Perspective on Variability in the Development of Human Action

Variability in Coordination During Changing Body Metrics

Humans encounter dramatic changes in the metrics of body size and relational metrics among body segments during early development. These metrics are changing as the infant is exploring and attempting mastery over patterns for mobility and other functional actions. Infants are initially “top heavy,” with large heads and upper bodies in comparison with their lower trunks and legs.² A newborn's head is 25% of the body length,³ which poses potential challenges for dynamically controlling the head. Head control is developing while the head circumference and weight are dynamically changing.

Bartlett³ investigated the effects of anthropometrics on gross motor skills during the first year of life. At 6 weeks of age, infants with proportionally larger heads (head size to body mass) had lower motor skills, specifically lower skills in prone head lifting. Body proportions change and the center of mass lowers as infants mature into toddlers. Infant weight, on average, more than doubles in the first year of life, and length increases, on average, approximately 50% while head circumference increases approximately 30%.⁴ However, toddlers remain top heavy as they stand and begin to take steps while managing a large head at the top of the body. The width of the head is greater than the width of the pelvis⁴ when an infant begins to toddle, and initially the infant has a wide base of support and makes variable steps and strides.

The coordination pattern that emerges from toddlers as they begin to walk is, in part, a function of the relation of head size to pelvis. These relational metrics between head and pelvis, however, are dynamically changing while walking is mastered. The walking patterns that emerge under these changing task demands will, of necessity, be variable—the constraints of the task are variable. Changes in movement capabilities are required to accommodate the metric changes, as well as to meet new environmental challenges.

Although these relational metrics contribute to variability in action, the relation of body metrics and motor development is not a one-to-one mapping. There is correlational evidence suggesting that top-heavy infants are later walkers,⁵ but the study by Bartlett³ and other studies^4–6 have failed to find relations between many of the tested anthropometrics, when considered independently, with crawling or walking. In order to fully address the relation of infant and environmental constraints on the emergence of patterns of mobility, studies need to be conducted that take into account the multiple constraints on the emergence of mobility patterns such as changes in strength, motivation, and cognition.

The plasticity of human biological systems supports the adaptation of our actions to these dynamic changes. This adaptive process requires variability in performance in order for the most optimal solutions for action to emerge.

Variability in Coordination During Changing Task Demands

The dynamic action of adults, as well as infants and toddlers, must accommodate a wide range of demands even when the metrics of change have stabilized or movement patterns have been mastered. Envision placing a heavy and expensive object on top of your head that is wider than your pelvis. Your first movement would most likely be to widen your base of support. Your first steps would be tentative as you explore your strength and coordination for this new task. Although you have mastered walking, the task context has now changed, and the constraints of this new task force exploration and adaptation of your actions in order to keep the precious object from falling. In creative experiments with infants, Adolph and colleagues^5–8 demonstrated that infants who had mastered walking used adaptive variation in coordination patterns under changing task demands. With a backpack loaded with feathers, infants navigated down a 20-degree sloping surface, but when the backpack was lead-weighted, infants refused to walk down the slope.⁶ In another study, experienced walkers easily navigated independently across wide bridges, but would only cross a narrow bridge if provided a sturdy, wooden handrail. Indeed, they reduced walking across the narrow bridge if the handrail was changed from sturdy wood to wobbly rubber.⁸ Details of the environment as well as feedback and consequences of our actions yield information about the appropriate action patterns and concomitant required forces and timing of forces that produce the most optimal coordination to achieve a goal. Goals for adults in their workplaces may include mastery of a wide variety of actions, including rapid computer keyboard skills, heavy lifting for construction, or precise eye-hand coordination for robotic surgery. Human action over a lifetime and within constraints of person, culture, geography, and climate requires adaptable systems with potential for abundant variability.

Typical Coordination Emerges Within Limits of Variability

Variability, however, must have definable limits. The number of possible task solutions available to us far exceeds the number of solutions that we typically use for action.^1,9,10 We reduce the scope of possible task solutions with the use of repetition of action and the development of perception of commonality across task environments. What we actually reduce and how this reduction occurs is an open debate, but the reduction of possible solutions to probable solutions and the reduction of variability within a solution appear to be common assumptions of theories of human action.^10–12 Although a set of probable solutions to daily actions may emerge, individuals who are healthy maintain a range of possible solutions that can be produced under changing task demands. For example, although I may have a typical walking speed, this speed can be increased (same pattern scaled up on speed) if I want to walk beside a faster-walking friend. However, the friend may be walking too fast, and scaling up on the speed of my walking pattern may not suffice; I may have to change strategies and begin jogging.

