Implications of a dynamical systems approach to understanding infant kicking behavior.

Traditionally, physical therapists have evaluated and treated infants at-risk for developmental disabilities based on the theoretical construct of maturation of the central nervous system (CNS). This belief purports that hierarchical development of the CNS results in inhibition of lower centers of the brain and the emergence of deliberate or voluntary movements. This intrinsic, or maturational, concept of development is based to a large extent on the writings of McGraw(l) and Gesell(2) in which the nervous system was thought to prescribe changes in movement. Motor development was depicted as a sequence of motor milestones occurring at specific ages. This sequence of motor development, which looked invariant, was taken to be reflective of the normal progression of the infant's developing nervous system. in the maturational theory, the biological make-up of the organism, specifically the CNS, was considered to be the effective force in determining the outcome of movement. Similar views on the role of neuromaturation persist today. in the development of walking, for example, Forssberg(3) argues that digitigrade walking gives way to plantigrade walking as higher centers of the CNS mature.

In addition to the concept of motor milestones depicting the development of the nervous system, the evolution of reflexes and reactions has also been related to the hierarchical development of the CNS.(4,5) From this perspective, early primitive reflexes are considered to be cortically inhibited by higher centers of the CNS, permitting postural reactions and voluntary motor control to develop.(6) Under this concept, reflexes and reactions are considered to be activated by sensory stimuli emanating from the external environment and to be the substrata of normal motor coordination.(5,7,8) Sensory inputs control motor outputs. The infant, therefore, is considered a passive recipient of environmental stimuli rather than an active participant in the process of development. With respect to the development of walking, the view that reflexes are the building blocks of later voluntary movement persists.(9) Zelazo and colleagues(9) emphasize the conversion of innate reflexive patterns to intentional and voluntary control of these movement patterns as a result of developing cognitive skills.

These theoretical perspectives, be they neuromaturational or cognitive, consider the development of movement to be the result of higher-order control of the nervous system and thus are prescriptive in nature. The conception of development is a gradual increase of cortical influence (ie, voluntary cerebral hemisphere activity that suppresses the subcortical reflexive display). Although cognitive theories incorporate interaction with the environment, emphasis is on the formation of progressive higher plans, instructions, and commands for producing behavior. Cognitive models, like strict neurological accounts of movement, assume that action arises solely from coded instructions. These constructs imply that somewhere in the body exist a set of commands that create and direct movement patterns. In the last few years, an alternative approach to the development of movement has been proposed, the dynamical systems theory. The CNS, according to this theory, is not viewed as the sole source of behavior but rather as one subsystem of many that dynamically interact to produce movement with respect to functional tasks. The framework for the dynamical systems theory was inspired by the work of Bemstein(10,11) and guided by the principles of nonequilibrium phenomena in physics.(12) These concepts have been elaborated on by various authors(13-17)and more recently expanded to the dynamic pattern theory.(18-20(The reader is referred to these works for more detailed accounts of the general principles underlying the dynamical systems theory and the dynamic pattern theory.) Thelen and colleagues(2l-25)have extended these concepts to the development of movement in humans.

The purpose of this article is threefold. First, I will review four assumptions of the dynamical systems perspective that pertain to the development of movement and that are relevant to work that I have done. Second, I will review my research studies with high- and low-risk preterm and full-term infants with respect to these theoretical assumptions. Last, I will offer suggestions on how the dynamical systems perspective may influence the practice of physical therapy in the evaluation and treatment of atypical infants. Theoretical Perspective Dynamical Systems Theory

The dynamical systems perspective provides a new way of conceptualizing motor development. Rather than viewing developing motor behaviors as the unfolding of predetermined or prescribed patterns in the CNS, this perspective sees motor behavior as emerging from the dynamic cooperation of many subsystems in a task specific context. That is, the subsystems self-organize to produce movement and do not depend on the prior existence of instructions embodied in one hierarchically important subsystem (like the CNS). I will present an outline of the principles of the dynamical systems perspective with respect to the development of movement to set the stage for the interpretation of my research with atypical infants. Readers are referred to the works of Thelen and colleagues(2l-25) for more detailed discussions of this theory and examples of this perspective as it pertains to normal motor development. Four critical assumptions of the dynamical systems perspective for this discussion are as follows: (1) Developing organisms are high-dimensional systems, and movement represents a compression of the multiple degrees of freedom; (2) movement emerges in a self organizing fashion as a function of the cooperation of the many subsystems in a task-specific context; (3) movement patterns are preferred, but not obligatory, and occupy preferred regions of their state space; and (4) new behavioral forms emerge in development as a series of phase shifts. Each of these key assumptions will be reviewed.

Assumption 1. Developing organisms are high-dimensional systems, and movement represents a compression of the multiple degrees freedom. A dynamical system is any system that changes over time. A high dimensional system is one that is composed of many degrees of freedom, that is, many elements. With respect to the human, the degrees of freedom, or the independent elements of the body, could be the muscles, bones, joints, neurons, and motor units. In this sense, the term "degrees of freedom" does not refer to spatial degrees of freedom such as flexion, extension, rotation, adduction, and abduction.

Bernstein(10,11) suggested that the many degrees of freedom inherent in the multiple muscles and joints of the body could be organized into larger functional groups or synergies constraining the muscles to act as a unit. These basic units of motor behavior came to be known as functional synergies (14) or coordinative structures.(14,26-28) Movement, however, is more than muscles and motoneurons. Other elements that participate in movement include sensory, perceptual, and integrative neural components; the respiratory and cardiac subsystems; and many levels of the autonomic subsystem. Thus, in more general terms, units of behavior may be described as coordinated patterns.(18)

How can we describe a movement composed of many elements, such as neurons, muscles, bones, and joints by only a few elements? How do you compress the many degrees of freedom inherent in a system to only a few degrees of freedom, or how do you go from a high-dimensional behavior to a low-dimensional description of that behavior? This low dimensional description of the behavior has been called a collective variable, or order parameter (19,21) Collective variables characterize movement patterns. They capture, in simpler terms, all the systems that cooperate to produce the movement. Collective variables allow us to quantify the coordination of movement of a single limb, such as intralimb kicking, or the coordination of movement between two limbs, such as interlimb walking. Examples of collective variables that quantitatively describe movement in a single limb are the timing of individual movement phases such as flexion and extension; phase lags, defined as the time between the onset of movement of one joint with respect to another joint; and the relationship of individual joints to each other.(29-31) At the interlimb level of analysis, as in walking, examples of collective variables are the relative temporal and spatial phasing between the two limbs.(21,31-35) Thus, collective variables are used to index changes in movement behavior.

