Motor coordination

• Jörn Diedrichsen, Institute of Cognitive Neuroscience

Coordination between two or more effectors (muscles, joints, limbs, or even different people) occurs when the motor commands to one effector depend (in a causal or statistical sense) on the state of the other effector(s). Coordination is goal-directed; the interdependency of movements promotes the achievement of a behavioral task.

Introduction

Coordination can be described in terms of the behavioral goals, and in terms of the underlying mechanisms. For goal achievement, coordination may be necessary because two effectors are mechanically linked (such as the upper and lower arm). In this case motor commands to one effector must actively compensate for torques induced by movements of the other. In other cases the effectors do not interact mechanically, but coordination may be necessary because multiple effectors act on the same task-relevant outcome variable (such as producing a tune with two hands when playing the piano).

Coordination arises through two mechanisms: feed-forward and feedback control. Using feed-forward control, the interdependence of the effectors is preplanned and is visible before sensory feedback arising from the movement can be utilized. Under feedback control, coordination arises during the correction of deviations from the intended movement during movement execution.

Feedback control

Feedback control enables the motor system to react to sensory input that indicates deviations form the planned movement. In contrast to local feedback control, the defining feature of coordinative feedback control is that motor commands to one effector depend on sensory feedback from another effector. Such dependency can, for example, be observed in the reaction of the shoulder in responses to perturbations of the elbow, or in the reaction of one hand in response to perturbations of the other. Coordinative feedback responses can be observed after very brief latencies (<70 ms) and are likely the result of motor cortical control circuits (Pruszynski et al. 2011).

Feed-forward control

Coordination is also achieved in a predictive, feed-forward manner. For example, to make a successful reaching movement, the muscular activity around the shoulder joint needs to be tightly coordinated with the muscular activity around the elbow joint to compensate for the interaction torques and to ensure a straight reaching trajectory. The term “synergy” is often introduced to explain coordination across different muscles. As a descriptive term, a synergy simply refers to the strong regularities in the spatiotemporal pattern of muscle commands, and the observation that large portions of the variance of recorded muscle activity can be described by a small number of linear components (d'Avella et al. 2006). As an explanatory term, a synergy refers to a neural controller that produces the correlated pattern of muscle activity. In the framework of Optimal Feedback Control, coordination in both feed-forward and feedback control is achieved by making the control policy of one effector dependent on an internal estimate of the state of another effector (Todorov et al. 2002, Diedrichsen et al. 2010). The difference between feed-forward or feedback control within this framework is gradual, and simply reflects the fact that the state estimate is informed by an internal prediction in the former, and actual sensory information in the latter case.

Redundancy, uncontrolled manifold, and minimal intervention principle

Figure 1: Example of redundancy, in which different combinations of two effectors achieve the behavioral goal. The observed motor commands (red ellipse) lie close to this solution space or uncontrolled manifold (line) . Feedback mechanisms that only correct task-relevant error (blue errors) can account for this pattern of variability, as variation along the uncontrolled manifold is allowed to accumulate over the movement (minimal intervention principle).

Coordination is tightly related to the concept of redundancy. Redundancy refers to the fact that the motor system usually has more degrees of freedom available than strictly necessary for the achievement of a task. For example, different combinations of elbow and wrist rotation may lead to the same desired position for the endpoint of a hand-held instrument (Vetter et al. 2002). Therefore, multiple combinations of movements from each of a set of effectors will ultimately lead to the same task outcome (Figure 1). Coordination is historically viewed as the solution to the “problem of redundancy” (Bernstein 1967), meaning the problem of choosing a movement from the set of redundant possibilities.

A typical characteristic of biological motor control is that movement variability along this task-irrelevant dimension is much higher than along task-relevant dimensions (red ellipse, Figure 1). The subspace formed by these task-irrelevant dimensions is therefore often referred to as the uncontrolled manifold (Scholz et al. 1999). Increased variability along the manifold can be explained by feedback mechanisms that follow the minimal intervention principle (Todorov et al. 2002), i.e. that only correct for task-relevant errors and do not reduce variability along the uncontrolled manifold, thereby reducing effort and signal-dependent noise. To achieve this task dependence, the feedback mechanism requires knowledge about the state of both effectors, and about how the movements of each effector influence the outcome variable or goal of the task. In this view, the system actively exploits the redundancy of the system to improve performance (Latash, 2012).

