Interlimb Coordination: Neural, Dynamical, and Cognitive Constraints

This comprehensive edited treatise discusses the neurological, physiological, and cognitive aspects of interlimb coordination. It is unique in promoting a multidisciplinary perspective through introductory chapter contributions from experts in the neurosciences, experimental and developmental psychology, and kinesiology. Beginning with chapters defining the neural basis of interlimb coordination in animals, the book progresses toward an understanding of human locomotor control and coordination and the underlying brain structures and nerves that make such control possible. Section two focuses on the dynamics of interlimb coordination and the physics of movement. The final section presents information on how practice and experience affect coordination, including general skill acquisition, learning to walk, and the process involved in rhythmic tapping.

• Language: English
• Publisher: Academic Press
• Release date: Oct 22, 2013
• ISBN: 9781483289243

Book preview

gratitude.

1

Topics on Interlimb Coordination

Stephan P. Swinnen,     Laboratorium Motorische Controle, Departement Kinantropologie, Katholieke Universiteit Leuven, Leuven, Belgium

Jean Massion,     Laboratoire de Neurosciences Fonctionelles, Centre National de la Recherche Scientifique, Marseille, France

H. Heuer,     Institut für Arbeitsphysiologie, Universität Dortmund, Dortmund, Germany

Publisher Summary

This chapter provides an overview of an elementary framework of the field of interlimb coordination. Humans are capable of coordinating various limbs and body parts with each other—for example, the left and right hands or thumbs, the hand and foot, and the head and arm. One of the most intensively studied examples of interlimb coordination in the field of neuroscience is animal and human locomotion. The neural substrate underlying limb coordination is structured in such a way that flexibility and differentiated activity is possible whenever required. Afferent information plays a crucial role in interlimb coordination, and it is at the basis of the phase-dependent modulation of patterns of muscle activity.

I Introduction

II Exploring the Neural Basis of Interlimb Coordination

A The Neural Control of Locomotion

B The Neural Control of Posture

C Coordination of Head, Body, and Limbs

D The Neural Control of Goal-Directed Arm Movements

E Bimanual Coupling and Decoupling: Central and Peripheral Contributions

III In Search of the Equations of Motion Underlying Interlimb Coordination: Interlimb Dynamics

IV Modulation of Coordination Patterns through Practice and Experience

A Development of Coordination Patterns

B Acquisition of Coordination Patterns

V Conclusion

References

I Introduction

Humans are capable of coordinating various limbs and body parts with each other, for example, the left and right hands or thumbs, the hand and foot, and the head and arm. A remarkable spatiotemporal coordination is evident in spite of the large differences in inertial characteristics of the effectors involved. This observation suggests the existence of some basic coordination principles that apply across widely different cooperative ensembles. Underlying this well-organized global behavior is the coordination of subcomponents at various levels of the movement apparatus: intrajoint, intralimb, and interlimb. At the level of the individual joint, coordination between muscles acting on a common joint is required, such as the interplay between agonists and antagonists. Within a limb, the various joints and muscles, acting on one or more of these joints, must be properly organized to function efficiently. Finally, interlimb coordination is necessary to perform the most essential animal functions like walking, swimming, and feeding.

Whereas particular expressions of interlimb coordination such as locomotion have been investigated intensively in the neurosciences during the past 30 years, the interest in coordination within the behavioral sciences is relatively recent. Due to the development of new movement registration technologies, increased computational power, and the search for new links with the neurosciences and biophysics, the way has been made free for the study of more complex motor behaviors. This is an important development, since the capability to coordinate our limbs is at the heart of everyday life.

Two scientists, who were already actively involved in interlimb coordination research more than half a century ago, can be considered pioneers in this field. First, we owe a great deal to the Russian physiologist and movement specialist Bernstein (1967) who was particularly interested in studying complex motor acts. He was mainly struck by the observation that the movement apparatus, with such a tremendous degree of multilayered complexity, can accomplish goal-directed behavior so effortlessly. This came to be known as the degrees-of-freedom problem. Second, the German behavioral physiologist Von Hoist collected miles of data on the coordination of fin movements in decapitated fish (Labrus). He divided the observed coordination patterns into two major categories: absolute and relative coordination. Absolute coordination is characterized by the maintenance of a fixed-phase relation and by frequency synchronization of the fin movements. Relative coordination refers to a larger group of coordination patterns characterized by less stringent coupling modes, that is, the component activities are neither completely independent of one another nor linked in a fixed mutual relationship. Whereas this distinction is theoretically relevant, Von Hoist remarked that both types are often observed intermittently in fish preparations. In addition, he derived two basic principles that pertained to these coordination modes. On the one hand, he observed a tendency for each fin pattern to maintain its own frequency, referred to as the maintenance tendency (Beharrungstendenz). On the other hand, a tendency for one fin pattern to impose its inherent frequency on the other fin was often evident. This form of cooperation or (mutual) attraction of the fin movements was referred to as the magnet effect (Magnet-effect). The latter effect was often associated with the superposition effect that pertains to attraction between rhythmic units in the amplitude domain.

Von Hoist argued that the magnet effect and the maintenance tendency are in mutual opposition: If the former predominates, then there is continuous agreement in frequency under absolute coordination; if the latter predominates, there is relative coordination—the frequencies of the rhythms differ, and the dependent rhythm, under the magnet effect of the dominant rhythm, exhibits periodicity whose form is determined by the reciprocal frequency relationship and whose extent is governed by the intensity of the magnet effect (Von Hoist, 1973, p. 63).