Coordinative patterns have the characteristics of spatial and temporal order and of stability and flexibility. The stability of the coordinative patterns, indexed by the collective variables, captures the spatial and temporal order of the movement that we recognize as kicking, walking, or throwing a ball. Flexibility allows the coordinative pattern to adjust to a variety of environmental demands.

Assumption 2. Movement emerges in a self-organizing fashion as a function of the cooperation of the many subsystems in a task-specific context. Behavior is multiply determined and task assembled. Humans are complex biological systems in which movement outcome is an interplay of all components of the system in real and developmental time, assembled by the task and context. Real time refers to the self-organization of a movement pattern that is currently ongoing (ie, what the organism does second by second, minute by minute). Developmental time is the self-organization of movement patterns that change over days, months, or years. According to the dynamical systems perspective, no prescription or code, such as a central pattern generator or motor program, exists in the CNS before the movement is exhibited that specifies all of the details of a particular movement pattern. Rather, the outcome of the movement pattern is the result of the dynamic interaction of contributing subsystems that organize with respect to the demands of the task and the environmental context of the system. Examples of subsystems that may interact to produce movement are the pattern generation or coordinated pattern, joint synchrony, postural control (balance and equilibrium), body constraints, muscle strength, extensor and flexor control, perceptual processes, cognition, and motivation.

With respect to developing systems, each of these interacting subsystems has its own course of development and each matures at its own developmental rate. Thus, at every stage of development, movement is assembled with whatever subsystems are maturationally available with respect to the particular environmental and task specific context. No one subsystem is predominant in the system. Rather, it is function (task and context) that is the catalyst for assembling the available and appropriate subsystems for movement, rather than preexisting instructions.

Assumption 3. Movement patterns are preferred, but not obligatory, and occupy preferred regions of their state space.

A general characteristic of complex dynamical systems is to settle into preferred, but not obligatory, movement patterns. "Preferred" means that the system contains no prescription ahead of time for the movement pattern, but that the system simply prefers or "wants" to move in a certain way. The system does not necessarily have to move that way, because the pattern of movement is not preassembled in a fixed manner. In other words, these movement patterns are not "hard-wired" and rigidly programmed in the CNS but are "softwired" and flexible, that is, responsive and adaptable to changing biological and environmental conditions and to varying task demands. Even reflexes, which have been considered to be hard-wired in the CNS, are quite variable and responsive to the arousal level of the infant. Preferred movement patterns act as a kind of dynamic magnet, or dynamic attractor, such that when the system is perturbed, it tends to return to the preferred movement pattern. These preferred movement patterns can be represented graphically by plotting the joint angle of a moving limb against another joint angle of the same limb (angle-angle plot) or the joint velocity of a limb against the joint angle of the same limb (phase plane plot). If the movement pattern is stable, it occupies a preferred space within the plot. In dynamical terms, the preferred space occupied by the movement pattern is called a preferred attractor state space, and the preferred movement pattern is called an attractor.(21) Assumption 4 New bebavioral forms emerge in development as a series of phase shifts.

According to the dynamical systems perspective, a shift from one stable movement form to another occurs without stable intermediate movement states. New forms of movement emerge as a result of changes in critical values of one parameter. An often cited example of this proposition is the change of the quadruped gait of the horse from walking to trotting.(36) in this example, the parameter that shifts the horse from walking to trotting is considered an increase in speed, and thus muscle power. Another example is infant kicking. When the infant is in a sleepy or drowsy state, little kicking is noted. As the infant becomes more aroused, the spatial and temporal pattern of kicking is observed. In the crying state, a new pattern emerges, a rigid coactivation of all the muscles into stiff immobility. Thus, behavioral state is considered the parameter that drives the system from little kicking to reciprocal kicking to coactivation of the muscles.(35,37) In dynamical terms, a parameter that shifts the movement from one form to another is called a control parameter. The word "control" does not mean a prescription for movement change, but rather a reorganization of the movement.(21) Control parameters may reside in the infant (eg, behavioral state), the environment (eg, gravity), the social environment (eg, the caretaker), or the goal or task.(21) During transitions, the preferred coordinative pattern becomes less stable and more easily perturbed by control parameters.

A dynamical systems view of development no longer allows us to give the CNS the preeminent role of change in behavior. There is no single cause or predetermined model, be it genetic, neurological, cognitive, or environmental, for behavioral change. No longer can we consider the infant and young child as passive recipients of information from the environment. We should view them as active participants in which movement self assembles from the many subsystems within the environmental context. Instead of reflexes and reactions demonstrating hierarchical organization and comprising the substrata of movement, we need to consider the concept of functional units of behavior (ie, coordinative patterns) that are heterarchically organized and dynamically interact with other subsystems to determine motor outcome. Although there is ample documentation of the sequence of motor development, both normal and abnormal, we do not yet understand how movement originates during development or how we can promote optimal movement for infants and young children who are at risk for movement dysfunction, Fundamental questions addressed by therapists who care for atypical or handicapped infants and children are as follows: (1) Is the movement of atypical or handicapped infants different from that of "normal" (typical or nonhandicapped) infants? (2) Is the development of movement of atypical or handicapped infants different from that of normal infants? and (3) What intervention or treatments can be used in the remediation of motor dysfunction for atypical infants? The dynamical systems theory may be useful as a theoretical model for providing new insights into the evaluation and treatment of atypical and handicapped infants. First, however, we need to determine whether the theory can be transferred or generalized to the understanding of movement and the development of movement of atypical infants. In discussing these issues, I will use the results of my research that address the dynamical systems perspective. The principles of dynamical systems were used to study intralimb kicking behavior in low- and high-risk preterm and full-term infants. Specifically, these studies asked the following questions: (1) Are movements organized in the preterm infant? (2) Do early forms of movement change with development? (3) What control parameters affect the outcome of movement at the different ages? (4) Are there differences in movement patterns between low-risk preterm infants and normal full-term infants? and (5) Are there differences in leg movements in low-risk versus high-risk preterm infants?