A coordinated person can continue to walk across a parking lot even if the surface changes from asphalt to stones. The action of walking continues, but we vary the characteristics within the act of walking. The typically coordinated person changes action patterns within the constraints of a new task environment. Kelso et al¹³ showed that increasing the speed of a treadmill will create a sudden switch from a walking pattern to a running pattern at a critical threshold of speed. In this experiment, the treadmill speed constrains the possible solutions to locomotion while remaining upright. Bilateral locomotion continues on the treadmill, but the locomotory pattern changes in form from walking to running. In the experiment by Kelso and colleagues and in other experiments, the variability within the pattern increases, that is, the pattern begins to become unstable and irregular just prior to and shortly after the pattern change occurs.^14–16 This instability can be seen, for example, as an increase in standard deviations of measured variables. The new pattern may have more variability immediately after the change and then “settles” into the new pattern, as evidenced by reduced variability in measured variables.^12–14

Variability Allows Exploration: Exploration Affords Skill Development

The processes at work in achieving task-specific actions that can be reproduced automatically and typically with reduced variability are processes that occur not only during development, but I would argue, these developmental processes also support the acquisition of skilled action across the life span. These life-span processes of skill development are dependent on the constraints of the individual, the environmental context for action, and the task that emerges at the interface of the individual in the environment.^17–20 Gibson^21–23 and other proponents^24–26 of the perception-action theory of human action argue that the exploratory actions of human infants provide the necessary variability through which infants explore the dynamics of their actions and the associated consequences. It is through this exploration of the constraints on action and the resultant variability and possible “error” that is produced that infants learn actions that succeed and actions that fail.

Observations on the acquisition of walking support the extraordinary amount of variable practice necessary to master walking during a time of rapid infant growth and changes in perception and cognition. Adolph²⁷ documented that each walking hour, 14-month-olds complete 9,000 steps, which is the approximate length of 29 football fields, and they experience 15 falls. All of this action occurs under the constraints of developing perception, cognition, and body metrics. This exploration of the dynamic variability of their movements provides infants with the embodied knowledge^28,29 for successful movement through the environment.

Skilled movements include not only the anticipation of the consequences of actions at the level of the task but also anticipation of the reactive forces (interaction and gravitational torques) that are produced with multi-joint actions. These forces must be anticipated to ensure successful functional action. In order to successfully anticipate the dynamics for successful action, infants must experience the variability of their actions and the failure of certain types and forms of their actions to achieve their valued goals. A successful reach and grasp can be achieved through many different combinations of joint motions. Muscle torques may vary, yet the hand as “end effector” may succeed in grasping that bottle of milk under a variety of combinations of joint angles, joint torques, and multiple starting points. Early infant reaches are characterized by indirect hand paths, inconsistent velocities, and multiple units bounded with accelerations and decelerations (movement units).^30–32 These early reaches typically fail to achieve the goal of either touching or retrieving the desired object. With practice, however, reaches become straighter in the path to the target, force is concentrated to the beginning of the reach, and forces are anticipated and generated to ensure a successful reach and grasp. The wide variations in reach characteristics associated with failed attempts are tuned to the successful characteristics of the dynamics that lead to successful reaching under similar circumstances.

Actions that become stable in an infant's repertoire have value to infant systems. Sporns and Edelman defined value as “constraints provided by value systems already specified during embryogenesis as a result of evolutionary selection.”^1(p968) Early action then is more fully explored as environmental feedback to this value system guides what is considered adaptive in relation to these species specific values. For example, internal “value” for high visual contrasts leads infants to visually search areas around hair lines, eyes, and lips in the faces they view.³³