Even though these principles were extracted from research on fin movements, they currently form a major source of inspiration for the study of human coordination (Kelso, Chapter 15, this volume; Turvey & Schmidt, Chapter 14, this volume). Today, many research laboratories across the world investigate these phenomena in a variety of different contexts. Others are more concerned with the study of discrete bimanual tasks in which the limbs assume differentiated roles to accomplish goal-directed behavior (Fagard, Chapter 21, this volume; Peters, Chapter 27, this volume; Walter & Swinnen, Chapter 23, this volume).

The present book consists of a series of introductory chapters, representing various levels of research and subdivisions of science that currently address interlimb coordination, for example, the neurosciences, the behavioral sciences, kinesiology, biomechanics, and dynamics. Even though each of these fields of science is characterized by a unique approach to the study of interlimb coordination, using its own techniques to acquire knowledge, all strive for a better understanding of how the human control system manages to organize the cooperation among the limbs. Neuroscientific approaches focus on the neuronal networks and pathways underlying rhythmic and discrete coordination patterns, in particular locomotion and bimanual coordination. Some chapter contributions concentrate on the identification of the locus of the central pattern generator underlying locomotion, whereas others are mainly concerned with the reflex modulation of these patterns as a result of sensory information. Scientists advocating a dynamical approach seek to uncover the equations of motion that govern movement coordination. They attempt to identify the dynamic states at which moving animals converge when provided enough time to settle down. Finally, some scientists are mainly concerned with a better understanding of goal-directed motor behavior and the changes in coordination that occur as a result of development and learning, that is, the modulation or overcoming of preexisting/preferred coordination modes with the goal of expanding the behavioral repertoire. Those who have a strong link with cognitive psychology direct their attention to a better understanding of the nature of the central representation underlying complex coordination and the movement features it comprises.

II Exploring the Neural Basis of Interlimb Coordination

A The Neural Control of Locomotion

One of the most intensively studied examples of interlimb coordination in the field of neuroscience is animal and human locomotion. Since Sherrington, three areas of interest have dominated experimental studies on locomotion: (1) the role of reflexes in locomotion; (2) the capability of the spinal cord to generate intrinsic rhythms; and (3) the control of the spinal cord by higher centers. At one time or another, attention has mainly been directed at one of these mechanisms for motor control. More recent studies have concentrated on the synthesis of these mechanisms into a general framework for nervous control (Shepherd, 1988). This also typifies the chapters on locomotion in this volume. The contributions refer to the study of locomotion in invertebrates such as the crayfish (Cattaert et al., Chapter 3) and higher vertebrates, including those using a quadrupedal gait, such as the cat (Kato, Chapter 4), and a bipedal gait, such as the human (Brooke et al., Chapter 6; Duysens & Tax, Chapter 5).

Pioneering work on locomotion was conducted by Grillner (1975, 1981) and Shik and co-workers (Shik, Severin, & Orlovsky, 1966), who spent considerable efforts in demonstrating the existence of a relatively autonomous neural network, called the central pattern generator (CPG) (see also Cattaert et al., Chapter 3, this volume; Duysens & Tax, Chapter 5, this volume). This confirmed earlier ideas put forward by T. Graham Brown in 1911, who demonstrated that the rhythmic alternation between flexion and extension is not reflex in origin but is generated by neurons located in the spinal cord. CPGs have been demonstrated in most locomotory networks found in invertebrates and vertebrates. The rhythm production generally results from both membrane properties of neurons and particular network connections.

Even though these patterns of interlimb coordination can be observed in the absence of afferent information, this should not be taken to imply that afference plays a minor role in normal locomotion. Brown (1911) was well aware of this when he suggested that afferent input was probably important in grading the component movements to the specific environmental contingencies. Since locomotion is a highly automated type of motor behavior, it is not surprising that interlimb reflexes have evolved to support the coordination of the limb movements during gait and to modulate the basic patterns during unexpected perturbations (Duysens & Tax, Chapter 5, this volume).

In Chapter 4 of this volume, Kato reviews his experimenal contributions of the past 15 years, which deal with locomotor coordination after horizontal and longitudinal separation of the spinal cord in spinal intact cats and spinal lesioned cats. Lateral hemisection of the spinal cord was carried out in order to disrupt descending and ascending long tracts unilaterally. These hemisected preparations do not show any differences in step length or in step time when compared with normal control cats. However, they do demonstrate evidence for less accurate foot placement responses when walking on grid surfaces. According to Kato, this suggests that interlimb reflex pathways serve to coordinate the spatial aspects of locomotion in quadrupedal gait. On the other hand, spinal transsection or double hemisection results in a disruption of the phase relations between the fore- and hindlimbs, and this points to the importance of descending signals from brainstem locomotor centers for achieving coordination among the limbs.

Some chapters specifically deal with the reflex modulation of locomotory activities. Cattaert and co-workers have investigated two locomotor systems in the crayfish (Crustacea): (1) swimming, accomplished by four pairs of paddles; and (2) walking, by means of four to five pairs of thoracic legs. They show some nice examples of sensory-motor interactions, which are analyzed at the cellular level. A comparison of both systems reveals the existence of similarities between their central commands. But, there exist essential differences as well, pertaining to the presence or absence of contact with rigid substrates, as is the case with the legs during walking. The control of ongoing movement during walking is strongly tied with the stance phase where force coding receptors are active. These determine the switch to the swing phase in the leg where the force coding receptors reside, as well as the onset of the stance phase of the ipsilateral legs. Thus, coordination in the actual walking system results from strong interactions between central and peripheral commands that allow for maximum adaptability to environmental demands (e.g., the quality of the terrain).