Methods Used in Current Studies Design

The research design of these studies was a longitudinal mixed design that addressed the description of infant leg movement, intralimb kicking, using kinematic analysis. The independent variables, identified as possible control parameters, were risk factors, age, environment, behavioral state, passive muscle tone, joint extensibility, and body-build measurements. The dependent variables were the kinematic variables of amplitude, peak velocity, duration and frequency of movement, and joint angles at the onset of flexion and at peak flexion.

Subjects

Forty-nine infants participated in these studies. Fifteen were full-term infants delivered at 39 to 41 weeks' gestational age (GA) who met the following criteria: (1) singleton birth; (2) vaginal vertex delivery; (3) appropriate for gestational age (AGA) with respect to head circumference, weight, and length; and (4) healthy as determined by a pediatric physical examination. Inclusion criteria for 10 low-risk preterm infants born at 34 to 36 weeks' GA were (1) singleton birth, (2) vertex presentation at birth, (3) AGA, (4) Apgar score of at least 4 at 1 minute and 7 at 5 minutes after birth, and (5) absence of any physical malformation. Full-term and low-risk preterm infants exhibiting severe respiratory distress, pneumothorax, apnea, cardiac failure, convulsions, intraventricular hemorrhage (IVH), hydrocephalus, perinatal asphyxia, chronic lung disease, or drug withdrawal were excluded from the study. Inclusion criteria for 24 high-risk preterm infants born at 34 weeks' GA or less were (1) singleton birth, (2) vertex presentation at birth, (3) AGA, and (4) IVH. High-risk preterm infants who exhibited drug withdrawal were excluded from the study. There were 5 infants with a grade I IVH, 8 with a grade 11 FVH, 5 with a grade Ill IVH, and 6 with a grade IV IVH. One infant with a grade IV IVH also exhibited hip dysplasia.

Instrumentation

Videography was used in these studies. A color video camera was positioned perpendicular to the top of the examining table to produce a lateral view of the infant's leg. The infant's leg movements were recorded on 1.27-cm (0.5-in) videotape, using a VHS portable video recording system. Tapes of the infants were transferred to 1.9-cm (0.75-in) videotapes, using a videocassette recorder with a recording speed of 60 frames/s. A video counter-timer (60 frames/s) was superimposed on the videotape.

Hand digitization of a selected movement sequence was conducted using a two-dimensional sonic digitizer. This process involved projecting the video image onto the video screen; placing the mounted sonic digitizer in front of the video screen; and touching the grid with a stylus, resulting in the XY coordinates being automatically transferred to a computer. The system is limited to the speed and accuracy of the human operator. My experience indicates that an experienced operator can convert an average of four coordinate XY pairs of 10 seconds of movement filmed at a speed of 60 frames/s in 2 hours.

Procedure

Full-term and low-risk preterm infants were initially videotaped for 3 minutes on the third day after birth. Lowrisk preterm infants were filmed again at 40 weeks' postgestational age (PGA) and at 4, 8, and 12 months' corrected age (PGA=GA+chronological age). High-risk infants were initially videotaped as soon as possible after birth after becoming medically stable. These infants were filmed again at 34 to 36 weeks' and 40 weeks' PGA and at 4, 8, and 12 months' corrected age. These ages were selected as possible time periods when transitions or phase shifts from one stable state to another may occur.

Infants were initially filmed in the neonatal intensive care unit (NICU) or the full-term nursery. Preterm infants were later filmed in the NICU or special infant care unit. Data collection after discharge was conducted in the physical therapy research laboratory, in the follow-up nursery clinic setting, or in the children's homes. To ensure awake infants, videotaping was conducted before feedings, when possible. With three exceptions, the right lower extremity was marked at the lateral border of the base of the fifth metatarsal head, the lateral malleolus, the lateral femoral condyle, and the lateral thigh at the hip area with a 0.64-cm (0.25-in) circle of dark blue tape affixed to a 1.27-cm (0.5-in) circle of white tape (Fig. 1). The left leg of three infants was filmed because of the presence of an intravenous line in their right leg at the time of filming. The infant's diaper was removed for marking and filming. After the infants were prepared for videotaping, they were allowed to kick spontaneously in the supine position for 3 minutes. No specific stimuli were presented to the infant to elicit kicking. The infant's head was supported in the midline position by the examiner's (CBH) hand. The other hand was placed on the infant's abdomen to maintain the trunk in the midline position. During the recording session, the infant's arousal state was recorded every 15 seconds on a six point scale (ie, I1asleep, 6=crying)(38) or on appropriate subdivisions of this scale for preterm infants.(39) After filming, passive muscle tone, joint extensibility, and body-build measurements were taken. Passive muscle tone was measured by the amount of (1) resistance to passive stretch of the leg flexors to extension and (2) recoil of the legs. joint extensibility was measured by the angles of the hip, knee, and ankle. Each joint angle was scored on an ordinal scale. Elicitation and criteria for passive muscle tone and joint extensibility were adapted from the clinical assessment tools of the Assessment of Preterm Infants' Behavior,(39) the Clinical Assessment of Gestational Age in the Newborn Infant,(40) the Neurological Assessment of the Preterm and Full-term Newborn Infant,(41) and the Neurological Evaluation of the Maturity of Newborn Infants.(42) Body-build measurements were taken of weight, crown-rump and crownheel length, and circumferences of the calf and thigh at their widest points. From these measurements, the ponderal index of Rohrer(43) and leg volume(44) were calculated. The ponderal index is a measure of relative stockiness and was calculated by dividing the infant's body weight (in grams) by the crown-heel length cubed and multiplied by 100. Leg volume was estimated by assuming the leg was a cylinder, with the height determined by crown-heel length minus crown-rump length and circumference estimated by averaging the thigh and calf measurements. The rates of growth with respect to weight, ponderal index, and leg volume were determined by dividing the difference between two ages by the measurement at the younger age.

