Motor Control & Sensory Integration

200 J. P. Piek reflexes are independent of developing motor control, with these two processes being both neurologically and developmentally distinct, and instead, running in parallel with each other. If indeed reflexes do not form the basis for motor coordination, what other resources are available for the infant to use as a basis for the development of the complex motor patterns that emerge as the infant grows? Over the last decade, considerable research has focused on infant spontaneous movements. These are movements, often rhythmical, that are prevalent in infancy but have virtually disappeared in normal infants once they have achieved independent walking at around 12 months of age. Many investigators now acknowledge that these movements may be important for motor development (Parker, 1992; Thelen et al., 1991; Turvey and Fitzpatrick, 1993). However, only rexently has any serious consideration been given to understanding the origin and role of spontaneous movements in the development of motor control. 2. ORIGINS OF SPONTANEOUS MOVEMENTS 2.1 Maturational Approach The development of motor coordination is dependent on the maturation of the neural pathways that are responsible for motor control. The newborn infant has the slowest developmental rate in the animal kingdom (Goodman, 1990) with only 23% cranial capacity present at birth. The thickness of the motor cortex, for example, does not reach adult proportion until around two and a half years of age (Cratty, 1986). The stages of postural and locomotor development have been direcdy related to the maturation of the CNS. A strict maturational view explains motor development purely in terms of the physiological changes that occur in the infant, without regard to environmental or cognitive influences. Development is depicted as a set of stages where different levels of motor control are thought to be determined through maturation of the appropriate neural pathways. Hence, more complex behaviours would be the consequence of the development of higher centres of the brain. Arnold Gesell (1933) was one of the earlier researchers who believed that biological factors had a dominant role in the development of not only motor control but also cognitive functioning. Using observational analysis to determine a set of invariant sequences of development, Gesell produced his Developmental Schedules which have had a profound influence on the testing and structuring of developmental tests to the present day. Many of

Acquisition of Motor Coordination in Infants 201 the infant tests that are still popular today were developed on the basis of Geselrs original research (eg., Griffiths, 1951; Bayley, 1969). A pure maturationist viewpoint would argue that different spontaneous movements emerge as the appropriate neural pathways mature in the infant. Rhythmical spontaneous movements in infants have been likened to spontaneous motor patterns in other species, such as quadrapedal gait, flying and swimming movements (Forssberg, 1985). It has been proposed that spontaneous movements are defined by innate pattern generators which determine the basic rhythm for later locomotor activity. Forssberg (1985) argues that the innate stepping response in the young infant creates the basic rhythm for the adult plantigrade gait, although in early infancy, this innate pattern generator appears to be programmed for a digitigrade pattern. This early movement pattern is gradually transformed into the plantigrade gait through an hierarchical system that utilizes additional neural mechanisms as the infant matures. 2.2 Cognitive Approach Central control mechanisms form the basis for the contemporary cognitive theories, many of which have developed from the theory of servo-mechanisms (Craik, 1947, 1948). This information processing approach describes motor control in terms of separate processes which can be categorized as input, central, output and feedback processes. This approach again emphasises the hierarchical nature of motor development, and attributes motor coordination to the establishment of schema, representations or motor programs in the higher centres that are responsible for the control of movement (Glencross, 1977, Schmidt, 1975). Peripheral feedback mechanisms are crucial for monitoring the performance and providing environmental information. According to this view, spontaneous movements would be represented by an internal schema that would specify the appropriate motor commands. The prescriptive approach has lost some degree of popularity over recent years, as several problems have been highlighted that need to be addressed by both the maturational and cognitive theories of motor control. Bemstein (1967) argued that the many joints and linkages, with an even greater number of muscles, are too numerous to be able to provide a prescriptive solution to each movement problem (i.e., the 'degrees of freedom' problem). Furthermore, in terms of neural signals, there is an anomaly in that there are innumerably more executable solutions for any movement than there are learned examples. This is Bernstein's (1967) notion of motor redundancy. Contextual changes cannot be accounted for

202 J.P. Piek by the notion that movements are stored as a set of specific neural commands since a small variation in context calls for a variation in motor commands. In addition, innately specified motor programs cannot account for the morphological and biomechanical changes in the infant as it rapidly grows and develops. Change must be accommodated throughout the process of motor development. Schmidt's (1980) notion of a generalized motor program addressed these issues by proposing that the invariant features of the movement are stored as a basic program, or general schema, where every movement class is determined by a generalized motor program. Situation-specific parameters such as time, overall force, muscle selection and response size are mapped in at a later stage based on the current context (Shapiro and Schmidt, 1982). In order to produce a movement, two independent memory states are needed, one to select the response specifications (the recall schema) and the other to generate the expected sensory consequences so that the correctness of the response can be determined (the recognition schema). Despite this solution, however, a new theoretical paradigm emerged in the late 70's that argued against the prescriptive approach to motor control. Instead, this new dynamic systems approach suggested that movements emerge from the properties of the underlying system, rather than the hierarchically organised central control. 2.3 DynamicSystemsApproach Thelen (1985) has attributed infant spontaneous activity to the dynamic control of muscle synergies or coordinative structures. This approach, often termed the ecological approach, has its origins in both the Gibsonian ecological theory of direct perception (Gibson, 1979) and Bernstein's (1967) notion of coordinative structures. Gibson (1979) emphasised the importance of a dynamic relationship between perception and action, arguing against an animal-environment dualism. Action and perception mutually guide each other in a process that detects invariants which define the properties of events, objects and places. Bernstein (1967) proposed a solution to the degrees of freexiom problem whereby linkages are formed between different sub-systems, thus reducing the multiple degrees of freedom present in the motor system. These cooperative sub-systems have been termed muscle synergies or coordinative structures. These linkages have dynamic properties and their functioning has been likened to mass-spring systems that are intrinsically self-equilibrating.

Acquisition of Motor Coordination in Infants 203 Ulrich and Ulrich (1993) argue that there are four basic principles which are fundamental in def'ming motor development from a dynamic systems perspective. Firstly, the organism must be a non-linear, self-organizing system that produces spontaneous pattern formation through the cooperation of many heterogeneous components. It is the task or context that determines the emergence of new behaviour patterns from available subsystems rather than a set of hard-wired, predetermined commands. Subsystems that are available include skeletal, muscular, nervous, and perceptual systems, as well as levels of arousal, energy and motivation. Secondly, movement patterns can be characterized by one or more \"collective variables\". These allow a complex pattern with multiple degress of freedom to be def'med in much simpler terms. Ulrich and Ulrich (1993) described this as \"a low dimensional description of a highly complex system\" (p.449). The third principle proposes that certain movement patterns or behavioural outcomes are preferred, given a particular status of subsystem or context, therefore acting as dynamic attractors under these circumstances. Finally, shifting from one attractor state to another is determined by a change in the stability induced by the control parameter. Such self-organizing systems are found in non-biological systems. For example, a shift from ice to water to steam is the result of altering the control parameter (i.e. temperature) beyond critical levels to induce spontaneous changes in the properties of the system, resulting in the phase transitions. Although the dynamic systems approach has gained considerable support, Von Hofsten (1989) argued that this approach tends to oversimplify the problem of motor coordination, as the role of the brain is very different from that of the physical constraints. \"Physical constraints, including body parameters, define the implementation problems of actions, whereas the constraints set up by the brain are there to solve these problems.\" (p.951). Several theorists believe that the diversity of these two approaches prohibits the possibility of a unified theory (eg., Savelsbergh, 1993a), but it seems that this may be the next step in resolving this paradigm crisis. A possible compromise between prescriptive and dynamic approaches has been proposed which suggests that a dynamically driven lower level system is integrated with a cognitively organized higher level (Glencross, Whiting & Abernethy, 1994; Summers, 1992). It seems that there must be a strong prescriptive element as evidenced by the essential similarity of all human movement, and an ever-present dynamic element to accommodate the rapid development of particular motor patterns to cope with specific motor problems.

204 J.P. Piek 2.4 Neuronal Selectionist Approach Spores and Edelman (1993) have argued that \"A satisfactory understanding of early human development can only be achieved within the context of a global theory of brain function\" (p. 960). They maintain that neither the current prescriptive nor the dynamic systems approach can adequately describe the development of motor control. The prescriptive approach has the degrees of fre~om and contextual difficulties addressed earlier, whereas the dynamic systems approach has failed to take into account the specific neural mechanisms that are responsible for motor control. As a result, Spores and Edelman (1993) proposed an alternative based on 'somatic selective processes'. They argue that there is not a hard-wiring of neuronal circuits during development. Instead, there are local collectives of interconnected neurones called 'neuronal groups' with dynamic variability that give rise to a diverse output. They outline the basic steps for the development of sensorimotor coordination. Firstly, during development, there is a basic repertoire of movements produced as a result of the 'spontaneous generation' of a variety of movement. This was the first principle of the dynamic systems approach described by Ulrich and Ulrich (1993). Neural selection occurs through the process of 'adaptive value', where the organism develops the ability to sense the outcome of different movements on the environment. Adaptive value provides the constraints when selecting the movements for various global mappings. Evidence for this has been provided through synthetic neural modeling. It is proposed that successive selection will produce a stable repertoire of movements that can be modified according to the changing demands of the growing infant and the environment. In contrast, Changeux and Dehaene (1989) have suggested that there is a progressive regression of synapses as the infant develops. Like the model of Sporns and Edelman (1993), the selection process is dependent upon activity of the appropriate pathways. That is, the eventual neural pathway is dependent upon the spontaneous or evoked activity that occurs prior to selection. However, Changeux (1983) has proposed that part of the learning process appears to be the regression of the multi-innervations at birth to the selection of a single motor pathway, which would consequently reduce the likelihood of subsequent variation. Apoptosis, or programmed cell death is an important principle of neuroembryology, where selective pruning is believed to result in greater specificity (Sarnat, 1994).

Acquisition of Motor Coordination in Infants 205 Turvey and Fitzpatrick (1993) have suggested that the somatic selectionist approach is analogous to \"chaos with feedback\". The \"chaos\" is accounted for by the random procedure of generating variations in synaptic patterns whereas the \"feedback\" provides the means to determine the effective patterns (i.e., Spores and Edelmen's \"adaptive value\"). 3. CHARACTERISTICS OF SPONTANEOUS MOVEMENTS Savelsbergh (1993b) suggested that a major factor contributing to the renewed interest in infant motor development over the last decade has been the introduction of the dynamic systems approach. This approach has now been adopted by many new and established researchers in the field. An equally important factor has been the technological advances that have occurred over the last decade. Whereas the majority of research on infant motor development prior to the 1980's depended on qualitative, observational analyses, there is now a substantial use of quantitative techniques such as motion analysis, EMG, accelerometers, force plates, etc. This new technology has the potential to examine 'real-world' behaviours such as infant spontaneous movements which previously could only be examined through observation. The following section briefly reviews the types of research that have been carded out, and the techniques employed, both qualitative and quantitative, to investigate the properties of spontaneous movements. 3.1 Description and Frequency Many of the earlier studies on spontaneous movements were based on observation (Bruner, 1969; Prechtl & Nolte, 1984; Thelen, 1979). These were important in describing and defining the population of spontaneous movements found in both the foetus and the infant. Foetal motility was examined as early as 1885 by Preyer (cited in Provine, 1993), who relied on observations of pregnant women, or listened with a stethoscope for palpations of the foetal movements through the abdominal cavity, in order to measure motility. Preyer attributed foetal movements to central ozigins rather than any peripheral influences, and concluded that the same movements were produced in the newborn. The development of the ultrasound in the 1970's provided an opportunity to examine foetal movements more extensively. A comprehensive analysis was carded out by deVries, Visser and Prechtl (1982, 1985, 1988), who detected startle and general movements as early