Brooke and Duysens and their respective co-workers concentrate their reviews on sensori-motor interactions during human walking. Brooke and co-workers have studied one of the most rapid autogenic pathways that is represented by the H-reflex. They show that rhythmic limb movements, as generated during pedaling, cause modulation of the spinal pathway of the Ia H-reflex. For soleus H-reflexes during symmetric pedaling, they demonstrate (1) that passive movement of the legs results in movement-induced inhibition of this spinal pathway; (2) that the extent of the reflex depression is positively related to the velocity of passive movement of the two legs; and finally (3) that the relationship to velocity is maintained when the legs move actively. During asymmetric coordination of movement of one leg while the other remains passive, the inhibition is retained in the leg being moved but also in the static contralateral limb, and this is again related to the velocity of passive movement of the opposite limb. They propose that the inhibition is transmitted through the spinal pathway and that it arises from the receptors generating movement afference.

Duysens and Tax (Chapter 5) review some slower acting pathways that may be involved in the production of corrective movements during locomotory events. Whether interlimb reflexes are in operation during normal locomotion or following perturbations, they subserve the important goals of minimizing instability and securing progression. The emerging corrective movements may involve limbs other than those upon which a particular perturbation or stimulus is evoked. Two examples clearly illustrate this effect. On one hand, prolongation of the swing phase of one leg results in a longer duration of the support phase in the contralateral leg. On the other hand, a shortening of the stance phase in one leg results in a shortening of the swing phase in the contralateral leg to take up the support function earlier in time. That stimuli- or perturbation-induced inter- and intralimb responses depend on the phase of the locomotory cycle is a true example of highly efficient and goal-dependent movement organization. Many examples of such phase-dependent responses in cat and human can be found in Chapter 5 by Duysens and Tax, who conclude that there are striking similarities in patterns of interlimb coordination among these species even though we are dealing here with quadrupedal versus bipedal walkers. Differences among animals and humans are to be sought in the greater dominance of supraspinal over spinal mechanisms in the latter (Dietz, 1992).

Even though CPGs are held responsible for interlimb coupling, Grillner (1985) has also underscored the existence of individual pattern generators for each limb (see also Cattaert et al., Chapter 3, this volume). Activity in one limb can be observed in the absence of activity in the other limbs. Kato supports these findings in cat preparations where all the impulses from contralateral as well as supralumbar regions are cut off. The ipsilateral hindlimb shows a walking pattern, independent of the other three legs (Chapter 4). Moreover, Grillner (1985) has suggested that these neural networks cannot only be made responsible for complex coordinations but also for the production of more specific movements. Only parts of the total unit are then activated for the volitional control of more specific ankle or knee movements. Within this perspective, the production of new movements may then require learning to combine and sequence specific fractions of the neuronal apparatus used to control the innate movement patterns in a novel way (Grillner, 1985, p. 148).

Whereas it has often been assumed that a CPG consists of a well-defined assemblage of neurons functionally distinguishable from others, Meyrand, Simmers, and Moulins (1991) have recently argued that CPGs may not be considered immutable functional entities. Instead, they show in Crustacea that neurons from different circuits can be reconfigured into a new circuit that enables a different function. This selective dismantling of preexisting networks, giving rise to the recruitment of a different functional network, provides us with an important clue toward a better understanding of the mysterious but enormous flexibility in motor coordination that can be found across the animal world. Additional evidence is required to demonstrate similar phenomena in the vertebrate nervous system, which evidently has a much larger number of neurons.

In summary, the neural substrate underlying interlimb coordination is structured in such a way that flexibility and differentiated activity is possible whenever required. Afferent information plays a crucial role in interlimb coordination and it is at the basis of the phase-dependent modulation of patterns of muscle activity.

B The Neural Control of Posture

Even though posture looks like a rather static activity at first sight, it is a true example of interlimb coordination. A major requirement for postural equilibrium is the maintenance of the center of gravity within the base of support. This process of balance control, or, of the regaining of balance, involves coordination of many muscles. Early studies already demonstrated that the production of focal arm movements is often preceded by bilateral leg activity in order to maintain balance. For example, Belenkii Gurfinkel and Paltsev (1967) and Paltsev and Elner (1967) investigated the interactions between arm lifting and lowering and postural activity. They found that activity in the m. deltoideus (responsible for the arm action) showed an increase in activity 130–140 msec after the sound signal, whereas the biceps femoris on the ipsilateral side was activated up to 40–50 msec before onset of deltoid activity. The biceps femoris on the contralateral side showed increased electrical activity up to 30–40 msec before its contralateral counterpart. These observations suggest that even small voluntary actions, limited to a particular body part, cause a complete reorganization of muscle activity in many body parts and this occurs with the appropriate timing of the events.