Data Reduction

Behavioral records were analyzed in two ways. First, the 3-minute records were analyzed with respect to the frequency of kicking. A kick was defined as hip and knee flexion that brought the knee closer to the chest. For a new kick to be counted, the knee must have reached a minimum of 90 degrees of extension before returning to a flexed position. The videotape was reviewed without sound so that the coder was not influenced by the arousal level of the infant.

Second, joint-angle changes for a 10second segment of continuous movement were coded. The samples were selected by viewing the entire 3-minute kicking session of each infant without sound and choosing a representative 10-second segment of continuous movement(29,30,45) or the first 10 seconds of continuous movement,(46) with the following restrictions: (1) The infant's legs were not out of the sagittal plane, or movement outside the sagittal plane was minimal; and (2) all four joint markers were visible.

The 10-second segment was then coded frame by frame (1 frame= 162/3 milliseconds), on a video recorder with a recording speed of 60 frames/s. Six hundred frames of movement were coded. Coders were trained to reliability (ie, intraclass correlation coefficient [2,11 >.80) prior to digitizing. The sonic digitizer was used to determine the XY spatial coordinates of the marked hip, knee, ankle, and fifth toe. joint angles of the hip, knee, and ankle were calculated from the coordinate data (Fig. 1). Because these joint angles reflect the relationships among body segments, they are not analogous to clinical goniometric measurements. In this study, 180 degrees was regarded as full extension at each joint (eg, full plantar flexion of the ankle). These joint angles were used as reference points to generate kinematic data of amplitude, velocity, and duration of movement.

I believe that measurement error, introduced into the system as a result of infant variability and the measurement and conversion process, was minimized by (1) selection of the 10second segment, (2) adhering to sound videographic procedures, (3) adequate training of personnel during the measurement and conversion process, and (4) subjecting the raw angle data to Fourier analysis and filtering the data using Butterworth-type lowpass digital filters.(47) Fourier analysis was used to quantify the frequency domains of both the signal XY coordinate data of kicking) and the noise (error measurement) to identify the frequency cutoff to be used to separate the signal from the noise. This cutoff frequency was used to develop Butterworth-type low-pass digital filters.(48) These low-pass filters selectively filter out the high-frequency noise end of the frequency spectrum while passing over the low-frequency components of the signal (kicking). The resultant computer readout indicated the frame number; the raw and filtered angles of the hip, knee, and ankle; the instantaneous velocity of the movement of the three joints; and the sequential time of the 10-second movement segment. Graphic displays of the kinematic data were generated. Graphs included time-dependent representations in which the patterns of joint displacement were plotted as a function of time (Fig. 2, top panel) and time independent representations (ie, amplitude-velocity phase plane trajectories of the knee [Fig. 2, bottom panel], and hip and ankle). These graphs resulted in visually permanent records of the movement that were used for evaluation purposes, compared with other movement sequences of the same infant and compared with movement patterns of other infants.

These plots provided useful interpretative data about the movement path of a body segment with respect to time as well as the movement path's velocity, direction, and magnitude.(49) The smoothed data were analyzed according to predetermined criteria.(50) The kick cycle was characterized by four phases. The flexion phase of the kick cycle lasted from the frame at which continuous leg movement (for at least five frames) in a horizontal plane toward the body was first noticed until the frame at which movement stopped or changed direction. The intrakick pause was the time interval between the cessation of the flexion phase and the initiation of the extension phase. In the extension phase, the infant's foot moved continuously away from the body until horizontal movement ceased. The interkick pause was the time interval between the end of extension and the initiation of the next flexion phase. The duration of each phase of the kick cycle was calculated. In addition, the phase lags, or time between the onset or termination of the movement of one joint and that of another joint, were calculated and normalized by the kick period, defined as the average duration of completed kick cycles. The peak velocities of both the flexion and extension movements were recorded. The joint angles at the onset and end of the flexion and extension movements were recorded, and the amplitude excursion was calculated.

Summary and Discussion of Results of Current Research

The following summary and discussion represent data obtained for infants at 34 weeks' GA and at 40 weeks' GA or PGA and for one high risk preterm infant through 12 months' corrected age. Details of the data for full-term and low-risk preterm infants are published elsewhere.(29,30) Preliminary data on highrisk infants have been presented at meetings(46,51) and are currently being readied for submission for publication. The summary of the results will be presented according to the principles of the dynamical systems theory discussed previously.

Assumption 1. Developing organisms are high-dimensional systems, and movement represents a compression of the multiple degrees of freedom. Are the movement patterns of intralimb kicking of atypical infants different from those of normal infants?

Are the collective variables for these movements the same or different? Intralimb kicking in all infants was similar, regardless of age, environment, or risk factors.(29,30,46) The multiple degrees of freedom were reduced to that of a few representing the observable movement of intralimb kicking, a coordinated pattern. The top panel of Figure 2 shows the conventional angle-time displacement of the joint angles of the hip, knee, and ankle plotted as a function of time for four infants at different GAs and PGAS. Although visual analysis indicates some differences in the movement trajectories among the infants, the pattern of intralimb kicking is similar. The quality of the visual analysis of the intralimb coordination can quantitatively be indexed by the following collective variables: (1) the relationships of individual joints to each other, (2) the phase lags between joints during kicking, and (3) the timing of the movement phases. The relationships among the joints (three pairs for each infant: hip and knee, hip and ankle, and knee and ankle) were strong.

In all infants, the hip-knee relationship was especially strong: When the hip flexed, the knee flexed, and when the hip extended, the knee extended. Thus, the movement of the joints of one leg are constrained to act as a unit of movement. A system comprising many degrees of freedom is reduced to only a few. The close synchrony of movements denoted by the high interjoint correlations was confirmed by small phase lags between the movements of the hip, knee, and ankle joints, demonstrating that the joints moved in near-perfect unison.(29,30) Another component of movement that reflects temporal stability of time is the timing of the flexion and extension phases of the kick cycle. In all infants, the timing of the movement phases were similar, with the extension phase being longer than the flexion phase. In addition, there was a trend for the younger or sicker infants to have longer flexion and extension phases. Thus, for intralimb kicking, the collective variables that described the movement pattern of intralimb kicking were strength of the joint correlations, phase lags between joints, and the movement time of flexion and extension.