206 J.P. Piek as 8 weeks, followed a week later by localized head and limb movements. Furthermore, it was surprising that although there is very little differentiated neural structure at this early age, the foetus produced very specific movement patterns. Prechtl (1984) provided three possible functions for these early foetal movements. Firstly, they are important for the infant's survival within the uterus, as they ensure that there are no adhesions or local stasis of circulation in the foetal skin. Secondly, they may anticipate postnatal functions, such as foetal breathing movements. Thirdly, they may have an important role in the shaping of the skeletal system, that is, producing mechanical influences on the bones and joints. Cioni and Prechtl (1988) demonstrated a natural progression of movements from the foetus to the newborn. As the sequence of occurrence of these movements was invariant across age, this was taken as further evidence for a maturational view of motor development. Prechtl (1990) argued that the systematic observations of spontaneous motor activity may be a useful tool in the neurological assessment of preterm infant. Ferrari, Cioni and Prechtl (1990) used \"general movements\", a category of spontaneous movements found in the neonate, as a means of determining the motor outcome of infants. The infants' movements were recorded on videotape and judgements made on the basis of qualitative observation. All but one of the 14 low risk infants examined were judged normal. Of the 29 brain-damaged infants assessed, all were judged as abnormal, and of these, 20 were found to have a long term disability. The movements of the brain-damaged infants were found to be less variable than those of the low-risk group, and lacked the fluency and complexity found for the normal group. Qualitative assessment of spontaneous movements in the new-born infants has proven to ~ a reliable contribution to neurological examination (Touwen, 1990; van Kranen-Mastenbroek et al., 1992). Prechtrs work focused on the neonate and infant up to around 18 weeks of age. Spontaneous movements are known to continue to around 12 months of age in normal infants. In her seminal paper, Thelen (1979) described a longitudinal study which classified 47 patterned movements termed 'rhythmical stereotypies' in 20 infants up to 12 months of age. According to Thelen's (1979) def'mition, these rhythmical spontaneous movements needed to be repeated at least three times in succession to be recorded. Thelen found that developmental profiles emerged when the movements were categorised into separate body parts or postures, and compared across age. It appeared that characteristic rhythmical patterns were the precursors to particular stages of motor development. For example, hands and knees rocking preceded crawling in infants. As a consequence of these f'mdings, Thelen

Acquisition of Motor Coordination in Infants 207 (1979) originally argued that spontaneous movements are defined by the level of maturation of the CNS, although her more recent work supports a dynamic systems approach. Thelen's original study was a major contribution to the categorizing of spontaneous movements, particularly in the stages of infancy following the neonatal period. In a recent study (Piek & Carman, 1994), Thelen's work has been extended using a cross-sectional analysis that employed videoanalysis to record the spontaneous movements of 50 infants from 2 weeks to 50 weeks of age. A cross-sectional procedure was used in order to provide a greater range of movements than that found for the 20 infants used by Thelen in her longitudinal study. In all, 53 different types of spontaneous movements were observed and classified. A less rigid definition of spontaneous movements was employed in this study which did not require the repetition of a movement at least three times in succession. The work of Prechtl et al. (1979) indicated that many of the spontaneous movements produced by young infants are single, isolated movements, and consequently it seemed important to include these in the study. As a result, a much higher percentage of movements were recorded for the young infants under 20 weeks of age as many of these were single, isolated leg kicks or ann waves as has been reported by Prechtl and colleagues. Figure 1 shows the frequency of spontaneous movements produced for each of the 50 infants across age. It can be seen from this figure that frequency is dependent on posture. As with Thelen (1979), developmental profiles for the frequency of occurrence demonstrated that the different types of movements were dependent on the posture and mobility of the infant rather than chronological age. Indeed, it seems that these spontaneous movements may well be the prerequisites for each level of motor development that requires postural changes. Equally, the pattern of spontaneous movement might be guided or constrained by the postural changes. Thelen et al. (1991) suggested that infant spontaneous movements \"form the neuromuscular bases from which skills such as reaching, sitting, and walking are built.\"(p.44). These qualitative studies provide valuable information on the nature and frequency of spontaneous movements and have lead to many speculations regarding their role in the acquisition of coordination. More recently, the characteristics of spontaneous movements have been quantified using a variety of approaches and techniques, in the hope that this will contribute to our understanding of their function.

208 J.P. Piek 1400. O mobile r a sitting 1200' A prone L-- } s~ne o I000 @ 0 800, 0 r rr r G 600 r @ .d @ @ r 0 0& *g, 0 400 0 Ca, m 200 0 = o [.0.. o oo 45 50 O~ 20 25 30 35 40 0 5 I0 15 AGE (Weeks) Figure 1. Totalnumberof spontaneousmovementsproducedper hour for each infant plotted as a functionof age. (Reprintedby permission of ElsevierScience Publishers Ireland Ltd, from Piek, J.P. & Carman, R. Developmentalprofilesof spontaneousmovementsin infants. EarlyHumanDevelopment, 1994). 3.2 Temporal Parameters An important feature that has emerged from the qualitative examination of spontaneous movements is the degree of rhythmicity in many of these movements. Rhythmical motor patterns have been identified in humans for many years. Wolff (1967, 1968) termed these types of movements 'stereotypic mannerisms' when investigating abnormal infants with Down's Syndrome, psychomotor retardation, blindness or schizophrenia. Rhythmical movements were found to persist in children with these conditions (Kravitz & Boehm, 1971; Wolff, 1968), and it was thought that they may be of value in the earlier diagnosis of such conditions. The rhythmicity of these types of movements was of particular interest to Wolff (1968) who suggested that they may be important in the development of complex temporal sequences as a result of phase-locking or interaction of several patterns. Rhythmicity has also been examined by Robertson (1990) in foetal movement patterns using spectral analysis. Cyclic motility was found to occur around midgestation by Robertson (1985) at a rate of approximately 1 to 4 minutes. The temporal characteristics of cyclic motor activity did not appear to show any changes during the prenatal period, and were also

Acquisition of Motor Coordination in Infants 209 maintained in the first two months. Robertson (1993) conducted a longitudinal study on newborns from 1 month to 4 months of age, examining the rate, strength, and irregularity of cyclic motility. Although the rhythmicity persisted after 2 months of age, there were significant changes in its characteristics. The rate of oscillation did not differ from prenatal to neonatal to post-neonatal age, but at 2 months of age there was an abrupt decline in the relative strength of these oscillations with a corresponding increase in the complexity of the temporal organisation. Robertson (1993) argued that at two months there was a new source of variation introduced that had a shorter time constant than earlier motility, and speculated that this may be a result of the coupling of movement and attention leading to greater exploratory powers for the infant to examine the environment. Hopkins and Prechtl (1984) and Prechtl and Hopkins (1986) have also identified changes in the nature of spontaneous movements in infants at around two to three months of age. The earlier gross movements that vary in speed, intensity and force (a slow 'writhing' quality) changed to small, elegant movements that involve primarily the limbs and head (a 'fidgety' quality). Consequently, they suggested that this was the emergence of the first transitional stage in postnatal neurological development. It was suggested that the changes observed in the nature of the spontaneous movements may have been the result of a decrease in the co- contraction of antagonist muscle groups with a corresonding increase in reciprocal activation (Hadders-Algra, Van Eykem, den Nieuwendijk & Prechtl, 1992). Using EMG correlates to test this hypothesis, Hadders-Algra et al. (1992) found that although co-activation of the antagonists was present for more than 70% of the time for both 'writhing' and 'fidgety' stages, there were substantial changes in the nature of the EMG activity. The phasic muscle activity became shorter, the amplitude was attenuated, and there was a reduction in the tonic background activity. It was suggested that, as a result of spinal and supraspinal reorganisation, there was a reduction in the motor unit sensitivity that would account for the observed changes in EMG activity. Hadders-Algra and Prechtl (1992) found a further developmental change in the third month with the occurrence of very rapid arm movements called 'swipes' or 'swats', and suggested that these may emerge as a result of the functional development of the supraspinal structures such as the basal ganglia, cerebellum and cerebral cortex. EMG recordings have also been used by Thelen and Fisher (1983a) to examine spontaneous leg kicks in infants 2 to 4 weeks of age. As a result, Thelen and Fisher (1983a) found that neither the braking of flexion nor the initiation of the extensor phase seemed to be

210 J.P. Piek controlled by active muscle intervention in the 2 and 4 week old infants. They suggested that the flexion phase is well developed virtually from birth compared with the extension phase. Leg extension appeared to be quite slow compared to the flexion phase and relied initially on passive forces for the extension. The relative invariance of the timing of flexion and extension in infant kicking was pointed out by Thelen and Fisher (1983b). Changes in the frequency and vigour of kicking were associated with changes in the level of arousal and context of the kicking. However, neither the timing of the leg flexion nor of leg extension appeared to differ when the context was varied. For example, no differences were found for these measures when infants were in an active, moving state compared to a state of crying (rhelen, Bradshaw & Ward, 1981), nor was the timing affected when infant kicking was reinforced by attaching the leg to a mobile (Tbelen & Fisher, 1983b). In a comparison of the two legs during alternating kicking, Thelen, Ridley-Johnson and Fisher (1983) found that the two legs appeared to form one coordinative structure. The interlimb latencies decreased between the ages of 2 and 26 weeks suggesting that this coupling of the limbs \"tightens\" with maturation. The relationship between the timing of flexion and extension was also examined in a cross-sectional study of infants from 2 to 26 weeks of age (Piek& Carman, 1994). Frame-by frame videoanalysis was used to measure the timing of the flexion and extension of spontaneous leg kicks and the lift and return of spontaneous ann waves. Developmental profiles for 24 infants up to 6 months of age were produced for the spontaneous leg kicks and arm waves. There was a significant negative correlation between age and the mean time taken for both leg flexion and leg extension, and for the arm lifts and returns. The extension phase was significantly longer than the flexion phase, and there was a significant difference in the variability of the leg flexion and extension. Overall, the extension or return phases of the movements were found to be slower than the flexion or lift phases, particularly for the younger infants. In the very young infants, it was observed that in many cases the limb appeared to rely on passive forces such as gravity to return to its original position. This supported the EMG findings of Thelen and Fisher (1983a) , and would account for the greater variance found in the extension or return phases of the movement as these movements are dependent to a greater degree on external forces acting upon them. As the internal active forces become stronger, the movements become faster and less variable. This could also be associated with an increase in co-contraction as observed by Hadders-Algra et al. (1992).

