning being repetition and favorable or sensually-pleasurable outcomes.  In this view the agent responded relatively passively to environmental changes. Bernstein’s work turned this paradigm on its head. Perceptual information is necessary to guide motor behavior, and motion is actively controlled on the basis of this perceptual information. Motion is not merely the result of motor habits or, as Bernstein put it, motor clichés, as the push-button-panel theory would have it. Implicit in the picture of perception being used to guide action is a capacity for making choices in the form of this-and-not-that, and the necessary basis for such choice is a goal, or, to use Bernstein’s phrase, a model of the desired future. The neurophysiology of the human being indicates that she is not primarily someone who reacts passively to the environment but rather someone who envisions possible future events, makes desirable potential events into goals, and formulates motor plans that are able to actualize those goals. Given this central tenet, it is not surprising that many of Bernstein’s intellectual descendants also subscribe to the ecological psychology of James Gibson (1904-1979), being concerned with how the person makes use of the environment and seeing perception as active and purpose-driven.9

This raises a central question for musicians: what form should the model of the desired future take? From what we have already seen we may conclude that it cannot take a spatial form; it is not enough simply to place a finger at a certain point on the fingerboard because the string may be out of tune. Nor is it adequate to define the model in terms of the spatio-temporal dynamics of motion; it is not enough to strike a key with a certain velocity profile because the piano being performed on may have completely different action from the one in the practice room. From a Bernsteinian perspective it is of primary importance to find a better answer to this question.

Repetition in Bernsteinian motor learning

Repetition is almost as prominent in Bernstein’s theory of motor learning as it is in Pavlovian theories, but it functions in very different ways. According to Bernstein, the control of motor behavior is distributed hierarchically among different brain regions. These regions are differentiated by evolutionary age, and each region has as its proper function the control of those motor processes and capabilities which co-emerged with it evolutionarily. The highest, most sophisticated, and evolutionarily most recent brain regions set the motor goals and monitor progress at the highest level, farming out subcomponents of the task to the brain areas that are best able to accomplish them. These lower brain areas generally operate without conscious intervention.

In Bernstein’s somewhat idealized picture of motor learning there are two phases. In the first phase the motor task is repeated with minimal variation and if possible in simplified form, all under the direct, conscious control of the highest motor level. The job of this first stage is to develop an automatism at a lower level that will, in the future, be able to take over the task, or at least significant aspects of the task. These initial attempts are characterized by clumsiness, high variability of outcome, much conscious attention, and an emphasis on the visual domain in the monitoring of sensory feedback. As lower-level automatisms become available for delegation of control, there will be abrupt transitions in the character of the movements: they will become both more graceful and more consistent; they will require much less conscious attention; and proprioceptive feedback will take over the leading role from vision in monitoring progress and guiding corrections. We have already encountered one of Bernstein’s favorite examples of this process, learning to ride a bicycle. In the beginning the rider has difficulty balancing, sways from side to side and falls frequently, and devotes conscious attention to the visual monitoring of balance. Once an automatism is ready to take over control of balance it will appear suddenly, and the formerly tenuous balance will become graceful and effortless. Neither vision nor conscious attention will be required for balance, which will be achieved automatically by a feedback process from the vestibular system to the various relevant muscles.  

Once this transition has taken place, a new phase of motor learning begins. In this phase the emphasis is again on repetition, but in varied circumstances. The goal is to make the automatisms more robust by training them to spontaneously find successful motor solutions to unforeseen changes in internal and external conditions. In a much quoted phrase, Bernstein described the training process as repetition without repetition.10 In the first phase of motor learning this refers to the transformation of motions through practice—though the approximate motions are repeated, their exact form changes. In the second phase of motor learning it refers to the development of the ability to find effective motor solutions for a wide range of task conditions—the basic activity is repeated, but the motions change to suit the circumstances. The concept of repetition without repetition emphasizes that the goal of motor training is not the stable production of nearly identical neural stimuli, nor of nearly identical muscle forces, nor even of nearly identical joint motions. Rather, the goal of motor training is the consistently successful realization of a motor goal under ecologically real conditions—that is, conditions that are subject to unpredictable changes.