Examples of coordination

Muscle coordination for index finger movements

Seven muscles, located around its base, pull the index finger in different directions. This muscular system is (at least for low forces) redundant. That is, multiple combinations of muscle activations can produce the same desired force at the tip of the index finger. The variability of activity during force production is not independent across muscles, thereby indicating that these muscles are coordinated. Variability is highest along the uncontrolled manifold, i.e., the subspace of muscle activations that does not change the force at the tip of the index finger. While this structured variability fulfills the criterion of synergy, the results can be explained by coordinative feedback corrections and the minimal intervention principle (Valero-Cuevas et al. 2009).

Temporal coordination of muscle commands

To achieve smooth and coordinated arm movements, the activity of all muscles must be precisely coordinated in time. For single-joint movements, a tri-phasic pattern of muscle activity is observed: After an initial burst of the agonist, the antagonist muscle breaks the movement, followed by an additional burst of the agonist that stabilizes the limb (Hallett et al., 1975). For multi-joint movements in different directions, the onset of shoulder and elbow muscles are finely coordinated in time (Karst and Hasan, 1991). Whether these temporal regularities are an emergent property of a coordination mechanisms, in which the command to each muscle depends on the estimated state of the movement (Diedrichsen et al., 2007), or whether they are caused by an internal timing mechanism, is unclear.

Multi-joint coordination of elbow and shoulder

Figure 2: A positional perturbation to the elbow (black arrow) leads to a local feedback response (elbow torque, red arrow) and a coordinative feedback response in the shoulder joint (shoulder torque, blue arrow). Adapted from (Pruszynski et al. 2011).

The two-joint arm (elbow, shoulder) provides an example, in which coordination is necessary because movement in one effector will result in torques acting on the other. In an elegant experiment, Pruszynski and colleagues (Pruszynski et al. 2011) provide evidence that fast feedback mechanisms actively compensate for this mechanical coupling. In one of their experiments, they delivered a positional perturbation to the elbow joint (Figure 2, dashed arrow). A local feedback mechanism would react with a torque generated at the elbow joint only (red arrow). However, due to the mechanical coupling in the joint, this would lead to movement of the shoulder. Therefore, to stabilize the shoulder, participants also produce a compensatory shoulder torque (blue arrow) at time latencies as short as 70 ms. Using single-unit recording in primary motor cortex, the authors also found that neurons whose firing rate correlates well with the shoulder torque under static conditions exhibit a response to elbow perturbations at a latency of 50 ms. These motor cortical neurons therefore provide the neuronal manifestation of fast coordinative feedback responses. Thus, the responses of one effector depend on sensory input signaling the perturbation of another effector.

Bimanual coordination

The ability to use both hands simultaneously to manipulate objects is one of the core examples of dexterous motor behaviors in humans. Studies of human bimanual coordination fall broadly into two perspectives. The first one focuses on the phenomenon that movements of the two upper limbs spontaneously synchronize in time and space. The second perspective attempts to illuminate how the two hands achieve coordination to accomplish specific tasks, in which a breaking of symmetry and synchrony is often required.

Coordination dynamics: the tendency to move in synchrony and symmetry

One of the persistent observations about movements of the two upper limbs is that – unless required by the task – humans tend to initiate movements of the upper limbs in close temporal proximity, often with onset asynchronies of less than 20 ms (Kelso et al. 1979). The tendency to synchronize can also be observed in cyclic movements, such as the repetitive abduction – adduction of the index fingers of the two hands. Under such conditions, the human motor system has a relatively persistent preference for patterns of mirror symmetric movements (Kelso 1984). At higher speeds, the less preferred coordination patterns become unstable and the system spontaneously falls into the mirror-symmetric mode. These phenomena have been successfully explained with a dynamical systems approach to coordination, in which the interaction between the hands is modeled as two coupled oscillators (Haken et al. 1985, Schoener et al. 1986).