Acquisition of Motor Coordination in Infants 211 The studies described in this section have utilised many different techniques to examine the temporal parameters of spontaneous movements, and have been guided by quite different theoretical perspectives. Whereas Prechfl's research, and Thelen's earlier study, have focused on the maturational viewpoint, studies by Robertson and more recent research by Thelen and colleagues have argued that their fmdings support a dynamic systems approach. The importance of understanding the temporal relationships is undeniable since their role in other areas of motor control has been extensively pursued. 3.3 Joint Angle Relationships One of the most exciting techniques that has been introduced to study spontaneous movements in recent years is that of motion analysis. This produces time series data for the joint angles examined which can then be analyzed to provide information on joint angle displacement, velocity and acceleration curves, and other dynamic characteristics of the movements observed. 3.3.1. Intralimb Coordination. Latash and Latash (1994) have recently acknowledged the notable contribution that Esther Thelen has made in the area of motor development based on the ideas of Bemstein. Using 3-D limb kinematics, Thelen and colleagues (Schneider, Zemicke, Ulrich, Jensen & Thelen, 1990; Thelen, 1985; Thelen et al., 1991) have examined quite extensively the different contributions of muscular, passive and gravitational torques on the ankle, knee and hip joints when infants are producing spontaneous leg kicks. Comparison of ankle, knee and hip joints by Thelen and Fisher (1983a) in two and four week old infants suggested that the leg acts synergistically during a leg kick as a result of the self-organization of active and passive forces. Thelen et al. (1991) argued that the dynamic and complex interplay of both active and passive forces that has been found implies that these movements could not be driven by a program that specified exact patterns of muscle activation. Electromyographic evidence of intralimb rigidity obtained by Thelen and Fisher (1983a) for leg kicks and by Hadders-Algra and Prechtl (1993) for ann movements was described earlier. Kinematic evidence has been provided by Thelen and Fisher (1983a) in their examination of the joint angle changes that occurred between the hip, knee and ankle in spontaneous leg kicks in 2 and 4 week old infants. Pair-wise cross correlations were used

212 J.P. Piek to determine the degree of synchrony between the joints of the limb. The findings suggested that the three joints were indeed tightly linked, especially for the 4 week olds. As a result, they argued that interjoint coordination was highly structured in the newborn infant. In addition, Thelen (1985) argued that from about 2 to 6 months of age there was a period of \"apparent disorganization as the joint action became individualized from the mass activity of the newborn\" (p. 10). From 5 to 8 months, there was another period of fight interjoint coupling involving larger functional groups. Cross-correlations can be evaluated to examine the degree of coupling between joints. Not only does this allow the joint angle time series to be examined at any given time, but it can also determine any phase lags present between different joints. That is, are the highest correlations produced when the time series are in-phase (i.e. not lagged), or can even stronger relationships be detected when one series is lagged in relation to the other? I am currently addressing this problem in a longitudinal study that is quantifying the characteristics of spontaneous movements in infants from birth to one year of age. Infants' movements are recorded using a MacReflex Motion Analysis System, and joint angle time series are determined for the ankle, knee, hip, shoulder and elbow for both the fight and left side of the infant's body. Examples of infant spontaneous leg kicking in supine position are presented in Figure 2. These time series graphs demonstrate typical leg kicks for a normal full-term infant at 4, 8 12 and 16 weeks of age. It can be seen that at 4 weeks, the infant produces minimal movement with the left leg and a strong single kick with the fight leg. Isolated leg kicks are one of the most prevalent spontaneous movements found at this age (Cioni & Prechtl, 1990; Piek& Carman, 1994). They typically have a rapid flexion phase followed by a slower extension phase (Piek& Carman, 1994; Thelen & Fisher, 1983a), which can be seen for the fight leg kick presented. The extension phase is believed to rely more on passive forces such as gravitation to return the leg to the floor. By 8 weeks of age, regular patterned leg kicks were initiated. In the example presented in Figure 2B, the two large kicks for the left leg are qualitatively similar, and the extension phase appears to be similar to the flexion phase. At this stage the kicking is not continuous as the two kicks are separated by a small resting phase. Although there is some evidence of movement for the fight leg, rhythmical kicking does not appear to be present. At 12 weeks of age (Figure 2C), the infant produced continuous, rhythmical kicking for both legs. The kicks appear to be regular and stereotyped with tittle interkick pause. Continuous

Acquisition of Motor Coordination in Infants 213 kicks were also present for both legs at 16 weeks of age (Figure 2D), but these were less rhythmical and stereotyped than those produced at 12 weeks of age. The relationship between the hip, knee and ankle for each limb can be determined by using cross correlation functions (CCF's). As the movements were recorded at 25hz, the relationship between pairs of angles could be examined out of phase by 40 ms shifts to the left or right for a range of 280 ms in each direction. One of the basic assumptions of the cross-correlation procedure ks that the time series are stationary, that is, the mean and variance of each series stay approximately the same over the length of the series (i.e., they do not drift). In order to ensure that the series is stationary, the usual procedure is to difference the series (SPSS for Windows, 1993). The original series is replaced by taking the differences between adjacent values in the original series. Table 1 is a summary of the findings for the time series data presented in Figure 2. Intralimb coordination is determined by examining the pair-wise cross correlations found between joints. At 4 weeks of age there was a strong synchronous relationship between the hip, knee and ankle. For the left leg, the joints were in-phase except for the hip and ankle, where the ankle led the hip by about 120ms. The joints for the fight leg were strongly correlated but not in-phase, as the hip lagged the knee by 120 ms, the ankle led the knee by 40 ms, and the ankle lagged the hip by around 80 ms. At 8 weeks, the joints of the left leg produced a fight synergy with the knee lagging the other joints by around 40 ms. A detailed analysis of 5 infants at 7 weeks (Piek, submitted) confirmed that there was a rigid, synchronous relationship between the leg joints at this age. In the example presented in Figure 2, however, the left leg did not demonstrate this synergy. Although the highest cross-correlations occurred at 0 lag suggesting that the joints were in-phase, the correlations were not significant. An important characteristic that is emerging from the longitudinal data is the different developmental patterns for the contralateral limbs. It cannot be assumed that the limbs develop at the same rate. By 12 weeks of age both legs demonstrate strong interjoint relationships which were also evident at 16 weeks of age. This appears to be contrary to Thelen's findings that there is a disassociation of the joints from around 8 weeks of age. Indeed, as pointed out in Section 3.2, Hadders-Algra et al (1992) found that coactivation of the antagonists was still present for at least 70% of the time for arm movements in infants 2 to 3 months of age. The longitudinal data presented here indicates that even at 16 weeks of age the joints are still tightly linked although not necessarily in-phase.

Left Side A. Aav e 4 w e e k 9 --I hip i I 9 -I knee ! 190 I I ankle i i # 49 9 170 ! 4444444444444444#444e444444444#4444#4444 \"~ ...... .. (U I k-- (~ 150 l r9-3 1 3 0 L -,,,'\" \",9 r~ v mmmmr9 t- l ,,,,,,,,,,,~ , .~- ;_ < (U 110 milUlllilgm nilllmlTinnlnmm ill Trr . , nifilllilllilll[i <:~ 90 - 70 ......................... -- ~ ~. ,,o ~o c~ ~ ~ ,,.o. o ~ ~ ,,o co e,i ~ c~ c5 o o .... ~ Time (Seconds) B. Age 8 wee 210 r i iX) 17o,L,,,,,,,,,........ , _.,,,,, ,-i........ 9 9 OL_ 9 LagL \" 9 a~ ~) '-',A i \" \".- -L'~- 9 ~ 9 J ~9\" ~ r , F2\"F'F'W-Pe- FW~- -- -%~vpIq.\" ~A.. O Tj.~J AL \"'~ c- t- 110 < < 90 o ~ ~ ,,o ~o ,-- ~ ~ ~o ~o e~ ~ ~. ,,q. ~. 0 0 0 C~ ~-- ~- ~- ~- C~ r C~ eq Time (Seconds)

Right Side ks [_--~r- hip knee 210 . ankle ] 190 170 150 130 11o 90 I 7 0 L. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . o ~ ~ ~o ~ ,.-.- ~ ~ ~o ~0 ~ e~ ~. ~. ~ t'Q 0,,I C,,I 0 0 0 C~ ,~-- ,,- ~,- '~- Time (Seconds) eks 190 r 150 ~ I|11 130 110 - 90 o ~ ,r co co ,-- ~ ,r co Qo ~ e~ ~. ,,q. C~ 0 0 0 ~-- ~ ~-- ~ r r C',,,I e',,i Time (Seconds)

C. Age 12 w ~\" 15o \"\"'\"** t .~ ~m ++ r'1-1-3 130 /! ,;~ ~3b 9 i' --.-- hip Q.) a 11o .] 9+ r L l -\" --m-- knee --'~En 90 --,-- ankle ~' -- / r ~ CO cO . ,., - - < 50 30 (\",,I \"~ CO <:30 ~-- C',,l ~ r,O 00 r 0 o o o ,~ .... Time (Seconds) D. Age 16 we 18o =&LrILII 9m ~ ~ 170 =- .411.I / LI~ llrltltltnltlLltlllllL ~ 160 .i \" +--.._ +:1~,,: 82 9 _.411 9 QL) 1 5 0 va . ... m.+ [ , \" , ; , , , : , , ; a , , ' , ' , + , : + ! . 140 -1- 3) r'hm 130 ~ % <I:E _ 110 ~~, ; , ,%\" ,+.-. . . . . . .-. -++::%. . ...++\" \\kt + ~~,\"*/ \"+~ ~ <I: 90 t -'\" 80 0 ~ ~ ~ ~ ~- ~ ~ ~ CO ~ ~ ~ ~ cO oooo ,- ,- ,-: ,- ~ ~i o,i ~ Time (Seconds) Figure 2. Three second time series samples of spontaneous leg kickin

eeks 170 ~3o 110 -- n[ ~ hip :;~ --~-- ankle =: 70 ~: ~\" 50 ,% 30 QO 0,I CO ~ cO CO t~l CO '~\" CO ~ ~ CO ~ '~ CO (\",,I CO ~\" 0 o ~ \"+ o=+ ``~ ,=~'~- . .'~ . . .9 . = ~o ,.0, ,~ . ,~.., .,.. . r,J Time (Seconds) eeks 180 1~o 160 150 140 130 ) ~1o 100 90 80 0 ~ ~ ~ ~ CO ~ ~ ~ ~ ~ ~ ~ ~ ~ ~~~ o ~ ~ ~. d ~d ,--\" ~,- ,~-- 9 ~,- ~,- o~ e,i ~ ,~~ ,e,,,.i ~c,0,i o o Time (Seconds) ng for a full-term infant at ages 4, 8, 12, and 16 weeks of age.

Table 1. Cross correlations (CC) and phase-lags (lag) for the highest correlati presented inFigure 2 (* p<.O1). Hip / Knee Ankle / Left Leg ..... Age Left Leg Right Leg CC Lag (Weeks) CC Lag CC Lag 0.76* 0 0.63* -3 0.57* 0 0.82* -1 0.37 0 0.88* -1 12 0.73* 0 0.79* 0 0.50* 0 16 0.82* -1 0.42* 0 0.83* 0

ions produced between the hip, knee and ankle time series data / Knee Ankle / Hip Right Leg Left Leg Right Leg CC Lag CC Lag CC Lag 0.57* 1 0.57* 3 0.58* -2 0.25 0 0.79* 0 0.32 0 0.56* 1 0.37 1 0.60* 3 0.70* 0 0.76* 1 0.66* 2