All infants, regardless of age, environment, or risk factors, demonstrated a coordinative pattern of kicking, indexed by collective variables that compressed the multiple degrees of freedom to a few, representing kicking. Thus, when grouped, the kicking behaviors of preterm infants (low- and high-risk) are similar to the kicking behaviors of normal infants and can be described by the same collective variables: joint correlations, phase lags, and timing of the movement phases. Although the high-risk preterm infants, as a group, demonstrated organized movement at 40 weeks' PGA, some infants showed individual profiles in which kicking appeared disorganized (Fig. 3). The time-dependent diagrams (Fig. 3, top panel) show that the ankle is out of phase with the knee and hip in infant A, a preterm infant with a grade IV IVH, and that the knee is out of phase with the hip and ankle in infant B, a preterm infant with a grade IV IVH and hip dysplasia. Although the timing of the movement phases of flexion and extension were similar to that of other infants, the correlations between the joints were weak and the phase lags were long. This was especially true for the hip ankle and knee-ankle relationships for infant A and for the hip-knee and knee-ankle relationships for infant B. Therefore, two order parameters that described the coordinative pattern of early intralimb kicking movements and that differentiated between organized and disorganized movement patterns (ie, joint correlations and phase lags between joints) could be used as evaluative tools for kicking.

Assumption 2. Movement emerges in a seff-organizing fashion as a function of the cooperation of the many subsystems in a task-specific context. Although the pattern of kicking was stable among low-risk preterm and full-term infants, differences were found in the amplitude and velocity of movement, the pauses during the kick cycle, and the joint angles at the initiation of flexion and at peak flexion.(29,30) With age, low-risk preterm infants showed decreased movement amplitude and peak velocity (Figs. 2A, 2B).(37) Older preterm infants also demonstrated a trend to kick more in 3 minutes than younger preterm infants, reflective of shorter pauses. At 40 weeks' PGA, all low-risk preterm infants were more extended at all joints, especially the ankle, in comparison with full-term infants (Figs. 2A, 2B; top panel).(30) These same infants also paused more during kicking in 3 minutes in comparison with full-term infants, resulting in longer kick periods and a trend toward less frequent kicking.

If the movement pattern is the same, what explains these differences? According to the dynamical systems theory, movement outcome is not determined by a pattern of strict muscle activity alone but by the pattern of neuromuscular activity and by mechanical and dynamic considerations such as measures of body build, passive viscoelastic properties of muscles, and muscle strength. Movement outcome is considered the result of the cooperative interaction of all the participating subsystems in developmental and real time and is not coded a priori anywhere in the nervous system. Because there are no preexisting instructions for defining intralimb kicking, the order of the kicking pattern is the result of the self organization of the participating subsystems. Kicking spontaneously arose from the interaction of the components.

Stepwise regression analysis of arousal level, passive muscle tone, joint extensibility, and measurements of body build on the kinematic variables of duration, amplitude, and velocity did explain differences in movement seen in infants between 34 to 36 weeks' GA and 40 weeks' PGA and between low-risk preterm infants at term equivalent age and full-term infants (CB Heriza, E Thelen; unpublished research).(45,52,53) Arousal level influenced the frequency of kicking and the joint angles at peak flexion (CB Heriza, E Thelen; unpublished research).(45) Infants who were highly aroused demonstrated short pauses, resulting in short kick cycles and an increase in the frequency of kicking. These same infants showed small joint angles at peak flexion. Body build also affected movement outcome (CB Heriza, E Thelen; unpublished research).45,52 Infants who became stocky and heavy rapidly or who developed chubby legs kicked less. These infants also demonstrated small-amplitude excursions associated with small peak velocities and large angles at the beginning of flexion and at peak flexion. These results may reflect asynchronous development of muscle strength and body fat(44) such that infants who were fat or who became fat quickly did not have the muscle strength to kick their heavier, stockier legs. Passive muscle tone influenced the interkick pause and joint angles at the beginning of kick flexion (CB Heriza, E Thelen; unpublished research).(45,53) Infants who had increased muscle tension demonstrated short pauses and small joint angles at the beginning of kick flexion.

Possible interpretations of the differences in kicking outcome are as follows: (1) Increased arousal level, which presumably is reflected in more energy delivered to the muscles and more motor output, including kicks, contributed to small joint angles at peak flexion and short pauses in the kick cycle, resulting in increased frequency of kicking; (2) with age, the larger masses of the limbs in a gravity-controlled environment contributed to decreased movement amplitude and velocity and inhibited the extreme flexion of the youngest preterm infants; and (3) the long confinement in the intrauterine space for full-term infants may bias muscle and joints toward flexion, contributing to the flexor dominance of the full-term infants. Movement change with age in low-risk preterm infants and differences in kicking between low-risk preterm infants at 40 weeks' PGA and full-term infants, therefore, may be largely nonneural and not caused by maturational changes in the CNS. The differences in kicking reflect the assumption of the dynamical systems perspective that the outcome of the movement is the result of the dynamical interaction of the subsystems in real and developmental time within a context, including the physical constraints and supports of the environment. At 34 to 36 weeks' GA, 3 days postintrauterine environment, the various subsystems dynamically organize in real time to produce infant kicking; at 40 weeks' PGA, these same subsystems, which have matured and experienced 6 weeks of extrauterine environment, self-organize to produce kicking; and at term 40 weeks' GA, these same subsystems, which have experienced the intrauterine environment for an additional 6 weeks, self organize for kicking. Thus, although the pattern of coordination of intralimb kicking of preterm infants at different ages and of full-term infants is similar, the context varied between time spent in the intrauterine or extrauterine environment and the components of the system (ie, body build, arousal level, passive muscle tone, and joint extensibility) varied with age. These variations may explain differences seen among the infants.

Assumption 3. Movement patterns are preferred, but not obligatory, and occupy preferred regions of their state space.

Are the movement patterns of intralimb kicking of atypical infants different from those of normal infants? Are these movement patterns preferred, and do they occupy preferred regions of their state space? The intralimb coordinated behavior of kicking can be described in terms of phase-plane trajectories of the movement by plotting kicking as a function of the position and velocity of a joint. The bottom panel of Figure 2 shows the phase-plane trajectories of the knee of the representative infants at different GAs and PGAS.