Dexterity, the highest achievement of human motor control

This emphasis on the spontaneous generation of successful motor actions dovetails with the centrality for Bernstein of dexterity, the motor capacity that he placed above strength, speed, and endurance. Bernstein defined dexterity as follows:

Dexterity is the ability to find a motor solution for any external situation, that is adequate to solve any emerging motor problem

correctly (i.e. adequately and accurately)

quickly (with respect to both decision making and achieving a correct result)

rationally (i.e., expediently and economically), and

resourcefully (i.e., quick-wittedly and initiatively [sic]).11

This definition appears near the end of On Dexterity and its Development, and it summarizes several central aspects of his approach: the active agency of the person moving, the crucial role of goals and intentions, and the variability both of task conditions and of the movements needed to achieve the goals. 

Bernsteinian Motor Learning for Musicians

We have seen that many musicians have implicitly-Pavlovian approaches to motor learning. They emphasize many exact repetitions of each task.  They strive to eliminate variations from their practice conditions, for example always trying to practice on the same piano or retuning their instruments frequently.  When they encounter a difficult task, they slow it down to the point at which it can be accomplished successfully, trying to strengthen desired neural connections by always playing correctly.

Much of this is often beneficial; certainly no complex task can be mastered apart from extensive, repetitious practice. Nonetheless, this picture of practice habits resembles that of a basketball player who practices only free-throw shots. This would be an athlete who faces unpredictable real-world situations, in which completely novel physical challenges must be overcome through real-time improvisation, and who prepares by practicing a task that is always exactly the same; rather than practicing shots from every conceivable posture and position on the court, she always stands at the free throw line. This is not a fair comparison; the imaginary basketball player is grossly misguided in a way that the equivalent musician is not. But it does underscore the point that some musicians do little to cultivate the kind of dexterity Bernstein described, even avoiding situations in which it would be needed.  How can Bernsteinian dexterity be cultivated?

The body: Pavlovian or Bernsteinian?

Before beginning to practice, a musician who wants to develop Bernsteinian dexterity should think about how he relates to his body. In the Pavlovian view, the musician’s body is treated like one of Pavlov’s dogs. It is an unintelligent servant; it cannot be trusted to do anything well on its own; and in order to perform successfully it must be given a very detailed set of instructions, with those instructions drilled insistently. All of the intelligence in the system resides at the level of the conscious planning and evaluation of practice.

The Bernsteinian view is completely different. The body is a mysterious partner with vast capacities. It can and must be trusted, because it will encounter situations that conscious intelligence can neither predict nor prepare it for exactly. Directive, conscious intelligence acts like the trainer of a champion, creating opportunities for the body to expand its abilities and reach its full potential. This is not to say that all musicians who focus on exact repetition take a low view of their own bodies; but a low view of the body seems to flow logically out of the Pavlovian approach, so that embracing the Pavlovian perspective could very easily lead to or reinforce just such a low view. Like all people, musicians have to decide whether they are fundamentally empty, in need of good things being added from the outside, or fundamentally full, needing to have the good things within them brought out and developed. Musicians who embrace the latter view will find significant resonance with the Bernsteinian understanding of motor learning. And they will probably want to take the next step, which is thinking about the form of the model of the desired future. 

The desired future: motion or sound?

After Cyrus Forough’s recital I asked him how he had overcome the acoustics, and he told me that that was the purpose of the dress rehearsal—to listen to the sound in the hall and to make whatever adjustments were needed to make the sound right. An emphasis on sound is also at the heart of the story about Perlman, with its opposition between putting the fingers in the usual places and playing in tune. In fact, all of the examples of variable task conditions given above focus on the end goal of producing a certain sound.