Acquisition of Motor Coordination in Infants 217 Preliminary fmdings indicate that when the leg kicks become more rhythmical, and produce sequential leg kicks, this rigidity is particularly evident. Infants develop strong, regular sequential kicking by their 10th week, and this continues until the infant becomes mobile (Piek & Carman, 1994). Although co-contraction is associated with the earlier spontaneous movements, there has been some indication that there is a persistence of joint synchrony within each limb in the older infants when performing rhythmical spontaneous movements. It is evident that these rigid couplings are distinct from the type of synergies produced for more coordinated control, where more complex joint-angle relationships would necessitate out-of phase relationships between joints. 3.3.2. Interlimb Coordination. Before infants can produce coordinated movements such as walking, running, swimming, etc., they need to learn how to coordinate their limbs to work together. This coordination develops over the first year. Interlimb coordination has received a great deal of attention over recent years (e.g., Swinnen et al., 1994), but very little research has been carded out in the area of infant motor development. Corbetta and Thelen (1994) pointed out that there has been extensive investigations of one-handed reaching in infants, but bimanual reaching plays an important role in the infant's development of reaching. Consequently, the investigation of bimanual coordination addresses issues such as, why there is an initial coupling of limbs before single limb coordination is developed, and how the infant's perceptual motor abilities change with this shift (Corbetta & Thelen, 1994). An increase in the coupling of the legs was demonstrated by Thelen, Ridley-Johnson and Fisher (1983) in infants between the ages of 2 and 26 weeks. Furthermore, Piek and Carman (1994) found that simultaneous kicking of both legs was the most frequent spontaneous movement in infants between 20 and 30 weeks of age. In order to understand the development of infant coordination, it is important to determine the interlimb relationships that are present in the young infant and how they change throughout development. The relationship between the left and fight legs for the kicks presented in Figure 2 was determined by comparing the time series data for the left and fight hip, knee and ankle (see Table 2). Synchronous or alternating kicks would be evidenced by a positive or negative correlation, respectively, with a 0 lag. Significant cross correlations occur between the two legs at all ages. At 4 weeks of age, there was a significant negative correlation between the

218 J.P. Piek two knees. This is neither synchronous nor alternating kicking, but rather implies that there is some relationship developing between the two limbs. At 8 weeks of age, both the hips and knees were significantly correlated, with the fight limb leading the left by 80 to 120ms. By 12 weeks of age, there is evidence of strong synchronous kicking for the two legs as both the hip and knee had significant correlations at 0 lag. Even higher correlations were found at 16 weeks, but the limbs are once again out of phase, this time with the left leg leading the fight. Table 2. CCF's for the pair-wise comparisons of left and fight kicks for the hip, knee and ankle time series given in Figure 2 (* p<.01). Age Left/Right Hip Left/Right Knee Left/Right Ankle (Weeks) CC Lag CC Lag CC Lag -0.26 -2 -0.53* -4 0.26 -4 0.50* -2 0.42* -3 -0.39 -6 12 0.57* 0 0.63* 0 0.34 0 16 0.79* 3 0.50* 4 0.44* 2 Thelen (1985) argued that interlimb coordination followed a similar pattern to intralimb coordination which implies a synchronous rather than sequential development of intralimb and interlimb coordination. Her findings suggested that there was a newborn synergy between the legs producing alternating kicking in the first month. This was followed by a period a asymmetry from 1 to 5 months where infants primarily kicked with one leg. Between 4 and 6 months, strong synchronous kicking appeared. WhitaU and Clark (1994) suggested that Thelen examined intralimb coordination in a more quantitative fashion than interlimb coordination. Certainly, earlier animal studies such as Bekoff (1981) have suggested that \"intrajoint coordination tends to precede interlimb coordination\" (Whitall &

Acquisition of Motor Coordination in Infants 219 Clark, 1994, p.404). Indeed, the current findings suggest that interlimb coordination may develop in a sequential rather than a synchronous fashion. Finally, the relationship between ipsilateral and contralateral limbs can also be investigated using CCF's. In Figure 3 the time series data for the right elbow has been added to the data presented in Figure 2 for the left leg kick at 16 weeks of age. The right elbow was significantly correlated (p< .01) with the left hip (r = .47) at a lag of 2, the left knee (r = .49) at a lag of 1 and the left ankle (r = .54) at 0 lag, suggesting the formation of more complex synergies as the infant develops. There was also a significant correlation between the fight elbow and fight hip (r = .54) at a lag 4. The progressive coupling of all body segments as the infant matures is another characteristic of spontaneous movements that requires further investigation. 170 - ~ - - r elbow \"i, -- 9 ..... I hip ,oo ..... 9 .... I k n e e - 9 - - I ankle ,A 9 9 ~A 9149149 :~ :~ -'~ :~'~ ~ \\ \\. / .J'~ :~\" ~'l~ r o - 70 • t ,q-+-t~F+@+-k-.F k-Fq-~-l--k++--l~l--l-k-~--i-4-]--l--l-+-l~-l~--l--4--L-+-§ IIIf - ~ 5 + @ ~ ~ - ~ O e,,I ~t ~D CO ~ r ~\" ~ CO ~ 04 ~1\" ~ tO ~0 r ~ \" tO ~O ~1\" r ~f r,D ~0 tO \" ~ ,-- ~ r ~1\" O ,r',,. ~ . ~ O 9r ~~ O , - ~ t ~ ~ . ~ ~ ~0. . ~ ~ e,i dddo ood. . . . . ~ ~ ~i TIME Figure 3. Time-series data for a full-term infant at 16 weeks of age demonstrating the relationship between the right elbow and left leg. 3.3.3. Investigations of Preterm Infants. Largo (1993) has recently pointed out that very few studies have been carded out to determine how motor development is influenced by prematurity in infants. Many of the behavioural differences found in the preterm infant are discussed in terms of neurological immaturity. Yet, an underlying difference in the development of these infants could be attributed to the environmental influences on these infants that are vastly different to the normal full-term infant which is protected within the

220 J.P. Piek uterus for the full 40 weeks. What effect does the change in environment have on an organism that is not at the level of neural maturation appropriate for the external influences? Although there have been numerous studies carded out investigating gross and free motor control in pre-term infants using such measures as gross motor milestones (Allen & Alexander, 1990), the Neuromotor Behavioral Inventory (Gorga, Stem & Ross, 1985), a combination of the Peabody Developmental Motor Scales and the Griffiths Mental Development Scales (Piper, Byrne, Darrah and Watt, 1989) and the Bayley Scales of Infant Development (Barrera, Rosenbaum & Cunningham, 1987; Valvano & DeGangi, 1986), very little research has been conducted to examine any kinematic differences, or changes in the interlimb or intersegmental coordination. These differences may not be detected using the observational assessment of the above tasks. In one such study, Katz et al. (1991) examined hip motion in pre-term infants using a goniometer. This study was designed to provide norms for hip motion for infants with different gestational ages. Heriza (1988a, b) compared spontaneous leg kicks in low-risk preterm infants and full- term infants at 40 weeks post-gestational age. The relationships between the hip, knee and ankle joints were examined using pair-wise correlations and phase lags in the key kinematic events, in addition to the timing of flexion and extension. It was argued that the extrauterine environmental events did not influence the highly organized synergies found for the spontaneous leg kicks. However, when Heriza (1991) examined the individual profiles of high and low risk preterm infants in a longitudinal study, some differences were evident, although not consistent between infants. One would expect a great deal of variability in this population of infants as has been exemplified when examining the aetiology of later developmental disorders in preterm infants. Still, it does suggest that these measures may provide some insight into the development of motor coordination in this population of infants. Hadders-Algra and Prechtl (1993) carded out a longitudinal study on two preterm infants with gestational ages between 28 and 33 weeks, comparing the EMG activity of the biceps and triceps brachii during general movements at ages 33-34 weeks gestational age, then at term age, and finally at 3 months post-term. They found characteristic differences in the spontaneous movements of preterm infants compared with the normal full-term infants. Whereas co-activation of the antagonists was found to be present in the full-term neonates

Acquisition of Motor Coordination in Infants 221 for more than 70% of the time, the two preterm infants examined demonstrated an increase in co-activation with increasing age. Co-activation had reached the level found in the fullterm infants by the post-term age of 3 months. Such findings suggest a need for further investigation of both low- and high-risk pre-term infants to determine the different temporal and spatial relationships that may be present in the spontaneous movements of these infants, especially as there has been research to indicate that pre-terms at school age have problems with motor coordination and control (Lukeman & Melvin, 1993). 4. CONCLUSION In a recent paper on rhyflu~icity of movements in children, Parker (1992) reviewed the work of some of the earlier researchers who have examined infant spontaneous movements. Many of these investigators argued that spontaneous movements are important as precursors for more complex motor development (Bruner, 1969; Thelen, 1979). The rhythmicity was noted as a crucial aspect of the movements and it was suggested that one of the main roles of spontaneous movements was its contribution to the later development of serial ordering for both cognitive and motor behaviour. (Wolff, 1967, 1968). In determining the importance of spontaneous movements in the development of motor coordination, it is necessary to take into account all the research, both qualitative and quantitative, as each provides a different perspective on the nature of these movements. A consistent finding that is evident across qualitative, physiological and kinematic data is the rigid, synergistic relationships that appear in spontaneous movements. Von Hofsten (1989) described such a relationship between the ann and hand in newborn infants. He suggested that the neuromotor structures are undifferentiated in the neonate, resulting in the infant's arm and hand flexing and extending in a synergy when the ann is extended. It appears that the arm movement locks the hands into a rigid coupling, despite the fact that the infant is capable of independent hand movements when not influenced by the ann extension. Bemstein (1967) suggested that the formation of rigid couplings in the early stages of skill acquisition can initially reduce the degrees of freedom. Vereijken, van Emmerik, Whiting and Newell (1992) called this concept \"free(z)ing the degrees of freedom\". In a study involving the acquisition of a slalom-like ski movement in adults, they argued that joints were rigidly fixed in the initial stages of learning, consequently reducing the number of degrees of

222 J.P. Piek freexlom that the subject was required to control. The tight couplings were indicated by relatively high cross-correlations between joints. As learning progressed, there was a release of this rigid coupling as the joints became increasingly independent of each other. There are important qualitative differences between the rigid couplings found early in learning, and the coordinative structures that develop with training. In the former, the joints appear to be held in a rigid manner that produce restricted movement of the joints. Coordinative structures, on the other hand, involve a dynamic, functional relationship between the appropriate joints. Turvey and Fitzpatrick (1993) suggested that spontaneous movements allow the exploration of the intrinsic dynamics of the limbs, providing a means of determining the stable patterns that constitute appropriate muscle synergies. It appears that appropriate pattern selection emerges as a result of the fight couplings that occur in spontaneous movements. By forming these rigid couplings, spontaneous movements reduce the degrees of freedom and provide, as Bruner (1969) and Fentress (1976) have pointed out, movements that require less processing capacity for control. Consequently, this .provides a medium that allows the system to learn the complex relationships associated with the control of coordinated voluntary movements without excessive \"noise\" from unnecessary movements. As a result, the infant learns appropriate relationships, and the influence of not only joint and muscular relationships, but other factors such as postural control, and the influence of cognitive and affective systems. It is not a rigid set of muscle commands that are determined through this procedure, but rather, the \"operational principles\" (Lashley, 1951) that are determined. By constraining the degrees of f~.xtom through spontaneous movements, the infant can develop and refine the appropriate relationships needed for normal motor control. Rigid co- contractions permit motor learning without the interference of complex, uncoordinated motor output. As the new dynamics are developed and integrated into voluntary motor control, there is a break-down of the fight intralimb and interlimb synergies. The neuronal theories of Spores and Edelman (1993) and Changeux (1983) have pointed out the importance of repetition in the development of appropriate neural pathways. Rhythmical spontaneous movements are both spatially and temporally repetitious, thus facilitating the selection process outlined by these theories. RePetitive innervation of a particular pathway has been linked to the preference of particular neural pathways in the maturational process. Sporns and Edelman (1993) argue that through successive selection, the infant will build up a stable repertoire of movements.