Visual analysis indicates that the movement trajectories of the knees of all infants are confined to a particular region of the plot and show a particular ordered pattern, having the form of simple closed curves with oscillations or loops at the beginning and end of the movement phases. The amplitudes and peak velocities of the kicking movements for each infant are similar, with the exception of the high-risk infant's movements, which show variability of amplitude and peak velocity for the four kicks. There are no self-intersections (ie, no self crossings or reversals of movement at zero velocity) during the flexion and extension movements. In other words, intralimb kicking can be said to occupy preferred, but not obligatory, regions of the state space. All infants, regardless of age, amount of time spent in the extrauterine environment, or the risk factor, demonstrated a preferred configuration of movement within the state space. Such preferred, repetitive cyclic behavior of kicking can be considered a stable attractor state.(54) Although the 40-week PGA high-risk preterm infants, as a group, demonstrated preferred behavioral patterns that acted as stable attractors, some infants showed individual profiles. Phase-plane diagrams of two of these infants Fig. 3, bottom panel) demonstrate differences in their state space. Infant A exhibited a grade Iv IVH; infant B exhibited a grade IV IVH and hip dysplasia, in general, the kicks show simple closed curves with oscillations at the extremes of the movement. There is variability in the amplitude and the velocity of the kicking movements in these infants, and, in contrast to the phase-plane plots of other infants, self-intersecting loops are present (ie, self-crossings or reversals of movement at zero velocity). These loops vary in size and are proportional to the amplitude or pace of the retrograde motion. These reversals tend to occur during the extension phase of the movement, especially in infant A. Thus, these movement paths appear to occupy various portions of the state space, rather than the preferred regions of the space, and the movement pattern could perhaps be said to be chaotic. High-risk infants are being followed to a corrected age of 12 months. It will be of interest to see whether the movement pattern remains disorganized or becomes adaptable for higher levels of function.

Assumption 4. New behavioral forms emerge in development as a series of phase shifts.

Is the development of movement of atypical or handicapped infants different from that of normal infants? One high-risk preterm infant has been followed during the first year after birth.(51) This infant, who was born at 26 weeks' GA and had a documented grade W IVH, showed organized movement at 40 weeks' PGA, as indexed by strong joint correlations. The infant's ankle, knee, and hip joints were closely synchronized in the kicking movement in that all joints moved in unison (ie, they moved into flexion together and into extension together) (Fig. 2D, top panel). Although this infant's close synchrony of the joints began to diminish, resulting in greater individual joint action, at 4 months' corrected age, as is seen in full-term infants,(31) the infant's joint correlations became stronger at 8 months' corrected age, with the hip and knee correlation remaining strong at 12 months' corrected age. The kicking pattern stayed close to the preferred pattern at 40 weeks' PGA. In this case, the pathology may be defined as maintenance of an early preferred attractor state with the inability to uncouple the early coordinated movement pattern. The infant did not show the expected pattern change with development. A critical question is, why not? What control parameters, be they intrinsic or extrinsic, prevented this transition, or what control parameters were missing? Future research is needed to answer these questions.

Summary

These data suggest that the dynamical systems perspective to understanding movement can be applied to atypical populations. The patterns of coordination seen in intralimb kicking demonstrated that a high-dimensional system with many degrees of freedom can be reduced to a unit of behavior with few degrees of freedom. Collective variables that indexed the system were joint correlations, phase lags, and the timing of the movement phases of kicking. These parameters did differentiate among normal and atypical infants. All components of the system (ie, musculoskeletal component, neurological component, arousal level, body build, passive elastic properties of muscles, and muscle strength) self-organized to produce preferred movement patterns in real and developmental time. These preferred movement patterns were assembled within the constraints of the immediate environment. These preferred movement patterns occupied preferred regions of the state space and thus can be considered stable dynamic attractors.

For intralimb kicking, movement dysfunction may have resulted because (1) the muscles and joints of one leg may not have been organized as a preferred movement pattern because additional variables interfered with the natural timing of the movement or (2) one or more control parameters may have rate-limited or masked another component in real or developmental time such that, in at least one infant, transitions to more mature movement patterns were prevented. Although the components of the system dynamically self-organized, the movement pattern may have been too restricted in some infants and too random and disorganized in others.

Clinical implications

The implications of the dynamical systems theory for the atypical or handicapped infant and child are just now being explored. The understanding of how this theory influences the evaluation and treatment of these infants is in its infancy. The theoretical constructs of the dynamical systems theory with respect to the development of movement, however, reflect the necessity to review and perhaps reinterpret traditional concepts of motor development.

Motor development is not prescriptive or hierarchical, and changes in movement with age cannot be attributed solely to maturation of the CNS. Rather, movement is emergent from the dynamic interaction of all subsystems within a task-specific context. In this view, the CNS is seen as a necessary, but not sufficient, component to explain movement changes. Other important subsystems are the infant's biomechanical, psychological, and social environments.