In Chapter 8, Dietz describes his work on the perturbation of stance, as applied to one or both legs in the same or in opposite directions. He underscores that a combination of afferent inputs provides the necessary information to control body equilibrium. He deals with some of the interactions among the various input sources. Dietz maintains that during bilateral leg displacements, the activity induced by the respective contralateral leg is linearly summed or subtracted, depending on leg displacement in the same or in opposite directions. Thus, following unilateral displacements, a bilateral activation takes place that results in cocontraction of the nondisplaced leg. This results in a more stable base from which to compensate for the perturbation. A displacement of the legs in opposite directions causes the body’s center of gravity to fall between the legs, and therefore, less invasive compensatory responses are needed. The resulting coordination is thought to be mediated by a spinal mechanism, due to the observed short latencies, which is under supraspinal control. Moreover, Dietz hypothesizes that extensor load receptors (Golgi tendon organs) signal changes with respect to the projection of the body’s center of mass in reference to the feet. This possibly represents a newly discovered function of these receptors in the regulation of stance and gait.

Lee and Russo address Bernstein’s degree-of-freedom problem through an attempt to characterize the constraints that underlie interlimb coordination during force pulling with the arms while standing upright (Chapter 25). They distinguish two ways to model how the system’s degrees of freedom may be constrained to achieve coordination. Local constraints relate to joint kinematic and kinetic variables or muscle variables. Global constraints relate to abstract goals of the system, such as maintaining or accelerating the center of mass to a particular location over the support base. They propose an analytical model that operationally defines global coordination in the task and that is partially supported by empirical work (except for very low force pulls).

In summary, what emerges from perturbation studies during locomotion and stance is that the corrective movements thus induced are not limited to the locus of the perturbation but consist of an overall response involving the coordination of many muscles and body parts within a kinematic chain. Interlimb reflexes serve to secure the principal goals of minimizing instability and, in the case of locomotion, of enabling progression. Dietz (1992) summarizes: Irrespective of the conditions under which stance and gait are investigated, the neuronal pattern evoked during a particular task is always directed to hold the body’s center of mass over the base of support (p. 48). For extensive reviews on the neural control of posture and movement, we refer the refer the reader to Massion (1992) and Dietz (1992).

C Coordination of Head, Body, and Limbs

Limb movements are dependent on or covary with head position, as becomes evident in labyrinthine reflexes and symmetric and asymmetric tonic neck reflexes. These reflexes often interact with each other during movement production. For example, head dorsiflexion induces extension of the forelimbs and flexion of the hindlimbs, whereas the opposite effect occurs during ventriflexion (symmetric tonic neck reflex). The neck reflexes can be observed in the newborn but gradually disappear during the first months of life. However, some have argued that such reflexes do not fully disappear and may show up during the production of movements in sports events or other voluntary activities (Keele, 1981). Fukuda (1961) captured many skilled performers on film and found various patterns of limb coordination that are congruent with those found in the tonic neck reflexes. In addition, Hellebrandt, Houtz, Partridge, and Walters (1956) underscored the role of reflexes during the production of forceful events and demonstrated that patterns in accordance with these reflexes augment work output. It is thus reasonable to assume that some patterns of coordination are built into the organism and become subsequently integrated into voluntary activities. Coaches are aware of the important role of head position during skill acquisition. Instructions often relate to particular head movements (look at the floor during a handstand) in order to promote these built-in patterns of activity. Sometimes, these patterns are conducive to the new coordination form to be acquired; at other times, they may impose persistent errors in performance that need to be suppressed (Walter & Swinnen, Chapter 23, this volume). More will be said about the two faces of preexisting coordination modes later in this chapter.

In their chapter on head an body coordination, Berthoz and Pozzo argue that the head serves as an important frame of reference during multilimb coordination (Chapter 7). Depending on the type of activity to be executed, the head is stabilized intermittently under the control of gaze. This stabilization allows the head to serve as an inertial guidance platform for the control of multilimb movement. They infer this from the strong tendency of performers to stabilize the head with respect to the sagittal as well as the frontal plane. Head stabilization may simplify the transformations necessary to set up a coherent internal representation of external space. During various types of walking, head angular displacement in the sagittal plane remains within a small range when compared with the movements of the other limbs. When trunk movements are limited, the head is locked onto the trunk. During complex balancing tasks (standing on a narrow beam or on a semicircular platform), the head is again stabilized whereas the trunk is making the compensatory movements. A remarkable stabilization of the head in the frontal plane can also be observed during downhill skiing.

D The Neural Control of Goal-Directed Arm Movements

Through evolution, humans have evolved from a quadrupedal to a bipedal gait. This has freed the upper limbs for various manipulatory activities. Depending on the task under consideration, both hands are being used in a symmetrical manner (Ohtsuki, Chapter 13, this volume), or one hand performs the focal movement while the other serves a stabilizing function, providing a positional reference (see Peters, Chapter 27, this volume). This particular type of interlimb cooperation is currently investigated within various subdivisions of science. Neuro-scientists are predominantly interested in identifying the brain structures that are uniquely involved in upper-limb control, whereas others focus on the identification and description of bimanual interactions through kinematic and kinetic analyses. On one hand, a strong tendency is observed for both limbs to be synchronized during their simultaneous performance (Kelso, Southard, & Goodman, 1979). On the other hand, bimanual activity can be differentiated to a great extent through practice and experience (Swinnen, Walter, & Shapiro, 1988a, Swinnen, Walter, Beirinckx, & Meugens, 1991a, Swinnen, Young, Walter, and Serrien, 1991b). When two tasks have to be performed that differ in their difficulty level, the left hand is usually assigned the easier and the right hand the more complex, attention-demanding task. According to Peters, this is the predominant allocation of tasks in righthanders, whereas the situation is more complicated in left-handers (Chapter 27, this volume).