If a musician wants to shift from an implicitly Pavlovian approach to a Bernsteinian one, the first step is to make sound the primary locus of attention. The Pavlovian musician essentially says, “As a musician I move in certain ways, and these motions produce musical sounds.” The Bernsteinian musician says, “As a musician I make sound, and my body knows or can figure out how to move so as to produce that sound.” Of course no musician would ever make the first statement; nonetheless, the practice habits of many musicians would seem to reflect just that implicit belief. All musicians have sound as their ultimate goal, but some have arrived at an almost absolute focus on the exact repetition of motion as a sufficient means to that end. From a Bernsteinian perspective this is misguided.  Given the many situational variants that can arise, it is important also to develop the ability to adapt spontaneously to changing circumstances.

A conceptual shift from an implicitly Pavlovian understanding of motor learning to a Bernsteinian one is primarily a matter of putting the model of the desired future in terms of sound, recognizing that it is inadequate to strive for a perfectly consistent set of movements (because a variety of real-world conditions can prevent these movements from producing the desired sounds). In order for this kind of conceptual shift to have a pervasive impact on performance it must lead to an expanded repertoire of practice techniques. For this we must return to Bernstein’s understanding of motor learning as a two-stage process and to repetition without repetition. The task of the first phase is the construction of an automatism—the musician must move away from a need for conscious attention and visual monitoring, and shift control down to lower level processes. The second phase must put that automatism through as many different situations as possible in order to strengthen it and develop its capacity for dexterity in Bernstein’s sense. This two-phase process of motor learning provides the conceptual framework for the exercises that I now describe.

Developing an Automatism

As might be expected, the second phase of Bernstein’s motor learning process will receive more attention than the first. But there are some ways in which the development of an automatism—the first phase—can be informed by Bernsteinian theory. It is this phase that receives the most attention from usual approaches to musical skill development. For example, approaches that focus on the development of “muscle memory” are concerned with developing automatisms. This also casts new light on the results of Bernstein’s experiments with Popowa, discussed above. For while slow playing may not allow a musician to observe the same motions that she makes in actual performance, the slower pace may indeed facilitate the visual observation and conscious, high-level control that characterizes the first phase of motor learning.

A more sympathetic view of slow practice softens but does not remove the force of the original critique; it remains the case that the dynamics of motion at a much slower tempo will often be quite different from those required in an actual performance. Slow practice is an attempt to implement the right idea—that in order to facilitate learning, something must be done to simplify the task. But as a drastic change of tempo often leads to significant changes in the nature of the task, it makes sense to look for other options.

One possible approach is to divide the passage into manageable segments. Rather than slowing down the passage and gradually speeding it up, split it up into very small pieces and then gradually reassemble the pieces. This is exactly the point of an exercise from the Dounis school of string playing, my knowledge of which comes from eight years of study with George Neikrug, one of D. C. Dounis’ foremost disciples. The exercise breaks passages into chunks of varying size, rotating through various possibilities for the points of segmentation. This allows each possible succession of notes to be practiced independently without slowing down the tempo. If a violinist were learning the running eighth notes near the beginning of Smetana’s Overture to The Bartered Bride (Example 2), using this exercise she would first break the passage into groups of two, playing the dotted rhythms as in Example 3a, with strong accents on each quarter beat.  As notated in Example 3 the tempo is about two times slower than in the original passage; thus the sixteenth notes in Example 3 are comparable with eighth notes in the original. The rhythmic notation in the exercise is only approximate; the intention is to snap the rhythm quickly and vigorously, training the fingers to produce brief sequences of motion as unified wholes.

Example 2. Smetana, Overture to The Bartered Bride, Violin 2, mm. 14-17


Example 3a. A Dounis approach to practicing the Smetana



So far this is a very standard exercise; the key to the Dounis version is lengthening the segments. After the violinist became comfortable with Example 3a, she would then practice groups of four, elongating each eighth note in rotation as in Example 3b. Again, the notated rhythm is approximate, the purpose being to toss the groups off as units. Finally, she would practice groups of eight, with the lengthening rotating through a whole measure (Example 3c). This exercise develops the note successions, while working toward a performance in which the music is felt to move in one.12