If function rather than instruction drives behavior, the infant and young child should be considered as active participants in development rather than passive recipients of environmental stimuli. Although reflexes and automatic reactions have been traditional evaluation tools, and indeed the movement pattern is probably the same as that seen in spontaneous movement,(55) the concept of stimulus response reflects the passive nature of such an approach. Instead of causing the response, the stimulus facilitates the production of motor patterns that also occur as spontaneous movement.(55) The task or goal within the environment, with its constraints and supports, assembles behavior as opposed to being prescribed a priori by the CNS. This does not mean that we discard the concept of reflexes and reactions, but it does mean that we expand our concepts of movement development such that we consider other factors that influence movement in addition to external stimuli. A new synthesis of motor development should build on what we know; we should not discard our database, but incorporate it and go beyond.(56,57) We must address the process of movement changes versus the end product of movement.(21,56) Motor milestones are important because they provide a measure of how the infant and young child are developing with respect to normative data and they provide information on function. Motor milestones, however, do not tell us how movement changes. We need to develop evaluation tools to investigate the process of movement change. The identification of collective variables may provide alternative ways to assess movement changes that occur in development or as the result of therapeutic intervention. Additionally, we need to utilize different research strategies to identify these changes. Traditional group designs often miss individual differences and tell us nothing of the process of change. Longitudinal studies as well as case studies may address these issues.(21,56) We may not be able to identify all infants at-risk for movement dysfunction early, because early movement, with respect to intralimb kicking, showed invariant patterns across ages, risk factors, and environments. Some infants may not be identified until phase transitions when infants are expected to disassociate and reorganize basic coordinative patterns to form functional movement synergies within the environment. We need to identify the key transitional phases and the key control parameters that constrain or facilitate movement. Are the transition periods and control parameters hypothesized by Thelen(22) for normal infants the same for atypical infants? Which parameters are most powerful at specific transition periods? If identified, can physical therapists utilize these parameters for intervention with infants with atypical movement? For example, can we perturb the system to promote disassociation of the rigid attractor state? What variables can we use to shift the system to a more functional adaptive synergy.? What are the parameters in the social and environmental contexts that influence leg movements during the first year after birth? Can we use these parameters in intervention programs? If control parameters are most influential at transitional stages,(21) is it imperative for us as therapists to intervene at these ages? It may not be as important at other ages during nontransitional stages. How early can infants with atypical movement be identified? What are the optimal time periods to evaluate? Are there different types of abnormal movement patterns (eg, obligatory preferred attractor states, disorganized attractor states)? If so, are different control parameters involved in the different types of movement dysfunction with respect to constraining developmental progression?