In their excellent review, Wiesendanger et al. (Chapter 9, this volume) discuss evidence concerning the unique involvement of particular cortical areas in bimanual activities. Their review is built on the distinction between movements in which both upper limbs are strongly synchronized and movements in which asymmetrical contributions are required to accomplish a goal. The former type of activity is often characterized by simultaneous initiation of movement, which may be achieved by direct and indirect bilateral connections between motor and premotor cortical areas and the spinal cord. Each hemisphere is in principle capable of exerting this bilateral control, thereby assuring a tight coupling of the limbs. More will be said about this later in this chapter. These movements are relatively well preserved following cortical lesions outside the primary motor cortex and they do not seem to require intact commissures when they are of limited complexity (Wiesendanger et al., Chapter 9).

Many goal-directed actions involve asymmetrical contributions from both arms. This may require suppressing the strong initial tendency for bimanual coupling or synchronization (Peters, Chapter 27, this volume; Swinnen et al., 1991b). Wiesendanger and collaborators suggest that various types of cortical lesions may impair the performance and/or learning of such differentiated bimanual activities: the parietal cortex, the premotor cortex, the anterior corpus callosum, and the medial frontal cortex (including the supplementary motor area). For that reason, they conclude that it is not justified by the current evidence to hypothesize a single cortical superarea functioning as a unifying structure. It is rather the case that widely distributed cortical association areas cooperate as an interconnected ensemble to produce goal-directed bimanual actions (Chapter 9).

Split-brains form a unique population to investigate the role of the corpus callosum in performing and learning bimanual skills and other movements involving both body sides. In Chapter 10, Berlucchi et al. review their work on reaction times for simple responses using axial, proximal, and distal muscles of the upper limbs in normal subjects and subjects with a callosal deficit. In comparison to normal subjects, they demonstrate that the time needed for interhemispheric integration of crossed responses becomes much larger in patients with a callosal deficit and this is the case for unilateral and bilateral distal responses and unilateral proximal responses. It is inferred from these findings that the intact corpus callosum allows for efficient interhemispheric communication in order to secure integration of these responses. However, when proximal bilateral responses and unilateral and bilateral axial responses are made, other pathways are hypothesized to be responsible for integration. The authors argue that a bilaterally distributed motor system secures the production of these symmetrical movements involving axial and proximal limb muscles, and this system can be called upon by each of both hemispheres. Indeed, the acallosal subjects show a strong tendency for bimanual synchronization under these circumstances. Thus, whereas a callosal contribution seems necessary for the bilateral synchronization of distal responses, it does not seem critical when proximal or axial movements are involved. In the latter case, synchronization between the two sides is made possible through the common origin of the motor commands and through the bilateral distribution of the pathways transmitting them.

This bilateral control system, to which Wiesendanger and co-workers also refer, has been demonstrated anatomically and physiologically in primates by Kuypers (1973, 1985; Brinkman & Kuypers, 1973). It is reviewed in detail by Shinoda and co-workers in Chapter 2, this volume. The descending pathways from the cerebral cortex and the brain stem to the spinal cord can be distinguished into those that connect contralaterally with the dorsolateral part of the spinal intermediate zone and those that connect bilaterally with its ventromedial parts. Cells of the dorsolateral part (the lateral descending motor tract group) seem to distribute preferentially to motoneurons of distal extremity muscles, whereas cells of the ventromedial part (the medial descending motor tract group) to motoneurons of axial and proximal limb muscles. The latter group is phylogenetically and ontogenetically older than the former group. Furthermore, the single long descending motor tract axons exert their effects on different groups of spinal neurons simultaneously through multiple axon collaterals, thereby enabling control of the excitability of multiple muscles at multisegmental levels concurrently. This wide degree of divergence that characterizes the single long descending axons may constitute the neural substrate underlying the appropriate combination of muscles into functional synergies (Shinoda et al., Chapter 2).

E Bimanual Coupling and Decoupling: Central and Peripheral Contributions

When both arms are moved simultaneously, a strong synchronization tendency becomes evident and this is interpreted by some to indicate that the limb musculature is constrained to act as a single functional unit or coordinative structure (Kelso et al., 1979). In the previous section, we have already elaborated on the neural substrate that may subserve these synchronization effects. Viewed from the perspective of the degrees-of-freedom problem, the existence of these preferred modes of coordination is considered an optimal solution. However, if bimanual movements would always be constrained in this way, humans would largely fail to comply with daily task requirements, which are often characterized by finely differentiated patterns of limb activity (role differentiation, see Fagard, Chapter 21, this volume; Peters, Chapter 27, this volume). Luckily, the human control system is endowed with adaptability and flexibility to overcome these intrinsic coordination tendencies. A nice example of differentiated bimanual movements is provided by Castiello and Stelmach (Chapter 28, this volume). When a precision grip with one hand is performed together with a prehension task involving the whole hand, they show unique grasp-related kinematic profiles that are captured within a common temporal metric. In general, timing is proposed to be a major parameter that constrains bimanual performance (Heuer, 1991).