Example 3b. A Dounis approach to practicing the Smetana



Example 3c. A Dounis approach to practicing the Smetana


The chemist-turned-philosopher Michael Polanyi has argued that all learning is a process of discovering for oneself, and that as such it always involves mysterious leaps across gaps from lack of understanding to understanding, ‘heuristic gaps’ in his terminology.13 This view resonates strongly with Bernstein’s observation of the sudden change in the way a task is performed when an automatism is formed. Polanyi believed that heuristic leaps could be narrowed but not eliminated. Similarly, Bernstein saw the formation of an automatism as something that goes on subconsciously, as a background process; he never suggested that anyone had the ability to build an automatism consciously, putting the pieces together one by one. From this perspective, the Dounis segmentation is similar to the usual approach of slowing down the tempo. Each, in different ways, narrows the heuristic gap that must be crossed. This is why slow playing is often effective. But neither can eliminate the gap; neither can replace the mysterious process of developing an automatism. Some approaches may well be more helpful than others, but this seems one area in which more is better—the more ways the gap can be narrowed, the more chances for the spark to fly across.

Repetition without Repetition

The ways in which existing automatisms can be strengthened fall into two categories; first I will discuss exercises pertaining to brain-to-body communication at the level of conscious intentions, and then exercises that introduce physical challenges, challenges that are not addressed primarily through conscious thought and problem-solving.  Some exercises are novel and others put familiar tasks in a new light.


Sight-reading is generally valued as an important skill, but it is often seen as a discrete one. To say that someone is not a good sight-reader is not, in our usual understanding, to criticize that person’s general abilities as a musician. But sight-reading highlights the distinction between rigid, Pavlovian motor skills, in which novel motion sequences require sustained practice, and adaptable, Bernsteinian motor skills that can generate new patterns of motion on-line.  Sight-reading has a broader value because it provides an opportunity to practice responding quickly to unforeseen motor demands, to develop dexterity in Bernstein’s sense.

Imitation by ear

The distinction between motion and sound as the main locus of attention raises questions about what notation means to a musician. We are familiar with the idea that notation encodes sound, but for most musicians notation will also encode movement; this means that some musicians may well read scores as proto-tablatures.14 If so, the sight-reading task may accomplish its goals only partially; while it certainly requires the assembly of novel sequences of motion in real time, it is possible that the sight-reader may think more in terms of sequences of motion than of sequences of sounds. (There is a straightforward test for this; as he is sight-reading, does the musician discover what the music sounds like by hearing himself play it, or can he auralize it in advance?)

If the effectiveness of the sight-reading task for the development of Bernsteinian dexterity is compromised because the musician is sight-reading not sounds but actions, then the obvious solution is to remove the visual stimulus and have the musician play from an auditory cue, imitating what is heard. There are a variety of ways to do this. Musicians may play in small groups, taking turns in a drill in which a leader improvises a short passage and the followers repeat it. Similar practice could be done with a recording, either playing back short excerpts or else imitating in real time. For the latter task, unfamiliar popular music would serve well, as the various sections tend to repeat, making playing along in real-time more feasible. For more challenging motion sequences, playing along with recordings of classical pieces that are very familiar by ear but which have not been studied could also be helpful. In addition to developing dexterity, these exercises would be valuable as ear training, and, especially when playing along with recordings, could foster improvement in intonation. Imitation by ear provides a helpful complement to sight-reading; most musicians will be able to sight-read more complex passages than they can imitate by ear, but imitation by ear targets the connection between imagined sound and the production of motion. Both activities strengthen the interconnections between action and perception.

Memorization of singing

The exercises above support the general development of dexterity; now we turn to exercises that develop dexterity in the performance of specific pieces. One of the most direct ways to interrogate—and to influence—the primary locus of attention involves memorization: is the piece memorized as a sequence of sounds, or as a sequence of motions? Musicians often do the latter, especially when they memorize spontaneously through practice with a score. This is revealed when they are asked to sing the music: if they cannot sing without making the motions of playing, then it is not sound that has been memorized but motion. (Obviously, it is not clear that this test could be meaningfully implemented for singers). This is, of course, the equivalent of the test for whether sight-reading was based mainly on motion or on sound; if musicians memorize motion, they discover what the music sounds like by making their instruments play it for them. If memorization has indeed reinforced motion at the expense of sound, there is an obvious solution, at least for instrumentalists: memorize singing the piece.15 From personal experience, I have found that a number of benefits come from memorizing singing. The freedom of not thinking about the instrument helps in exploring interpretive possibilities, and once choices have been made, it also helps the process of focusing expressive shapes and musical gestures, making their features more sharply profiled. And in post-tonal music it certainly helps with intonation. More than any other exercise, memorizing through singing helps to establish sound as the primary locus of attention.16