What are the particulars? Explanation awaits descriptive and experimental manipulations to identify crucial points of transitions and to identify control parameters that cause the system to seek new coordinative functional synergies. Acknowledgments I wish to express my appreciation to Mary Lou Barnes, EdD, PT; Carolyn Crutchfield, EdD, PT; julie High, MS, PT; Randy Richter, MS, PT; and john Scholz, PhD, PT, for their valuable comments and suggestions on an earlier version of this manuscript. References 1 McGraw MB. Tbe Neuromuscular Maturation of the Human Infant. New York, NY: Hafner Press; 1945. 2 Gesell A. The ontogenesis of infant behavior. In: Carmichael L, ed. Manual of Child Psychology. 2nd ed. New York, NY: john Wiley & Sons Inc; 1954:335-373. 3 Forssberg H. Ontogeny of human locomotor control, 1: infant stepping, supported locomotion and transition to independent locomotion. Exp Brain Res. 1985;57:480-493. 4 Bobath K. The Motor Deficit in Patients with Cerebral Palsy. Philadelphia, Pa: JB Lippincott Co; 1966. 5 Fiorentino MR. A Basis for Sensorimotor Development: Normal and Abnormal Springfield, Ill: Charles C Thomas, Publisher; 1981. 6 Capute AJ, Accardo Pj, Vining EPG, et al. Primitive Reflex Profile. Baltimore, Md: University Park Press; 1978. 7 Capute AJ, Shapiro BK, Accardo Pj, et al. Motor functions: associated primitive reflex profiles. Dev Med Child Neurol 1982;24:662-669. 8 Kottke Fj. From reflex to skill: the training of coordination. Arch Phys Med Rebabil. 1980;61:551-561. 9 Zelazo PR, Weiss Mj, Leonard E. The development of unaided walking: the acquisition of higher order control. In: Zelazo PR, Barr RG, eds. Challenges to Developmental Paradigms.. implications for Theory, Assessment and Treatment. Hillsdale, Nj: Lawrence Erlbaum Associates Inc; 1989:139-165. 10 Bernstein N. Coordination and Regulation of Movements. New York, NY: Pergamon Press Inc; 1967. 11 Whiting HTA, ed. Human Motor Actions: Bernstein Reassessed. New York, NY; Elsevier Science Publishing Co Inc; 1984. 12 Haken H. Synergetics, an Introduction: Non-equilibrium Phase Transitions and Set(Organization in Physics, Chemist and Biology 3rd ed. New York, NY: Springer-Verlag New York Inc; 1983. 13 Kelso JAS, Holt KG, Kugler PN, Turvey MT. on the concept of coordinative structures as dissipative structures, II: empirical lines of convergence. In: Stelmach GE, Requin J, eds. Tutorials in Motor Behavior Amsterdam, the Netherlands: Elsevier Science Publishers BV; 1980;49-70. 14 Kelso JAS, Tuller B. A dynamical basis for action systems. In: Gassaniga M, ed. Handbook of Cognitive Neuroscience. New York, NY: Plenum Publishing Corp; 1984:321-356. 15 Kugler PN, Turvey MT: Information, Natural Law and the Self-Assembly of Rhythmic Movement. Hillsdale, Nj: Lawrence Erlbaum Associates Inc; 1987. 16 Kugler PN, Kelso JAS, Turvey MT. On the concept of coordinative structures as dissipative structures, 1: theoretical lines of convergence. In: Stelmach GE, Requin J, eds, Tutorials in Motor Behavior Amsterdam, the Netherlands: Elsevier Science Publishers BV; 1980:3-47. 17 Kugler PN, Kelso JAS, Turvey MT. On coordination and control in naturally developing systems. In: Kelso JAS, Clark JE, eds. The Development of Movement Coordination and Control New York, NY: john Wiley & Sons Inc; 1982:5-78. 18 Kelso JAS, Schoner G. Self-organization of coordinative movement patterns. Human Movement Science. 1988;7:27-46. 19 Schoner G, Kelso JAS. Dynamic pattern generation in behavioral and neural systems. Science. 1988;239:1513-1520. 20 Scholz JP. Dynamic pattern theory: some implications for therapeutics. Phys Ther. 1991;71:827-843. 21 Thelen E. Self-organization in developmental processes: can systems approaches work? In: Gunnar M, Thelen E, eds. Systems and Development.. The Minnesota Symposium on Child Psychology. Hillsdale, Nj: Lawrence Erlbaum Associates Inc; 1989;22:77-117. 22 Thelen E. Development of coordinated movement: implications for early human development. In: Wade MG, Whiting HTA, eds. Motor Development in Children: Aspects of Coordination and Control. Dordrecht, the Netherlands: Martinus Nijhoff/Dr W junk Publishers; 1986:107-120. 23 Thelen E, Fogel A. Toward an action-based theory of infant development. in: Lockman J, Hazen N, eds. Action in Social Context. New York, NY: Plenum Publishing Corp; 1989: 23-63. 24 Thelen E, Kelso JAS, Fogel A. Self organizing systems and infant motor development. Developmental Review. 1987;7:39-65. 25 Kamm K, Thelen E, jensen JL. A dynamical systems approach to motor development. Phys Ther. 1991;71:763-775. 26 Easton TA. On the normal use of reflexes. American Scientist. 1972;60:591-599. 27 Kelso JAS, Southards DL, Goodman D. On the nature of human interlimb coordination. Science, 1979;203:1029-1031. 28 Turvey MT. Preliminaries to a theory of action with reference to vision. In: Shaw R, Biansford J, eds. Perceiving, Acting, and Knowing.- Toward an Ecological Psychology. New York, NY: john Wiley & Sons Inc; 1977:211265. 29 Heriza CB. Organization of leg movements in preterm infants. Phys Ther. 1988;68:13401346. 30 Heriza CB. Comparison of leg movements in preterm infants at term with healthy full term infants. Phys Ther. 1988;68:1687-1693. 31 Thelen E. Developmental origins of motor coordination: leg movements in human infants. Dev Psychobiol 1985;18:1-22. 32 Clark JE, Truly TL, Phillips Sj. A dynamical systems approach to understanding the development of lower limb coordination in locomotion. In: Block H, Bertenthal B, eds. Sensory-Motor Organization and Development in Infancy and Early Childbood Hingham, Mass: Kluwer Academic Publishers. In press. 33 Clark JE, Whitall J, Phillips Sj. Human interlimb coordination: the first 6 months of independent walking. Dev Psychobiol 1988;21: 445-456. 34 Thelen E. Treadmill-elicited stepping in seven-month-old infants. Child Dev. 1986;57:1498-1506. 35 Thelen E, Ridley-johnson R, Fisher DM. Shifting patterns of bilateral coordination and lateral dominance in the leg movements of young infants. Dev Psychobiol. 1983;16:29-46. 36 Tuller B, Turvey MT, Fitch HL. The Bernstein perspective, 11: the concept of muscle linkage or coordinative structure. in: Kelso JAS, ed. Human Motor Behavior.. An Introduction. Hillsdale, Nj: Lawrence Erlbaum Associates Inc; 1982:253-270. 37 Thelen E, Fisher DM, Ridley-johnson R, Griffin Nj. The effects of body build and arousalon newborn infant stepping. Dev Psychobiol. 1982;15:447-453. 38 Brazelton TB. Neonatal Behavioral Assessment Scale. 2nd ed. Philadelphia, Pa: JB Lippincott Co; 1984. 39 Als H, Lester BM, Tronick EZ. Toward a research instrument for the assessment of preterm infants' behavior (APIB). In: Fitzgerald HE, Lester BM, Yogman MW, eds. Theory and Research in Behavioral Pediatrics. New York, NY: Plenum Publishing Corp; 1982;1:35-132. 40 Dubowitz LMS, Dubowitz V, Goldberg C. Clinical assessment of gestational age in the newborn infant. j Pediatr. 1970;77: 1-10. 41 Dubowitz LMS, Dubowitz V. The Neurological Assessment of the Preterm and Full-term Newborn Infant. Philadelphia, Pa: JB Lippincott Co; 1981. 42 Amiel-Tison C. Neurological evaluation of the maturity of newborn infants. Arch Dis Child. 1968;43:89-93. 43 Brandt J. Postnatal growth of preterm and full-term infants. In: johnston FE, Roche AF, Suzanne C, eds. Human Physical Growth and Maturation. New York, NY: Plenum Publishing Corp; 1980:139-159. 44 Thelen E, Fisher DM, Ridley-johnson R. The relationship between physical growth and a newborn reflex. Infant Behavior and Development. 1984;7:479-493. 45 Heriza CB. A Kinematic Analysis of leg Movements in Prematature and Full-Term Infants. Edwardsville, Ill: Southern Illinois University at Edwardsville; 1986. Doctoral dissertation. 46 Heriza CB. Kinematic analysis of spontaneous leg movements in preterm infants with documented hemorrhage. Presented at the 64th Annual Conference of the American Physical Therapy Association; june 11-15, 1989; Nashville, Tenn. 47 Thelen E, Fisher DM. The organization of spontaneous leg movements in newborn infants Journal of Motor Behavior. 1983;15:353377, 48 Winter DA. Biomechanics of human Movement. New York, NY: john Wiley & Sons Inc; 1979. 49 Harris SA, Heriza CB. Measuring infant movement: clinical and technological assessment techniques. Phys Ther. 1987;67:18771880. 50 Thelen E, Bradshaw G, Ward JA. Spontaneous kicking in month-old infants: manifestation of a human central locomotor program. Behav Neural Biol 1981;32:45-53. 51 Heriza CB. Developmental changes in leg movements during the first year in an infant with TVH. Presented at the 64th Annual Conference of the American Physical Therapy Association; june 11-15, 1989; Nashville, Tenn. 52 Heriza CB. Effect of body build on infant kicking. Presented at the 53rd Annual Conference of the American Physical Therapy Association; June 28-july 2, 1987; San Antonio, Tex. 53 Heriza CB: Effects of passive muscle tone and joint extensibility on infant kicking. Presented at the 1988 joint Congress of the American Physical Therapy Association and the Canadian Physiotherapy Association; june 12-16, 1988; Las Vegas, Nev. 54 Thelen E. On the nature of developing motor systems and the transition from prenatal to postnatal life. In: Smotheran W, Robinson SA, eds. Fetal Behavior. Bloomfield, Nj: The Telford Press. In press. 55 Wolf PH. The Development of bebavioral States and the Expression of emotions in Early Infancy. Chicago, Ill: University of Chicago Press; 1987. 56 Horowitz FD. Exploying Developmental Theories: Toward a Structural/Behavioral Model of development. Hillsdale, Nj: Lawrence Eribaum Associates Inc; 1987. 57 Thelen E. The role of motor development in developmental psychology: a view of the past and an agenda for the future. In: Eisenberg N, ed. Contemporary Topics in Developmental Psychology. New York, NY: john Wiley & Sons Inc; 1989:3-33.