Neural studies of bimanual coordination dynamics indicate that this coupling arises at multiple levels of the nervous system. The preference for mirror symmetry appears to occur largely in visual reference frame and is strongly influenced by visual feedback (Mechsner et al. 2001). This indicates that bimanual coupling of spatial aspects occurs not necessarily between homologous muscles, but between the spatial plans for the two hands. Furthermore, spatial coupling appears to have a cortical origin, as the preference for bimanual symmetric movements is severely diminished after surgical removal of the corpus callosum (Kennerley et al. 2002). In contrast, the temporal coupling of discrete movements is largely preserved after callosotomy, suggesting a subcortical origin (Diedrichsen et al. 2010).

Bimanual load lifting task and asymmetric role distribution

Many every day activities are inherently asymmetric, with one hand (often the dominant hand) taking an active and the other taking a more supportive role. Humans often use the non-dominant hand to hold an object and then manipulate it with the dominant hand, for example when typing on a smart phone. In this situation, the “passive” hand predicts the mechanical perturbation induced by the active hand, compensates for the perturbation, and stabilizes the object. This class of behaviors has been studied using the bimanual load-lifting task, in which one hand holds an object, which is then lifted with the other hand. In anticipation of the unloading, the activity of the load-bearing arm muscles decreases predictively, ~50 ms before the weight on the arm starts to decrease (Massion 1984). As in other anticipatory postural adjustments, the anticipatory activity only occurs if the load is lifted by a self-initiated movement, but not when it is lifted by an external force, even if the time of unloading is perfectly predictable (Diedrichsen et al. 2003).

Even for tasks that appear symmetric, the motor system assigns a leading role to one hand and a compensatory role to the other. For example, when uncapping a pen, one hand holds the pen, the other the cap, and the two hands then exert simultaneous and opposing forces to remove the cap. Despite this symmetry, one hand assumes a leading role, an assignment that can be flexibly switched between hands. In a recent study (Johansson et al. 2006), participants manipulated an object bimanually. To move a cursor leftwards on a computer screen, one hand had to exert a leftward force, the other a rightward force onto and object. Because the object was held freely and had to be kept stable, the task required two equal and opposing forces. However, the hand for which the force direction was congruent with the cursor movement assumed a leading role, which was measureable as an earlier and larger EMG signal from that hand’s muscles, and by an increased BOLD signal from the contralateral motor cortex.

Bimanual reaching and feedback control

Figure 3: Task-dependent bimanual feedback control. In the two-cursor task the two hands move separate cursors to two targets. In the one-cursor task, they move a single cursor, displayed at the spatial midpoint between the hands, to a common target. The unperturbed movement (black solid line) is identical in the two situations. When applying a mechanical perturbation to the left hand (dashed line), participants react with a local feedback response by the left hand in the two-cursor task (red arrow). However, in the one-cursor task, a coordinative feedback response can be observed both from the left and right hands (red and blue arrow). Adapted from (Diedrichsen 2007).

Bimanual coordination also involves fast and highly task-dependent feedback control mechanisms. Changes in bimanual feedback mechanisms can be observed when comparing the bimanual manipulation of a common object to the manipulation of two individual objects (Dimitriou et al. 2012). This phenomenon can even be observed when the coupling between the hands is virtual rather than mechanical, as for example, when the two hands jointly control a single cursor displayed at their spatial midpoint (Diedrichsen 2007). The task is to move this cursor to a single target by moving both hands forward (Figure 3). This one-cursor task can be compared with a biomechanically identical two-cursor task, in which both hands move separate cursors to separate targets. While the feed-forward commands (black line) are identical in both situations, the feedback corrections are very different. In the two-cursor task, a mechanical perturbation to one hand causes only a local feedback correction in that hand (red arrow). In the one-cursor task, a perturbation of one hand leads to a coordinated response in the unperturbed hand (blue arrow). This response is independent of actual visual feedback from the cursor. Thus, it depends only on proprioceptive feedback from the other arm. The coordinative feedback response is fast and is measurable in the EMG as early as 70 ms after the perturbation (Mutha et al. 2009, Dimitriou et al. 2012). Finally, the response is highly task-dependent and can be switched on the basis of a visual instruction cue alone (Diedrichsen et al. 2009) .