Practicing variants

It is a standard practice technique to introduce variants—changing articulations, introducing uneven rhythms, rapidly alternating dynamic level, etc.—and the usual goal is a stronger motor program that can handle changes of various sorts. This is already very much in line with the exercises described here, and the only contribution of a Bernsteinian perspective would be an emphasis on practicing these variants spontaneously, rather than using normal practice methods in what amounts to learning a different version of the piece. The latter approach is no doubt helpful, but it targets Bernsteinian dexterity less precisely than deciding on the fly what kind of variation to play and seeing what happens.

This kind of practice may require a special kind of mental or even emotional discipline. Some musicians do all they can never to sound bad; fearing that they may be training themselves to play poorly, they insist on simplifying the task at hand to the point at which it can be done well and building from there. This is probably rooted in rather Pavlovian thinking comparable to Suzuki’s insistence that practice makes permanent. Certainly it is counterproductive to practice poorly or inattentively, but tolerating inaccuracy while attempting something just beyond one’s current skill level is a different matter. Just as no one learns to ride a bicycle without spending time riding poorly, the exercise described here may well lead to some time spent playing less well than would otherwise be desired. For some musicians this will feel very uncomfortable, and they will want to stop and practice the variations more systematically, as they would the piece itself. This kind of practice may well be beneficial, but it will be differently beneficial than pressing on with the attempt to modify a motor plan on the fly. Are we willing to let ourselves fail for a while, trusting that our bodies may actually be in the middle of a learning process whose results aren’t visible yet?

Playing familiar pieces in other keys

An exercise that both strengthens sound as the form of the model of the desired future and creates novel motor demands is playing a memorized piece in a new key.  Such a piece must be noticeably below the musician’s current maximum level of difficulty (substantially below for keyboard players). This capitalizes on existing knowledge of sound sequences in a way that frustrates existing knowledge of motion sequences, requiring the on-line construction of novel motor plans.


For jazz and popular musicians, improvisation is practiced extremely frequently. For classical musicians, though, suggesting improvising may seem pointless: this is often taken as one of the defining contrasts between classical and jazz musicians, and many classical musicians feel paralyzed when asked to improvise. But improvisation was an important skill when much of the core repertory was being composed, and this suggests that it could still be a helpful skill for classical musicians. To make this less intimidating, it may be helpful to focus on improvising exercises and etudes rather than improvising “real music.” In his treatise on violin playing, Joseph Szigeti writes of the importance of being able to “improvise one of the innumerable variants that every violinist should be able to invent for himself when he encounters some difficulty in one of the masterpieces, old or contemporary.”17 The basic idea is that, for a difficult passage, instead of practicing exactly and only that passage, it can be extremely helpful to practice a number of passages that place similar kinds of challenges in different contexts (some of David Popper’s etudes in his High School of Cello Playing, op. 73, are particularly clear examples of this). The desired flexibility reflects a conception of dexterity very much in accord with Bernstein’s, especially if some of the exercises are extemporized.18

Introducing novel physical challenges

The second category of exercise involves introducing novel kinds of physical challenges. I encountered this kind of exercise in my study with George Neikrug, who would have me hold the bow at the tip instead of the frog or lean my viola against a wall and play without allowing the thumb of my left hand to make contact with the instrument. These exercises had very specific purposes, but they also served to enhance dexterity by changing aspects of the task’s motor demands. With these exercises, and especially with the ones that go beyond the sorts of things routinely experienced in performance situations, musicians should carefully assess the potential for injury and err on the side of caution in choosing pieces for the exercises.