Acquisition of Motor Coordinationin Infants 223 In conclusion then, it would not only be chronological age or neural maturation that is the determining factor of motor development, but equally important are the opportunities that the infant has to perform the movements that are essential for this learning process. These early processes would therefore be crucial for the development of normal motor control. ACKNOWLEDGMENTS I wish to thank Peter and David Livesey for their valuable advice, Nick O'Dwyer for his helpful comments on this paper, and Natalie Gasson for her assistance in data collection and analyses. This research was supported by the NH&MRC PHRDC of Australia. REFERENCES Allen, M.C., & Alexander, G.R. (1990). Gross motor milestones in preterm infants: Correction for degree of prematurity. The Journal of Pediatrics, 116, 955-959. Barrera, M., Rosenbaum, P., & Cunningham, C. (1987). Corrected and uncorrected bayley scores: Longitudinal developmental patterns in low and high birth weight preterm infants. Infant Behavior and Development, 10, 337-346. Basmajian, J.V. (1977). Motor learning and control: A working hypothesis. Archives of Physical Medicine & Rehabilitation, 58, 38-41. Bayley, N. (1969). Manual for the Bayley Scales of Infant Development. New York: Psychological Corporation. Bekoff, A. (1981). Embryonic development of the neural circuitry underlying motor coordination. In W.M. Cowan (Ed.), Studies in developmental neurobiology (pp. 134- 170). New York: Oxford Uni Press. Bemstein, N. (1967). The co-ordination and regulation of movements. Sydney: Pergamon Press. Bruner, J.S. (1969). On voluntary action and its hierarchical structure. In A. Koesfler & J.R. Smythies (Eds.), Beyond reductionism: New perspectives in the life sciences. London: Hutchinson & Co. Changeux, J.P. (1983). Concluding remarks on the \"singularity\" of nerve cells and its ontogenesis. Progress in Brain Research, 58, 465-478. Changeux, J.P., & Dehaene, S. (1989). Neuronal models of cognitive functions. Cognition, 33, 63-109.

224 J.P. Piek Cioni, G., & Prechtl, H.F.R. (1988). Development of posture and motility in preterm infants. In C. von Euler, H. Forssberg & H. Lagercrantz (Eds), Neurobiology of Early Infant Behaviour (pp.69-77). Stockholm: Stockton Press. Cioni, G., & Prechtl, H.F.R. (1990). Preterm and early postterm motor behaviour in low-risk premature infants. EarlyHuman Development, 23, 159-191. Corbetta, D., & Thelen, E. (1994). Shifting patterns of interlimb coordination in infants' reaching: A case study. In S. Swinnen, H. Heuer, J. Massion & P. Casaer (Eds.), Interlimb coordination:Neural dynamical and cognitive constraints (pp.413-438). San Diego: Academic Press. Craik, K.J.W. (1947). Theory of the human operator in control systems. I. The operator as an engineering system. British Journal of Psychology, 38, 56-61. Craik, K.J.W. (1948). Theory of the human operator in control systems. II. Man as an element on a control system. British Journal of Psychology, 38, 142-148. Cratty, B.J. (1986). Perceptual & Motor Development in Infants & Children (3rd Ed). NJ: Prentice Hall. de Vries, J.I.P., Visser, G.H.A., & Prechtl, H.F.R. (1982). The emergence of fetal behaviour. I. Qualitative aspects. EarlyHuman Development, 7, 301-322. de Vries, J.I.P., Visser, G.H.A., & Prechtl, H.F.R. (1985). The emergence of fetal behaviour. 1I. Quantitative aspects. EarlyHuman Development, 12, 99-120. de Vries, J.I.P., Visser, G.H.A., & Prechtl, H.F.R. (1982). The emergence of fetal behaviour. III. Individual differences and consistencies. EarlyHuman Development, 16, 85-103. Fentress, J.C., (1976). Dynamic boundaries or patterned behaviour: Interactions and self- organization. In P.P.G. Bateson & R.A. Hindle (Eds.), Growing Points in Ethology (pp. 135-170). Cambridge: Cambridge University Press. Ferrari, F., Cioni, G., & Prechd, H.F.R. (1990). Qualitative changes of general movements in preterm infants with brain lesions. EarlyHuman Development, 23, 193-231. Forrsberg, H. (1985). Ontogeny of human locomotor control I. Infant stepping, supported locomotion and transition to independent locomotion. Experimental Brain Research, 57, 480-493. Gatev, V. (1972). Role of inhibition on the development of motor coordination in early childhood. DevelopmentalMedicine and ChildNeurology, 14, 336-341.

Acquisitionof Motor Coordinationin Infants 225 GeseU, A. (1933). Maturation and the patterning of behavior. In C. Murchison (Ed.), A handbook of child psychology (2nd Ed. pp. 209-235). Worcester, MA: Clark University Press. Gibson, J.J. (1979). The ecological approach to visual perception. Boston: Houghton- Mifflin. Glencross, D.J. (1977). Control of skilled movements. PsychologicalBulletin, 84, 14-29. Glencross, D.J., Whiting, H.T.A, & Abernethy, B. (1994). Motor control, motor learning and the acquisition of skill: Historical trends and future directions. International Journal of Sport Psychology, 25, 32-52. Goodman, J.F. (1990). Infant intelligence: Do we, can we, should we assess it? In C.R. Reynolds & R.W. Kamphaus (Eds), Handbook of Psychological and Educational Assessment of Children: Intelligence and Achievement (pp. 183-208). New York: Guilford Press. Gorga, D., Stem, F.M., & Ross, G. (1985). Trends in neuromotor behavior of preterm and fuUterm infants in the first year of life: A preliminary report. Developmental Medicine and ChildNeurology, 27, 756-766. Griffiths, R. (1951). The Griffiths Mental DevelopmentScalefor Testing Babiesfrom Birth to Two Years. London: Child Development Research Centre. Hadders-Algra, M., & Prechtl, H.F.R. (1992). Developmental course of general movements in early infancy. I. Descriptive analysis of change in form. Early Human Development, 28, 201-213. Hadders-Algra, M. & Prechtl, H.R.R. (1993). EMG correlates of general movements in healthy preterm infants. Journal of Physiology, 459, 330P. Hadders-Algra, M., Van Eykem, L.A., Klip-Van den Nieuwendijk, A.W.J., & Prechtl, H.F.R. (1992). Developmental course of general movements in early infancy. II. EMG correlates. EarlyHuman Development, 28, 231-251. Heriza, C. B. (1988). Organization of leg movements in preterm infants. Physical Therapy, 68, 1340-1346. Heriza, C.B. (1988). Comparison of leg movements in preterm infants at term with healthy fuU-term infants,Physical Therapy, 68, 1687-1693. Heriza, C.B. (1991). Implications of a dynamical systems approach to understanding infant kicking behavior. Physical Therapy, 71,222-235.

226 J.P. Piek Hopkins, B. & Prechtl, H.F.R. (1984). A qualitative approach to the development of movements during early infancy. In H.F.R. Prechtl (Ed.), Continuity of Neural Functions from Prenatal to Posmatal Life. Clinics in Developmental Medicine No 94 (pp. 179- 197). Oxford: Blackwell Scientific. Jensen, J.L., Thelen, E., & Ulrich, B.D. (1989). Constraints on multi-joint movements: From the spontaneity of infancy to the skill of adults. Human Movement Science, 8, 393-402. Katz, K., Davidson, S., Dulitzky, F., Naor, N., Wieluasky, E., & Merlob, P. (1991). Normal ranges of hip motion of preterm infants. Developmental Medicine and Child Neurology, 33, 993-997. Kelso, J.A.S., & TuUer, M.T. (1984). A dynamical basis for action systems. In M. Gassaniga (Ed.) Handbook of Cognitive Neuroscience (pp. 321-356), New York: Plenum Publishing Corp. Kravitz, H., & Boehm, J.J. (1971). Rhythmic habit patterns in infancy: Their sequence, age of onset, and frequency. Child Development, 42, 399-413. Largo, R.H. (1993). Early motor development in preterm children. In G.J.P. Savelsbergh (Ed.), The Development of Coordination in Infancy (pp. 425-444). Amsterdam: Elsevier Science. Lashley, K.S. (1951). The problem of serial order in behavior. In L.A. Jeffress (Ed.), Cerebral Mechanisms in Behavior (pp.112-136). New York: John Wiley & Sons. Latash, L.P., & Latash, M.L. (1994). A new book by N.A. Bemstein: \"On Dexterity and Its Development\". Journal of Motor Behavior, 26, 56-62. Lukeman, D., & Melvin, D. (1993). Annotation: The preterm infant: Psychological issues in childhood. Journal of Child Psychology and Psychiatry,34, 837-849. McDonnell, P.M., & Corkum, V.L. (1991). The role of reflexes in the patterning of limb movements in the first six months of life. In J. Fagard & P.H. Wolff (Eds.), The development of timing control and temporal organization in coordinated action. Amsterdam: Elsevier Science. McDonnell, P.M., Corkum, V.L., & Wilson, D.L. (1989). Patterns of movement in the fast 6 months of life: New directions. Canadian Journal of Psychology, 43, 320-339. Parker, H. (1992). Children's motor rhythm and timing: A dynamical approach. In J.J. Summers (Ed.), Approaches to the Study of Motor Control and Learning (pp. 163-194). Amsterdam: Elsevier Science.

Acquisition of Motor Coordinationin Infants 227 Piek, J.P. A quantitative analysis of spontaneous movements in infants. Manuscript submitted for publication. Piek, J.P., & Carman, R.C. (1994). Developmental prof'des of spontaneous movements in infants. Early Human Development, 39,109-126. Piper, M.C., Byrne, P.J., Darrah, J., & Watt, M.J. (1989). Gross and fine motor development of preterm infants at eight and 12 months of age. Developmental Medicine and Child Neurology, 31, 591-597. Prechtl, H.F.R. (1984). Continuity and change in early neural development. In H.F.R. Prechtl (Ed.), Continuity of Neural Functions from Prenatal to Postnatal Life, Clinics in Developmental Medicine Vo194. Oxford: Blackwell Scientific. Prechtl, H.F.R. (1990). Qualitative changes of spontaneous movements in fetus and preterm infant are a marker of neurological dysfunction. Early Human Development, 23, 151- 158. Prechtl, H.F.R., Fargel, J.W., Weinmann, H.M., & Bakker, H.H. (1979). Postures, Motility and Respiration of Low-risk Pre-term Infants. Developmental Medicine and Child Neurology, 21, 3-27. Prechtl, H.F.R., & Hopkins, B. (1986). Developmental transformations of spontaneous movements in early infancy. Early Human Development, 14, 233-238. Prechtl, H.F.R., & Nolte, R. (1984). Motor behaviour of preterm infants. In H.F.R. Prechtl (Ed), Continuity of Neural Functions from Prenatal to Posmatal Life, Clinics in Developmental Medicine, Vol. 94 (pp. 79-92). Oxford: Blackwell Scientific. Provine, R.R. (1993). Prenatal behavior development: Ontogenetic adaptations and non- linear processes. In G.J.P. Savelsbergh (Ed.), The Development of Coordination in Infancy (pp. 204-236). Amsterdam: Elsevier Science. Purves, D. (1988). Body and brain. Cambridge, MA: Harvard University Press. Robertson, S.S. (1985). Cyclic motor activity in the human fetus after midgestation. Developmental Psychobiology, 18, 411-419. Robertson, S.S. (1990). Temporal organization in fetal and newborn movement. In H. Bloch & B.I. Bertenthal (Eds.), Sensory-Motor Organizations and development in Infancy and Early Childhood (pp. 105-122). Amsterdam: Kluwer Academic. Robertson, S.S. (1993). Oscillation and complexity in early infant behavior. Child Development, 64, 1022-1035.