The same task also provides an example of how coordinative feedback responses can lead to increased variability along the uncontrolled manifold, even during unperturbed movements (see Figure 1). In the two-cursor task, the endpoint variability of the two hands in the horizontal direction is almost entirely independent. In the one-cursor task, there is a negative correlation between the endpoints, with lateral deviations of one hand being compensated for by a mirror-symmetric lateral deviation of the other hand. Simulations show that this synergy can be explained by task-dependent feedback mechanisms identified during the mechanical perturbation (Diedrichsen 2007). This task-dependent correlation structure arises gradually over the time course of the movement, consistent with the role of feedback processes.

Neural representation of bimanual movements

The existence of task-flexible coordinative bimanual control demands that the neural areas that control each hand have access to information about the other arm. However, the movements of each arm are nearly exclusively controlled by the contra-lateral motor cortex – at least in terms of cortical projections activating spinal motor neurons (Soteropoulos et al. 2011). Despite this contra-lateral output organization, it is clear that motor cortical neurons receive input about ongoing movements on the ipsilateral side. During unimanual movements, finger (Diedrichsen et al. 2012) and arm (Ganguly et al. 2009) movements can be decoded from the ipsilateral motor cortex. During bimanual movement, motor cortical cells show nonlinear tuning for the combination of the contra- and ipsi-lateral movement (Donchin et al. 1998). One possible function of this arrangement is to endow motor cortex with the ability to make the motor commands to the contralateral arm dependent on the state – sensed or predicted – of the ipsilateral arm. The finding that each arm can learn force fields that are non-linearly dependent on the movements of the other arm is consistent with this organization pattern (Yokoi et al. 2011). In this way, and despite the mainly contralateral organization of behavioral outputs, the two motor cortices jointly form a coordinative bimanual controller.

Anticipatory postural adjustments (APA)

A large class of coordinative phenomena is related to the control of posture. In feed-forward control, postural coordination can be observed whenever participants execute an action that changes the body’s center of gravity. For example, when rapidly raising an arm, the ankle flexors are activated shortly before the arm muscles to shift the body forward and thereby neutralize the interaction torques (Bouisset et al. 1987). Similarly, when dropping a weight that is held with an outstretched arm, the ankle and hip compensate for the anticipated shift of the centre of gravity (Slijper et al. 2002). APAs are acquired through learning; for some behaviors the full adult-like pattern does not emerge until the age of 8 years (Schmitz et al. 2002). Well-learned APAs may be controlled from motorcortical (Viallet et al., 1992), but also brainstem circuits (Prentice & Drew, 2001). However, even in adulthood, APAs remain flexible and are modifiable by learning and contextual information (Morton et al. 2001). The ability to make these adjustments - but not their storage - depends on the integrity of the cerebellum (Diedrichsen et al. 2005).

Arm-finger coordination

Another well-studied example of coordination is the adjustment of the grip force applied to an object when moving the load-bearing arm. Upwards acceleration of the arm increases the load force on the object, therefore necessitating an increase in the strength with which the hand grips the object. Downward acceleration decreases the load force and allows for reductions of the grip force. The grip-force load-force coupling is very tight and clearly predictive (Flanagan et al. 1993). Like other coordinative behaviors, it is highly adaptable, and only occurs when the load changes are caused by an individual’s own actions (Blakemore et al. 1998).

A second form of coordination between arm and finger movement occurs when reaching out to grasp an object. The opening of the fingers depends strongly and reliably on the state of the arm movement, and is adjusted rapidly following perturbations to the arm (Haggard 1997).