Recreating standard challenges of performance

The most straightforward exercises that introduce physical challenges simply recreate conditions that are commonly found in performance. For example, pianists often have to play on unfamiliar instruments; although they often attempt to practice on the same piano, they can seek out different instruments, and especially ones that are not in optimal condition. Similarly, many instruments go out of tune during performance, but some musicians are always retuning during practice. Instead, they can sometimes deliberately put their instruments out of tune, changing the tuning whenever they have adapted to the current one. For a particular challenge, they can ask a friend to retune the instrument and then, without checking the intonation, discover through playing repertoire what adjustments are needed. In addition to this exercise, reed players could deliberately play with poor or damaged reeds.

It is also possible to create novel challenges by switching instruments. Beyond the case of pianists, musicians whose instruments are portable, who do not usually need to play unfamiliar instruments, could trade for a day with a friend or colleague, forcing their bodies to figure out how to adapt. And as suggested by the story about Cyrus Forough, a challenging new acoustic environment can necessitate on-the-fly adjustments as well.

More unusual kinds of physical challenges

There are, of course, many other ways in which novel physical challenges could be created. One could try to simulate the effects of nervousness as a way of preparing for the challenges of performance, for example imitating sweaty or clammy hands by putting something sticky or greasy on the fingers to change the character of the interface between body and instrument (clearly this would need to be done with an instrument that could acceptably sustain some damage). Strapping a palm sander to the upper arm to simulate nervous shakes would be a different path to the same goal. I offer these ideas not to suggest that anyone actually try them but rather in the spirit of a brainstorming session, to give a sense of the scope of the possibilities. Extreme though they may be, they offer examples of just how far Bernsteinian conceptions of motor learning for musicians could move from more standard, implicitly Pavlovian ones. These new ideas flow from making sound the primary locus of attention and emphasizing forms of practice that involve a different kind of dexterity and a different kind of repetition—repetition without repetition.


Abernethy, Bruce and W. A. Sparrow. “The Rise and Fall of Dominant Paradigms in Motor Behaviour Research.” In Approaches to the Study of Motor Control and Learning, edited by Jeffery J. Summers, 3‒45. Amsterdam: North-Holland, 1992.

Altenmüller, Eckart, Mario Wiesendanger, and Jürg Kesselring, eds. Music, Motor Control and the Brain. Oxford: Oxford University Press, 2006.

Bernstein, Nikolai Aleksandrovitsch. “On Dexterity and its Development.” In Dexterity and its Development, edited by Mark L. Latash and Michael T. Turvey, 1‒244. Mahwah, NJ: Lawrence Erlbaum Associates, 1996.

–––. “The Problem of Interrelations between Coordination and Localization.” In Classics in Movement Science, edited by Mark L. Latash and Vladimir M. Zatsiorsky, 64‒84. Champaign, IL: Human Kinetics, 2001. Originally published 1935.

Bernstein, Nikolai Aleksandrovitsch and Tatiana Popowa. “Untersuchung über die Biodynamik des Klavieranschlags.” Arbeitsphysiologie 1 (1929): 396‒432.

Feigenberg, I. M. and L. P. Latash. “N. A. Bernstein: The Reformer of Neuroscience.” In Dexterity and its Development, edited by Mark L. Latash and Michael T. Turvey, 247‒275. Mahwah, NJ: Lawrence Erlbaum Associates, 1996.

Gibson, James J. The Ecological Approach to Visual Perception. Boston: Houghton Mifflin, 1979.

Ito, John Paul. “Impulse Structure in Tonal Music: A Theory of the Metrical Coordination of Motor Behavior in Performers.” Ph.D. diss., Columbia University, 2004.

Latash, Mark L. and Michael T. Turvey, eds. Dexterity and its Development. Mahwah, NJ: Lawrence Erlbaum Associates, 1996.

Meijer, Onno G. and Klaus Roth, eds. Complex Movement Behaviour: ‘The’ Motor-Action Controversy. Amsterdam: North-Holland, 1988.