228 J.P. Piek Sarnat, H.B. (1994). Maturation of the nervous system and malformations as disturbances in normal developmental processes. Transactions of the 15th Annual Neuroscience Colloquium of Western Australia, Perth, Western Australia. Savelsbergh, G.J.P. (1993a). The state of the art: A personal view on the development of movement coordination. In G.J.P. Savelsbergh (Ed.), The Development of Coordination in Infancy (pp. 519-524). Amsterdam: Elsevier Science. Savelsbergh, G.J.P. (1993b). Perspectives on the development of movement coordination in infancy. In G.J.P. Savelsbergh (Ed.), The Development of Coordination in Infancy (pp. 1-9). Amsterdam: Elsevier Science. Schmidt, R.A. (1975). A Schema theory of discrete motor skills learning. Psychological Review, 82, 255-260. Schmidt, R.A. (1980). Past and future issues in motor programming. The Research Quarterly for Exercise and Sport, 51,122-140. Schneider, K., Zemicke, R.F., Ulrich, B.D., Jensen, J.L., & Thelen, E. (1990). Understanding movement control in infants through the analysis of limb intersegmental dynamics. Journal of Motor Behaviour, 22, 493-520. Shapiro, D.C., & Schmidt, R.A. (1982). The schema theory: recent evidence and developmental implications. In J.A.S. Kelso & J.E. Clark (Eds), The Development of Movement Control and Coordination (pp.113-150). New York: Wiley. Sporns, O., & Edelman, G.M. (1993). Solving Bernstein's problem: A proposal for the development of coordinated movement by selection. Child Development, 64, 960-981. SPSS for Windows: Trends (1993). Chicago: SPSS Inc. Summers, J.J. (1992). Movement behaviour: A field in crisis? In J.J. Summers (Ed.), Approaches to the Study of Motor Control and Learning (pp. 551-562). Amsterdam: Elsevier Science. Swinnen, S., Heuer, H., Massion, J., & Casaer, P. (1994). Interlimb coordination: Neural dynamical, and cognitive constraints. San Diego: Academic Press. Thelen, E. (1979). Rhythmicalstereotypies in normal human infants. Animal Behaviour, 27, 699-715. Thelen, E. (1981). Kicking, rocking, and waving: Contextual analysis of rhyOunical stereotypies in normal human infants. Animal Behaviour, 29, 3-11.

Acquisition of Motor Coordinationin Infants 229 Thelen, E. (1985). Developmental origins of motor coordination: Leg movements in human infants. Developmental Psychobiology, 18, 1-22. Thelen, E. (1992). Development of locomotion from a dynamic systems approach. In H. Forssberg and H. Hirschfeld (Eds.), Movement Disorders in Children (pp.169-173). Medicine and Sport Science, Vol.36. Basel: Karger. Thelen, E., Bradshaw, G., & Ward, J.A. (1981). Spontaneous kicking in month-old infants: Manifestation of a human central locomotor program. Behavioral and Neural Biology, 32, 45-53. Thelen, E., & Fisher, D.M. (1983a). The organisation of spontaneous leg movements in newborn infants. Journal of Motor Behaviour, 15, 353-377. Thelen, E., & Fisher, D.M. (1983b). From spontaneous to instrumental behaviour: Kinematic analysis of movement changes during very early learning. Child Development, 54, 129-140. Thelen, E., Jensen, J., Kamm, K., Corbetta, D., Schneider, K., & Zemicke, R.F. (1991). Infant motor development: Implications for motor neuroscience. In J. Requin and G.E. Stelmach (Eds.), Tutorials in Motor Neuroscience (pp. 43-57). Amsterdam: Kluwer Academic. Thelen, E., Ridley-Johnson, R., & Fisher, D. (1983). Shifting patterns of bilateral coordination and lateral dominance in the leg movements of young infants. Developmental Psychobiology, 16, 29-46. Touwen, B.C.L. (1990). Variability and stereotypy of spontaneous motility as a predictor of neurological development of preterm infants. Developmental Medicine & Child Neurology, 32, 501-508. Turvey, M.T., & Fitzpatrick, P. (1993). Commentary: Development of perception-action systems and general principles of pattern formation. Child Development, 64, 1175-1190. Ulrich, B.D., & Ulrich, D.A. (1993). Dynamic systems approach to understanding motor delay in infants with Down syndrome. In G.J.P. Savelsbergh (Ed.), The Development of Coordination in Infancy (p. 445-459). Amsterdam: Elsevier Science. Valvano, J., & DeGangi, G.A. (1986). Atypical posture and movement findings in high risk pre-term infants. Physical & Occupational Therapy in Pediatrics, 6, 71-85.

230 J.P. Piek van Kranen-Mastenbroek, V., van Oostenbrugge, R., Palmans, L., Stevens, A., Kingma, H., Blanco, C., Hasaart, T., & Vies, J. (1992). Inter- and intra-observer agreement in the assessment of the quality of spontaneous movements in the newborn. Brain & Development, 14, 289-293. Vereijken, B., van Emmerik, R.E.A., Whiting, H.T.A., & Newell, K.M. (1992). Free(z)ing Degrees of Freedom in Skill Acquisition. Journal of Motor Behavior, 24, 133-142. von Hofsten, C. (1989). Mastering reaching and grasping: The development of manual skills in infancy. In S.A. Wallace (Ed.), Perspectives on the Coordination of Movement (pp. 223-258). Amsterdam: Elsevier Science. Whitall, J., & Clark, J.E. (1994). The development of bipedal interlimb coordination. In S. Swinnen, H. Heuer, J. Mission, & P. Casaer (Eds.), lnterlimb coordination: Neural dynamical, and cognitive constraint (pp. 391-411). San Diego: Academic Press. Wolff, P.H. (1967). The role of biological rhythms in early psychological development. Bulletin of the Menninger Clinic, 31,197-218. Wolff, P.H. (1968). Stereotypic behaviour and development. The Canadian Psychologist, 9, 474-483.

Motor Control and Sensory Motor Integration: Issues and Directions 231 D.J. Glencross and J.P. Piek (Editors) 9 1995 Elsevier Science B.V. All rights reserved. Chapter 9 GOING AROUND IN CIRCLES: THE DYNAMICS OF BIMANUAL CIRCLING JJ.Summers Department of Psychology, University of Southern Queensland A.Semjen, Laboratoire de Neurosciences Cognitives, CNRS, Marseille, France R.G. Carson, Department of Human Movement Studies, University of Queensland, Brisbane, Australia. J. Thomas Department of Psychology, University of Southern Queensland, Toowoomba, Australia. In the study of the dynamics of motor control tasks requiring multijoint intralimb coordination have been rare and considerations of multijoint bimanual coordination almost completely absent. This chapter reviews some recent research we have conducted using a bimanual circle drawing task to elucidate the dynamics of intralimb and interlimb coordination. Right-handed and left-handed subjects were required to produce circles in the horizontal plane with both hands simultaneously in either a symmetrical mode (i.e., one hand moving clockwise, the other counter-clockwise) or in an asymmetrical mode (i.e., both hands moving clockwise or counter-clockwise). These movements were performed at various frequencies. The main findings from the studies were: (1) symmetrical circling was a much easier task than asymmetrical circling; (2) in the asymmetrical mode, distortions of movement trajectories, suggestive of loss of stability, and phase transitions, were evident as movement frequency was increased; and (3) distortions of circular trajectories were most prominent for movements of the nondominant hand and appeared to precipitate movement reversals. The results are discussed in terms of manual asymmetries and the putative neural mechanisms underlying the bimanual circling effects. 1. INTRODUCTION How the limbs are coordinated is a central issue in the study of movement control that has been addressed from neural, dynamic and cognitive perspectives (eg., Swinnen, Heuer,

232 J.J. Summers et al. Massion, & Casaer, 1994). Much of this work has focused on the constraints on interlimb coordination that are evident when we attempt to perform two motor tasks at the same time. Patting the head while rubbing the stomach is the standard example but, more formally, interference between concurrent motor activities has been demonstrated in a variety of bimanual tasks. For example, great difficulty is experienced in the performance of similar movements (eg., tapping) by the two hands when the movement streams form nonharmonically related rhythms, as in polyrhythms (eg., Peters & Schwartz, 1989; Summers & Kennedy, 1992; Summers & Pressing, 1994). Mutual interference is also observed when the hands perform in the same time frame different movements such as drawing circles with one hand and straight lines with the other hand (eg., Franz, Zelaznik, & McCabe, 1991; Swinnen, Walter, & Shapiro, 1988; Swinnen, Walter, Beinrinckx, & Meugens, 1991). In all the above situations the interference between limbs is usually manifested as a tendency to synchronise the movements of the two limbs. The tendency toward synchronisation of the limbs in rhythmic motor tasks has been extensively demonstrated in the finger oscillation task developed by Kelso and colleagues (e.g., Scholz & Kelso, 1989). In this task subjects are required to cycle their index fingers at the same frequency adopting either an in-phase coordination mode (simultaneous flexion and extension of both fingers) or an anti-phase mode (one fmger extends while the other flexes). Typically, the antiphase mode is less stable than the in-phase and cannot be maintained beyond a critical frequency at which spontaneous transition to the in-phase mode is commonly observed. Spontaneous transitions between modes of coordination have now been demonstrated in a variety of bimanual and unilateral motor tasks including wrist oscillations (e.g., Byblow, Carson & Goodman, 1994), coordination of wrist and ankle (eg., Baldissera, Cavallari, Maafini, & Tassone, 1991; Carson, Goodman, Kelso, & Elliott, 1993), elbow and wrist (eg., Kelso, Buchanan, & Wallace, 1991), and elbow and knee (Kelso & Jeka, 1992). The transition phenomena have formed the cornerstone of the 'dynamical pattern approach' to motor coordination. This approach has modelled the transitions between coordinative patterns using concepts from synergetics and nonlinear oscillator theory (Haken, Kelso, & Bunz, 1985). According to this conceptualisation coordinative states are characterised in terms of collective variables or order parameters which are abstract properties that describe the relationship between the individual components in a system and also distinguish between different stable patterns of coordination (e.g., Kelso & Jeka,

The Dynamics of Bimanual Circling 233 1992). Of particular interest in this approach are situations which induce a qualitative change in system behaviour or 'phase transitions'. Examination of pre and post transition behaviour allows the identification of the relevant collective variables. At the level of limb kinematics, an appropriate collective variable has been found to be the relative phase angle between two oscillating limbs. Control parameters are parameters responsible for inducing changes in the coordinative pattern by leading the system into states where the existing pattern is unstable. In the lrmger oscillation studies, for a system prepared in the anti-phase mode increasing the cycling rate produced, at a critical frequency, a spontaneous shift to an in-phase pattern (Kelso, et al., 1981). Frequency, because it induces phase transitions is regarded, therefore, as a control parameter. Patterns that occur spontaneously during functional behaviours are considered to be intrinsic to the system (Schoner & Kelso, 1988). Intrinsic dynamics reflect the internal coordination constraints which are unspecific to the behavioural pattern produced, but determine the stable collective states toward which behaviour is spontaneously attracted. The switching between coordinative modes evident in the finger oscillation tasks has been modelled successfully using the tools of synergetics by Haken, Kelso & Bunz (1985). One feature of the Haken et al., model that has proved to be unsupported is the assumption of symmetry between component oscillators. For example, there is growing evidence that in one-dimensional oscillatory movements of the fingers and hands when a transition from anti-phase coordination to in-phase coordination occurs it is usually precipitated by the nondominant hand falling into the phase of the dominant hand (Byblow, Carson, & Goodman, 1994; Carson, Byblow, & Goodman, 1994; MacKenzie & Patla, 1983). Recent extensions to the Haken et al., (1985) model included a symmetry breaking term to account for situations in which the inherent frequencies of the component oscillators are different, as is the case in arm-leg coordination (Kelso & Jeka, 1992; Kelso, Delcolle & Schoner, 1990). Alternatively, manual asymmetries have been treated as reflecting differences in the strength of the coupling between the left and fight limbs (Byblow et al., 1994; Carson et al., 1993; see also Treffner & Turvey, in press). There is also a great deal of evidence of manual asymmetries in bimanual coordination when the two hands perform different tasks. For example, strong asymmetry effects are found when subjects are asked to tap out a simple rhythm with one hand while tapping fast or slow with the other hand (Ibbotson & Morton, 1981; Peters, 1981, 1985). In right- handers performance in these dual tasks is better when the fight hand takes the activity that