In a number of recent studies, one of us in collaboration with colleagues has demonstrated that learning results in a progressive dissociation of limb movements that are initially synchronized (Swinnen, Walter, Lee, & Serrien, in press; 1988, 1991b, Walter & Swinnen, 1992, Chapter 23, this volume). Dissociation of movement is not a smooth process but requires subjects to overcome the (mutual) interference that is often evident at the start of practice. Patterns of activity in one limb show up in the other limb or vice versa. One is prompted to ask what the locus of this interference is. Among other things, Cohen (1970) hypothesized that dual-task interference may result from excessive demands in monitoring the discordant proprioceptive information generated by both tasks. However, Teasdale and collaborators (Chapter 12, this volume) refute this explanation. They contend that the locus of the observed interference is to be sought at the level of movement organization and programming. Their argument is based on a comparison of the performance of normal subjects with that of a unique deafferented patient suffering from a total loss of the large sensory myelinated fibers while still possessing an intact peripheral motor system. Similar to Cohen, they have their subjects perform a continuous wrist pronation/supination movement while performing a discrete secondary task upon presentation of an auditory stimulus, for example, a wrist extension or verbal response. If Cohen’s hypothesis is correct, bimanual performance should be less impaired in the deafferented patient. Instead, they show that the absence of proprioceptive information yields a more pronounced interference in performance, and the recovery from the perturbation takes longer. Thus, Teasdale and co-workers’ observations suggest that dual-task interference is not a direct consequence of the monitoring of movement afference. Rather, it arises at the efferent level or during the stage of movement programming and organization. This is in agreement with observations in discrete bimanual tasks, where interference between the electrical activity of the biceps muscles of both upper limbs can already be detected before any movement has taken place (and presumably before any appreciable movement-related afferent information can be generated) (Swinnen et al., 1988, 1991b). In addition, reaction-time studies conducted by Heuer (1990) have provided evidence that intermanual interactions can already be established at the central level of control.

In summary, two important conclusions can be drawn from these observations. On one hand, interference during dual-task performance occurs in spite of the absence of proprioceptive information and can therefore be located at the level of movement planning, organization, and possibly efferent control. On the other hand, afferent information from the moving limbs is important for sustaining interlimb coordination in the face of upcoming distortions or perturbations. This is nicely demonstrated by Teasdale and co-workers, who show that the recovery from interference is less successful in the deafferented patient. More generally, we propose that interlimb coordination for various limb combinations and coupling modes depends on the monitoring of afferent information from the moving limbs. This point is also made by Baldissera and co-workers with respect to ipsilateral control of hand and feet movements (Baldissera, Cavallari, Marini, & Tassone, 1991, Baldissera et al., Chapter 11, this volume). On the basis of their findings with various limb-loading techniques, they argue that kinesthetic afferences are important for sustaining interlimb coordination, especially for anti-phase coupling. They conclude that in-phase and anti-phase coupling of the ipsilateral hand and foot are possibly controlled by means of different feedback systems and different degrees of elaboration of these systems. During in-phase coupling, the feedback system is presumably operating in a less stringent fashion, whereas during antiphase coordination, feedback monitoring from the moving limbs is important and is argued to require constant attention. In addition, one of us has recently observed considerable phase destabilization in blindfolded subjects during the production of a homolateral coordination pattern (forearm-lower leg) as a result of passive movement generated in the contralateral side (Swinnen, Serrien, & Daelman, 1993). Presumably, passive movement generates afferent information that is discordant with the afferent information generated in the actively coordinated limbs.

III In Search of the Equations of Motion Underlying Interlimb Coordination: Interlimb Dynamics

The dynamical approach to interlimb coordination has undergone a major development in the past 15 years (for a review, see Turvey, 1990). In the present context, dynamics is not to be understood in the strict traditional sense as the study of how objects move under the action of forces (the force-mass-acceleration approach). Dynamics is a field emerging between mathematics and the sciences. It deals with changes in systems and tries to express its existing and evolving states. In principle, it is best suited for application to cyclical tasks even though Schöner undertakes an attempt to show how a dynamical approach may address the problem of trajectory formation of a single limb (Schöner, Chapter 17, this volume). A dynamical system consists of two parts: (1) the essential information about a system or the notions of a state; and (2) a rule that describes how the state evolves with time (the dynamic) (Crutchfield, Farmer, Packard, & Shaw, 1986).

Even though the simplest of coordination tasks (like moving two fingers together) is rather complex when considering the cooperation that is required among the multiple subcomponents at various layers of the motor apparatus, the dynamical approach is directed at uncovering the basic principles or laws that characterize interlimb coordination. These principles capture the cooperative behavior of the ensemble and cannot necessarily be inferred from the individual behavior of the subcomponents. As Schöner (Chapter 17) notes, the coordination dynamics are not to be equated with the physical dynamics of the biomechanical system, even though they may contribute to those dynamics. Coordination is argued to be a consequence of evolving processes of self-organization or pattern formation, concepts that figure predominantly in Haken’s synergetics (Haken, 1983; Kelso, Chapter 15, this volume).

Central to the approach is the identification of relevant macroscopic variables and their equations of motion, built around the concept of relative phase. Within the domain of interlimb coordination, relative phase can be defined as the phase difference between two oscillatory signals (in the present case, limb movements). Phase refers to the point of advancement of the signal within a cycle, that is a description of the stage that a periodic motion has reached. Relative phase is thus a useful variable for the assessment of spatiotemporal coordination. It does not provide information of each signal separately but uniquely characterizes the way two signals relate to each other. For a long time, neuroscientists interested in locomotion have used measures of phase angles to determine modes of intra- and intergirdle coordination in animals. It is only recently, however, that scientists have further explored and modeled its characteristic dynamics.