Another internal value, that of the value for maternal sound,³⁴ may create action that combines orienting to the maternal voice and searching the female face around hair lines, eyes, and lips. This initial searching does not have to include the concept of intention on the part of the infant. The internal value for high contrast and high-frequency sound couples to scaffold the initial actions. This turning to the female voice and searching the high-contrast areas of the face is adaptive, species-specific behavior that affords a variable interaction between infant and female; this female often is the mother. This face-searching behavior on the part of the infant contributes to the interpersonal dynamic of mother-infant interaction, which eventually may become intentional action on the part of the infant. This interaction is shaped as the infant explores in the context of varying maternal feedback and a contingency develops; either the baby or the mother can initiate this contingent interaction, but both are required for it to be sustained. This interaction requires an exploratory process that includes the cycle of variability within the dynamics of a given solution and the exploration of the possible range of solutions to maintain the interaction, all of which requires variability of the action of head, face, and vocalizations.

The still-face paradigm³⁵ is an example of the environmental elicitation of variability within a solution and the potential for new and variable solutions on the part of the infant to participate in mother-infant interaction when new constraints emerge and the expected contingent responses from the mother are not forthcoming. In this experimental paradigm, the mother is asked to interact with her infant, and then at the experimenter's cue, she stops the interaction and holds her face still. This is an extraordinary maternal task! When the infant is met by a “still-faced” mother, he or she may scale on a particular solution (eg, raising the eyebrows higher or widening the eyes) and eventually the infant may explore new movement solutions by including head movements or adding vocalizations to his or her repertoire.³⁶ Eventually the infant may even dissolve into crying, a new solution, if the mother continues to refrain from interaction. The solutions used to elicit the mother's expected feedback will have dynamic variability within each solution as the infant manages the forces of these new movement solutions. Adding movement solutions and variability within a solution are explored by the infant to elicit the interaction. High variability and the often consequent failure to achieve task goals are essential to mastery of skilled interaction. Failure, as experienced through variability of response, is essential to successful action.

The development of successful reaching and grasping is another action that includes and requires rich exploratory variability. Successful functional reach and grasp emerges during the first half year of life from actions that are initially only grossly directed toward objects.³⁰ Thelen and colleagues^37,38 contended that “new skills must arise from the interplay of new task demands with the already existing movement dynamics.”^37(p1060) Early arm actions are highly variable in terms of path, speed, and accuracy^31,39 but within the first year these early, unsuccessful actions develop into successful reach and grasp behaviors.⁴⁰ Precise pincher grasp is finally achieved after extensive exploration and repeated failures to achieve a functional outcome. A certain amount of variability in the path, speed, or functional success of reach and grasp will continue to be inherent in this flexible action, but if the reach and grasp action is not functional or not efficient, variability will be either increased or decreased, depending on the success of the dynamic action.

Variability as Problem Rather Than Solution

This raises the issue of movement variability as problem rather than a necessary part of the solution to the development of successful skilled actions. Variability as a source of information is expected as new functional skills are acquired.²² Indeed, if variability is restricted, the acquisition of skilled action also will be restricted. This lack of variability within a solution and of variability of adaptive solutions for task goals has been identified as a hallmark of atypical actions for both adults and children with neurological insults and resultant movement dysfunction.^20,41,42 Just as the lack of variability of action is a hindrance to the development of functional action, excessive variability can be a hindrance to the development of automatic, dependable, and typical functional action.^43–45

Consequences of Too Little and Too Much

Van Emmerick and colleagues⁴¹ demonstrated smaller changes in mean relative phase and lower variability between the pelvis and thorax (transverse plane) during walking in patients with Parkinson disease (PD) (due to rigidity leading to trunk or axial stiffness) compared with controls who were healthy. Relative phase refers to the relative position of joint angles during a movement. A common example is the relation of the hip angle to the knee angle during gait. In the study cited here, the angles created by the pelvis and thorax are compared. People with PD walk with a very “stable” pattern, which limits variability when the task demands change (eg, walking across the parking lot when the surface changes).

In their study of postural control in people with PD, Dimitrova et al⁴² demonstrated that people with PD were impaired in their ability to increase electromyographic responses when they transitioned from a wide to narrow base of support and they demonstrated co-contraction of agonist to perturbations. Participants could not vary their responses to appropriately meet the new task demands. This increase in co-contraction resulted in a fixed posture—a posture that is not an adaptive solution for maintaining balance in response to a perturbation.