234 J.J. Summers et al. is more demanding in terms of attention. Peters (1990, 1994a) has argued that such asymmetries do not simply reflect basic unilateral asymmetries but are due to asymmetrical allocation of focal attention in dual task situations. 2. BIMANUAL CIRCLING Although considerable advances have been made in understanding the spatial and temporal constraints in the dynamics of motor control, much of our knowledge has been obtained from bimanual tasks requiring the coordination of homologous joint pairs, or unilateral tasks involving heterolateral or homolateral joints. Studies of multijoint intralimb coordination have been less common and examination of multijoint bimanual coordination almost entirely absent. Recently, we have conducted a series of experiments involving a task requiting multijoint intralimb and interlimb coordination. The task involves tracing circles with both hands simultaneously in either a symmetrical mode (i.e., one hand moving clockwise and the other counter-clockwise) or asymmetrical mode (i.e., both hands moving clockwise or counter-clockwise). When you attempt this task the first thing you notice is that symmetrical circling is easier to perform than asymmetrical circling. Of particular interest was the question of whether speeding up the circling movements would produce a phase transition, i.e., a switch from the asymmetrical mode of coordination to the symmetrical mode, and whether the transitions occurred preferentially in the nondominant hand. 3. EXPERIMENT I In the first experiment (Semjen, Summers & Cattaert, in press) we examined circle drawing in two right-handed and two left-handed subjects. Circle drawing was compared under three conditions: (1) producing circles in clockwise and anti-clockwise directions with each hand alone; (2) bimanual circling in a symmetrical mode in either an 'inwards' direction (the left hand moving clockwise and the right hand anti-clockwise) or 'outwards' direction (the left hand moving anti-clockwise and the right hand clockwise), and (3) bimanual circling in an asymmetrical mode with both hands moving either clockwise or anti-clockwise.

The Dynamics of Bimanual Circling 235 Subjects were required to trace continuously with their index fingertip(s) the contour of 10 cm diameter circle(s) drawn on a paper sheet fixed to a table's surface. They were instructed to execute as well-formed circles as possible at either their preferred speed or as fast as possible. No instructions were given as to whether phase transitions should be resisted. Each trial was 10 seconds in duration and fingertip trajectories were recorded with an infrared camera system (ELITE system). A light reflecting marker was secured on the end of the index finger of each hand. For data analyses, the circular paths of the fingertips were decomposed into oscillations along the Y axis (parallel to the sagittal plane) and the X axis (perpendicular to the Y axis). The main fmdings were that close synchronisation between the hands was possible in symmetrical circling movements. In contrast, large distortions of hand trajectory and transient reversals of movement direction were observed when asymmetrical movements were performed as fast as possible. Furthermore, these effects were confined to the nondominant hand. In Figure 1 the f'mgertip trajectories during an anti-clockwise asymmetrical trial under speed instructions are shown for a left-handed subject. During the initial few cycles (Figure 1, 1-200) the subject is able to perform the task satisfactorily, although the circles made with the nondominant (fight) hand are larger and more elliptical than those produced by the dominant hand. The trajectories produced by the nondominant hand became more variable and the ellipses increasingly elongated (Figure 1,201-400) until after approximately 4 secs of circling the nondominant hand exhibited a direction reversal into the symmetrical (outwards) mode. The reversal was followed almost immediately by a return by the nondominant hand to the required asymmetrical movement (Figure 1, 401- 600). This correction took the form of abrupt switching of the nondominant hand's oscillations along the X axis from symmetrical phase into the asymmetrical phase while the phase relationship between the Y components remained relatively unaffected. A period of stable asymmetrical circling then ensued until about 7 secs into the trial another abrupt direction reversal by the nondominant hand occurred followed almost immediately by an attempt to return to asymmetrical circling and then another direction reversal (Figure 1, 601-800). The reversal was again quickly corrected and the subject was maintaining the anti-clockwise direction of movement up to the end of the trial, but the nondominant hand's trajectory was considerably distorted (Figure 1, 801-1000).

236 J.J. Summers et al. 401-600 1-200 201-400 O0 O@ I- i- ' - L ..,, .. , , _. , 801-1000 601-800 00 Figure 1. Handtrajectoriesof a left-handedsubjectduring a 10 see. trial of asymmetrical(anti-clockwise) circling at the maximumrate. Each of the five trajectoryplots (e.g., 1-200) represents200 recordingpoints (2 secs). Severe trajectory distortions of the nondominant hand, including segments of nearly straight lines and diagonals crossing the workplane, were exhibited to a greater or lesser extent by all subjects when producing asymmetrical movements at their maximum rate. Furthermore, breakdowns in the control of the movement trajectory often culminated in a reversal of the movement direction. There were, however, numerous cases of trajectory distortions occurring without subsequent transition to symmetrical circling. It is also noteworthy that during these trials the dominant hand of both right- and left-handers was relatively unaffected by the wild excursions of the nondominant hand. Some insight into the processes underlying these effects was gained through examining the phase relationships between the oscillatory displacements of the left and right hands along the X and Y axes. Phase differences (signed deviations from synchrony) were measured for every cycle, with perfect synchronisation between the hands being indicated by a value of zero. Positive phase differences indicate that the dominant hand was lagging behind the nondominant hand, whereas negative values indicate that the dominant hand was leading the nondominant hand (see Figure 2).

9The Dynamicsof Bimanual Circling 237 At the preferred rate, the phase difference for all subjects was close to zero (i.e., synchrony) during bimanual symmetrical movements and tended to remain stable over successive movement cycles. During asymmetrical movements, the most commonly observed pattern was an increasing deviation from synchrony with the phase difference values becoming more negative over cycles. Sym in --- Y Sym out 0 . 5 ,'-..................... \"....................... +X 0 ~ i i i i i i i ~ .... 0.5 -\" .......... :.......... \"..................... :- -0.5 -;..................... ;..................... 0 / o 0 15 30 t- LO_ .e2:_0 0 e- 13. -0.5 ; ..................... ; ..................... ~1 0 15 30 Asym ccw Asym cw 0 . 5 -'- ..................... : ..................... '- 0.5 \"'. ..................... \".................... \"\" O t fio 0 ~~ : i 0 -~~-0 ............................................ 9i ~t -0.5 -; ..................... ;..................... ;- -0.5 ,.. . . . . ;........~ ..... 0 12 24 0 12 24 Cycles Cycles Figure 2. Phase difference between the hands on the Y and X axes as a function of movement cycles, during four individual trials of bimanual circling at the maximum rate. Top panels show symmetrical inwards (Sym in) and outwards (Sym out) movements, lower panels asymmetrical anticlockwise (Asym ccw) and clockwise (Asym cw). When subjects performed the bimanual movements as fast as possible, symmetrical movements were again produced with a phase difference close to zero. This is illustrated in Figure 2 which shows individual symmetrical trials (Sym in, Sym out) for a right-handed subject. In contrast, there appeared three broad patterns of behaviour when subjects

238 J.J. Summers et al. performed asymmetrical movements at maximum speed. In one, the hands immediately adopted an asynchrony, usually with the dominant hand leading, which remained stable across cycles. In the second pattern (Figure 2, Asym ccw), the hands exhibited increasing asynchrony (again, dominant hand typically leading) over cycles with phase slippages often occurring towards the end of the trial. The final pattern was observed when the two hands were moving at vastly different rates (i.e., no frequency locking). In all cases, it was the dominant hand that was moving the faster and continuous variations in phase difference values occurred over successive movement cycles (Figure 2, Asym cw). In summary, when both fight- and left-handed subjects were asked to increase the rate of bimanual symmetrical movements little disruption to the coordination between the hands was observed. All subjects were able to maintain synchrony between the hands during symmetrical circling movements. Increasing the rate of asymmetrical movements, in contrast, produced dramatic effects on the phase of the hands. Asynchrony between the hands, with the dominant hand leading, was frequently observed as were breakdowns in coordination manifest by phase slippages, non 1:1 frequency relations, and transitions from asymmetrical circling to symmetrical circling. These phenomena indicate that at a high circling frequency the stability of asymmetrical coordination was lost and the system was attracted toward the more stable symmetrical mode. The trajectory distortions may be seen as an expression of the conflict between these differentially stable modes of coordination. In the Semjen, Summers and Cattaert (in press) study the speed at which subjects performed the circling movements was not directly controlled. That is, subjects were free to interpret the rate instructions as they wished. As a consequence, when asked to go as fast as possible the frequency produced by the dominant hand ranged across subjects from 2.53 Hz to 4.33 Hz. If frequency acted as a control parameter driving the system through its different coordinative states, then it is not surprising that the subject showing the highest number of direction reversals also exhibited the shortest movement periods. Furthermore, some subjects appeared to avoid a direction reversal by subtle modifications to the circling movements, such as slowing down. 4. EXPERIMENT H In a second study, therefore, we attempted to control movement frequency by requiting subjects to synchronise their movements to an auditory metronome which

The Dynamics of Bimanual Cdrcling 239 increased in frequency over a trial. Six right-handed subjects were instructed to produce one full cycle of movement for each beat of the metronome while maintaining the prescribed mode of coordination. The frequency of the metronome was scaled from 1.50 Hz to 3.00 Hz in 7 steps of 8 seconds duration. Thus, each trial lasted approximately 60 secs and there were 6 trials per condition. The conditions were identical to those used in the first study: Single limb, bimanual symmetrical, and bimanual asymmetrical, with two directions of movement being examined for each condition. Again, no explicit instructions were given regarding a possible pattern change. 4.1 Results (1) Temporal Aspects Frequency Deviation. In Figure 3 is shown the extent to which the frequency of oscillations of each hand (collapsed over X and Y dimensions) deviated from the frequency of the metronome. As can be seen, the frequency of the right hand tended to be slightly greater than that of the metronome (positive deviations) in all conditions. In contrast, movement frequency of the nondominant (left) hand was substantially below that of the metronome (negative deviations) on both the single limb and bimanual asymmetrical conditions. The movement frequency of the left hand, however, was considerably enhanced when in a symmetrical relationship with the dominant hand and closely matched the Frequency Deviation: Hand by Condition 0.15 I Left Hand 0.1 1 I~ RightHand -'O.~- 0.05- .t'~l > 1a2~1 0 >, - t- -0.05 - i,,.. u_ -0.1- -0.15 J Single Asymmetrical Symmetrical Condition Figure 3. Meanfrequency deviation scores in Hz (Mean produced frequency - metronome frequency) for the left hand and right hand across the three coordination conditions.