References

Bernstein, N. A. (1967) The co-ordination and regulation of movement, Oxford, Pergamon.

Blakemore, S. J., Goodbody, S. J. & Wolpert, D. M. (1998) Predicting the consequences of our own actions: the role of sensorimotor context estimation. Journal of Neuroscience, 18, 7511-8.

Bouisset, S. & Zattara, M. (1987) Biomechanical study of the programming of anticipatory postural adjustments associated with voluntary movement. J Biomech, 20, 735-42.

D'avella, A., Portone, A., Fernandez, L. & Lacquaniti, F. (2006) Control of fast-reaching movements by muscle synergy combinations. J Neurosci, 26, 7791-810.

Diedrichsen, J. (2007) Optimal task-dependent changes of bimanual feedback control and adaptation. Curr Biol, 17, 1675-9.

Diedrichsen, J., Criscimagna-Hemminger, S.E., & Shadmehr, R. (2007). Dissociating timing and coordination as functions of the cerebellum. "J Neurosci", 27, 6291-6301.

Diedrichsen, J. & Gush, S. (2009) Reversal of bimanual feedback responses with changes in task goal. J Neurophysiol, 101, 283-8.

Diedrichsen, J., Shadmehr, R. & Ivry, R. B. (2010) The coordination of movement: optimal feedback control and beyond. Trends Cogn Sci, 14, 31-9.

Diedrichsen, J., Verstynen, T., Hon, A., Lehman, S. L. & Ivry, R. B. (2003) Anticipatory adjustments in the unloading task: Is an efference copy necessary for learning? Exp Brain Res, 148, 272-276.

Diedrichsen, J., Verstynen, T., Lehman, S. L. & Ivry, R. B. (2005) Cerebellar involvement in anticipating the consequences of self-produced actions during bimanual movements. J Neurophysiol, 93, 801-12.

Diedrichsen, J., Wiestler, T. & Krakauer, J. W. (2012) Two Distinct Ipsilateral Cortical Representations for Individuated Finger Movements. Cereb Cortex.

Dimitriou, M., Franklin, D. W. & Wolpert, D. M. (2012) Task-dependent coordination of rapid bimanual motor responses. J Neurophysiol, 107, 890-901.

Donchin, O., Gribova, A., Steinberg, O., Bergman, H. & Vaadia, E. (1998) Primary motor cortex is involved in bimanual coordination. Nature, 395, 274-8.

Flanagan, J. R. & Wing, A. M. (1993) Modulation of grip force with load force during point-to-point arm movements. Exp Brain Res, 95, 131-43.

Ganguly, K., Secundo, L., Ranade, G., Orsborn, A., Chang, E. F., Dimitrov, D. F., Wallis, J. D., Barbaro, N. M., Knight, R. T. & Carmena, J. M. (2009) Cortical representation of ipsilateral arm movements in monkey and man. J Neurosci, 29, 12948-56.

Haggard, P. (1997) Coordinating actions. Q J Exp Psychol A, 50, 707-25.

Haken, H., Kelso, J. A. S. & Bunz, H. (1985) A theoretical model of phase transitions in human hand movements. Biological Cybernetics, 51, 347-356.

Hallett, M., Shahani, B.T., & Young, R.R. (1975). EMG analysis of stereotyped voluntary movements in man. "J Neurol Neurosurg Psychiatry", 38, 1154-1162.

Johansson, R. S., Theorin, A., Westling, G., Andersson, M., Ohki, Y. & Nyberg, L. (2006) How a lateralized brain supports symmetrical bimanual tasks. PLoS Biol, 4, e158.

Karst, G.M., & Hasan, Z. (1991) Timing and magnitude of electromyographic activity for two-joint arm movements in different directions. "J Neurophysiol", 66, 1594-1604.

Kelso, J. A. S. (1984) Phase transitions and critical behavior in human bimanual coordination. Am J Physiol, 246, R1000-4.