Advocates of the dynamical approach make a distinction between order parameters and control parameters. Order parameters (also called collective variables) characterize the behavioral patterns of interest. Relative phase has been proposed as a primary candidate in this respect (Kelso, Chapter 15; Schöner, Chapter 17; Turvey & Schmidt, Chapter 14). Control parameters induce changes that can be observed and characterized at the level of the order parameters even though they do not contain any specific information about the potential changes that order parameters undergo. A well-known example that aids in clarifying this distinction is the performance of cyclical wrist or finger movements under increased cycling frequency conditions (see also Carson et al., Chapter 16). You can easily try this out yourself: When you start moving both fingers in front of you in the same direction (also called anti-phase or 180° out-of-phase) and you progressively increase the cycling frequency, you will suddenly experience a transition or shift toward symmetrical movements of both fingers (called in-phase), unless you intentionally oppose this rather spontaneous change. Thus, relative phase is said to capture the dynamics of this coordination system, whereas the control parameter (cycling frequency) allows these dynamics to emerge.

A number of additional points are to be made from the aforementioned example. First, although not all behavioral changes take the form of phase transitions, advocates of the dynamical approach argue that phase transitions provide a window into understanding behavioral patterns and, more generally, the behavior of living things. The system evolves from one level of organization to another more stable level and the transition is often preceded by instabilities. A phase transition is the simplest type of self-organization in physics. The changes from a liquid to a gas or solid are most familiar to us. Apparently, no superordinate command structure initiates or controls this transition. Currently, phase transitions are not only viewed within the small time scales evident in the aforementioned examples. Their existence is also presumed along the longer time scales of development and learning (see Section IV). Whereas the traditional experimental approach to human functioning has often concentrated on the study of stable behavioral features, the dynamical approach suggests an alternative perspective whereby instabilities are the major focus of attention. Instabilities serve to demarcate behavioral patterns. Phase transitions that arise out of these instabilities constitute a special entry point for developing a language upon which to build a deeper understanding of the behavior of living things.

Observable behavioral patterns are mapped onto attractors of the order parameter dynamics. Roughly speaking, attractors are what the behavior of a system is attracted too when allowed enough time to settle down. The attractor can be: (1) a single point; (2) a closed curve (or limit cycle), which describes a system with periodic behavior; or (3) a fractal or strange attractor for a system exhibiting chaos. From an experimental point of view, signatures of the dynamics underlying interlimb coordination can be found in the stability of coordination patterns and various stability measures can be used for that purpose. For the specific case of isofrequency finger, hand, or arm movements, these attractors refer to the in-phase and anti-phase coordination mode. The behavioral implications are a natural tendency to fall into these modes even when attempting alternative coordination patterns. Attraction toward in-phase coordination is generally stronger than anti-phase coordination (Kelso, 1984). In the specific case of arm movements in the frontal plane, cyclical movements in different directions (moving toward or away from the body midline) are produced more stably than movements in the same direction. However, a different pattern of findings can be observed for hand (lower arm) and foot (lower leg) movements in the sagittal plane (see Baldissera et al., Chapter 11, this volume; Kelso & Jeka, 1992; Swinnen et al., 1993). Here, movements in the same direction (both up or down) are performed with a higher degree of stability than movements in different directions (one up, one down) and this effect is not muscle specific (Baldissera, Cavallari, & Civaschi, 1982). A similar principle holds for intersegmental coordination, that is movements at the elbow and wrist joints (Kelso, Buchanan, & Wallace, 1991). This suggests that the mutual direction of limb movements is an important determinant of coordinative stability (see also Kelso & Jeka, 1992). Additional research is required to verify the generalizability of this principle to other limb combinations and planes of motion. For some of these coordination patterns, biomechanical interactions between the segments or postural disturbance effects cannot be excluded as partial accounts for differential stability. Proponents of the dynamical approach are predominantly concerned with developing equations of motion that capture the coordination dynamics, whereas neuroscientists focus at unraveling the architecture of the neural networks and pathways that give rise to these preexisting preferred coordination modes.

The aforementioned examples pertain to the specific case of interlimb coordination with frequency and phase locking. Von Hoist referred to this as absolute coordination. However, humans and other species are endowed with a great deal of adaptability and flexibility to defy these elementary coordination modes. This implies that a variety of interlimb patterns can be explored even though tendencies toward certain frequency ratios and toward in- and anti-phase coordination will remain evident (relative coordination) Von Hoist (1973/1937) gave the example of a child walking along with his father. When walking independently, both have different speed and stride amplitude preferences. When walking together, short events of phase and frequency locking will occur, alternated with periods of desynchronization and adjustments toward synchronization. In other words, there is an interplay between cooperation and competition, or in Von Hoist’s terms, between the magnet effect and the maintenance tendency.