240 J.J. Summers et al. metronome frequency. Examination of these trends across levels of pacing frequency showed that the left hand decrement in both the single limb and asymmetrical conditions was an increasing function of pacing frequency (F5,30 = 2.237, p < 0.001). Coefficient of Variation. The variability of movement frequency of each hand in each condition is shown in Figure 4, in terms of coefficient of variation (Standard deviations in relation to the mean, expressed as percentages). Movements of the fight hand were significantly less variable than those of the left hand in all conditions. Furthermore, the degree of variability for movements of the fight hand did not vary as a function of condition (F5,30 = 1.595, p > 0.05). In contrast, variability of the left hand was much greater in the asymmetrical condition than in the other two conditions. It is interesting to note the equivalence in variability of the single limb and symmetrical conditions given the marked difference in frequency deviation (Figure 3) between the two conditions. CV Frequency: Hand by Condition e 12 Left Hand 10 L~! Right Hand >, 8 ~6 m U. o4 Single I i\" Asymmetrical Symmetrical Condition Figure 4. Coefficient of variation for the left hand and right hand across the three coordination conditions. Betweenhand significant differences indicated by asterisk. Examination of these trends across levels of pacing frequency (Figure 5) indicates that the variability of movements of the right hand was little affected by increases in movement frequency. For movements of the left hand, however, in the asymmetrical condition there was a linear increase in variability with increases in pacing frequency (F5,30 = 3.23, p < 0.0001).

The Dynamics of Bimanual Orcling 241 CV Frequency:Conditionby Frequency(Left Hand) 18- .-.0- Single 16 w Asymmetrical 14 - - B - Symmetrical g~o II \"1 fgt. 8 2 2.25 2.5 2.75 3 6 Frequency (Hz) 4 2 1.5 1.75 18 CV Frequency:Conditionby Frequency(Right Hand) 16 Single Asymmetrical 14 Symmetrical >, o 12 o-10 2 LL \" o> 8-9 IIIIII I 1.5 1.75 2 2.25 2.5 2.75 3 Frequency (Hz) Figure 5. Coefficient of variation for the left hand and right hand in the three coordination conditions as a function of pacing frequency. (2) Spatial Aspects The circling task imposes both temporal and spatial constraints. As we have seen, coordination mode had a clear effect on the timing variability of circling movements. To examine the spatial aspects of the task we developed an Aspect Ratio (Walters & Carson, 1994) to indicate the degree of circularity of the movement trajectory. The aspect ratio considers all points in the data set to define the closed contours of the movement trajectory. The data points comprising each contour are linearly connected and treated as forming the boundary of a surface area. The area, the first and second moments of the area, the principal moments of inertia and the principal axes of inertia of each surface area (about

242 J.J. Summers et al. centroidal axes) are calculated. These give all the parameters required to completely characterise the best fitting elipse without the need for scaling (see Waiters & Carson, 1994). The aspect ratio is simply the ratio of the minor and major axes- a value of 1 indicates a perfect circle, a value of 0 indicates a straight line. As is evident from Figure 6, trajectories produced by the fight hand were more circular than those produced by the left hand in all conditions (F1,4 =17.892, p<0.05). Furthermore, the degree of circularity did not change across conditions for the fight hand (all F < 1.0). In contrast, movements made by the left hand were less circular in the asymmetrical condition than those made in both the single limb and bimanual symmetrical conditions (all p < 0.01). Aspect Ratio: Hand by Condition 0.85 - Left Hand I~1 Right Hand 0.8 \\\\'~\\ . m0 \\\\\\\\ rr \\\\\\\\ \"5 0.75 \\\\\\\\ \\\\\\\\ r \\\\\\\\ \\\\\\\\ < 0.7- \\\\\\\\ \\\\\\\\ 0.65 \\\\\\\\ \\\\\\\\ \\\\\\\\ \\\\\\\\ \\\\\\\\ \\\\\\\\ \\\\\\\\ \\\\\\\\ \\\\\\\\ \\\\\\\\ 0.6 i i ---I Single Asymmetrical Symmetrical Condition Figure 6. Meanaspect ratio for the left hand and right hand across the three coordination conditions. Betweenhand significant differencesindicatedby asterisk. Interestingly, while there was a linear decrease in the degree of circularity with increases in pacing frequency for the left hand, this was expressed equally across all conditions. That is, for the nondominant hand the differences in circularity between the conditions did not change as a function of frequency (Figure 7). For the right hand, however, the shape of the trajectory was affected by increasing frequency to a smaller

The Dynamics of Bimanual Orcling 243 degree than the left hand. This effect was evident in the interaction of hand with condition (F2,8 = 6.842, p < 0.05). Aspect Ratio: Condition by Frequency(Left Hand) 0.9- Single Asymmetrical 0.85 Symmetrical o 0.8 ~ 0.75 < 0.7. 0.65 0.6 I 1 I I I I 1.5 1.75 2 2.25 2.5 2.75 3 Frequency (Hz) Aspect Ratio: Condition by Frequency(Right Hand) 0.9- Asymmetrical - 4 - - Symmetrical 0.85. O 0.8. n-\" c~.0\"75 u) < 0.7 0.65 0.6 ! I I I I I 1.5 1.75 2 2.25 2.5 2.75 3 Frequency (Hz) Figure 7. Aspect ratio for the left hand and the right hand in the three coordination conditions as a function of pacing frequency. The second experiment, therefore, confirmed the main phenomena reported by Semjen, Summers, and Cattaert (in press). That is, relative to single limb movements, when subjects gradually increased the rate of circling movements while adopting an asymmetrical coordination mode the nondominant hand was unable to maintain the required frequency, exhibited large variability and a loss of spatial accuracy. In contrast, when the two hands circled in a symmetrical relationship the nondominant hand performed no worse than under

244 J.J. Summers et al. single limb conditions and was actually facilitated in its ability to maintain the metronome frequency. In both studies when subjects drew circles with a single limb, faster and more regular movements were performed with the dominant hand than with the nondominant hand. It is possible, therefore, that the temporal asynchrony frequently observed between the hands in the performance of bimanual asymmetrical movements may reflect differences in the characteristic frequencies of the two hands. A number of features of the asymmetrical circling data were consistent with a coordination system in which the subcomponents have different inherent frequencies. First, frequency relations other than 1:1 (e.g., 2:3) sometimes emerged during asymmetrical circling at maximum speed. Second, f'Lxed phase relations other than absolute in-phase or anti-phase synchronization (fixed point drift) were observed. Third, when phase asynchrony occurred the nondominant hand almost always lagged the dominant hand. Recently, Stucchi and Viviani (1993) reported that in a bimanual task involving the tracing of ellipses in the vertical plane the dominant hand consistently led the nondominant hand. Similarly, Jardin, Swinnen and Meulenbroek (1994) using a bimanual circling task found a phase lag of 8 to 10 degrees between the dominant and nondominant limb when in the symmetrical coordination mode. In both these studies, however, the fastest movement rates examined were equivalent to the slowest rates (1 - 1.5 Hz) used in our studies. There were, however, other features of the data from the two studies that suggest that the observed manual asymmetries may not arise solely from differences in characteristic frequencies of the two hands. First, in the symmetrical mode of coordination the hands remained closely synchronised across different movement rates and non 1:1 frequency ratios and fixed point drift were not observed. Second, differences between the mean frequencies of the hands in the asymmetrical mode at the preferred rate were quite small and increased as movement rate became greater. Furthermore, increases in movement rate were accompanied by trajectory distortions which occurred particularly in the nondominant hand during asymmetrical circling. Thus, the temporal differences evident between the hands during asymmetrical circling at the faster rates may reflect a slowing down of the nondominant hand due to the production of distorted trajectories and transitions and also differences in the intrinsic frequencies of the two hands. A different or additional factor contributing to the manual asymmetries observed in the bimanual circling task may be the differential coupling between the limbs. Specifically, it

The Dynamics of Bimanual Ording 245 has been argued that in the region of a transition the magnitude of the coupling influence of the dominant limb on the nondominant limb is greater than that in the opposite direction (Byblow, Carson, & Goodman, 1994; Carson, 1993; Carson, Byblow, & Goodman, 1994). Furthermore, it is assumed that coupling strengths are inversely proportional to the frequency of oscillation. That is, the mutual coupling influences of the dominant and nondominant limbs are equal and opposite at lower movement frequencies, whereas at higher movement frequencies the influence of the dominant limb on the nondominant exceeds that acting in the opposite direction. The above description does capture many of the manual asymmetry features observed in the bimanual circling task, such as the increase in phase difference between the hands with increasing movement rate and that movement reversals during asymmetrical circling were always precipitated by the nondominant limb. At another level of description attention has been suggested as an intervening variable in manual asymmetries. From this perspective, limitations on the ability to perform concurrent movements of the two hands reflect the problem of dividing focal attention. Furthermore, Peters (1985, 1990, 1994a) has argued that in the performance of bimanual activities there is an inherent attention asymmetry. That is, \"the preferred hand normally receives focused attention (figure) while the nonpreferred hand is not directly attended to (ground), or receives only subsidiary attention\" (Peters, 1994a, p. 599). The differential focusing of attention would be most evident when bimanual task demands prohibit the rapid switching of attention between the hands or the integration of the movements of the two hands into a common temporal chain (Peters, 1994b). Thus, the increased variability exhibited by the nondominant hand as movement rate increased during asymmetrical circling may reflect the withdrawal of attention from that hand. Certainly, in both bimanual circling experiments subjects frequently reported that maintaining the asymmetrical coordination mode required great mental \"effort\". Research examining the role of attention factors in the circling task is currently in progress. 5. NEURAL BASIS OF BIMANUAL CIRCLING EFFECTS The three main findings from the bimanual circling studies that need explanation are: (1) circling the two hands in a symmetrical mode is a much easier task than circling the hands

246 J.J. Summers et al. in an asymmetrical coordination mode, (2) there are manual asymmetries evident in bimanual circling which are exaggerated when the two hands are moved with increasing frequency in the same direction (ie. clockwise or anti-clockwise), and (3) when performing asymmetrical circling under speed instructions, the nondominant hand lags behind the dominant hand and exhibits trajectory distortions and movement reversals. 5.1 Coordination Mode An important difference between the two coordination modes examined in this chapter is that the symmetrical movements were generated by simultaneous activation of homologous muscles, whereas asymmetrical movements involve at least some simultaneous activation of nonhomologous muscle groups. It has been suggested that the body's bilaterally symmetrical subsystems may be strongly coupled by concomitant activation of homologous muscle groups (Cohen, 1971). At the supraspinal level there are bilateral connections between the sensory and motor cortices for proximal limb musculature, while at lower levels there is bilateral distribution of some descending motor pathways (Swinnen et al., 1991). There is evidence that each half of the brain has full contralateral control over arm, hand and finger movements and also ipsilateral control of arm movements through bilateral as well as unilateral projections from the cerebral cortex through the brain stem to spinal cord (Brinkman & Kuypers, 1972, 1973; Swinnen, et al., 1991). It appears, therefore, that when different actions have to be performed by the limbs at the same time the tendency toward co-activation of homologous muscle groups must be overcome. Swinnen and colleagues, for example, in a series of studies have shown that learning to produce two different movement trajectories in both upper-limbs simultaneously involves the gradual elimination of mutual interference or neural \"crosstalk\" (see Swinnen & Walter, 1988; Walter & Swinnen, 1994 for reviews). Although the corpus callosum is an obvious candidate for the transmission of symmetrical control, greater attraction to in- phase and anti-phase patterns was evident in 'split-brain' patients (Tuller & Kelso, 1989) suggesting that other sub-cortical pathways may be involved in interhemispheric transmission. The split-brain research also suggests that the corpus callosum may be important for the inhibitory feedback which is required for the production of independent limb movements. Baldissera, Cavallari, Marini and Tassone (1991) have argued that kinesthetic afferent signals are important in the production of phase relationships of interlimb segments.