Palmer, Caroline. “The Nature of Memory for Music Performance Skills.” In Music, Motor Control and the Brain, edited by Eckart Altenmüller, Mario Wiesendanger, and Jürg Kesselring, 39‒53. Oxford: Oxford University Press, 2006.

Polanyi, Michael. Personal Knowledge: Towards a Post-Critical Philosophy. Chicago: University of Chicago Press, 1962.

Reed, Edward S. and Blandine Bril. “The Primacy of Action in Development.” In Dexterity and its Development, edited by Mark L. Latash and Michael T. Turvey, 431‒451. Mahwah, NJ: Lawrence Erlbaum Associates, 1996.

Summers, Jeffery J. and Anson, J. Greg. “Current Status of the Motor Program: Revisited.” Human Movement Science 28 (2009): 566‒77.

Szigeti, Joseph. Szigeti on the Violin. New York: Frederick A. Praeger, 1970.

Thompson, Sam and Andreas C. Lehmann. “Strategies for Sight-Reading and Improvising Music.” In Musical Excellence: Strategies and Techniques to Enhance Performance, edited by Aaron Williamon, 143‒159. Oxford: Oxford University Press, 2004.


1Feigenberg and Latash, "Bernstein," 260.

2My discussion of Bernstein's work draws primarily on three sources: the English translation of Bernstein's book On Dexterity and its Development, first published in Russian in 1991 (the original publication having been halted for political reasons in 1950); the abridged, edited, and translated version of Bernstein's paper "The Problem of Interrelations Between Coordination and Localization," included in the collection Classics of Movement Science and originally published in 1935; and the commentaries on the relationships between On Dexterity and its Development and contemporary human movement science that are published together with that book in a volume edited by Latash andTurvey.

3The two competing approaches are the motor-programming approach and the dynamic-systems approach. I offer a non-technical overview of the conflict in "Impulse Structure," 150-157. Other helpful sources include Meijer and Roth, Complex Movement Behaviour, Abernethy and Sparrow, "Dominant Paradigms," and Summers and Anson, "Current Status." For a good survey of music and motor control that achieves about as much general applicability as is currently possible, see Altenmüller, Wiesendanger, and Kesselring, eds., Music, Motor Control and the Brain.

4Bernstein, "The Problem of Interrelations," 67.

5Ibid., 68.

6Reed and Bril, "Primacy of Action," 435.

7Bernstein and Popowa, "Untersuchung über die Biodynamik des Klavieranschlags."

8As it happens, Bernstein and Popowa also disproved a second theory, that of playing with the weight of the arms alone—that is, with no active downward muscular forces applied. They demonstrated that gravity did not provide sufficient acceleration for playing at medium and fast tempi, and they observed that at slow tempi playing with weight alone occurred significantly less frequently than some pianists claimed.

9Gibson, Ecological Approach.

10Bernstein, On Dexterity and its Development, 204.

11Ibid., 228.

12For a comprehensive theory of just what it means for music to move "in one," see Ito, "Impulse Structure."

13Polanyi, Personal Knowledge, chapters 5-6, esp. 123, 142‒43.

14The most commonly found tablatures today are probably the guitar tabs circulated among guitar players who don't read lead sheets; each chord is represented by a schematic picture of a fretboard with circles to indicate finger placement.

15For keyboard players this attempt to divorce sonic representation from physical representation will probably be only partially successful. At least for complex pieces, few keyboard players will be able to auralize the entire texture precisely, and few will be entirely able to remember the sound of inner voices apart from remembering the motions that produce them.

16For a discussion of empirical research that investigates the differences between cognitive representations of performance and motor representations, see Palmer, "Nature of Memory."

17Szigeti, Szigeti on the Violin, 7. I am indebted to Daniel Barolsky for bringing this passage to my attention.

18For a further discussion of improvisation aimed partly at classical musicians, one that connects it with sight reading, see Thompson and Lehmann, "Strategies for Sight-Reading and Improvising Music."

31845 Last modified on March 6, 2019