Fetters et al⁴⁶ identified significantly higher intralimb joint correlations in infants born prematurely with white matter disease in comparison with infants born prematurely without these lesions and with infants born at full-term. This more “cramped” movement of the infants with lesions also was less variable. This reduced variability is shown in the Table as reduced standard deviations in the more distal joint couplings for the infants with white matter disease. Infants at this age are typically “decoupling” joints and beginning to combine flexion in one joint with relative extension at another. In all of these studies,^41,42,46 a decreased variability in response to changing task demands may have limited adaptability and skill acquisition.

The work by Sanger and colleagues^43,44 with children with dystonia demonstrates the challenge of excessive variability for the development of dependable, repeatable skilled action. Movement toward specific goals is continually accompanied by highly variable and inefficient movements. The child with certain types of dystonia must harness and reduce the variability in order to produce successful actions. Children with developmental coordination disorder, previously described as “clumsy,”⁴⁷ also demonstrate what can appear to be excessive variability in their functional actions.⁴⁵ They have difficulty reproducing efficient and effective actions across task contexts.^48,49 This inability to depend on movement capabilities paradoxically often leads these children to restrict movements to those they have mastered in one context. The movements may not be functional in the new context, and this lack of adaptability creates frustration and further decreases the incentives to explore with movements.

Evaluation of Variability

Both excessive and constrained variability present challenges for examination and evaluation of infants and children. Physical therapist clinical examination and evaluation of infants and young children have a primary goal to identify action challenges and counsel families on appropriate action. Identifying challenges requires our deep working understanding of the typical task solutions that emerge during early development and the expected variability in these solutions. Large databases form the basis of norm-referenced infant tests that define actions such as walking. The definition of walking then defines the variability in the onset of walking age. This is typically expressed as the average age of walking onset and the standard deviation for this age. Physical therapists are knowledgeable about this variation in the onset of walking and know when to alert parents that the variation of typical has been exceeded. However, we also have knowledge of the variations that are typical within walking. A child with spastic diplegia may succeed at walking, but the pattern of walking will be atypical and the variability of patterns of walking may be limited. In the first situation, we are alerted that variability of age of onset of locomotion is restricted, and in the second situation, we are alerted that variability within locomotion is restricted. The child with dystonia may have achieved locomotion at a typical age, but the within-pattern variability may be too large to be functional or efficient.

Intervention for Atypical Variability

Active problem solving—or, as Gibson²¹ termed it, “discovery learning”—provides a foundation for intervention for problems of variability of task solutions (too few) and hypovariability or hypervariability within solutions (too little and too much). The roles of the physical therapist are to “set the stage” for discovery learning and to contribute possible constraints to be eliminated or added in order to enhance the learning.²⁰ If variability is not a part of the system, it must be sought in therapy, varying context, limbs used, and motivations. Self-produced action must be sought with its inherent variability without the initial physical assistance of the physical therapist. This struggling, exploratory action on the part of the infant and child must be coupled with eventual success. The competent therapist designs the environment so that success can be supported while variable action is required. This approach typically requires a continual “update” of the demands of the environment while keeping the goal of the success of the action.

If variability is excessive and actions are not produced reliably and efficiently, it must be reduced in therapy. Reduced variability can be enhanced with a variety of environmental constraints, including increased proprioception through weighted limbs or vests.^50–52 Recent creative research with the use of exoskeletons has shown promise for improved gait for adults following stroke.⁵³ These exoskeletons create force fields to limit variability in action and “guide” the person toward a successful trajectory.

Body-weight–supported treadmill training has been used successfully with a variety of movement problems and patient groups, including children with Down syndrome,^54–56 cerebral palsy,⁵⁷ and myelomeningocele⁵⁸ and infants born with very low birth weight.⁵⁹ Constraints on speed of walking and various parameters of gait can be varied with control within the training and training can be individualized to support each infant or child's unique patterns of variability during supported walking.

In summary, variability is a necessary constraint for successful human actions. The lack of variability and excessive variability are hallmarks of the movement patterns produced by people across the life span following neurological insult. Active problem solving as therapy, with its inherent error as a part of the therapeutic process, is critical to the successful learning of functional actions. The role of the physical therapist is to create movement environments and provide personal and environmental constraints that elicit and support self-produced functional actions.