Kelso, J. A. S., Southard, D. L. & Goodman, D. (1979) On the coordination of two-handed movements. Journal of Experimental Psychology: Human Perception & Performance, 5, 229-238.

Kennerley, S., Diedrichsen, J., Hazeltine, E., Semjen, A. & Ivry, R. B. (2002) Callosotomy patients exhibit temporal and spatial uncoupling during continuous bimanual movements. Nature Neuroscience, 5, 376-381.

Latash, M. L. (2012) The bliss (not the problem) of motor abundance (not redundancy). Exp Brain Res, 217, 1-5.

Massion, J. (1984) Postural changes accompanying voluntary movements. Normal and pathological aspects. Human Neurobiology, 2, 261-7.

Mechsner, F., Kerzel, D., Knoblich, G. & Prinz, W. (2001) Perceptual basis of bimanual coordination. Nature, 414, 69-73.

Morton, S. M., Lang, C. E. & Bastian, A. J. (2001) Inter- and intra-limb generalization of adaptation during catching. Exp Brain Res, 141, 438-45.

Mutha, P. K. & Sainburg, R. L. (2009) Shared Bimanual Tasks Elicit Bimanual Reflexes during Movement. J Neurophysiol.

Prentice S.D. & Drew T. (2001) Contributions of the reticulospinal system to the postural adjustments occurring during voluntary gait modifications. "J Neurophysiol", 85, 679–698.

Pruszynski, J. A., Kurtzer, I., Nashed, J. Y., Omrani, M., Brouwer, B. & Scott, S. H. (2011) Primary motor cortex underlies multi-joint integration for fast feedback control. Nature, 478, 387-90.

Schmitz, C., Martin, N. & Assaiante, C. (2002) Building anticipatory postural adjustment during childhood: a kinematic and electromyographic analysis of unloading in children from 4 to 8 years of age. Exp Brain Res, 142, 354-64.

Schoener, G., Haken, H. & Kelso, J. A. S. (1986) A Stochastic theory of phase transitions in human hand movement. Biological Cybernetics, 53, 247-257.

Scholz, J. P. & Schoner, G. (1999) The uncontrolled manifold concept: identifying control variables for a functional task. Exp Brain Res, 126, 289-306.

Slijper, H., Latash, M. L. & Mordkoff, J. T. (2002) Anticipatory postural adjustments under simple and choice reaction time conditions. Brain Res, 924, 184-97.

Soteropoulos, D. S., Edgley, S. A. & Baker, S. N. (2011) Lack of evidence for direct corticospinal contributions to control of the ipsilateral forelimb in monkey. J Neurosci, 31, 11208-19.

Todorov, E. & Jordan, M. I. (2002) Optimal feedback control as a theory of motor coordination. Nat Neurosci, 5, 1226-35.

Valero-Cuevas, F. J., Venkadesan, M. & Todorov, E. (2009) Structured variability of muscle activations supports the minimal intervention principle of motor control. J Neurophysiol, 102, 59-68.

Vetter, P., Flash, T. & Wolpert, D. M. (2002) Planning movements in a simple redundant task. Curr Biol, 12, 488-91.

Viallet, F., Massion, J., Massarino, R., & Khalil, R. (1992) Coordination between posture and movement in a bimanual load lifting task: putative role of a medial frontal region including the supplementary motor area. "Exp Brain Res", 88, 674-684.

Yokoi, A., Hirashima, M. & Nozaki, D. (2011) Gain field encoding of the kinematics of both arms in the internal model enables flexible bimanual action. J Neurosci, 31, 17058-68.

see also

Coordination Dynamics Eye-hand coordination Motor cortex: Control Optimal Feedback Control

Sponsored by: Prof. Daniel Wolpert, University of Cambridge, UK

Reviewed by: Anonymous (via Prof. Daniel Wolpert, University of Cambridge, UK)

Reviewed by: Anonymous (via Mr. Travis DeWolf, Computational Neuroscience Research Group, University of Waterloo, ON, Canada)

Accepted on: 2012-12-09 12:20:56 GMT