Less stringent modes of coupling are more common when the effectors involved are asymmetric. The sources for asymmetry can be manifold, for example, when coordinating different effectors (the arm and leg), or when inertial differences are artificially imposed on the same effectors. In Chapter 14, Turvey and Schmidt describe an experimental task setup in which subjects oscillate hand-held pendulums that can be varied in length. These length variations affect the degree of dissociation between the pendulums’ eigenfrequencies. Under these circumstances, they still show a tendency toward in-phase and anti-phase attraction even though relative phase becomes more destabilized when compared to the symmetrical conditions. The stability differences between the in-phase and anti-phase mode, as mentioned before, tend to become smaller as the difference between the pendulums’ eigenfrequencies becomes larger (greater differences in their length). In addition, these phenomena interact with the overall cycling frequency at which these oscillatory movements are performed (Turvey & Schmidt, Chapter 14). As such, their work elucidates the cooperative and competitive phenomena that characterize interlimb coordination. Both are represented in the equations of motion the authors discuss in their chapter. Similar to Turvey and Schmidt, Carson and associates focus on the component oscillators to improve understanding of their combined operation (Chapter 16, this volume). They conclude that the interaction between intrinsic upper-limb asymmetries and informational and mechanical constraints may have a significant influence on the coupling dynamics.

In summary, the dynamical approach has as its main objective the development of equations of motion governing interlimb coordination and the identification of the dynamics underlying coordination. The tendency toward phase and frequency synchronization, which is ubiquitous in many species at various layers of the motor control apparatus, is thereby underscored. Whereas a strong emphasis was initially placed on the production of movement patterns with a 1:1 frequency ratio (absolute coordination), attention has recently shifted to other forms of coordination, characterized by a less stringent coupling of the limbs (relative coordination). It is now becoming evident that a great variety of coordination modes can be flexibly accomplished even though preferred coordination tendencies remain present. These are the behavioral expressions of a few elementary coordination principles. As will be discussed later, the learning of new coordination patterns can partly be understood against the backdrop of these preexisting coordination modes. Not only intrinsic or preferred coordination modes but also intended, memorized, or learned behavioral patterns are expressed in terms of the dynamics of coordination (see Zanone & Kelso, Chapter 22, this volume).

The dynamical approach brings a new challenge to the reductionist view that focuses on the study of a stystem’s subcomponents. It is also committed to applying a rather universal language for describing both living and nonliving systems. Phase transitions thereby constitute a special entry point for developing a language upon which to build a better understanding of the behavior of living things. However, this approach is only in its initial stage of development and many questions remain as yet unanswered. For example, in addition to manipulating cycling frequency, there are possibly various alternative ways to probe a motor system and elucidate its archaic coordination modes. One of us has observed phase transitions during two-limb (arm-leg) coordination as a result of addition of a third limb or through passive movement of a third limb. What is perhaps common across these various manipulations is that they overload or perturb the system, causing it to regress to its most rudimentary coordination modes. The dynamical approach provides some tools to describe and characterize observed patterns and eventual transitions among patterns. The archaic movement forms that the system settles in when stressed, may also reflect the central nervous system’s most easily potentiated patterns of neural wiring.

IV Modulation of Coordination Patterns through Practice and Experience

In previous sections, we have already hinted at the flexibility and adaptability of coordination patterns. These capabilities become more prevalent as humans develop and learn. Modulation of coordination patterns is evident in the developing child, who undergoes a remarkable evolution in its basic coordination patterns during the first year of life. This pertains to the development of postural control (Woollacott & Sveistrup, Chapter 18, this volume), and quadrupedal and bipedal locomotion, with the major goal of supporting and transporting the body in a gravitational field (Whitall & Clark, Chapter 19). It is also evident, however, in the development of the upper limbs, which become specialized for various manipulative functions. (Corbetta & Thelen, Chapter 20; Fagard, Chapter 21). Even though the time scales underlying changes in behavior can be markedly different, the study of motor development and learning share many commonalities. Both involve the gradual mastering and control of the degrees of freedom inherent in the motor system. Often, preexisting or preferred coordination tendencies must be overcome. Development and learning can then be understood against the background of these preexisting patterns, which form the basic building blocks for the creation of more differentiated patterns of activity. Woollacott and Sveistrup address the degrees of freedom at the muscle synergy level, whereas the other contributors focus on the interlimb level.

A Development of Coordination Patterns

Woollacott and Sveistrup describe the development of posture control during the first year of life, which takes place in a cephalocaudal direction (Chapter 18). They demonstrate that the calibration of input-output relationships between sensory inputs controlling posture and the neck muscles occurs before calibration of the trunk muscles. Later, during the development of pull to stand behavior, the muscles become activated in a distal to proximal sequence, that is from tibialis anterior to quadriceps and abdominal muscles. The postural synergies appear to be temporally organized in an adultlike fashion at about the onset of independent stance. Postural control is a prerequisite for the exploration of the environment and the development of locomotory behavior, discussed next.

Whitall and Clark (Chapter 19, this volume) describe the development of bipedal locomotion (walking, running, and galloping) between the first and second year of life. From the onset of walking behavior, the anti-phase (alternating) pattern is adopted, even though the phasing is more variable than in adults. Adultlike variability is evident after about three months of walking experience. Following that period, other forms of locomotion emerge, such as running and galloping. Galloping is an interesting example in that it requires the lower limbs to be out of phase about 90–120°. Thus, the preferred anti-phase mode of coordination is overcome to explore new movement forms. That small children acquire this skill relatively easy is rather striking in view of recent findings pointing to the relative difficulty of acquiring a 90° out-of-phase pattern with the upper limbs in adults (Lee, Swinnen, & Verschueren, 1993; Zanone & Kelso, 1992, Chapter 22, this